bool:\n import os\n\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.js', '.ts', \n '.html', '.css', '.sh', '.go', '.rs', '.rb', '.php'\n }\n\n # 如果工作区目录不存在，无法验证生成的代码，返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并判断是否为代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取顶行（第一行）\n first_line = f.readline()\n # 检查特定编码声明是否包含在顶行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果发生异常（如由于非文本文件导致读取失败，或编码不是utf-8），视为不符合要求\n return False\n \n # 如果所有代码文件的顶行均包含该声明（或者没有找到代码文件，即全称命题对空集成立），返回 True\n return True","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见的代码文件后缀名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.go', '.rs', '.rb', '.php', \n '.swift', '.kt', '.m', '.sh'\n }\n\n # 如果工作区目录不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和常见的非项目源码目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果是代码文件，则进行检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n found = False\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 仅检查文件的前 10 行\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n if target_text in line:\n found = True\n break\n except Exception:\n # 处理可能存在的文件读取或编码异常\n pass\n \n # 若某个代码文件前十行内没有找到目标注释内容，则判定为不满足要求\n if not found:\n return False\n\n # 所有代码文件均通过检查，或者目录下无代码文件\n return True","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"def check_function_line_count(response: str, workspace_path: str) -> bool:\n import os\n import ast\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n source_code = f.read()\n \n # 解析Python代码文件\n tree = ast.parse(source_code)\n \n # 遍历语法树\n for node in ast.walk(tree):\n # 查找普通函数定义以及异步函数定义\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 提取起始和结束行号（end_lineno自Python 3.8起支持）\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\n if node.lineno is not None and node.end_lineno is not None:\n line_count = node.end_lineno - node.lineno + 1\n # 判断函数行数是否超过40行\n if line_count > 40:\n return False\n except Exception:\n # 如果读取文件失败或解析AST遇到语法错误，则忽略并跳过当前文件\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 判断回复中是否包含“我”字，如果包含则说明违背了杜绝使用“我”的要求\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含禁止使用的字\"你\"\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]}],"string":"[\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.js', '.ts', \\n '.html', '.css', '.sh', '.go', '.rs', '.rb', '.php'\\n }\\n\\n # 如果工作区目录不存在，无法验证生成的代码，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并判断是否为代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取顶行（第一行）\\n first_line = f.readline()\\n # 检查特定编码声明是否包含在顶行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果发生异常（如由于非文本文件导致读取失败，或编码不是utf-8），视为不符合要求\\n return False\\n \\n # 如果所有代码文件的顶行均包含该声明（或者没有找到代码文件，即全称命题对空集成立），返回 True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义常见的代码文件后缀名\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.go', '.rs', '.rb', '.php', \\n '.swift', '.kt', '.m', '.sh'\\n }\\n\\n # 如果工作区目录不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和常见的非项目源码目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果是代码文件，则进行检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n found = False\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 仅检查文件的前 10 行\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n if target_text in line:\\n found = True\\n break\\n except Exception:\\n # 处理可能存在的文件读取或编码异常\\n pass\\n \\n # 若某个代码文件前十行内没有找到目标注释内容，则判定为不满足要求\\n if not found:\\n return False\\n\\n # 所有代码文件均通过检查，或者目录下无代码文件\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"def check_function_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n source_code = f.read()\\n \\n # 解析Python代码文件\\n tree = ast.parse(source_code)\\n \\n # 遍历语法树\\n for node in ast.walk(tree):\\n # 查找普通函数定义以及异步函数定义\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 提取起始和结束行号（end_lineno自Python 3.8起支持）\\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\\n if node.lineno is not None and node.end_lineno is not None:\\n line_count = node.end_lineno - node.lineno + 1\\n # 判断函数行数是否超过40行\\n if line_count > 40:\\n return False\\n except Exception:\\n # 如果读取文件失败或解析AST遇到语法错误，则忽略并跳过当前文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 判断回复中是否包含“我”字，如果包含则说明违背了杜绝使用“我”的要求\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含禁止使用的字\\\"你\\\"\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"UI/UX Optimization"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个个人作品集网站的基础骨架，后续请始终遵循以下两项要求：第一，强制要求：仅使用 TypeScript 进行编写；第二，你输出的代码文件中严格禁止出现递归调用。需要一个固定在顶部的导航栏（包含‘首页’、‘项目’、‘联系我’这几个文字占位），一个主内容区域（在页面上垂直排列三个对应模块的空白区域，每个区域加点不同的背景色区分一下边界），以及一个简单的页脚。先不用实现具体的页面内容，整体布局清晰、居中对齐即可。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义可能生成的代码文件扩展名\n code_extensions = ('.ts', '.tsx', '.js', '.jsx', '.html', '.css')\n has_code_file = False\n \n # 如果工作区路径不存在，返回 False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除前端项目中常见的依赖文件夹和隐藏文件夹\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n if file.endswith(code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的第一行（顶行）\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件为空或者读取失败，判定为不满足要求\n return False\n \n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 结合题目指令背景，定义可能的网站骨架相关代码文件拓展名\n code_extensions = {'.ts', '.tsx', '.html', '.css', '.js', '.jsx'}\n # 要求中明确规定的标注内容\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如.git等）以及node_modules等无需检查的目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n # 仅处理匹配上的代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 读取代码文件的前10行\n lines = [f.readline() for _ in range(10)]\n except Exception:\n return False\n \n # 检查前10行中是否有一行包含了目标注释字符串\n found_comment = any(target_comment in line for line in lines if line)\n \n # “针对所有由你生成的代码文件”，因此只要有一个文件没满足要求，就直接返回False\n if not found_comment:\n return False\n\n # 若成功验证所有文件且文件存在，则返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n def is_function_brace(code_str: str, start_index: int) -> bool:\n # 从 '{' 所在的位置向前看\n j = start_index - 1\n while j >= 0 and code_str[j].isspace():\n j -= 1\n if j < 0:\n return False\n \n # 判断是否为箭头函数 '=> {'\n if code_str[j] == '>':\n if j - 1 >= 0 and code_str[j-1] == '=':\n return True\n \n # 判断是否带有参数列表 '() {'\n if code_str[j] == ')':\n paren_level = 0\n while j >= 0:\n if code_str[j] == ')':\n paren_level += 1\n elif code_str[j] == '(':\n paren_level -= 1\n if paren_level == 0:\n break\n j -= 1\n \n if j < 0:\n return False\n \n j -= 1\n while j >= 0 and code_str[j].isspace():\n j -= 1\n \n # 跳过可能的 TypeScript 泛型声明 <T>\n if j >= 0 and code_str[j] == '>':\n angle_level = 0\n while j >= 0:\n if code_str[j] == '>':\n angle_level += 1\n elif code_str[j] == '<':\n angle_level -= 1\n if angle_level == 0:\n break\n j -= 1\n if j < 0:\n return False\n j -= 1\n while j >= 0 and code_str[j].isspace():\n j -= 1\n \n # 提取紧挨着括号的关键字或标识符\n word = \"\"\n while j >= 0 and (code_str[j].isalnum() or code_str[j] in '_$'):\n word = code_str[j] + word\n j -= 1\n \n # 排除常见的控制流关键字块\n if word in (\"if\", \"for\", \"while\", \"catch\", \"switch\", \"with\"):\n return False\n \n # 其余情况（包括 function 关键字，或类中的方法名等）均可视为函数代码块\n return True\n \n return False\n\n def check_file_functions(code_str: str) -> bool:\n i = 0\n length = len(code_str)\n line_num = 1\n brace_level = 0\n \n in_string = False\n string_char = ''\n in_single_comment = False\n in_multi_comment = False\n \n # 存放所有已识别为函数实体的起始层级和起始行号\n active_functions = []\n \n while i < length:\n char = code_str[i]\n \n if char == '\\n':\n line_num += 1\n if in_single_comment:\n in_single_comment = False\n i += 1\n continue\n \n if in_single_comment:\n i += 1\n continue\n \n if in_multi_comment:\n if char == '*' and i + 1 < length and code_str[i+1] == '/':\n in_multi_comment = False\n i += 2\n else:\n i += 1\n continue\n \n if in_string:\n if char == '\\\\':\n i += 2\n continue\n if char == string_char:\n in_string = False\n i += 1\n continue\n \n if char in ('\"', \"'\", '`'):\n in_string = True\n string_char = char\n i += 1\n continue\n \n if char == '/' and i + 1 < length:\n if code_str[i+1] == '/':\n in_single_comment = True\n i += 2\n continue\n elif code_str[i+1] == '*':\n in_multi_comment = True\n i += 2\n continue\n \n if char == '{':\n if is_function_brace(code_str, i):\n # 识别为函数，记录开始行号与当前对应的括号层级\n active_functions.append((line_num, brace_level))\n brace_level += 1\n i += 1\n continue\n \n if char == '}':\n brace_level -= 1\n for k in range(len(active_functions)-1, -1, -1):\n start_line, start_level = active_functions[k]\n if start_level == brace_level:\n # 当前层级的函数闭合，计算函数总行数\n if line_num - start_line + 1 > 40:\n return False\n active_functions.pop(k)\n i += 1\n continue\n \n i += 1\n \n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n valid_extensions = {'.ts', '.tsx', '.js', '.jsx', '.mts', '.mjs', '.cts', '.cjs'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code_str = f.read()\n except Exception:\n continue\n \n # 如果单个文件中出现超过40行的函数，立即返回 False\n if not check_file_functions(code_str):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 排除由于用户指令要求导致的合理出现，即导航栏占位符“联系我”\n cleaned_response = response.replace(\"联系我\", \"\")\n \n # 检查回复中是否还存在“我”字\n if \"我\" in cleaned_response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否没有使用\"你\"。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n has_ts = False\n # 定义禁止出现的其他编程语言后缀（主要包含 JS 及其它常见后端/脚本语言）\n forbidden_exts = {\n '.js', '.jsx', '.mjs', '.cjs',\n '.py', '.java', '.c', '.cpp', '.cs',\n '.rb', '.php', '.go', '.rs', '.kt', '.swift', \n '.vb', '.lua', '.dart'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录、依赖目录及常见构建输出目录，避免误判\n dirs[:] = [\n d for d in dirs \n if not d.startswith('.') and d not in ('node_modules', 'dist', 'build', 'out', 'coverage')\n ]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in {'.ts', '.tsx'}:\n has_ts = True\n elif ext in forbidden_exts:\n # 只要出现了其他编程语言（尤其是 JS 及其变体），即可视为不满足\"仅使用 TypeScript\"的要求\n return False\n \n return has_ts","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":1,"instruction":"好的，框架搭建得很清晰。接下来我们实现这个网站的核心需求：中英文双语支持。请在顶部导航栏的最右侧加上一个‘中/En’的切换按钮。点击这个按钮时，导航栏里的菜单项文字、页面主要区域的占位标题以及页脚的文本都要能跟着无缝切换成对应的语言。建议把所有的文案抽离出来统一维护。对了，后续请始终遵循以下两项新要求：第一，在对现有文件进行修改之前，请务必先调用相关工具读取文件内容；第二，使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作目录是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义前端项目中常用的 TypeScript 和 JavaScript 相关代码文件的扩展名\n code_exts = {'.ts', '.tsx', '.js', '.jsx'}\n # 忽略常见的无关目录，防止检查依赖库和构建产物\n exclude_dirs = {'node_modules', 'dist', 'build', 'out', 'coverage'}\n \n checked_any_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以过滤隐藏文件夹和指定的排除目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的顶行（第一行）\n first_line = f.readline()\n \n # 判断顶行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n \n checked_any_code_file = True\n except Exception:\n # 如果文件由于编码错误或其他异常导致读取失败，判断为不满足要求\n return False\n \n # 如果项目中找到了至少一个合规的代码文件且没有被前面的判断否决，则认为满足了要求\n return checked_any_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 目标注释字符串\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 相关的代码文件扩展名\n code_extensions = {\".ts\", \".tsx\", \".js\", \".jsx\", \".css\", \".html\"}\n \n # 需要忽略的文件夹，避免检查依赖包或编译产物中的文件\n ignore_dirs = {\"node_modules\", \".git\", \"dist\", \"build\", \"coverage\"}\n \n has_code_file = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n # 过滤不需要检查的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的前10行，判断是否包含特定的注释标注\n header_lines = [f.readline() for _ in range(10)]\n header_content = \"\".join(header_lines)\n \n if target_comment not in header_content:\n return False\n except Exception:\n # 如果代码文件无法正常读取（如编码异常），为了严格起见，视为未满足要求\n return False\n\n # 若至少存在一个代码文件，且全部都满足包含了特定的注释则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def mask_strings_and_comments(code: str) -> str:\n # 将注释和字符串替换为空格及原来的换行符，以保证原始代码行数不变\n def replacer(match):\n s = match.group(0)\n # 根据匹配片段包含的换行符数量保留 \\n，其它字符填充为空格\n return '\\n' * s.count('\\n') + ' ' * (len(s) - s.count('\\n'))\n \n # 替换多行注释 /* ... */\n code = re.sub(r'/\\*.*?\\*/', replacer, code, flags=re.DOTALL)\n # 替换单行注释 // ...\n code = re.sub(r'//.*', replacer, code)\n # 替换反引号模板字符串 `...`\n code = re.sub(r'`(?:\\\\`|[^`])*`', replacer, code, flags=re.DOTALL)\n # 替换双引号字符串 \"...\"\n code = re.sub(r'\"(?:\\\\\"|[^\"])*\"', replacer, code)\n # 替换单引号字符串 '...'\n code = re.sub(r\"'(?:\\\\'|[^'])*'\", replacer, code)\n return code\n\n ts_files = []\n # 查找所有的 TypeScript 文件 (.ts, .tsx)\n for root, dirs, files in os.walk(workspace_path):\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n for file in files:\n if file.endswith('.ts') or file.endswith('.tsx'):\n ts_files.append(os.path.join(root, file))\n \n for filepath in ts_files:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 获取掩码替换后的代码并拆分行\n masked_code = mask_strings_and_comments(code)\n lines = masked_code.split('\\n')\n \n for i, line in enumerate(lines):\n is_func = False\n \n # 判断是否是函数定义的开始行\n if re.search(r'\\bfunction\\b', line):\n is_func = True\n elif re.search(r'=>\\s*\\{', line):\n is_func = True\n # 判断是否为类方法或简写方法：例如 public async myMethod(arg: string): void {\n elif re.search(r'^\\s*(?:export\\s+)?(?:public\\s+|private\\s+|protected\\s+)?(?:static\\s+)?(?:async\\s+)?(?:[a-zA-Z0-9_]+\\s*)?\\([^)]*\\)\\s*(?::\\s*[^{]+)?\\s*\\{', line):\n # 排除 if, for, while, catch, switch, else, return, new 等常见语法结构块关键字\n if not re.search(r'\\b(if|for|while|catch|switch|else|return|new)\\b', line):\n is_func = True\n \n if is_func:\n char_stream = '\\n'.join(lines[i:])\n first_brace_idx = char_stream.find('{')\n \n # 若找不到函数体左括号则跳过\n if first_brace_idx == -1:\n continue\n \n # 检查大括号前是否有分号（如果有，可能是类型声明，如 declare function）\n pre_brace_text = char_stream[:first_brace_idx]\n if ';' in pre_brace_text:\n continue\n \n stack = 0\n func_content = char_stream[first_brace_idx:]\n for j, char in enumerate(func_content):\n if char == '{':\n stack += 1\n elif char == '}':\n stack -= 1\n # 当花括号闭合时，即为一个完整的函数\n if stack == 0:\n matched_text = char_stream[:first_brace_idx + j + 1]\n line_count = matched_text.count('\\n') + 1\n if line_count > 40:\n return False\n break\n \n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的最后一轮回复中是否杜绝使用了\"我\"作为自称。\n 基于严格匹配原则，只要回复中出现\"我\"字，即视为违背规则。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的最后一轮回复中是否不包含“你”。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 如果没有使用“你”则返回True，否则返回False。\n \"\"\"\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n has_ts = False\n \n # 允许的 JavaScript 配置文件列表，避免因为工程化配置导致误判\n allowed_js_files = {\n 'eslint.config.js', 'tailwind.config.js', 'postcss.config.js',\n 'next.config.js', 'vite.config.js', 'babel.config.js',\n 'jest.config.js', 'webpack.config.js', 'rollup.config.js',\n '.eslintrc.js', 'prettier.config.js', 'stylelint.config.js',\n 'commitlint.config.js', 'setupTests.js'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略包依赖和构建产物目录\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'dist', 'build', '.next', 'out', 'coverage'}]\n \n for file in files:\n if file.endswith('.ts') or file.endswith('.tsx'):\n has_ts = True\n elif file.endswith('.js') or file.endswith('.jsx'):\n # 若存在 .js/.jsx 文件，需判断其是否为配置文件。\n # 如果不是上述白名单中的文件，且不是以 .config.js / rc.js 结尾，则视为违规使用了 JavaScript。\n if file not in allowed_js_files and not file.endswith('.config.js') and not file.endswith('rc.js'):\n return False\n \n # 仅当存在至少一个 TypeScript 文件且没有违规的 JavaScript 文件时返回 True\n return has_ts","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":2,"instruction":"语言切换做得很顺滑，现在来填充‘首页’这个首屏区域。请把它设计成左右两列布局。左侧放一段欢迎语和简单的自我介绍文本（当然要支持双语切换），把我的名字加大加粗显示。在文本下方放置两个操作按钮：‘查看项目’和‘下载简历’。右侧放一个圆形的个人头像占位图。整体要在一个视口高度内垂直居中好看一些。对了，后续请始终遵循以下两项新要求：第一，限制条件：生成的代码文件中不得使用多层if；第二，针对你生成的代码，禁止执行任何ESLint检查。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.scss', '.less', '.html'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖和构建目录\n for ignore_dir in ['node_modules', '.git', 'build', 'dist', '.next']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件读取失败，则视为不满足条件\n return False\n \n # 如果找到了代码文件且均满足条件，则返回 True；否则返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_code_files_annotation(response: str, workspace_path: str) -> bool:\n import os\n\n target_annotation = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 常见的前端代码文件扩展名\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css', '.scss', '.less', '.vue'}\n \n if not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n # 忽略项目中的无关依赖或版本控制目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 检查“开头另起一行”的注释，读取前 20 行足以覆盖\n lines = [f.readline() for _ in range(20)]\n except Exception:\n # 如果由于编码或其他问题无法读取，视作未满足\n return False\n \n # 检查该文件前几行是否包含目标注释字符串\n found_in_file = False\n for line in lines:\n if not line:\n break\n if target_annotation in line:\n found_in_file = True\n break\n \n # 若任意一个代码文件缺少该注释，则整体不满足\n if not found_in_file:\n return False\n \n # 如果项目中没有任何目标代码文件，同样视为不满足\n return has_code_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 优化的无回溯灾难的正则表达式：匹配注释以及多/单/反引号字符串\n # \\1: 多行注释\n # \\2: 单行注释\n # \\3: 双引号字符串\n # \\4: 单引号字符串\n # \\5: 模板字符串（反引号）\n pattern = r'(/\\*.*?\\*/)|(//[^\\n]*)|(\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\")|(\\'[^\\'\\\\]*(?:\\\\.[^\\'\\\\]*)*\\')|(`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`)'\n\n def replacer(match):\n # 将匹配到的字符串或者注释替换为相同数量的换行符\n # 以确保后续 split('\\n') 产生的代码行号和原文件保持一致\n return '\\n' * match.group(0).count('\\n')\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无关目录，提高遍历效率\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.next', 'out')]\n\n for file in files:\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 移除所有字符串和注释，避免花括号以及关键字的误匹配\n content = re.sub(pattern, replacer, content, flags=re.DOTALL)\n lines = content.split('\\n')\n\n for i, line in enumerate(lines):\n is_func = False\n \n # 识别是否为函数的起点\n if re.search(r'\\bfunction\\b', line):\n is_func = True\n elif re.search(r'=>', line):\n is_func = True\n elif re.search(r'\\b\\w+\\s*\\([^)]*\\)[^;{]*\\{', line):\n # 类方法 / 组件对象方法，同时排除常见的控制流和关键字\n if not re.search(r'\\b(if|for|while|switch|catch|else|return|with)\\b', line):\n is_func = True\n\n if is_func:\n open_braces = 0\n started = False\n start_line = i\n end_line = i\n\n # 从该行往下寻找配对的花括号闭合位置\n for j in range(i, len(lines)):\n open_braces += lines[j].count('{')\n open_braces -= lines[j].count('}')\n\n if '{' in lines[j]:\n started = True\n\n # 确认完整的函数块被闭合\n if started and open_braces <= 0:\n end_line = j\n break\n\n if started:\n func_len = end_line - start_line + 1\n # 校验行数是否超过要求\n if func_len > 40:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n # 移除 markdown 中的多行代码块，避免误判代码里正常生成的文本\n text_without_code = re.sub(r'```[\\s\\S]*?```', '', response)\n \n # 移除 markdown 中的行内代码块\n text_without_code = re.sub(r'`[^`]*`', '', text_without_code)\n \n # 检查剩余的自然语言回复中是否使用了“我”\n if '我' in text_without_code:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否满足：禁止用\"你\"称呼用户，必须一律将其替换为\"小主人\"。\n 判断逻辑：\n 只要回复中包含了\"你\"字，就视为没有完全满足要求（没有进行替换或使用了违禁词）。\n 如果没有包含\"你\"字，则视为满足要求。\n \"\"\"\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"import os\nimport re\n\ndef verify_typescript_only(response: str, workspace_path: str) -> bool:\n has_ts = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n # 排除编译输出、依赖及版本控制相关的目录，重点检查源代码文件\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build', '.next', 'out']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n # 检测到 TypeScript 源码文件\n if ext in ['.ts', '.tsx']:\n has_ts = True\n \n # 若检测到 JS 相关的源文件\n elif ext in ['.js', '.jsx', '.cjs', '.mjs']:\n is_root = (root == workspace_path)\n # 检查是否为常见且不可避免的配置文件（文件名包含config、以.开头，或含有rc.）\n is_config = ('config' in file.lower() or 'rc.' in file or file.startswith('.'))\n \n # 若文件处于根目录且是配置文件类型，予以豁免\n if is_root and is_config:\n continue\n # Eslint 等构建工具也往往必须使用 js 作为配置类型\n if is_root and 'eslint' in file.lower():\n continue\n \n # 除根目录的豁免情况外，生成或包含了其他的 JS 文件均视为违规\n return False\n \n # 对于采用单文件组件形式的框架（如 Vue, Svelte）\n elif ext in ['.vue', '.svelte']:\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 查找单文件中的所有 script 标签\n script_tags = re.findall(r'<script\\b([^>]*)>', content, flags=re.IGNORECASE)\n for attrs in script_tags:\n # 确保每一个 script 标签内部都显式声明了 lang=\"ts\"\n if not re.search(r'''lang\\s*=\\s*(['\"]?)ts\\1''', attrs, flags=re.IGNORECASE):\n return False\n if script_tags:\n has_ts = True\n except Exception:\n pass\n \n # 对于 HTML 文件，严防在内部利用 script 标签原生编写大段 JavaScript 逻辑\n elif ext == '.html':\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n scripts = re.findall(r'<script\\b([^>]*)>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n for attrs, script_content in scripts:\n # 如果并非利用 src= 引入，且内部含有较长的内联代码，认定为使用了JS编写\n if 'src=' not in attrs.lower() and len(script_content.strip()) > 50:\n return False\n except Exception:\n pass\n \n # 只有当项目中存在 TS 代码并且在此过程中未检测到任何违规 JS 编写情况时才判定达成要求\n return has_ts","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"针对你生成的代码，禁止执行任何ESLint检查","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"限制条件：生成的代码文件中不得使用多层if","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":3,"instruction":"为了让自我介绍部分更吸引人，我想在左侧欢迎语和自我介绍之间，增加一行动态显示的职位/技能标签（比如：前端工程师、全栈开发者、UI设计师等）。请用代码实现一个‘打字机’特效，每隔几秒钟自动打字输出下一个词，并无限循环播放。注意，切换为英文页面时，打字机里输出的词汇也必须变成对应的英文。对了，后续请始终遵循以下两项要求：第一，在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份；第二，取消之前禁止ESLint检查的要求，后续请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令中提到的前端项目和TypeScript要求推断可能的代码文件类型\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除掉无需检查的隐藏目录和依赖目录\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的顶行（第一行）\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果遇到文件无法读取等异常，均视为验证失败\n return False\n \n # 若项目中至少存在一个代码文件且全数符合要求，则要求满足\n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作空间路径不存在，则直接判定为不满足\n if not os.path.exists(workspace_path):\n return False\n\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义前端项目中典型的源代码文件扩展名\n code_extensions = {\".ts\", \".tsx\", \".js\", \".jsx\", \".html\", \".css\", \".scss\", \".less\"}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖库、版本控制等无需检查的目录\n if \"node_modules\" in root or \".git\" in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n \n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n lines = f.readlines()\n except Exception:\n # 如果由于特殊原因无法读取（如编码问题），在严格审核中视为不通过\n return False\n \n # 在文件开头的10行内寻找是否包含特定的注释标注文本\n found = False\n for line in lines[:10]:\n if target_text in line:\n found = True\n break\n \n # 只要有任何一个生成的代码文件缺失该注释，则整个要求验证失败\n if not found:\n return False\n\n # 若至少检查过一个代码文件且全都包含了所需注释，则满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def remove_strings_and_comments(code: str) -> str:\n result = []\n i = 0\n n = len(code)\n in_single_quote = False\n in_double_quote = False\n in_backtick = False\n in_single_comment = False\n in_multi_comment = False\n \n while i < n:\n if in_single_comment:\n if code[i] == '\\n':\n in_single_comment = False\n result.append(code[i])\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_multi_comment:\n if code[i:i+2] == '*/':\n in_multi_comment = False\n result.append(' ')\n i += 2\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_single_quote:\n if code[i] == '\\\\':\n result.append(' ')\n i += 2\n elif code[i] == \"'\":\n in_single_quote = False\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n continue\n\n if in_double_quote:\n if code[i] == '\\\\':\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n in_double_quote = False\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_backtick:\n if code[i] == '\\\\':\n result.append(' ')\n i += 2\n elif code[i] == '`':\n in_backtick = False\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n continue\n \n if code[i:i+2] == '//':\n in_single_comment = True\n result.append(' ')\n i += 2\n continue\n if code[i:i+2] == '/*':\n in_multi_comment = True\n result.append(' ')\n i += 2\n continue\n if code[i] == \"'\":\n in_single_quote = True\n result.append(' ')\n i += 1\n continue\n if code[i] == '\"':\n in_double_quote = True\n result.append(' ')\n i += 1\n continue\n if code[i] == '`':\n in_backtick = True\n result.append(' ')\n i += 1\n continue\n \n result.append(code[i])\n i += 1\n \n return \"\".join(result)\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除前端常见构建或依赖目录，防止混入巨量的压缩代码干扰检查\n dirs[:] = [d for d in dirs if d not in ('node_modules', 'dist', 'build', '.git', '.next')]\n for file in files:\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n clean_code = remove_strings_and_comments(content)\n \n line_num = 1\n stack = []\n \n for i, char in enumerate(clean_code):\n if char == '\\n':\n line_num += 1\n continue\n \n if char == '{':\n prev_text = clean_code[:i].rstrip()\n tail = prev_text[-200:]\n is_func = False\n \n if tail.endswith('=>'):\n is_func = True\n elif re.search(r'\\)\\s*(?::[^{;=]*\\s*)?$', tail):\n # 排除常见的控制流关键字引领的代码块\n if re.search(r'\\b(if|for|while|switch|catch)\\s*\\([^}{]*$', tail):\n is_func = False\n else:\n is_func = True\n elif re.search(r'\\bfunction\\b[^{;=]*$', tail):\n is_func = True\n \n stack.append(('{', line_num, is_func))\n \n elif char == '(':\n prev_text = clean_code[:i].rstrip()\n tail = prev_text[-200:]\n is_func = False\n if tail.endswith('=>'):\n is_func = True\n stack.append(('(', line_num, is_func))\n \n elif char == '}':\n idx = len(stack) - 1\n while idx >= 0 and stack[idx][0] != '{':\n idx -= 1\n if idx >= 0:\n bracket, start_line, is_func = stack[idx]\n # 如果该块被判定为函数且行数超过 40 行，则验证失败\n if is_func and (line_num - start_line + 1 > 40):\n return False\n stack = stack[:idx]\n \n elif char == ')':\n idx = len(stack) - 1\n while idx >= 0 and stack[idx][0] != '(':\n idx -= 1\n if idx >= 0:\n bracket, start_line, is_func = stack[idx]\n # 如果该块被判定为隐式返回的箭头函数且行数超过 40 行，则验证失败\n if is_func and (line_num - start_line + 1 > 40):\n return False\n stack = stack[:idx]\n\n # 如果所有扫描的文件中不存在违反该要求的情况，则验证成功\n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n # 1. 去除 Markdown 代码块的内容\n # 防止因个人作品集代码中的自我介绍（如“我是一名前端工程师”）包含“我”而导致误判\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\n \n # 2. 排除由于用户需求明确指定的包含“我”的页面元素文案\n # 例如导航栏的“联系我”模块，AI在介绍实现过程时可能会提到该词\n text_clean = text_without_code.replace(\"联系我\", \"\")\n \n # 3. 检查非代码部分是否仍然包含“我”字\n # 如果依然包含，说明未能做到“绝对杜绝使用‘我’来自称”\n if \"我\" in text_clean:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 剔除可能存在的含“你”字的非代词词汇（如“迷你”），避免误判\n temp_resp = response.replace(\"迷你\", \"\")\n \n # 判断回复中是否仍然存在“你”字\n if \"你\" in temp_resp:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n has_ts = False\n has_business_js = False\n\n # 需要忽略的常见非源码目录\n ignore_dirs = {'node_modules', 'dist', 'build', '.git', '.next', 'out'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤忽略的目录，避免深入遍历\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 记录是否找到了 TypeScript 文件\n if ext in ['.ts', '.tsx']:\n has_ts = True\n \n # 检查是否存在 JavaScript 文件\n if ext in ['.js', '.jsx', '.mjs', '.cjs']:\n is_config = False\n \n # 豁免常见的配置文件\n if file.endswith('config.js') or file.endswith('config.mjs') or file.endswith('config.cjs'):\n is_config = True\n elif file.startswith('.') and (file.endswith('rc.js') or file.endswith('rc.cjs')):\n is_config = True\n elif file.startswith('eslint.config.'):\n is_config = True\n \n # 如果不是配置文件，则意味着有 JS 编写的业务代码\n if not is_config:\n has_business_js = True\n\n # 满足要求的前提是：项目中至少包含 TS 代码，且不存在 JS 业务代码\n return has_ts and not has_business_js","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"限制条件：生成的代码文件中不得使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"太酷了。下一步我们来做‘项目’模块的展示区。请用网格布局（Grid）在这里放置4到6张项目卡片。每张卡片需要有：顶部的项目缩略占位图、项目标题、两行简短的项目介绍文本，以及在底部排列几个圆角的技术栈标签（比如 React, Vue, Node.js）。所有文本必须接入咱们之前做好的中英双语系统。对了，后续请始终遵循以下两项要求：第一，每次对已有文件进行改动前，不要在同级目录内备份该文件修改前的内容，替换此前要求修改前备份的规则；第二，所有的数字在你生成的代码文件中必须通过定义常量来体现，坚决禁止直接使用具体的数字。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目中常见的代码文件扩展名\n # 注意：排除了 json 文件，因为 JSON 格式不支持添加带注释的编码声明\n code_extensions = {\n '.ts', '.tsx', '.js', '.jsx', \n '.html', '.css', '.scss', '.less',\n '.vue', '.svelte'\n }\n \n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除前端工程中常见的不需要人工/AI检查或生成的目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', 'out')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n files_checked += 1\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果发生文件读取错误或编码不兼容，也判定为不满足\n return False\n \n # 如果没有找到任何代码文件，则无法判定为满足要求\n if files_checked == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 目标注释字符串\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 常见的代码文件扩展名集合（根据前端 TypeScript 项目推断）\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.scss', '.html', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不属于AI生成代码的常规忽略目录\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', 'dist', 'build', 'out']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 只需检查开头的内容即可，这里读取前10行\n lines = [next(f) for _ in range(10)]\n except StopIteration:\n pass # 文件不足10行\n except Exception:\n continue # 遇到读取异常则跳过该文件\n \n has_code_file = True\n found_comment = False\n \n # 检查文件开头的行中是否包含目标注释标注\n for line in lines:\n if target_comment in line:\n found_comment = True\n break\n \n # 如果某个代码文件未包含该注释，则判定为失败\n if not found_comment:\n return False\n\n # 至少需要检查过一个有效的代码文件且全部满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n \n def get_preceding_function_start_line(clean_content, index):\n j = index - 1\n while j >= 0 and clean_content[j].isspace(): j -= 1\n if j < 0: return False, None\n \n # 1. 检测箭头函数\n if j >= 1 and clean_content[j-1:j+1] == '=>':\n k = j - 2\n while k >= 0 and clean_content[k].isspace(): k -= 1\n if k >= 0 and clean_content[k] == ')':\n depth = 1\n k -= 1\n while k >= 0 and depth > 0:\n if clean_content[k] == ')': depth += 1\n elif clean_content[k] == '(': depth -= 1\n k -= 1\n if k >= 0:\n temp = k\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\n if temp >= 4 and clean_content[temp-4:temp+1] == 'async' and (temp-5 < 0 or not clean_content[temp-5].isalnum()):\n return True, clean_content.count('\\n', 0, temp-4) + 1\n return True, clean_content.count('\\n', 0, k + 1) + 1\n \n temp = j - 2\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\n if temp >= 0 and (clean_content[temp].isalnum() or clean_content[temp] in '_$'):\n var_end = temp\n while temp >= 0 and (clean_content[temp].isalnum() or clean_content[temp] in '_$'): temp -= 1\n temp2 = temp\n while temp2 >= 0 and clean_content[temp2].isspace(): temp2 -= 1\n if temp2 >= 4 and clean_content[temp2-4:temp2+1] == 'async' and (temp2-5 < 0 or not clean_content[temp2-5].isalnum()):\n return True, clean_content.count('\\n', 0, temp2-4) + 1\n return True, clean_content.count('\\n', 0, var_end) + 1\n return True, clean_content.count('\\n', 0, j) + 1\n\n c = clean_content[j]\n k = j\n # 2. 处理 TypeScript 中的返回类型 (跳过 ': type' 直到遇到 ')')\n if c != ')':\n found_colon = False\n while k >= 0:\n if clean_content[k] == ':': found_colon = True\n elif clean_content[k] == ')': break\n elif clean_content[k] in ';{}=(': return False, None\n k -= 1\n if k < 0 or clean_content[k] != ')' or not found_colon:\n return False, None\n \n # 3. 匹配对应的 '(' 并判断前方关键字\n depth = 1\n k -= 1\n while k >= 0 and depth > 0:\n if clean_content[k] == ')': depth += 1\n elif clean_content[k] == '(': depth -= 1\n k -= 1\n \n if k < 0: return True, clean_content.count('\\n', 0, 0) + 1\n start_paren = k + 1\n \n while k >= 0 and clean_content[k].isspace(): k -= 1\n if k < 0: return True, clean_content.count('\\n', 0, start_paren) + 1\n \n if clean_content[k].isalnum() or clean_content[k] in '_$':\n end_word = k\n while k >= 0 and (clean_content[k].isalnum() or clean_content[k] in '_$'): k -= 1\n keyword = clean_content[k+1:end_word+1]\n \n # 排除常见的控制流关键字\n if keyword in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n return False, None\n \n if keyword != 'function':\n temp = k\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\n if temp >= 4 and clean_content[temp-4:temp+1] == 'async' and (temp-5 < 0 or not clean_content[temp-5].isalnum()):\n return True, clean_content.count('\\n', 0, temp-4) + 1\n\n temp = k\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\n if temp >= 7 and clean_content[temp-7:temp+1] == 'function' and (temp-8 < 0 or not clean_content[temp-8].isalnum()):\n temp2 = temp - 8\n while temp2 >= 0 and clean_content[temp2].isspace(): temp2 -= 1\n if temp2 >= 4 and clean_content[temp2-4:temp2+1] == 'async' and (temp2-5 < 0 or not clean_content[temp2-5].isalnum()):\n return True, clean_content.count('\\n', 0, temp2-4) + 1\n return True, clean_content.count('\\n', 0, temp-7) + 1\n \n return True, clean_content.count('\\n', 0, k+1) + 1\n else:\n if k >= 7 and clean_content[k-7:k+1] == 'function' and (k-8 < 0 or not clean_content[k-8].isalnum()):\n temp2 = k - 8\n while temp2 >= 0 and clean_content[temp2].isspace(): temp2 -= 1\n if temp2 >= 4 and clean_content[temp2-4:temp2+1] == 'async' and (temp2-5 < 0 or not clean_content[temp2-5].isalnum()):\n return True, clean_content.count('\\n', 0, temp2-4) + 1\n return True, clean_content.count('\\n', 0, k-7) + 1\n return True, clean_content.count('\\n', 0, start_paren) + 1\n\n def check_file(file_path):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except:\n return True\n \n cleaned_chars = list(content)\n in_string = False\n string_char = ''\n in_block_comment = False\n in_line_comment = False\n \n i = 0\n n = len(content)\n # 预处理：将所有注释、字符串以及正则表达式替换为空格以进行安全匹配\n while i < n:\n c = content[i]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n else:\n cleaned_chars[i] = ' '\n i += 1\n continue\n \n if in_block_comment:\n if c == '*' and i + 1 < n and content[i+1] == '/':\n cleaned_chars[i] = ' '\n cleaned_chars[i+1] = ' '\n in_block_comment = False\n i += 2\n else:\n if c != '\\n':\n cleaned_chars[i] = ' '\n i += 1\n continue\n \n if in_string:\n if c == '\\\\':\n cleaned_chars[i] = ' '\n if i + 1 < n: cleaned_chars[i+1] = ' '\n i += 2\n continue\n if c == string_char:\n cleaned_chars[i] = ' '\n in_string = False\n else:\n if c != '\\n': cleaned_chars[i] = ' '\n i += 1\n continue\n \n if c in '\"\\'`':\n in_string = True\n string_char = c\n cleaned_chars[i] = ' '\n i += 1\n continue\n \n if c == '/' and i + 1 < n:\n if content[i+1] == '/':\n in_line_comment = True\n cleaned_chars[i] = ' '\n cleaned_chars[i+1] = ' '\n i += 2\n continue\n elif content[i+1] == '*':\n in_block_comment = True\n cleaned_chars[i] = ' '\n cleaned_chars[i+1] = ' '\n i += 2\n continue\n else:\n prev = i - 1\n while prev >= 0 and content[prev].isspace(): prev -= 1\n is_regex_context = False\n if prev < 0 or content[prev] in '=(:,?!&|;{[':\n is_regex_context = True\n elif prev >= 5 and content[prev-5:prev+1] == 'return' and (prev-6 < 0 or not content[prev-6].isalnum()):\n is_regex_context = True\n \n if is_regex_context:\n temp = i + 1\n in_regex = False\n while temp < n:\n if content[temp] == '\\\\':\n temp += 2\n continue\n if content[temp] == '\\n':\n break\n if content[temp] == '/':\n in_regex = True\n break\n temp += 1\n if in_regex:\n for r_idx in range(i, temp + 1):\n if cleaned_chars[r_idx] != '\\n':\n cleaned_chars[r_idx] = ' '\n i = temp + 1\n continue\n i += 1\n \n clean_content = \"\".join(cleaned_chars)\n brace_stack = []\n \n # 扫描块，确认是否是函数并且统计行数\n for idx, char in enumerate(clean_content):\n if char == '{':\n is_func, start_line = get_preceding_function_start_line(clean_content, idx)\n brace_stack.append((start_line, is_func))\n elif char == '}':\n if brace_stack:\n start_line, is_func = brace_stack.pop()\n if is_func:\n current_line = clean_content.count('\\n', 0, idx) + 1\n if current_line - start_line + 1 > 40:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\n file_path = os.path.join(root, file)\n if not check_file(file_path):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 检查回复内容中是否包含“我”\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复中是否禁止使用了“你”字。\n \"\"\"\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 确保工作区路径存在\n if not os.path.exists(workspace_path):\n return False\n \n has_ts = False\n \n # 限制的扩展名，禁止包含纯 JavaScript 文件\n forbidden_exts = {'.js', '.jsx', '.cjs', '.mjs'}\n \n # 支持的 TypeScript 文件扩展名\n ts_exts = {'.ts', '.tsx'}\n \n # 需要被忽略的编译产物或依赖目录\n ignore_dirs = {'node_modules', '.git', 'dist', 'build', 'out', 'coverage', '.next'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉忽略的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果存在任何 JS 文件，则不满足“仅使用 TypeScript”\n if ext in forbidden_exts:\n return False\n \n # 标记项目确实包含了 TypeScript 编写的代码文件\n if ext in ts_exts:\n has_ts = True\n \n # 判断标准：必须完全没有 JS 相关文件，且至少包含一个 TS/TSX 源码文件\n return has_ts","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"限制条件：生成的代码文件中不得使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次对已有文件进行改动前，不要在同级目录内备份该文件修改前的内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有的数字在你生成的代码文件中必须通过定义常量来体现，坚决禁止直接使用具体的数字","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":5,"instruction":"现在的项目列表全是静态的，来加点高级交互。首先，在项目展示区最上方加几个分类标签（例如‘全部’、‘前端’、‘后端’），点击标签可以动态筛选并平滑过滤出对应的项目卡片。其次，点击任意一张项目卡片时，弹出一个带有深色遮罩层的模态框（Modal），里面展示该项目的放大版图片、详细介绍长文本以及一个‘访问网站’的模拟按钮。模态框右上角记得加个关闭按钮。对了，后续请始终遵循以下两项新要求：第一，替换此前要求所有数字必须通过定义常量体现的规则，在编写生成的代码文件时，务必直接使用数字，禁止为任何数字创建常量定义；第二，在你生成的所有代码文件中，注释内容必须统一使用中文编写。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令背景，界定可能生成的网站/TypeScript项目的常见代码文件扩展名\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css'}\n found_code_file = False\n \n # 如果工作目录不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的顶行内容\n first_line = f.readline()\n \n # 检查第一行是否包含强制要求中的特定编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果遇到文件无法读取、编码错误等异常情况，直接判定为未满足要求\n return False\n \n # 只有当项目中存在至少一个代码文件并且全满足时，才返回True\n return found_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 目标注释字符串\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 用户明确要求使用 TypeScript 进行编写，也可能包含配套的样式或模板文件\n code_extensions = ['.ts', '.tsx', '.css', '.scss', '.html', '.js', '.jsx']\n found_any_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖目录和版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n # 判断是否为常见的代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n found_any_file = True\n file_path = os.path.join(root, file)\n try:\n has_comment = False\n with open(file_path, 'r', encoding='utf-8') as f:\n # 只需检查文件开头的几行，此处取前 10 行\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n if target_comment in line:\n has_comment = True\n break\n \n # 如果任一代码文件没有发现目标注释，则要求未满足\n if not has_comment:\n return False\n except Exception:\n # 如果遇到无法读取的文件，为了严谨起见，直接忽略或视为不影响总体验证\n continue\n \n # 如果项目中没有任何代码文件，或者所有的代码文件都正确包含了注释，返回对应的布尔结果\n return found_any_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 遍历项目文件\n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的构建和依赖目录\n parts = root.replace('\\\\', '/').split('/')\n if any(ignored in parts for ignored in ['node_modules', '.git', 'dist', 'build', 'out']):\n continue\n \n for file in files:\n # 根据用户指令第1轮可推断项目使用TypeScript，这里包含常见的JS/TS扩展名\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 第一步：词法简化。剔除注释和字符串内容，以免其中的大括号干扰匹配。\n # 严格保留 \\n 确保行号不变。\n clean_code_chars = []\n i = 0\n n = len(code)\n state = 'NORMAL'\n while i < n:\n if state == 'NORMAL':\n if code[i:i+2] == '//':\n state = 'LINE_COMMENT'\n clean_code_chars.append(' ')\n i += 2\n elif code[i:i+2] == '/*':\n state = 'BLOCK_COMMENT'\n clean_code_chars.append(' ')\n i += 2\n elif code[i] == \"'\":\n state = 'STRING1'\n clean_code_chars.append(' ')\n i += 1\n elif code[i] == '\"':\n state = 'STRING2'\n clean_code_chars.append(' ')\n i += 1\n elif code[i] == '`':\n state = 'STRING3'\n clean_code_chars.append(' ')\n i += 1\n else:\n clean_code_chars.append(code[i])\n i += 1\n elif state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state = 'NORMAL'\n clean_code_chars.append('\\n')\n else:\n clean_code_chars.append(' ')\n i += 1\n elif state == 'BLOCK_COMMENT':\n if code[i:i+2] == '*/':\n state = 'NORMAL'\n clean_code_chars.append(' ')\n i += 2\n else:\n if code[i] == '\\n':\n clean_code_chars.append('\\n')\n else:\n clean_code_chars.append(' ')\n i += 1\n elif state in ['STRING1', 'STRING2', 'STRING3']:\n if state == 'STRING1' and code[i] == \"'\":\n state = 'NORMAL'\n clean_code_chars.append(' ')\n i += 1\n elif state == 'STRING2' and code[i] == '\"':\n state = 'NORMAL'\n clean_code_chars.append(' ')\n i += 1\n elif state == 'STRING3' and code[i] == '`':\n state = 'NORMAL'\n clean_code_chars.append(' ')\n i += 1\n elif code[i] == '\\\\':\n if i + 1 < n:\n if code[i+1] == '\\n':\n clean_code_chars.append(' \\n')\n else:\n clean_code_chars.append(' ')\n else:\n clean_code_chars.append(' ')\n i += 2\n elif code[i] == '\\n':\n clean_code_chars.append('\\n')\n i += 1\n else:\n clean_code_chars.append(' ')\n i += 1\n \n clean_text = \"\".join(clean_code_chars)\n \n # 第二步：找到所有的括号位置并区分普通块与函数块\n brace_positions = []\n for idx, char in enumerate(clean_text):\n if char == '{':\n p = idx - 1\n while p >= 0 and clean_text[p].isspace():\n p -= 1\n if p < 0:\n brace_positions.append(('OTHER_START', idx))\n continue\n \n last_char = clean_text[p]\n is_func = False\n \n # 匹配箭头函数\n if p >= 1 and clean_text[p-1:p+1] == '=>':\n is_func = True\n else:\n # 判断大括号前是否为合法的函数结束符号(如 ), 字母, >, 等)\n valid_end_chars = set(')>]|&}')\n if last_char in valid_end_chars or last_char.isalnum() or last_char in '_$':\n temp_p = p\n found_paren = False\n brace_count = 0\n bracket_count = 0\n \n # 回溯寻找匹配的参数闭合列表 ')'\n while temp_p >= 0:\n c = clean_text[temp_p]\n if c == '}':\n brace_count += 1\n elif c == '{':\n brace_count -= 1\n elif c == ']':\n bracket_count += 1\n elif c == '[':\n bracket_count -= 1\n elif c == ')':\n if brace_count == 0 and bracket_count == 0:\n found_paren = True\n break\n elif c in ';=':\n if brace_count == 0 and bracket_count == 0:\n break\n temp_p -= 1\n \n if found_paren:\n paren_count = 0\n k = temp_p\n # 配对找到对应的 '('\n while k >= 0:\n if clean_text[k] == ')':\n paren_count += 1\n elif clean_text[k] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n k -= 1\n \n if k >= 0:\n q = k - 1\n while q >= 0 and clean_text[q].isspace():\n q -= 1\n if q >= 0:\n word_end = q + 1\n while q >= 0 and (clean_text[q].isalnum() or clean_text[q] in '_$'):\n q -= 1\n word = clean_text[q+1:word_end]\n \n # 如果 '(' 前的关键字非控制流，即认为是函数\n if word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n is_func = True\n\n if is_func:\n brace_positions.append(('FUNC_START', idx))\n else:\n brace_positions.append(('OTHER_START', idx))\n \n elif char == '}':\n brace_positions.append(('END', idx))\n \n # 第三步：使用栈来进行结构匹配，并验证函数长度\n stack = []\n for item in brace_positions:\n if item[0] in ['FUNC_START', 'OTHER_START']:\n stack.append(item)\n elif item[0] == 'END':\n if stack:\n start_item = stack.pop()\n # 仅统计函数的长度\n if start_item[0] == 'FUNC_START':\n start_idx = start_item[1]\n end_idx = item[1]\n start_line = clean_text.count('\\n', 0, start_idx) + 1\n end_line = clean_text.count('\\n', 0, end_idx) + 1\n # 检查函数行数是否超过 40 行\n if (end_line - start_line + 1) > 40:\n return False\n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 判断回复中是否包含被禁止使用的字“我”\n if '我' in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断 AI 的最后一轮回复中是否遵守了禁止使用“你”的规则。\n \"\"\"\n # 如果回复中存在“你”字，则说明违反了禁止称呼和替换的规则\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n has_ts = False\n has_js = False\n \n # 需要忽略的常见编译产物与依赖目录\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'coverage', 'out', '.next', '.cache'}\n \n # 允许部分特定前端构建工具和 linter 使用 JavaScript 配置文件的扩展名（豁免清单）\n allowed_js_configs = {\n '.eslintrc.js', '.eslintrc.cjs', 'postcss.config.js', \n 'tailwind.config.js', 'next.config.js', 'vite.config.js',\n 'jest.config.js', 'babel.config.js', 'rollup.config.js',\n 'webpack.config.js', 'vue.config.js', 'remix.config.js'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以排除不需要检查的目录和隐藏目录（如 .vscode, .idea 等）\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\n \n for file in files:\n if file.endswith(('.ts', '.tsx')):\n has_ts = True\n elif file.endswith(('.js', '.jsx', '.cjs', '.mjs')):\n # 豁免以 config.js/cjs/mjs 结尾的通用配置文件或在白名单中的工程化配置文件\n if file.endswith(('.config.js', '.config.cjs', '.config.mjs')) or file in allowed_js_configs:\n continue\n \n # 如果出现豁免清单以外的 JS 文件，则视为使用了 JavaScript 编写核心代码\n has_js = True\n break\n \n if has_js:\n break\n \n # 必须至少包含一个 ts/tsx 文件，并且没有任何非配置类的 js/jsx 文件\n return has_ts and not has_js","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"限制条件：生成的代码文件中不得使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次对已有文件进行改动前，不要在同级目录内备份该文件修改前的内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在编写生成的代码文件时，务必直接使用数字，禁止为任何数字创建常量定义","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":6,"instruction":"项目模块交互很棒。现在完善页面最底部的‘联系我’模块。我需要一个好看的联系表单，包含‘姓名’、‘邮箱’和‘留言内容’（多行文本框）三个输入框，以及一个‘发送留言’按钮。重点是加上前端的表单验证逻辑：如果漏填任意必填项，或者邮箱格式不正确，要在对应的输入框下方立即显示红色的错误提示语（提示语也要支持双语）。全部验证通过后点击发送，清空表单并弹出一个‘发送成功’的绿色提示框。对了，后续请始终遵循以下两项要求：第一，替换此前要求代码注释统一使用中文的规则，无论生成哪个代码文件，里面的注释都只能使用英文；第二，替换此前修改已有文件前不需要备份的规则，只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_encoding_declaration(response: str, workspace_path: str) -> bool:\n # 常见的前端源码相关文件扩展名\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.scss', '.less', '.html', '.vue'}\n # 过滤掉非手写的依赖文件夹或构建产物文件夹\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'out', '.next', 'coverage'}\n \n found_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 判断顶行（第一行）是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果无法读取或是其他读取异常，视为未能满足生成有效代码的要求\n return False\n \n # 如果项目中没有任何代码文件，说明未生成有效代码结构，返回 False\n return found_code_files","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 前端开发中常见的代码文件扩展名\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css', '.scss', '.less', '.vue'}\n \n file_checked = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的非业务代码依赖与配置目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n if not lines:\n continue\n\n file_checked = True\n \n # 检查文件开头（前10行内）是否存在目标注释\n has_comment = False\n for line in lines[:10]:\n if target_comment in line:\n has_comment = True\n break\n \n # 如果有任意一个代码文件的开头不包含该注释，则不满足要求\n if not has_comment:\n return False\n\n # 必须至少检查过一个文件，否则说明没有生成相关代码文件\n if not file_checked:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return True\n\n # 替换函数：将注释和字符串字面量中的非换行字符均转化为空格\n # 目的是去掉花括号干扰并保证清洗后的代码行号结构和原代码完全对齐\n def replacer(m):\n match_str = m.group(0)\n return ''.join(['\\n' if ch == '\\n' else ' ' for ch in match_str])\n\n # 匹配规则：双引号字符串、单引号字符串、反引号字符串、单行注释、多行注释\n pattern = re.compile(\n r'\"(?:\\\\.|[^\\\\\"])*\"|\\'(?:\\\\.|[^\\\\\\'])*\\'|`(?:\\\\.|[^\\\\`])*`|//.*?$|/\\*.*?\\*/',\n re.DOTALL | re.MULTILINE\n )\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(('.ts', '.tsx', '.js', '.jsx')):\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n\n clean_text = pattern.sub(replacer, code)\n stack = []\n \n for i, char in enumerate(clean_text):\n if char == '{':\n start = i - 1\n depth = 0\n # 往前寻找这段声明部分的起始边界\n while start >= 0:\n ch = clean_text[start]\n if ch == ')':\n depth += 1\n elif ch == '(':\n depth -= 1\n \n # 在非圆括号嵌套中，遇到语句终结符或作用域界限代表当前声明的起点界限\n if depth == 0 and ch in '{};':\n break\n start -= 1\n \n decl_start = start + 1\n declaration = clean_text[decl_start:i]\n decl_stripped = declaration.strip()\n \n is_function = False\n \n # 1. 包含 => (箭头函数) 或 function 关键字\n if '=>' in declaration or 'function' in declaration:\n is_function = True\n # 2. 如果不含，但以 ')' 结尾，这极有可能是方法如 foo(a, b) {\n elif decl_stripped.endswith(')'):\n d = 0\n idx = len(decl_stripped) - 1\n while idx >= 0:\n if decl_stripped[idx] == ')':\n d += 1\n elif decl_stripped[idx] == '(':\n d -= 1\n if d == 0:\n break\n idx -= 1\n \n if idx >= 0:\n prefix = decl_stripped[:idx].strip()\n # 提取括号前的最后一个标识词\n words = re.findall(r'[a-zA-Z_]\\w*', prefix)\n if words:\n last_word = words[-1]\n # 排除常见的语法控制流关键字\n if last_word not in ('if', 'for', 'while', 'catch', 'switch', 'with'):\n is_function = True\n \n # 压栈：记录当前花括号的位置、它是否属于一个函数、以及函数声明语句的起始索引\n stack.append((i, is_function, decl_start))\n \n elif char == '}':\n if stack:\n brace_idx, is_function, decl_start = stack.pop()\n if is_function:\n # 根据提取出来的完整函数范围计算其实际跨越的行数\n func_text = clean_text[decl_start:i+1].strip()\n lines_count = func_text.count('\\n') + 1\n if lines_count > 40:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断AI助手的回复中是否杜绝使用了\"我\"。\n 如果回复中包含\"我\"字，则判定为不满足要求返回False；否则返回True。\n \"\"\"\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI的回复中是否遵守了禁止使用“你”称呼用户的要求。\n \"\"\"\n # 剔除中文中极少数包含“你”字但并非指代用户的固定词汇，避免误判\n clean_response = response.replace(\"迷你\", \"\")\n \n # 检查回复内容中是否包含“你”或“您”\n if \"你\" in clean_response or \"您\" in clean_response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def check_typescript_only_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n has_ts_or_tsx = False\n has_unauthorized_js_or_jsx = False\n \n # 需要跳过的无关文件夹\n exclude_dirs = {'node_modules', 'dist', 'build', 'out', '.git', '.vscode', '.idea', 'coverage'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以跳过特定目录，不再向下遍历\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 检测到 TS/TSX 代码文件\n if ext in ['.ts', '.tsx']:\n has_ts_or_tsx = True\n \n # 检测到 JS/JSX 代码文件\n elif ext in ['.js', '.jsx']:\n file_lower = file.lower()\n # 豁免常见的 JS 配置文件\n is_config_file = (\n 'config' in file_lower or \n 'rc.js' in file_lower or \n file_lower.startswith('.eslint') or \n file_lower.startswith('.prettier')\n )\n \n # 如果不是被豁免的配置文件，则视为违反了 TypeScript 限定的要求\n if not is_config_file:\n has_unauthorized_js_or_jsx = True\n\n # 满足条件：项目中有 TypeScript 源码，且没有未授权的 JavaScript 源码\n return has_ts_or_tsx and not has_unauthorized_js_or_jsx","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"限制条件：生成的代码文件中不得使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在编写生成的代码文件时，务必直接使用数字，禁止为任何数字创建常量定义","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"无论生成哪个代码文件，里面的注释都只能使用英文","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":7,"instruction":"最后一步，我们来优化响应式体验和整体导航体验。首先做移动端适配：当屏幕宽度小于768px时，顶部导航栏的文字菜单请折叠隐藏，显示为一个‘汉堡图标’，点击后能弹出下拉菜单；同时所有的分列布局（如首屏的左右布局、项目的网格布局）都要自动变为单列展示。其次，点击顶部导航栏的任意菜单项时，页面不要生硬跳转，而是带有平滑滚动动画（Smooth Scroll）滑动到下方对应的模块区域。对了，后续请始终遵循以下两项要求：第一，替换此前要求代码中务必直接使用数字、禁止定义数字常量的规则，你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用；第二，保持你生成的代码文件里每个函数的圈复杂度均在3以内。","instruction_following_checklist":[{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n file_checked = False\n # 根据用户指令中“仅使用 TypeScript 进行编写”的要求，限定检查的文件扩展名\n valid_extensions = {'.ts', '.tsx'}\n # 排除常见的无关或依赖目录\n exclude_dirs = {'node_modules', '.git', 'dist', 'build'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1]\n if ext in valid_extensions:\n file_checked = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的第一行（顶行）\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件读取发生异常，视为不满足\n return False\n\n # 只有当至少检查了一个文件且全部满足要求时，才返回 True\n return file_checked","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.html'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的依赖与构建目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.next')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 仅需读取前几行进行校验即可\n lines = [f.readline() for _ in range(5)]\n \n # 检查文件头部（取前5行以兼容第一行的特定编码声明）\n is_comment_found = False\n for line in lines:\n if not line: # 遇到文件末尾\n break\n if target_comment in line:\n is_comment_found = True\n break\n \n # 如果某个代码文件开头没有找到该特定注释，则不满足要求\n if not is_comment_found:\n return False\n \n except Exception:\n # 如果遇到文件读取异常（如非法编码等），保守判定为不满足\n return False\n \n # 如果项目中没有任何对应的代码文件，说明没有生成，无法满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return True\n\n # 替换注释和字符串为空白内容（保持原有换行符，以便精确计算行数）\n def replacer(match):\n s = match.group(0)\n return '\\n' * s.count('\\n') + ' ' * (len(s) - s.count('\\n'))\n\n pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"|`(?:\\\\.|[^\\\\`])*`',\n re.DOTALL | re.MULTILINE\n )\n\n # 移除 prefix 中的局部平衡大括号，防止解构赋值对象等带来的正则干扰\n def remove_balanced_braces(s: str) -> str:\n while True:\n new_s = re.sub(r'\\{[^{}]*\\}', 'X', s)\n if new_s == s:\n break\n s = new_s\n return s\n\n blacklist = {'if', 'for', 'while', 'switch', 'catch', 'with'}\n\n # 正则表达式用于匹配在 '{' 之前的不同类型的函数签名\n arrow_pattern = re.compile(r'=>\\s*$')\n anon_func_pattern = re.compile(r'\\bfunction\\b\\s*(?:\\w+\\s*)?(?:<[^;{=]*>\\s*)?\\([^;{]*\\)\\s*(?::[^;{=]*)?$')\n func_pattern = re.compile(r'\\b([a-zA-Z_]\\w*)\\s*(?:<[^;{=]*>\\s*)?\\([^;{]*\\)\\s*(?::[^;{=]*)?$')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(('.ts', '.tsx', '.js', '.jsx')):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n clean_content = pattern.sub(replacer, content)\n \n # 计算通过 idx 到行号的映射\n lines = clean_content.split('\\n')\n idx_to_line = []\n for r, line in enumerate(lines):\n idx_to_line.extend([r + 1] * (len(line) + 1))\n \n stack = []\n blocks = []\n \n # 找到所有的成对的大括号，计算块行数\n for i, char in enumerate(clean_content):\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start_i = stack.pop()\n start_line = idx_to_line[start_i]\n end_line = idx_to_line[i]\n \n # 如果块的行数超过了 40 行，认为是一个过长的块\n if end_line - start_line + 1 > 40:\n blocks.append(start_i)\n\n # 对所有的大块进行函数特征判断\n for start_i in blocks:\n prefix_start = max(0, start_i - 500)\n prefix = clean_content[prefix_start:start_i]\n \n prefix_clean = remove_balanced_braces(prefix)\n \n # 若是箭头函数\n if arrow_pattern.search(prefix_clean):\n return False\n \n # 若是 function 关键字声明的函数\n if anon_func_pattern.search(prefix_clean):\n return False\n \n # 若是类方法或是带有名称的普通函数形式\n match = func_pattern.search(prefix_clean)\n if match:\n word = match.group(1)\n if word not in blacklist:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"回答时，必须绝对杜绝使用\"我\"来自称，而是永远使用\"小岚\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 检查助手的回复文本中是否包含“我”字\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"小主人\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含被禁止的“你”字\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"强制要求：仅使用 TypeScript 进行编写","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 1. 检查回复中的代码块，如果存在 JS/JSX 代码块，则直接返回 False\n js_pattern = r'```(javascript|js|jsx)\\b'\n if re.search(js_pattern, response, re.IGNORECASE):\n return False\n \n has_ts = False\n \n # 检查回复中是否包含 TS/TSX 代码块作为备用依据\n ts_pattern = r'```(typescript|ts|tsx)\\b'\n if re.search(ts_pattern, response, re.IGNORECASE):\n has_ts = True\n\n # 2. 检查工作目录中的实际文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 排除非源代码相关的常见文件夹以防止误判\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', 'out', 'coverage'}\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果发现 js/jsx 文件，需要判断是否属于常见配置文件\n if ext in ['.js', '.jsx']:\n # 容错处理：配置文件（例如 tailwind.config.js, .eslintrc.js, .prettierrc.js）可以宽容放行\n is_config = (\n file.endswith('.config.js') or \n file.startswith('.eslintrc') or \n file.endswith('rc.js')\n )\n # 如果不是这些常见的配置文件，则说明违背了“仅使用 TypeScript 进行编写”的指令\n if not is_config:\n return False\n \n # 如果发现 ts/tsx 文件，则将 has_ts 标记为 True\n elif ext in ['.ts', '.tsx']:\n has_ts = True\n\n return has_ts","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"限制条件：生成的代码文件中不得使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"无论生成哪个代码文件，里面的注释都只能使用英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个个人作品集网站的基础骨架，后续请始终遵循以下两项要求：第一，强制要求：仅使用 TypeScript 进行编写；第二，你输出的代码文件中严格禁止出现递归调用。需要一个固定在顶部的导航栏（包含‘首页’、‘项目’、‘联系我’这几个文字占位），一个主内容区域（在页面上垂直排列三个对应模块的空白区域，每个区域加点不同的背景色区分一下边界），以及一个简单的页脚。先不用实现具体的页面内容，整体布局清晰、居中对齐即可。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义可能生成的代码文件扩展名\\n code_extensions = ('.ts', '.tsx', '.js', '.jsx', '.html', '.css')\\n has_code_file = False\\n \\n # 如果工作区路径不存在，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除前端项目中常见的依赖文件夹和隐藏文件夹\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n if file.endswith(code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的第一行（顶行）\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件为空或者读取失败，判定为不满足要求\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 结合题目指令背景，定义可能的网站骨架相关代码文件拓展名\\n code_extensions = {'.ts', '.tsx', '.html', '.css', '.js', '.jsx'}\\n # 要求中明确规定的标注内容\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如.git等）以及node_modules等无需检查的目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n # 仅处理匹配上的代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 读取代码文件的前10行\\n lines = [f.readline() for _ in range(10)]\\n except Exception:\\n return False\\n \\n # 检查前10行中是否有一行包含了目标注释字符串\\n found_comment = any(target_comment in line for line in lines if line)\\n \\n # “针对所有由你生成的代码文件”，因此只要有一个文件没满足要求，就直接返回False\\n if not found_comment:\\n return False\\n\\n # 若成功验证所有文件且文件存在，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n def is_function_brace(code_str: str, start_index: int) -> bool:\\n # 从 '{' 所在的位置向前看\\n j = start_index - 1\\n while j >= 0 and code_str[j].isspace():\\n j -= 1\\n if j < 0:\\n return False\\n \\n # 判断是否为箭头函数 '=> {'\\n if code_str[j] == '>':\\n if j - 1 >= 0 and code_str[j-1] == '=':\\n return True\\n \\n # 判断是否带有参数列表 '() {'\\n if code_str[j] == ')':\\n paren_level = 0\\n while j >= 0:\\n if code_str[j] == ')':\\n paren_level += 1\\n elif code_str[j] == '(':\\n paren_level -= 1\\n if paren_level == 0:\\n break\\n j -= 1\\n \\n if j < 0:\\n return False\\n \\n j -= 1\\n while j >= 0 and code_str[j].isspace():\\n j -= 1\\n \\n # 跳过可能的 TypeScript 泛型声明 <T>\\n if j >= 0 and code_str[j] == '>':\\n angle_level = 0\\n while j >= 0:\\n if code_str[j] == '>':\\n angle_level += 1\\n elif code_str[j] == '<':\\n angle_level -= 1\\n if angle_level == 0:\\n break\\n j -= 1\\n if j < 0:\\n return False\\n j -= 1\\n while j >= 0 and code_str[j].isspace():\\n j -= 1\\n \\n # 提取紧挨着括号的关键字或标识符\\n word = \\\"\\\"\\n while j >= 0 and (code_str[j].isalnum() or code_str[j] in '_$'):\\n word = code_str[j] + word\\n j -= 1\\n \\n # 排除常见的控制流关键字块\\n if word in (\\\"if\\\", \\\"for\\\", \\\"while\\\", \\\"catch\\\", \\\"switch\\\", \\\"with\\\"):\\n return False\\n \\n # 其余情况（包括 function 关键字，或类中的方法名等）均可视为函数代码块\\n return True\\n \\n return False\\n\\n def check_file_functions(code_str: str) -> bool:\\n i = 0\\n length = len(code_str)\\n line_num = 1\\n brace_level = 0\\n \\n in_string = False\\n string_char = ''\\n in_single_comment = False\\n in_multi_comment = False\\n \\n # 存放所有已识别为函数实体的起始层级和起始行号\\n active_functions = []\\n \\n while i < length:\\n char = code_str[i]\\n \\n if char == '\\\\n':\\n line_num += 1\\n if in_single_comment:\\n in_single_comment = False\\n i += 1\\n continue\\n \\n if in_single_comment:\\n i += 1\\n continue\\n \\n if in_multi_comment:\\n if char == '*' and i + 1 < length and code_str[i+1] == '/':\\n in_multi_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n if in_string:\\n if char == '\\\\\\\\':\\n i += 2\\n continue\\n if char == string_char:\\n in_string = False\\n i += 1\\n continue\\n \\n if char in ('\\\"', \\\"'\\\", '`'):\\n in_string = True\\n string_char = char\\n i += 1\\n continue\\n \\n if char == '/' and i + 1 < length:\\n if code_str[i+1] == '/':\\n in_single_comment = True\\n i += 2\\n continue\\n elif code_str[i+1] == '*':\\n in_multi_comment = True\\n i += 2\\n continue\\n \\n if char == '{':\\n if is_function_brace(code_str, i):\\n # 识别为函数，记录开始行号与当前对应的括号层级\\n active_functions.append((line_num, brace_level))\\n brace_level += 1\\n i += 1\\n continue\\n \\n if char == '}':\\n brace_level -= 1\\n for k in range(len(active_functions)-1, -1, -1):\\n start_line, start_level = active_functions[k]\\n if start_level == brace_level:\\n # 当前层级的函数闭合，计算函数总行数\\n if line_num - start_line + 1 > 40:\\n return False\\n active_functions.pop(k)\\n i += 1\\n continue\\n \\n i += 1\\n \\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n valid_extensions = {'.ts', '.tsx', '.js', '.jsx', '.mts', '.mjs', '.cts', '.cjs'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n except Exception:\\n continue\\n \\n # 如果单个文件中出现超过40行的函数，立即返回 False\\n if not check_file_functions(code_str):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 排除由于用户指令要求导致的合理出现，即导航栏占位符“联系我”\\n cleaned_response = response.replace(\\\"联系我\\\", \\\"\\\")\\n \\n # 检查回复中是否还存在“我”字\\n if \\\"我\\\" in cleaned_response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否没有使用\\\"你\\\"。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_ts = False\\n # 定义禁止出现的其他编程语言后缀（主要包含 JS 及其它常见后端/脚本语言）\\n forbidden_exts = {\\n '.js', '.jsx', '.mjs', '.cjs',\\n '.py', '.java', '.c', '.cpp', '.cs',\\n '.rb', '.php', '.go', '.rs', '.kt', '.swift', \\n '.vb', '.lua', '.dart'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录、依赖目录及常见构建输出目录，避免误判\\n dirs[:] = [\\n d for d in dirs \\n if not d.startswith('.') and d not in ('node_modules', 'dist', 'build', 'out', 'coverage')\\n ]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in {'.ts', '.tsx'}:\\n has_ts = True\\n elif ext in forbidden_exts:\\n # 只要出现了其他编程语言（尤其是 JS 及其变体），即可视为不满足\\\"仅使用 TypeScript\\\"的要求\\n return False\\n \\n return has_ts\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好的，框架搭建得很清晰。接下来我们实现这个网站的核心需求：中英文双语支持。请在顶部导航栏的最右侧加上一个‘中/En’的切换按钮。点击这个按钮时，导航栏里的菜单项文字、页面主要区域的占位标题以及页脚的文本都要能跟着无缝切换成对应的语言。建议把所有的文案抽离出来统一维护。对了，后续请始终遵循以下两项新要求：第一，在对现有文件进行修改之前，请务必先调用相关工具读取文件内容；第二，使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作目录是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义前端项目中常用的 TypeScript 和 JavaScript 相关代码文件的扩展名\\n code_exts = {'.ts', '.tsx', '.js', '.jsx'}\\n # 忽略常见的无关目录，防止检查依赖库和构建产物\\n exclude_dirs = {'node_modules', 'dist', 'build', 'out', 'coverage'}\\n \\n checked_any_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以过滤隐藏文件夹和指定的排除目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的顶行（第一行）\\n first_line = f.readline()\\n \\n # 判断顶行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n \\n checked_any_code_file = True\\n except Exception:\\n # 如果文件由于编码错误或其他异常导致读取失败，判断为不满足要求\\n return False\\n \\n # 如果项目中找到了至少一个合规的代码文件且没有被前面的判断否决，则认为满足了要求\\n return checked_any_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 目标注释字符串\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 相关的代码文件扩展名\\n code_extensions = {\\\".ts\\\", \\\".tsx\\\", \\\".js\\\", \\\".jsx\\\", \\\".css\\\", \\\".html\\\"}\\n \\n # 需要忽略的文件夹，避免检查依赖包或编译产物中的文件\\n ignore_dirs = {\\\"node_modules\\\", \\\".git\\\", \\\"dist\\\", \\\"build\\\", \\\"coverage\\\"}\\n \\n has_code_file = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的前10行，判断是否包含特定的注释标注\\n header_lines = [f.readline() for _ in range(10)]\\n header_content = \\\"\\\".join(header_lines)\\n \\n if target_comment not in header_content:\\n return False\\n except Exception:\\n # 如果代码文件无法正常读取（如编码异常），为了严格起见，视为未满足要求\\n return False\\n\\n # 若至少存在一个代码文件，且全部都满足包含了特定的注释则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def mask_strings_and_comments(code: str) -> str:\\n # 将注释和字符串替换为空格及原来的换行符，以保证原始代码行数不变\\n def replacer(match):\\n s = match.group(0)\\n # 根据匹配片段包含的换行符数量保留 \\\\n，其它字符填充为空格\\n return '\\\\n' * s.count('\\\\n') + ' ' * (len(s) - s.count('\\\\n'))\\n \\n # 替换多行注释 /* ... */\\n code = re.sub(r'/\\\\*.*?\\\\*/', replacer, code, flags=re.DOTALL)\\n # 替换单行注释 // ...\\n code = re.sub(r'//.*', replacer, code)\\n # 替换反引号模板字符串 `...`\\n code = re.sub(r'`(?:\\\\\\\\`|[^`])*`', replacer, code, flags=re.DOTALL)\\n # 替换双引号字符串 \\\"...\\\"\\n code = re.sub(r'\\\"(?:\\\\\\\\\\\"|[^\\\"])*\\\"', replacer, code)\\n # 替换单引号字符串 '...'\\n code = re.sub(r\\\"'(?:\\\\\\\\'|[^'])*'\\\", replacer, code)\\n return code\\n\\n ts_files = []\\n # 查找所有的 TypeScript 文件 (.ts, .tsx)\\n for root, dirs, files in os.walk(workspace_path):\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n for file in files:\\n if file.endswith('.ts') or file.endswith('.tsx'):\\n ts_files.append(os.path.join(root, file))\\n \\n for filepath in ts_files:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 获取掩码替换后的代码并拆分行\\n masked_code = mask_strings_and_comments(code)\\n lines = masked_code.split('\\\\n')\\n \\n for i, line in enumerate(lines):\\n is_func = False\\n \\n # 判断是否是函数定义的开始行\\n if re.search(r'\\\\bfunction\\\\b', line):\\n is_func = True\\n elif re.search(r'=>\\\\s*\\\\{', line):\\n is_func = True\\n # 判断是否为类方法或简写方法：例如 public async myMethod(arg: string): void {\\n elif re.search(r'^\\\\s*(?:export\\\\s+)?(?:public\\\\s+|private\\\\s+|protected\\\\s+)?(?:static\\\\s+)?(?:async\\\\s+)?(?:[a-zA-Z0-9_]+\\\\s*)?\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^{]+)?\\\\s*\\\\{', line):\\n # 排除 if, for, while, catch, switch, else, return, new 等常见语法结构块关键字\\n if not re.search(r'\\\\b(if|for|while|catch|switch|else|return|new)\\\\b', line):\\n is_func = True\\n \\n if is_func:\\n char_stream = '\\\\n'.join(lines[i:])\\n first_brace_idx = char_stream.find('{')\\n \\n # 若找不到函数体左括号则跳过\\n if first_brace_idx == -1:\\n continue\\n \\n # 检查大括号前是否有分号（如果有，可能是类型声明，如 declare function）\\n pre_brace_text = char_stream[:first_brace_idx]\\n if ';' in pre_brace_text:\\n continue\\n \\n stack = 0\\n func_content = char_stream[first_brace_idx:]\\n for j, char in enumerate(func_content):\\n if char == '{':\\n stack += 1\\n elif char == '}':\\n stack -= 1\\n # 当花括号闭合时，即为一个完整的函数\\n if stack == 0:\\n matched_text = char_stream[:first_brace_idx + j + 1]\\n line_count = matched_text.count('\\\\n') + 1\\n if line_count > 40:\\n return False\\n break\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的最后一轮回复中是否杜绝使用了\\\"我\\\"作为自称。\\n 基于严格匹配原则，只要回复中出现\\\"我\\\"字，即视为违背规则。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的最后一轮回复中是否不包含“你”。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 如果没有使用“你”则返回True，否则返回False。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_ts = False\\n \\n # 允许的 JavaScript 配置文件列表，避免因为工程化配置导致误判\\n allowed_js_files = {\\n 'eslint.config.js', 'tailwind.config.js', 'postcss.config.js',\\n 'next.config.js', 'vite.config.js', 'babel.config.js',\\n 'jest.config.js', 'webpack.config.js', 'rollup.config.js',\\n '.eslintrc.js', 'prettier.config.js', 'stylelint.config.js',\\n 'commitlint.config.js', 'setupTests.js'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略包依赖和构建产物目录\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'dist', 'build', '.next', 'out', 'coverage'}]\\n \\n for file in files:\\n if file.endswith('.ts') or file.endswith('.tsx'):\\n has_ts = True\\n elif file.endswith('.js') or file.endswith('.jsx'):\\n # 若存在 .js/.jsx 文件，需判断其是否为配置文件。\\n # 如果不是上述白名单中的文件，且不是以 .config.js / rc.js 结尾，则视为违规使用了 JavaScript。\\n if file not in allowed_js_files and not file.endswith('.config.js') and not file.endswith('rc.js'):\\n return False\\n \\n # 仅当存在至少一个 TypeScript 文件且没有违规的 JavaScript 文件时返回 True\\n return has_ts\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"语言切换做得很顺滑，现在来填充‘首页’这个首屏区域。请把它设计成左右两列布局。左侧放一段欢迎语和简单的自我介绍文本（当然要支持双语切换），把我的名字加大加粗显示。在文本下方放置两个操作按钮：‘查看项目’和‘下载简历’。右侧放一个圆形的个人头像占位图。整体要在一个视口高度内垂直居中好看一些。对了，后续请始终遵循以下两项新要求：第一，限制条件：生成的代码文件中不得使用多层if；第二，针对你生成的代码，禁止执行任何ESLint检查。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.scss', '.less', '.html'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖和构建目录\\n for ignore_dir in ['node_modules', '.git', 'build', 'dist', '.next']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取失败，则视为不满足条件\\n return False\\n \\n # 如果找到了代码文件且均满足条件，则返回 True；否则返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_code_files_annotation(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_annotation = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 常见的前端代码文件扩展名\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css', '.scss', '.less', '.vue'}\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略项目中的无关依赖或版本控制目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 检查“开头另起一行”的注释，读取前 20 行足以覆盖\\n lines = [f.readline() for _ in range(20)]\\n except Exception:\\n # 如果由于编码或其他问题无法读取，视作未满足\\n return False\\n \\n # 检查该文件前几行是否包含目标注释字符串\\n found_in_file = False\\n for line in lines:\\n if not line:\\n break\\n if target_annotation in line:\\n found_in_file = True\\n break\\n \\n # 若任意一个代码文件缺少该注释，则整体不满足\\n if not found_in_file:\\n return False\\n \\n # 如果项目中没有任何目标代码文件，同样视为不满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 优化的无回溯灾难的正则表达式：匹配注释以及多/单/反引号字符串\\n # \\\\1: 多行注释\\n # \\\\2: 单行注释\\n # \\\\3: 双引号字符串\\n # \\\\4: 单引号字符串\\n # \\\\5: 模板字符串（反引号）\\n pattern = r'(/\\\\*.*?\\\\*/)|(//[^\\\\n]*)|(\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\")|(\\\\'[^\\\\'\\\\\\\\]*(?:\\\\\\\\.[^\\\\'\\\\\\\\]*)*\\\\')|(`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`)'\\n\\n def replacer(match):\\n # 将匹配到的字符串或者注释替换为相同数量的换行符\\n # 以确保后续 split('\\\\n') 产生的代码行号和原文件保持一致\\n return '\\\\n' * match.group(0).count('\\\\n')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关目录，提高遍历效率\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.next', 'out')]\\n\\n for file in files:\\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 移除所有字符串和注释，避免花括号以及关键字的误匹配\\n content = re.sub(pattern, replacer, content, flags=re.DOTALL)\\n lines = content.split('\\\\n')\\n\\n for i, line in enumerate(lines):\\n is_func = False\\n \\n # 识别是否为函数的起点\\n if re.search(r'\\\\bfunction\\\\b', line):\\n is_func = True\\n elif re.search(r'=>', line):\\n is_func = True\\n elif re.search(r'\\\\b\\\\w+\\\\s*\\\\([^)]*\\\\)[^;{]*\\\\{', line):\\n # 类方法 / 组件对象方法，同时排除常见的控制流和关键字\\n if not re.search(r'\\\\b(if|for|while|switch|catch|else|return|with)\\\\b', line):\\n is_func = True\\n\\n if is_func:\\n open_braces = 0\\n started = False\\n start_line = i\\n end_line = i\\n\\n # 从该行往下寻找配对的花括号闭合位置\\n for j in range(i, len(lines)):\\n open_braces += lines[j].count('{')\\n open_braces -= lines[j].count('}')\\n\\n if '{' in lines[j]:\\n started = True\\n\\n # 确认完整的函数块被闭合\\n if started and open_braces <= 0:\\n end_line = j\\n break\\n\\n if started:\\n func_len = end_line - start_line + 1\\n # 校验行数是否超过要求\\n if func_len > 40:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 移除 markdown 中的多行代码块，避免误判代码里正常生成的文本\\n text_without_code = re.sub(r'```[\\\\s\\\\S]*?```', '', response)\\n \\n # 移除 markdown 中的行内代码块\\n text_without_code = re.sub(r'`[^`]*`', '', text_without_code)\\n \\n # 检查剩余的自然语言回复中是否使用了“我”\\n if '我' in text_without_code:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否满足：禁止用\\\"你\\\"称呼用户，必须一律将其替换为\\\"小主人\\\"。\\n 判断逻辑：\\n 只要回复中包含了\\\"你\\\"字，就视为没有完全满足要求（没有进行替换或使用了违禁词）。\\n 如果没有包含\\\"你\\\"字，则视为满足要求。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify_typescript_only(response: str, workspace_path: str) -> bool:\\n has_ts = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除编译输出、依赖及版本控制相关的目录，重点检查源代码文件\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build', '.next', 'out']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n # 检测到 TypeScript 源码文件\\n if ext in ['.ts', '.tsx']:\\n has_ts = True\\n \\n # 若检测到 JS 相关的源文件\\n elif ext in ['.js', '.jsx', '.cjs', '.mjs']:\\n is_root = (root == workspace_path)\\n # 检查是否为常见且不可避免的配置文件（文件名包含config、以.开头，或含有rc.）\\n is_config = ('config' in file.lower() or 'rc.' in file or file.startswith('.'))\\n \\n # 若文件处于根目录且是配置文件类型，予以豁免\\n if is_root and is_config:\\n continue\\n # Eslint 等构建工具也往往必须使用 js 作为配置类型\\n if is_root and 'eslint' in file.lower():\\n continue\\n \\n # 除根目录的豁免情况外，生成或包含了其他的 JS 文件均视为违规\\n return False\\n \\n # 对于采用单文件组件形式的框架（如 Vue, Svelte）\\n elif ext in ['.vue', '.svelte']:\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 查找单文件中的所有 script 标签\\n script_tags = re.findall(r'<script\\\\b([^>]*)>', content, flags=re.IGNORECASE)\\n for attrs in script_tags:\\n # 确保每一个 script 标签内部都显式声明了 lang=\\\"ts\\\"\\n if not re.search(r'''lang\\\\s*=\\\\s*(['\\\"]?)ts\\\\1''', attrs, flags=re.IGNORECASE):\\n return False\\n if script_tags:\\n has_ts = True\\n except Exception:\\n pass\\n \\n # 对于 HTML 文件，严防在内部利用 script 标签原生编写大段 JavaScript 逻辑\\n elif ext == '.html':\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n scripts = re.findall(r'<script\\\\b([^>]*)>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n for attrs, script_content in scripts:\\n # 如果并非利用 src= 引入，且内部含有较长的内联代码，认定为使用了JS编写\\n if 'src=' not in attrs.lower() and len(script_content.strip()) > 50:\\n return False\\n except Exception:\\n pass\\n \\n # 只有当项目中存在 TS 代码并且在此过程中未检测到任何违规 JS 编写情况时才判定达成要求\\n return has_ts\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"针对你生成的代码，禁止执行任何ESLint检查\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"限制条件：生成的代码文件中不得使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"为了让自我介绍部分更吸引人，我想在左侧欢迎语和自我介绍之间，增加一行动态显示的职位/技能标签（比如：前端工程师、全栈开发者、UI设计师等）。请用代码实现一个‘打字机’特效，每隔几秒钟自动打字输出下一个词，并无限循环播放。注意，切换为英文页面时，打字机里输出的词汇也必须变成对应的英文。对了，后续请始终遵循以下两项要求：第一，在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份；第二，取消之前禁止ESLint检查的要求，后续请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令中提到的前端项目和TypeScript要求推断可能的代码文件类型\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉无需检查的隐藏目录和依赖目录\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的顶行（第一行）\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果遇到文件无法读取等异常，均视为验证失败\\n return False\\n \\n # 若项目中至少存在一个代码文件且全数符合要求，则要求满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作空间路径不存在，则直接判定为不满足\\n if not os.path.exists(workspace_path):\\n return False\\n\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义前端项目中典型的源代码文件扩展名\\n code_extensions = {\\\".ts\\\", \\\".tsx\\\", \\\".js\\\", \\\".jsx\\\", \\\".html\\\", \\\".css\\\", \\\".scss\\\", \\\".less\\\"}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖库、版本控制等无需检查的目录\\n if \\\"node_modules\\\" in root or \\\".git\\\" in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n \\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于特殊原因无法读取（如编码问题），在严格审核中视为不通过\\n return False\\n \\n # 在文件开头的10行内寻找是否包含特定的注释标注文本\\n found = False\\n for line in lines[:10]:\\n if target_text in line:\\n found = True\\n break\\n \\n # 只要有任何一个生成的代码文件缺失该注释，则整个要求验证失败\\n if not found:\\n return False\\n\\n # 若至少检查过一个代码文件且全都包含了所需注释，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def remove_strings_and_comments(code: str) -> str:\\n result = []\\n i = 0\\n n = len(code)\\n in_single_quote = False\\n in_double_quote = False\\n in_backtick = False\\n in_single_comment = False\\n in_multi_comment = False\\n \\n while i < n:\\n if in_single_comment:\\n if code[i] == '\\\\n':\\n in_single_comment = False\\n result.append(code[i])\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_multi_comment:\\n if code[i:i+2] == '*/':\\n in_multi_comment = False\\n result.append(' ')\\n i += 2\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_single_quote:\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif code[i] == \\\"'\\\":\\n in_single_quote = False\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n\\n if in_double_quote:\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n in_double_quote = False\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_backtick:\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif code[i] == '`':\\n in_backtick = False\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if code[i:i+2] == '//':\\n in_single_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n if code[i:i+2] == '/*':\\n in_multi_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n if code[i] == \\\"'\\\":\\n in_single_quote = True\\n result.append(' ')\\n i += 1\\n continue\\n if code[i] == '\\\"':\\n in_double_quote = True\\n result.append(' ')\\n i += 1\\n continue\\n if code[i] == '`':\\n in_backtick = True\\n result.append(' ')\\n i += 1\\n continue\\n \\n result.append(code[i])\\n i += 1\\n \\n return \\\"\\\".join(result)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除前端常见构建或依赖目录，防止混入巨量的压缩代码干扰检查\\n dirs[:] = [d for d in dirs if d not in ('node_modules', 'dist', 'build', '.git', '.next')]\\n for file in files:\\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n clean_code = remove_strings_and_comments(content)\\n \\n line_num = 1\\n stack = []\\n \\n for i, char in enumerate(clean_code):\\n if char == '\\\\n':\\n line_num += 1\\n continue\\n \\n if char == '{':\\n prev_text = clean_code[:i].rstrip()\\n tail = prev_text[-200:]\\n is_func = False\\n \\n if tail.endswith('=>'):\\n is_func = True\\n elif re.search(r'\\\\)\\\\s*(?::[^{;=]*\\\\s*)?$', tail):\\n # 排除常见的控制流关键字引领的代码块\\n if re.search(r'\\\\b(if|for|while|switch|catch)\\\\s*\\\\([^}{]*$', tail):\\n is_func = False\\n else:\\n is_func = True\\n elif re.search(r'\\\\bfunction\\\\b[^{;=]*$', tail):\\n is_func = True\\n \\n stack.append(('{', line_num, is_func))\\n \\n elif char == '(':\\n prev_text = clean_code[:i].rstrip()\\n tail = prev_text[-200:]\\n is_func = False\\n if tail.endswith('=>'):\\n is_func = True\\n stack.append(('(', line_num, is_func))\\n \\n elif char == '}':\\n idx = len(stack) - 1\\n while idx >= 0 and stack[idx][0] != '{':\\n idx -= 1\\n if idx >= 0:\\n bracket, start_line, is_func = stack[idx]\\n # 如果该块被判定为函数且行数超过 40 行，则验证失败\\n if is_func and (line_num - start_line + 1 > 40):\\n return False\\n stack = stack[:idx]\\n \\n elif char == ')':\\n idx = len(stack) - 1\\n while idx >= 0 and stack[idx][0] != '(':\\n idx -= 1\\n if idx >= 0:\\n bracket, start_line, is_func = stack[idx]\\n # 如果该块被判定为隐式返回的箭头函数且行数超过 40 行，则验证失败\\n if is_func and (line_num - start_line + 1 > 40):\\n return False\\n stack = stack[:idx]\\n\\n # 如果所有扫描的文件中不存在违反该要求的情况，则验证成功\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 1. 去除 Markdown 代码块的内容\\n # 防止因个人作品集代码中的自我介绍（如“我是一名前端工程师”）包含“我”而导致误判\\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\\n \\n # 2. 排除由于用户需求明确指定的包含“我”的页面元素文案\\n # 例如导航栏的“联系我”模块，AI在介绍实现过程时可能会提到该词\\n text_clean = text_without_code.replace(\\\"联系我\\\", \\\"\\\")\\n \\n # 3. 检查非代码部分是否仍然包含“我”字\\n # 如果依然包含，说明未能做到“绝对杜绝使用‘我’来自称”\\n if \\\"我\\\" in text_clean:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 剔除可能存在的含“你”字的非代词词汇（如“迷你”），避免误判\\n temp_resp = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 判断回复中是否仍然存在“你”字\\n if \\\"你\\\" in temp_resp:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n has_ts = False\\n has_business_js = False\\n\\n # 需要忽略的常见非源码目录\\n ignore_dirs = {'node_modules', 'dist', 'build', '.git', '.next', 'out'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤忽略的目录，避免深入遍历\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 记录是否找到了 TypeScript 文件\\n if ext in ['.ts', '.tsx']:\\n has_ts = True\\n \\n # 检查是否存在 JavaScript 文件\\n if ext in ['.js', '.jsx', '.mjs', '.cjs']:\\n is_config = False\\n \\n # 豁免常见的配置文件\\n if file.endswith('config.js') or file.endswith('config.mjs') or file.endswith('config.cjs'):\\n is_config = True\\n elif file.startswith('.') and (file.endswith('rc.js') or file.endswith('rc.cjs')):\\n is_config = True\\n elif file.startswith('eslint.config.'):\\n is_config = True\\n \\n # 如果不是配置文件，则意味着有 JS 编写的业务代码\\n if not is_config:\\n has_business_js = True\\n\\n # 满足要求的前提是：项目中至少包含 TS 代码，且不存在 JS 业务代码\\n return has_ts and not has_business_js\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"限制条件：生成的代码文件中不得使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"太酷了。下一步我们来做‘项目’模块的展示区。请用网格布局（Grid）在这里放置4到6张项目卡片。每张卡片需要有：顶部的项目缩略占位图、项目标题、两行简短的项目介绍文本，以及在底部排列几个圆角的技术栈标签（比如 React, Vue, Node.js）。所有文本必须接入咱们之前做好的中英双语系统。对了，后续请始终遵循以下两项要求：第一，每次对已有文件进行改动前，不要在同级目录内备份该文件修改前的内容，替换此前要求修改前备份的规则；第二，所有的数字在你生成的代码文件中必须通过定义常量来体现，坚决禁止直接使用具体的数字。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目中常见的代码文件扩展名\\n # 注意：排除了 json 文件，因为 JSON 格式不支持添加带注释的编码声明\\n code_extensions = {\\n '.ts', '.tsx', '.js', '.jsx', \\n '.html', '.css', '.scss', '.less',\\n '.vue', '.svelte'\\n }\\n \\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除前端工程中常见的不需要人工/AI检查或生成的目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', 'out')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n files_checked += 1\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果发生文件读取错误或编码不兼容，也判定为不满足\\n return False\\n \\n # 如果没有找到任何代码文件，则无法判定为满足要求\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 目标注释字符串\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 常见的代码文件扩展名集合（根据前端 TypeScript 项目推断）\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.scss', '.html', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不属于AI生成代码的常规忽略目录\\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', 'dist', 'build', 'out']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 只需检查开头的内容即可，这里读取前10行\\n lines = [next(f) for _ in range(10)]\\n except StopIteration:\\n pass # 文件不足10行\\n except Exception:\\n continue # 遇到读取异常则跳过该文件\\n \\n has_code_file = True\\n found_comment = False\\n \\n # 检查文件开头的行中是否包含目标注释标注\\n for line in lines:\\n if target_comment in line:\\n found_comment = True\\n break\\n \\n # 如果某个代码文件未包含该注释，则判定为失败\\n if not found_comment:\\n return False\\n\\n # 至少需要检查过一个有效的代码文件且全部满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\n def get_preceding_function_start_line(clean_content, index):\\n j = index - 1\\n while j >= 0 and clean_content[j].isspace(): j -= 1\\n if j < 0: return False, None\\n \\n # 1. 检测箭头函数\\n if j >= 1 and clean_content[j-1:j+1] == '=>':\\n k = j - 2\\n while k >= 0 and clean_content[k].isspace(): k -= 1\\n if k >= 0 and clean_content[k] == ')':\\n depth = 1\\n k -= 1\\n while k >= 0 and depth > 0:\\n if clean_content[k] == ')': depth += 1\\n elif clean_content[k] == '(': depth -= 1\\n k -= 1\\n if k >= 0:\\n temp = k\\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\\n if temp >= 4 and clean_content[temp-4:temp+1] == 'async' and (temp-5 < 0 or not clean_content[temp-5].isalnum()):\\n return True, clean_content.count('\\\\n', 0, temp-4) + 1\\n return True, clean_content.count('\\\\n', 0, k + 1) + 1\\n \\n temp = j - 2\\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\\n if temp >= 0 and (clean_content[temp].isalnum() or clean_content[temp] in '_$'):\\n var_end = temp\\n while temp >= 0 and (clean_content[temp].isalnum() or clean_content[temp] in '_$'): temp -= 1\\n temp2 = temp\\n while temp2 >= 0 and clean_content[temp2].isspace(): temp2 -= 1\\n if temp2 >= 4 and clean_content[temp2-4:temp2+1] == 'async' and (temp2-5 < 0 or not clean_content[temp2-5].isalnum()):\\n return True, clean_content.count('\\\\n', 0, temp2-4) + 1\\n return True, clean_content.count('\\\\n', 0, var_end) + 1\\n return True, clean_content.count('\\\\n', 0, j) + 1\\n\\n c = clean_content[j]\\n k = j\\n # 2. 处理 TypeScript 中的返回类型 (跳过 ': type' 直到遇到 ')')\\n if c != ')':\\n found_colon = False\\n while k >= 0:\\n if clean_content[k] == ':': found_colon = True\\n elif clean_content[k] == ')': break\\n elif clean_content[k] in ';{}=(': return False, None\\n k -= 1\\n if k < 0 or clean_content[k] != ')' or not found_colon:\\n return False, None\\n \\n # 3. 匹配对应的 '(' 并判断前方关键字\\n depth = 1\\n k -= 1\\n while k >= 0 and depth > 0:\\n if clean_content[k] == ')': depth += 1\\n elif clean_content[k] == '(': depth -= 1\\n k -= 1\\n \\n if k < 0: return True, clean_content.count('\\\\n', 0, 0) + 1\\n start_paren = k + 1\\n \\n while k >= 0 and clean_content[k].isspace(): k -= 1\\n if k < 0: return True, clean_content.count('\\\\n', 0, start_paren) + 1\\n \\n if clean_content[k].isalnum() or clean_content[k] in '_$':\\n end_word = k\\n while k >= 0 and (clean_content[k].isalnum() or clean_content[k] in '_$'): k -= 1\\n keyword = clean_content[k+1:end_word+1]\\n \\n # 排除常见的控制流关键字\\n if keyword in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n return False, None\\n \\n if keyword != 'function':\\n temp = k\\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\\n if temp >= 4 and clean_content[temp-4:temp+1] == 'async' and (temp-5 < 0 or not clean_content[temp-5].isalnum()):\\n return True, clean_content.count('\\\\n', 0, temp-4) + 1\\n\\n temp = k\\n while temp >= 0 and clean_content[temp].isspace(): temp -= 1\\n if temp >= 7 and clean_content[temp-7:temp+1] == 'function' and (temp-8 < 0 or not clean_content[temp-8].isalnum()):\\n temp2 = temp - 8\\n while temp2 >= 0 and clean_content[temp2].isspace(): temp2 -= 1\\n if temp2 >= 4 and clean_content[temp2-4:temp2+1] == 'async' and (temp2-5 < 0 or not clean_content[temp2-5].isalnum()):\\n return True, clean_content.count('\\\\n', 0, temp2-4) + 1\\n return True, clean_content.count('\\\\n', 0, temp-7) + 1\\n \\n return True, clean_content.count('\\\\n', 0, k+1) + 1\\n else:\\n if k >= 7 and clean_content[k-7:k+1] == 'function' and (k-8 < 0 or not clean_content[k-8].isalnum()):\\n temp2 = k - 8\\n while temp2 >= 0 and clean_content[temp2].isspace(): temp2 -= 1\\n if temp2 >= 4 and clean_content[temp2-4:temp2+1] == 'async' and (temp2-5 < 0 or not clean_content[temp2-5].isalnum()):\\n return True, clean_content.count('\\\\n', 0, temp2-4) + 1\\n return True, clean_content.count('\\\\n', 0, k-7) + 1\\n return True, clean_content.count('\\\\n', 0, start_paren) + 1\\n\\n def check_file(file_path):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except:\\n return True\\n \\n cleaned_chars = list(content)\\n in_string = False\\n string_char = ''\\n in_block_comment = False\\n in_line_comment = False\\n \\n i = 0\\n n = len(content)\\n # 预处理：将所有注释、字符串以及正则表达式替换为空格以进行安全匹配\\n while i < n:\\n c = content[i]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n else:\\n cleaned_chars[i] = ' '\\n i += 1\\n continue\\n \\n if in_block_comment:\\n if c == '*' and i + 1 < n and content[i+1] == '/':\\n cleaned_chars[i] = ' '\\n cleaned_chars[i+1] = ' '\\n in_block_comment = False\\n i += 2\\n else:\\n if c != '\\\\n':\\n cleaned_chars[i] = ' '\\n i += 1\\n continue\\n \\n if in_string:\\n if c == '\\\\\\\\':\\n cleaned_chars[i] = ' '\\n if i + 1 < n: cleaned_chars[i+1] = ' '\\n i += 2\\n continue\\n if c == string_char:\\n cleaned_chars[i] = ' '\\n in_string = False\\n else:\\n if c != '\\\\n': cleaned_chars[i] = ' '\\n i += 1\\n continue\\n \\n if c in '\\\"\\\\'`':\\n in_string = True\\n string_char = c\\n cleaned_chars[i] = ' '\\n i += 1\\n continue\\n \\n if c == '/' and i + 1 < n:\\n if content[i+1] == '/':\\n in_line_comment = True\\n cleaned_chars[i] = ' '\\n cleaned_chars[i+1] = ' '\\n i += 2\\n continue\\n elif content[i+1] == '*':\\n in_block_comment = True\\n cleaned_chars[i] = ' '\\n cleaned_chars[i+1] = ' '\\n i += 2\\n continue\\n else:\\n prev = i - 1\\n while prev >= 0 and content[prev].isspace(): prev -= 1\\n is_regex_context = False\\n if prev < 0 or content[prev] in '=(:,?!&|;{[':\\n is_regex_context = True\\n elif prev >= 5 and content[prev-5:prev+1] == 'return' and (prev-6 < 0 or not content[prev-6].isalnum()):\\n is_regex_context = True\\n \\n if is_regex_context:\\n temp = i + 1\\n in_regex = False\\n while temp < n:\\n if content[temp] == '\\\\\\\\':\\n temp += 2\\n continue\\n if content[temp] == '\\\\n':\\n break\\n if content[temp] == '/':\\n in_regex = True\\n break\\n temp += 1\\n if in_regex:\\n for r_idx in range(i, temp + 1):\\n if cleaned_chars[r_idx] != '\\\\n':\\n cleaned_chars[r_idx] = ' '\\n i = temp + 1\\n continue\\n i += 1\\n \\n clean_content = \\\"\\\".join(cleaned_chars)\\n brace_stack = []\\n \\n # 扫描块，确认是否是函数并且统计行数\\n for idx, char in enumerate(clean_content):\\n if char == '{':\\n is_func, start_line = get_preceding_function_start_line(clean_content, idx)\\n brace_stack.append((start_line, is_func))\\n elif char == '}':\\n if brace_stack:\\n start_line, is_func = brace_stack.pop()\\n if is_func:\\n current_line = clean_content.count('\\\\n', 0, idx) + 1\\n if current_line - start_line + 1 > 40:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\\n file_path = os.path.join(root, file)\\n if not check_file(file_path):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 检查回复内容中是否包含“我”\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复中是否禁止使用了“你”字。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 确保工作区路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_ts = False\\n \\n # 限制的扩展名，禁止包含纯 JavaScript 文件\\n forbidden_exts = {'.js', '.jsx', '.cjs', '.mjs'}\\n \\n # 支持的 TypeScript 文件扩展名\\n ts_exts = {'.ts', '.tsx'}\\n \\n # 需要被忽略的编译产物或依赖目录\\n ignore_dirs = {'node_modules', '.git', 'dist', 'build', 'out', 'coverage', '.next'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉忽略的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果存在任何 JS 文件，则不满足“仅使用 TypeScript”\\n if ext in forbidden_exts:\\n return False\\n \\n # 标记项目确实包含了 TypeScript 编写的代码文件\\n if ext in ts_exts:\\n has_ts = True\\n \\n # 判断标准：必须完全没有 JS 相关文件，且至少包含一个 TS/TSX 源码文件\\n return has_ts\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"限制条件：生成的代码文件中不得使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对已有文件进行改动前，不要在同级目录内备份该文件修改前的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有的数字在你生成的代码文件中必须通过定义常量来体现，坚决禁止直接使用具体的数字\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"现在的项目列表全是静态的，来加点高级交互。首先，在项目展示区最上方加几个分类标签（例如‘全部’、‘前端’、‘后端’），点击标签可以动态筛选并平滑过滤出对应的项目卡片。其次，点击任意一张项目卡片时，弹出一个带有深色遮罩层的模态框（Modal），里面展示该项目的放大版图片、详细介绍长文本以及一个‘访问网站’的模拟按钮。模态框右上角记得加个关闭按钮。对了，后续请始终遵循以下两项新要求：第一，替换此前要求所有数字必须通过定义常量体现的规则，在编写生成的代码文件时，务必直接使用数字，禁止为任何数字创建常量定义；第二，在你生成的所有代码文件中，注释内容必须统一使用中文编写。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令背景，界定可能生成的网站/TypeScript项目的常见代码文件扩展名\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css'}\\n found_code_file = False\\n \\n # 如果工作目录不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的顶行内容\\n first_line = f.readline()\\n \\n # 检查第一行是否包含强制要求中的特定编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果遇到文件无法读取、编码错误等异常情况，直接判定为未满足要求\\n return False\\n \\n # 只有当项目中存在至少一个代码文件并且全满足时，才返回True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 目标注释字符串\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 用户明确要求使用 TypeScript 进行编写，也可能包含配套的样式或模板文件\\n code_extensions = ['.ts', '.tsx', '.css', '.scss', '.html', '.js', '.jsx']\\n found_any_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖目录和版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n # 判断是否为常见的代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n found_any_file = True\\n file_path = os.path.join(root, file)\\n try:\\n has_comment = False\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 只需检查文件开头的几行，此处取前 10 行\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n if target_comment in line:\\n has_comment = True\\n break\\n \\n # 如果任一代码文件没有发现目标注释，则要求未满足\\n if not has_comment:\\n return False\\n except Exception:\\n # 如果遇到无法读取的文件，为了严谨起见，直接忽略或视为不影响总体验证\\n continue\\n \\n # 如果项目中没有任何代码文件，或者所有的代码文件都正确包含了注释，返回对应的布尔结果\\n return found_any_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 遍历项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的构建和依赖目录\\n parts = root.replace('\\\\\\\\', '/').split('/')\\n if any(ignored in parts for ignored in ['node_modules', '.git', 'dist', 'build', 'out']):\\n continue\\n \\n for file in files:\\n # 根据用户指令第1轮可推断项目使用TypeScript，这里包含常见的JS/TS扩展名\\n if file.endswith(('.ts', '.tsx', '.js', '.jsx')):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 第一步：词法简化。剔除注释和字符串内容，以免其中的大括号干扰匹配。\\n # 严格保留 \\\\n 确保行号不变。\\n clean_code_chars = []\\n i = 0\\n n = len(code)\\n state = 'NORMAL'\\n while i < n:\\n if state == 'NORMAL':\\n if code[i:i+2] == '//':\\n state = 'LINE_COMMENT'\\n clean_code_chars.append(' ')\\n i += 2\\n elif code[i:i+2] == '/*':\\n state = 'BLOCK_COMMENT'\\n clean_code_chars.append(' ')\\n i += 2\\n elif code[i] == \\\"'\\\":\\n state = 'STRING1'\\n clean_code_chars.append(' ')\\n i += 1\\n elif code[i] == '\\\"':\\n state = 'STRING2'\\n clean_code_chars.append(' ')\\n i += 1\\n elif code[i] == '`':\\n state = 'STRING3'\\n clean_code_chars.append(' ')\\n i += 1\\n else:\\n clean_code_chars.append(code[i])\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state = 'NORMAL'\\n clean_code_chars.append('\\\\n')\\n else:\\n clean_code_chars.append(' ')\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if code[i:i+2] == '*/':\\n state = 'NORMAL'\\n clean_code_chars.append(' ')\\n i += 2\\n else:\\n if code[i] == '\\\\n':\\n clean_code_chars.append('\\\\n')\\n else:\\n clean_code_chars.append(' ')\\n i += 1\\n elif state in ['STRING1', 'STRING2', 'STRING3']:\\n if state == 'STRING1' and code[i] == \\\"'\\\":\\n state = 'NORMAL'\\n clean_code_chars.append(' ')\\n i += 1\\n elif state == 'STRING2' and code[i] == '\\\"':\\n state = 'NORMAL'\\n clean_code_chars.append(' ')\\n i += 1\\n elif state == 'STRING3' and code[i] == '`':\\n state = 'NORMAL'\\n clean_code_chars.append(' ')\\n i += 1\\n elif code[i] == '\\\\\\\\':\\n if i + 1 < n:\\n if code[i+1] == '\\\\n':\\n clean_code_chars.append(' \\\\n')\\n else:\\n clean_code_chars.append(' ')\\n else:\\n clean_code_chars.append(' ')\\n i += 2\\n elif code[i] == '\\\\n':\\n clean_code_chars.append('\\\\n')\\n i += 1\\n else:\\n clean_code_chars.append(' ')\\n i += 1\\n \\n clean_text = \\\"\\\".join(clean_code_chars)\\n \\n # 第二步：找到所有的括号位置并区分普通块与函数块\\n brace_positions = []\\n for idx, char in enumerate(clean_text):\\n if char == '{':\\n p = idx - 1\\n while p >= 0 and clean_text[p].isspace():\\n p -= 1\\n if p < 0:\\n brace_positions.append(('OTHER_START', idx))\\n continue\\n \\n last_char = clean_text[p]\\n is_func = False\\n \\n # 匹配箭头函数\\n if p >= 1 and clean_text[p-1:p+1] == '=>':\\n is_func = True\\n else:\\n # 判断大括号前是否为合法的函数结束符号(如 ), 字母, >, 等)\\n valid_end_chars = set(')>]|&}')\\n if last_char in valid_end_chars or last_char.isalnum() or last_char in '_$':\\n temp_p = p\\n found_paren = False\\n brace_count = 0\\n bracket_count = 0\\n \\n # 回溯寻找匹配的参数闭合列表 ')'\\n while temp_p >= 0:\\n c = clean_text[temp_p]\\n if c == '}':\\n brace_count += 1\\n elif c == '{':\\n brace_count -= 1\\n elif c == ']':\\n bracket_count += 1\\n elif c == '[':\\n bracket_count -= 1\\n elif c == ')':\\n if brace_count == 0 and bracket_count == 0:\\n found_paren = True\\n break\\n elif c in ';=':\\n if brace_count == 0 and bracket_count == 0:\\n break\\n temp_p -= 1\\n \\n if found_paren:\\n paren_count = 0\\n k = temp_p\\n # 配对找到对应的 '('\\n while k >= 0:\\n if clean_text[k] == ')':\\n paren_count += 1\\n elif clean_text[k] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n k -= 1\\n \\n if k >= 0:\\n q = k - 1\\n while q >= 0 and clean_text[q].isspace():\\n q -= 1\\n if q >= 0:\\n word_end = q + 1\\n while q >= 0 and (clean_text[q].isalnum() or clean_text[q] in '_$'):\\n q -= 1\\n word = clean_text[q+1:word_end]\\n \\n # 如果 '(' 前的关键字非控制流，即认为是函数\\n if word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n is_func = True\\n\\n if is_func:\\n brace_positions.append(('FUNC_START', idx))\\n else:\\n brace_positions.append(('OTHER_START', idx))\\n \\n elif char == '}':\\n brace_positions.append(('END', idx))\\n \\n # 第三步：使用栈来进行结构匹配，并验证函数长度\\n stack = []\\n for item in brace_positions:\\n if item[0] in ['FUNC_START', 'OTHER_START']:\\n stack.append(item)\\n elif item[0] == 'END':\\n if stack:\\n start_item = stack.pop()\\n # 仅统计函数的长度\\n if start_item[0] == 'FUNC_START':\\n start_idx = start_item[1]\\n end_idx = item[1]\\n start_line = clean_text.count('\\\\n', 0, start_idx) + 1\\n end_line = clean_text.count('\\\\n', 0, end_idx) + 1\\n # 检查函数行数是否超过 40 行\\n if (end_line - start_line + 1) > 40:\\n return False\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 判断回复中是否包含被禁止使用的字“我”\\n if '我' in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断 AI 的最后一轮回复中是否遵守了禁止使用“你”的规则。\\n \\\"\\\"\\\"\\n # 如果回复中存在“你”字，则说明违反了禁止称呼和替换的规则\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_ts = False\\n has_js = False\\n \\n # 需要忽略的常见编译产物与依赖目录\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'coverage', 'out', '.next', '.cache'}\\n \\n # 允许部分特定前端构建工具和 linter 使用 JavaScript 配置文件的扩展名（豁免清单）\\n allowed_js_configs = {\\n '.eslintrc.js', '.eslintrc.cjs', 'postcss.config.js', \\n 'tailwind.config.js', 'next.config.js', 'vite.config.js',\\n 'jest.config.js', 'babel.config.js', 'rollup.config.js',\\n 'webpack.config.js', 'vue.config.js', 'remix.config.js'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以排除不需要检查的目录和隐藏目录（如 .vscode, .idea 等）\\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\\n \\n for file in files:\\n if file.endswith(('.ts', '.tsx')):\\n has_ts = True\\n elif file.endswith(('.js', '.jsx', '.cjs', '.mjs')):\\n # 豁免以 config.js/cjs/mjs 结尾的通用配置文件或在白名单中的工程化配置文件\\n if file.endswith(('.config.js', '.config.cjs', '.config.mjs')) or file in allowed_js_configs:\\n continue\\n \\n # 如果出现豁免清单以外的 JS 文件，则视为使用了 JavaScript 编写核心代码\\n has_js = True\\n break\\n \\n if has_js:\\n break\\n \\n # 必须至少包含一个 ts/tsx 文件，并且没有任何非配置类的 js/jsx 文件\\n return has_ts and not has_js\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"限制条件：生成的代码文件中不得使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对已有文件进行改动前，不要在同级目录内备份该文件修改前的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在编写生成的代码文件时，务必直接使用数字，禁止为任何数字创建常量定义\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"项目模块交互很棒。现在完善页面最底部的‘联系我’模块。我需要一个好看的联系表单，包含‘姓名’、‘邮箱’和‘留言内容’（多行文本框）三个输入框，以及一个‘发送留言’按钮。重点是加上前端的表单验证逻辑：如果漏填任意必填项，或者邮箱格式不正确，要在对应的输入框下方立即显示红色的错误提示语（提示语也要支持双语）。全部验证通过后点击发送，清空表单并弹出一个‘发送成功’的绿色提示框。对了，后续请始终遵循以下两项要求：第一，替换此前要求代码注释统一使用中文的规则，无论生成哪个代码文件，里面的注释都只能使用英文；第二，替换此前修改已有文件前不需要备份的规则，只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n # 常见的前端源码相关文件扩展名\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.scss', '.less', '.html', '.vue'}\\n # 过滤掉非手写的依赖文件夹或构建产物文件夹\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'out', '.next', 'coverage'}\\n \\n found_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 判断顶行（第一行）是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取或是其他读取异常，视为未能满足生成有效代码的要求\\n return False\\n \\n # 如果项目中没有任何代码文件，说明未生成有效代码结构，返回 False\\n return found_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 前端开发中常见的代码文件扩展名\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.html', '.css', '.scss', '.less', '.vue'}\\n \\n file_checked = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的非业务代码依赖与配置目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n if not lines:\\n continue\\n\\n file_checked = True\\n \\n # 检查文件开头（前10行内）是否存在目标注释\\n has_comment = False\\n for line in lines[:10]:\\n if target_comment in line:\\n has_comment = True\\n break\\n \\n # 如果有任意一个代码文件的开头不包含该注释，则不满足要求\\n if not has_comment:\\n return False\\n\\n # 必须至少检查过一个文件，否则说明没有生成相关代码文件\\n if not file_checked:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 替换函数：将注释和字符串字面量中的非换行字符均转化为空格\\n # 目的是去掉花括号干扰并保证清洗后的代码行号结构和原代码完全对齐\\n def replacer(m):\\n match_str = m.group(0)\\n return ''.join(['\\\\n' if ch == '\\\\n' else ' ' for ch in match_str])\\n\\n # 匹配规则：双引号字符串、单引号字符串、反引号字符串、单行注释、多行注释\\n pattern = re.compile(\\n r'\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|`(?:\\\\\\\\.|[^\\\\\\\\`])*`|//.*?$|/\\\\*.*?\\\\*/',\\n re.DOTALL | re.MULTILINE\\n )\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(('.ts', '.tsx', '.js', '.jsx')):\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n\\n clean_text = pattern.sub(replacer, code)\\n stack = []\\n \\n for i, char in enumerate(clean_text):\\n if char == '{':\\n start = i - 1\\n depth = 0\\n # 往前寻找这段声明部分的起始边界\\n while start >= 0:\\n ch = clean_text[start]\\n if ch == ')':\\n depth += 1\\n elif ch == '(':\\n depth -= 1\\n \\n # 在非圆括号嵌套中，遇到语句终结符或作用域界限代表当前声明的起点界限\\n if depth == 0 and ch in '{};':\\n break\\n start -= 1\\n \\n decl_start = start + 1\\n declaration = clean_text[decl_start:i]\\n decl_stripped = declaration.strip()\\n \\n is_function = False\\n \\n # 1. 包含 => (箭头函数) 或 function 关键字\\n if '=>' in declaration or 'function' in declaration:\\n is_function = True\\n # 2. 如果不含，但以 ')' 结尾，这极有可能是方法如 foo(a, b) {\\n elif decl_stripped.endswith(')'):\\n d = 0\\n idx = len(decl_stripped) - 1\\n while idx >= 0:\\n if decl_stripped[idx] == ')':\\n d += 1\\n elif decl_stripped[idx] == '(':\\n d -= 1\\n if d == 0:\\n break\\n idx -= 1\\n \\n if idx >= 0:\\n prefix = decl_stripped[:idx].strip()\\n # 提取括号前的最后一个标识词\\n words = re.findall(r'[a-zA-Z_]\\\\w*', prefix)\\n if words:\\n last_word = words[-1]\\n # 排除常见的语法控制流关键字\\n if last_word not in ('if', 'for', 'while', 'catch', 'switch', 'with'):\\n is_function = True\\n \\n # 压栈：记录当前花括号的位置、它是否属于一个函数、以及函数声明语句的起始索引\\n stack.append((i, is_function, decl_start))\\n \\n elif char == '}':\\n if stack:\\n brace_idx, is_function, decl_start = stack.pop()\\n if is_function:\\n # 根据提取出来的完整函数范围计算其实际跨越的行数\\n func_text = clean_text[decl_start:i+1].strip()\\n lines_count = func_text.count('\\\\n') + 1\\n if lines_count > 40:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断AI助手的回复中是否杜绝使用了\\\"我\\\"。\\n 如果回复中包含\\\"我\\\"字，则判定为不满足要求返回False；否则返回True。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI的回复中是否遵守了禁止使用“你”称呼用户的要求。\\n \\\"\\\"\\\"\\n # 剔除中文中极少数包含“你”字但并非指代用户的固定词汇，避免误判\\n clean_response = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 检查回复内容中是否包含“你”或“您”\\n if \\\"你\\\" in clean_response or \\\"您\\\" in clean_response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def check_typescript_only_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n has_ts_or_tsx = False\\n has_unauthorized_js_or_jsx = False\\n \\n # 需要跳过的无关文件夹\\n exclude_dirs = {'node_modules', 'dist', 'build', 'out', '.git', '.vscode', '.idea', 'coverage'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以跳过特定目录，不再向下遍历\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 检测到 TS/TSX 代码文件\\n if ext in ['.ts', '.tsx']:\\n has_ts_or_tsx = True\\n \\n # 检测到 JS/JSX 代码文件\\n elif ext in ['.js', '.jsx']:\\n file_lower = file.lower()\\n # 豁免常见的 JS 配置文件\\n is_config_file = (\\n 'config' in file_lower or \\n 'rc.js' in file_lower or \\n file_lower.startswith('.eslint') or \\n file_lower.startswith('.prettier')\\n )\\n \\n # 如果不是被豁免的配置文件，则视为违反了 TypeScript 限定的要求\\n if not is_config_file:\\n has_unauthorized_js_or_jsx = True\\n\\n # 满足条件：项目中有 TypeScript 源码，且没有未授权的 JavaScript 源码\\n return has_ts_or_tsx and not has_unauthorized_js_or_jsx\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"限制条件：生成的代码文件中不得使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在编写生成的代码文件时，务必直接使用数字，禁止为任何数字创建常量定义\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，里面的注释都只能使用英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"最后一步，我们来优化响应式体验和整体导航体验。首先做移动端适配：当屏幕宽度小于768px时，顶部导航栏的文字菜单请折叠隐藏，显示为一个‘汉堡图标’，点击后能弹出下拉菜单；同时所有的分列布局（如首屏的左右布局、项目的网格布局）都要自动变为单列展示。其次，点击顶部导航栏的任意菜单项时，页面不要生硬跳转，而是带有平滑滚动动画（Smooth Scroll）滑动到下方对应的模块区域。对了，后续请始终遵循以下两项要求：第一，替换此前要求代码中务必直接使用数字、禁止定义数字常量的规则，你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用；第二，保持你生成的代码文件里每个函数的圈复杂度均在3以内。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在进行任何文件操作的第一步，是运行`pwd`命令查看当前所在目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n file_checked = False\\n # 根据用户指令中“仅使用 TypeScript 进行编写”的要求，限定检查的文件扩展名\\n valid_extensions = {'.ts', '.tsx'}\\n # 排除常见的无关或依赖目录\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1]\\n if ext in valid_extensions:\\n file_checked = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的第一行（顶行）\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取发生异常，视为不满足\\n return False\\n\\n # 只有当至少检查了一个文件且全部满足要求时，才返回 True\\n return file_checked\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在开头另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n code_extensions = {'.ts', '.tsx', '.js', '.jsx', '.css', '.html'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的依赖与构建目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.next')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 仅需读取前几行进行校验即可\\n lines = [f.readline() for _ in range(5)]\\n \\n # 检查文件头部（取前5行以兼容第一行的特定编码声明）\\n is_comment_found = False\\n for line in lines:\\n if not line: # 遇到文件末尾\\n break\\n if target_comment in line:\\n is_comment_found = True\\n break\\n \\n # 如果某个代码文件开头没有找到该特定注释，则不满足要求\\n if not is_comment_found:\\n return False\\n \\n except Exception:\\n # 如果遇到文件读取异常（如非法编码等），保守判定为不满足\\n return False\\n \\n # 如果项目中没有任何对应的代码文件，说明没有生成，无法满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中的所有函数，各自的行数都不超过 40 行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 替换注释和字符串为空白内容（保持原有换行符，以便精确计算行数）\\n def replacer(match):\\n s = match.group(0)\\n return '\\\\n' * s.count('\\\\n') + ' ' * (len(s) - s.count('\\\\n'))\\n\\n pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.DOTALL | re.MULTILINE\\n )\\n\\n # 移除 prefix 中的局部平衡大括号，防止解构赋值对象等带来的正则干扰\\n def remove_balanced_braces(s: str) -> str:\\n while True:\\n new_s = re.sub(r'\\\\{[^{}]*\\\\}', 'X', s)\\n if new_s == s:\\n break\\n s = new_s\\n return s\\n\\n blacklist = {'if', 'for', 'while', 'switch', 'catch', 'with'}\\n\\n # 正则表达式用于匹配在 '{' 之前的不同类型的函数签名\\n arrow_pattern = re.compile(r'=>\\\\s*$')\\n anon_func_pattern = re.compile(r'\\\\bfunction\\\\b\\\\s*(?:\\\\w+\\\\s*)?(?:<[^;{=]*>\\\\s*)?\\\\([^;{]*\\\\)\\\\s*(?::[^;{=]*)?$')\\n func_pattern = re.compile(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*(?:<[^;{=]*>\\\\s*)?\\\\([^;{]*\\\\)\\\\s*(?::[^;{=]*)?$')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(('.ts', '.tsx', '.js', '.jsx')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n clean_content = pattern.sub(replacer, content)\\n \\n # 计算通过 idx 到行号的映射\\n lines = clean_content.split('\\\\n')\\n idx_to_line = []\\n for r, line in enumerate(lines):\\n idx_to_line.extend([r + 1] * (len(line) + 1))\\n \\n stack = []\\n blocks = []\\n \\n # 找到所有的成对的大括号，计算块行数\\n for i, char in enumerate(clean_content):\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start_i = stack.pop()\\n start_line = idx_to_line[start_i]\\n end_line = idx_to_line[i]\\n \\n # 如果块的行数超过了 40 行，认为是一个过长的块\\n if end_line - start_line + 1 > 40:\\n blocks.append(start_i)\\n\\n # 对所有的大块进行函数特征判断\\n for start_i in blocks:\\n prefix_start = max(0, start_i - 500)\\n prefix = clean_content[prefix_start:start_i]\\n \\n prefix_clean = remove_balanced_braces(prefix)\\n \\n # 若是箭头函数\\n if arrow_pattern.search(prefix_clean):\\n return False\\n \\n # 若是 function 关键字声明的函数\\n if anon_func_pattern.search(prefix_clean):\\n return False\\n \\n # 若是类方法或是带有名称的普通函数形式\\n match = func_pattern.search(prefix_clean)\\n if match:\\n word = match.group(1)\\n if word not in blacklist:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"回答时，必须绝对杜绝使用\\\"我\\\"来自称，而是永远使用\\\"小岚\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 检查助手的回复文本中是否包含“我”字\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"小主人\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含被禁止的“你”字\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"强制要求：仅使用 TypeScript 进行编写\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 1. 检查回复中的代码块，如果存在 JS/JSX 代码块，则直接返回 False\\n js_pattern = r'```(javascript|js|jsx)\\\\b'\\n if re.search(js_pattern, response, re.IGNORECASE):\\n return False\\n \\n has_ts = False\\n \\n # 检查回复中是否包含 TS/TSX 代码块作为备用依据\\n ts_pattern = r'```(typescript|ts|tsx)\\\\b'\\n if re.search(ts_pattern, response, re.IGNORECASE):\\n has_ts = True\\n\\n # 2. 检查工作目录中的实际文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除非源代码相关的常见文件夹以防止误判\\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', 'out', 'coverage'}\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果发现 js/jsx 文件，需要判断是否属于常见配置文件\\n if ext in ['.js', '.jsx']:\\n # 容错处理：配置文件（例如 tailwind.config.js, .eslintrc.js, .prettierrc.js）可以宽容放行\\n is_config = (\\n file.endswith('.config.js') or \\n file.startswith('.eslintrc') or \\n file.endswith('rc.js')\\n )\\n # 如果不是这些常见的配置文件，则说明违背了“仅使用 TypeScript 进行编写”的指令\\n if not is_config:\\n return False\\n \\n # 如果发现 ts/tsx 文件，则将 has_ts 标记为 True\\n elif ext in ['.ts', '.tsx']:\\n has_ts = True\\n\\n return has_ts\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"限制条件：生成的代码文件中不得使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码经过 ESLint 校验，并修复其中的所有报错，直至 ESLint 检查结果不包含任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，里面的注释都只能使用英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["向下滚动网页，观察顶部导航栏是否始终固定在页面顶部不会随页面滚动而消失，同时检查导航栏中是否包含‘首页’、‘项目’、‘联系我’这几个文字占位；并检查页面主体是否包含垂直排列的三个不同背景色的区域，并在最下方包含一个页脚。","点击顶部导航栏最右侧的'中/En'切换按钮，检查整个网页的静态文字（包含导航栏菜单项、页面标题、介绍文本、页脚等）是否无缝切换为另一门语言。","观察大屏幕下的首屏内容，检查整个区域是否在一个视口高度内垂直居中，并呈现左右两列布局：左侧包含加粗放大的姓名、介绍文本及'查看项目'和'下载简历'两个按钮；右侧为一个圆形的占位头像图。","观察首屏左侧的介绍区域，检查在欢迎语和自我介绍文本之间是否有一个自动逐字显示的'打字机'特效文字，每隔几秒输出并替换成下一个词组，且能无限循环播放。","在展示打字机特效时，点击顶部的中英文切换按钮，观察打字机当前或后续输出的词汇是否能跟随页面成功切换为对应的目标语言词汇。","滚动到'项目'模块，检查是否以网格布局（Grid）展示了4到6张项目卡片，且每张卡片上方显示图片占位，下方包含项目标题、两行文本介绍及底部带圆角的技术栈标签。此时点击顶部中/En切换按钮，检查这些项目卡片的文本是否也能成功切换为目标语言。","在项目模块顶部找到分类标签（如'全部'、'前端'等），点击不同的分类标签，观察下方的项目卡片列表是否动态进行了过滤筛选。","点击项目列表中的任意一张卡片，检查页面是否弹出一个带有深色半透明遮罩层的模态框（Modal），并且模态框内部包含大尺寸的图片、长介绍文本、'访问网站'按钮以及右上角的关闭按钮。","在模态框处于打开的状态下，点击模态框右上角的关闭按钮，检查模态框和对应的遮罩层是否被成功关闭隐藏。","滚动到页面底部的'联系我'表单处，保持表单全部空白或输入错误格式的邮箱字符，直接点击'发送留言'按钮，检查未填写的输入框或错误邮箱下方是否会立即显示出红色的错误提示语。","触发表单验证并保持红色的错误提示语处于显示状态，点击顶部的中英文语言切换按钮，检查这些红色错误提示文字是否跟着成功切换了语言。","在表单中填写完整的姓名、正确格式的邮箱以及留言文本后，点击'发送留言'按钮，检查表单的所有输入框是否被清空，同时页面是否弹出一个绿色的'发送成功'提示框。","将浏览器窗口宽度缩小至768px以下（或开启移动设备模拟），观察顶部导航栏的菜单项文字是否自动隐藏折叠，变为一个'汉堡图标'；点击该汉堡图标，检查是否能向下弹出展开完整的下拉菜单。","在窗口宽度小于768px的移动端模式下向下滚动页面，观察首屏原有的左右两列布局以及项目模块的多列网格卡片布局，检查它们是否全部自动响应转换为上下排列的单列布局。","在顶部导航栏点击'项目'或'联系我'等其他菜单项，检查页面是否伴随平滑过渡的滚动动画（Smooth Scroll）自动往下滑动至对应的模块区域，而不是瞬间跳跃式切换。"],"string":"[\n \"向下滚动网页，观察顶部导航栏是否始终固定在页面顶部不会随页面滚动而消失，同时检查导航栏中是否包含‘首页’、‘项目’、‘联系我’这几个文字占位；并检查页面主体是否包含垂直排列的三个不同背景色的区域，并在最下方包含一个页脚。\",\n \"点击顶部导航栏最右侧的'中/En'切换按钮，检查整个网页的静态文字（包含导航栏菜单项、页面标题、介绍文本、页脚等）是否无缝切换为另一门语言。\",\n \"观察大屏幕下的首屏内容，检查整个区域是否在一个视口高度内垂直居中，并呈现左右两列布局：左侧包含加粗放大的姓名、介绍文本及'查看项目'和'下载简历'两个按钮；右侧为一个圆形的占位头像图。\",\n \"观察首屏左侧的介绍区域，检查在欢迎语和自我介绍文本之间是否有一个自动逐字显示的'打字机'特效文字，每隔几秒输出并替换成下一个词组，且能无限循环播放。\",\n \"在展示打字机特效时，点击顶部的中英文切换按钮，观察打字机当前或后续输出的词汇是否能跟随页面成功切换为对应的目标语言词汇。\",\n \"滚动到'项目'模块，检查是否以网格布局（Grid）展示了4到6张项目卡片，且每张卡片上方显示图片占位，下方包含项目标题、两行文本介绍及底部带圆角的技术栈标签。此时点击顶部中/En切换按钮，检查这些项目卡片的文本是否也能成功切换为目标语言。\",\n \"在项目模块顶部找到分类标签（如'全部'、'前端'等），点击不同的分类标签，观察下方的项目卡片列表是否动态进行了过滤筛选。\",\n \"点击项目列表中的任意一张卡片，检查页面是否弹出一个带有深色半透明遮罩层的模态框（Modal），并且模态框内部包含大尺寸的图片、长介绍文本、'访问网站'按钮以及右上角的关闭按钮。\",\n \"在模态框处于打开的状态下，点击模态框右上角的关闭按钮，检查模态框和对应的遮罩层是否被成功关闭隐藏。\",\n \"滚动到页面底部的'联系我'表单处，保持表单全部空白或输入错误格式的邮箱字符，直接点击'发送留言'按钮，检查未填写的输入框或错误邮箱下方是否会立即显示出红色的错误提示语。\",\n \"触发表单验证并保持红色的错误提示语处于显示状态，点击顶部的中英文语言切换按钮，检查这些红色错误提示文字是否跟着成功切换了语言。\",\n \"在表单中填写完整的姓名、正确格式的邮箱以及留言文本后，点击'发送留言'按钮，检查表单的所有输入框是否被清空，同时页面是否弹出一个绿色的'发送成功'提示框。\",\n \"将浏览器窗口宽度缩小至768px以下（或开启移动设备模拟），观察顶部导航栏的菜单项文字是否自动隐藏折叠，变为一个'汉堡图标'；点击该汉堡图标，检查是否能向下弹出展开完整的下拉菜单。\",\n \"在窗口宽度小于768px的移动端模式下向下滚动页面，观察首屏原有的左右两列布局以及项目模块的多列网格卡片布局，检查它们是否全部自动响应转换为上下排列的单列布局。\",\n \"在顶部导航栏点击'项目'或'联系我'等其他菜单项，检查页面是否伴随平滑过渡的滚动动画（Smooth Scroll）自动往下滑动至对应的模块区域，而不是瞬间跳跃式切换。\"\n]","isConversational":false}}},{"rowIdx":47,"cells":{"task_id":{"kind":"number","value":48,"string":"48"},"repository_policy":{"kind":"string","value":"请你在后续所有交互轮次中严格遵循以下规则：\n1. 无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词。\n2. 所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"。\n3. 确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"。\n4. 请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"。\n5. 无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码。\n6. 生成的各个代码文件内，注释部分需完全使用英文撰写。\n7. 你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n8. 运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复最开头是否带有\"喵～\"语气词。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，代表是否满足要求\n \"\"\"\n # 去除首尾可能的空白字符，判断开头是否为\"喵～\"\n return response.strip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复末尾是否带有语气词\"喵～\"。\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容两端的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 检查是否以\"喵～\"结尾\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查助手的直接文字回复中是否包含被禁止的 \"我\"\n if \"我\" in response:\n return False\n \n # 检查工作区生成的代码项目文件中是否包含 \"我\"\n if workspace_path and os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if \"我\" in content:\n return False\n except Exception:\n # 忽略无法用 utf-8 读取的文件（例如二进制文件、图片等）\n continue\n \n return True","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复中是否使用了\"小可爱\"称呼用户，并且完全替代了\"你\"。\n \"\"\"\n # 检查回复中是否包含\"小可爱\"\n if \"小可爱\" not in response:\n return False\n \n # 检查回复中是否完全没有\"你\"字\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见代码文件的扩展名集合，避免检查到二进制或无关文本文件\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', \n '.go', '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.m'\n }\n \n if not os.path.exists(workspace_path):\n return True\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 检查起始第一行是否包含该编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法打开或发生其他异常情况，视为未能满足要求\n return False\n \n # 所有代码文件第一行均符合要求，或没生成任何代码文件，返回True\n return True","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 定义常见的代码文件后缀\n code_extensions = {\n '.py', '.js', '.ts', '.html', '.css', '.cpp', '.c', '.h', \n '.java', '.go', '.rs', '.php', '.rb', '.sh', '.cs'\n }\n\n # 遍历工作区目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read().strip()\n \n if not content:\n # 代码文件为空，没有包含所需的注释\n return False\n\n # 按行分割并去除空行\n lines = [line.strip() for line in content.split('\\n') if line.strip()]\n \n if not lines:\n return False\n\n # 考虑到不同语言的多行注释可能导致闭合标签(如 */ 或 -->)独占最后一行，\n # 故检查结尾最后3个非空行中是否包含目标声明。\n found_statement = False\n for line in lines[-3:]:\n if target_statement in line:\n found_statement = True\n break\n \n if not found_statement:\n return False\n\n except Exception:\n # 如果由于读取权限或文件编码等原因失败，默认判定为不符合\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]}],"string":"[\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，代表是否满足要求\\n \\\"\\\"\\\"\\n # 去除首尾可能的空白字符，判断开头是否为\\\"喵～\\\"\\n return response.strip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复末尾是否带有语气词\\\"喵～\\\"。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容两端的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 检查是否以\\\"喵～\\\"结尾\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查助手的直接文字回复中是否包含被禁止的 \\\"我\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n # 检查工作区生成的代码项目文件中是否包含 \\\"我\\\"\\n if workspace_path and os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if \\\"我\\\" in content:\\n return False\\n except Exception:\\n # 忽略无法用 utf-8 读取的文件（例如二进制文件、图片等）\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复中是否使用了\\\"小可爱\\\"称呼用户，并且完全替代了\\\"你\\\"。\\n \\\"\\\"\\\"\\n # 检查回复中是否包含\\\"小可爱\\\"\\n if \\\"小可爱\\\" not in response:\\n return False\\n \\n # 检查回复中是否完全没有\\\"你\\\"字\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见代码文件的扩展名集合，避免检查到二进制或无关文本文件\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', \\n '.go', '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.m'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 检查起始第一行是否包含该编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法打开或发生其他异常情况，视为未能满足要求\\n return False\\n \\n # 所有代码文件第一行均符合要求，或没生成任何代码文件，返回True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 定义常见的代码文件后缀\\n code_extensions = {\\n '.py', '.js', '.ts', '.html', '.css', '.cpp', '.c', '.h', \\n '.java', '.go', '.rs', '.php', '.rb', '.sh', '.cs'\\n }\\n\\n # 遍历工作区目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read().strip()\\n \\n if not content:\\n # 代码文件为空，没有包含所需的注释\\n return False\\n\\n # 按行分割并去除空行\\n lines = [line.strip() for line in content.split('\\\\n') if line.strip()]\\n \\n if not lines:\\n return False\\n\\n # 考虑到不同语言的多行注释可能导致闭合标签(如 */ 或 -->)独占最后一行，\\n # 故检查结尾最后3个非空行中是否包含目标声明。\\n found_statement = False\\n for line in lines[-3:]:\\n if target_statement in line:\\n found_statement = True\\n break\\n \\n if not found_statement:\\n return False\\n\\n except Exception:\\n # 如果由于读取权限或文件编码等原因失败，默认判定为不符合\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我开发一个基于Web的“自由落体物理模拟实验工具”。第一步，请先搭建好基础的页面骨架。需求如下：1. 页面分为左右两栏布局，左侧是占据主要屏幕空间的“实验舱”，请使用 Canvas 或带边框的大块 DOM 区域，底部画一条粗线代表“地面”，实验舱靠上的位置放置一个静止的小球作为实验对象。2. 右侧是“控制面板”，包含三个结构清晰的卡片占位：实验参数设置区、播放控制区（放置开始、暂停、重置三个按钮，暂不实现功能）、实时数据看板。3. 整体界面风格保持现代、极简。目前阶段不需要编写任何物理运动计算逻辑，只要把静态排版和基本UI元素渲染出来即可。\n另外请注意，后续所有轮次都需要遵循以下要求：\n1. 每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。\n2. 你生成给用户的回复内容，必须一直使用中文。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否在最开头带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 忽略可能的前导空白字符，并判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 考虑到助手回复末尾可能带有空白字符或标点符号，先去除末尾常见的空白符及标点\n cleaned_response = response.rstrip(\" \\n\\r\\t。！？.!?\")\n \n # 判断处理后的回复是否以指定的语气词结尾\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否没有使用\"我\"。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，如果回复中没有出现\"我\"，则返回True，否则返回False。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复中是否完全没有出现“你”字\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义需要检查的常见代码文件扩展名\n code_extensions = {\n '.html', '.htm', \n '.css', '.scss', '.less', \n '.js', '.mjs', '.cjs', \n '.ts', '.tsx', '.jsx', \n '.vue', '.svelte',\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.php', '.rb', '.rs'\n }\n \n # 忽略非生成的主流系统或依赖目录\n exclude_dirs = {'.git', 'node_modules', '.vscode', '.idea', '__pycache__'}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要遍历的子目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline().strip()\n # 检查严格匹配的子串是否出现在第一行\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果遇到文件不可读或编码异常的情况，也判定为不满足\n return False\n \n # 如果没有生成任何代码文件，代表骨架尚未生成，也判定为未满足\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 题目要求的必须包含的目标声明字符串\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常见的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \n '.rb', '.php', '.sh', '.swift', '.kt', '.rs'\n }\n \n found_any_file = False\n \n # 如果工作区目录不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_any_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于任何权限或读取问题导致读取失败，则算作未满足\n return False\n \n # 过滤出所有包含内容的非空行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n # 若代码文件内容完全为空，则无法包含要求的注释\n if not non_empty_lines:\n return False\n \n # 考虑到块状注释可能导致文件最后一行仅存在 `-->` 或 `*/` 等符号\n # 取末尾的 5 个非空行进行校验，确认是否在结尾处包含了该行注释声明\n found_target = False\n for line in non_empty_lines[-5:]:\n if target_str in line:\n found_target = True\n break\n \n if not found_target:\n return False\n \n # 如果未检测到任何代码文件，说明助手未能生成代码骨架，返回 False\n return found_any_file","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]}]},{"round_id":1,"instruction":"很好，骨架搭出来了。接下来我们实现核心的物理引擎和基础控制逻辑。请为左侧实验舱建立一个“像素与物理现实”的映射关系（例如设定 50 像素等于 1 米，并在界面角落标出比例尺）。给小球赋予固定的重力加速度（g=9.8m/s²）。实现控制面板里的“开始”、“暂停”和“重置”按钮逻辑：点击开始后，小球开始受重力影响做加速下落运动；触碰到地面时，小球绝对不能穿模，并且需要以 0.7 的弹性系数向上反弹，反复弹跳直到速度趋近于零最终静止；重置按钮能清空状态并让小球回到页面顶部的初始位置。\n另外请注意，后续所有轮次都需要遵循以下新增要求：1. 生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径。2. 你创建的每一个代码文件，算上空行后的总行数都必须是奇数值。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容前面的空白字符，以防AI生成的回复开头带有空格或换行\n cleaned_response = response.lstrip()\n \n # 检查最开头是否带有\"喵～\"\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n cleaned_response = response.strip()\n \n # 定义句末可能的标点符号集合（中英文常见结束标点及引号）\n punctuations = \"。！？.!?\\\"'”’\"\n \n # 循环去除末尾的标点符号及空白字符\n while cleaned_response and cleaned_response[-1] in punctuations:\n cleaned_response = cleaned_response[:-1].strip()\n \n # 判断去除标点后的回复末尾是否为指定的语气词\n return cleaned_response.endswith(\"喵～\") or cleaned_response.endswith(\"喵~\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 1. 检查回复内容中是否包含自称“我”\n if \"我\" in response:\n return False\n \n # 2. 检查生成的工作区代码文件中是否包含“我”\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if \"我\" in content:\n return False\n except Exception:\n # 忽略非文本或无法读取的文件\n pass\n\n return True","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if \"小可爱\" in response and \"你\" not in response:\n return True\n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 推测的可能的代码文件类型，贴合前端Web开发任务\n code_exts = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的非业务代码文件夹\n if '.git' in root or 'node_modules' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查是否为空文件以及首行是否严格包含声明\n if not first_line or \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n \n has_code_file = True\n except Exception:\n # 如果发生文件读取或解码错误，判定为验证失败\n return False\n \n # 只要成功验证了至少一个生成的代码文件且全都符合要求，即返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 结合Web页面开发需求推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n found_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n found_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 文件读取失败时判定未满足\n return False\n \n # 查找倒序的第一个非空行作为最后一行的内容\n last_line = \"\"\n for line in reversed(lines):\n if line.strip():\n last_line = line.strip()\n break\n \n # 验证结尾的新行注释中是否包含了指定的声明内容\n if target_statement not in last_line:\n return False\n\n # 若未找到任何代码文件，则无法证明其满足了要求\n return found_code_files","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 常见的Web及其他代码文件的后缀集合\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符切分统计行数，能够把所有空行（含末尾空行）都算入其中\n lines_count = len(content.split('\\n'))\n \n # 如果行数是偶数，说明不满足要求\n if lines_count % 2 == 0:\n return False\n except Exception:\n # 如果有异常（如编码非utf-8或无法读取），为避免程序崩溃此处做跳过处理\n continue\n \n # 如果遍历检查完没有返回 False 且确保找到了代码文件，则符合所有条件\n return has_code_file","tags":["Quantity","Exact Value"]}]},{"round_id":2,"instruction":"现在来完善参数控制和数据监控。在右侧的“实验参数设置区”中，添加三个可交互的输入框，允许用户自定义：初始高度（范围 1-100 米）、重力加速度（允许修改，以便模拟月球等不同星体环境）、小球质量（千克）。同时，在“实时数据看板”中，请使用清晰的字体实时更新并显示小球在运动过程中的核心物理量：已用时间（秒，保留两位小数）、当前速度（米/秒）和离地高度（米）。要求：当用户在运行前拖动滑块或修改“初始高度”参数时，左侧实验舱里小球的垂直位置需要实时同步调整。\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数。2. 生成的代码文件中，任何函数的圈复杂度都不能超过3。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除前导空白字符后，判断是否以“喵～”开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除字符串首尾的空白字符\n cleaned_response = response.strip()\n \n # 如果直接以\"喵～\"结尾，则满足条件\n if cleaned_response.endswith(\"喵～\"):\n return True\n \n # 考虑到末尾可能带有标点符号，去除常见的结束标点后再进行判断\n cleaned_response_without_punct = cleaned_response.rstrip(\"。！？.!?\")\n \n return cleaned_response_without_punct.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复中是否没有出现\"我\"字。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 如果回复中不包含\"我\"，返回True；否则返回False\n \"\"\"\n return \"我\" not in response","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含\"小可爱\"，且是否完全没有出现\"你\"字\n if \"小可爱\" in response and \"你\" not in response:\n return True\n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义可能生成的前端及相关代码文件扩展名\n code_extensions = {\".html\", \".htm\", \".css\", \".js\", \".ts\", \".jsx\", \".tsx\", \".py\"}\n has_code_file = False\n \n # 遍历代码项目路径\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 如果文件为空或者第一行没有包含规定的编码声明，则不满足要求\n if not first_line or \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若文件读取发生异常，则无法确认其满足要求，返回False\n return False\n \n # 如果项目中没有任何代码文件，说明未成功生成代码，返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json'}\n file_checked_count = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能存在的第三方依赖或隐藏文件夹\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_checked_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 寻找文件末尾的最后有效非空行\n last_valid_line = \"\"\n for line in reversed(lines):\n if line.strip():\n last_valid_line = line.strip()\n break\n \n # 检查最后非空行中是否包含了要求的声明文本\n if target_statement not in last_valid_line:\n return False\n except Exception:\n # 如果读取报错（如编码问题等），判定为未满足要求\n return False\n \n # 至少要验证了一个文件才算真正满足了该指令\n return file_checked_count > 0","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断该Web项目可能涉及的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.py'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前工作目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n has_code_file = True\n \n # 计入空行的总行数统计方式：按换行符分隔\n # 若文件为空字符，行数记为0；否则通过split('\\n')统计物理行数\n line_count = len(content.split('\\n')) if content else 0\n \n # 判断行数是否为偶数\n if line_count % 2 != 0:\n return False\n except Exception:\n # 忽略读取失败的非文本/非UTF-8文件\n pass\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过3","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":3,"instruction":"为了让物理模拟更真实且具备探究价值，下一步我们需要打破理想环境，引入“空气阻力”模型。请在参数设置区新增一个“材质与形状”的下拉菜单（选项包含：实心铁球、木球、带降落伞的沙袋）。选择不同的选项时，系统在后台自动赋予小球不同的风阻系数(C)和迎风面积(A)。在小球的下落运动演算中加入空气阻力公式（F = 0.5 * ρ * v² * C * A，空气密度ρ设为常数）。由于阻力与速度平方成正比，小球最终应能达到终端速度。另外，请在实时数据看板中新增一项，实时显示当前小球受到的空气阻力大小（牛顿）。\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 生成的代码文件中，任何函数的圈复杂度都不能超过2。2. 每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\"pass\", \"fail\", 或 \"error\"）。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复最开头是否带有\"喵～\"语气词。\n \"\"\"\n # 考虑到文本最前方可能会包含空格或换行等不可见字符，使用lstrip()去除左侧空白后判断\n if response.lstrip().startswith(\"喵～\"):\n return True\n return False","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复末尾是否附带语气词\"喵～\"\n \"\"\"\n # 移除回复内容末尾可能包含的空白字符（如换行符、空格等）\n cleaned_response = response.strip()\n \n # 检查结尾是否为指定语气词\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否遵循了不使用“我”作为自我称呼的要求。\n 出于自动化评测的准确性和鲁棒性，这里采用严格匹配：回复中不得出现“我”字。\n \"\"\"\n # 如果回复中存在“我”字，则判定为不满足要求\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复内容中是否包含禁止使用的“你”字\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义可能生成的常见代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.py', '.ts', '.jsx', '.tsx'}\n found_code_file = False\n \n # 遍历当前项目的工作区路径\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行，去除可能包含的头尾空白字符与换行符\n first_line = f.readline().strip()\n \n # 检查声明编码的标识是否存在于该文件的第一行\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于任何异常（如非UTF-8编码导致读取失败），判定为不符合要求\n return False\n \n # 如果遍历完毕未触发False，且至少找到了一个代码文件，则符合要求\n return found_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户指令推断可能涉及到的代码文件后缀，主要包含网页前端、JSON测试结果等\n code_extensions = {'.html', '.css', '.js', '.json', '.py', '.ts', '.jsx', '.tsx'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的无关文件夹\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于编码或其他原因无法读取，默认认为不满足格式要求\n return False\n \n # 寻找代码文件中的倒数第一行（忽略结尾的纯空行）\n last_non_empty_line = \"\"\n for line in reversed(lines):\n if line.strip():\n last_non_empty_line = line\n break\n \n # 如果最后非空行不存在，或者指定的声明文字未出现在最后一行，则验证失败\n if target_statement not in last_non_empty_line:\n return False\n \n # 只要生成了代码文件并且均通过了检查，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，说明未创建任何文件\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 根据项目推断可能的代码文件类型后缀\n target_extensions = {'.html', '.css', '.js', '.json', '.ts', '.jsx', '.tsx', '.vue'}\n checked_files = 0\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的包管理目录或版本控制目录\n if any(ignored in root.split(os.sep) for ignored in ['.git', 'node_modules', '__pycache__']):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用 split('\\n') 进行分割可以精确计算计入空行后的总行数\n line_count = len(content.split('\\n'))\n \n # 如果有任何一个代码文件的行数为奇数，则不满足要求\n if line_count % 2 != 0:\n return False\n \n checked_files += 1\n except Exception:\n # 忽略读取异常的文件（例如非utf-8编码等个别意外情况）\n continue\n\n # 至少需要检查过一个代码文件才算真正验证通过\n return checked_files > 0","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\"pass\", \"fail\", 或 \"error\"）","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"做得不错！现在我想增强工具的视觉表现和横向对比能力。请在实验舱增加“频闪打点”功能：小球下落时，每隔 0.2 秒在它当前所在位置留下一个半透明的残影（轨迹点），并在残影旁边用微小字体标注出那一刻的时间戳。此外，请将系统升级为“双球对比”模式：在实验舱里再放入第二个不同颜色的小球（球B），控制面板里的参数区进行左右分栏，允许用户分别对球A和球B独立设置不同的材质、质量或初始高度。点击开始时，双球在同一个场景下同时下落，让用户可以直观观察到空气阻力或质量差异对轨迹的影响。\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 生成的代码文件中，任何函数的圈复杂度都不能超过1。2. 修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 考虑到文本最开头可能存在空格、换行等空白字符，先去除前置空白符再进行判断\n cleaned_response = response.lstrip()\n \n # 判断回复最开头是否带有\"喵～\"语气词\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复结尾的空白字符（包括空格、换行符等），检查最后是否以“喵～”结尾\n cleaned_response = response.rstrip()\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n # 使用正则表达式搜索回复中是否存在“我”字。\n # 为了避免把“自我”等词汇误判为第一人称自称，使用反向否定预查 (?<!自) 排除前面是“自”的情况。\n if re.search(r'(?<!自)我', response):\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否使用了\"小可爱\"并且完全替代了\"你\"。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n # 检查是否包含指定的称谓\"小可爱\"\n if \"小可爱\" not in response:\n return False\n \n # 检查是否完全替代了\"你\"（即回复中不能再出现\"你\"字）\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据Web实验工具项目的上下文，推断出涉及的代码文件后缀\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n\n # 若工作区路径无效，直接返回 False\n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n # 仅校验推测出的代码文件类型\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取首行\n first_line = f.readline()\n # 判断是否包含要求的声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果发生文件读取或编码错误，保守判定不符合要求\n return False\n\n # 若遍历结束后找到了代码文件且全符合条件，即为True；若未生成任何代码文件，则认为未完成要求。\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 指定要求中需要检查的结尾声明文本\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 预测并推断当前Web项目可能的代码文件类型\n code_extensions = {\".html\", \".css\", \".js\", \".ts\", \".jsx\", \".vue\", \".py\"}\n \n found_code_files = False\n \n # 遍历当前项目的工作区代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 若存在无法以utf-8读取的文本文件，则默认验证失败\n return False\n \n # 剔除文件结尾纯空白的换行，定位到真正包含内容的最后一行\n while lines and not lines[-1].strip():\n lines.pop()\n \n # 如果文件为空，必然是不符合条件的\n if not lines:\n return False\n \n # 检查最后一行的文本是否包含目标字符串\n last_line = lines[-1].strip()\n if target_statement not in last_line:\n return False\n\n # 若至少验证了一个代码文件且都通过，则返回True\n return found_code_files","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断可能的代码文件类型后缀\n code_extensions = {'.html', '.js', '.css', '.py', '.json'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # readlines() 会按行读取所有内容，包含空行，符合“计入空行”的要求\n lines = f.readlines()\n \n has_code_file = True\n \n # 检查总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果由于非utf-8编码等原因导致无法读取，安全起见忽略或继续检测\n pass\n \n # 要求至少存在代码文件且所有文件行数均为偶数才算真正满足条件\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过1","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\"pass\", \"fail\", 或 \"error\"）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":5,"instruction":"最后一轮，我们加入能量可视化分析和实验导出功能来提升工具的专业度。请在页面底部新增一个横向展开的面板，使用原生 DOM 元素的高度动态变化来模拟柱状图。为每个小球提供三根并排的动态柱子，实时显示其：动能(Ek)、重力势能(Ep)和总机械能(E)。用户可以直观看到由于空气阻力和地面反弹做功导致的机械能衰减过程。最后，在控制面板最下方增加一个“导出实验报告”按钮，点击后能将整个下落过程（以 0.1 秒为采样间隔）中两个球的时间、高度、速度和总机械能数据拼接成一份完整的 CSV 文件，并自动触发浏览器下载。\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 针对所有生成的代码文件有一项限制：每一个文件包含空行在内的总行数不可为偶数，必须是奇数。2. 对于生成的代码文件，请控制行宽，使得每一行不超过80个字符。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除可能存在的首部空白字符，然后判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有发给用户的回复，其末尾必须强制附带语气词\"喵～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容末尾可能出现的空白符以及常见的标点符号\n cleaned_response = response.rstrip(\" \\t\\n\\r。，！？.,!?\")\n \n # 判断清理后的字符串是否以指定的语气词结尾\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"浮浮酱\"而不是\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否遵守了“自我称呼一律使用'浮浮酱'而不是'我'”的要求。\n 通过严格检查回复文本中是否包含“我”字来进行判定。\n \"\"\"\n # 如果回复中出现了“我”字，说明违反了不使用“我”的规定\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"请保持在回复里使用\"小可爱\"对用户进行称谓，以此完全替代\"你\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查是否使用了\"小可爱\"\n if \"小可爱\" not in response:\n return False\n \n # 检查是否完全替代了\"你\"（不能出现\"你\"字）\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的主流代码文件扩展名，排除像 .json、.csv 这样的不支持注释或强要求格式的数据文件\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.py'}\n file_checked = False\n \n # 优先遍历和检查当前项目工作区下的物理文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_checked = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行并去除两端空格及换行符\n first_line = f.readline().strip()\n # 判断是否包含目标编码声明要求\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 出现读取异常（如编码错误或没有访问权限等），稳妥起见按未满足处理\n return False\n \n # 若工作区内没有任何物理形式的代码文件，则转而检查回复响应中的 Markdown 代码块\n if not file_checked:\n # 只匹配那些带有明确语言标注的代码块，过滤掉 bash/sh 等常见终端输入或纯文本记录输出\n code_blocks = re.findall(r'```(\\w+)\\n(.*?)```', response, re.DOTALL)\n for lang, block in code_blocks:\n lang = lang.lower().strip()\n # 常见的前端Web或通用编程语言标记\n if lang in ['html', 'javascript', 'js', 'css', 'typescript', 'ts', 'vue', 'react', 'python', 'py']:\n lines = block.lstrip('\\n').split('\\n')\n if lines and len(lines) > 0:\n first_line = lines[0].strip()\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n file_checked = True\n \n # 如果找到并检查了至少一个代码文件/代码块，且都没有触发 False 判定，则为满足。\n # 反之如果没有找到任何代码，代表未能实际生成带有编码声明的代码文件，返回 False。\n return file_checked","tags":["Content","Format"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 目标要求必须包含的声明字符串\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据用户指令推断，基于Web的物理模拟项目主要包含以下代码文件类型\n code_extensions = ['.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue']\n \n if not os.path.exists(workspace_path):\n return False\n \n code_files = []\n # 遍历工作区，寻找所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果项目中没有任何代码文件，则无法满足约束条件\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 从末尾开始，去掉无意义的空行，锁定最后一个有效行\n while lines and not lines[-1].strip():\n lines.pop()\n \n # 如果代码文件为空，显然没有包含对应注释\n if not lines:\n return False\n \n last_line = lines[-1].strip()\n \n # 判断最后一行是否包含了目标声明子串\n if target_comment not in last_line:\n return False\n \n except Exception:\n # 出现编码错误等读取异常，视为验证失败\n return False\n \n # 所有代码文件都经过检验且符合要求\n return True","tags":["Content","Format"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"针对所有生成的代码文件有一项限制：每一个文件包含空行在内的总行数不可为偶数，必须是奇数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能的Web开发相关代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理推测出的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 以换行符对文件内容进行分割，可以准确统计在编辑器中显示的总行数(包含了首尾的空行)\n lines_count = len(content.split('\\n'))\n \n # 检查总行数是否为偶数\n if lines_count % 2 == 0:\n return False\n \n has_code_file = True\n except Exception:\n # 如果遇到文件读取等异常，跳过继续处理其他文件\n continue\n\n # 如果有处理到代码文件并且都没有触发偶数行的判断，返回 True；否则返回 False\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过1","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\"pass\", \"fail\", 或 \"error\"）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"对于生成的代码文件，请控制行宽，使得每一行不超过80个字符","validation_code":"def check_code_line_width(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 推断可能的代码文件类型：基于Web项目的背景，主要是前端三剑客文件，也可以兼容可能的python脚本\n code_extensions = {'.html', '.js', '.css', '.py'}\n \n # 1. 检查物理工作目录中的相关代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选符合推断的扩展名的代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去掉行尾的换行符后，严格检查字符数是否超过80个字符\n if len(line.rstrip('\\r\\n')) > 80:\n return False\n except Exception:\n # 忽略可能存在读取异常的文件\n continue\n \n # 2. 兜底策略：检查回复内容中的 Markdown 代码块\n code_blocks = re.findall(r'```(?:[a-zA-Z0-9]+)?\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n lines = block.split('\\n')\n for line in lines:\n if len(line.rstrip('\\r\\n')) > 80:\n return False\n \n return True","tags":["Quantity","Range"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我开发一个基于Web的“自由落体物理模拟实验工具”。第一步，请先搭建好基础的页面骨架。需求如下：1. 页面分为左右两栏布局，左侧是占据主要屏幕空间的“实验舱”，请使用 Canvas 或带边框的大块 DOM 区域，底部画一条粗线代表“地面”，实验舱靠上的位置放置一个静止的小球作为实验对象。2. 右侧是“控制面板”，包含三个结构清晰的卡片占位：实验参数设置区、播放控制区（放置开始、暂停、重置三个按钮，暂不实现功能）、实时数据看板。3. 整体界面风格保持现代、极简。目前阶段不需要编写任何物理运动计算逻辑，只要把静态排版和基本UI元素渲染出来即可。\\n另外请注意，后续所有轮次都需要遵循以下要求：\\n1. 每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。\\n2. 你生成给用户的回复内容，必须一直使用中文。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否在最开头带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 忽略可能的前导空白字符，并判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 考虑到助手回复末尾可能带有空白字符或标点符号，先去除末尾常见的空白符及标点\\n cleaned_response = response.rstrip(\\\" \\\\n\\\\r\\\\t。！？.!?\\\")\\n \\n # 判断处理后的回复是否以指定的语气词结尾\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否没有使用\\\"我\\\"。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，如果回复中没有出现\\\"我\\\"，则返回True，否则返回False。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否完全没有出现“你”字\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', \\n '.css', '.scss', '.less', \\n '.js', '.mjs', '.cjs', \\n '.ts', '.tsx', '.jsx', \\n '.vue', '.svelte',\\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.php', '.rb', '.rs'\\n }\\n \\n # 忽略非生成的主流系统或依赖目录\\n exclude_dirs = {'.git', 'node_modules', '.vscode', '.idea', '__pycache__'}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要遍历的子目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline().strip()\\n # 检查严格匹配的子串是否出现在第一行\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果遇到文件不可读或编码异常的情况，也判定为不满足\\n return False\\n \\n # 如果没有生成任何代码文件，代表骨架尚未生成，也判定为未满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 题目要求的必须包含的目标声明字符串\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \\n '.rb', '.php', '.sh', '.swift', '.kt', '.rs'\\n }\\n \\n found_any_file = False\\n \\n # 如果工作区目录不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_any_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于任何权限或读取问题导致读取失败，则算作未满足\\n return False\\n \\n # 过滤出所有包含内容的非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n # 若代码文件内容完全为空，则无法包含要求的注释\\n if not non_empty_lines:\\n return False\\n \\n # 考虑到块状注释可能导致文件最后一行仅存在 `-->` 或 `*/` 等符号\\n # 取末尾的 5 个非空行进行校验，确认是否在结尾处包含了该行注释声明\\n found_target = False\\n for line in non_empty_lines[-5:]:\\n if target_str in line:\\n found_target = True\\n break\\n \\n if not found_target:\\n return False\\n \\n # 如果未检测到任何代码文件，说明助手未能生成代码骨架，返回 False\\n return found_any_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"很好，骨架搭出来了。接下来我们实现核心的物理引擎和基础控制逻辑。请为左侧实验舱建立一个“像素与物理现实”的映射关系（例如设定 50 像素等于 1 米，并在界面角落标出比例尺）。给小球赋予固定的重力加速度（g=9.8m/s²）。实现控制面板里的“开始”、“暂停”和“重置”按钮逻辑：点击开始后，小球开始受重力影响做加速下落运动；触碰到地面时，小球绝对不能穿模，并且需要以 0.7 的弹性系数向上反弹，反复弹跳直到速度趋近于零最终静止；重置按钮能清空状态并让小球回到页面顶部的初始位置。\\n另外请注意，后续所有轮次都需要遵循以下新增要求：1. 生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径。2. 你创建的每一个代码文件，算上空行后的总行数都必须是奇数值。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容前面的空白字符，以防AI生成的回复开头带有空格或换行\\n cleaned_response = response.lstrip()\\n \\n # 检查最开头是否带有\\\"喵～\\\"\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n cleaned_response = response.strip()\\n \\n # 定义句末可能的标点符号集合（中英文常见结束标点及引号）\\n punctuations = \\\"。！？.!?\\\\\\\"'”’\\\"\\n \\n # 循环去除末尾的标点符号及空白字符\\n while cleaned_response and cleaned_response[-1] in punctuations:\\n cleaned_response = cleaned_response[:-1].strip()\\n \\n # 判断去除标点后的回复末尾是否为指定的语气词\\n return cleaned_response.endswith(\\\"喵～\\\") or cleaned_response.endswith(\\\"喵~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 1. 检查回复内容中是否包含自称“我”\\n if \\\"我\\\" in response:\\n return False\\n \\n # 2. 检查生成的工作区代码文件中是否包含“我”\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if \\\"我\\\" in content:\\n return False\\n except Exception:\\n # 忽略非文本或无法读取的文件\\n pass\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if \\\"小可爱\\\" in response and \\\"你\\\" not in response:\\n return True\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 推测的可能的代码文件类型，贴合前端Web开发任务\\n code_exts = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的非业务代码文件夹\\n if '.git' in root or 'node_modules' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查是否为空文件以及首行是否严格包含声明\\n if not first_line or \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果发生文件读取或解码错误，判定为验证失败\\n return False\\n \\n # 只要成功验证了至少一个生成的代码文件且全都符合要求，即返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 结合Web页面开发需求推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n found_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n found_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 文件读取失败时判定未满足\\n return False\\n \\n # 查找倒序的第一个非空行作为最后一行的内容\\n last_line = \\\"\\\"\\n for line in reversed(lines):\\n if line.strip():\\n last_line = line.strip()\\n break\\n \\n # 验证结尾的新行注释中是否包含了指定的声明内容\\n if target_statement not in last_line:\\n return False\\n\\n # 若未找到任何代码文件，则无法证明其满足了要求\\n return found_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 常见的Web及其他代码文件的后缀集合\\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符切分统计行数，能够把所有空行（含末尾空行）都算入其中\\n lines_count = len(content.split('\\\\n'))\\n \\n # 如果行数是偶数，说明不满足要求\\n if lines_count % 2 == 0:\\n return False\\n except Exception:\\n # 如果有异常（如编码非utf-8或无法读取），为避免程序崩溃此处做跳过处理\\n continue\\n \\n # 如果遍历检查完没有返回 False 且确保找到了代码文件，则符合所有条件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来完善参数控制和数据监控。在右侧的“实验参数设置区”中，添加三个可交互的输入框，允许用户自定义：初始高度（范围 1-100 米）、重力加速度（允许修改，以便模拟月球等不同星体环境）、小球质量（千克）。同时，在“实时数据看板”中，请使用清晰的字体实时更新并显示小球在运动过程中的核心物理量：已用时间（秒，保留两位小数）、当前速度（米/秒）和离地高度（米）。要求：当用户在运行前拖动滑块或修改“初始高度”参数时，左侧实验舱里小球的垂直位置需要实时同步调整。\\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数。2. 生成的代码文件中，任何函数的圈复杂度都不能超过3。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除前导空白字符后，判断是否以“喵～”开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除字符串首尾的空白字符\\n cleaned_response = response.strip()\\n \\n # 如果直接以\\\"喵～\\\"结尾，则满足条件\\n if cleaned_response.endswith(\\\"喵～\\\"):\\n return True\\n \\n # 考虑到末尾可能带有标点符号，去除常见的结束标点后再进行判断\\n cleaned_response_without_punct = cleaned_response.rstrip(\\\"。！？.!?\\\")\\n \\n return cleaned_response_without_punct.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的回复中是否没有出现\\\"我\\\"字。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 如果回复中不包含\\\"我\\\"，返回True；否则返回False\\n \\\"\\\"\\\"\\n return \\\"我\\\" not in response\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含\\\"小可爱\\\"，且是否完全没有出现\\\"你\\\"字\\n if \\\"小可爱\\\" in response and \\\"你\\\" not in response:\\n return True\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义可能生成的前端及相关代码文件扩展名\\n code_extensions = {\\\".html\\\", \\\".htm\\\", \\\".css\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".py\\\"}\\n has_code_file = False\\n \\n # 遍历代码项目路径\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 如果文件为空或者第一行没有包含规定的编码声明，则不满足要求\\n if not first_line or \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若文件读取发生异常，则无法确认其满足要求，返回False\\n return False\\n \\n # 如果项目中没有任何代码文件，说明未成功生成代码，返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json'}\\n file_checked_count = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能存在的第三方依赖或隐藏文件夹\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_checked_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 寻找文件末尾的最后有效非空行\\n last_valid_line = \\\"\\\"\\n for line in reversed(lines):\\n if line.strip():\\n last_valid_line = line.strip()\\n break\\n \\n # 检查最后非空行中是否包含了要求的声明文本\\n if target_statement not in last_valid_line:\\n return False\\n except Exception:\\n # 如果读取报错（如编码问题等），判定为未满足要求\\n return False\\n \\n # 至少要验证了一个文件才算真正满足了该指令\\n return file_checked_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断该Web项目可能涉及的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.py'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前工作目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n has_code_file = True\\n \\n # 计入空行的总行数统计方式：按换行符分隔\\n # 若文件为空字符，行数记为0；否则通过split('\\\\n')统计物理行数\\n line_count = len(content.split('\\\\n')) if content else 0\\n \\n # 判断行数是否为偶数\\n if line_count % 2 != 0:\\n return False\\n except Exception:\\n # 忽略读取失败的非文本/非UTF-8文件\\n pass\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过3\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"为了让物理模拟更真实且具备探究价值，下一步我们需要打破理想环境，引入“空气阻力”模型。请在参数设置区新增一个“材质与形状”的下拉菜单（选项包含：实心铁球、木球、带降落伞的沙袋）。选择不同的选项时，系统在后台自动赋予小球不同的风阻系数(C)和迎风面积(A)。在小球的下落运动演算中加入空气阻力公式（F = 0.5 * ρ * v² * C * A，空气密度ρ设为常数）。由于阻力与速度平方成正比，小球最终应能达到终端速度。另外，请在实时数据看板中新增一项，实时显示当前小球受到的空气阻力大小（牛顿）。\\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 生成的代码文件中，任何函数的圈复杂度都不能超过2。2. 每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\\\"pass\\\", \\\"fail\\\", 或 \\\"error\\\"）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n # 考虑到文本最前方可能会包含空格或换行等不可见字符，使用lstrip()去除左侧空白后判断\\n if response.lstrip().startswith(\\\"喵～\\\"):\\n return True\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的回复末尾是否附带语气词\\\"喵～\\\"\\n \\\"\\\"\\\"\\n # 移除回复内容末尾可能包含的空白字符（如换行符、空格等）\\n cleaned_response = response.strip()\\n \\n # 检查结尾是否为指定语气词\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否遵循了不使用“我”作为自我称呼的要求。\\n 出于自动化评测的准确性和鲁棒性，这里采用严格匹配：回复中不得出现“我”字。\\n \\\"\\\"\\\"\\n # 如果回复中存在“我”字，则判定为不满足要求\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复内容中是否包含禁止使用的“你”字\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义可能生成的常见代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.py', '.ts', '.jsx', '.tsx'}\\n found_code_file = False\\n \\n # 遍历当前项目的工作区路径\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行，去除可能包含的头尾空白字符与换行符\\n first_line = f.readline().strip()\\n \\n # 检查声明编码的标识是否存在于该文件的第一行\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何异常（如非UTF-8编码导致读取失败），判定为不符合要求\\n return False\\n \\n # 如果遍历完毕未触发False，且至少找到了一个代码文件，则符合要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户指令推断可能涉及到的代码文件后缀，主要包含网页前端、JSON测试结果等\\n code_extensions = {'.html', '.css', '.js', '.json', '.py', '.ts', '.jsx', '.tsx'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的无关文件夹\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于编码或其他原因无法读取，默认认为不满足格式要求\\n return False\\n \\n # 寻找代码文件中的倒数第一行（忽略结尾的纯空行）\\n last_non_empty_line = \\\"\\\"\\n for line in reversed(lines):\\n if line.strip():\\n last_non_empty_line = line\\n break\\n \\n # 如果最后非空行不存在，或者指定的声明文字未出现在最后一行，则验证失败\\n if target_statement not in last_non_empty_line:\\n return False\\n \\n # 只要生成了代码文件并且均通过了检查，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，说明未创建任何文件\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据项目推断可能的代码文件类型后缀\\n target_extensions = {'.html', '.css', '.js', '.json', '.ts', '.jsx', '.tsx', '.vue'}\\n checked_files = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的包管理目录或版本控制目录\\n if any(ignored in root.split(os.sep) for ignored in ['.git', 'node_modules', '__pycache__']):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用 split('\\\\n') 进行分割可以精确计算计入空行后的总行数\\n line_count = len(content.split('\\\\n'))\\n \\n # 如果有任何一个代码文件的行数为奇数，则不满足要求\\n if line_count % 2 != 0:\\n return False\\n \\n checked_files += 1\\n except Exception:\\n # 忽略读取异常的文件（例如非utf-8编码等个别意外情况）\\n continue\\n\\n # 至少需要检查过一个代码文件才算真正验证通过\\n return checked_files > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\\\"pass\\\", \\\"fail\\\", 或 \\\"error\\\"）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"做得不错！现在我想增强工具的视觉表现和横向对比能力。请在实验舱增加“频闪打点”功能：小球下落时，每隔 0.2 秒在它当前所在位置留下一个半透明的残影（轨迹点），并在残影旁边用微小字体标注出那一刻的时间戳。此外，请将系统升级为“双球对比”模式：在实验舱里再放入第二个不同颜色的小球（球B），控制面板里的参数区进行左右分栏，允许用户分别对球A和球B独立设置不同的材质、质量或初始高度。点击开始时，双球在同一个场景下同时下落，让用户可以直观观察到空气阻力或质量差异对轨迹的影响。\\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 生成的代码文件中，任何函数的圈复杂度都不能超过1。2. 修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 考虑到文本最开头可能存在空格、换行等空白字符，先去除前置空白符再进行判断\\n cleaned_response = response.lstrip()\\n \\n # 判断回复最开头是否带有\\\"喵～\\\"语气词\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复结尾的空白字符（包括空格、换行符等），检查最后是否以“喵～”结尾\\n cleaned_response = response.rstrip()\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 使用正则表达式搜索回复中是否存在“我”字。\\n # 为了避免把“自我”等词汇误判为第一人称自称，使用反向否定预查 (?<!自) 排除前面是“自”的情况。\\n if re.search(r'(?<!自)我', response):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否使用了\\\"小可爱\\\"并且完全替代了\\\"你\\\"。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n # 检查是否包含指定的称谓\\\"小可爱\\\"\\n if \\\"小可爱\\\" not in response:\\n return False\\n \\n # 检查是否完全替代了\\\"你\\\"（即回复中不能再出现\\\"你\\\"字）\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据Web实验工具项目的上下文，推断出涉及的代码文件后缀\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n\\n # 若工作区路径无效，直接返回 False\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n # 仅校验推测出的代码文件类型\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取首行\\n first_line = f.readline()\\n # 判断是否包含要求的声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果发生文件读取或编码错误，保守判定不符合要求\\n return False\\n\\n # 若遍历结束后找到了代码文件且全符合条件，即为True；若未生成任何代码文件，则认为未完成要求。\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 指定要求中需要检查的结尾声明文本\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 预测并推断当前Web项目可能的代码文件类型\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".vue\\\", \\\".py\\\"}\\n \\n found_code_files = False\\n \\n # 遍历当前项目的工作区代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 若存在无法以utf-8读取的文本文件，则默认验证失败\\n return False\\n \\n # 剔除文件结尾纯空白的换行，定位到真正包含内容的最后一行\\n while lines and not lines[-1].strip():\\n lines.pop()\\n \\n # 如果文件为空，必然是不符合条件的\\n if not lines:\\n return False\\n \\n # 检查最后一行的文本是否包含目标字符串\\n last_line = lines[-1].strip()\\n if target_statement not in last_line:\\n return False\\n\\n # 若至少验证了一个代码文件且都通过，则返回True\\n return found_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断可能的代码文件类型后缀\\n code_extensions = {'.html', '.js', '.css', '.py', '.json'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # readlines() 会按行读取所有内容，包含空行，符合“计入空行”的要求\\n lines = f.readlines()\\n \\n has_code_file = True\\n \\n # 检查总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果由于非utf-8编码等原因导致无法读取，安全起见忽略或继续检测\\n pass\\n \\n # 要求至少存在代码文件且所有文件行数均为偶数才算真正满足条件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过1\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\\\"pass\\\", \\\"fail\\\", 或 \\\"error\\\"）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后一轮，我们加入能量可视化分析和实验导出功能来提升工具的专业度。请在页面底部新增一个横向展开的面板，使用原生 DOM 元素的高度动态变化来模拟柱状图。为每个小球提供三根并排的动态柱子，实时显示其：动能(Ek)、重力势能(Ep)和总机械能(E)。用户可以直观看到由于空气阻力和地面反弹做功导致的机械能衰减过程。最后，在控制面板最下方增加一个“导出实验报告”按钮，点击后能将整个下落过程（以 0.1 秒为采样间隔）中两个球的时间、高度、速度和总机械能数据拼接成一份完整的 CSV 文件，并自动触发浏览器下载。\\n另外请注意，后续所有轮次都需要遵循以下调整和新增要求：1. 针对所有生成的代码文件有一项限制：每一个文件包含空行在内的总行数不可为偶数，必须是奇数。2. 对于生成的代码文件，请控制行宽，使得每一行不超过80个字符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除可能存在的首部空白字符，然后判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有发给用户的回复，其末尾必须强制附带语气词\\\"喵～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容末尾可能出现的空白符以及常见的标点符号\\n cleaned_response = response.rstrip(\\\" \\\\t\\\\n\\\\r。，！？.,!?\\\")\\n \\n # 判断清理后的字符串是否以指定的语气词结尾\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"浮浮酱\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否遵守了“自我称呼一律使用'浮浮酱'而不是'我'”的要求。\\n 通过严格检查回复文本中是否包含“我”字来进行判定。\\n \\\"\\\"\\\"\\n # 如果回复中出现了“我”字，说明违反了不使用“我”的规定\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保持在回复里使用\\\"小可爱\\\"对用户进行称谓，以此完全替代\\\"你\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查是否使用了\\\"小可爱\\\"\\n if \\\"小可爱\\\" not in response:\\n return False\\n \\n # 检查是否完全替代了\\\"你\\\"（不能出现\\\"你\\\"字）\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的主流代码文件扩展名，排除像 .json、.csv 这样的不支持注释或强要求格式的数据文件\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.py'}\\n file_checked = False\\n \\n # 优先遍历和检查当前项目工作区下的物理文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_checked = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行并去除两端空格及换行符\\n first_line = f.readline().strip()\\n # 判断是否包含目标编码声明要求\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 出现读取异常（如编码错误或没有访问权限等），稳妥起见按未满足处理\\n return False\\n \\n # 若工作区内没有任何物理形式的代码文件，则转而检查回复响应中的 Markdown 代码块\\n if not file_checked:\\n # 只匹配那些带有明确语言标注的代码块，过滤掉 bash/sh 等常见终端输入或纯文本记录输出\\n code_blocks = re.findall(r'```(\\\\w+)\\\\n(.*?)```', response, re.DOTALL)\\n for lang, block in code_blocks:\\n lang = lang.lower().strip()\\n # 常见的前端Web或通用编程语言标记\\n if lang in ['html', 'javascript', 'js', 'css', 'typescript', 'ts', 'vue', 'react', 'python', 'py']:\\n lines = block.lstrip('\\\\n').split('\\\\n')\\n if lines and len(lines) > 0:\\n first_line = lines[0].strip()\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n file_checked = True\\n \\n # 如果找到并检查了至少一个代码文件/代码块，且都没有触发 False 判定，则为满足。\\n # 反之如果没有找到任何代码，代表未能实际生成带有编码声明的代码文件，返回 False。\\n return file_checked\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 目标要求必须包含的声明字符串\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据用户指令推断，基于Web的物理模拟项目主要包含以下代码文件类型\\n code_extensions = ['.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue']\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files = []\\n # 遍历工作区，寻找所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果项目中没有任何代码文件，则无法满足约束条件\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 从末尾开始，去掉无意义的空行，锁定最后一个有效行\\n while lines and not lines[-1].strip():\\n lines.pop()\\n \\n # 如果代码文件为空，显然没有包含对应注释\\n if not lines:\\n return False\\n \\n last_line = lines[-1].strip()\\n \\n # 判断最后一行是否包含了目标声明子串\\n if target_comment not in last_line:\\n return False\\n \\n except Exception:\\n # 出现编码错误等读取异常，视为验证失败\\n return False\\n \\n # 所有代码文件都经过检验且符合要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"针对所有生成的代码文件有一项限制：每一个文件包含空行在内的总行数不可为偶数，必须是奇数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能的Web开发相关代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理推测出的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 以换行符对文件内容进行分割，可以准确统计在编辑器中显示的总行数(包含了首尾的空行)\\n lines_count = len(content.split('\\\\n'))\\n \\n # 检查总行数是否为偶数\\n if lines_count % 2 == 0:\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果遇到文件读取等异常，跳过继续处理其他文件\\n continue\\n\\n # 如果有处理到代码文件并且都没有触发偶数行的判断，返回 True；否则返回 False\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过1\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每完成一次代码生成的测试，都必须将结果更新至test_results.json文件。文件类型须为List[Dict]，每个字典元素规定包含两个属性：description记载本次测试内容，status标识测试状态（仅允许设为\\\"pass\\\", \\\"fail\\\", 或 \\\"error\\\"）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，请控制行宽，使得每一行不超过80个字符\",\n \"validation_code\": \"def check_code_line_width(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 推断可能的代码文件类型：基于Web项目的背景，主要是前端三剑客文件，也可以兼容可能的python脚本\\n code_extensions = {'.html', '.js', '.css', '.py'}\\n \\n # 1. 检查物理工作目录中的相关代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选符合推断的扩展名的代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去掉行尾的换行符后，严格检查字符数是否超过80个字符\\n if len(line.rstrip('\\\\r\\\\n')) > 80:\\n return False\\n except Exception:\\n # 忽略可能存在读取异常的文件\\n continue\\n \\n # 2. 兜底策略：检查回复内容中的 Markdown 代码块\\n code_blocks = re.findall(r'```(?:[a-zA-Z0-9]+)?\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n for line in lines:\\n if len(line.rstrip('\\\\r\\\\n')) > 80:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面整体静态排版，确认左侧存在主要区域“实验舱”（含底部代表地面的粗线和界面角落的比例尺），实验舱内有两颗颜色不同的小球；右侧有控制面板；页面底部包含能量分析的柱状图面板。","查看右侧“实验参数设置区”，确认其采用了分栏布局以分别控制两个小球，且每个小球独立具备：初始高度（范围1-100米）、重力加速度、质量的输入框或滑块，以及“材质与形状”（如实心铁球、木球、带降落伞的沙袋）下拉菜单。","在未启动实验（静止状态）下，通过数值输入或滑块拖动调整任一小球的“初始高度”，确认左侧实验舱中该小球的垂直位置会实时上下移动并与设定的高度对应。","将球A的“重力加速度”设为极大值（如20或更高），球B保持9.8，点击“开始”按钮，确认两个小球同时从各自初始位置开始下落，且重力加速度大的球下落速度明显更快。","将球A材质选为“实心铁球”，球B材质选为“带降落伞的沙袋”，高度相同，点击“开始”按钮。观察下落过程，确认引入空气阻力后，球B的下落速度由于空气阻力较大而明显慢于球A。","点击“开始”让小球下落，在半空中点击“暂停”按钮，确认所有小球瞬间悬停在空中不再移动；再次点击“开始”，小球从原悬停位置继续加速下落。","点击“开始”按钮，紧盯小球触底的瞬间，确认小球没有穿透底部代表地面的粗线（不穿模），而是向上发生反弹，且第一次反弹高度约为初始下落高度的0.49倍（弹性系数0.7），每次反弹的最高点逐渐降低，最后完全静止在地面粗线上。","在小球正在下落或已静止在底部的状态下，点击“重置”按钮，确认左侧实验舱的小球立刻返回页面顶部的初始设定高度位置，且处于静止状态；同时已用时间、速度、轨迹残影、能量柱状图等所有运动状态被完全清空。","点击“开始”按钮让小球下落，观察左侧实验舱，确认两个小球在下落路线上每隔0.2秒会留下一个半透明的残影轨迹点，且每个残影旁边都有微小字体标注的具体时间戳数值。","点击“开始”按钮后，观察右侧“实时数据看板”，确认面板中的数字在小球运动期间持续快速刷新，且分别显示两个小球的：已用时间（保留两位小数）、当前速度、离地高度、空气阻力大小（牛顿）。","点击“开始”按钮后，观察页面底部的能量柱状图面板，确认每个小球对应的三根柱子（动能、重力势能、总机械能）的高度随下落实时动态变化，且能在下落过程中观察到重力势能柱缩短、动能柱长高、总机械能柱逐渐降低（机械能衰减）的现象。","点击控制面板最下方的“导出实验报告”按钮，检查浏览器是否成功触发文件下载弹窗，且下载的文件后缀为 .csv 格式；打开文件确认包含两个球的时间、高度、速度和总机械能数据，且采样间隔严格为0.1秒。","选择“带降落伞的沙袋”材质，点击“开始”按钮，观察实时数据看板中的速度值，确认小球下落一段时间后速度趋于稳定（达到终端速度）。","为球A和球B设置不同的重力加速度、材质和质量参数，点击“开始”按钮，确认两个小球的运动过程分别符合各自的参数设置，参数独立生效互不干扰。"],"string":"[\n \"查看页面整体静态排版，确认左侧存在主要区域“实验舱”（含底部代表地面的粗线和界面角落的比例尺），实验舱内有两颗颜色不同的小球；右侧有控制面板；页面底部包含能量分析的柱状图面板。\",\n \"查看右侧“实验参数设置区”，确认其采用了分栏布局以分别控制两个小球，且每个小球独立具备：初始高度（范围1-100米）、重力加速度、质量的输入框或滑块，以及“材质与形状”（如实心铁球、木球、带降落伞的沙袋）下拉菜单。\",\n \"在未启动实验（静止状态）下，通过数值输入或滑块拖动调整任一小球的“初始高度”，确认左侧实验舱中该小球的垂直位置会实时上下移动并与设定的高度对应。\",\n \"将球A的“重力加速度”设为极大值（如20或更高），球B保持9.8，点击“开始”按钮，确认两个小球同时从各自初始位置开始下落，且重力加速度大的球下落速度明显更快。\",\n \"将球A材质选为“实心铁球”，球B材质选为“带降落伞的沙袋”，高度相同，点击“开始”按钮。观察下落过程，确认引入空气阻力后，球B的下落速度由于空气阻力较大而明显慢于球A。\",\n \"点击“开始”让小球下落，在半空中点击“暂停”按钮，确认所有小球瞬间悬停在空中不再移动；再次点击“开始”，小球从原悬停位置继续加速下落。\",\n \"点击“开始”按钮，紧盯小球触底的瞬间，确认小球没有穿透底部代表地面的粗线（不穿模），而是向上发生反弹，且第一次反弹高度约为初始下落高度的0.49倍（弹性系数0.7），每次反弹的最高点逐渐降低，最后完全静止在地面粗线上。\",\n \"在小球正在下落或已静止在底部的状态下，点击“重置”按钮，确认左侧实验舱的小球立刻返回页面顶部的初始设定高度位置，且处于静止状态；同时已用时间、速度、轨迹残影、能量柱状图等所有运动状态被完全清空。\",\n \"点击“开始”按钮让小球下落，观察左侧实验舱，确认两个小球在下落路线上每隔0.2秒会留下一个半透明的残影轨迹点，且每个残影旁边都有微小字体标注的具体时间戳数值。\",\n \"点击“开始”按钮后，观察右侧“实时数据看板”，确认面板中的数字在小球运动期间持续快速刷新，且分别显示两个小球的：已用时间（保留两位小数）、当前速度、离地高度、空气阻力大小（牛顿）。\",\n \"点击“开始”按钮后，观察页面底部的能量柱状图面板，确认每个小球对应的三根柱子（动能、重力势能、总机械能）的高度随下落实时动态变化，且能在下落过程中观察到重力势能柱缩短、动能柱长高、总机械能柱逐渐降低（机械能衰减）的现象。\",\n \"点击控制面板最下方的“导出实验报告”按钮，检查浏览器是否成功触发文件下载弹窗，且下载的文件后缀为 .csv 格式；打开文件确认包含两个球的时间、高度、速度和总机械能数据，且采样间隔严格为0.1秒。\",\n \"选择“带降落伞的沙袋”材质，点击“开始”按钮，观察实时数据看板中的速度值，确认小球下落一段时间后速度趋于稳定（达到终端速度）。\",\n \"为球A和球B设置不同的重力加速度、材质和质量参数，点击“开始”按钮，确认两个小球的运动过程分别符合各自的参数设置，参数独立生效互不干扰。\"\n]","isConversational":false}}},{"rowIdx":48,"cells":{"task_id":{"kind":"number","value":49,"string":"49"},"repository_policy":{"kind":"string","value":"你后续所有操作及所有轮次回复均需严格遵循以下约束，不得违反：\n1. 修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它。\n2. 每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件。\n3. 你所生成的所有代码文件都必须使用UTF-8进行编码。\n4. 生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n5. 切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式。\n6. 你提供的代码项目，其目录层级最高不能超过2层。\n7. 每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0。\n8. 每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的扩展名集合\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.scss', '.less',\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.kts', '.m', '.mm',\n '.sh', '.bash', '.bat', '.ps1', '.json', '.xml', '.yaml', '.yml', '.sql'\n }\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码检测\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试将二进制数据解码为 utf-8\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明文件不符合 UTF-8 编码（例如为 GBK 等其他编码）\n return False\n except Exception:\n # 忽略诸如权限不足之类的其他读取异常\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见的代码文件扩展名，根据实际需求推断\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cs', '.js', '.ts', '.html', '.css', \n '.go', '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.m', '.h', '.hpp'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n found = False\n # 检查文件开头处（为增加鲁棒性，检查前10行以防止有shebang或空行等）\n for line in lines[:10]:\n if target_comment in line:\n found = True\n break\n \n # 如果有任何一个代码文件不包含该注释，则判定要求未满足\n if not found:\n return False\n except Exception:\n # 如果文件读取失败或不是UTF-8编码，判定失败\n return False\n \n # 如果存在至少一个代码文件且都通过了检查，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"import os\nfrom pathlib import Path\n\ndef verify(response: str, workspace_path: str) -> bool:\n workspace = Path(workspace_path)\n \n # 确保给定的项目路径确实存在且为目录\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for root, dirs, files in os.walk(workspace):\n # 计算当前遍历目录相对于工作区根目录的相对路径\n rel_path = Path(root).relative_to(workspace)\n \n # 根目录自身的 parts 长度为 0（第1层）\n # 一级子目录的 parts 长度为 1（第2层）\n # 如果 parts 长度 >= 2，说明该目录为二级或更深的子目录（第3层或以上），违反了最高2层的规定\n if len(rel_path.parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]}],"string":"[\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的扩展名集合\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.scss', '.less',\\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.kts', '.m', '.mm',\\n '.sh', '.bash', '.bat', '.ps1', '.json', '.xml', '.yaml', '.yml', '.sql'\\n }\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码检测\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试将二进制数据解码为 utf-8\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明文件不符合 UTF-8 编码（例如为 GBK 等其他编码）\\n return False\\n except Exception:\\n # 忽略诸如权限不足之类的其他读取异常\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义常见的代码文件扩展名，根据实际需求推断\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cs', '.js', '.ts', '.html', '.css', \\n '.go', '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.m', '.h', '.hpp'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n found = False\\n # 检查文件开头处（为增加鲁棒性，检查前10行以防止有shebang或空行等）\\n for line in lines[:10]:\\n if target_comment in line:\\n found = True\\n break\\n \\n # 如果有任何一个代码文件不包含该注释，则判定要求未满足\\n if not found:\\n return False\\n except Exception:\\n # 如果文件读取失败或不是UTF-8编码，判定失败\\n return False\\n \\n # 如果存在至少一个代码文件且都通过了检查，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n workspace = Path(workspace_path)\\n \\n # 确保给定的项目路径确实存在且为目录\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for root, dirs, files in os.walk(workspace):\\n # 计算当前遍历目录相对于工作区根目录的相对路径\\n rel_path = Path(root).relative_to(workspace)\\n \\n # 根目录自身的 parts 长度为 0（第1层）\\n # 一级子目录的 parts 长度为 1（第2层）\\n # 如果 parts 长度 >= 2，说明该目录为二级或更深的子目录（第3层或以上），违反了最高2层的规定\\n if len(rel_path.parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"UI/UX Optimization"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我搭一个页面通知组件（Notification）的基础框架。页面中间放一个'发送通知'的按钮。点击按钮后，在屏幕右下角能弹出一个简单的黑底白字消息卡片。现在暂时不需要做复杂的动画和业务逻辑，只要让它能渲染在右下角，并且在卡片上提供一个'X'按钮让我能点击销毁它就可以了。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _notif 作为结尾；二是你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，页面通知组件的代码大概率涉及以下前端文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 如果有其他异常（如文件无法读取），保守跳过或视作无效情况处理，这里不做判定失败\n continue\n \n # 所有代码文件都顺利通过了 UTF-8 编码测试\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = []\n # 提取文件开头的前5个非空行\n for line in f:\n if line.strip():\n lines.append(line)\n if len(lines) >= 5:\n break\n \n # 检查前几个非空行中是否包含要求中的注释字符串\n found_comment = False\n for line in lines:\n if target_comment in line:\n found_comment = True\n break\n \n if not found_comment:\n return False\n except Exception:\n # 如果遇到无法以UTF-8读取的文件或其他异常情况，默认视为验证失败\n return False\n \n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 计算当前目录相对于根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录属于第1层，直接跳过\n if rel_path == '.':\n continue\n \n # 统一路径分隔符为 '/'，以此切分相对路径\n parts = rel_path.replace('\\\\', '/').split('/')\n \n # 如果相对路径包含至少两部分（如 dir1/dir2），说明目录达到了第3层，即超过了2层\n if len(parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _notif 作为结尾","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能的生成的代码文件类型\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.py'}\n \n # 单行正则匹配模式\n patterns = [\n # JavaScript / TypeScript /等传统 function 声明: function foo(...)\n r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(',\n # 箭头函数赋值声明: const foo = (...) => 或 const foo = arg =>\n r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>',\n # 匿名函数赋值声明: const foo = function(...)\n r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s*)?function\\b',\n # 对象属性 function声明: foo: function()\n r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s*)?function\\b',\n # 对象属性 箭头函数声明: foo: () =>\n r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>',\n ]\n\n # 多行正则匹配模式\n multiline_patterns = [\n # Python: def foo(...)\n r'^[ \\t]*def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(',\n # JavaScript/TypeScript 类或对象的方法简写: foo() { ... } 或 async foo() { ... }\n r'^[ \\t]*(?:async[ \\t]+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{',\n # JavaScript/TypeScript 类的箭头函数属性: foo = () => { ... }\n r'^[ \\t]*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>'\n ]\n \n # 过滤掉由于简单的正则可能误判的编程语言关键字\n keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'catch', 'function',\n 'constructor', 'return', 'import', 'export', 'super', 'with', 'new', 'typeof',\n 'await', 'yield'\n }\n\n # 遍历项目目录结构\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见依赖、日志和编译输出缓存目录\n for ignore_dir in ['node_modules', '.git', '__pycache__', 'dist', 'build']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n func_names = set()\n \n # 收集函数名\n for p in patterns:\n for match in re.finditer(p, code):\n func_names.add(match.group(1))\n \n for p in multiline_patterns:\n for match in re.finditer(p, code, re.MULTILINE):\n func_names.add(match.group(1))\n \n # 验证函数名是否符合结尾要求\n for name in func_names:\n # 忽略控制流/原生关键字的混淆\n if name in keywords:\n continue\n \n if not name.endswith('_notif'):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"import os\nimport re\n\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个数字是否是质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 如果工作区不存在，无法验证具体代码，视作没有违规的类名\n if not os.path.exists(workspace_path):\n return True\n\n # 正则表达式定义\n # 1. 匹配面向对象语言中的类名 (如 class NotificationBox)\n class_def_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_]\\w*)')\n \n # 2. 匹配 HTML/JSX/Vue 中的 class 属性 (如 class=\"card-item\" 或 className=\"card-item\")\n html_class_pattern = re.compile(r'\\b(?:class|className)\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']')\n \n # 3. 匹配 CSS 文件中定义的类名 (如 .card-item)\n css_class_pattern = re.compile(r'\\.([a-zA-Z_][a-zA-Z0-9_-]*)')\n\n # 支持的前端及常规代码文件类型\n valid_extensions = ('.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.py')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(valid_extensions):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 1. 检查面向对象的类名\n for class_name in class_def_pattern.findall(content):\n if is_prime(len(class_name)):\n return False\n \n # 2. 检查 HTML/React 中的类名属性\n for class_str in html_class_pattern.findall(content):\n # class 属性里可能用空格分隔了多个类名，如 class=\"btn primary\"\n class_names = class_str.strip().split()\n for class_name in class_names:\n if not class_name:\n continue\n if is_prime(len(class_name)):\n return False\n \n # 3. 仅在 .css 文件中严格检查 CSS 类名定义（避免在 JS 文件中把 object.property 误认为 css 类）\n if file.endswith('.css'):\n for class_name in css_class_pattern.findall(content):\n if is_prime(len(class_name)):\n return False\n \n except Exception:\n # 忽略读取或编码异常的文件\n pass\n\n # 遍历结束没有发现任何长度为质数的类名，则满足“每个类名的长度恰好是一个非质数”的要求\n return True","tags":["Style","Naming"]}]},{"round_id":1,"instruction":"不错，现在我们给通知加上生命周期和基础动画。要求通知出现时，是从屏幕右侧平滑滑入并带有淡入（fade in）效果；显示停留3秒后，它要能自动向右滑动淡出并从DOM中彻底移除。如果我在3秒内手动点击那个关闭按钮，它也要播放同样的退出动画后再销毁。另外请你后续轮次始终遵循以下两条更新规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_anim 作为结尾；二是请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见前端页面组件相关的代码文件后缀\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n found_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试用UTF-8解码文件内容，若非UTF-8编码（且包含非ASCII字符）将抛出UnicodeDecodeError\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 发现无法使用UTF-8解码的代码文件，违反了要求\n return False\n except Exception:\n # 遇到其他读取错误（如权限等），保守视作验证失败\n return False\n \n # 只要发现了代码文件，且全部通过了UTF-8解码测试，即认为满足要求\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 待校验的准确注释目标字符串\n target_string = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据页面组件开发场景，可能包含的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.php', '.java', '.py', '.c', '.cpp', '.cs', '.go', '.rb'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 只针对代码文件进行校验，忽略诸如.log文件或没有后缀名的特殊文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 检查文件开头：读取前5行内容\n # 留有前5行的余量是为了兼容类似于多行注释的第一行可能是 `<!--` 或 `/*` 等情况\n lines = [f.readline() for _ in range(5)]\n \n found = False\n for line in lines:\n if target_string in line:\n found = True\n break\n \n # 只要有一个代码文件没在开头带有该注释，直接返回False\n if not found:\n return False\n has_code_file = True\n except Exception:\n # 文件读取失败等情况，保守判定为未通过\n return False\n \n # 如果项目中没有找到任何有效代码文件，自然也算作未完成该需求\n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n \n workspace = Path(workspace_path)\n \n # 如果工作区不存在或不是一个目录，无法验证项目，视为不满足条件\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for root, dirs, files in os.walk(workspace):\n # 过滤掉以 '.' 开头的隐藏目录和特殊的 '__pycache__' 缓存目录，防止误判\n # 注意：使用切片赋值 dirs[:] 可以原地修改列表，从而阻止 os.walk 继续遍历这些被排除的目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = Path(root).relative_to(workspace)\n \n # rel_path.parts 的长度代表了当前目录的层级深度：\n # - 长度为 0 表示当前目录是 workspace 根目录本身\n # - 长度为 1 表示当前目录是一级目录（如 workspace/src）\n # - 长度为 2 表示当前目录是二级目录（如 workspace/src/components）\n # - 长度 > 2 表示存在三级或更深的目录，这违反了“层级最高不能超过2层”的要求\n if len(rel_path.parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_anim 作为结尾","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}\n \n # 正则1：匹配常规声明 `function name(...)` 或 generator函数 `function* name(...)`\n pattern_func_decl = re.compile(r'\\bfunction\\b\\s*\\*?\\s*([a-zA-Z_$]\\w*)\\s*\\(')\n \n # 正则2：匹配显式变量赋值定义的函数 `const/let/var name = function()` 或 `const name = () =>`\n pattern_var_decl = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$]\\w*)\\s*=\\s*(?:async\\s+)?(?:function\\b|[^;{=]*?=>)')\n \n # 正则3：匹配在类属性中独立声明的箭头函数形式 `name = () =>`（限制行首与修饰符，避免误伤JSX属性）\n pattern_class_prop = re.compile(r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+|readonly\\s+)*([a-zA-Z_$]\\w*)\\s*=\\s*(?:async\\s+)?(?:function\\b|[^;{=]*?=>)', re.MULTILINE)\n \n # 正则4：匹配类方法/对象方法 `name() { ... }` \n # 使用 \\([^()]*\\) 确保圆括号内不能再嵌套内部的左括号，防止误判了回调函数调用(如 setTimeout(function(){)\n pattern_method = re.compile(r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+|async\\s+|get\\s+|set\\s+)*([a-zA-Z_$]\\w*)\\s*(?:<[^>]*>\\s*)?\\([^()]*\\)\\s*\\{', re.MULTILINE)\n \n # 语言关键字与系统字词，防方法声明误伤黑名单\n keywords = {\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'constructor', 'super', \n 'function', 'import', 'export', 'default', 'new', 'await', \n 'yield', 'typeof', 'instanceof', 'void', 'delete', 'in', 'do', 'class', 'get', 'set'\n }\n \n has_target_files = False\n funcs_found = []\n \n # 扫描生成的代码项目\n for root, dirs, files in os.walk(workspace_path):\n parts = root.split(os.sep)\n # 排除可能包含海量第三方无关代码的目录\n if 'node_modules' in parts or '.git' in parts or 'dist' in parts or 'build' in parts:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_target_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 搜集文件中所有定义的函数名\n funcs_found.extend(pattern_func_decl.findall(content))\n funcs_found.extend(pattern_var_decl.findall(content))\n funcs_found.extend(pattern_class_prop.findall(content))\n funcs_found.extend(pattern_method.findall(content))\n except Exception:\n continue\n \n # 没有可验证的前端代码文件时视作不满足\n if not has_target_files:\n return False\n \n valid_func_count = 0\n for name in funcs_found:\n if name in keywords:\n continue\n \n valid_func_count += 1\n \n # 检查收集到的函数名是否严格落实了 \"_noti_anim\" 后缀\n if not name.endswith('_noti_anim'):\n return False\n \n # 组件代码中如果没有任何提取到的函数，意味着核心业务逻辑压根未落实，同样返回 False\n if valid_func_count == 0:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"def check_class_name_length_is_non_prime(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 正则表达式：匹配 `class` 关键字后面的类名（兼容JavaScript/TypeScript/Python等常见类名定义）\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)')\n \n # 根据场景定义可能存在的代码文件扩展名\n valid_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.html', \n '.vue', '.svelte', '.py', '.java', '.cpp', '.hpp', '.h', '.cs'\n }\n\n # 遍历工作区目录下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n # 按照系统提示要求，生成的文件均使用 UTF-8 编码\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取出文件中所有的类名\n classes = class_pattern.findall(content)\n for cls_name in classes:\n # 如果类名的长度是质数，则违背了“长度恰好是一个非质数”的要求\n if is_prime(len(cls_name)):\n return False\n\n # 若所有类名长度均为非质数，或未找到任何类名，均算作满足要求\n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 判断数字是否为质数的辅助函数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n ** 0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 针对前端文件的简单变量声明正则表达式\n js_var_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n # 针对 Python 文件的普通赋值操作正则表达式（排除双等号 == 的情况）\n py_var_pattern = re.compile(r'\\b([a-zA-Z_][a-zA-Z0-9_]*)\\s*=[^=]')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n filepath = os.path.join(root, file)\n \n # 过滤常见可能包含变量的代码文件后缀名\n if ext in ['.js', '.ts', '.html', '.jsx', '.tsx', '.vue', '.py']:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n variables = []\n # 匹配前端代码或含 `<script>` 的 HTML 的变量\n if ext in ['.js', '.ts', '.html', '.jsx', '.tsx', '.vue']:\n variables.extend(js_var_pattern.findall(content))\n # 匹配 Python 代码的变量\n if ext == '.py':\n variables.extend(py_var_pattern.findall(content))\n \n # 判断所提取到的各个变量名的长度是否均为质数\n for var_name in variables:\n if not is_prime(len(var_name)):\n return False\n except Exception:\n # 避免个别文件由于非正常格式等因素干扰读取，直接跳过处理\n continue\n\n return True","tags":["Style","Naming"]}]},{"round_id":2,"instruction":"接下来丰富一下通知的样式。我希望支持四种状态：Success（绿色）、Error（红色）、Warning（黄色）、Info（蓝色）。请在页面中间增加几个对应的触发按钮方便我测试。每种状态的通知卡片要配上对应的左侧Icon图标，另外，在每张通知卡片的底部需要加一根极细的进度条，用来可视化那3秒的倒计时，进度条长度要在3秒内平滑地缩减到0。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_style 作为结尾；二是进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令中关于页面通知组件、DOM操作等内容，推断可能的代码文件后缀名\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试将二进制内容用 utf-8 解码，以验证文件是否采用 UTF-8 编码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦捕获到解码错误，说明存在非 UTF-8 编码的文件\n return False\n except Exception:\n # 遇到其他读取错误也认为未满足条件\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 目标注释字符串\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 针对前端项目推断的常见代码文件后缀\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果无法使用utf-8编码读取代码文件，视为失败\n return False\n \n # 检查文件头部（前10行）是否包含该注释\n found = False\n for line in lines[:10]:\n if target_comment in line:\n found = True\n break\n \n # 若某个代码文件开头未发现该注释，则不满足要求\n if not found:\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"import os\nfrom pathlib import Path\n\ndef check_directory_level(response: str, workspace_path: str) -> bool:\n # 保证 workspace_path 存在\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，避免如 .git 等环境辅助目录影响层级判断\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 获取当前目录相对于项目根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 忽略根目录本身\n if rel_path == '.':\n continue\n \n # 将相对路径分解成各级目录的元组\n parts = Path(rel_path).parts\n \n # 如果 parts 的长度等于 1，例如 parts == ('src',)，代表这是第2层目录\n # 如果 parts 的长度大于等于 2，例如 parts == ('src', 'components')，代表出现了第3层及以上的目录\n if len(parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_style 作为结尾","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方依赖等可能影响判断的文件夹\n if 'node_modules' in root.split(os.sep):\n continue\n \n for file in files:\n # 排除压缩文件\n if '.min.' in file:\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 去除单行和多行注释以避免误识别\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n \n func_names = []\n \n # 匹配方式1: 显式具名函数声明 -> function funcName(...)\n for m in re.finditer(r'\\bfunction\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\(', code):\n func_names.append(m.group(1))\n \n # 匹配方式2: 赋值形式的函数 (箭头函数表达式、普通函数表达式或对象属性等)\n pattern2 = r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*[:=]\\s*(?:async\\s+)?(?:function\\s*(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\s*\\(|(?:<[^>]*>\\s*)?\\([^)]*\\)\\s*=>|[a-zA-Z_$][0-9a-zA-Z_$]*\\s*=>)'\n for m in re.finditer(pattern2, code):\n func_names.append(m.group(1))\n \n # 匹配方式3: 对象简写方法或类成员方法 -> funcName(...) {\n pattern3 = r'^\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*(?:<[^>]*>\\s*)?\\([^)]*\\)\\s*\\{'\n for m in re.finditer(pattern3, code, re.MULTILINE):\n name = m.group(1)\n # 过滤常见的控制流关键字与保留字\n if name not in {'if', 'for', 'while', 'switch', 'catch', 'return', 'function', 'constructor', 'super', 'this', 'import', 'export'}:\n func_names.append(name)\n \n # 检查提取到的所有函数名是否都符合指定后缀\n for name in set(func_names):\n if not name.endswith('_noti_style'):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断一个自然数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 选取涉及前端及常见编程语言的代码文件\n valid_extensions = ('.js', '.jsx', '.ts', '.tsx', '.html', '.vue', '.py', '.java', '.cs', '.cpp')\n if not file.endswith(valid_extensions):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取面向对象编程中的类名（匹配如 \"class NotificationComponent\" 或 \"export class Noti\" 等格式）\n class_matches = re.findall(r'\\bclass\\s+([a-zA-Z_]\\w*)', content)\n \n for cls_name in class_matches:\n # 题目要求长度必须恰好是一个\"非质数\"\n # 如果是质数，则表示不满足用户要求，直接返回 False\n if is_prime(len(cls_name)):\n return False\n \n # 如果没找到类名，或者所有类名长度全为非质数，则满足要求\n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def check_variables_length_prime(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n def get_vars_from_js(js_code: str):\n vars_found = []\n # 清除单行和多行注释以避免干扰匹配\n js_code = re.sub(r'//.*', '', js_code)\n js_code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', js_code)\n \n # 1. 匹配常规的 var, let, const 声明\n for match in re.finditer(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', js_code):\n vars_found.append(match.group(1))\n \n # 2. 匹配简易的数组解构赋值，例如 const [a, b] = [1, 2]\n for match in re.finditer(r'\\b(?:var|let|const)\\s+\\[(.*?)\\]', js_code):\n inner = match.group(1)\n for item in inner.split(','):\n item = item.strip()\n if item:\n # 过滤掉默认值部分\n var_name = item.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n vars_found.append(var_name)\n \n # 3. 匹配简易的对象解构赋值，例如 const {a, b} = obj\n for match in re.finditer(r'\\b(?:var|let|const)\\s+\\{(.*?)\\}', js_code):\n inner = match.group(1)\n for item in inner.split(','):\n item = item.strip()\n if item:\n # 如果包含别名，例如 a: b，实际变量名为 b\n if ':' in item:\n item = item.split(':')[1].strip()\n # 过滤掉可能存在的默认值\n var_name = item.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n vars_found.append(var_name)\n \n return vars_found\n\n all_vars = set()\n found_any_code = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\n found_any_code = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 针对网页和组件文件，仅提取 script 标签中的代码部分\n if file.endswith(('.html', '.htm', '.vue')):\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for script_content in scripts:\n all_vars.update(get_vars_from_js(script_content))\n else:\n all_vars.update(get_vars_from_js(content))\n except Exception:\n continue\n\n # 如果没有找到任何可识别的代码文件，或者项目中无任何变量声明，则按无法通过验证处理\n if not found_any_code or not all_vars:\n return False\n \n # 逐一校验提取出的每个变量名长度是否均为质数\n for var in all_vars:\n if not is_prime(len(var)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"现在来处理多条通知并发的情况，这比较有挑战。如果我连续点击按钮多次，这些通知需要在右下角的容器里自下而上地堆叠排列。最关键的是，当一条新的通知插进来，或者中间某一条因为超时/被点击而消失时，剩下的通知必须要有一个平滑的上下位移过渡动画，让它们顺滑地挤占空缺位置，绝不能出现瞬间闪烁或硬切跳跃的情况。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_stack 作为结尾；二是确保最终生成的代码文件中完全避开for循环的使用。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目背景，定义前端常用的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取整个文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 出现 UnicodeDecodeError 表示文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略权限不足或其他非编码导致的读取错误\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常用代码文件扩展名白名单，排除 .log, .json 等非代码或不支持常规注释的文件\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.mjs', '.cjs'\n }\n \n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n found_comment = False\n \n # 检查每个代码文件开头的前10行（兼容文件首部可能有 DOCTYPE 声明或 shebang 等内容的情况）\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n if target_comment in line:\n found_comment = True\n break\n \n # 若发现有任何一个代码文件未包含指定的注释说明，则返回 False\n if not found_comment:\n return False\n except Exception:\n # 如果遇到文件异常无法打开读取的情况，视作验证失败\n return False\n \n # 如果项目中至少存在一个代码文件，且每个代码文件都满足了要求，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"import os\nfrom pathlib import Path\n\ndef check_directory_level(response: str, workspace_path: str) -> bool:\n workspace = Path(workspace_path).resolve()\n \n # 如果工作区不存在或不是一个目录，直接返回 False\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for root, dirs, files in os.walk(workspace):\n root_path = Path(root).resolve()\n \n try:\n rel_parts = root_path.relative_to(workspace).parts\n except ValueError:\n # 理论上 os.walk(workspace) 产出的 root 都会在 workspace 下，防御性捕获异常\n continue\n \n # 排除隐藏目录带来的干扰，比如 .git 等环境或工具生成的目录\n if any(part.startswith('.') for part in rel_parts):\n continue\n \n # 如果相对路径的层级达到 2 (即意味着存在 workspace/dir/subdir 结构)\n # 则说明目录层级达到了 3 层，超过了“最高不超过2层”的限制\n if len(rel_parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_stack 作为结尾","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 设定需要检查的代码文件扩展名，排除系统或配置的无关文件\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.py', '.vue', '.svelte'}\n exclude_dirs = {'node_modules', '.git', 'venv', 'env', '__pycache__', 'dist', 'build'}\n \n # 关键字列表，用于避免将JS控制流或者系统保留字误判为函数名\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'constructor', 'with', 'import', 'export', 'super'}\n \n # Python 的函数匹配正则\n pattern_py = r'\\bdef\\s+([a-zA-Z_]\\w*)\\s*\\('\n \n # JS/TS 传统 function 函数名匹配正则\n pattern_js_func = r'\\bfunction\\s+([a-zA-Z_$][\\w$]*)\\s*\\('\n \n # JS/TS 变量形式的函数赋值匹配正则，例如 const myFunc = () => 或者 let myFunc = function()\n pattern_js_arrow = r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)[^=;\\n]*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)'\n \n # JS/TS 对象属性形式的函数赋值正则，例如 myFunc: () => 或者 myFunc: function()\n pattern_obj_func = r'\\b([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)'\n \n # JS/TS 类的属性函数正则，例如 myFunc = () => \n pattern_class_prop_func = r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+)?([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)'\n \n # JS/TS 类/对象的方法简写形式匹配正则，例如 myFunc() { ... }\n pattern_method = r'^\\s*(?:async\\s+)?(?:(?:get|set)\\s+)?([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{'\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤排除的目录及隐藏目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 预处理：移除注释片段以防止在注释中的示例函数造成误判\n if ext == '.py':\n content = re.sub(r'#.*', '', content)\n content = re.sub(r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'', '', content)\n content = re.sub(r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"', '', content)\n else:\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n if ext in {'.html', '.vue', '.svelte'}:\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n \n # 按照语言对函数进行匹配校验\n if ext == '.py':\n for m in re.finditer(pattern_py, content):\n func_name = m.group(1)\n if not func_name.endswith('_noti_stack'):\n return False\n else:\n # JavaScript / TypeScript / Frontend 的各种变体匹配\n for m in re.finditer(pattern_js_func, content):\n func_name = m.group(1)\n if not func_name.endswith('_noti_stack'):\n return False\n \n for m in re.finditer(pattern_js_arrow, content):\n func_name = m.group(1)\n if not func_name.endswith('_noti_stack'):\n return False\n \n for m in re.finditer(pattern_obj_func, content):\n func_name = m.group(1)\n if not func_name.endswith('_noti_stack'):\n return False\n \n for m in re.finditer(pattern_class_prop_func, content, re.MULTILINE):\n func_name = m.group(1)\n if func_name not in keywords:\n if not func_name.endswith('_noti_stack'):\n return False\n \n for m in re.finditer(pattern_method, content, re.MULTILINE):\n func_name = m.group(1)\n if func_name not in keywords:\n if not func_name.endswith('_noti_stack'):\n return False\n except Exception:\n # 遇到无法解码的文件跳过即可\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 正则表达式匹配面向对象编程语言中的类名（适用于 JS, TS, Python 等）\n # 匹配模式：class关键字 + 至少一个空白字符 + 类名标识符\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)(?:\\s|\\(|{|:|$)')\n\n # 定义需要检查的常见代码文件扩展名\n target_extensions = ('.js', '.ts', '.jsx', '.tsx', '.py', '.html', '.htm', '.vue', '.svelte')\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖模块和版本控制目录，避免误判非AI生成的代码\n if 'node_modules' in root or '.git' in root or '__pycache__' in root:\n continue\n \n for file in files:\n if file.endswith(target_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找当前文件内所有声明的类名\n classes = class_pattern.findall(content)\n \n for cls_name in classes:\n cls_len = len(cls_name)\n # 如果发现任一类名长度是质数，则违背了“长度恰好是非质数”的要求\n if is_prime(cls_len):\n return False\n \n except Exception:\n # 忽略个别无法读取或编码异常的文件\n continue\n\n # 若所有类名长度均为非质数，或项目中未显式定义OOP类(不违规)，则判定为满足要求\n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def verify_variable_name_lengths_are_prime(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 质数判断函数\n def is_prime(n):\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 收集所有的前端代码文件\n js_files = []\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.vue')):\n js_files.append(os.path.join(root, file))\n\n if not js_files:\n return False\n \n var_names = set()\n for file in js_files:\n with open(file, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 若是 HTML 或 Vue 文件，重点提取 <script> 标签内代码\n if file.endswith('.html') or file.endswith('.vue'):\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n content = '\\n'.join(scripts)\n \n # 过滤掉多行和单行注释，防止非代码文本干扰提取\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n \n # 1. 匹配简单的单变量声明 (例: let abc = 1; 或 const def = 2)\n matches1 = re.findall(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n var_names.update(matches1)\n \n # 2. 匹配对象解构赋值变量 (例: const { a, b: c } = obj)\n matches2 = re.findall(r'\\b(?:let|const|var)\\s*\\{([^}]+)\\}', content)\n for m in matches2:\n parts = m.split(',')\n for p in parts:\n p = p.strip()\n if p:\n if ':' in p:\n name = p.split(':')[1].strip()\n else:\n name = p\n # 滤除可能的默认值\n name = name.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', name):\n var_names.add(name)\n \n # 3. 匹配数组解构赋值变量 (例: let [x, , y=2] = arr)\n matches3 = re.findall(r'\\b(?:let|const|var)\\s*\\[([^\\]]+)\\]', content)\n for m in matches3:\n parts = m.split(',')\n for p in parts:\n p = p.strip()\n if p:\n name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', name):\n var_names.add(name)\n \n # 4. 匹配传统函数的形参 (例: function foo(arg1, arg2))\n funcs = re.findall(r'function\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', content)\n for f in funcs:\n args = f.split(',')\n for arg in args:\n arg = arg.strip().split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', arg):\n var_names.add(arg)\n \n # 5. 匹配多参数箭头函数的形参 (例: (arg1, arg2) => )\n arrow_funcs = re.findall(r'\\(([^)]*)\\)\\s*=>', content)\n for f in arrow_funcs:\n args = f.split(',')\n for arg in args:\n arg = arg.strip().split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', arg):\n var_names.add(arg)\n \n # 6. 匹配单参数箭头函数的形参 (例: arg => )\n arrow_funcs2 = re.findall(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', content)\n var_names.update(arrow_funcs2)\n\n # 只要提取不到任何变量，说明生成的内容并非典型的规范 JS 代码\n if not var_names:\n return False\n \n # 判断每一个提取到的变量名的长度是否均为质数\n for var in var_names:\n if not is_prime(len(var)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保最终生成的代码文件中完全避开for循环的使用","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能包含前端项目或者Python代码的文件后缀\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', \n '.vue', '.svelte', \n '.html', '.htm', \n '.css', '.scss', '.sass',\n '.py'\n }\n \n def remove_comments_and_strings(code: str, ext: str) -> str:\n # 正则表达式用于移除各类语言的字符串和注释\n patterns = [\n r\"'(?:\\\\.|[^'\\\\])*'\", # 单引号字符串\n r'\"(?:\\\\.|[^\"\\\\])*\"', # 双引号字符串\n r\"`(?:\\\\.|[^`\\\\])*`\", # 反引号模板字符串\n r\"/\\*.*?\\*/\", # 多行注释 /* ... */\n r\"//.*?$\", # 单行注释 // ...\n r\"<!--.*?-->\" # HTML 注释 <!-- ... -->\n ]\n \n # 针对Python特有的注释和多行字符串形式\n if ext == '.py':\n patterns.extend([\n r\"\\#.*?$\", # Python 单行注释\n r'\"\"\".*?\"\"\"', # Python 双引号多行字符串\n r\"'''.*?'''\" # Python 单引号多行字符串\n ])\n \n regex = \"|\".join(patterns)\n pattern = re.compile(regex, re.DOTALL | re.MULTILINE)\n return pattern.sub('', code)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除注释和字符串，以便进行更安全的关键字匹配\n clean_content = remove_comments_and_strings(content, ext)\n \n # 1. 检测 JS/TS/Svelte/HTML 原生 for 循环，包括 for(...), for...in, for...of, for await(...)\n # 使用 (?<!\\.) 排除可能存在的方法调用，如 arr.for()\n if re.search(r'(?<!\\.)\\bfor\\s*(?:await\\s*)?\\(', clean_content):\n return False\n \n # 2. 检测 Vue 或 Angular 的模板内置 for 循环指令\n if re.search(r'\\bv-for\\b|\\*ngFor\\b', clean_content):\n return False\n \n # 3. 检测 Sass/SCSS 的 @for 循环\n if re.search(r'@for\\b', clean_content):\n return False\n \n # 4. 检测 Python 原生 for 循环\n if ext == '.py' and re.search(r'\\bfor\\s+.*?\\s+in\\b', clean_content):\n return False\n \n except Exception:\n # 如果由于编码或其他问题读取失败，跳过该文件继续处理其他文件\n continue\n\n # 如果所有代码文件都没检测出 for 循环相关语法，则验证通过\n return True","tags":["Style","Paradigm"]}]},{"round_id":4,"instruction":"加一点交互细节，这个对用户体验很关键。现在3秒自动关闭有时来不及看清长文本内容，我希望当我的鼠标悬浮（Hover）在某条通知卡片上时，该通知的自动销毁倒计时要立刻暂停，底部的进度条动画也要随之停住；当我鼠标移开后，倒计时和进度条再接着从刚才暂停的地方继续往下走，直到消失。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_hover 作为结尾；二是你构建的代码文件中必须排除对while循环的使用。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\n # 定义常见的前端与常规代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.vue', '.json', '.py', \n '.java', '.c', '.cpp', '.h', '.xml', '.txt'\n }\n \n # 如果工作区目录不存在，可以直接视为失败\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅验证代码文件，忽略日志、备份（如.bak）或二进制文件等\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制形式读取文件内容\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试强制用UTF-8进行解码\n raw_data.decode('utf-8')\n \n except UnicodeDecodeError:\n # 只要有任何一个代码文件无法使用UTF-8解码，便不满足要求\n return False\n except Exception:\n # 遇到诸如权限等无法读取的异常情况，也视为无法满足验证\n return False\n \n # 如果检查了所有代码文件都没发生 UnicodeDecodeError 异常，则说明全部满足UTF-8编码要求\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 目标要求包含的具体字符串\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 结合用户指令推断，包含常见的前端代码扩展名，加入部分通用后台扩展名增强兼容性\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.h'}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = []\n # 仅读取前10行来验证是否在文件的“开头处”\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n lines.append(line)\n \n header_content = \"\".join(lines)\n \n # 如果没有在该代码文件的开头部分找到指定的注释文本，则不满足要求\n if target_str not in header_content:\n return False\n except Exception:\n # 对于非纯文本或编码非UTF-8等异常无法读取的文件，跳过处理\n continue\n \n # 如果项目下完全没有找到任何代码文件，判定未完成生成代码的任务\n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"import os\nfrom pathlib import Path\n\ndef verify(response: str, workspace_path: str) -> bool:\n try:\n workspace = Path(workspace_path).resolve()\n \n # 确保工作区路径存在且是一个目录\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for root, dirs, files in os.walk(workspace):\n # 原地修改 dirs 列表以过滤掉以 '.' 开头的隐藏目录（如 .git, .vscode 等）\n # 这样 os.walk 就不会继续向下遍历这些隐藏目录\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 获取当前遍历到的目录相对于根目录的相对路径\n rel_path = Path(root).relative_to(workspace)\n \n # rel_path.parts 会将相对路径按层级拆分为元组\n # - 对于 workspace 本身，parts 为空元组 ()，长度 0 （属于第1层）\n # - 对于 workspace/dir1，parts 为 ('dir1',)，长度 1 （属于第2层）\n # - 对于 workspace/dir1/dir2，parts 为 ('dir1', 'dir2')，长度 2 （属于第3层）\n # 要求最高不超过2层，因此长度达到2时即视为违规\n if len(rel_path.parts) >= 2:\n return False\n \n return True\n except Exception:\n return False","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_hover 作为结尾","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断所有生成的代码文件中的函数是否均以 _noti_hover 结尾。\n \"\"\"\n # 针对 JavaScript/TypeScript 常见函数定义写出匹配正则\n js_patterns = [\n # 1. 常规函数声明 function foo(...)\n re.compile(r'\\bfunction\\s*(?:\\*\\s*)?([a-zA-Z_$][\\w$]*)\\s*\\('),\n # 2. 匿名函数赋值 const/let/var foo = function(...)\n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?function\\b'),\n # 3. 箭头函数赋值 const/let/var foo = (...) => ... 或者 foo = e => ...\n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][\\w$]*)\\s*=>'),\n # 4. 对象方法 { foo: function(...) } 或 { foo: (...) => ... }\n re.compile(r'([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|(?:\\([^)]*\\)|[a-zA-Z_$][\\w$]*)\\s*=>)'),\n # 5. Class中的类方法声明（多行模式匹配开头）\n re.compile(r'^\\s*(?:public\\s+|private\\s+|protected\\s+)?(?:abstract\\s+)?(?:async\\s+)?(?:static\\s+)?(?:get\\s+|set\\s+)?([a-zA-Z_$][\\w$]*)\\s*(?:<[^>]+>)?\\s*\\([^)]*\\)\\s*(?::\\s*[^{]+)?\\{', re.MULTILINE)\n ]\n # 针对 Python 函数定义的正则\n py_pattern = re.compile(r'\\bdef\\s+([a-zA-Z_]\\w*)\\s*\\(')\n\n # 排除容易在正则匹配（尤其是第5个正则）中作为代码块或者关键字被误抓取的保留字\n ignore_names = {\n 'if', 'for', 'while', 'switch', 'catch', 'return', 'constructor', 'render', \n 'super', 'get', 'set', 'try', 'finally', 'else', 'do', 'with', 'typeof', \n 'instanceof', 'new', 'delete', 'await', 'yield', 'import', 'export', 'function'\n }\n \n found_any_func = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue', '.svelte', '.py']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n code_blocks = []\n # 如果是包含标签的前端页面或组件，抽离出script块中的代码再校验\n if ext in ['.html', '.htm', '.vue', '.svelte']:\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n code_blocks.extend(scripts)\n else:\n code_blocks.append(content)\n \n for code in code_blocks:\n # 移除单行及多行注释，避免注释内的伪函数或旧代码干扰\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n funcs = []\n if ext == '.py':\n funcs.extend(py_pattern.findall(code))\n else:\n for p in js_patterns:\n funcs.extend(p.findall(code))\n \n for func in funcs:\n func = func.strip()\n # 忽略无效函数名或系统保留字\n if not func or func in ignore_names:\n continue\n \n found_any_func = True\n # 只要存在不以 _noti_hover 结尾的函数，则直接认为不符合规范\n if not func.endswith('_noti_hover'):\n return False\n\n # 若搜寻到了至少一个函数，并且没有提前触发False抛出，说明满足题意；否则如果连函数都没有，也视为违反业务逻辑返回False\n return True if found_any_func else False","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助判断素数，用于判断非质数（素数返回 True，非素数/非质数返回 False）\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 匹配面向对象中的类名，如 class NotificationManager\n oop_class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\b')\n # 匹配 HTML / JSX 中的 CSS 类名，如 class=\"toast\" 或 className='msg-card'\n html_class_pattern = re.compile(r'\\b(?:class|className)\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']')\n\n # 根据项目背景推断常见的前端代码文件类型\n valid_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n\n # 验证 OOP 类名\n for match in oop_class_pattern.findall(content):\n # 依照指令限制，因未明确说明\"字符数\"，必须使用 count_word 获取计数值\n if is_prime(count_word(match)):\n return False\n\n # 验证 HTML/JSX 中的 CSS 类名\n for match in html_class_pattern.findall(content):\n # CSS类名可能包含多个并用空格分隔\n class_names = match.strip().split()\n for c_name in class_names:\n # 同样依照约束，必须使用 count_word 而不是 len\n if is_prime(count_word(c_name)):\n return False\n except Exception:\n # 忽略读取或编码过程中的异常文件\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def check_variable_lengths_are_prime(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n def extract_js_variables(code_text: str) -> set:\n # 去除字符串字面量（避免字符串内部文本带来干扰）\n code_text = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '\"\"', code_text)\n code_text = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", \"''\", code_text)\n code_text = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '``', code_text, flags=re.DOTALL)\n # 去除注释\n code_text = re.sub(r'//.*', '', code_text)\n code_text = re.sub(r'/\\*.*?\\*/', '', code_text, flags=re.DOTALL)\n \n variables = []\n # 提取 let / const / var 声明的直接标识符\n pattern = r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b'\n variables.extend(re.findall(pattern, code_text))\n \n # 提取常规 function 后面的参数\n func_pattern = r'\\bfunction\\s+[a-zA-Z_$0-9]*\\s*\\((.*?)\\)'\n for params_str in re.findall(func_pattern, code_text):\n params = [p.strip() for p in params_str.split(',') if p.strip()]\n for p in params:\n p = p.split('=')[0].strip() # 处理可能存在的默认参数\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p):\n variables.append(p)\n \n # 提取箭头函数的单参数模式 (例如: arg => ...)\n arrow_pattern_1 = r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>'\n variables.extend(re.findall(arrow_pattern_1, code_text))\n \n # 提取箭头函数的多参数模式 (例如: (arg1, arg2) => ...)\n arrow_pattern_2 = r'\\((.*?)\\)\\s*=>'\n for params_str in re.findall(arrow_pattern_2, code_text):\n params = [p.strip() for p in params_str.split(',') if p.strip()]\n for p in params:\n p = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p):\n variables.append(p)\n \n return set(variables)\n\n def extract_py_variables(code_text: str) -> set:\n variables = set()\n try:\n tree = ast.parse(code_text)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n variables.add(node.id)\n elif isinstance(node, ast.arg):\n variables.add(node.arg)\n except Exception:\n pass\n return variables\n\n if not os.path.exists(workspace_path):\n return False\n\n has_variables = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n vars_found = set()\n if file.endswith('.py'):\n vars_found = extract_py_variables(content)\n elif file.endswith(('.html', '.htm')):\n # 对于 HTML 文件，提取其中所有的 <script> 标签内容\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n vars_found.update(extract_js_variables(script))\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n vars_found = extract_js_variables(content)\n \n if vars_found:\n has_variables = True\n for var_name in vars_found:\n # 检查任何一个提取到的变量名，其长度是否为质数\n if not is_prime(len(var_name)):\n return False\n except Exception:\n continue\n\n # 只有当成功匹配到至少一个变量且所有变量长度均为质数时，才认定为符合条件\n return has_variables","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保最终生成的代码文件中完全避开for循环的使用","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查生成的代码项目中是否完全避开了 for 循环的使用。\n 返回 True 表示满足要求（没有 for 循环），False 表示存在 for 循环。\n \"\"\"\n if not os.path.exists(workspace_path):\n return False\n \n # 定义正则用于去除字符串常量，以防止字符串中的 \"for\" 干扰\n string_pattern1 = r'\"(?:\\\\.|[^\"\\\\])*\"'\n string_pattern2 = r\"'(?:\\\\.|[^'\\\\])*'\"\n string_pattern3 = r\"`(?:\\\\.|[^`\\\\])*`\"\n \n # 定义正则用于去除各类注释\n html_comment = r\"<!--[\\s\\S]*?-->\"\n comment_pattern_multi = r\"/\\*[\\s\\S]*?\\*/\"\n comment_pattern_single = r\"//.*\"\n py_comment_single = r\"#.*\"\n \n def clean_code(code: str) -> str:\n # 先去除所有的字符串内容\n code = re.sub(string_pattern1, \"\", code)\n code = re.sub(string_pattern2, \"\", code)\n code = re.sub(string_pattern3, \"\", code)\n # 再去除所有的注释\n code = re.sub(html_comment, \"\", code)\n code = re.sub(comment_pattern_multi, \"\", code)\n code = re.sub(comment_pattern_single, \"\", code)\n code = re.sub(py_comment_single, \"\", code)\n return code\n\n # 前端及常见的代码文件后缀\n code_extensions = (\n '.js', '.ts', '.jsx', '.tsx', \n '.html', '.htm', '.vue', '.svelte', \n '.py', '.css', '.scss', '.less'\n )\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n if not file.lower().endswith(code_extensions):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 文件无法读取则跳过\n continue\n \n # 清理代码内容中的注释和字符串\n cleaned = clean_code(content)\n \n # 1. 匹配常规 C-like 和 JS/TS 的 for 循环 (for, for...in, for...of, for await...of)\n if re.search(r'\\bfor\\s*(?:await\\s*)?\\(', cleaned):\n return False\n \n # 2. 匹配 Python 中的 for 循环 (for...in)\n if file.endswith('.py') and re.search(r'\\bfor\\b\\s+.*?\\s+\\bin\\b', cleaned):\n return False\n \n # 3. 匹配 Vue 和 Angular 等模板中的循环指令\n if re.search(r'\\bv-for\\b|\\*ngFor\\b', cleaned):\n return False\n \n # 4. 匹配 Jinja/Django 等模板引擎的 for 循环\n if re.search(r'\\{%\\s*for\\b', cleaned):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"你构建的代码文件中必须排除对while循环的使用","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义可能包含逻辑循环的代码文件扩展名\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 针对HTML或Vue文件，为防止普通文本或DOM中恰好出现 \"while\"，仅截取 script 标签里的逻辑代码进行检查\n if ext in {'.html', '.vue'}:\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n content = '\\n'.join(scripts)\n \n # 正则匹配规则：去除多行注释、双斜杠单行注释、井号单行注释，以及反引号、单引号、双引号包裹的字符串（支持内部转义）\n pattern = re.compile(\n r'/\\*.*?\\*/|//.*?$|#.*?$|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"|`(?:\\\\.|[^\\\\`])*`',\n re.DOTALL | re.MULTILINE\n )\n cleaned_content = re.sub(pattern, ' ', content)\n \n # 搜索是否存在独立关键字 while\n if re.search(r'\\bwhile\\b', cleaned_content):\n return False\n except Exception:\n # 若文件读取或解析存在异常，跳过继续验证其他文件\n continue\n \n return True","tags":["Style","Paradigm"]}]},{"round_id":5,"instruction":"好，现在我们引入核心的'优先级'控制机制。请在页面上的触发区域加一个优先级下拉框选项（低、中、高）。同级别的通知按先后顺序正常排队。但如果我触发了一个'高优先级'的通知，它绝不能排在队列末尾，而是必须立刻强行插队到当前可见通知列表的最顶部（最新鲜、最显眼的位置），并且高优先级的通知进场时，需要带有一个引人注目的左右'震动（Shake）'特效。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_priority 作为结尾；二是对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 针对前端页面通知组件任务，枚举可能的代码文件后缀名\n code_extensions = {\n '.html', '.htm', \n '.css', '.scss', '.less', \n '.js', '.ts', \n '.jsx', '.tsx', \n '.vue', '.svelte', \n '.json'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对推断为代码类型的文件进行 UTF-8 编码格式检查\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n # 使用二进制模式读取文件原始字节\n with open(filepath, 'rb') as f:\n raw_data = f.read()\n # 测试是否可以合法被 utf-8 解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码出现错误，说明文件非有效 UTF-8 编码\n return False\n except Exception:\n # 若因无权限等原因无法读取，默认视作验证失败\n return False\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 针对前端组件常见代码文件扩展名进行推断与限制，排除日志等非代码文件\n code_extensions = {\".html\", \".css\", \".js\", \".jsx\", \".ts\", \".tsx\", \".vue\"}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取前 10 行以检查文件开头处是否存在目标注释\n lines = [f.readline() for _ in range(10)]\n \n # 检查是否有任意一行包含目标注释文本\n found = any(target_comment in line for line in lines)\n \n if not found:\n return False\n except Exception:\n # 如果读取遇到异常（比如编码错误或无法访问），则视为无法验证该文件，返回 False\n return False\n \n # 如果找到了至少一个代码文件，并且所有代码文件都包含该注释，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"def check_directory_levels(response: str, workspace_path: str) -> bool:\n import os\n import pathlib\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，防止评测系统环境预置的目录（如 .git 等）影响层级判断\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n if rel_path == '.':\n depth = 0\n else:\n # 获取相对路径包含的层级数\n # 比如 \"src\" 的 parts 是 ('src',), 长度 1 (代表第 2 层目录)\n # 比如 \"src/utils\" 的 parts 是 ('src', 'utils'), 长度 2 (代表第 3 层目录)\n depth = len(pathlib.Path(rel_path).parts)\n \n # 相对深度为 2 意味着已经到达了第 3 层目录，不符合最高 2 层要求\n if depth >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_priority 作为结尾","validation_code":"import os\nimport re\n\ndef check_function_names(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n patterns_regex = [\n # 1. 匹配标准函数声明：function funcName(...) 或 function *funcName(...)\n r'\\bfunction\\s*(?:\\*\\s*)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:<[^>]*>)?\\s*\\(',\n \n # 2. 匹配 Python 函数声明：def funcName(...)\n r'\\bdef\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(',\n \n # 3. 匹配箭头函数或匿名函数的变量赋值：const/let/var funcName = (...) => { ... } \n r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:(?:<[^>]*>)?\\s*function\\b|(?:<[^>]*>)?\\s*\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)',\n \n # 4. 匹配类方法或对象方法：methodName(args) { ... }\n # 排除各类控制流和语言保留关键字，防止将 if(...){, for(...){ 等错误当作函数捕获\n r'\\b(?!(?:if|catch|for|while|switch|return|function|def|async|await|class|interface|type|new|super|this|import|export|with|typeof|instanceof|delete|void|yield)\\b)([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:<[^>]*>)?\\s*\\([^()]*\\)\\s*(?::[^{=;]*)?\\{'\n ]\n \n found_functions = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n is_python = file.endswith('.py')\n # 常见前端代码文件及 Python\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 预处理清洗内容：移除字符串字面量、各类型注释及 <style> 块，严防正则陷入死角或误匹配\n if is_python:\n sanitization_patterns = [\n r'#[^\\n]*', # Python 单行注释\n r'(?s:\"\"\".*?\"\"\")', # Python 三引号注释/字符串\n r\"(?s:'''.*?''')\", # Python 三单引号注释/字符串\n r'(?s:\"(?:\\\\.|[^\"\\\\])*\")', # 双引号字符串\n r\"(?s:'(?:\\\\.|[^'\\\\])*')\" # 单引号字符串\n ]\n else:\n sanitization_patterns = [\n r'(?i)(?s:<style[^>]*>.*?</style>)', # HTML/Vue style 块（CSS）\n r'(?s:<!--.*?-->)', # HTML 注释\n r'(?s:/\\*.*?\\*/)', # JS/TS 块级多行注释\n r'//[^\\n]*', # JS/TS 单行注释\n r'(?s:\"(?:\\\\.|[^\"\\\\])*\")', # 双引号字符串\n r\"(?s:'(?:\\\\.|[^'\\\\])*')\", # 单引号字符串\n r'(?s:`(?:\\\\.|[^`\\\\])*`)' # 模板字符串\n ]\n \n combined_pattern = '|'.join(sanitization_patterns)\n \n def replacer(match):\n s = match.group(0)\n # 若匹配结果为字符串字面量，保留空引号避免代码结构黏连（如导致连续词法分析错误）\n if s.startswith(('\"', \"'\", '`')):\n if s.startswith('\"\"\"'): return '\"\"'\n if s.startswith(\"'''\"): return \"''\"\n return s[0] + s[0]\n # 若为注释或样式块，则一律由空格取代\n else:\n return \" \"\n \n content = re.sub(combined_pattern, replacer, content)\n \n # 开始使用预定义正则全面提取函数名称\n for pattern in patterns_regex:\n for match in re.finditer(pattern, content):\n func_name = match.group(1)\n # 双重保险：过滤可能遗漏的硬性关键字\n if func_name in ['if', 'catch', 'for', 'while', 'switch', 'return', 'function', 'def']:\n continue\n \n found_functions = True\n if not func_name.endswith('_noti_priority'):\n return False\n\n return found_functions","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断一个自然数是否为质数\n def is_prime(n):\n if n <= 1:\n return False\n if n == 2 or n == 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n if not os.path.exists(workspace_path):\n return False\n \n # 匹配各类主流编程语言中的类声明关键字及类名\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z0-9_$]+)')\n \n # 定义可能包含类定义的源代码文件扩展名\n valid_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \n '.vue', '.py', '.java', '.cs', '.cpp', '.hpp', '.h'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件读取失败则跳过该文件继续验证\n continue\n \n # 粗略去除代码注释，防止注释中的示例代码干扰解析\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n \n # 寻找文件中的所有类名\n classes = class_pattern.findall(content)\n for class_name in classes:\n # 如果任一类名的长度是质数，则违背了“必须是一个非质数”的要求\n if is_prime(len(class_name)):\n return False\n \n # 所有文件验证完毕，均未发现长度为质数的类名，符合要求\n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def verify_variables_length_prime(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 内部辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n if n in (2, 3):\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 正则表达式：用于匹配 JavaScript / TypeScript / HTML 等前端文件中 let, const, var 声明的变量\n # 仅匹配单个基础标识符，避免复杂的解构赋值引发正则误判\n js_var_pattern = re.compile(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n # 正则表达式：用于匹配 Python 代码中基础的行首变量赋值\n py_var_pattern = re.compile(r'^[ \\t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \\t]*=[ \\t]*[^=]', re.MULTILINE)\n\n # 需要扫描的代码文件扩展名\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\n\n # 遍历代码项目目录\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n\n variables = []\n \n # 针对不同语言环境提取变量名\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\n variables.extend(js_var_pattern.findall(content))\n if ext == '.py':\n variables.extend(py_var_pattern.findall(content))\n\n # 验证每一个提取出的变量名长度是否均为质数\n for var_name in variables:\n if not is_prime(len(var_name)):\n return False\n \n except Exception:\n # 忽略无法正常以utf-8读取的文件\n continue\n\n return True","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保最终生成的代码文件中完全避开for循环的使用","validation_code":"def check_for_loop_absence(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 预编译针对不同for循环的正则表达式\n # 匹配 JavaScript / TypeScript 及类似C的for循环: for(...) 或 for await(...)\n js_for_regex = re.compile(r'\\bfor\\s*(?:await\\s*)?\\(')\n # 匹配 Python 的 for 循环: for x in obj:\n py_for_regex = re.compile(r'\\bfor\\s+[a-zA-Z0-9_,\\s\\(\\)]+\\s+in\\b')\n # 匹配 Vue 的循环语法: v-for=\n vue_for_regex = re.compile(r'\\bv-for\\s*=')\n # 匹配 Angular 的循环语法: *ngFor=\n ng_for_regex = re.compile(r'\\*ngFor\\s*=')\n # 匹配 传统模版引擎(如Jinja2/Django/Liquid等)的循环语法: {% for %}\n tpl_for_regex = re.compile(r'\\{%\\s*for\\b')\n\n # 可能包含for循环的前端与脚本代码文件后缀\n valid_extensions = {\n \".js\", \".ts\", \".jsx\", \".tsx\", \".mjs\", \".cjs\", \n \".vue\", \".html\", \".htm\", \".svelte\", \".py\"\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n\n # 移除代码内容中的注释，以防止注释中的 for 词汇产生误伤\n # 1. 移除 JS/TS/Py 的单行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n # 2. 移除 JS/TS 的多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 3. 移除 HTML 的注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n\n # 检查是否包含任何一种for循环\n if (js_for_regex.search(content) or\n py_for_regex.search(content) or\n vue_for_regex.search(content) or\n ng_for_regex.search(content) or\n tpl_for_regex.search(content)):\n return False # 发现了 for 循环，验证失败\n\n except Exception:\n # 如果某些文件读取出错（如权限问题或编码非UTF-8），则直接跳过该文件继续验证\n continue\n\n # 如果遍历检查完所有符合条件的文件都没有发现for循环，说明已完全避开for循环\n return True","tags":["Style","Paradigm"]},{"constraint":"你构建的代码文件中必须排除对while循环的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能包含前端代码的文件类型\n valid_extensions = {'.js', '.ts', '.html', '.jsx', '.tsx', '.vue', '.css', '.svelte'}\n \n # 构造正则表达式以匹配字符串和注释\n # string 分组：正确处理双引号、单引号、反引号构成的字符串及内部可能的转义字符\n # comment 分组：匹配 JS 的单行注释、多行注释，以及 HTML 的注释\n pattern_str = (\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|'\n r\"'(?:\\\\.|[^'\\\\])*'|\"\n r'`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<comment>//[^\\n]*|/\\*.*?\\*/|<!--.*?-->)'\n )\n pattern = re.compile(pattern_str, re.DOTALL)\n\n # 替换逻辑：字符串被替换为空字符串字面量 \"\" 避免意外破坏后续语法结构，注释则直接被清空\n def replacer(match):\n if match.group('string'):\n return '\"\"'\n else:\n return ''\n\n # 遍历项目目录下的所有相关代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 过滤掉内容中的字符串和注释文本，避免内容中的 \"while\" 文本引发误判\n clean_content = pattern.sub(replacer, content)\n \n # 检查干净的代码中是否包含 while 循环语法特征\n # 在 JS/TS 等前端语言中，while 循环一定符合 'while' 后紧跟可选空格和左括号 '(' 的格式\n if re.search(r'\\bwhile\\s*\\(', clean_content):\n return False\n except Exception:\n # 如果遇到文件无法读取的情况，跳过并继续处理其他文件\n pass\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符","validation_code":"def check_code_line_width(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，当前任务属于Web前端相关的DOM/组件操作\n # 因此设定常见的Web前端代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）和常见的依赖包目录（如 node_modules）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行末的换行符，并统计剩余字符的长度\n if len(line.rstrip('\\r\\n')) > 80:\n return False\n except Exception:\n # 遇到读取异常(例如非文本文件等)，直接跳过继续检查\n continue\n \n # 如果项目中存在代码文件并且所有代码行的长度均未超过 80 个字符，则满足要求\n return has_code_file","tags":["Quantity","Range"]}]},{"round_id":6,"instruction":"现在的堆叠如果发了十几条屏幕就占满了。我们来做一个'堆叠折叠'模式。规定当屏幕上的通知超过3条时，第4条及以后的旧通知不再向下铺开，而是向后退缩，折叠成一叠'纸牌'的效果（后面的卡片层级降低，逐渐缩小、变暗、并被前面的部分遮挡）。同时在容器最下方显示一个'展开全部（共X条）'的文字按钮，点击后这些纸牌动画展开成普通的长列表，再点一次则重新折叠。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_fold 作为结尾；二是请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.vue', '.json'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名并统一转换为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 仅验证代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦触发 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如文件权限等其他非编码相关的异常\n continue\n\n # 所有代码文件都能成功以 UTF-8 解码，或未发现代码文件\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据用户构建页面通知组件的要求，推断生成的前端代码文件可能的扩展名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \n '.css', '.scss', '.less', '.vue', '.svelte'\n }\n \n found_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和前端常见的第三方依赖/构建目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'dist', 'build')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n has_comment = False\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 检查文件开头部分（前20行），确认是否加上了要求的注释\n for _ in range(20):\n line = f.readline()\n if not line:\n break\n if target_comment in line:\n has_comment = True\n break\n except Exception:\n # 若文件读取失败或编码非UTF-8，判定为未满足要求\n return False\n \n # 如果任何一个代码文件开头没有该注释，则不满足要求\n if not has_comment:\n return False\n\n return found_code_file","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"def verify_directory_level(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n \n workspace = Path(workspace_path).resolve()\n \n # 若提供的工作区路径不存在或不是目录，则无法满足要求\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n # 递归遍历工作区内的所有文件和文件夹\n for current_path in workspace.rglob('*'):\n # 仅针对目录进行层级判断\n if current_path.is_dir():\n # 获取该目录相对于项目根目录（workspace）的相对路径\n rel_parts = current_path.relative_to(workspace).parts\n \n # 若相对路径的分段数量 >= 2，说明出现了第3层及以上的目录\n # (例如：workspace/dir1/dir2 -> rel_parts = ('dir1', 'dir2')，长度为 2)\n if len(rel_parts) >= 2:\n return False\n \n # 遍历结束均未发现层级越界情况，满足要求\n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_fold 作为结尾","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 提取 Python 函数定义的正则\n py_func_pattern = re.compile(r'^\\s*def\\s+([a-zA-Z_]\\w*)\\s*\\(', re.MULTILINE)\n \n # 提取 JS/TS 普通函数声明的正则\n js_func_pattern = re.compile(r'\\bfunction\\s+([a-zA-Z_]\\w*)\\s*\\(')\n \n # 提取 JS/TS 赋值给变量的箭头函数或匿名函数的正则\n js_var_func_pattern = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_]\\w*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_]\\w*\\s*=>)')\n \n # 提取 JS/TS 类方法或对象方法的正则（带有开括号与花括号特征）\n js_method_pattern = re.compile(r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+|async\\s+)*([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*(?::\\s*[^{]+)?\\s*\\{', re.MULTILINE)\n js_keywords = {'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', 'super', 'function', 'else'}\n\n # 排除不需要检查的系统库/第三方库目录\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 如果当前路径属于被忽略的目录结构中，则跳过\n if any(d in root.split(os.sep) for d in ignore_dirs):\n continue\n \n for file in files:\n # 仅处理常见的前后端代码文件\n if not file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.py')):\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n functions = []\n \n # 区分 Python 和 前端JS类文件\n if file.endswith('.py'):\n functions.extend(py_func_pattern.findall(content))\n else:\n text_to_search = content\n # 若为 HTML / Vue，则提取 script 标签内容做匹配\n if file.endswith(('.html', '.vue')):\n script_pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n scripts = script_pattern.findall(content)\n text_to_search = '\\n'.join(scripts)\n \n # 匹配多种 JS/TS 函数形式\n functions.extend(js_func_pattern.findall(text_to_search))\n functions.extend(js_var_func_pattern.findall(text_to_search))\n \n # 匹配方法，并排除干扰关键字\n methods = js_method_pattern.findall(text_to_search)\n for m in methods:\n if m not in js_keywords and not m.startswith('if') and not m.startswith('for') and not m.startswith('while') and not m.startswith('switch'):\n functions.append(m)\n \n # 对所有搜集到的函数名进行验证\n for func_name in functions:\n if not func_name.endswith('_noti_fold'):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个自然数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n if not os.path.isdir(workspace_path):\n return True\n\n # 正则表达式：匹配 `class` 关键字后紧跟的类名标识符\n # 适用于绝大多数主流语言 (JS, TS, Python, Java, C++, PHP 等)\n pattern = re.compile(r'\\bclass\\s+([a-zA-Z_][a-zA-Z0-9_]*)')\n \n # 前端开发及常见后端开发的代码文件后缀\n valid_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.py', '.java', \n '.cpp', '.h', '.hpp', '.cs', '.php', '.html', '.vue', '.svelte'\n }\n\n # 遍历代码项目\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略读取失败的文件，或非UTF-8编码的文件\n continue\n \n # 提取文件中所有的类名\n class_names = pattern.findall(content)\n for cls_name in class_names:\n cls_len = len(cls_name)\n # 如果某个类名的长度刚好是质数，则判定为不满足要求\n if is_prime(cls_len):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个数字是否为质数\n def is_prime(n):\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n all_vars = set()\n # 过滤掉合法的 JS/TS 关键字，以防匹配误伤\n keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'break', 'continue', 'return', \n 'function', 'class', 'extends', 'super', 'new', 'this', 'var', 'let', 'const', 'try', \n 'catch', 'finally', 'throw', 'typeof', 'instanceof', 'void', 'delete', 'in', 'of', \n 'yield', 'await', 'async', 'import', 'export', 'default', 'true', 'false', 'null', \n 'undefined', 'NaN', 'Infinity'\n }\n\n # 一系列正则用于提取JS/TS变量与参数\n var_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n param_pattern = re.compile(r'\\bfunction\\s*[a-zA-Z_$]*\\s*\\(([^)]*)\\)')\n arrow_param_pattern = re.compile(r'(?:(?:const|let|var)\\s+[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=\\s*)?\\(([^)]*)\\)\\s*=>')\n arrow_single_param_pattern = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>')\n \n def extract_from_js(code):\n # 移除单行及多行注释，避免对注释内容误提取\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n # 1. var, let, const 单个声明\n for match in var_pattern.findall(code):\n all_vars.add(match)\n \n # 2. 对象解构\n for match in re.findall(r'\\b(?:var|let|const)\\s*\\{([^}]+)\\}', code):\n for p in match.split(','):\n p_name = p\n # 处理带有别名重命名的解构 {a: b}\n if ':' in p:\n p_name = p.split(':')[1]\n p_name = p_name.split('=')[0].strip()\n if p_name:\n all_vars.add(p_name)\n \n # 3. 数组解构\n for match in re.findall(r'\\b(?:var|let|const)\\s*\\[([^\\]]+)\\]', code):\n for p in match.split(','):\n p_name = p.split('=')[0].strip()\n if p_name:\n all_vars.add(p_name)\n \n # 4. 普通函数参数\n for match in param_pattern.findall(code):\n for p in match.split(','):\n if p.strip():\n all_vars.add(p.strip())\n \n # 5. 箭头函数参数 (多参数和无参数)\n for match in arrow_param_pattern.findall(code):\n for p in match.split(','):\n if p.strip():\n all_vars.add(p.strip())\n \n # 6. 箭头函数参数 (单参数)\n for match in arrow_single_param_pattern.findall(code):\n if match.strip():\n all_vars.add(match.strip())\n \n # 7. catch 块异常变量\n for match in re.findall(r'\\bcatch\\s*\\(\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\)', code):\n all_vars.add(match)\n \n # 8. import 导入的默认变量\n for match in re.findall(r'\\bimport\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s+from', code):\n all_vars.add(match)\n \n # 9. import 导入的解构变量\n for match in re.findall(r'\\bimport\\s*\\{([^}]+)\\}', code):\n for p in match.split(','):\n p_name = p\n if ' as ' in p:\n p_name = p.split(' as ')[1]\n p_name = p_name.strip()\n if p_name:\n all_vars.add(p_name)\n\n def extract_from_py(code):\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.Assign):\n for target in node.targets:\n if isinstance(target, ast.Name):\n all_vars.add(target.id)\n elif isinstance(node, ast.FunctionDef) or isinstance(node, ast.AsyncFunctionDef):\n for arg in node.args.args:\n all_vars.add(arg.arg)\n if node.args.vararg:\n all_vars.add(node.args.vararg.arg)\n if node.args.kwarg:\n all_vars.add(node.args.kwarg.arg)\n except:\n pass\n\n has_files = False\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except:\n continue\n \n # 识别对应类型的文件并分别提取其中变量名\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\n has_files = True\n extract_from_js(content)\n elif ext in ['.html', '.htm', '.vue']:\n has_files = True\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n extract_from_js(script)\n elif ext == '.py':\n has_files = True\n extract_from_py(content)\n \n if not has_files:\n return False\n \n valid_vars = set()\n for var in all_vars:\n # 清理TS类型注解及默认值，提取最核心的变量名称\n var_name = var.split(':')[0].split('=')[0].strip()\n # 清除变长参数的前缀 `...`\n var_name = var_name.lstrip('.')\n \n # 确保只匹配到合法的标识符部分\n match = re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*', var_name)\n if match:\n v = match.group(0)\n if v not in keywords:\n valid_vars.add(v)\n\n # 如未找到任何变量，项目结构逻辑可能存在异常或未包含脚本代码，故当做未完成约束处理\n if not valid_vars:\n return False\n\n # 检查所有被提取到的变量名的长度，是否恰好都等于质数\n for var in valid_vars:\n if not is_prime(len(var)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保最终生成的代码文件中完全避开for循环的使用","validation_code":"def check_for_loops(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对不同语言环境定义 for 循环匹配正则\n # 匹配 JS/TS/Java/C++ 等的 for 循环: for ( 或者 for await (，且排除作为属性调用的情况 (如 obj.for)\n js_for_pattern = re.compile(r'(?<!\\.)\\bfor\\s*(?:await\\s+)?\\(')\n # 匹配 Vue 的指令循环\n vue_for_pattern = re.compile(r'\\bv-for\\b')\n # 匹配 CSS/SCSS 的循环\n css_for_pattern = re.compile(r'@for\\s+')\n # 匹配 Python 的 for ... in 循环及推导式\n py_for_pattern = re.compile(r'(?<!\\.)\\bfor\\s+.*?\\s+in\\b', re.DOTALL)\n \n # 去除字符串字面量的正则表达式，避免字符串中恰好存在关键字导致误判\n str_double = re.compile(r'\"([^\"\\\\]|\\\\.)*\"')\n str_single = re.compile(r\"'([^'\\\\]|\\\\.)*'\")\n str_backtick = re.compile(r\"`([^`\\\\]|\\\\.)*`\")\n \n # 去除各语言常见注释的正则表达式\n comment_single = re.compile(r'//.*')\n comment_multi = re.compile(r'/\\*.*?\\*/', re.DOTALL)\n comment_html = re.compile(r'<!--.*?-->', re.DOTALL)\n comment_py = re.compile(r'#.*')\n \n # 分类定义需要扫描的文件扩展名\n js_exts = {'.js', '.ts', '.jsx', '.tsx'}\n html_exts = {'.html', '.htm', '.vue'}\n css_exts = {'.css', '.scss', '.less'}\n py_exts = {'.py'}\n other_exts = {'.java', '.c', '.cpp', '.cs'}\n valid_exts = js_exts | html_exts | css_exts | py_exts | other_exts\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 第一步：先去除字符串内容，避免字符串里的注释或代码干扰\n code = str_double.sub('\"\"', code)\n code = str_single.sub(\"''\", code)\n code = str_backtick.sub(\"``\", code)\n \n # 第二步：去除所有的注释文本\n code = comment_single.sub('', code)\n code = comment_multi.sub('', code)\n code = comment_html.sub('', code)\n code = comment_py.sub('', code)\n \n # 第三步：按照文件类型，执行对应的 for 循环安全检查\n if ext in js_exts | html_exts | other_exts:\n if js_for_pattern.search(code):\n return False\n \n if ext in html_exts:\n if vue_for_pattern.search(code):\n return False\n \n if ext in css_exts | html_exts:\n if css_for_pattern.search(code):\n return False\n \n if ext in py_exts:\n if py_for_pattern.search(code):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"你构建的代码文件中必须排除对while循环的使用","validation_code":"import os\nimport re\n\ndef verify_no_while_loop(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证代码项目中是否排除了对 while 循环的使用。\n \"\"\"\n \n def clean_code(code: str) -> str:\n # 正则匹配并移除注释和字符串，防止误判注释或字符串中的 'while' 单词\n pattern = re.compile(\n r'<!--.*?-->|' # HTML 注释\n r'/\\*.*?\\*/|' # C/JS 多行注释\n r'//.*?$|' # JS 单行注释\n r'#.*?$|' # Python/Shell 单行注释\n r'\"\"\".*?\"\"\"|' # Python 多行双引号字符串\n r\"'''.*?'''|\" # Python 多行单引号字符串\n r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"|' # 双引号字符串\n r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'|\" # 单引号字符串\n r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', # JS 模板字符串\n re.DOTALL | re.MULTILINE\n )\n return pattern.sub('', code)\n\n # 常见的前端和可能涉及的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.svelte', '.py'}\n\n if not os.path.exists(workspace_path):\n # 如果项目路径不存在，无法验证是否有代码文件，视作未违反\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n # 剔除代码中的注释和字符串\n cleaned_code = clean_code(code)\n \n # 检查清理后的代码中是否存在完整的 'while' 关键字\n if re.search(r'\\bwhile\\b', cleaned_code):\n return False\n except Exception:\n # 忽略无法用 utf-8 读取的文件或其他文件读取异常\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对前端通知组件项目，定义可能存在的常见代码文件后缀\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h'\n }\n \n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅校验代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 指令中要求所有文件须为UTF-8编码，使用 errors='ignore' 避免偶然的特殊字符引发读取崩溃\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n # 剔除行尾的回车与换行符，保留行内的所有字符(包含空格)\n line_content = line.rstrip('\\r\\n')\n # 当某一行字符数超出80时，直接返回False\n if len(line_content) > 80:\n return False\n except Exception:\n # 若文件由于某种意外原因完全无法读取，则保守判定为失败\n return False\n \n # 如果存在代码文件且均通过了行宽检查，则代表满足要求\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":7,"instruction":"我们再增加一项高级手势交互：支持拖拽移除。允许我用鼠标按住任意一条通知向右拖拽滑动。拖拽过程中，通知卡片要跟随鼠标横向移动，且透明度随着滑动距离的增加而逐渐降低；如果我拖拽超过卡片宽度的40%然后松手，卡片直接顺势飞出屏幕并触发销毁逻辑；如果拖动没超过40%就松手，卡片需要带有弹性回弹动画（Spring）恢复到原位。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_drag 作为结尾；二是对于你输出的每一个代码文件，请控制其连同空行在内的总行数是一个奇数。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前端及后端代码文件扩展名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.vue', '.json', '.py', '.java', '.c', '.cpp', '.h', \n '.sh', '.md', '.txt', '.xml', '.yaml', '.yml'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs，避免进入隐藏目录（如 .git, .idea 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试用 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 只要有一个文件出现 UTF-8 解码错误，直接返回 False\n return False\n except Exception:\n # 忽略因权限等导致的其他读取错误\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据用户指令推断，这是一个前端UI组件项目，定义常见的前端代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n code_file_count = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅针对代码文件进行检查，排除诸如 .log 这样的非代码文件\n if ext in code_extensions:\n code_file_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n found_comment = False\n # 检查文件开头处（设定前5行内）是否包含目标注释字符串\n for line_idx, line in enumerate(f):\n if line_idx >= 5:\n break\n if target_comment in line:\n found_comment = True\n break\n \n if not found_comment:\n return False\n except Exception:\n # 如果由于编码或其他问题导致文件无法正常读取，判定为失败\n return False\n \n # 如果没有找到任何相关的代码文件，也视作未满足生成代码的检查前提\n if code_file_count == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n workspace = Path(workspace_path).resolve()\n \n for root, dirs, files in os.walk(workspace_path):\n current_dir = Path(root).resolve()\n \n try:\n # 计算相对于项目根目录的相对路径\n rel_path = current_dir.relative_to(workspace)\n except ValueError:\n continue\n \n # 忽略以 '.' 开头的隐藏目录及其子目录（如 .git, .vscode 等）\n if any(part.startswith('.') for part in rel_path.parts):\n continue\n \n # rel_path.parts 的长度代表了该目录的层级\n # - 项目根目录本身 parts 为 ()，长度为 0\n # - 第一层子目录 (如 src) parts 为 ('src',)，长度为 1\n # - 第二层子目录 (如 src/components) parts 为 ('src', 'components')，长度为 2\n # - 第三层及以上子目录会导致长度 > 2，这意味着超过了2层目录的限制\n if len(rel_path.parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_drag 作为结尾","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对前端或后端页面通知组件可能用到的后缀名\n exts = ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte', '.py']\n \n # 关键字黑名单：防止将流程控制、内置词或关键字错误识别为函数名\n ignore_names = {\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'render', \n 'return', 'super', 'function', 'class', 'import', 'export', 'require',\n 'setTimeout', 'setInterval', 'clearTimeout', 'clearInterval', 'Promise',\n 'map', 'filter', 'reduce', 'forEach', 'then', 'catch', 'finally',\n 'get', 'set', 'static', 'async', 'await', 'new', 'delete', 'typeof',\n 'in', 'instanceof', 'void', 'with', 'yield', 'let', 'var', 'const'\n }\n \n # 常见代码范式下的函数名提取正则\n patterns = [\n # JS/TS/HTML: function name(...) 或 async function name(...)\n re.compile(r'\\b(?:async\\s+)?function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('),\n # JS/TS/HTML: const/let/var name = [async] function / (...) => / x =>\n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'),\n # JS/TS class method: [static] [async] [get/set] name(...) {\n re.compile(r'^\\s*(?:static\\s+)?(?:async\\s+)?(?:get\\s+|set\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{', re.MULTILINE),\n # JS/TS class property arrow function: name = [async] (...) =>\n re.compile(r'^\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)', re.MULTILINE),\n # Python: def name(...) 或 async def name(...)\n re.compile(r'\\b(?:async\\s+)?def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n ]\n \n for root, dirs, files in os.walk(workspace_path):\n # 避开通常不包含AI直接生成代码的常见第三方库或环境目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n if any(file.endswith(ext) for ext in exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 依序使用不同正则匹配并校验其中的函数名称\n for pattern in patterns:\n for match in pattern.finditer(content):\n func_name = match.group(1)\n # 过滤关键字或保留字\n if func_name in ignore_names:\n continue\n # 检查后缀约束，不符合则直接返回 False\n if not func_name.endswith('_noti_drag'):\n return False\n except Exception:\n # 对于不可读或编解码等异常问题忽略，继续排查下一个文件\n continue\n \n # 未发现任何违反要求的函数命名，返回 True\n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断一个数字是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 匹配类定义的正则表达式：匹配 `class ` 后跟随的合法标识符作为类名\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z_][A-Za-z0-9_]*)\\b')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 读取文件，若为非文本或编码不匹配则忽略\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 找到当前文件中的所有类名\n matches = class_pattern.findall(content)\n for class_name in matches:\n # 如果存在类名长度为质数，则违背了“长度为非质数”的要求\n if is_prime(len(class_name)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def verify_variable_length_is_prime(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n variables = set()\n\n # 用于匹配JS/TS等前端语言中通过 var/let/const 声明的变量\n var_pattern = r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:=|;|:|,)'\n # 用于匹配JS/TS等前端语言中基本的函数参数变量\n func_pattern = r'\\bfunction\\s*(?:[a-zA-Z0-9_$]+)?\\s*\\(([^)]*)\\)'\n\n # 1. 扫描工作区文件提取变量\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n\n if ext in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if ext == '.py':\n # Python 文件使用 ast 精准提取变量\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n variables.add(node.id)\n elif isinstance(node, ast.arg):\n variables.add(node.arg)\n except SyntaxError:\n pass\n else:\n # 前端文件使用正则提取变量\n for match in re.finditer(var_pattern, content):\n variables.add(match.group(1))\n \n for match in re.finditer(func_pattern, content):\n args_str = match.group(1)\n if args_str.strip():\n for arg in args_str.split(','):\n m = re.search(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg.strip())\n if m:\n variables.add(m.group(0))\n\n # 2. 作为补充，扫描回复文本中的代码块以防文件未写入\n code_blocks = re.findall(r'```(?:javascript|js|typescript|ts|html|vue)\\s*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n for match in re.finditer(var_pattern, block):\n variables.add(match.group(1))\n for match in re.finditer(func_pattern, block):\n args_str = match.group(1)\n if args_str.strip():\n for arg in args_str.split(','):\n m = re.search(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg.strip())\n if m:\n variables.add(m.group(0))\n\n # 如果没有找到任何变量，则无法判定其满足了要求，返回False\n if not variables:\n return False\n\n # 3. 校验所有提取出来的变量长度是否全为质数\n for var in variables:\n if not is_prime(len(var)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保最终生成的代码文件中完全避开for循环的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def strip_js_code(code):\n # 移除单行注释、多行注释以及单/双/反引号字符串，避免匹配到其中的 for 单词字样\n # 限制单/双引号匹配不能跨行，以避免因语法残缺漏写闭合引号而误吞大段代码\n pattern = re.compile(\n r'//[^\\n]*|' # 单行注释\n r'/\\*[\\s\\S]*?\\*/|' # 多行注释\n r\"'(?:\\\\.|[^'\\n\\r])*'|\" # 单引号字符串\n r'\"(?:\\\\.|[^\"\\n\\r])*\"|' # 双引号字符串\n r'`(?:\\\\.|[^`])*`' # 模板字符串（允许跨行）\n )\n return pattern.sub('', code)\n\n def strip_py_code(code):\n # 移除 Python 格式的注释与字符串\n pattern = re.compile(\n r'#[^\\n]*|'\n r\"'''[\\s\\S]*?'''|\"\n r'\"\"\"[\\s\\S]*?\"\"\"|'\n r\"'(?:\\\\.|[^'\\n\\r])*'|\"\n r'\"(?:\\\\.|[^\"\\n\\r])*\"'\n )\n return pattern.sub('', code)\n\n # 匹配 JavaScript / TypeScript 的 for 循环结构\n # 匹配: for( 且紧跟循环体的合理起始特征：let, var, const, ;, {, [, +, - 或合法的变量名配合后续的操作符(如 =, in, of 等)\n js_for_pattern = re.compile(\n r'\\bfor\\s*(?:await\\s*)?\\(\\s*(?:'\n r'let\\b|var\\b|const\\b|;|\\{|\\[|[+\\-]|'\n r'[A-Za-z_$][\\w$]*\\s*(?:[=+\\-*/%&|^<>!(.\\[]|\\b(?:in|of)\\b)'\n r')'\n )\n \n # 匹配 Python 的 for 循环结构\n py_for_pattern = re.compile(\n r'\\bfor\\s+(?:[A-Za-z_]\\w*(?:\\s*,\\s*[A-Za-z_]\\w*)*)\\s+in\\s+'\n )\n \n # 匹配各类常见前端模板中的 for 循环指令\n template_for_pattern = re.compile(\n r'\\bv-for\\s*=|' # Vue v-for\n r'\\bngFor\\b|' # Angular (*ngFor)\n r'\\{%\\s*for\\s+' # Jinja / Django 等模板语法\n )\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.html', '.py', '.scss', '.sass']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 检测前端模板中的特有循环指令 (保留原文检测，因为指令通常以字符串格式作为属性值)\n if template_for_pattern.search(content):\n return False\n \n # 检测 SCSS/SASS 中的 @for 循环\n if ext in ['.scss', '.sass']:\n if re.search(r'@for\\s+', content):\n return False\n continue\n\n # 检测 Python 中的 for 循环\n if ext == '.py':\n stripped = strip_py_code(content)\n if py_for_pattern.search(stripped):\n return False\n continue\n\n # 提取 HTML, Vue, Svelte 中的 <script> 区域内容\n # 防止 <template> 等非脚本区域的普通文本内容对后面纯 JS 正则校验造成干扰\n if ext in ['.html', '.vue', '.svelte']:\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n script_content = '\\n'.join(scripts)\n else:\n script_content = content\n\n # 处理剥离注释及纯文本字符串后的 JS/TS 代码排查 for 循环关键字结构\n if script_content.strip():\n stripped_script = strip_js_code(script_content)\n if js_for_pattern.search(stripped_script):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"你构建的代码文件中必须排除对while循环的使用","validation_code":"import os\nimport re\n\ndef verify_no_while_loop(response: str, workspace_path: str) -> bool:\n # 定义前端项目中及常见后端可能用到的代码文件后缀\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.css', '.py', '.java', '.c', '.cpp', '.cs'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n in_single_comment = False\n in_multi_comment = False\n in_string = False\n string_char = ''\n i = 0\n n = len(content)\n clean_code_chars = []\n \n # 使用简易状态机去除注释与字符串的干扰\n while i < n:\n c = content[i]\n \n # 处于单行注释中\n if in_single_comment:\n if c == '\\n':\n in_single_comment = False\n clean_code_chars.append('\\n')\n i += 1\n continue\n \n # 处于多行注释中\n if in_multi_comment:\n if c == '*' and i + 1 < n and content[i+1] == '/':\n in_multi_comment = False\n i += 2\n else:\n i += 1\n continue\n \n # 处于字符串之中\n if in_string:\n if c == '\\\\': # 遇到转义字符直接跳过\n if i + 1 < n:\n i += 2\n else:\n i += 1\n continue\n if c == string_char: # 匹配到字符串结尾引号\n in_string = False\n clean_code_chars.append(c)\n i += 1\n continue\n \n # 检测字符串开头\n if c in (\"'\", '\"', '`'):\n in_string = True\n string_char = c\n clean_code_chars.append(c)\n i += 1\n continue\n \n # 检测 JS/TS/Java/C 等语言单行注释的开头 (//)\n if c == '/' and i + 1 < n and content[i+1] == '/':\n in_single_comment = True\n i += 2\n continue\n \n # 检测 JS/TS/CSS 等语言多行注释的开头 (/*)\n if c == '/' and i + 1 < n and content[i+1] == '*':\n in_multi_comment = True\n i += 2\n continue\n \n # 检测 HTML/Vue 的注释开头 (<!--)\n if c == '<' and content[i:i+4] == '<!--':\n end_idx = content.find('-->', i + 4)\n if end_idx != -1:\n i = end_idx + 3\n else:\n i = n\n continue\n \n # 检测 Python/Ruby/Shell 等的单行注释 (#)\n if ext in {'.py', '.rb', '.sh'} and c == '#':\n in_single_comment = True\n i += 1\n continue\n \n # 如果不在任何注释或字符串中，加入清洗后的干净字符集合\n clean_code_chars.append(c)\n i += 1\n \n clean_str = \"\".join(clean_code_chars)\n \n # 使用 \\b 确保匹配到的是严格作为单一关键字的 while\n if re.search(r'\\bwhile\\b', clean_str):\n return False\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令（开发页面通知组件）推断可能的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.vue', '.scss', '.less'\n }\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为推断的代码文件类型\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 仅去除行末尾的换行符，以便准确计算这一行的实际行宽\n line_content = line.rstrip('\\r\\n')\n \n # 要求中明确说明“字符”，故直接使用len函数进行统计\n if len(line_content) > 80:\n return False\n except Exception:\n # 若文件由于某种异常（如非文本格式、编码问题）无法读取，则安全跳过\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于你输出的每一个代码文件，请控制其连同空行在内的总行数是一个奇数","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.go', '.rs', '.php', '.rb'\n }\n \n has_verified_any = False\n \n # 1. 优先检查 workspace_path 下生成的代码文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n # 统计包括空行在内的总行数，如果是偶数则不符合要求\n if len(lines) % 2 == 0:\n return False\n has_verified_any = True\n except Exception:\n # 忽略无法以 utf-8 读取或者出现其他异常的文件\n continue\n \n # 2. 如果 workspace_path 中没有检测到任何代码文件，则回退检查 response 中的代码块\n if not has_verified_any:\n # 使用正则表达式提取 Markdown 格式的代码块\n code_blocks = re.findall(r'```[a-zA-Z]*\\n(.*?)```', response, re.DOTALL)\n if not code_blocks:\n # 既没有检测到代码文件也没有发现代码块，无法证实满足了条件\n return False\n \n for block in code_blocks:\n # 兼容正则匹配时末尾自带的单一换行符，避免 split('\\n') 时多出最后一行空行\n # 这旨在使统计逻辑与 f.readlines() 的正常表现对齐\n if block.endswith('\\n'):\n block = block[:-1]\n \n lines = block.split('\\n')\n \n if len(lines) % 2 == 0:\n return False\n has_verified_any = True\n \n return has_verified_any","tags":["Quantity","Exact Value"]}]},{"round_id":8,"instruction":"最后一步，完善极限情况下的等待队列管理。限制DOM树中最多只允许渲染5条通知节点（无论是展开还是折叠状态）。如果在短时间内触发了超过5个，多出来的不能直接渲染，而是要进入内存里的'后台等待队列'。一旦当前可见的某条通知被销毁腾出了名额，等待队列里的通知要立刻根据优先级（高优优先，同级先进先出）自动被调度进场。请在页面上加个统计面板，让我能实时看到'当前显示: M条，后台排队: N条'来验证这个逻辑。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_queue 作为结尾；二是对于你生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-。","instruction_following_checklist":[{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 依据用户指令中构建页面通知组件等要求，推断可能的前端代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n # 遍历当前代码项目工作区\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 utf-8 编码完整读取并解码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 抛出 UnicodeDecodeError 说明文件并非合法的 UTF-8 编码\n return False\n except Exception:\n # 若出现其他无法读取的情况，保守起见视作不满足要求\n return False\n\n # 需确保项目内确实生成了代码文件且全部验证通过，才判定为满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_string = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 针对前端或常见项目生成的代码文件类型\n valid_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs'\n }\n \n # 优先检查工作区生成的物理文件\n if workspace_path and os.path.exists(workspace_path):\n has_code_files = False\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in valid_extensions:\n has_code_files = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = []\n # 只取前10行来寻找是否在文件“开头处”包含该声明注释\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n lines.append(line)\n \n found = any(target_string in line for line in lines)\n # 一旦有代码文件前10行内没有规定注释，则判为未满足\n if not found:\n return False \n except Exception:\n return False\n \n # 若检查了所有的物理代码文件都没问题，则返回通过\n if has_code_files:\n return True\n\n # 兜底方案：如果工作区路径失效或无物理文件，则检查最后一轮回复当中的代码块\n code_blocks = re.findall(r'```[\\w]*\\n(.*?)```', response, re.DOTALL)\n if not code_blocks:\n return False\n \n has_valid_blocks = False\n for block in code_blocks:\n lines = block.strip().split('\\n')\n if not lines:\n continue\n has_valid_blocks = True\n found = any(target_string in line for line in lines[:10])\n if not found:\n return False\n \n return has_valid_blocks","tags":["Content","Format"]},{"constraint":"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过2层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉以 '.' 开头的隐藏目录（如 .git, .vscode 等），这些不计入项目真实层级\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n\n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n\n # 根目录 '.' 为第 1 层\n if rel_path == '.':\n depth = 1\n else:\n # 通过系统路径分隔符拆分相对路径，计算所在层级\n # 例如相对路径 'folder1' -> ['folder1']，深度为 1 + 1 = 2\n # 相对路径 'folder1/folder2' -> ['folder1', 'folder2']，深度为 2 + 1 = 3\n depth = len(rel_path.split(os.sep)) + 1\n\n # 若发现任何目录深度超过 2 层，则判定为不符合要求\n if depth > 2:\n return False\n\n # 遍历结束未发现超过2层的目录，则符合要求\n return True","tags":["Quantity","Complexity"]},{"constraint":"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_queue 作为结尾","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 涵盖不同语言（JS/TS, Python等）常见的函数定义模式\n patterns = [\n # Python: def foo()\n r'\\bdef\\s+([a-zA-Z_]\\w*)\\s*\\(',\n \n # JS/TS: function foo()\n r'\\bfunction\\s*\\*?\\s+([a-zA-Z_$][\\w$]*)\\s*\\(',\n \n # JS/TS: const/let/var foo = function()\n r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?function\\b',\n \n # JS/TS: const/let/var foo = () =>\n r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][\\w$]*)\\s*=>',\n \n # JS/TS 对象属性函数: foo: function()\n r'\\b([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?function\\b',\n \n # JS/TS 对象属性箭头函数: foo: () =>\n r'\\b([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][\\w$]*)\\s*=>',\n \n # JS/TS 类或对象的简写方法: foo() {\n r'^\\s*(?:async\\s+)?(?:get\\s+|set\\s+)?([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{',\n \n # JS/TS 类属性箭头函数: foo = () =>\n r'^\\s*([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][\\w$]*)\\s*=>'\n ]\n \n # 语言关键字黑名单（为了避免大括号形式的控制流语句被误判为函数名）\n reserved_words = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', \n 'function', 'constructor', 'return', 'super', 'with', 'typeof',\n 'instanceof', 'new', 'delete', 'yield', 'await', 'get', 'set'\n }\n\n # 常见代码文件格式后缀\n extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\n \n found_functions = []\n \n if not os.path.exists(workspace_path):\n return False\n \n has_files = False\n # 遍历当前工作空间内的所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in extensions:\n has_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过各正则模式提取该文件内的函数名\n for pattern in patterns:\n matches = re.finditer(pattern, content, re.MULTILINE)\n for match in matches:\n func_name = match.group(1)\n if func_name not in reserved_words:\n found_functions.append(func_name)\n except Exception:\n pass\n\n # 至少要解析到代码文件才能判定通过\n if not has_files:\n return False\n \n if not found_functions:\n return True \n \n # 遍历所有的函数名，验证其是否满足以 `_noti_queue` 结尾\n for func in found_functions:\n if not func.endswith('_noti_queue'):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断给定正整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 预编译常见的类名正则匹配器\n # 1. 面向对象的类声明 (如: class MyClass)\n decl_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_]\\w*)')\n # 2. HTML / JSX 中的类属性 (如: class=\"my-class container\")\n attr_pattern = re.compile(r'\\bclass(?:Name)?\\s*=\\s*[\\'\"]([^\\'\"]+)[\\'\"]')\n # 3. CSS 规则中的类选择器 (如: .my-class)\n css_pattern = re.compile(r'\\.([a-zA-Z_][a-zA-Z0-9_-]*)')\n # 4. 匹配 HTML 中的 <style> 内容\n style_tag_pattern = re.compile(r'<style[^>]*>(.*?)</style>', re.DOTALL | re.IGNORECASE)\n\n # 包含通常会出现类名的相关代码文件类型\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.svelte', '.py', '.java'}\n \n found_any_class = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 1. 检查代码的类定义声明\n for match in decl_pattern.finditer(content):\n cls_name = match.group(1)\n found_any_class = True\n if is_prime(len(cls_name)):\n return False\n \n # 2. 检查 HTML / JSX 标签的类名属性\n for match in attr_pattern.finditer(content):\n classes_str = match.group(1)\n for cls_name in classes_str.split():\n cls_name = cls_name.strip()\n if cls_name:\n found_any_class = True\n if is_prime(len(cls_name)):\n return False\n \n # 3. 检查 CSS 样式表及其样式块中的类选择器\n if ext == '.css':\n for match in css_pattern.finditer(content):\n cls_name = match.group(1)\n found_any_class = True\n if is_prime(len(cls_name)):\n return False\n elif ext in {'.html', '.vue', '.svelte'}:\n for style_match in style_tag_pattern.finditer(content):\n style_content = style_match.group(1)\n for match in css_pattern.finditer(style_content):\n cls_name = match.group(1)\n found_any_class = True\n if is_prime(len(cls_name)):\n return False\n\n # 若检测到了至少一个类名，且没有触发由于质数长度抛出的 False，则证明所有类名均符合规则\n return found_any_class","tags":["Style","Naming"]},{"constraint":"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def is_prime(n: int) -> bool:\n \"\"\"判断一个正整数是否为质数\"\"\"\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 排除隐藏文件\n if file_name.startswith('.'):\n continue\n \n file_path = os.path.join(root, file_name)\n ext = os.path.splitext(file_name)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n # 针对前端或页面代码文件\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\n # 利用正则匹配 JavaScript / TypeScript 中通过 let, const, var 声明的变量名\n # 这里主要抓取声明语句紧跟的第一个标识符作为变量名代表\n matches = re.findall(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', content)\n for var_name in matches:\n if not is_prime(len(var_name)):\n return False\n \n # 针对 Python 代码文件（尽管主要需求是前端页面，但也可能包含 Python 提供服务）\n elif ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 检查赋值产生的变量\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if not is_prime(len(node.id)):\n return False\n # 检查函数参数变量\n elif isinstance(node, ast.arg):\n if not is_prime(len(node.arg)):\n return False\n except SyntaxError:\n # 忽略存在语法错误的 Python 文件\n pass\n\n return True","tags":["Style","Naming"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保最终生成的代码文件中完全避开for循环的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义常见的前端及后端源代码文件扩展名\n extensions = {\n '.js', '.ts', '.jsx', '.tsx', \n '.vue', '.svelte', '.html', \n '.py', '.java', '.c', '.cpp', \n '.cs', '.go', '.rb', '.php', \n '.swift', '.kt', '.mjs', '.cjs'\n }\n \n # 构建匹配多种语言环境下 for 循环语法的正则表达式\n # 1. (?<![\\.\\?])\\bfor\\s*\\( 匹配 JS, TS, Java, C++, C# 等的 for 循环，同时排除类似 Symbol.for() 这样的对象方法调用\n # 2. (?<![\\.\\?])\\bfor\\s+[a-zA-Z0-9_,\\s]+\\s+in\\b 匹配 Python 等语言的 for ... in ... 循环结构\n # 3. \\bv-for\\s*= 匹配 Vue 框架中的列表渲染指令\n # 4. \\*ngFor\\s*= 匹配 Angular 框架中的列表渲染指令\n # 5. (?<![\\.\\?])\\bfor\\s+[a-zA-Z0-9_,\\s]+\\s*:= 匹配 Go 语言的 for 循环推导\n # 6. (?<![\\.\\?])\\bfor\\s*\\{ 匹配 Go 语言的 for 无限循环\n for_pattern = re.compile(\n r'(?<![\\.\\?])\\bfor\\s*\\(|' \n r'(?<![\\.\\?])\\bfor\\s+[a-zA-Z0-9_,\\s]+\\s+in\\b|' \n r'\\bv-for\\s*=|' \n r'\\*ngFor\\s*=|' \n r'(?<![\\.\\?])\\bfor\\s+[a-zA-Z0-9_,\\s]+\\s*:=|' \n r'(?<![\\.\\?])\\bfor\\s*\\{' \n )\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除字符串字面量（避免字符串内存在 'for(' 等字眼干扰判断）\n # 处理常见的双引号和单引号，允许内部包含转义字符\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', content)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", content)\n # 处理反引号模板字符串（可能跨行使用）\n content = re.sub(r'`([^`\\\\]|\\\\.)*`', '``', content, flags=re.DOTALL)\n \n # 移除各种语言的注释内容（单行与多行注释），避免注释中的内容引发误判\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 如果在清理后的代码正文中检测到了任何 for 循环语句结构，说明未避开 for 循环的使用\n if for_pattern.search(content):\n return False\n except Exception:\n # 忽略读取异常的文件，继续检查下一个文件\n continue\n \n # 所有的代码文件均未检测到 for 循环，满足用户指令的要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你构建的代码文件中必须排除对while循环的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义各类字符串字面量的正则表达式（处理转义字符的情况）\n str_double = r'\"(?:\\\\.|[^\"\\\\])*\"'\n str_single = r\"'(?:\\\\.|[^'\\\\])*'\"\n str_backtick = r\"`(?:\\\\.|[^`\\\\])*`\"\n \n # 定义前端各类注释的正则表达式\n comment_html = r'<!--[\\s\\S]*?-->'\n comment_multi = r'/\\*[\\s\\S]*?\\*/'\n comment_single = r'//.*'\n\n # 针对 JS/TS/HTML/CSS 等前端文件的清理规则（剔除字符串和常规前端注释）\n strip_pattern_js = re.compile('|'.join([\n str_double, str_single, str_backtick,\n comment_html, comment_multi, comment_single\n ]))\n\n # 针对 Python 文件的清理规则（剔除字符串和 # 注释）\n comment_py = r'#.*'\n strip_pattern_py = re.compile('|'.join([\n str_double, str_single, str_backtick,\n comment_py\n ]))\n\n # 待检查的常见代码文件后缀类型\n code_extensions = {'.html', '.htm', '.js', '.ts', '.jsx', '.tsx', '.vue', '.css', '.scss', '.py'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 根据文件类型应用不同的文本清理策略\n if ext == '.py':\n clean_content = strip_pattern_py.sub('', content)\n else:\n clean_content = strip_pattern_js.sub('', content)\n \n # 查找 while 关键字，利用 (?<!\\.) 排除作为对象属性的情形（如 object.while）\n if re.search(r'(?<!\\.)\\bwhile\\b', clean_content):\n return False\n except Exception:\n # 忽略存在读取权限或编码错误问题的文件\n continue\n\n # 遍历结束无任何 while 关键字检出，视为满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令中对“DOM树、页面通知组件”的描述，推断可能的前端代码文件类型\n target_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n if not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历工作区目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去除行末的换行符和回车符以严格计算行内的真实字符数\n clean_line = line.rstrip('\\r\\n')\n # 要求明确说明统计“字符数”，因此遵循提示使用len()函数统计\n if len(clean_line) > 80:\n return False\n except Exception:\n # 若因编码问题等无法读取文件流，则保守返回 False\n return False\n \n # 只有在找到了前端代码文件，并且没有一行的字符数大于80的情况下，才视为满足该要求\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于你输出的每一个代码文件，请控制其连同空行在内的总行数是一个奇数","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 常见的代码文件后缀名集合\n code_extensions = {\n '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rb', '.php', '.swift', '.kt'\n }\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历工作区内所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符切分获取文本编辑器意义上的总行数（连同空行）\n lines = content.split('\\n')\n total_lines = len(lines)\n \n # 检查是否为奇数\n if total_lines % 2 == 0:\n return False\n except Exception:\n # 如果由于编码等问题无法读取文件，默认未能满足验证\n return False\n \n # 如果有代码文件并且全都满足奇数行条件，则返回True，否则返回False\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目及其他常见代码文件的扩展名集合\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', '.scss', '.less',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.sh'\n }\n \n if not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的首行，并转换为小写以作宽容匹配\n first_line = f.readline().strip().lower()\n # 判断首行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法以 UTF-8 编码读取或发生其他读取异常，视为不满足要求\n return False\n \n # 如果项目中没有任何代码文件，同样视为未达到要求\n return has_code_file","tags":["Content","Format"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我搭一个页面通知组件（Notification）的基础框架。页面中间放一个'发送通知'的按钮。点击按钮后，在屏幕右下角能弹出一个简单的黑底白字消息卡片。现在暂时不需要做复杂的动画和业务逻辑，只要让它能渲染在右下角，并且在卡片上提供一个'X'按钮让我能点击销毁它就可以了。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _notif 作为结尾；二是你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，页面通知组件的代码大概率涉及以下前端文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 如果有其他异常（如文件无法读取），保守跳过或视作无效情况处理，这里不做判定失败\\n continue\\n \\n # 所有代码文件都顺利通过了 UTF-8 编码测试\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = []\\n # 提取文件开头的前5个非空行\\n for line in f:\\n if line.strip():\\n lines.append(line)\\n if len(lines) >= 5:\\n break\\n \\n # 检查前几个非空行中是否包含要求中的注释字符串\\n found_comment = False\\n for line in lines:\\n if target_comment in line:\\n found_comment = True\\n break\\n \\n if not found_comment:\\n return False\\n except Exception:\\n # 如果遇到无法以UTF-8读取的文件或其他异常情况，默认视为验证失败\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 计算当前目录相对于根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录属于第1层，直接跳过\\n if rel_path == '.':\\n continue\\n \\n # 统一路径分隔符为 '/'，以此切分相对路径\\n parts = rel_path.replace('\\\\\\\\', '/').split('/')\\n \\n # 如果相对路径包含至少两部分（如 dir1/dir2），说明目录达到了第3层，即超过了2层\\n if len(parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _notif 作为结尾\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能的生成的代码文件类型\\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.py'}\\n \\n # 单行正则匹配模式\\n patterns = [\\n # JavaScript / TypeScript /等传统 function 声明: function foo(...)\\n r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(',\\n # 箭头函数赋值声明: const foo = (...) => 或 const foo = arg =>\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>',\\n # 匿名函数赋值声明: const foo = function(...)\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s*)?function\\\\b',\\n # 对象属性 function声明: foo: function()\\n r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s*)?function\\\\b',\\n # 对象属性 箭头函数声明: foo: () =>\\n r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>',\\n ]\\n\\n # 多行正则匹配模式\\n multiline_patterns = [\\n # Python: def foo(...)\\n r'^[ \\\\t]*def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(',\\n # JavaScript/TypeScript 类或对象的方法简写: foo() { ... } 或 async foo() { ... }\\n r'^[ \\\\t]*(?:async[ \\\\t]+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{',\\n # JavaScript/TypeScript 类的箭头函数属性: foo = () => { ... }\\n r'^[ \\\\t]*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>'\\n ]\\n \\n # 过滤掉由于简单的正则可能误判的编程语言关键字\\n keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'catch', 'function',\\n 'constructor', 'return', 'import', 'export', 'super', 'with', 'new', 'typeof',\\n 'await', 'yield'\\n }\\n\\n # 遍历项目目录结构\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见依赖、日志和编译输出缓存目录\\n for ignore_dir in ['node_modules', '.git', '__pycache__', 'dist', 'build']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n func_names = set()\\n \\n # 收集函数名\\n for p in patterns:\\n for match in re.finditer(p, code):\\n func_names.add(match.group(1))\\n \\n for p in multiline_patterns:\\n for match in re.finditer(p, code, re.MULTILINE):\\n func_names.add(match.group(1))\\n \\n # 验证函数名是否符合结尾要求\\n for name in func_names:\\n # 忽略控制流/原生关键字的混淆\\n if name in keywords:\\n continue\\n \\n if not name.endswith('_notif'):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个数字是否是质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 如果工作区不存在，无法验证具体代码，视作没有违规的类名\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 正则表达式定义\\n # 1. 匹配面向对象语言中的类名 (如 class NotificationBox)\\n class_def_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)')\\n \\n # 2. 匹配 HTML/JSX/Vue 中的 class 属性 (如 class=\\\"card-item\\\" 或 className=\\\"card-item\\\")\\n html_class_pattern = re.compile(r'\\\\b(?:class|className)\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']')\\n \\n # 3. 匹配 CSS 文件中定义的类名 (如 .card-item)\\n css_class_pattern = re.compile(r'\\\\.([a-zA-Z_][a-zA-Z0-9_-]*)')\\n\\n # 支持的前端及常规代码文件类型\\n valid_extensions = ('.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.py')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(valid_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 1. 检查面向对象的类名\\n for class_name in class_def_pattern.findall(content):\\n if is_prime(len(class_name)):\\n return False\\n \\n # 2. 检查 HTML/React 中的类名属性\\n for class_str in html_class_pattern.findall(content):\\n # class 属性里可能用空格分隔了多个类名，如 class=\\\"btn primary\\\"\\n class_names = class_str.strip().split()\\n for class_name in class_names:\\n if not class_name:\\n continue\\n if is_prime(len(class_name)):\\n return False\\n \\n # 3. 仅在 .css 文件中严格检查 CSS 类名定义（避免在 JS 文件中把 object.property 误认为 css 类）\\n if file.endswith('.css'):\\n for class_name in css_class_pattern.findall(content):\\n if is_prime(len(class_name)):\\n return False\\n \\n except Exception:\\n # 忽略读取或编码异常的文件\\n pass\\n\\n # 遍历结束没有发现任何长度为质数的类名，则满足“每个类名的长度恰好是一个非质数”的要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，现在我们给通知加上生命周期和基础动画。要求通知出现时，是从屏幕右侧平滑滑入并带有淡入（fade in）效果；显示停留3秒后，它要能自动向右滑动淡出并从DOM中彻底移除。如果我在3秒内手动点击那个关闭按钮，它也要播放同样的退出动画后再销毁。另外请你后续轮次始终遵循以下两条更新规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_anim 作为结尾；二是请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见前端页面组件相关的代码文件后缀\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n found_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试用UTF-8解码文件内容，若非UTF-8编码（且包含非ASCII字符）将抛出UnicodeDecodeError\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 发现无法使用UTF-8解码的代码文件，违反了要求\\n return False\\n except Exception:\\n # 遇到其他读取错误（如权限等），保守视作验证失败\\n return False\\n \\n # 只要发现了代码文件，且全部通过了UTF-8解码测试，即认为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 待校验的准确注释目标字符串\\n target_string = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据页面组件开发场景，可能包含的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.php', '.java', '.py', '.c', '.cpp', '.cs', '.go', '.rb'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 只针对代码文件进行校验，忽略诸如.log文件或没有后缀名的特殊文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 检查文件开头：读取前5行内容\\n # 留有前5行的余量是为了兼容类似于多行注释的第一行可能是 `<!--` 或 `/*` 等情况\\n lines = [f.readline() for _ in range(5)]\\n \\n found = False\\n for line in lines:\\n if target_string in line:\\n found = True\\n break\\n \\n # 只要有一个代码文件没在开头带有该注释，直接返回False\\n if not found:\\n return False\\n has_code_file = True\\n except Exception:\\n # 文件读取失败等情况，保守判定为未通过\\n return False\\n \\n # 如果项目中没有找到任何有效代码文件，自然也算作未完成该需求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n \\n workspace = Path(workspace_path)\\n \\n # 如果工作区不存在或不是一个目录，无法验证项目，视为不满足条件\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for root, dirs, files in os.walk(workspace):\\n # 过滤掉以 '.' 开头的隐藏目录和特殊的 '__pycache__' 缓存目录，防止误判\\n # 注意：使用切片赋值 dirs[:] 可以原地修改列表，从而阻止 os.walk 继续遍历这些被排除的目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = Path(root).relative_to(workspace)\\n \\n # rel_path.parts 的长度代表了当前目录的层级深度：\\n # - 长度为 0 表示当前目录是 workspace 根目录本身\\n # - 长度为 1 表示当前目录是一级目录（如 workspace/src）\\n # - 长度为 2 表示当前目录是二级目录（如 workspace/src/components）\\n # - 长度 > 2 表示存在三级或更深的目录，这违反了“层级最高不能超过2层”的要求\\n if len(rel_path.parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_anim 作为结尾\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}\\n \\n # 正则1：匹配常规声明 `function name(...)` 或 generator函数 `function* name(...)`\\n pattern_func_decl = re.compile(r'\\\\bfunction\\\\b\\\\s*\\\\*?\\\\s*([a-zA-Z_$]\\\\w*)\\\\s*\\\\(')\\n \\n # 正则2：匹配显式变量赋值定义的函数 `const/let/var name = function()` 或 `const name = () =>`\\n pattern_var_decl = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|[^;{=]*?=>)')\\n \\n # 正则3：匹配在类属性中独立声明的箭头函数形式 `name = () =>`（限制行首与修饰符，避免误伤JSX属性）\\n pattern_class_prop = re.compile(r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+|readonly\\\\s+)*([a-zA-Z_$]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|[^;{=]*?=>)', re.MULTILINE)\\n \\n # 正则4：匹配类方法/对象方法 `name() { ... }` \\n # 使用 \\\\([^()]*\\\\) 确保圆括号内不能再嵌套内部的左括号，防止误判了回调函数调用(如 setTimeout(function(){)\\n pattern_method = re.compile(r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+|async\\\\s+|get\\\\s+|set\\\\s+)*([a-zA-Z_$]\\\\w*)\\\\s*(?:<[^>]*>\\\\s*)?\\\\([^()]*\\\\)\\\\s*\\\\{', re.MULTILINE)\\n \\n # 语言关键字与系统字词，防方法声明误伤黑名单\\n keywords = {\\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'constructor', 'super', \\n 'function', 'import', 'export', 'default', 'new', 'await', \\n 'yield', 'typeof', 'instanceof', 'void', 'delete', 'in', 'do', 'class', 'get', 'set'\\n }\\n \\n has_target_files = False\\n funcs_found = []\\n \\n # 扫描生成的代码项目\\n for root, dirs, files in os.walk(workspace_path):\\n parts = root.split(os.sep)\\n # 排除可能包含海量第三方无关代码的目录\\n if 'node_modules' in parts or '.git' in parts or 'dist' in parts or 'build' in parts:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_target_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 搜集文件中所有定义的函数名\\n funcs_found.extend(pattern_func_decl.findall(content))\\n funcs_found.extend(pattern_var_decl.findall(content))\\n funcs_found.extend(pattern_class_prop.findall(content))\\n funcs_found.extend(pattern_method.findall(content))\\n except Exception:\\n continue\\n \\n # 没有可验证的前端代码文件时视作不满足\\n if not has_target_files:\\n return False\\n \\n valid_func_count = 0\\n for name in funcs_found:\\n if name in keywords:\\n continue\\n \\n valid_func_count += 1\\n \\n # 检查收集到的函数名是否严格落实了 \\\"_noti_anim\\\" 后缀\\n if not name.endswith('_noti_anim'):\\n return False\\n \\n # 组件代码中如果没有任何提取到的函数，意味着核心业务逻辑压根未落实，同样返回 False\\n if valid_func_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"def check_class_name_length_is_non_prime(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 正则表达式：匹配 `class` 关键字后面的类名（兼容JavaScript/TypeScript/Python等常见类名定义）\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)')\\n \\n # 根据场景定义可能存在的代码文件扩展名\\n valid_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', \\n '.vue', '.svelte', '.py', '.java', '.cpp', '.hpp', '.h', '.cs'\\n }\\n\\n # 遍历工作区目录下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 按照系统提示要求，生成的文件均使用 UTF-8 编码\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取出文件中所有的类名\\n classes = class_pattern.findall(content)\\n for cls_name in classes:\\n # 如果类名的长度是质数，则违背了“长度恰好是一个非质数”的要求\\n if is_prime(len(cls_name)):\\n return False\\n\\n # 若所有类名长度均为非质数，或未找到任何类名，均算作满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 判断数字是否为质数的辅助函数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n ** 0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 针对前端文件的简单变量声明正则表达式\\n js_var_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n # 针对 Python 文件的普通赋值操作正则表达式（排除双等号 == 的情况）\\n py_var_pattern = re.compile(r'\\\\b([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=[^=]')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n filepath = os.path.join(root, file)\\n \\n # 过滤常见可能包含变量的代码文件后缀名\\n if ext in ['.js', '.ts', '.html', '.jsx', '.tsx', '.vue', '.py']:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n variables = []\\n # 匹配前端代码或含 `<script>` 的 HTML 的变量\\n if ext in ['.js', '.ts', '.html', '.jsx', '.tsx', '.vue']:\\n variables.extend(js_var_pattern.findall(content))\\n # 匹配 Python 代码的变量\\n if ext == '.py':\\n variables.extend(py_var_pattern.findall(content))\\n \\n # 判断所提取到的各个变量名的长度是否均为质数\\n for var_name in variables:\\n if not is_prime(len(var_name)):\\n return False\\n except Exception:\\n # 避免个别文件由于非正常格式等因素干扰读取，直接跳过处理\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来丰富一下通知的样式。我希望支持四种状态：Success（绿色）、Error（红色）、Warning（黄色）、Info（蓝色）。请在页面中间增加几个对应的触发按钮方便我测试。每种状态的通知卡片要配上对应的左侧Icon图标，另外，在每张通知卡片的底部需要加一根极细的进度条，用来可视化那3秒的倒计时，进度条长度要在3秒内平滑地缩减到0。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_style 作为结尾；二是进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令中关于页面通知组件、DOM操作等内容，推断可能的代码文件后缀名\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试将二进制内容用 utf-8 解码，以验证文件是否采用 UTF-8 编码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦捕获到解码错误，说明存在非 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 遇到其他读取错误也认为未满足条件\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 目标注释字符串\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 针对前端项目推断的常见代码文件后缀\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果无法使用utf-8编码读取代码文件，视为失败\\n return False\\n \\n # 检查文件头部（前10行）是否包含该注释\\n found = False\\n for line in lines[:10]:\\n if target_comment in line:\\n found = True\\n break\\n \\n # 若某个代码文件开头未发现该注释，则不满足要求\\n if not found:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef check_directory_level(response: str, workspace_path: str) -> bool:\\n # 保证 workspace_path 存在\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，避免如 .git 等环境辅助目录影响层级判断\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 获取当前目录相对于项目根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 忽略根目录本身\\n if rel_path == '.':\\n continue\\n \\n # 将相对路径分解成各级目录的元组\\n parts = Path(rel_path).parts\\n \\n # 如果 parts 的长度等于 1，例如 parts == ('src',)，代表这是第2层目录\\n # 如果 parts 的长度大于等于 2，例如 parts == ('src', 'components')，代表出现了第3层及以上的目录\\n if len(parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_style 作为结尾\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方依赖等可能影响判断的文件夹\\n if 'node_modules' in root.split(os.sep):\\n continue\\n \\n for file in files:\\n # 排除压缩文件\\n if '.min.' in file:\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 去除单行和多行注释以避免误识别\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n \\n func_names = []\\n \\n # 匹配方式1: 显式具名函数声明 -> function funcName(...)\\n for m in re.finditer(r'\\\\bfunction\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\(', code):\\n func_names.append(m.group(1))\\n \\n # 匹配方式2: 赋值形式的函数 (箭头函数表达式、普通函数表达式或对象属性等)\\n pattern2 = r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*[:=]\\\\s*(?:async\\\\s+)?(?:function\\\\s*(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\\\s*\\\\(|(?:<[^>]*>\\\\s*)?\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][0-9a-zA-Z_$]*\\\\s*=>)'\\n for m in re.finditer(pattern2, code):\\n func_names.append(m.group(1))\\n \\n # 匹配方式3: 对象简写方法或类成员方法 -> funcName(...) {\\n pattern3 = r'^\\\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*(?:<[^>]*>\\\\s*)?\\\\([^)]*\\\\)\\\\s*\\\\{'\\n for m in re.finditer(pattern3, code, re.MULTILINE):\\n name = m.group(1)\\n # 过滤常见的控制流关键字与保留字\\n if name not in {'if', 'for', 'while', 'switch', 'catch', 'return', 'function', 'constructor', 'super', 'this', 'import', 'export'}:\\n func_names.append(name)\\n \\n # 检查提取到的所有函数名是否都符合指定后缀\\n for name in set(func_names):\\n if not name.endswith('_noti_style'):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断一个自然数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 选取涉及前端及常见编程语言的代码文件\\n valid_extensions = ('.js', '.jsx', '.ts', '.tsx', '.html', '.vue', '.py', '.java', '.cs', '.cpp')\\n if not file.endswith(valid_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取面向对象编程中的类名（匹配如 \\\"class NotificationComponent\\\" 或 \\\"export class Noti\\\" 等格式）\\n class_matches = re.findall(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)', content)\\n \\n for cls_name in class_matches:\\n # 题目要求长度必须恰好是一个\\\"非质数\\\"\\n # 如果是质数，则表示不满足用户要求，直接返回 False\\n if is_prime(len(cls_name)):\\n return False\\n \\n # 如果没找到类名，或者所有类名长度全为非质数，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def check_variables_length_prime(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n def get_vars_from_js(js_code: str):\\n vars_found = []\\n # 清除单行和多行注释以避免干扰匹配\\n js_code = re.sub(r'//.*', '', js_code)\\n js_code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', js_code)\\n \\n # 1. 匹配常规的 var, let, const 声明\\n for match in re.finditer(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', js_code):\\n vars_found.append(match.group(1))\\n \\n # 2. 匹配简易的数组解构赋值，例如 const [a, b] = [1, 2]\\n for match in re.finditer(r'\\\\b(?:var|let|const)\\\\s+\\\\[(.*?)\\\\]', js_code):\\n inner = match.group(1)\\n for item in inner.split(','):\\n item = item.strip()\\n if item:\\n # 过滤掉默认值部分\\n var_name = item.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n vars_found.append(var_name)\\n \\n # 3. 匹配简易的对象解构赋值，例如 const {a, b} = obj\\n for match in re.finditer(r'\\\\b(?:var|let|const)\\\\s+\\\\{(.*?)\\\\}', js_code):\\n inner = match.group(1)\\n for item in inner.split(','):\\n item = item.strip()\\n if item:\\n # 如果包含别名，例如 a: b，实际变量名为 b\\n if ':' in item:\\n item = item.split(':')[1].strip()\\n # 过滤掉可能存在的默认值\\n var_name = item.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n vars_found.append(var_name)\\n \\n return vars_found\\n\\n all_vars = set()\\n found_any_code = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\\n found_any_code = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 针对网页和组件文件，仅提取 script 标签中的代码部分\\n if file.endswith(('.html', '.htm', '.vue')):\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for script_content in scripts:\\n all_vars.update(get_vars_from_js(script_content))\\n else:\\n all_vars.update(get_vars_from_js(content))\\n except Exception:\\n continue\\n\\n # 如果没有找到任何可识别的代码文件，或者项目中无任何变量声明，则按无法通过验证处理\\n if not found_any_code or not all_vars:\\n return False\\n \\n # 逐一校验提取出的每个变量名长度是否均为质数\\n for var in all_vars:\\n if not is_prime(len(var)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在来处理多条通知并发的情况，这比较有挑战。如果我连续点击按钮多次，这些通知需要在右下角的容器里自下而上地堆叠排列。最关键的是，当一条新的通知插进来，或者中间某一条因为超时/被点击而消失时，剩下的通知必须要有一个平滑的上下位移过渡动画，让它们顺滑地挤占空缺位置，绝不能出现瞬间闪烁或硬切跳跃的情况。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_stack 作为结尾；二是确保最终生成的代码文件中完全避开for循环的使用。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目背景，定义前端常用的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取整个文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 出现 UnicodeDecodeError 表示文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略权限不足或其他非编码导致的读取错误\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常用代码文件扩展名白名单，排除 .log, .json 等非代码或不支持常规注释的文件\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.mjs', '.cjs'\\n }\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n found_comment = False\\n \\n # 检查每个代码文件开头的前10行（兼容文件首部可能有 DOCTYPE 声明或 shebang 等内容的情况）\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n if target_comment in line:\\n found_comment = True\\n break\\n \\n # 若发现有任何一个代码文件未包含指定的注释说明，则返回 False\\n if not found_comment:\\n return False\\n except Exception:\\n # 如果遇到文件异常无法打开读取的情况，视作验证失败\\n return False\\n \\n # 如果项目中至少存在一个代码文件，且每个代码文件都满足了要求，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef check_directory_level(response: str, workspace_path: str) -> bool:\\n workspace = Path(workspace_path).resolve()\\n \\n # 如果工作区不存在或不是一个目录，直接返回 False\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for root, dirs, files in os.walk(workspace):\\n root_path = Path(root).resolve()\\n \\n try:\\n rel_parts = root_path.relative_to(workspace).parts\\n except ValueError:\\n # 理论上 os.walk(workspace) 产出的 root 都会在 workspace 下，防御性捕获异常\\n continue\\n \\n # 排除隐藏目录带来的干扰，比如 .git 等环境或工具生成的目录\\n if any(part.startswith('.') for part in rel_parts):\\n continue\\n \\n # 如果相对路径的层级达到 2 (即意味着存在 workspace/dir/subdir 结构)\\n # 则说明目录层级达到了 3 层，超过了“最高不超过2层”的限制\\n if len(rel_parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_stack 作为结尾\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 设定需要检查的代码文件扩展名，排除系统或配置的无关文件\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.py', '.vue', '.svelte'}\\n exclude_dirs = {'node_modules', '.git', 'venv', 'env', '__pycache__', 'dist', 'build'}\\n \\n # 关键字列表，用于避免将JS控制流或者系统保留字误判为函数名\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'constructor', 'with', 'import', 'export', 'super'}\\n \\n # Python 的函数匹配正则\\n pattern_py = r'\\\\bdef\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\('\\n \\n # JS/TS 传统 function 函数名匹配正则\\n pattern_js_func = r'\\\\bfunction\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\('\\n \\n # JS/TS 变量形式的函数赋值匹配正则，例如 const myFunc = () => 或者 let myFunc = function()\\n pattern_js_arrow = r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)[^=;\\\\n]*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)'\\n \\n # JS/TS 对象属性形式的函数赋值正则，例如 myFunc: () => 或者 myFunc: function()\\n pattern_obj_func = r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)'\\n \\n # JS/TS 类的属性函数正则，例如 myFunc = () => \\n pattern_class_prop_func = r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+)?([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)'\\n \\n # JS/TS 类/对象的方法简写形式匹配正则，例如 myFunc() { ... }\\n pattern_method = r'^\\\\s*(?:async\\\\s+)?(?:(?:get|set)\\\\s+)?([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤排除的目录及隐藏目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 预处理：移除注释片段以防止在注释中的示例函数造成误判\\n if ext == '.py':\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'', '', content)\\n content = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"', '', content)\\n else:\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n if ext in {'.html', '.vue', '.svelte'}:\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n \\n # 按照语言对函数进行匹配校验\\n if ext == '.py':\\n for m in re.finditer(pattern_py, content):\\n func_name = m.group(1)\\n if not func_name.endswith('_noti_stack'):\\n return False\\n else:\\n # JavaScript / TypeScript / Frontend 的各种变体匹配\\n for m in re.finditer(pattern_js_func, content):\\n func_name = m.group(1)\\n if not func_name.endswith('_noti_stack'):\\n return False\\n \\n for m in re.finditer(pattern_js_arrow, content):\\n func_name = m.group(1)\\n if not func_name.endswith('_noti_stack'):\\n return False\\n \\n for m in re.finditer(pattern_obj_func, content):\\n func_name = m.group(1)\\n if not func_name.endswith('_noti_stack'):\\n return False\\n \\n for m in re.finditer(pattern_class_prop_func, content, re.MULTILINE):\\n func_name = m.group(1)\\n if func_name not in keywords:\\n if not func_name.endswith('_noti_stack'):\\n return False\\n \\n for m in re.finditer(pattern_method, content, re.MULTILINE):\\n func_name = m.group(1)\\n if func_name not in keywords:\\n if not func_name.endswith('_noti_stack'):\\n return False\\n except Exception:\\n # 遇到无法解码的文件跳过即可\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 正则表达式匹配面向对象编程语言中的类名（适用于 JS, TS, Python 等）\\n # 匹配模式：class关键字 + 至少一个空白字符 + 类名标识符\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)(?:\\\\s|\\\\(|{|:|$)')\\n\\n # 定义需要检查的常见代码文件扩展名\\n target_extensions = ('.js', '.ts', '.jsx', '.tsx', '.py', '.html', '.htm', '.vue', '.svelte')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖模块和版本控制目录，避免误判非AI生成的代码\\n if 'node_modules' in root or '.git' in root or '__pycache__' in root:\\n continue\\n \\n for file in files:\\n if file.endswith(target_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找当前文件内所有声明的类名\\n classes = class_pattern.findall(content)\\n \\n for cls_name in classes:\\n cls_len = len(cls_name)\\n # 如果发现任一类名长度是质数，则违背了“长度恰好是非质数”的要求\\n if is_prime(cls_len):\\n return False\\n \\n except Exception:\\n # 忽略个别无法读取或编码异常的文件\\n continue\\n\\n # 若所有类名长度均为非质数，或项目中未显式定义OOP类(不违规)，则判定为满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def verify_variable_name_lengths_are_prime(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 质数判断函数\\n def is_prime(n):\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 收集所有的前端代码文件\\n js_files = []\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.vue')):\\n js_files.append(os.path.join(root, file))\\n\\n if not js_files:\\n return False\\n \\n var_names = set()\\n for file in js_files:\\n with open(file, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 若是 HTML 或 Vue 文件，重点提取 <script> 标签内代码\\n if file.endswith('.html') or file.endswith('.vue'):\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n content = '\\\\n'.join(scripts)\\n \\n # 过滤掉多行和单行注释，防止非代码文本干扰提取\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n \\n # 1. 匹配简单的单变量声明 (例: let abc = 1; 或 const def = 2)\\n matches1 = re.findall(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n var_names.update(matches1)\\n \\n # 2. 匹配对象解构赋值变量 (例: const { a, b: c } = obj)\\n matches2 = re.findall(r'\\\\b(?:let|const|var)\\\\s*\\\\{([^}]+)\\\\}', content)\\n for m in matches2:\\n parts = m.split(',')\\n for p in parts:\\n p = p.strip()\\n if p:\\n if ':' in p:\\n name = p.split(':')[1].strip()\\n else:\\n name = p\\n # 滤除可能的默认值\\n name = name.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', name):\\n var_names.add(name)\\n \\n # 3. 匹配数组解构赋值变量 (例: let [x, , y=2] = arr)\\n matches3 = re.findall(r'\\\\b(?:let|const|var)\\\\s*\\\\[([^\\\\]]+)\\\\]', content)\\n for m in matches3:\\n parts = m.split(',')\\n for p in parts:\\n p = p.strip()\\n if p:\\n name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', name):\\n var_names.add(name)\\n \\n # 4. 匹配传统函数的形参 (例: function foo(arg1, arg2))\\n funcs = re.findall(r'function\\\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', content)\\n for f in funcs:\\n args = f.split(',')\\n for arg in args:\\n arg = arg.strip().split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', arg):\\n var_names.add(arg)\\n \\n # 5. 匹配多参数箭头函数的形参 (例: (arg1, arg2) => )\\n arrow_funcs = re.findall(r'\\\\(([^)]*)\\\\)\\\\s*=>', content)\\n for f in arrow_funcs:\\n args = f.split(',')\\n for arg in args:\\n arg = arg.strip().split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', arg):\\n var_names.add(arg)\\n \\n # 6. 匹配单参数箭头函数的形参 (例: arg => )\\n arrow_funcs2 = re.findall(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', content)\\n var_names.update(arrow_funcs2)\\n\\n # 只要提取不到任何变量，说明生成的内容并非典型的规范 JS 代码\\n if not var_names:\\n return False\\n \\n # 判断每一个提取到的变量名的长度是否均为质数\\n for var in var_names:\\n if not is_prime(len(var)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保最终生成的代码文件中完全避开for循环的使用\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能包含前端项目或者Python代码的文件后缀\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', \\n '.vue', '.svelte', \\n '.html', '.htm', \\n '.css', '.scss', '.sass',\\n '.py'\\n }\\n \\n def remove_comments_and_strings(code: str, ext: str) -> str:\\n # 正则表达式用于移除各类语言的字符串和注释\\n patterns = [\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", # 单引号字符串\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', # 双引号字符串\\n r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`\\\", # 反引号模板字符串\\n r\\\"/\\\\*.*?\\\\*/\\\", # 多行注释 /* ... */\\n r\\\"//.*?$\\\", # 单行注释 // ...\\n r\\\"<!--.*?-->\\\" # HTML 注释 <!-- ... -->\\n ]\\n \\n # 针对Python特有的注释和多行字符串形式\\n if ext == '.py':\\n patterns.extend([\\n r\\\"\\\\#.*?$\\\", # Python 单行注释\\n r'\\\"\\\"\\\".*?\\\"\\\"\\\"', # Python 双引号多行字符串\\n r\\\"'''.*?'''\\\" # Python 单引号多行字符串\\n ])\\n \\n regex = \\\"|\\\".join(patterns)\\n pattern = re.compile(regex, re.DOTALL | re.MULTILINE)\\n return pattern.sub('', code)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除注释和字符串，以便进行更安全的关键字匹配\\n clean_content = remove_comments_and_strings(content, ext)\\n \\n # 1. 检测 JS/TS/Svelte/HTML 原生 for 循环，包括 for(...), for...in, for...of, for await(...)\\n # 使用 (?<!\\\\.) 排除可能存在的方法调用，如 arr.for()\\n if re.search(r'(?<!\\\\.)\\\\bfor\\\\s*(?:await\\\\s*)?\\\\(', clean_content):\\n return False\\n \\n # 2. 检测 Vue 或 Angular 的模板内置 for 循环指令\\n if re.search(r'\\\\bv-for\\\\b|\\\\*ngFor\\\\b', clean_content):\\n return False\\n \\n # 3. 检测 Sass/SCSS 的 @for 循环\\n if re.search(r'@for\\\\b', clean_content):\\n return False\\n \\n # 4. 检测 Python 原生 for 循环\\n if ext == '.py' and re.search(r'\\\\bfor\\\\s+.*?\\\\s+in\\\\b', clean_content):\\n return False\\n \\n except Exception:\\n # 如果由于编码或其他问题读取失败，跳过该文件继续处理其他文件\\n continue\\n\\n # 如果所有代码文件都没检测出 for 循环相关语法，则验证通过\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"加一点交互细节，这个对用户体验很关键。现在3秒自动关闭有时来不及看清长文本内容，我希望当我的鼠标悬浮（Hover）在某条通知卡片上时，该通知的自动销毁倒计时要立刻暂停，底部的进度条动画也要随之停住；当我鼠标移开后，倒计时和进度条再接着从刚才暂停的地方继续往下走，直到消失。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_hover 作为结尾；二是你构建的代码文件中必须排除对while循环的使用。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端与常规代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.vue', '.json', '.py', \\n '.java', '.c', '.cpp', '.h', '.xml', '.txt'\\n }\\n \\n # 如果工作区目录不存在，可以直接视为失败\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅验证代码文件，忽略日志、备份（如.bak）或二进制文件等\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制形式读取文件内容\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试强制用UTF-8进行解码\\n raw_data.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 只要有任何一个代码文件无法使用UTF-8解码，便不满足要求\\n return False\\n except Exception:\\n # 遇到诸如权限等无法读取的异常情况，也视为无法满足验证\\n return False\\n \\n # 如果检查了所有代码文件都没发生 UnicodeDecodeError 异常，则说明全部满足UTF-8编码要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 目标要求包含的具体字符串\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 结合用户指令推断，包含常见的前端代码扩展名，加入部分通用后台扩展名增强兼容性\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.h'}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = []\\n # 仅读取前10行来验证是否在文件的“开头处”\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n lines.append(line)\\n \\n header_content = \\\"\\\".join(lines)\\n \\n # 如果没有在该代码文件的开头部分找到指定的注释文本，则不满足要求\\n if target_str not in header_content:\\n return False\\n except Exception:\\n # 对于非纯文本或编码非UTF-8等异常无法读取的文件，跳过处理\\n continue\\n \\n # 如果项目下完全没有找到任何代码文件，判定未完成生成代码的任务\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n try:\\n workspace = Path(workspace_path).resolve()\\n \\n # 确保工作区路径存在且是一个目录\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for root, dirs, files in os.walk(workspace):\\n # 原地修改 dirs 列表以过滤掉以 '.' 开头的隐藏目录（如 .git, .vscode 等）\\n # 这样 os.walk 就不会继续向下遍历这些隐藏目录\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 获取当前遍历到的目录相对于根目录的相对路径\\n rel_path = Path(root).relative_to(workspace)\\n \\n # rel_path.parts 会将相对路径按层级拆分为元组\\n # - 对于 workspace 本身，parts 为空元组 ()，长度 0 （属于第1层）\\n # - 对于 workspace/dir1，parts 为 ('dir1',)，长度 1 （属于第2层）\\n # - 对于 workspace/dir1/dir2，parts 为 ('dir1', 'dir2')，长度 2 （属于第3层）\\n # 要求最高不超过2层，因此长度达到2时即视为违规\\n if len(rel_path.parts) >= 2:\\n return False\\n \\n return True\\n except Exception:\\n return False\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_hover 作为结尾\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断所有生成的代码文件中的函数是否均以 _noti_hover 结尾。\\n \\\"\\\"\\\"\\n # 针对 JavaScript/TypeScript 常见函数定义写出匹配正则\\n js_patterns = [\\n # 1. 常规函数声明 function foo(...)\\n re.compile(r'\\\\bfunction\\\\s*(?:\\\\*\\\\s*)?([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\('),\\n # 2. 匿名函数赋值 const/let/var foo = function(...)\\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\b'),\\n # 3. 箭头函数赋值 const/let/var foo = (...) => ... 或者 foo = e => ...\\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][\\\\w$]*)\\\\s*=>'),\\n # 4. 对象方法 { foo: function(...) } 或 { foo: (...) => ... }\\n re.compile(r'([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|(?:\\\\([^)]*\\\\)|[a-zA-Z_$][\\\\w$]*)\\\\s*=>)'),\\n # 5. Class中的类方法声明（多行模式匹配开头）\\n re.compile(r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+)?(?:abstract\\\\s+)?(?:async\\\\s+)?(?:static\\\\s+)?(?:get\\\\s+|set\\\\s+)?([a-zA-Z_$][\\\\w$]*)\\\\s*(?:<[^>]+>)?\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^{]+)?\\\\{', re.MULTILINE)\\n ]\\n # 针对 Python 函数定义的正则\\n py_pattern = re.compile(r'\\\\bdef\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n\\n # 排除容易在正则匹配（尤其是第5个正则）中作为代码块或者关键字被误抓取的保留字\\n ignore_names = {\\n 'if', 'for', 'while', 'switch', 'catch', 'return', 'constructor', 'render', \\n 'super', 'get', 'set', 'try', 'finally', 'else', 'do', 'with', 'typeof', \\n 'instanceof', 'new', 'delete', 'await', 'yield', 'import', 'export', 'function'\\n }\\n \\n found_any_func = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue', '.svelte', '.py']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n code_blocks = []\\n # 如果是包含标签的前端页面或组件，抽离出script块中的代码再校验\\n if ext in ['.html', '.htm', '.vue', '.svelte']:\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n code_blocks.extend(scripts)\\n else:\\n code_blocks.append(content)\\n \\n for code in code_blocks:\\n # 移除单行及多行注释，避免注释内的伪函数或旧代码干扰\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n funcs = []\\n if ext == '.py':\\n funcs.extend(py_pattern.findall(code))\\n else:\\n for p in js_patterns:\\n funcs.extend(p.findall(code))\\n \\n for func in funcs:\\n func = func.strip()\\n # 忽略无效函数名或系统保留字\\n if not func or func in ignore_names:\\n continue\\n \\n found_any_func = True\\n # 只要存在不以 _noti_hover 结尾的函数，则直接认为不符合规范\\n if not func.endswith('_noti_hover'):\\n return False\\n\\n # 若搜寻到了至少一个函数，并且没有提前触发False抛出，说明满足题意；否则如果连函数都没有，也视为违反业务逻辑返回False\\n return True if found_any_func else False\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助判断素数，用于判断非质数（素数返回 True，非素数/非质数返回 False）\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 匹配面向对象中的类名，如 class NotificationManager\\n oop_class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\b')\\n # 匹配 HTML / JSX 中的 CSS 类名，如 class=\\\"toast\\\" 或 className='msg-card'\\n html_class_pattern = re.compile(r'\\\\b(?:class|className)\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']')\\n\\n # 根据项目背景推断常见的前端代码文件类型\\n valid_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n # 验证 OOP 类名\\n for match in oop_class_pattern.findall(content):\\n # 依照指令限制，因未明确说明\\\"字符数\\\"，必须使用 count_word 获取计数值\\n if is_prime(count_word(match)):\\n return False\\n\\n # 验证 HTML/JSX 中的 CSS 类名\\n for match in html_class_pattern.findall(content):\\n # CSS类名可能包含多个并用空格分隔\\n class_names = match.strip().split()\\n for c_name in class_names:\\n # 同样依照约束，必须使用 count_word 而不是 len\\n if is_prime(count_word(c_name)):\\n return False\\n except Exception:\\n # 忽略读取或编码过程中的异常文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def check_variable_lengths_are_prime(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n def extract_js_variables(code_text: str) -> set:\\n # 去除字符串字面量（避免字符串内部文本带来干扰）\\n code_text = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '\\\"\\\"', code_text)\\n code_text = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", \\\"''\\\", code_text)\\n code_text = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '``', code_text, flags=re.DOTALL)\\n # 去除注释\\n code_text = re.sub(r'//.*', '', code_text)\\n code_text = re.sub(r'/\\\\*.*?\\\\*/', '', code_text, flags=re.DOTALL)\\n \\n variables = []\\n # 提取 let / const / var 声明的直接标识符\\n pattern = r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b'\\n variables.extend(re.findall(pattern, code_text))\\n \\n # 提取常规 function 后面的参数\\n func_pattern = r'\\\\bfunction\\\\s+[a-zA-Z_$0-9]*\\\\s*\\\\((.*?)\\\\)'\\n for params_str in re.findall(func_pattern, code_text):\\n params = [p.strip() for p in params_str.split(',') if p.strip()]\\n for p in params:\\n p = p.split('=')[0].strip() # 处理可能存在的默认参数\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p):\\n variables.append(p)\\n \\n # 提取箭头函数的单参数模式 (例如: arg => ...)\\n arrow_pattern_1 = r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>'\\n variables.extend(re.findall(arrow_pattern_1, code_text))\\n \\n # 提取箭头函数的多参数模式 (例如: (arg1, arg2) => ...)\\n arrow_pattern_2 = r'\\\\((.*?)\\\\)\\\\s*=>'\\n for params_str in re.findall(arrow_pattern_2, code_text):\\n params = [p.strip() for p in params_str.split(',') if p.strip()]\\n for p in params:\\n p = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p):\\n variables.append(p)\\n \\n return set(variables)\\n\\n def extract_py_variables(code_text: str) -> set:\\n variables = set()\\n try:\\n tree = ast.parse(code_text)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n variables.add(node.id)\\n elif isinstance(node, ast.arg):\\n variables.add(node.arg)\\n except Exception:\\n pass\\n return variables\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_variables = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n vars_found = set()\\n if file.endswith('.py'):\\n vars_found = extract_py_variables(content)\\n elif file.endswith(('.html', '.htm')):\\n # 对于 HTML 文件，提取其中所有的 <script> 标签内容\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n vars_found.update(extract_js_variables(script))\\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n vars_found = extract_js_variables(content)\\n \\n if vars_found:\\n has_variables = True\\n for var_name in vars_found:\\n # 检查任何一个提取到的变量名，其长度是否为质数\\n if not is_prime(len(var_name)):\\n return False\\n except Exception:\\n continue\\n\\n # 只有当成功匹配到至少一个变量且所有变量长度均为质数时，才认定为符合条件\\n return has_variables\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保最终生成的代码文件中完全避开for循环的使用\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查生成的代码项目中是否完全避开了 for 循环的使用。\\n 返回 True 表示满足要求（没有 for 循环），False 表示存在 for 循环。\\n \\\"\\\"\\\"\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义正则用于去除字符串常量，以防止字符串中的 \\\"for\\\" 干扰\\n string_pattern1 = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"'\\n string_pattern2 = r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\"\\n string_pattern3 = r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`\\\"\\n \\n # 定义正则用于去除各类注释\\n html_comment = r\\\"<!--[\\\\s\\\\S]*?-->\\\"\\n comment_pattern_multi = r\\\"/\\\\*[\\\\s\\\\S]*?\\\\*/\\\"\\n comment_pattern_single = r\\\"//.*\\\"\\n py_comment_single = r\\\"#.*\\\"\\n \\n def clean_code(code: str) -> str:\\n # 先去除所有的字符串内容\\n code = re.sub(string_pattern1, \\\"\\\", code)\\n code = re.sub(string_pattern2, \\\"\\\", code)\\n code = re.sub(string_pattern3, \\\"\\\", code)\\n # 再去除所有的注释\\n code = re.sub(html_comment, \\\"\\\", code)\\n code = re.sub(comment_pattern_multi, \\\"\\\", code)\\n code = re.sub(comment_pattern_single, \\\"\\\", code)\\n code = re.sub(py_comment_single, \\\"\\\", code)\\n return code\\n\\n # 前端及常见的代码文件后缀\\n code_extensions = (\\n '.js', '.ts', '.jsx', '.tsx', \\n '.html', '.htm', '.vue', '.svelte', \\n '.py', '.css', '.scss', '.less'\\n )\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if not file.lower().endswith(code_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 文件无法读取则跳过\\n continue\\n \\n # 清理代码内容中的注释和字符串\\n cleaned = clean_code(content)\\n \\n # 1. 匹配常规 C-like 和 JS/TS 的 for 循环 (for, for...in, for...of, for await...of)\\n if re.search(r'\\\\bfor\\\\s*(?:await\\\\s*)?\\\\(', cleaned):\\n return False\\n \\n # 2. 匹配 Python 中的 for 循环 (for...in)\\n if file.endswith('.py') and re.search(r'\\\\bfor\\\\b\\\\s+.*?\\\\s+\\\\bin\\\\b', cleaned):\\n return False\\n \\n # 3. 匹配 Vue 和 Angular 等模板中的循环指令\\n if re.search(r'\\\\bv-for\\\\b|\\\\*ngFor\\\\b', cleaned):\\n return False\\n \\n # 4. 匹配 Jinja/Django 等模板引擎的 for 循环\\n if re.search(r'\\\\{%\\\\s*for\\\\b', cleaned):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你构建的代码文件中必须排除对while循环的使用\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义可能包含逻辑循环的代码文件扩展名\\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 针对HTML或Vue文件，为防止普通文本或DOM中恰好出现 \\\"while\\\"，仅截取 script 标签里的逻辑代码进行检查\\n if ext in {'.html', '.vue'}:\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n content = '\\\\n'.join(scripts)\\n \\n # 正则匹配规则：去除多行注释、双斜杠单行注释、井号单行注释，以及反引号、单引号、双引号包裹的字符串（支持内部转义）\\n pattern = re.compile(\\n r'/\\\\*.*?\\\\*/|//.*?$|#.*?$|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.DOTALL | re.MULTILINE\\n )\\n cleaned_content = re.sub(pattern, ' ', content)\\n \\n # 搜索是否存在独立关键字 while\\n if re.search(r'\\\\bwhile\\\\b', cleaned_content):\\n return False\\n except Exception:\\n # 若文件读取或解析存在异常，跳过继续验证其他文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"好，现在我们引入核心的'优先级'控制机制。请在页面上的触发区域加一个优先级下拉框选项（低、中、高）。同级别的通知按先后顺序正常排队。但如果我触发了一个'高优先级'的通知，它绝不能排在队列末尾，而是必须立刻强行插队到当前可见通知列表的最顶部（最新鲜、最显眼的位置），并且高优先级的通知进场时，需要带有一个引人注目的左右'震动（Shake）'特效。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_priority 作为结尾；二是对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 针对前端页面通知组件任务，枚举可能的代码文件后缀名\\n code_extensions = {\\n '.html', '.htm', \\n '.css', '.scss', '.less', \\n '.js', '.ts', \\n '.jsx', '.tsx', \\n '.vue', '.svelte', \\n '.json'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对推断为代码类型的文件进行 UTF-8 编码格式检查\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n # 使用二进制模式读取文件原始字节\\n with open(filepath, 'rb') as f:\\n raw_data = f.read()\\n # 测试是否可以合法被 utf-8 解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码出现错误，说明文件非有效 UTF-8 编码\\n return False\\n except Exception:\\n # 若因无权限等原因无法读取，默认视作验证失败\\n return False\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 针对前端组件常见代码文件扩展名进行推断与限制，排除日志等非代码文件\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".jsx\\\", \\\".ts\\\", \\\".tsx\\\", \\\".vue\\\"}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取前 10 行以检查文件开头处是否存在目标注释\\n lines = [f.readline() for _ in range(10)]\\n \\n # 检查是否有任意一行包含目标注释文本\\n found = any(target_comment in line for line in lines)\\n \\n if not found:\\n return False\\n except Exception:\\n # 如果读取遇到异常（比如编码错误或无法访问），则视为无法验证该文件，返回 False\\n return False\\n \\n # 如果找到了至少一个代码文件，并且所有代码文件都包含该注释，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"def check_directory_levels(response: str, workspace_path: str) -> bool:\\n import os\\n import pathlib\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，防止评测系统环境预置的目录（如 .git 等）影响层级判断\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n if rel_path == '.':\\n depth = 0\\n else:\\n # 获取相对路径包含的层级数\\n # 比如 \\\"src\\\" 的 parts 是 ('src',), 长度 1 (代表第 2 层目录)\\n # 比如 \\\"src/utils\\\" 的 parts 是 ('src', 'utils'), 长度 2 (代表第 3 层目录)\\n depth = len(pathlib.Path(rel_path).parts)\\n \\n # 相对深度为 2 意味着已经到达了第 3 层目录，不符合最高 2 层要求\\n if depth >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_priority 作为结尾\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_function_names(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n patterns_regex = [\\n # 1. 匹配标准函数声明：function funcName(...) 或 function *funcName(...)\\n r'\\\\bfunction\\\\s*(?:\\\\*\\\\s*)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:<[^>]*>)?\\\\s*\\\\(',\\n \\n # 2. 匹配 Python 函数声明：def funcName(...)\\n r'\\\\bdef\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(',\\n \\n # 3. 匹配箭头函数或匿名函数的变量赋值：const/let/var funcName = (...) => { ... } \\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:(?:<[^>]*>)?\\\\s*function\\\\b|(?:<[^>]*>)?\\\\s*\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)',\\n \\n # 4. 匹配类方法或对象方法：methodName(args) { ... }\\n # 排除各类控制流和语言保留关键字，防止将 if(...){, for(...){ 等错误当作函数捕获\\n r'\\\\b(?!(?:if|catch|for|while|switch|return|function|def|async|await|class|interface|type|new|super|this|import|export|with|typeof|instanceof|delete|void|yield)\\\\b)([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:<[^>]*>)?\\\\s*\\\\([^()]*\\\\)\\\\s*(?::[^{=;]*)?\\\\{'\\n ]\\n \\n found_functions = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n is_python = file.endswith('.py')\\n # 常见前端代码文件及 Python\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 预处理清洗内容：移除字符串字面量、各类型注释及 <style> 块，严防正则陷入死角或误匹配\\n if is_python:\\n sanitization_patterns = [\\n r'#[^\\\\n]*', # Python 单行注释\\n r'(?s:\\\"\\\"\\\".*?\\\"\\\"\\\")', # Python 三引号注释/字符串\\n r\\\"(?s:'''.*?''')\\\", # Python 三单引号注释/字符串\\n r'(?s:\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")', # 双引号字符串\\n r\\\"(?s:'(?:\\\\\\\\.|[^'\\\\\\\\])*')\\\" # 单引号字符串\\n ]\\n else:\\n sanitization_patterns = [\\n r'(?i)(?s:<style[^>]*>.*?</style>)', # HTML/Vue style 块（CSS）\\n r'(?s:<!--.*?-->)', # HTML 注释\\n r'(?s:/\\\\*.*?\\\\*/)', # JS/TS 块级多行注释\\n r'//[^\\\\n]*', # JS/TS 单行注释\\n r'(?s:\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")', # 双引号字符串\\n r\\\"(?s:'(?:\\\\\\\\.|[^'\\\\\\\\])*')\\\", # 单引号字符串\\n r'(?s:`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' # 模板字符串\\n ]\\n \\n combined_pattern = '|'.join(sanitization_patterns)\\n \\n def replacer(match):\\n s = match.group(0)\\n # 若匹配结果为字符串字面量，保留空引号避免代码结构黏连（如导致连续词法分析错误）\\n if s.startswith(('\\\"', \\\"'\\\", '`')):\\n if s.startswith('\\\"\\\"\\\"'): return '\\\"\\\"'\\n if s.startswith(\\\"'''\\\"): return \\\"''\\\"\\n return s[0] + s[0]\\n # 若为注释或样式块，则一律由空格取代\\n else:\\n return \\\" \\\"\\n \\n content = re.sub(combined_pattern, replacer, content)\\n \\n # 开始使用预定义正则全面提取函数名称\\n for pattern in patterns_regex:\\n for match in re.finditer(pattern, content):\\n func_name = match.group(1)\\n # 双重保险：过滤可能遗漏的硬性关键字\\n if func_name in ['if', 'catch', 'for', 'while', 'switch', 'return', 'function', 'def']:\\n continue\\n \\n found_functions = True\\n if not func_name.endswith('_noti_priority'):\\n return False\\n\\n return found_functions\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断一个自然数是否为质数\\n def is_prime(n):\\n if n <= 1:\\n return False\\n if n == 2 or n == 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 匹配各类主流编程语言中的类声明关键字及类名\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z0-9_$]+)')\\n \\n # 定义可能包含类定义的源代码文件扩展名\\n valid_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \\n '.vue', '.py', '.java', '.cs', '.cpp', '.hpp', '.h'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件读取失败则跳过该文件继续验证\\n continue\\n \\n # 粗略去除代码注释，防止注释中的示例代码干扰解析\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n \\n # 寻找文件中的所有类名\\n classes = class_pattern.findall(content)\\n for class_name in classes:\\n # 如果任一类名的长度是质数，则违背了“必须是一个非质数”的要求\\n if is_prime(len(class_name)):\\n return False\\n \\n # 所有文件验证完毕，均未发现长度为质数的类名，符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def verify_variables_length_prime(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 内部辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n if n in (2, 3):\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 正则表达式：用于匹配 JavaScript / TypeScript / HTML 等前端文件中 let, const, var 声明的变量\\n # 仅匹配单个基础标识符，避免复杂的解构赋值引发正则误判\\n js_var_pattern = re.compile(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n # 正则表达式：用于匹配 Python 代码中基础的行首变量赋值\\n py_var_pattern = re.compile(r'^[ \\\\t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \\\\t]*=[ \\\\t]*[^=]', re.MULTILINE)\\n\\n # 需要扫描的代码文件扩展名\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\\n\\n # 遍历代码项目目录\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n variables = []\\n \\n # 针对不同语言环境提取变量名\\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\\n variables.extend(js_var_pattern.findall(content))\\n if ext == '.py':\\n variables.extend(py_var_pattern.findall(content))\\n\\n # 验证每一个提取出的变量名长度是否均为质数\\n for var_name in variables:\\n if not is_prime(len(var_name)):\\n return False\\n \\n except Exception:\\n # 忽略无法正常以utf-8读取的文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保最终生成的代码文件中完全避开for循环的使用\",\n \"validation_code\": \"def check_for_loop_absence(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 预编译针对不同for循环的正则表达式\\n # 匹配 JavaScript / TypeScript 及类似C的for循环: for(...) 或 for await(...)\\n js_for_regex = re.compile(r'\\\\bfor\\\\s*(?:await\\\\s*)?\\\\(')\\n # 匹配 Python 的 for 循环: for x in obj:\\n py_for_regex = re.compile(r'\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s\\\\(\\\\)]+\\\\s+in\\\\b')\\n # 匹配 Vue 的循环语法: v-for=\\n vue_for_regex = re.compile(r'\\\\bv-for\\\\s*=')\\n # 匹配 Angular 的循环语法: *ngFor=\\n ng_for_regex = re.compile(r'\\\\*ngFor\\\\s*=')\\n # 匹配 传统模版引擎(如Jinja2/Django/Liquid等)的循环语法: {% for %}\\n tpl_for_regex = re.compile(r'\\\\{%\\\\s*for\\\\b')\\n\\n # 可能包含for循环的前端与脚本代码文件后缀\\n valid_extensions = {\\n \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".mjs\\\", \\\".cjs\\\", \\n \\\".vue\\\", \\\".html\\\", \\\".htm\\\", \\\".svelte\\\", \\\".py\\\"\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n # 移除代码内容中的注释，以防止注释中的 for 词汇产生误伤\\n # 1. 移除 JS/TS/Py 的单行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n # 2. 移除 JS/TS 的多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 3. 移除 HTML 的注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n\\n # 检查是否包含任何一种for循环\\n if (js_for_regex.search(content) or\\n py_for_regex.search(content) or\\n vue_for_regex.search(content) or\\n ng_for_regex.search(content) or\\n tpl_for_regex.search(content)):\\n return False # 发现了 for 循环，验证失败\\n\\n except Exception:\\n # 如果某些文件读取出错（如权限问题或编码非UTF-8），则直接跳过该文件继续验证\\n continue\\n\\n # 如果遍历检查完所有符合条件的文件都没有发现for循环，说明已完全避开for循环\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你构建的代码文件中必须排除对while循环的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能包含前端代码的文件类型\\n valid_extensions = {'.js', '.ts', '.html', '.jsx', '.tsx', '.vue', '.css', '.svelte'}\\n \\n # 构造正则表达式以匹配字符串和注释\\n # string 分组：正确处理双引号、单引号、反引号构成的字符串及内部可能的转义字符\\n # comment 分组：匹配 JS 的单行注释、多行注释，以及 HTML 的注释\\n pattern_str = (\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|'\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\"\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<comment>//[^\\\\n]*|/\\\\*.*?\\\\*/|<!--.*?-->)'\\n )\\n pattern = re.compile(pattern_str, re.DOTALL)\\n\\n # 替换逻辑：字符串被替换为空字符串字面量 \\\"\\\" 避免意外破坏后续语法结构，注释则直接被清空\\n def replacer(match):\\n if match.group('string'):\\n return '\\\"\\\"'\\n else:\\n return ''\\n\\n # 遍历项目目录下的所有相关代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 过滤掉内容中的字符串和注释文本，避免内容中的 \\\"while\\\" 文本引发误判\\n clean_content = pattern.sub(replacer, content)\\n \\n # 检查干净的代码中是否包含 while 循环语法特征\\n # 在 JS/TS 等前端语言中，while 循环一定符合 'while' 后紧跟可选空格和左括号 '(' 的格式\\n if re.search(r'\\\\bwhile\\\\s*\\\\(', clean_content):\\n return False\\n except Exception:\\n # 如果遇到文件无法读取的情况，跳过并继续处理其他文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符\",\n \"validation_code\": \"def check_code_line_width(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，当前任务属于Web前端相关的DOM/组件操作\\n # 因此设定常见的Web前端代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）和常见的依赖包目录（如 node_modules）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行末的换行符，并统计剩余字符的长度\\n if len(line.rstrip('\\\\r\\\\n')) > 80:\\n return False\\n except Exception:\\n # 遇到读取异常(例如非文本文件等)，直接跳过继续检查\\n continue\\n \\n # 如果项目中存在代码文件并且所有代码行的长度均未超过 80 个字符，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"现在的堆叠如果发了十几条屏幕就占满了。我们来做一个'堆叠折叠'模式。规定当屏幕上的通知超过3条时，第4条及以后的旧通知不再向下铺开，而是向后退缩，折叠成一叠'纸牌'的效果（后面的卡片层级降低，逐渐缩小、变暗、并被前面的部分遮挡）。同时在容器最下方显示一个'展开全部（共X条）'的文字按钮，点击后这些纸牌动画展开成普通的长列表，再点一次则重新折叠。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_fold 作为结尾；二是请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.vue', '.json'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名并统一转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅验证代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦触发 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如文件权限等其他非编码相关的异常\\n continue\\n\\n # 所有代码文件都能成功以 UTF-8 解码，或未发现代码文件\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据用户构建页面通知组件的要求，推断生成的前端代码文件可能的扩展名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \\n '.css', '.scss', '.less', '.vue', '.svelte'\\n }\\n \\n found_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和前端常见的第三方依赖/构建目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'dist', 'build')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n has_comment = False\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 检查文件开头部分（前20行），确认是否加上了要求的注释\\n for _ in range(20):\\n line = f.readline()\\n if not line:\\n break\\n if target_comment in line:\\n has_comment = True\\n break\\n except Exception:\\n # 若文件读取失败或编码非UTF-8，判定为未满足要求\\n return False\\n \\n # 如果任何一个代码文件开头没有该注释，则不满足要求\\n if not has_comment:\\n return False\\n\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"def verify_directory_level(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n \\n workspace = Path(workspace_path).resolve()\\n \\n # 若提供的工作区路径不存在或不是目录，则无法满足要求\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n # 递归遍历工作区内的所有文件和文件夹\\n for current_path in workspace.rglob('*'):\\n # 仅针对目录进行层级判断\\n if current_path.is_dir():\\n # 获取该目录相对于项目根目录（workspace）的相对路径\\n rel_parts = current_path.relative_to(workspace).parts\\n \\n # 若相对路径的分段数量 >= 2，说明出现了第3层及以上的目录\\n # (例如：workspace/dir1/dir2 -> rel_parts = ('dir1', 'dir2')，长度为 2)\\n if len(rel_parts) >= 2:\\n return False\\n \\n # 遍历结束均未发现层级越界情况，满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_fold 作为结尾\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 提取 Python 函数定义的正则\\n py_func_pattern = re.compile(r'^\\\\s*def\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(', re.MULTILINE)\\n \\n # 提取 JS/TS 普通函数声明的正则\\n js_func_pattern = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n \\n # 提取 JS/TS 赋值给变量的箭头函数或匿名函数的正则\\n js_var_func_pattern = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_]\\\\w*\\\\s*=>)')\\n \\n # 提取 JS/TS 类方法或对象方法的正则（带有开括号与花括号特征）\\n js_method_pattern = re.compile(r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+|async\\\\s+)*([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^{]+)?\\\\s*\\\\{', re.MULTILINE)\\n js_keywords = {'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', 'super', 'function', 'else'}\\n\\n # 排除不需要检查的系统库/第三方库目录\\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 如果当前路径属于被忽略的目录结构中，则跳过\\n if any(d in root.split(os.sep) for d in ignore_dirs):\\n continue\\n \\n for file in files:\\n # 仅处理常见的前后端代码文件\\n if not file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.py')):\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n functions = []\\n \\n # 区分 Python 和 前端JS类文件\\n if file.endswith('.py'):\\n functions.extend(py_func_pattern.findall(content))\\n else:\\n text_to_search = content\\n # 若为 HTML / Vue，则提取 script 标签内容做匹配\\n if file.endswith(('.html', '.vue')):\\n script_pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n scripts = script_pattern.findall(content)\\n text_to_search = '\\\\n'.join(scripts)\\n \\n # 匹配多种 JS/TS 函数形式\\n functions.extend(js_func_pattern.findall(text_to_search))\\n functions.extend(js_var_func_pattern.findall(text_to_search))\\n \\n # 匹配方法，并排除干扰关键字\\n methods = js_method_pattern.findall(text_to_search)\\n for m in methods:\\n if m not in js_keywords and not m.startswith('if') and not m.startswith('for') and not m.startswith('while') and not m.startswith('switch'):\\n functions.append(m)\\n \\n # 对所有搜集到的函数名进行验证\\n for func_name in functions:\\n if not func_name.endswith('_noti_fold'):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个自然数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 正则表达式：匹配 `class` 关键字后紧跟的类名标识符\\n # 适用于绝大多数主流语言 (JS, TS, Python, Java, C++, PHP 等)\\n pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)')\\n \\n # 前端开发及常见后端开发的代码文件后缀\\n valid_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.py', '.java', \\n '.cpp', '.h', '.hpp', '.cs', '.php', '.html', '.vue', '.svelte'\\n }\\n\\n # 遍历代码项目\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略读取失败的文件，或非UTF-8编码的文件\\n continue\\n \\n # 提取文件中所有的类名\\n class_names = pattern.findall(content)\\n for cls_name in class_names:\\n cls_len = len(cls_name)\\n # 如果某个类名的长度刚好是质数，则判定为不满足要求\\n if is_prime(cls_len):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个数字是否为质数\\n def is_prime(n):\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n all_vars = set()\\n # 过滤掉合法的 JS/TS 关键字，以防匹配误伤\\n keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'break', 'continue', 'return', \\n 'function', 'class', 'extends', 'super', 'new', 'this', 'var', 'let', 'const', 'try', \\n 'catch', 'finally', 'throw', 'typeof', 'instanceof', 'void', 'delete', 'in', 'of', \\n 'yield', 'await', 'async', 'import', 'export', 'default', 'true', 'false', 'null', \\n 'undefined', 'NaN', 'Infinity'\\n }\\n\\n # 一系列正则用于提取JS/TS变量与参数\\n var_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n param_pattern = re.compile(r'\\\\bfunction\\\\s*[a-zA-Z_$]*\\\\s*\\\\(([^)]*)\\\\)')\\n arrow_param_pattern = re.compile(r'(?:(?:const|let|var)\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=\\\\s*)?\\\\(([^)]*)\\\\)\\\\s*=>')\\n arrow_single_param_pattern = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>')\\n \\n def extract_from_js(code):\\n # 移除单行及多行注释，避免对注释内容误提取\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n # 1. var, let, const 单个声明\\n for match in var_pattern.findall(code):\\n all_vars.add(match)\\n \\n # 2. 对象解构\\n for match in re.findall(r'\\\\b(?:var|let|const)\\\\s*\\\\{([^}]+)\\\\}', code):\\n for p in match.split(','):\\n p_name = p\\n # 处理带有别名重命名的解构 {a: b}\\n if ':' in p:\\n p_name = p.split(':')[1]\\n p_name = p_name.split('=')[0].strip()\\n if p_name:\\n all_vars.add(p_name)\\n \\n # 3. 数组解构\\n for match in re.findall(r'\\\\b(?:var|let|const)\\\\s*\\\\[([^\\\\]]+)\\\\]', code):\\n for p in match.split(','):\\n p_name = p.split('=')[0].strip()\\n if p_name:\\n all_vars.add(p_name)\\n \\n # 4. 普通函数参数\\n for match in param_pattern.findall(code):\\n for p in match.split(','):\\n if p.strip():\\n all_vars.add(p.strip())\\n \\n # 5. 箭头函数参数 (多参数和无参数)\\n for match in arrow_param_pattern.findall(code):\\n for p in match.split(','):\\n if p.strip():\\n all_vars.add(p.strip())\\n \\n # 6. 箭头函数参数 (单参数)\\n for match in arrow_single_param_pattern.findall(code):\\n if match.strip():\\n all_vars.add(match.strip())\\n \\n # 7. catch 块异常变量\\n for match in re.findall(r'\\\\bcatch\\\\s*\\\\(\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\)', code):\\n all_vars.add(match)\\n \\n # 8. import 导入的默认变量\\n for match in re.findall(r'\\\\bimport\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s+from', code):\\n all_vars.add(match)\\n \\n # 9. import 导入的解构变量\\n for match in re.findall(r'\\\\bimport\\\\s*\\\\{([^}]+)\\\\}', code):\\n for p in match.split(','):\\n p_name = p\\n if ' as ' in p:\\n p_name = p.split(' as ')[1]\\n p_name = p_name.strip()\\n if p_name:\\n all_vars.add(p_name)\\n\\n def extract_from_py(code):\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Assign):\\n for target in node.targets:\\n if isinstance(target, ast.Name):\\n all_vars.add(target.id)\\n elif isinstance(node, ast.FunctionDef) or isinstance(node, ast.AsyncFunctionDef):\\n for arg in node.args.args:\\n all_vars.add(arg.arg)\\n if node.args.vararg:\\n all_vars.add(node.args.vararg.arg)\\n if node.args.kwarg:\\n all_vars.add(node.args.kwarg.arg)\\n except:\\n pass\\n\\n has_files = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except:\\n continue\\n \\n # 识别对应类型的文件并分别提取其中变量名\\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\\n has_files = True\\n extract_from_js(content)\\n elif ext in ['.html', '.htm', '.vue']:\\n has_files = True\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n extract_from_js(script)\\n elif ext == '.py':\\n has_files = True\\n extract_from_py(content)\\n \\n if not has_files:\\n return False\\n \\n valid_vars = set()\\n for var in all_vars:\\n # 清理TS类型注解及默认值，提取最核心的变量名称\\n var_name = var.split(':')[0].split('=')[0].strip()\\n # 清除变长参数的前缀 `...`\\n var_name = var_name.lstrip('.')\\n \\n # 确保只匹配到合法的标识符部分\\n match = re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*', var_name)\\n if match:\\n v = match.group(0)\\n if v not in keywords:\\n valid_vars.add(v)\\n\\n # 如未找到任何变量，项目结构逻辑可能存在异常或未包含脚本代码，故当做未完成约束处理\\n if not valid_vars:\\n return False\\n\\n # 检查所有被提取到的变量名的长度，是否恰好都等于质数\\n for var in valid_vars:\\n if not is_prime(len(var)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保最终生成的代码文件中完全避开for循环的使用\",\n \"validation_code\": \"def check_for_loops(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对不同语言环境定义 for 循环匹配正则\\n # 匹配 JS/TS/Java/C++ 等的 for 循环: for ( 或者 for await (，且排除作为属性调用的情况 (如 obj.for)\\n js_for_pattern = re.compile(r'(?<!\\\\.)\\\\bfor\\\\s*(?:await\\\\s+)?\\\\(')\\n # 匹配 Vue 的指令循环\\n vue_for_pattern = re.compile(r'\\\\bv-for\\\\b')\\n # 匹配 CSS/SCSS 的循环\\n css_for_pattern = re.compile(r'@for\\\\s+')\\n # 匹配 Python 的 for ... in 循环及推导式\\n py_for_pattern = re.compile(r'(?<!\\\\.)\\\\bfor\\\\s+.*?\\\\s+in\\\\b', re.DOTALL)\\n \\n # 去除字符串字面量的正则表达式，避免字符串中恰好存在关键字导致误判\\n str_double = re.compile(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"')\\n str_single = re.compile(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\")\\n str_backtick = re.compile(r\\\"`([^`\\\\\\\\]|\\\\\\\\.)*`\\\")\\n \\n # 去除各语言常见注释的正则表达式\\n comment_single = re.compile(r'//.*')\\n comment_multi = re.compile(r'/\\\\*.*?\\\\*/', re.DOTALL)\\n comment_html = re.compile(r'<!--.*?-->', re.DOTALL)\\n comment_py = re.compile(r'#.*')\\n \\n # 分类定义需要扫描的文件扩展名\\n js_exts = {'.js', '.ts', '.jsx', '.tsx'}\\n html_exts = {'.html', '.htm', '.vue'}\\n css_exts = {'.css', '.scss', '.less'}\\n py_exts = {'.py'}\\n other_exts = {'.java', '.c', '.cpp', '.cs'}\\n valid_exts = js_exts | html_exts | css_exts | py_exts | other_exts\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 第一步：先去除字符串内容，避免字符串里的注释或代码干扰\\n code = str_double.sub('\\\"\\\"', code)\\n code = str_single.sub(\\\"''\\\", code)\\n code = str_backtick.sub(\\\"``\\\", code)\\n \\n # 第二步：去除所有的注释文本\\n code = comment_single.sub('', code)\\n code = comment_multi.sub('', code)\\n code = comment_html.sub('', code)\\n code = comment_py.sub('', code)\\n \\n # 第三步：按照文件类型，执行对应的 for 循环安全检查\\n if ext in js_exts | html_exts | other_exts:\\n if js_for_pattern.search(code):\\n return False\\n \\n if ext in html_exts:\\n if vue_for_pattern.search(code):\\n return False\\n \\n if ext in css_exts | html_exts:\\n if css_for_pattern.search(code):\\n return False\\n \\n if ext in py_exts:\\n if py_for_pattern.search(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你构建的代码文件中必须排除对while循环的使用\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify_no_while_loop(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证代码项目中是否排除了对 while 循环的使用。\\n \\\"\\\"\\\"\\n \\n def clean_code(code: str) -> str:\\n # 正则匹配并移除注释和字符串，防止误判注释或字符串中的 'while' 单词\\n pattern = re.compile(\\n r'<!--.*?-->|' # HTML 注释\\n r'/\\\\*.*?\\\\*/|' # C/JS 多行注释\\n r'//.*?$|' # JS 单行注释\\n r'#.*?$|' # Python/Shell 单行注释\\n r'\\\"\\\"\\\".*?\\\"\\\"\\\"|' # Python 多行双引号字符串\\n r\\\"'''.*?'''|\\\" # Python 多行单引号字符串\\n r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"|' # 双引号字符串\\n r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'|\\\" # 单引号字符串\\n r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', # JS 模板字符串\\n re.DOTALL | re.MULTILINE\\n )\\n return pattern.sub('', code)\\n\\n # 常见的前端和可能涉及的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.svelte', '.py'}\\n\\n if not os.path.exists(workspace_path):\\n # 如果项目路径不存在，无法验证是否有代码文件，视作未违反\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n # 剔除代码中的注释和字符串\\n cleaned_code = clean_code(code)\\n \\n # 检查清理后的代码中是否存在完整的 'while' 关键字\\n if re.search(r'\\\\bwhile\\\\b', cleaned_code):\\n return False\\n except Exception:\\n # 忽略无法用 utf-8 读取的文件或其他文件读取异常\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对前端通知组件项目，定义可能存在的常见代码文件后缀\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h'\\n }\\n \\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅校验代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 指令中要求所有文件须为UTF-8编码，使用 errors='ignore' 避免偶然的特殊字符引发读取崩溃\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n # 剔除行尾的回车与换行符，保留行内的所有字符(包含空格)\\n line_content = line.rstrip('\\\\r\\\\n')\\n # 当某一行字符数超出80时，直接返回False\\n if len(line_content) > 80:\\n return False\\n except Exception:\\n # 若文件由于某种意外原因完全无法读取，则保守判定为失败\\n return False\\n \\n # 如果存在代码文件且均通过了行宽检查，则代表满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"我们再增加一项高级手势交互：支持拖拽移除。允许我用鼠标按住任意一条通知向右拖拽滑动。拖拽过程中，通知卡片要跟随鼠标横向移动，且透明度随着滑动距离的增加而逐渐降低；如果我拖拽超过卡片宽度的40%然后松手，卡片直接顺势飞出屏幕并触发销毁逻辑；如果拖动没超过40%就松手，卡片需要带有弹性回弹动画（Spring）恢复到原位。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_drag 作为结尾；二是对于你输出的每一个代码文件，请控制其连同空行在内的总行数是一个奇数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前端及后端代码文件扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.vue', '.json', '.py', '.java', '.c', '.cpp', '.h', \\n '.sh', '.md', '.txt', '.xml', '.yaml', '.yml'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs，避免进入隐藏目录（如 .git, .idea 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试用 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 只要有一个文件出现 UTF-8 解码错误，直接返回 False\\n return False\\n except Exception:\\n # 忽略因权限等导致的其他读取错误\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据用户指令推断，这是一个前端UI组件项目，定义常见的前端代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n code_file_count = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅针对代码文件进行检查，排除诸如 .log 这样的非代码文件\\n if ext in code_extensions:\\n code_file_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n found_comment = False\\n # 检查文件开头处（设定前5行内）是否包含目标注释字符串\\n for line_idx, line in enumerate(f):\\n if line_idx >= 5:\\n break\\n if target_comment in line:\\n found_comment = True\\n break\\n \\n if not found_comment:\\n return False\\n except Exception:\\n # 如果由于编码或其他问题导致文件无法正常读取，判定为失败\\n return False\\n \\n # 如果没有找到任何相关的代码文件，也视作未满足生成代码的检查前提\\n if code_file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n workspace = Path(workspace_path).resolve()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n current_dir = Path(root).resolve()\\n \\n try:\\n # 计算相对于项目根目录的相对路径\\n rel_path = current_dir.relative_to(workspace)\\n except ValueError:\\n continue\\n \\n # 忽略以 '.' 开头的隐藏目录及其子目录（如 .git, .vscode 等）\\n if any(part.startswith('.') for part in rel_path.parts):\\n continue\\n \\n # rel_path.parts 的长度代表了该目录的层级\\n # - 项目根目录本身 parts 为 ()，长度为 0\\n # - 第一层子目录 (如 src) parts 为 ('src',)，长度为 1\\n # - 第二层子目录 (如 src/components) parts 为 ('src', 'components')，长度为 2\\n # - 第三层及以上子目录会导致长度 > 2，这意味着超过了2层目录的限制\\n if len(rel_path.parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_drag 作为结尾\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对前端或后端页面通知组件可能用到的后缀名\\n exts = ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte', '.py']\\n \\n # 关键字黑名单：防止将流程控制、内置词或关键字错误识别为函数名\\n ignore_names = {\\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'render', \\n 'return', 'super', 'function', 'class', 'import', 'export', 'require',\\n 'setTimeout', 'setInterval', 'clearTimeout', 'clearInterval', 'Promise',\\n 'map', 'filter', 'reduce', 'forEach', 'then', 'catch', 'finally',\\n 'get', 'set', 'static', 'async', 'await', 'new', 'delete', 'typeof',\\n 'in', 'instanceof', 'void', 'with', 'yield', 'let', 'var', 'const'\\n }\\n \\n # 常见代码范式下的函数名提取正则\\n patterns = [\\n # JS/TS/HTML: function name(...) 或 async function name(...)\\n re.compile(r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('),\\n # JS/TS/HTML: const/let/var name = [async] function / (...) => / x =>\\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'),\\n # JS/TS class method: [static] [async] [get/set] name(...) {\\n re.compile(r'^\\\\s*(?:static\\\\s+)?(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', re.MULTILINE),\\n # JS/TS class property arrow function: name = [async] (...) =>\\n re.compile(r'^\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)', re.MULTILINE),\\n # Python: def name(...) 或 async def name(...)\\n re.compile(r'\\\\b(?:async\\\\s+)?def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n ]\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 避开通常不包含AI直接生成代码的常见第三方库或环境目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n if any(file.endswith(ext) for ext in exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 依序使用不同正则匹配并校验其中的函数名称\\n for pattern in patterns:\\n for match in pattern.finditer(content):\\n func_name = match.group(1)\\n # 过滤关键字或保留字\\n if func_name in ignore_names:\\n continue\\n # 检查后缀约束，不符合则直接返回 False\\n if not func_name.endswith('_noti_drag'):\\n return False\\n except Exception:\\n # 对于不可读或编解码等异常问题忽略，继续排查下一个文件\\n continue\\n \\n # 未发现任何违反要求的函数命名，返回 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断一个数字是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 匹配类定义的正则表达式：匹配 `class ` 后跟随的合法标识符作为类名\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_][A-Za-z0-9_]*)\\\\b')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 读取文件，若为非文本或编码不匹配则忽略\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 找到当前文件中的所有类名\\n matches = class_pattern.findall(content)\\n for class_name in matches:\\n # 如果存在类名长度为质数，则违背了“长度为非质数”的要求\\n if is_prime(len(class_name)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def verify_variable_length_is_prime(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n variables = set()\\n\\n # 用于匹配JS/TS等前端语言中通过 var/let/const 声明的变量\\n var_pattern = r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:=|;|:|,)'\\n # 用于匹配JS/TS等前端语言中基本的函数参数变量\\n func_pattern = r'\\\\bfunction\\\\s*(?:[a-zA-Z0-9_$]+)?\\\\s*\\\\(([^)]*)\\\\)'\\n\\n # 1. 扫描工作区文件提取变量\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n\\n if ext in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if ext == '.py':\\n # Python 文件使用 ast 精准提取变量\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n variables.add(node.id)\\n elif isinstance(node, ast.arg):\\n variables.add(node.arg)\\n except SyntaxError:\\n pass\\n else:\\n # 前端文件使用正则提取变量\\n for match in re.finditer(var_pattern, content):\\n variables.add(match.group(1))\\n \\n for match in re.finditer(func_pattern, content):\\n args_str = match.group(1)\\n if args_str.strip():\\n for arg in args_str.split(','):\\n m = re.search(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg.strip())\\n if m:\\n variables.add(m.group(0))\\n\\n # 2. 作为补充，扫描回复文本中的代码块以防文件未写入\\n code_blocks = re.findall(r'```(?:javascript|js|typescript|ts|html|vue)\\\\s*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n for match in re.finditer(var_pattern, block):\\n variables.add(match.group(1))\\n for match in re.finditer(func_pattern, block):\\n args_str = match.group(1)\\n if args_str.strip():\\n for arg in args_str.split(','):\\n m = re.search(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg.strip())\\n if m:\\n variables.add(m.group(0))\\n\\n # 如果没有找到任何变量，则无法判定其满足了要求，返回False\\n if not variables:\\n return False\\n\\n # 3. 校验所有提取出来的变量长度是否全为质数\\n for var in variables:\\n if not is_prime(len(var)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保最终生成的代码文件中完全避开for循环的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def strip_js_code(code):\\n # 移除单行注释、多行注释以及单/双/反引号字符串，避免匹配到其中的 for 单词字样\\n # 限制单/双引号匹配不能跨行，以避免因语法残缺漏写闭合引号而误吞大段代码\\n pattern = re.compile(\\n r'//[^\\\\n]*|' # 单行注释\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|' # 多行注释\\n r\\\"'(?:\\\\\\\\.|[^'\\\\n\\\\r])*'|\\\" # 单引号字符串\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\n\\\\r])*\\\"|' # 双引号字符串\\n r'`(?:\\\\\\\\.|[^`])*`' # 模板字符串（允许跨行）\\n )\\n return pattern.sub('', code)\\n\\n def strip_py_code(code):\\n # 移除 Python 格式的注释与字符串\\n pattern = re.compile(\\n r'#[^\\\\n]*|'\\n r\\\"'''[\\\\s\\\\S]*?'''|\\\"\\n r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|'\\n r\\\"'(?:\\\\\\\\.|[^'\\\\n\\\\r])*'|\\\"\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\n\\\\r])*\\\"'\\n )\\n return pattern.sub('', code)\\n\\n # 匹配 JavaScript / TypeScript 的 for 循环结构\\n # 匹配: for( 且紧跟循环体的合理起始特征：let, var, const, ;, {, [, +, - 或合法的变量名配合后续的操作符(如 =, in, of 等)\\n js_for_pattern = re.compile(\\n r'\\\\bfor\\\\s*(?:await\\\\s*)?\\\\(\\\\s*(?:'\\n r'let\\\\b|var\\\\b|const\\\\b|;|\\\\{|\\\\[|[+\\\\-]|'\\n r'[A-Za-z_$][\\\\w$]*\\\\s*(?:[=+\\\\-*/%&|^<>!(.\\\\[]|\\\\b(?:in|of)\\\\b)'\\n r')'\\n )\\n \\n # 匹配 Python 的 for 循环结构\\n py_for_pattern = re.compile(\\n r'\\\\bfor\\\\s+(?:[A-Za-z_]\\\\w*(?:\\\\s*,\\\\s*[A-Za-z_]\\\\w*)*)\\\\s+in\\\\s+'\\n )\\n \\n # 匹配各类常见前端模板中的 for 循环指令\\n template_for_pattern = re.compile(\\n r'\\\\bv-for\\\\s*=|' # Vue v-for\\n r'\\\\bngFor\\\\b|' # Angular (*ngFor)\\n r'\\\\{%\\\\s*for\\\\s+' # Jinja / Django 等模板语法\\n )\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.html', '.py', '.scss', '.sass']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 检测前端模板中的特有循环指令 (保留原文检测，因为指令通常以字符串格式作为属性值)\\n if template_for_pattern.search(content):\\n return False\\n \\n # 检测 SCSS/SASS 中的 @for 循环\\n if ext in ['.scss', '.sass']:\\n if re.search(r'@for\\\\s+', content):\\n return False\\n continue\\n\\n # 检测 Python 中的 for 循环\\n if ext == '.py':\\n stripped = strip_py_code(content)\\n if py_for_pattern.search(stripped):\\n return False\\n continue\\n\\n # 提取 HTML, Vue, Svelte 中的 <script> 区域内容\\n # 防止 <template> 等非脚本区域的普通文本内容对后面纯 JS 正则校验造成干扰\\n if ext in ['.html', '.vue', '.svelte']:\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n script_content = '\\\\n'.join(scripts)\\n else:\\n script_content = content\\n\\n # 处理剥离注释及纯文本字符串后的 JS/TS 代码排查 for 循环关键字结构\\n if script_content.strip():\\n stripped_script = strip_js_code(script_content)\\n if js_for_pattern.search(stripped_script):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你构建的代码文件中必须排除对while循环的使用\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify_no_while_loop(response: str, workspace_path: str) -> bool:\\n # 定义前端项目中及常见后端可能用到的代码文件后缀\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.css', '.py', '.java', '.c', '.cpp', '.cs'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n in_single_comment = False\\n in_multi_comment = False\\n in_string = False\\n string_char = ''\\n i = 0\\n n = len(content)\\n clean_code_chars = []\\n \\n # 使用简易状态机去除注释与字符串的干扰\\n while i < n:\\n c = content[i]\\n \\n # 处于单行注释中\\n if in_single_comment:\\n if c == '\\\\n':\\n in_single_comment = False\\n clean_code_chars.append('\\\\n')\\n i += 1\\n continue\\n \\n # 处于多行注释中\\n if in_multi_comment:\\n if c == '*' and i + 1 < n and content[i+1] == '/':\\n in_multi_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n # 处于字符串之中\\n if in_string:\\n if c == '\\\\\\\\': # 遇到转义字符直接跳过\\n if i + 1 < n:\\n i += 2\\n else:\\n i += 1\\n continue\\n if c == string_char: # 匹配到字符串结尾引号\\n in_string = False\\n clean_code_chars.append(c)\\n i += 1\\n continue\\n \\n # 检测字符串开头\\n if c in (\\\"'\\\", '\\\"', '`'):\\n in_string = True\\n string_char = c\\n clean_code_chars.append(c)\\n i += 1\\n continue\\n \\n # 检测 JS/TS/Java/C 等语言单行注释的开头 (//)\\n if c == '/' and i + 1 < n and content[i+1] == '/':\\n in_single_comment = True\\n i += 2\\n continue\\n \\n # 检测 JS/TS/CSS 等语言多行注释的开头 (/*)\\n if c == '/' and i + 1 < n and content[i+1] == '*':\\n in_multi_comment = True\\n i += 2\\n continue\\n \\n # 检测 HTML/Vue 的注释开头 (<!--)\\n if c == '<' and content[i:i+4] == '<!--':\\n end_idx = content.find('-->', i + 4)\\n if end_idx != -1:\\n i = end_idx + 3\\n else:\\n i = n\\n continue\\n \\n # 检测 Python/Ruby/Shell 等的单行注释 (#)\\n if ext in {'.py', '.rb', '.sh'} and c == '#':\\n in_single_comment = True\\n i += 1\\n continue\\n \\n # 如果不在任何注释或字符串中，加入清洗后的干净字符集合\\n clean_code_chars.append(c)\\n i += 1\\n \\n clean_str = \\\"\\\".join(clean_code_chars)\\n \\n # 使用 \\\\b 确保匹配到的是严格作为单一关键字的 while\\n if re.search(r'\\\\bwhile\\\\b', clean_str):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令（开发页面通知组件）推断可能的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.vue', '.scss', '.less'\\n }\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为推断的代码文件类型\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 仅去除行末尾的换行符，以便准确计算这一行的实际行宽\\n line_content = line.rstrip('\\\\r\\\\n')\\n \\n # 要求中明确说明“字符”，故直接使用len函数进行统计\\n if len(line_content) > 80:\\n return False\\n except Exception:\\n # 若文件由于某种异常（如非文本格式、编码问题）无法读取，则安全跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请控制其连同空行在内的总行数是一个奇数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.go', '.rs', '.php', '.rb'\\n }\\n \\n has_verified_any = False\\n \\n # 1. 优先检查 workspace_path 下生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n # 统计包括空行在内的总行数，如果是偶数则不符合要求\\n if len(lines) % 2 == 0:\\n return False\\n has_verified_any = True\\n except Exception:\\n # 忽略无法以 utf-8 读取或者出现其他异常的文件\\n continue\\n \\n # 2. 如果 workspace_path 中没有检测到任何代码文件，则回退检查 response 中的代码块\\n if not has_verified_any:\\n # 使用正则表达式提取 Markdown 格式的代码块\\n code_blocks = re.findall(r'```[a-zA-Z]*\\\\n(.*?)```', response, re.DOTALL)\\n if not code_blocks:\\n # 既没有检测到代码文件也没有发现代码块，无法证实满足了条件\\n return False\\n \\n for block in code_blocks:\\n # 兼容正则匹配时末尾自带的单一换行符，避免 split('\\\\n') 时多出最后一行空行\\n # 这旨在使统计逻辑与 f.readlines() 的正常表现对齐\\n if block.endswith('\\\\n'):\\n block = block[:-1]\\n \\n lines = block.split('\\\\n')\\n \\n if len(lines) % 2 == 0:\\n return False\\n has_verified_any = True\\n \\n return has_verified_any\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"最后一步，完善极限情况下的等待队列管理。限制DOM树中最多只允许渲染5条通知节点（无论是展开还是折叠状态）。如果在短时间内触发了超过5个，多出来的不能直接渲染，而是要进入内存里的'后台等待队列'。一旦当前可见的某条通知被销毁腾出了名额，等待队列里的通知要立刻根据优先级（高优优先，同级先进先出）自动被调度进场。请在页面上加个统计面板，让我能实时看到'当前显示: M条，后台排队: N条'来验证这个逻辑。另外请你后续轮次始终遵循以下两条规则：一是对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_queue 作为结尾；二是对于你生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 依据用户指令中构建页面通知组件等要求，推断可能的前端代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n # 遍历当前代码项目工作区\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 utf-8 编码完整读取并解码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 抛出 UnicodeDecodeError 说明文件并非合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 若出现其他无法读取的情况，保守起见视作不满足要求\\n return False\\n\\n # 需确保项目内确实生成了代码文件且全部验证通过，才判定为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请在每个文件的开头处新起一行，加上清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_string = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 针对前端或常见项目生成的代码文件类型\\n valid_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs'\\n }\\n \\n # 优先检查工作区生成的物理文件\\n if workspace_path and os.path.exists(workspace_path):\\n has_code_files = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in valid_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = []\\n # 只取前10行来寻找是否在文件“开头处”包含该声明注释\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n lines.append(line)\\n \\n found = any(target_string in line for line in lines)\\n # 一旦有代码文件前10行内没有规定注释，则判为未满足\\n if not found:\\n return False \\n except Exception:\\n return False\\n \\n # 若检查了所有的物理代码文件都没问题，则返回通过\\n if has_code_files:\\n return True\\n\\n # 兜底方案：如果工作区路径失效或无物理文件，则检查最后一轮回复当中的代码块\\n code_blocks = re.findall(r'```[\\\\w]*\\\\n(.*?)```', response, re.DOTALL)\\n if not code_blocks:\\n return False\\n \\n has_valid_blocks = False\\n for block in code_blocks:\\n lines = block.strip().split('\\\\n')\\n if not lines:\\n continue\\n has_valid_blocks = True\\n found = any(target_string in line for line in lines[:10])\\n if not found:\\n return False\\n \\n return has_valid_blocks\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"切记，针对已有文件的修改，不得直接输出全量代码替换，必须输出严格的Unified Diff格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过2层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉以 '.' 开头的隐藏目录（如 .git, .vscode 等），这些不计入项目真实层级\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n\\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n\\n # 根目录 '.' 为第 1 层\\n if rel_path == '.':\\n depth = 1\\n else:\\n # 通过系统路径分隔符拆分相对路径，计算所在层级\\n # 例如相对路径 'folder1' -> ['folder1']，深度为 1 + 1 = 2\\n # 相对路径 'folder1/folder2' -> ['folder1', 'folder2']，深度为 2 + 1 = 3\\n depth = len(rel_path.split(os.sep)) + 1\\n\\n # 若发现任何目录深度超过 2 层，则判定为不符合要求\\n if depth > 2:\\n return False\\n\\n # 遍历结束未发现超过2层的目录，则符合要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次对项目做出更改后，需通过新建文件 change_log_X.log 的方式来记录这次变动，其中 X 为改动次数的编号，起点为 0\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对生成的代码文件有一项命名限制：所有函数的名字必须以 _noti_queue 作为结尾\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 涵盖不同语言（JS/TS, Python等）常见的函数定义模式\\n patterns = [\\n # Python: def foo()\\n r'\\\\bdef\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(',\\n \\n # JS/TS: function foo()\\n r'\\\\bfunction\\\\s*\\\\*?\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\(',\\n \\n # JS/TS: const/let/var foo = function()\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\b',\\n \\n # JS/TS: const/let/var foo = () =>\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][\\\\w$]*)\\\\s*=>',\\n \\n # JS/TS 对象属性函数: foo: function()\\n r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?function\\\\b',\\n \\n # JS/TS 对象属性箭头函数: foo: () =>\\n r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][\\\\w$]*)\\\\s*=>',\\n \\n # JS/TS 类或对象的简写方法: foo() {\\n r'^\\\\s*(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+)?([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{',\\n \\n # JS/TS 类属性箭头函数: foo = () =>\\n r'^\\\\s*([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][\\\\w$]*)\\\\s*=>'\\n ]\\n \\n # 语言关键字黑名单（为了避免大括号形式的控制流语句被误判为函数名）\\n reserved_words = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', \\n 'function', 'constructor', 'return', 'super', 'with', 'typeof',\\n 'instanceof', 'new', 'delete', 'yield', 'await', 'get', 'set'\\n }\\n\\n # 常见代码文件格式后缀\\n extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\\n \\n found_functions = []\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_files = False\\n # 遍历当前工作空间内的所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in extensions:\\n has_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过各正则模式提取该文件内的函数名\\n for pattern in patterns:\\n matches = re.finditer(pattern, content, re.MULTILINE)\\n for match in matches:\\n func_name = match.group(1)\\n if func_name not in reserved_words:\\n found_functions.append(func_name)\\n except Exception:\\n pass\\n\\n # 至少要解析到代码文件才能判定通过\\n if not has_files:\\n return False\\n \\n if not found_functions:\\n return True \\n \\n # 遍历所有的函数名，验证其是否满足以 `_noti_queue` 结尾\\n for func in found_functions:\\n if not func.endswith('_noti_queue'):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你给出的代码文件当中，每个类名的长度值必须恰好是一个非质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断给定正整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 预编译常见的类名正则匹配器\\n # 1. 面向对象的类声明 (如: class MyClass)\\n decl_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)')\\n # 2. HTML / JSX 中的类属性 (如: class=\\\"my-class container\\\")\\n attr_pattern = re.compile(r'\\\\bclass(?:Name)?\\\\s*=\\\\s*[\\\\'\\\"]([^\\\\'\\\"]+)[\\\\'\\\"]')\\n # 3. CSS 规则中的类选择器 (如: .my-class)\\n css_pattern = re.compile(r'\\\\.([a-zA-Z_][a-zA-Z0-9_-]*)')\\n # 4. 匹配 HTML 中的 <style> 内容\\n style_tag_pattern = re.compile(r'<style[^>]*>(.*?)</style>', re.DOTALL | re.IGNORECASE)\\n\\n # 包含通常会出现类名的相关代码文件类型\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.svelte', '.py', '.java'}\\n \\n found_any_class = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 1. 检查代码的类定义声明\\n for match in decl_pattern.finditer(content):\\n cls_name = match.group(1)\\n found_any_class = True\\n if is_prime(len(cls_name)):\\n return False\\n \\n # 2. 检查 HTML / JSX 标签的类名属性\\n for match in attr_pattern.finditer(content):\\n classes_str = match.group(1)\\n for cls_name in classes_str.split():\\n cls_name = cls_name.strip()\\n if cls_name:\\n found_any_class = True\\n if is_prime(len(cls_name)):\\n return False\\n \\n # 3. 检查 CSS 样式表及其样式块中的类选择器\\n if ext == '.css':\\n for match in css_pattern.finditer(content):\\n cls_name = match.group(1)\\n found_any_class = True\\n if is_prime(len(cls_name)):\\n return False\\n elif ext in {'.html', '.vue', '.svelte'}:\\n for style_match in style_tag_pattern.finditer(content):\\n style_content = style_match.group(1)\\n for match in css_pattern.finditer(style_content):\\n cls_name = match.group(1)\\n found_any_class = True\\n if is_prime(len(cls_name)):\\n return False\\n\\n # 若检测到了至少一个类名，且没有触发由于质数长度抛出的 False，则证明所有类名均符合规则\\n return found_any_class\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请使得生成的代码文件里包含的所有变量名，其长度都恰好等于质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"判断一个正整数是否为质数\\\"\\\"\\\"\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 排除隐藏文件\\n if file_name.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, file_name)\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 针对前端或页面代码文件\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\\n # 利用正则匹配 JavaScript / TypeScript 中通过 let, const, var 声明的变量名\\n # 这里主要抓取声明语句紧跟的第一个标识符作为变量名代表\\n matches = re.findall(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', content)\\n for var_name in matches:\\n if not is_prime(len(var_name)):\\n return False\\n \\n # 针对 Python 代码文件（尽管主要需求是前端页面，但也可能包含 Python 提供服务）\\n elif ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 检查赋值产生的变量\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if not is_prime(len(node.id)):\\n return False\\n # 检查函数参数变量\\n elif isinstance(node, ast.arg):\\n if not is_prime(len(node.arg)):\\n return False\\n except SyntaxError:\\n # 忽略存在语法错误的 Python 文件\\n pass\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保最终生成的代码文件中完全避开for循环的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义常见的前端及后端源代码文件扩展名\\n extensions = {\\n '.js', '.ts', '.jsx', '.tsx', \\n '.vue', '.svelte', '.html', \\n '.py', '.java', '.c', '.cpp', \\n '.cs', '.go', '.rb', '.php', \\n '.swift', '.kt', '.mjs', '.cjs'\\n }\\n \\n # 构建匹配多种语言环境下 for 循环语法的正则表达式\\n # 1. (?<![\\\\.\\\\?])\\\\bfor\\\\s*\\\\( 匹配 JS, TS, Java, C++, C# 等的 for 循环，同时排除类似 Symbol.for() 这样的对象方法调用\\n # 2. (?<![\\\\.\\\\?])\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s]+\\\\s+in\\\\b 匹配 Python 等语言的 for ... in ... 循环结构\\n # 3. \\\\bv-for\\\\s*= 匹配 Vue 框架中的列表渲染指令\\n # 4. \\\\*ngFor\\\\s*= 匹配 Angular 框架中的列表渲染指令\\n # 5. (?<![\\\\.\\\\?])\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s]+\\\\s*:= 匹配 Go 语言的 for 循环推导\\n # 6. (?<![\\\\.\\\\?])\\\\bfor\\\\s*\\\\{ 匹配 Go 语言的 for 无限循环\\n for_pattern = re.compile(\\n r'(?<![\\\\.\\\\?])\\\\bfor\\\\s*\\\\(|' \\n r'(?<![\\\\.\\\\?])\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s]+\\\\s+in\\\\b|' \\n r'\\\\bv-for\\\\s*=|' \\n r'\\\\*ngFor\\\\s*=|' \\n r'(?<![\\\\.\\\\?])\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s]+\\\\s*:=|' \\n r'(?<![\\\\.\\\\?])\\\\bfor\\\\s*\\\\{' \\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除字符串字面量（避免字符串内存在 'for(' 等字眼干扰判断）\\n # 处理常见的双引号和单引号，允许内部包含转义字符\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content)\\n # 处理反引号模板字符串（可能跨行使用）\\n content = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', '``', content, flags=re.DOTALL)\\n \\n # 移除各种语言的注释内容（单行与多行注释），避免注释中的内容引发误判\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 如果在清理后的代码正文中检测到了任何 for 循环语句结构，说明未避开 for 循环的使用\\n if for_pattern.search(content):\\n return False\\n except Exception:\\n # 忽略读取异常的文件，继续检查下一个文件\\n continue\\n \\n # 所有的代码文件均未检测到 for 循环，满足用户指令的要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你构建的代码文件中必须排除对while循环的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义各类字符串字面量的正则表达式（处理转义字符的情况）\\n str_double = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"'\\n str_single = r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\"\\n str_backtick = r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`\\\"\\n \\n # 定义前端各类注释的正则表达式\\n comment_html = r'<!--[\\\\s\\\\S]*?-->'\\n comment_multi = r'/\\\\*[\\\\s\\\\S]*?\\\\*/'\\n comment_single = r'//.*'\\n\\n # 针对 JS/TS/HTML/CSS 等前端文件的清理规则（剔除字符串和常规前端注释）\\n strip_pattern_js = re.compile('|'.join([\\n str_double, str_single, str_backtick,\\n comment_html, comment_multi, comment_single\\n ]))\\n\\n # 针对 Python 文件的清理规则（剔除字符串和 # 注释）\\n comment_py = r'#.*'\\n strip_pattern_py = re.compile('|'.join([\\n str_double, str_single, str_backtick,\\n comment_py\\n ]))\\n\\n # 待检查的常见代码文件后缀类型\\n code_extensions = {'.html', '.htm', '.js', '.ts', '.jsx', '.tsx', '.vue', '.css', '.scss', '.py'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 根据文件类型应用不同的文本清理策略\\n if ext == '.py':\\n clean_content = strip_pattern_py.sub('', content)\\n else:\\n clean_content = strip_pattern_js.sub('', content)\\n \\n # 查找 while 关键字，利用 (?<!\\\\.) 排除作为对象属性的情形（如 object.while）\\n if re.search(r'(?<!\\\\.)\\\\bwhile\\\\b', clean_content):\\n return False\\n except Exception:\\n # 忽略存在读取权限或编码错误问题的文件\\n continue\\n\\n # 遍历结束无任何 while 关键字检出，视为满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，请控制行宽，使得每一行不超过 80 个字符\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令中对“DOM树、页面通知组件”的描述，推断可能的前端代码文件类型\\n target_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历工作区目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去除行末的换行符和回车符以严格计算行内的真实字符数\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 要求明确说明统计“字符数”，因此遵循提示使用len()函数统计\\n if len(clean_line) > 80:\\n return False\\n except Exception:\\n # 若因编码问题等无法读取文件流，则保守返回 False\\n return False\\n \\n # 只有在找到了前端代码文件，并且没有一行的字符数大于80的情况下，才视为满足该要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请控制其连同空行在内的总行数是一个奇数\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件后缀名集合\\n code_extensions = {\\n '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rb', '.php', '.swift', '.kt'\\n }\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历工作区内所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符切分获取文本编辑器意义上的总行数（连同空行）\\n lines = content.split('\\\\n')\\n total_lines = len(lines)\\n \\n # 检查是否为奇数\\n if total_lines % 2 == 0:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取文件，默认未能满足验证\\n return False\\n \\n # 如果有代码文件并且全都满足奇数行条件，则返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目及其他常见代码文件的扩展名集合\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', '.scss', '.less',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的首行，并转换为小写以作宽容匹配\\n first_line = f.readline().strip().lower()\\n # 判断首行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法以 UTF-8 编码读取或发生其他读取异常，视为不满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，同样视为未达到要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查基础渲染与状态样式：页面中间包含发通知相关的触发按钮及优先级下拉框。分别点击Success、Error、Warning、Info测试按钮，观察右下角是否弹出带有左侧对应图标的卡片，且卡片颜色/样式（绿、红、黄、蓝）分别对应。","检查通知进出场动画与进度条：点击触发一条通知，观察通知必须从右侧平滑滑入并淡入；注意卡片底部需有极细进度条平滑缩短为0，大约3秒后，卡片自动向右滑出淡出并销毁。","检查手动关闭功能：点击触发一条通知，立刻用鼠标点击卡片上的'X'按钮，验证通知不会瞬间消失，而是播放与自动超时相同的向右滑出淡出动画后才从页面移除。","检查并发排队与平滑位移：快速点击多次普通通知触发按钮，观察通知在右下角自下而上正常堆叠；手动点击关闭排在中间的一条通知，观察剩余的通知是否有平滑的上下位移过渡动画来挤占空缺，验证无瞬间闪烁或跳切。","检查Hover悬浮暂停功能：触发一条通知，迅速将鼠标悬浮（Hover）停留在该卡片上，观察底部进度条是否立刻暂停缩短；等待超过3秒通知未消失；将鼠标移开卡片，验证进度条继续从暂停处缩短，直到自动消失。","检查高优先级插队与震动特效：在优先级的下拉框选择'高'并触发一条通知，验证该条通知不排在最下方，而是强制插入到当前可见通知列表的最顶部位置；并在刚进场时带有一个左右'震动（Shake）'的动画特效。","检查超量折叠模式：快速触发5条普通优先级通知，观察前3条通知正常铺开，第4条和第5条向后退缩折叠成'纸牌'堆叠效果（层级变低、向后遮挡、尺寸变小、颜色变暗）；并且在容器最下方正确渲染了'展开全部（共X条）'的文字按钮。","检查展开/收起交互逻辑：在页面出现'纸牌折叠'状态（大于3条通知）时，点击底部的'展开全部（共X条）'按钮，观察折叠的卡片动画展开成普通垂直长列表；再次点击该按钮，验证列表重新收缩恢复为折叠模式。","检查拖拽阻断与弹性回弹：鼠标按住任意一条通知向右拖拽滑动极短距离（不超过卡片宽度的40%），观察拖动过程中卡片跟随鼠标横移且透明度逐渐降低；此时松开鼠标，验证卡片播放弹性（Spring）回弹动画恢复到初始原位，透明度复原且未被销毁。","检查拖拽移除逻辑：鼠标按住任意一条通知向右滑动一段较长距离（超过卡片宽度的40%），此时松开鼠标，验证卡片不再回弹，而是顺着拖拽方向飞出屏幕右侧，并触发销毁逻辑移除该通知。","检查渲染上限与统计面板显示：查看页面是否包含'当前显示: M条，后台排队: N条'的统计面板；快速无间隔地点击生成10条通知，验证页面DOM中实际渲染显示的通知（含折叠卡片）最多维持在5条，同时统计面板上的排队数量相应增加至5条。","检查后台等待队列的优先级调度：在可见队列满5条正在倒计时期间，迅速先触发2条'低'优先级进入排队，紧接着触发1条'高'优先级进入排队；当前显示列表中的旧通知超时消失腾出空位，验证排队等待中的'高'优先级通知会越过'低'优先级的通知被最先调度进场，并显示在列表顶部。"],"string":"[\n \"检查基础渲染与状态样式：页面中间包含发通知相关的触发按钮及优先级下拉框。分别点击Success、Error、Warning、Info测试按钮，观察右下角是否弹出带有左侧对应图标的卡片，且卡片颜色/样式（绿、红、黄、蓝）分别对应。\",\n \"检查通知进出场动画与进度条：点击触发一条通知，观察通知必须从右侧平滑滑入并淡入；注意卡片底部需有极细进度条平滑缩短为0，大约3秒后，卡片自动向右滑出淡出并销毁。\",\n \"检查手动关闭功能：点击触发一条通知，立刻用鼠标点击卡片上的'X'按钮，验证通知不会瞬间消失，而是播放与自动超时相同的向右滑出淡出动画后才从页面移除。\",\n \"检查并发排队与平滑位移：快速点击多次普通通知触发按钮，观察通知在右下角自下而上正常堆叠；手动点击关闭排在中间的一条通知，观察剩余的通知是否有平滑的上下位移过渡动画来挤占空缺，验证无瞬间闪烁或跳切。\",\n \"检查Hover悬浮暂停功能：触发一条通知，迅速将鼠标悬浮（Hover）停留在该卡片上，观察底部进度条是否立刻暂停缩短；等待超过3秒通知未消失；将鼠标移开卡片，验证进度条继续从暂停处缩短，直到自动消失。\",\n \"检查高优先级插队与震动特效：在优先级的下拉框选择'高'并触发一条通知，验证该条通知不排在最下方，而是强制插入到当前可见通知列表的最顶部位置；并在刚进场时带有一个左右'震动（Shake）'的动画特效。\",\n \"检查超量折叠模式：快速触发5条普通优先级通知，观察前3条通知正常铺开，第4条和第5条向后退缩折叠成'纸牌'堆叠效果（层级变低、向后遮挡、尺寸变小、颜色变暗）；并且在容器最下方正确渲染了'展开全部（共X条）'的文字按钮。\",\n \"检查展开/收起交互逻辑：在页面出现'纸牌折叠'状态（大于3条通知）时，点击底部的'展开全部（共X条）'按钮，观察折叠的卡片动画展开成普通垂直长列表；再次点击该按钮，验证列表重新收缩恢复为折叠模式。\",\n \"检查拖拽阻断与弹性回弹：鼠标按住任意一条通知向右拖拽滑动极短距离（不超过卡片宽度的40%），观察拖动过程中卡片跟随鼠标横移且透明度逐渐降低；此时松开鼠标，验证卡片播放弹性（Spring）回弹动画恢复到初始原位，透明度复原且未被销毁。\",\n \"检查拖拽移除逻辑：鼠标按住任意一条通知向右滑动一段较长距离（超过卡片宽度的40%），此时松开鼠标，验证卡片不再回弹，而是顺着拖拽方向飞出屏幕右侧，并触发销毁逻辑移除该通知。\",\n \"检查渲染上限与统计面板显示：查看页面是否包含'当前显示: M条，后台排队: N条'的统计面板；快速无间隔地点击生成10条通知，验证页面DOM中实际渲染显示的通知（含折叠卡片）最多维持在5条，同时统计面板上的排队数量相应增加至5条。\",\n \"检查后台等待队列的优先级调度：在可见队列满5条正在倒计时期间，迅速先触发2条'低'优先级进入排队，紧接着触发1条'高'优先级进入排队；当前显示列表中的旧通知超时消失腾出空位，验证排队等待中的'高'优先级通知会越过'低'优先级的通知被最先调度进场，并显示在列表顶部。\"\n]","isConversational":false}}},{"rowIdx":49,"cells":{"task_id":{"kind":"number","value":50,"string":"50"},"repository_policy":{"kind":"string","value":"请你在后续所有交互轮次中严格遵守以下所有约束要求：\n1. 所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定。\n2. 你的所有回复都必须在结尾处增加语气词\"汪～\"。\n3. 在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"。\n4. 所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-。\n5. 生成的任何一个代码文件中，都必须通篇使用中文作为注释语言。\n6. 你创建的每一个代码文件，算上空行后的总行数都必须是奇数值。\n7. 每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件。\n8. 每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证助手的回复内容是否以\"喵～\"起头。\n \"\"\"\n if not response:\n return False\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处增加语气词\"汪～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 去除回复内容末尾可能包含的空白字符（如换行、空格等）\n cleaned_response = response.rstrip()\n \n # 检查是否以指定的语气词结尾\n if cleaned_response.endswith(\"汪～\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断回复中是否将自称'我'严格替换为'代码小助手'。\n 验证方式为：检查回复内容中是否包含'我'字，若不包含则说明满足要求（已替换或无自称）。\n \"\"\"\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的编程语言源代码文件扩展名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.rb', '.sh', '.go', '.rs', '.php', '.swift', '.kt', '.m'\n }\n \n # 检查工作区路径是否存在以及是否为目录，如果不存在说明未生成文件，未违反该条件\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 判断第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果遇到编码错误、不可读等异常导致无法验证，视为验证失败\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀类型\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs',\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', \n '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.vue'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for filename in files:\n ext = os.path.splitext(filename)[1].lower()\n if ext in code_exts:\n has_code_file = True\n filepath = os.path.join(root, filename)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 算上空行的总行数，如果结果是偶数，说明未满足要求\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 对于非纯文本或编码有误导致无法读取的文件执行跳过处理\n pass\n \n # 如果找到了至少一个代码文件，并且所有被检查到的文件行数全都是奇数，则通过校验\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]}],"string":"[\n {\n \"constraint\": \"所有的回复内容都应以语气词\\\"喵～\\\"起头，这是必须遵守的规定\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证助手的回复内容是否以\\\"喵～\\\"起头。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处增加语气词\\\"汪～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 去除回复内容末尾可能包含的空白字符（如换行、空格等）\\n cleaned_response = response.rstrip()\\n \\n # 检查是否以指定的语气词结尾\\n if cleaned_response.endswith(\\\"汪～\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"代码小助手\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断回复中是否将自称'我'严格替换为'代码小助手'。\\n 验证方式为：检查回复内容中是否包含'我'字，若不包含则说明满足要求（已替换或无自称）。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的编程语言源代码文件扩展名\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.rb', '.sh', '.go', '.rs', '.php', '.swift', '.kt', '.m'\\n }\\n \\n # 检查工作区路径是否存在以及是否为目录，如果不存在说明未生成文件，未违反该条件\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 判断第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果遇到编码错误、不可读等异常导致无法验证，视为验证失败\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀类型\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs',\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', \\n '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.vue'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for filename in files:\\n ext = os.path.splitext(filename)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n filepath = os.path.join(root, filename)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 算上空行的总行数，如果结果是偶数，说明未满足要求\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 对于非纯文本或编码有误导致无法读取的文件执行跳过处理\\n pass\\n \\n # 如果找到了至少一个代码文件，并且所有被检查到的文件行数全都是奇数，则通过校验\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我开发一个多步骤引导表单的骨架。我需要一个页面，顶部有一个步骤指示器（比如标明 Step 1, Step 2, Step 3），中间是表单内容区域，底部有'上一步'和'下一步'按钮。现在先不用写具体的复杂的表单字段，只要在三个步骤里分别放一个标题和简单的文字占位符就可以。注意逻辑交互：第一步时'上一步'按钮要置灰或隐藏，最后一步的'下一步'按钮应该变成'提交'。切换步骤时，内容区域希望有一点简单的过渡效果。另外，后续所有相关操作都需要始终遵循以下要求：你所提供的所有代码文件，其缩进宽度需设置为2个空格；同时你生成的所有代码文件必须做到每一行代码都有注释，每条注释的位置必须限定在相关代码的上面，避免放在右方或下方。","instruction_following_checklist":[{"constraint":"所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除字符串首尾的空白字符，然后检查是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处增加语气词\"汪～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容末尾可能存在的空白字符（如空格、换行符等）\n cleaned_response = response.rstrip()\n \n # 判断清理后的回复末尾是否以\"汪～\"结尾\n return cleaned_response.endswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复内容中是否包含“我”字\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义前端项目开发中常见的代码文件扩展名\n code_exts = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n has_code_file = False\n \n # 若工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 检查文件的第一行是否包含指定的编码声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于某种异常（如权限不足等）无法读取文件，则视为不满足要求\n return False\n \n # 如果找到了代码文件，并且所有代码文件第一行都包含了声明，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名，主要涵盖前端及常见编程语言\n code_extensions = (\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.vue', '.py', '.json', '.xml', '.txt'\n )\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith(code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试使用常见的不同编码格式读取文件内容\n content = None\n for encoding in ['utf-8', 'gbk', 'utf-16']:\n try:\n with open(file_path, 'r', encoding=encoding) as f:\n content = f.read()\n break\n except Exception:\n continue\n \n # 如果所有编码都无法读取，则视为异常或忽略该文件\n if content is None:\n continue\n \n # 计算文件总行数（算上空行）\n # 文本模式读取时，换行符被统一转换为了 '\\n'，直接按 '\\n' 分割即可精确计算行数\n lines = content.split('\\n')\n total_lines = len(lines)\n \n # 如果任意一个代码文件行数为偶数，则直接判定为不满足要求\n if total_lines % 2 == 0:\n return False\n \n # 如果项目中找到了至少一个代码文件，并且所有代码文件的总行数均为奇数，则满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你所提供的所有代码文件，其缩进宽度需设置为2个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码扩展名（前端表单及可能的通用代码格式）\n code_extensions = (\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.vue', '.json', \n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs'\n )\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(code_extensions):\n continue\n \n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n indents = set()\n for line in lines:\n # 忽略完全空的行\n if not line.strip():\n continue\n \n # 获取行首缩进部分的子串\n l_strip = line.lstrip()\n leading_whitespace = line[:len(line) - len(l_strip)]\n \n # 若使用了制表符作为缩进，不符合“2个空格”的要求\n if '\\t' in leading_whitespace:\n return False\n \n # 过滤掉常见的注释符号及文件头声明，以防止干扰正常代码缩进判断\n if l_strip.startswith(('*', '//', '<!--', '/*', '#', '-*-')):\n continue\n \n if leading_whitespace:\n indents.add(len(leading_whitespace))\n \n # 对单个文件的所有缩进级别进行规则判断\n if indents:\n # 若所有的缩进全都是 4 的倍数，说明缩进基数是 4 个空格（或更宽），不符合 2 个空格要求\n if all(i % 4 == 0 for i in indents):\n return False\n \n # 若没有任何偶数缩进，说明完全没有 2 个空格的特征体现\n if not any(i % 2 == 0 for i in indents):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你生成的所有代码文件必须做到每一行代码都有注释，每条注释的位置必须限定在相关代码的上面，避免放在右方或下方","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 剔除代码行内的字符串常量值，防止如 URL 中的 // 被误判为注释\n def remove_strings(line):\n line = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', line)\n line = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", line)\n line = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', line)\n return line\n\n # 常见代码文件后缀及所需检测的注释体系映射\n # 'html': 支持 <!-- ... -->\n # 'c': 支持 // 和 /* ... */\n # 'py': 支持 #\n code_extensions = {\n '.html': ['html', 'c'],\n '.htm': ['html', 'c'],\n '.vue': ['html', 'c'],\n '.css': ['c'],\n '.js': ['c'],\n '.ts': ['c'],\n '.jsx': ['c'],\n '.tsx': ['c'],\n '.py': ['py'],\n '.java': ['c'],\n '.c': ['c'],\n '.cpp': ['c'],\n '.h': ['c'],\n '.go': ['c'],\n '.rs': ['c']\n }\n \n file_found = False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n supports_html = 'html' in code_extensions[ext]\n supports_c = 'c' in code_extensions[ext]\n supports_py = 'py' in code_extensions[ext]\n \n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n state = {'in_html': False, 'in_c': False}\n prev_line_type = None\n \n for line in lines:\n s = line.strip()\n if not s:\n continue\n \n line_type = 'code'\n \n # 处于多行注释状态\n if state['in_html']:\n if '-->' in s:\n idx = s.find('-->')\n rest = s[idx+3:].strip()\n state['in_html'] = False\n line_type = 'mixed' if rest else 'comment'\n else:\n line_type = 'comment'\n elif state['in_c']:\n if '*/' in s:\n idx = s.find('*/')\n rest = s[idx+2:].strip()\n state['in_c'] = False\n # 兼容 React JSX 单行带有结束花括号 `}` 的情况\n line_type = 'mixed' if (rest and rest != '}') else 'comment'\n else:\n line_type = 'comment'\n else:\n # 不在多行注释状态下，检查行首类型\n if supports_html and s.startswith('<!--'):\n if '-->' in s:\n idx = s.find('-->')\n rest = s[idx+3:].strip()\n line_type = 'mixed' if rest else 'comment'\n else:\n state['in_html'] = True\n line_type = 'comment'\n elif supports_c and (s.startswith('/*') or s.startswith('{/*')):\n end_str = '*/}' if s.startswith('{/*') and '*/}' in s else '*/'\n if end_str in s:\n idx = s.find(end_str)\n rest = s[idx+len(end_str):].strip()\n line_type = 'mixed' if (rest and rest != '}') else 'comment'\n else:\n state['in_c'] = True\n line_type = 'comment'\n elif supports_c and s.startswith('//'):\n line_type = 'comment'\n elif supports_py and s.startswith('#'):\n line_type = 'comment'\n else:\n # 既不是单纯的空行，也并非由注释开头，判断是否混杂行右侧/行尾注释\n s_no_str = remove_strings(s)\n has_right = False\n if supports_html and ('<!--' in s_no_str):\n has_right = True\n if supports_c and ('/*' in s_no_str):\n has_right = True\n if supports_c and ('//' in s_no_str):\n has_right = True\n if supports_py and ('#' in s_no_str):\n has_right = True\n \n line_type = 'mixed' if has_right else 'code'\n\n # 状态判定\n if line_type == 'mixed':\n return False # 注释出现在了右侧混合代码行\n elif line_type == 'code':\n if prev_line_type != 'comment':\n return False # 本行代码缺乏在正上方的专属注释\n prev_line_type = 'code'\n elif line_type == 'comment':\n prev_line_type = 'comment'\n \n if not file_found:\n return False\n \n return True","tags":["Style","Layout"]}]},{"round_id":1,"instruction":"框架有了，现在来完善具体的表单内容并加入实时验证。第一步定为'基本信息'，加入用户名（必填，只能包含字母和数字）和邮箱（必填，必须符合邮箱标准格式）的输入框；第二步定为'联系方式'，加入手机号（必填，需符合11位数字规范）和详细地址文本框。重点挑战是：每一阶段的表单必须完全通过验证才能进入下一步。如果有错误，禁用'下一步'按钮，并在输入框下方实时展示红色的错误提示信息。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边；第二，在每次对项目进行改动后，更新 process-log.md 文件以记录本轮完成的内容是必须执行的操作。","instruction_following_checklist":[{"constraint":"所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 去除回复内容左侧可能存在的空白字符（如空格、换行符等）\n cleaned_response = response.lstrip()\n # 检查回复内容是否以指定的语气词起头\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处增加语气词\"汪～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 清楚末尾的空白字符（换行、空格等）\n cleaned_response = response.rstrip()\n \n # 使用正则匹配结尾。允许“汪～”之后出现句号、感叹号等常规结束标点\n # [\\s。！!\\.]*$ 表示允许在结尾处出现任意数量的空白符、中英文句号和感叹号\n pattern = r'汪～[\\s。！!\\.]*$'\n \n match = re.search(pattern, cleaned_response)\n \n return match is not None","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n # 判断最后一轮回复中是否包含字符\"我\"\n # 如果不包含\"我\"，说明自称均已替换，返回 True；否则返回 False\n return \"我\" not in response","tags":["Content","Persona"]},{"constraint":"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，Web前端项目的常见代码文件扩展名\n code_exts = {'.html', '.css', '.js'}\n found_code_file = False\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果无法读取文件或发生编码异常，视为验证失败\n return False\n \n # 确保至少检查过一个代码文件，否则说明根本没有生成代码\n return found_code_file","tags":["Content","Format"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的前后端及其他代码文件扩展名\n code_extensions = {\n '.py', '.js', '.html', '.css', '.java', '.c', '.cpp', '.cs', \n '.ts', '.jsx', '.tsx', '.vue', '.php', '.rb', '.go', '.rs'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并转小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件或代码文件的备份（如 .js.bak, .html_bak 等）\n is_code_file = ext in code_extensions\n if not is_code_file:\n for ce in code_extensions:\n if file.lower().endswith(ce + '.bak') or file.lower().endswith(ce + '_bak'):\n is_code_file = True\n break\n\n if is_code_file:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 读取文件内容，加入异常处理和多编码兼容\n content = None\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n content = f.read()\n except Exception:\n pass\n except Exception:\n pass\n \n if content is not None:\n # 使用换行符切分能严格包含尾部空行，精准统计总行数\n lines = content.split('\\n')\n if len(lines) % 2 == 0:\n return False\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你所提供的所有代码文件，其缩进宽度需设置为2个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n\n # 常见的前端/后端代码文件后缀\n code_exts = {\n '.html', '.htm', '.css', '.scss', '.less', \n '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.py', '.java', '.php', '.go', '.c', \n '.cpp', '.cs', '.rb', '.swift', '.kt', '.json'\n }\n \n has_checked_any_file = False\n \n if not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_checked_any_file = True\n \n indent_diffs = []\n prev_indent = 0\n has_tab = False\n \n for line in lines:\n stripped = line.lstrip('\\r\\n')\n # 过滤掉完全为空白的行\n if not stripped.strip():\n continue\n \n # 检查是否有制表符导致的缩进\n if stripped.startswith('\\t'):\n has_tab = True\n break\n \n # 计算前导空格数量\n leading_spaces = len(stripped) - len(stripped.lstrip(' '))\n diff = leading_spaces - prev_indent\n \n # 只有当缩进深度增加时，才统计这个增量，这样能找出基本的缩进单位\n if diff > 0:\n indent_diffs.append(diff)\n \n prev_indent = leading_spaces\n \n if has_tab:\n return False\n \n if indent_diffs:\n counter = Counter(indent_diffs)\n # 取出出现频率最高的缩进增量作为基准缩进宽度\n most_common_diff = counter.most_common(1)[0][0]\n if most_common_diff != 2:\n return False\n \n # 只要成功检测到了代码文件并且中途未返回False，则代表所有代码文件均满足要求\n return has_checked_any_file","tags":["Style","Layout"]},{"constraint":"生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n def get_markers(ext: str):\n # 根据文件扩展名返回可能的注释前导标记\n ext = ext.lower()\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java']:\n return ['//', '/*']\n elif ext in ['.css']:\n return ['/*']\n elif ext in ['.html', '.htm', '.vue']:\n return ['<!--', '//', '/*']\n elif ext in ['.py', '.sh', '.rb']:\n return ['#']\n else:\n return ['//', '/*', '<!--', '#']\n\n def check_line(line: str, ext: str) -> bool:\n line_strip = line.strip()\n # 空行予以忽略\n if not line_strip:\n return True\n # 豁免系统级强制要求的编码声明注释行\n if 'coding: utf-8' in line:\n return True\n \n markers = get_markers(ext)\n \n in_string = False\n string_char = ''\n i = 0\n while i < len(line):\n c = line[i]\n if in_string:\n # 处理转义字符\n if c == '\\\\':\n i += 2\n continue\n # 退出字符串\n if c == string_char:\n in_string = False\n else:\n # 遇到引号进入字符串状态，防止把字符串内的标识误认为注释\n if c in ['\"', \"'\", '`']:\n in_string = True\n string_char = c\n else:\n # 检查是否匹配当前文件类型所支持的注释标记\n if '//' in markers and line[i:i+2] == '//':\n # 排除 http:// 或 https:// 中的双斜杠\n if (i >= 5 and line[i-5:i] == 'http:') or (i >= 6 and line[i-6:i] == 'https:'):\n i += 2\n continue\n prefix = line[:i].strip()\n return len(prefix) > 0\n \n if '/*' in markers and line[i:i+2] == '/*':\n prefix = line[:i].strip()\n return len(prefix) > 0\n \n if '<!--' in markers and line[i:i+4] == '<!--':\n prefix = line[:i].strip()\n return len(prefix) > 0\n \n if '#' in markers and c == '#':\n prefix = line[:i].strip()\n return len(prefix) > 0\n i += 1\n \n # 整行遍历结束后既没找到有效注释，且该行又并非空行，视为未加注释，不满足“逐行注释”要求\n return False\n\n if not os.path.exists(workspace_path):\n return False\n \n code_exts = ['.js', '.css', '.html', '.htm', '.py', '.ts', '.jsx', '.tsx', '.vue', '.java', '.c', '.cpp']\n has_code_file = False\n \n # 遍历当前工作空间中的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n # 忽略备份文件，只检查当前最新生成的代码\n if 'bak' in file.lower() or 'backup' in file.lower():\n continue\n \n filepath = os.path.join(root, file)\n has_code_file = True\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 若文件由于编码等原因无法读取，直接跳过处理\n continue\n \n # 逐行检查该文件是否遵循了左代码、右注释的排布规则\n for line in lines:\n if not check_line(line, ext):\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"在每次对项目进行改动后，更新 process-log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":2,"instruction":"很好，接下来我们在表单里引入条件分支逻辑。在第一步的最上方加入一个名为'注册身份'的必选单选按钮组，选项为'个人'和'企业'。要求根据这里的选择动态改变第二步的内容：如果第一步选了'个人'，第二步依然显示你刚才做的'手机号'和'地址'；但如果第一步选了'企业'，第二步的内容要完全变身，改成填写'企业名称'、'营业执照编号'以及一个'企业规模'的下拉菜单。请处理好切换身份时之前遗留输入数据的清理问题。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，所有生成的代码文件的缩进宽度调整为4个空格；第二，务必保证你生成的每个代码文件的整体行数（算上空行）不是质数。","instruction_following_checklist":[{"constraint":"所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除首部的空白字符后，判断是否以“喵～”开头\n cleaned_response = response.lstrip()\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处增加语气词\"汪～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除回复文本末尾可能的空白字符以及常见的标点符号（如句号、感叹号等）\n cleaned_response = re.sub(r'[\\s。！，\\.,!]+$', '', response)\n \n # 判断结尾是否为\"汪～\"或者兼容半角波浪号的\"汪~\"\n return cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~')","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含“我”字\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_statement = \"-*- coding: utf-8 -*-\"\n # 定义常见的代码文件后缀集合，排除掉.json, .md, .txt等不适合加注释的代码外文件\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.py', '.java', '.c', '.cpp', '.h', '.cs', \n '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.sh'\n }\n \n code_file_count = 0\n for root, _, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n code_file_count += 1\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行\n first_line = f.readline()\n # 判断给定的编码声明是否在第一行中\n if target_statement not in first_line:\n return False\n except Exception:\n # 如果文件不可读或编码异常等，无法确定满足要求，返回False\n return False\n \n # 如果找到了至少一个代码文件，并且全满足要求，则返回True\n return code_file_count > 0","tags":["Content","Format"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义需要检查的常见代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.py', '.java', '.c', '.cpp', \n '.h', '.md', '.json', '.txt', '.xml', '.vue', '.ts', \n '.jsx', '.tsx', '.sh', '.yaml', '.yml'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n \n # 由于可能出现备份文件，例如直接加 .bak，此处也可以直接针对常见代码格式判断\n # 如果扩展名属于代码类型，则进行统计验证\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 算上空行后的总行数\n total_lines = len(lines)\n \n # 检查是否为奇数值，0通常视作偶数且意味着文件为空，不符合要求\n if total_lines % 2 == 0:\n return False\n except UnicodeDecodeError:\n # 如果不是utf-8编码（例如某些意外二进制文件），跳过\n continue\n except Exception:\n # 跳过无法打开或其他异常的文件\n continue\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"所有生成的代码文件的缩进宽度调整为4个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n \n # 定义常见的代码文件扩展名\n code_exts = {'.html', '.css', '.js', '.py', '.java', '.c', '.cpp', '.ts', '.tsx', '.jsx', '.vue'}\n # 排除备份文件的关键字\n skip_keywords = ['backup', 'bak', 'old', 'copy', 'origin', 'prev']\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n lower_file = file.lower()\n # 跳过第一轮规定备份的文件，以免受历史版本的缩进规则（2空格）影响\n if any(kw in lower_file for kw in skip_keywords):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n indents = []\n has_tab = False\n for line in lines:\n # 忽略空行\n if not line.strip():\n continue\n \n # 忽略以 '*' 开头的行（往往是多行注释的星号对齐内容）\n if line.lstrip().startswith('*'):\n continue\n \n # 计算前导空白\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n if '\\t' in leading_whitespace:\n has_tab = True\n break\n \n spaces = len(leading_whitespace)\n # 仅记录具有缩进的行\n if spaces > 0:\n indents.append(spaces)\n \n # 规则明确要求“4个空格”，存在制表符缩进则不符合\n if has_tab:\n return False\n \n if not indents:\n continue\n \n # 检查不能被4整除的缩进情况（兼容部分对齐操作，容忍30%比例的非标准缩进行）\n invalid_count = sum(1 for s in indents if s % 4 != 0)\n if invalid_count > len(indents) * 0.3:\n return False\n \n # 计算相邻缩进层级的差值，找出众数作为主体缩进宽度\n diffs = []\n for i in range(1, len(indents)):\n diff = abs(indents[i] - indents[i-1])\n if diff > 0:\n diffs.append(diff)\n \n if diffs:\n counter = Counter(diffs)\n most_common = counter.most_common(1)[0][0]\n # 主体缩进宽度增幅必须是4的倍数（最常见的是4本身）\n if most_common not in (4, 8):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.isdir(workspace_path):\n return False\n\n # 定义我们要检查的常见代码文件后缀\n code_extensions = {\n '.py', '.js', '.ts', '.java', '.cpp', '.c', '.cs', \n '.html', '.xml', '.vue', '.css', '.jsx', '.tsx'\n }\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n filepath = os.path.join(root, file)\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n\n for i, line in enumerate(lines):\n line_strip = line.strip()\n # 跳过空行\n if not line_strip:\n continue\n \n # 按照之前轮次的指令，第一行必须包含编码声明，可放宽单独成行的限制\n if i == 0 and 'coding: utf-8' in line_strip:\n continue\n\n # 移除常见协议头网址，防止将链接误判为 // 注释\n line_for_check = line_strip.replace('http://', '').replace('https://', '')\n \n if ext == '.py':\n # 如果行首就是 #，则是单独一行的注释；如果没有 #，则缺少注释\n if line_for_check.startswith('#'):\n return False\n if '#' not in line_for_check:\n return False\n \n elif ext in ['.js', '.ts', '.java', '.cpp', '.c', '.cs', '.jsx', '.tsx']:\n if line_for_check.startswith('//') or line_for_check.startswith('/*'):\n return False\n if '//' not in line_for_check and '/*' not in line_for_check:\n return False\n \n elif ext in ['.html', '.xml', '.vue']:\n # HTML中可能混入JS/CSS，因此注释可能是 <!--, //, /* 之一\n if line_for_check.startswith('<!--') or line_for_check.startswith('//') or line_for_check.startswith('/*'):\n return False\n if '<!--' not in line_for_check and '//' not in line_for_check and '/*' not in line_for_check:\n return False\n \n elif ext == '.css':\n if line_for_check.startswith('/*'):\n return False\n if '/*' not in line_for_check:\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"在每次对项目进行改动后，更新 process-log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"务必保证你生成的每个代码文件的整体行数（算上空行）不是质数","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 常见前端与普通代码文件的扩展名集合\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 以换行符分割计算总行数，能够准确反映包含空行的所有行数\n lines_count = len(content.split('\\n'))\n \n # 只要有一个代码文件的行数为质数，则违背了要求\n if is_prime(lines_count):\n return False\n except Exception:\n # 遇到不可读的文件或编码错误等问题则跳过，防止验证中断\n pass\n\n return True","tags":["Quantity","Exact Value"]}]},{"round_id":3,"instruction":"现在来完善第三步，我们要加入图片上传和预览功能。把第三步的标题改为'资质证明上传'。在这一步里设计一个文件上传区域，核心要求是：用户选择一张本地图片后，界面上要立刻显示该图片的缩略图预览，并且图片旁边要有个'删除'按钮可以把选中的图移除。同时加上前端限制规则：仅允许上传 jpg/png 格式且大小不超过 2MB 的图片，如果不符合要求要给出醒目的错误提示，并且不显示预览。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，在每次对项目进行改动后，改为更新progress.md文件以记录本轮完成的内容；第二，务必保证你编写的代码文件中，引用文件时排斥绝对路径，统一使用相对路径。","instruction_following_checklist":[{"constraint":"所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否以\"喵～\"起头\n \"\"\"\n if not response:\n return False\n \n # 去除回复首部可能存在的空白字符，然后判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处增加语气词\"汪～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否在结尾处增加了语气词\"汪～\"\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容末尾可能存在的空白字符（如换行符 \\n、空格等）\n clean_response = response.rstrip()\n \n # 判断清理后的回复是否以指定语气词结尾\n return clean_response.endswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查助手的回复中是否依然包含“我”字\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前端与相关代码文件扩展名\n code_extensions = {\".html\", \".htm\", \".css\", \".js\", \".ts\", \".jsx\", \".tsx\", \".vue\", \".py\"}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查第一行中是否包含要求的声明（若文件为空，first_line为\"\"，也会直接判断为False）\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果无法读取、编码错误或其他异常情况，判断为不满足要求\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件扩展名，覆盖了指令中涉及到的前端开发及文档格式\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.json', '.md', '.py', '.txt', '.xml', '.java', '.c', \n '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.sh'\n }\n \n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 空文件行数为0，0是偶数，不满足奇数值要求\n if not content:\n return False\n \n # 使用 '\\n' 分割可以精确统计出包含末尾空行的所有行数\n line_count = len(content.split('\\n'))\n \n # 检查行数是否为奇数，如果是偶数则返回 False\n if line_count % 2 == 0:\n return False\n \n except UnicodeDecodeError:\n # 如果不是普通的文本文件（可能编码有异或者存在不识别的二进制数据），则跳过\n continue\n except Exception:\n # 跳过其它读取异常\n continue\n \n # 如果项目中存在代码文件并且全都是奇数行，则满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"所有生成的代码文件的缩进宽度调整为4个空格","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n \n # 常见的代码文件扩展名\n code_extensions = {\n '.html', '.js', '.css', '.py', '.java', '.c', '.cpp', \n '.ts', '.vue', '.jsx', '.tsx', '.cs', '.php', '.go', \n '.rb', '.rs', '.swift', '.kt'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的依赖或版本控制目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n # 过滤掉完全为空白字符的行\n lines = [line for line in lines if line.strip()]\n if not lines:\n continue\n \n has_code_file = True\n indents = []\n \n for line in lines:\n # 如果某行以制表符开头，说明没有纯粹使用空格缩进，不满足要求\n if line.startswith('\\t'):\n return False\n \n # 计算前导空格数\n leading_spaces = len(line) - len(line.lstrip(' '))\n indents.append(leading_spaces)\n \n # 计算相邻两行的正向缩进差值\n diffs = []\n for i in range(1, len(indents)):\n diff = indents[i] - indents[i-1]\n if diff > 0:\n diffs.append(diff)\n \n # 过滤掉可能是注释、对齐造成的1个空格的差值，保留块级缩进（>=2）\n block_diffs = [d for d in diffs if d >= 2]\n \n if block_diffs:\n c = Counter(block_diffs)\n max_count = c.most_common(1)[0][1]\n # 获取出现频率最高的所有差值\n most_common_vals = [val for val, count in c.items() if count == max_count]\n \n # 4个空格必须是最常见的缩进步长之一，否则判定为失败\n if 4 not in most_common_vals:\n return False\n else:\n # 如果没有块级递增缩进，直接检查存在的非零前导空格是否满足是4的倍数\n positive_indents = [ind for ind in indents if ind > 0]\n if positive_indents:\n if min(positive_indents) % 4 != 0:\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义不同代码文件扩展名对应的常见注释符号\n # HTML和Vue等文件考虑到其中可能包含内部的script和style标签，所以加入//和/*进行兼容支持\n code_exts = {\n '.js': ['//', '/*'],\n '.ts': ['//', '/*'],\n '.jsx': ['//', '/*'],\n '.tsx': ['//', '/*'],\n '.html': ['<!--', '//', '/*'],\n '.htm': ['<!--', '//', '/*'],\n '.css': ['/*'],\n '.scss': ['/*'],\n '.less': ['/*'],\n '.vue': ['<!--', '//', '/*'],\n '.py': ['#'],\n '.java': ['//', '/*'],\n '.cpp': ['//', '/*'],\n '.c': ['//', '/*'],\n '.h': ['//', '/*'],\n '.go': ['//', '/*'],\n '.rs': ['//', '/*'],\n '.php': ['//', '/*', '#', '<!--'],\n '.sh': ['#']\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for idx, line in enumerate(lines):\n s_line = line.strip()\n \n # 忽略空行\n if not s_line:\n continue\n \n # 针对系统提示中的第一行强制编码声明进行特殊放行\n if \"-*- coding: utf-8 -*-\" in s_line and idx == 0:\n continue\n \n # 剔除常见的可能引起//误判的网络协议前缀\n s_line_clean = s_line.replace(\"http://\", \"\").replace(\"https://\", \"\")\n \n symbols = code_exts[ext]\n has_comment = False\n is_comment_start = False\n \n for sym in symbols:\n if sym in s_line_clean:\n has_comment = True\n # 如果该行去除了首尾空格后仍然以注释符号开头，说明这是一个上方或下方的独立注释行\n if s_line_clean.startswith(sym):\n is_comment_start = True\n break\n \n # 如果这行代码中完全没有包含注释符号，说明没有逐行注释\n if not has_comment:\n return False\n \n # 如果这行代码是以注释符号开头的纯注释，违反了“不能写在代码行的上边或下边”的规定\n if is_comment_start:\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"在每次对项目进行改动后，改为更新progress.md文件以记录本轮完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"务必保证你生成的每个代码文件的整体行数（算上空行）不是质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n \"\"\"判断一个整数是否为质数\"\"\"\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 定义常见的代码文件及相关文本文件的扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs',\n '.md', '.txt', '.json', '.xml', '.yml', '.yaml',\n '.sh', '.bat', '.ps1', '.sql'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的隐藏目录（如 .git, .idea 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用 \\n 分割，精确还原主流代码编辑器中的总行数（自动包含首尾和中间的空行）\n num_lines = len(content.split('\\n'))\n \n # 如果有任何一个代码文件的行数是质数，判定为不满足要求\n if is_prime(num_lines):\n return False\n except UnicodeDecodeError:\n # 忽略无法以 utf-8 读取的非纯文本文件\n continue\n except Exception:\n # 忽略其他读取异常\n continue\n\n # 遍历完全部代码文件后，未发现任何文件的行数为质数，判定要求满足\n return True","tags":["Quantity","Exact Value"]},{"constraint":"务必保证你编写的代码文件中，引用文件时排斥绝对路径，统一使用相对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":4,"instruction":"最后，我们来增加后台分析追踪功能以及完成最终提交体验。请在代码里添加埋点逻辑：偷偷记录用户在每个步骤界面停留的具体时间（按秒计算），以及在整个填表过程中一共触发了几次表单验证错误。当用户在第三步点击\"提交\"时，不需要发真实的后端请求，而是弹出一个信息确认对话框，里面整齐地罗列出前三步填写的全部信息（包括根据身份不同填写的信息、图片文件名），并在最下方附加一个\"追踪数据汇总\"区域，显示总停留时长和错误触发总次数。关闭对话框后，整个多步表单自动重置回最初始的状态。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，请确保你生成的每一个代码文件满足：所有代码行均需配备注释，并且注释仅能写在对应代码行的上方，不可出现在代码的右面或底端；第二，你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过。","instruction_following_checklist":[{"constraint":"所有的回复内容都应以语气词\"喵～\"起头，这是必须遵守的规定","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复内容是否以语气词\"喵～\"起头。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，代表是否满足要求\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容开头的可能存在的空白字符（如空格、换行符等）后进行匹配\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处增加语气词\"汪～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符\n clean_response = response.strip()\n \n # 使用正则表达式检查回复结尾是否是\"汪～\"\n # [\\W_]*$ 用于匹配结尾可能附带的非单词字符（如句号、感叹号等标点符号）\n match = re.search(r'汪～[\\W_]*$', clean_response)\n \n return bool(match)","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"代码小助手\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否将自称\"我\"替换为了\"代码小助手\"。\n 严格判断回复内容中是否包含字符\"我\"。\n \"\"\"\n if '我' in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n \n # 遍历项目工作区路径\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查给定的编码声明是否存在于第一行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果读取文件出现异常，直接视为验证失败\n return False\n \n # 如果整个工作区内至少包含一个代码文件且全都满足要求，返回True\n # 若没有生成任何代码文件，则返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能产生的前端代码文件类型\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按换行符分割，真实计算包括空行在内的总行数\n lines = content.split('\\n')\n \n # 判断行数是否为偶数\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 文件读取遇到异常时跳过（如非常规文本格式）\n continue\n \n # 如果项目中至少包含一个代码文件且全部代码文件行数均为奇数，则返回True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"所有生成的代码文件的缩进宽度调整为4个空格","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义需要检查的代码文件的后缀名\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.html', \n '.css', '.vue', '.json', '.java', '.cpp', '.c', \n '.h', '.cs', '.go', '.rb', '.php'\n }\n \n # 忽略掉常见的无关文件夹\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'venv', 'env'}\n file_checked = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以便 os.walk 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n count_4 = 0 # 前导空格为 4 的倍数的行数 (例如 4, 8, 12)\n count_2_not_4 = 0 # 前导空格为非 4 的倍数的偶数行数 (例如 2, 6, 10)\n count_tab = 0 # 包含 Tab 的行数\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 忽略空白或纯换行符的行\n if not line.strip():\n continue\n \n # 提取行首空白字符\n leading_whitespace = \"\"\n for char in line:\n if char in (' ', '\\t'):\n leading_whitespace += char\n else:\n break\n \n # 分析缩进行为\n if '\\t' in leading_whitespace:\n count_tab += 1\n else:\n spaces = len(leading_whitespace)\n if spaces > 0:\n if spaces % 4 == 0:\n count_4 += 1\n elif spaces % 2 == 0:\n count_2_not_4 += 1\n except Exception:\n # 如果由于编码或其他问题无法读取文件，则跳过\n continue\n \n total_even_spaces = count_4 + count_2_not_4\n \n # 判断当前代码文件是否满足4个空格缩进的要求\n if total_even_spaces > 0:\n file_checked = True\n \n if total_even_spaces < 3:\n # 对于极其简短的文件，直接比较两者数量即可\n if count_4 < count_2_not_4:\n return False\n else:\n # 对于常规文件，若 4的倍数 行数占比少于 60%，意味着使用了大量的如 2空格 缩进\n if count_4 / total_even_spaces < 0.6:\n return False\n \n # 如果文件使用了 Tab 作为主要缩进方式，认为违反 4个空格要求\n if count_tab > 0 and count_tab >= max(total_even_spaces, 1):\n return False\n\n # 若至少检查过一个代码文件且全部满足要求，则返回 True\n return file_checked","tags":["Style","Layout"]},{"constraint":"请确保你生成的每一个代码文件满足：所有代码行均需配备注释，并且注释仅能写在对应代码行的上方，不可出现在代码的右面或底端","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n has_valid_files = False\n \n # 支持的前端及后端常见代码文件后缀\n valid_extensions = ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.py']\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_extensions:\n continue\n \n has_valid_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n line_is_comment_list = []\n in_multiline_comment = False\n \n for i, line in enumerate(lines):\n stripped = line.strip()\n # 遇到空行直接记录并跳过，不将其算作任何逻辑判断层面的内容\n if not stripped:\n line_is_comment_list.append(False)\n continue\n \n is_comment_line = False\n \n # 判断当前行是否属于注释（包含多行注释内部判断）\n if in_multiline_comment:\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.vue'] and '*/' in stripped:\n in_multiline_comment = False\n # 确保注释闭合标签之后没有伴随同行代码\n if stripped.rsplit('*/', 1)[1].strip():\n return False \n is_comment_line = True\n elif ext in ['.html', '.vue'] and '-->' in stripped:\n in_multiline_comment = False\n # 确保注释闭合标签之后没有伴随同行代码\n if stripped.rsplit('-->', 1)[1].strip():\n return False\n is_comment_line = True\n else:\n is_comment_line = True\n else:\n if stripped.startswith('//'):\n is_comment_line = True\n elif ext == '.py' and stripped.startswith('#'):\n is_comment_line = True\n elif stripped.startswith('/*'):\n if '*/' not in stripped:\n is_comment_line = True\n in_multiline_comment = True\n else:\n if stripped.rsplit('*/', 1)[1].strip():\n return False\n is_comment_line = True\n elif stripped.startswith('<!--'):\n if '-->' not in stripped:\n is_comment_line = True\n in_multiline_comment = True\n else:\n if stripped.rsplit('-->', 1)[1].strip():\n return False\n is_comment_line = True\n elif stripped.startswith('-*- coding') or stripped.startswith('# -*- coding'):\n is_comment_line = True\n \n # 如果当前行为代码行，需验证\"不允许出现同行末尾注释\" 及 \"上一非空行必须是注释\"\n if not is_comment_line:\n # 使用正则过滤掉所有字符串常量内容以及常见的URL结构，避免把代码中合法的 `//` 或字符串里的 `/*` 误认为是行尾注释\n temp_line = re.sub(r'\".*?\"', '\"\"', stripped)\n temp_line = re.sub(r\"'.*?'\", \"''\", temp_line)\n temp_line = re.sub(r'`.*?`', '``', temp_line)\n temp_line = temp_line.replace('http://', '').replace('https://', '')\n \n if ext == '.py' and '#' in temp_line:\n return False\n if ext != '.py':\n if '//' in temp_line or '/*' in temp_line or '<!--' in temp_line:\n return False\n \n prev_is_comment = False\n # 往回找距离当前最近的那个\"非空行\"，检查其是否为注释\n for j in range(i-1, -1, -1):\n if not lines[j].strip():\n continue\n prev_is_comment = line_is_comment_list[j]\n break\n \n # 如果上方的有效行不是注释，则说明这行代码没有配备上方注释，即刻判定失败\n if not prev_is_comment:\n return False\n \n line_is_comment_list.append(is_comment_line)\n \n return has_valid_files","tags":["Style","Layout"]},{"constraint":"在每次对项目进行改动后，改为更新progress.md文件以记录本轮完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"务必保证你生成的每个代码文件的整体行数（算上空行）不是质数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n ** 0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n has_checked_any_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按 \\n 分割，能够准确计算出包含空行的总行数\n num_lines = len(content.split('\\n'))\n \n # 如果有任何一个代码文件的行数是质数，则直接违反了要求\n if is_prime(num_lines):\n return False\n \n has_checked_any_file = True\n except Exception:\n # 忽略读取失败的文件（例如编码问题等）\n continue\n\n # 如果没有任何代码文件，通常认为无法证实满足要求；如果有则说明全部非质数，满足条件\n return has_checked_any_file","tags":["Quantity","Exact Value"]},{"constraint":"务必保证你编写的代码文件中，引用文件时排斥绝对路径，统一使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我开发一个多步骤引导表单的骨架。我需要一个页面，顶部有一个步骤指示器（比如标明 Step 1, Step 2, Step 3），中间是表单内容区域，底部有'上一步'和'下一步'按钮。现在先不用写具体的复杂的表单字段，只要在三个步骤里分别放一个标题和简单的文字占位符就可以。注意逻辑交互：第一步时'上一步'按钮要置灰或隐藏，最后一步的'下一步'按钮应该变成'提交'。切换步骤时，内容区域希望有一点简单的过渡效果。另外，后续所有相关操作都需要始终遵循以下要求：你所提供的所有代码文件，其缩进宽度需设置为2个空格；同时你生成的所有代码文件必须做到每一行代码都有注释，每条注释的位置必须限定在相关代码的上面，避免放在右方或下方。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"所有的回复内容都应以语气词\\\"喵～\\\"起头，这是必须遵守的规定\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除字符串首尾的空白字符，然后检查是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处增加语气词\\\"汪～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容末尾可能存在的空白字符（如空格、换行符等）\\n cleaned_response = response.rstrip()\\n \\n # 判断清理后的回复末尾是否以\\\"汪～\\\"结尾\\n return cleaned_response.endswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"代码小助手\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复内容中是否包含“我”字\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义前端项目开发中常见的代码文件扩展名\\n code_exts = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n has_code_file = False\\n \\n # 若工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 检查文件的第一行是否包含指定的编码声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于某种异常（如权限不足等）无法读取文件，则视为不满足要求\\n return False\\n \\n # 如果找到了代码文件，并且所有代码文件第一行都包含了声明，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名，主要涵盖前端及常见编程语言\\n code_extensions = (\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.vue', '.py', '.json', '.xml', '.txt'\\n )\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith(code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试使用常见的不同编码格式读取文件内容\\n content = None\\n for encoding in ['utf-8', 'gbk', 'utf-16']:\\n try:\\n with open(file_path, 'r', encoding=encoding) as f:\\n content = f.read()\\n break\\n except Exception:\\n continue\\n \\n # 如果所有编码都无法读取，则视为异常或忽略该文件\\n if content is None:\\n continue\\n \\n # 计算文件总行数（算上空行）\\n # 文本模式读取时，换行符被统一转换为了 '\\\\n'，直接按 '\\\\n' 分割即可精确计算行数\\n lines = content.split('\\\\n')\\n total_lines = len(lines)\\n \\n # 如果任意一个代码文件行数为偶数，则直接判定为不满足要求\\n if total_lines % 2 == 0:\\n return False\\n \\n # 如果项目中找到了至少一个代码文件，并且所有代码文件的总行数均为奇数，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你所提供的所有代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码扩展名（前端表单及可能的通用代码格式）\\n code_extensions = (\\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.vue', '.json', \\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs'\\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(code_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n indents = set()\\n for line in lines:\\n # 忽略完全空的行\\n if not line.strip():\\n continue\\n \\n # 获取行首缩进部分的子串\\n l_strip = line.lstrip()\\n leading_whitespace = line[:len(line) - len(l_strip)]\\n \\n # 若使用了制表符作为缩进，不符合“2个空格”的要求\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 过滤掉常见的注释符号及文件头声明，以防止干扰正常代码缩进判断\\n if l_strip.startswith(('*', '//', '<!--', '/*', '#', '-*-')):\\n continue\\n \\n if leading_whitespace:\\n indents.add(len(leading_whitespace))\\n \\n # 对单个文件的所有缩进级别进行规则判断\\n if indents:\\n # 若所有的缩进全都是 4 的倍数，说明缩进基数是 4 个空格（或更宽），不符合 2 个空格要求\\n if all(i % 4 == 0 for i in indents):\\n return False\\n \\n # 若没有任何偶数缩进，说明完全没有 2 个空格的特征体现\\n if not any(i % 2 == 0 for i in indents):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件必须做到每一行代码都有注释，每条注释的位置必须限定在相关代码的上面，避免放在右方或下方\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 剔除代码行内的字符串常量值，防止如 URL 中的 // 被误判为注释\\n def remove_strings(line):\\n line = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', line)\\n line = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", line)\\n line = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', line)\\n return line\\n\\n # 常见代码文件后缀及所需检测的注释体系映射\\n # 'html': 支持 <!-- ... -->\\n # 'c': 支持 // 和 /* ... */\\n # 'py': 支持 #\\n code_extensions = {\\n '.html': ['html', 'c'],\\n '.htm': ['html', 'c'],\\n '.vue': ['html', 'c'],\\n '.css': ['c'],\\n '.js': ['c'],\\n '.ts': ['c'],\\n '.jsx': ['c'],\\n '.tsx': ['c'],\\n '.py': ['py'],\\n '.java': ['c'],\\n '.c': ['c'],\\n '.cpp': ['c'],\\n '.h': ['c'],\\n '.go': ['c'],\\n '.rs': ['c']\\n }\\n \\n file_found = False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n supports_html = 'html' in code_extensions[ext]\\n supports_c = 'c' in code_extensions[ext]\\n supports_py = 'py' in code_extensions[ext]\\n \\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n state = {'in_html': False, 'in_c': False}\\n prev_line_type = None\\n \\n for line in lines:\\n s = line.strip()\\n if not s:\\n continue\\n \\n line_type = 'code'\\n \\n # 处于多行注释状态\\n if state['in_html']:\\n if '-->' in s:\\n idx = s.find('-->')\\n rest = s[idx+3:].strip()\\n state['in_html'] = False\\n line_type = 'mixed' if rest else 'comment'\\n else:\\n line_type = 'comment'\\n elif state['in_c']:\\n if '*/' in s:\\n idx = s.find('*/')\\n rest = s[idx+2:].strip()\\n state['in_c'] = False\\n # 兼容 React JSX 单行带有结束花括号 `}` 的情况\\n line_type = 'mixed' if (rest and rest != '}') else 'comment'\\n else:\\n line_type = 'comment'\\n else:\\n # 不在多行注释状态下，检查行首类型\\n if supports_html and s.startswith('<!--'):\\n if '-->' in s:\\n idx = s.find('-->')\\n rest = s[idx+3:].strip()\\n line_type = 'mixed' if rest else 'comment'\\n else:\\n state['in_html'] = True\\n line_type = 'comment'\\n elif supports_c and (s.startswith('/*') or s.startswith('{/*')):\\n end_str = '*/}' if s.startswith('{/*') and '*/}' in s else '*/'\\n if end_str in s:\\n idx = s.find(end_str)\\n rest = s[idx+len(end_str):].strip()\\n line_type = 'mixed' if (rest and rest != '}') else 'comment'\\n else:\\n state['in_c'] = True\\n line_type = 'comment'\\n elif supports_c and s.startswith('//'):\\n line_type = 'comment'\\n elif supports_py and s.startswith('#'):\\n line_type = 'comment'\\n else:\\n # 既不是单纯的空行，也并非由注释开头，判断是否混杂行右侧/行尾注释\\n s_no_str = remove_strings(s)\\n has_right = False\\n if supports_html and ('<!--' in s_no_str):\\n has_right = True\\n if supports_c and ('/*' in s_no_str):\\n has_right = True\\n if supports_c and ('//' in s_no_str):\\n has_right = True\\n if supports_py and ('#' in s_no_str):\\n has_right = True\\n \\n line_type = 'mixed' if has_right else 'code'\\n\\n # 状态判定\\n if line_type == 'mixed':\\n return False # 注释出现在了右侧混合代码行\\n elif line_type == 'code':\\n if prev_line_type != 'comment':\\n return False # 本行代码缺乏在正上方的专属注释\\n prev_line_type = 'code'\\n elif line_type == 'comment':\\n prev_line_type = 'comment'\\n \\n if not file_found:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"框架有了，现在来完善具体的表单内容并加入实时验证。第一步定为'基本信息'，加入用户名（必填，只能包含字母和数字）和邮箱（必填，必须符合邮箱标准格式）的输入框；第二步定为'联系方式'，加入手机号（必填，需符合11位数字规范）和详细地址文本框。重点挑战是：每一阶段的表单必须完全通过验证才能进入下一步。如果有错误，禁用'下一步'按钮，并在输入框下方实时展示红色的错误提示信息。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边；第二，在每次对项目进行改动后，更新 process-log.md 文件以记录本轮完成的内容是必须执行的操作。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"所有的回复内容都应以语气词\\\"喵～\\\"起头，这是必须遵守的规定\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 去除回复内容左侧可能存在的空白字符（如空格、换行符等）\\n cleaned_response = response.lstrip()\\n # 检查回复内容是否以指定的语气词起头\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处增加语气词\\\"汪～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 清楚末尾的空白字符（换行、空格等）\\n cleaned_response = response.rstrip()\\n \\n # 使用正则匹配结尾。允许“汪～”之后出现句号、感叹号等常规结束标点\\n # [\\\\s。！!\\\\.]*$ 表示允许在结尾处出现任意数量的空白符、中英文句号和感叹号\\n pattern = r'汪～[\\\\s。！!\\\\.]*$'\\n \\n match = re.search(pattern, cleaned_response)\\n \\n return match is not None\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"代码小助手\\\"\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n # 判断最后一轮回复中是否包含字符\\\"我\\\"\\n # 如果不包含\\\"我\\\"，说明自称均已替换，返回 True；否则返回 False\\n return \\\"我\\\" not in response\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，Web前端项目的常见代码文件扩展名\\n code_exts = {'.html', '.css', '.js'}\\n found_code_file = False\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取文件或发生编码异常，视为验证失败\\n return False\\n \\n # 确保至少检查过一个代码文件，否则说明根本没有生成代码\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的前后端及其他代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.html', '.css', '.java', '.c', '.cpp', '.cs', \\n '.ts', '.jsx', '.tsx', '.vue', '.php', '.rb', '.go', '.rs'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并转小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件或代码文件的备份（如 .js.bak, .html_bak 等）\\n is_code_file = ext in code_extensions\\n if not is_code_file:\\n for ce in code_extensions:\\n if file.lower().endswith(ce + '.bak') or file.lower().endswith(ce + '_bak'):\\n is_code_file = True\\n break\\n\\n if is_code_file:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 读取文件内容，加入异常处理和多编码兼容\\n content = None\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n content = f.read()\\n except Exception:\\n pass\\n except Exception:\\n pass\\n \\n if content is not None:\\n # 使用换行符切分能严格包含尾部空行，精准统计总行数\\n lines = content.split('\\\\n')\\n if len(lines) % 2 == 0:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你所提供的所有代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n\\n # 常见的前端/后端代码文件后缀\\n code_exts = {\\n '.html', '.htm', '.css', '.scss', '.less', \\n '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.py', '.java', '.php', '.go', '.c', \\n '.cpp', '.cs', '.rb', '.swift', '.kt', '.json'\\n }\\n \\n has_checked_any_file = False\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_checked_any_file = True\\n \\n indent_diffs = []\\n prev_indent = 0\\n has_tab = False\\n \\n for line in lines:\\n stripped = line.lstrip('\\\\r\\\\n')\\n # 过滤掉完全为空白的行\\n if not stripped.strip():\\n continue\\n \\n # 检查是否有制表符导致的缩进\\n if stripped.startswith('\\\\t'):\\n has_tab = True\\n break\\n \\n # 计算前导空格数量\\n leading_spaces = len(stripped) - len(stripped.lstrip(' '))\\n diff = leading_spaces - prev_indent\\n \\n # 只有当缩进深度增加时，才统计这个增量，这样能找出基本的缩进单位\\n if diff > 0:\\n indent_diffs.append(diff)\\n \\n prev_indent = leading_spaces\\n \\n if has_tab:\\n return False\\n \\n if indent_diffs:\\n counter = Counter(indent_diffs)\\n # 取出出现频率最高的缩进增量作为基准缩进宽度\\n most_common_diff = counter.most_common(1)[0][0]\\n if most_common_diff != 2:\\n return False\\n \\n # 只要成功检测到了代码文件并且中途未返回False，则代表所有代码文件均满足要求\\n return has_checked_any_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n def get_markers(ext: str):\\n # 根据文件扩展名返回可能的注释前导标记\\n ext = ext.lower()\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java']:\\n return ['//', '/*']\\n elif ext in ['.css']:\\n return ['/*']\\n elif ext in ['.html', '.htm', '.vue']:\\n return ['<!--', '//', '/*']\\n elif ext in ['.py', '.sh', '.rb']:\\n return ['#']\\n else:\\n return ['//', '/*', '<!--', '#']\\n\\n def check_line(line: str, ext: str) -> bool:\\n line_strip = line.strip()\\n # 空行予以忽略\\n if not line_strip:\\n return True\\n # 豁免系统级强制要求的编码声明注释行\\n if 'coding: utf-8' in line:\\n return True\\n \\n markers = get_markers(ext)\\n \\n in_string = False\\n string_char = ''\\n i = 0\\n while i < len(line):\\n c = line[i]\\n if in_string:\\n # 处理转义字符\\n if c == '\\\\\\\\':\\n i += 2\\n continue\\n # 退出字符串\\n if c == string_char:\\n in_string = False\\n else:\\n # 遇到引号进入字符串状态，防止把字符串内的标识误认为注释\\n if c in ['\\\"', \\\"'\\\", '`']:\\n in_string = True\\n string_char = c\\n else:\\n # 检查是否匹配当前文件类型所支持的注释标记\\n if '//' in markers and line[i:i+2] == '//':\\n # 排除 http:// 或 https:// 中的双斜杠\\n if (i >= 5 and line[i-5:i] == 'http:') or (i >= 6 and line[i-6:i] == 'https:'):\\n i += 2\\n continue\\n prefix = line[:i].strip()\\n return len(prefix) > 0\\n \\n if '/*' in markers and line[i:i+2] == '/*':\\n prefix = line[:i].strip()\\n return len(prefix) > 0\\n \\n if '<!--' in markers and line[i:i+4] == '<!--':\\n prefix = line[:i].strip()\\n return len(prefix) > 0\\n \\n if '#' in markers and c == '#':\\n prefix = line[:i].strip()\\n return len(prefix) > 0\\n i += 1\\n \\n # 整行遍历结束后既没找到有效注释，且该行又并非空行，视为未加注释，不满足“逐行注释”要求\\n return False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_exts = ['.js', '.css', '.html', '.htm', '.py', '.ts', '.jsx', '.tsx', '.vue', '.java', '.c', '.cpp']\\n has_code_file = False\\n \\n # 遍历当前工作空间中的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n # 忽略备份文件，只检查当前最新生成的代码\\n if 'bak' in file.lower() or 'backup' in file.lower():\\n continue\\n \\n filepath = os.path.join(root, file)\\n has_code_file = True\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 若文件由于编码等原因无法读取，直接跳过处理\\n continue\\n \\n # 逐行检查该文件是否遵循了左代码、右注释的排布规则\\n for line in lines:\\n if not check_line(line, ext):\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 process-log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"很好，接下来我们在表单里引入条件分支逻辑。在第一步的最上方加入一个名为'注册身份'的必选单选按钮组，选项为'个人'和'企业'。要求根据这里的选择动态改变第二步的内容：如果第一步选了'个人'，第二步依然显示你刚才做的'手机号'和'地址'；但如果第一步选了'企业'，第二步的内容要完全变身，改成填写'企业名称'、'营业执照编号'以及一个'企业规模'的下拉菜单。请处理好切换身份时之前遗留输入数据的清理问题。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，所有生成的代码文件的缩进宽度调整为4个空格；第二，务必保证你生成的每个代码文件的整体行数（算上空行）不是质数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"所有的回复内容都应以语气词\\\"喵～\\\"起头，这是必须遵守的规定\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除首部的空白字符后，判断是否以“喵～”开头\\n cleaned_response = response.lstrip()\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处增加语气词\\\"汪～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除回复文本末尾可能的空白字符以及常见的标点符号（如句号、感叹号等）\\n cleaned_response = re.sub(r'[\\\\s。！，\\\\.,!]+$', '', response)\\n \\n # 判断结尾是否为\\\"汪～\\\"或者兼容半角波浪号的\\\"汪~\\\"\\n return cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"代码小助手\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含“我”字\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"-*- coding: utf-8 -*-\\\"\\n # 定义常见的代码文件后缀集合，排除掉.json, .md, .txt等不适合加注释的代码外文件\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.py', '.java', '.c', '.cpp', '.h', '.cs', \\n '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.sh'\\n }\\n \\n code_file_count = 0\\n for root, _, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n code_file_count += 1\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行\\n first_line = f.readline()\\n # 判断给定的编码声明是否在第一行中\\n if target_statement not in first_line:\\n return False\\n except Exception:\\n # 如果文件不可读或编码异常等，无法确定满足要求，返回False\\n return False\\n \\n # 如果找到了至少一个代码文件，并且全满足要求，则返回True\\n return code_file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.py', '.java', '.c', '.cpp', \\n '.h', '.md', '.json', '.txt', '.xml', '.vue', '.ts', \\n '.jsx', '.tsx', '.sh', '.yaml', '.yml'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 由于可能出现备份文件，例如直接加 .bak，此处也可以直接针对常见代码格式判断\\n # 如果扩展名属于代码类型，则进行统计验证\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 算上空行后的总行数\\n total_lines = len(lines)\\n \\n # 检查是否为奇数值，0通常视作偶数且意味着文件为空，不符合要求\\n if total_lines % 2 == 0:\\n return False\\n except UnicodeDecodeError:\\n # 如果不是utf-8编码（例如某些意外二进制文件），跳过\\n continue\\n except Exception:\\n # 跳过无法打开或其他异常的文件\\n continue\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件的缩进宽度调整为4个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n \\n # 定义常见的代码文件扩展名\\n code_exts = {'.html', '.css', '.js', '.py', '.java', '.c', '.cpp', '.ts', '.tsx', '.jsx', '.vue'}\\n # 排除备份文件的关键字\\n skip_keywords = ['backup', 'bak', 'old', 'copy', 'origin', 'prev']\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n lower_file = file.lower()\\n # 跳过第一轮规定备份的文件，以免受历史版本的缩进规则（2空格）影响\\n if any(kw in lower_file for kw in skip_keywords):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n indents = []\\n has_tab = False\\n for line in lines:\\n # 忽略空行\\n if not line.strip():\\n continue\\n \\n # 忽略以 '*' 开头的行（往往是多行注释的星号对齐内容）\\n if line.lstrip().startswith('*'):\\n continue\\n \\n # 计算前导空白\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n if '\\\\t' in leading_whitespace:\\n has_tab = True\\n break\\n \\n spaces = len(leading_whitespace)\\n # 仅记录具有缩进的行\\n if spaces > 0:\\n indents.append(spaces)\\n \\n # 规则明确要求“4个空格”，存在制表符缩进则不符合\\n if has_tab:\\n return False\\n \\n if not indents:\\n continue\\n \\n # 检查不能被4整除的缩进情况（兼容部分对齐操作，容忍30%比例的非标准缩进行）\\n invalid_count = sum(1 for s in indents if s % 4 != 0)\\n if invalid_count > len(indents) * 0.3:\\n return False\\n \\n # 计算相邻缩进层级的差值，找出众数作为主体缩进宽度\\n diffs = []\\n for i in range(1, len(indents)):\\n diff = abs(indents[i] - indents[i-1])\\n if diff > 0:\\n diffs.append(diff)\\n \\n if diffs:\\n counter = Counter(diffs)\\n most_common = counter.most_common(1)[0][0]\\n # 主体缩进宽度增幅必须是4的倍数（最常见的是4本身）\\n if most_common not in (4, 8):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义我们要检查的常见代码文件后缀\\n code_extensions = {\\n '.py', '.js', '.ts', '.java', '.cpp', '.c', '.cs', \\n '.html', '.xml', '.vue', '.css', '.jsx', '.tsx'\\n }\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n\\n for i, line in enumerate(lines):\\n line_strip = line.strip()\\n # 跳过空行\\n if not line_strip:\\n continue\\n \\n # 按照之前轮次的指令，第一行必须包含编码声明，可放宽单独成行的限制\\n if i == 0 and 'coding: utf-8' in line_strip:\\n continue\\n\\n # 移除常见协议头网址，防止将链接误判为 // 注释\\n line_for_check = line_strip.replace('http://', '').replace('https://', '')\\n \\n if ext == '.py':\\n # 如果行首就是 #，则是单独一行的注释；如果没有 #，则缺少注释\\n if line_for_check.startswith('#'):\\n return False\\n if '#' not in line_for_check:\\n return False\\n \\n elif ext in ['.js', '.ts', '.java', '.cpp', '.c', '.cs', '.jsx', '.tsx']:\\n if line_for_check.startswith('//') or line_for_check.startswith('/*'):\\n return False\\n if '//' not in line_for_check and '/*' not in line_for_check:\\n return False\\n \\n elif ext in ['.html', '.xml', '.vue']:\\n # HTML中可能混入JS/CSS，因此注释可能是 <!--, //, /* 之一\\n if line_for_check.startswith('<!--') or line_for_check.startswith('//') or line_for_check.startswith('/*'):\\n return False\\n if '<!--' not in line_for_check and '//' not in line_for_check and '/*' not in line_for_check:\\n return False\\n \\n elif ext == '.css':\\n if line_for_check.startswith('/*'):\\n return False\\n if '/*' not in line_for_check:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 process-log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"务必保证你生成的每个代码文件的整体行数（算上空行）不是质数\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 常见前端与普通代码文件的扩展名集合\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 以换行符分割计算总行数，能够准确反映包含空行的所有行数\\n lines_count = len(content.split('\\\\n'))\\n \\n # 只要有一个代码文件的行数为质数，则违背了要求\\n if is_prime(lines_count):\\n return False\\n except Exception:\\n # 遇到不可读的文件或编码错误等问题则跳过，防止验证中断\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在来完善第三步，我们要加入图片上传和预览功能。把第三步的标题改为'资质证明上传'。在这一步里设计一个文件上传区域，核心要求是：用户选择一张本地图片后，界面上要立刻显示该图片的缩略图预览，并且图片旁边要有个'删除'按钮可以把选中的图移除。同时加上前端限制规则：仅允许上传 jpg/png 格式且大小不超过 2MB 的图片，如果不符合要求要给出醒目的错误提示，并且不显示预览。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，在每次对项目进行改动后，改为更新progress.md文件以记录本轮完成的内容；第二，务必保证你编写的代码文件中，引用文件时排斥绝对路径，统一使用相对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"所有的回复内容都应以语气词\\\"喵～\\\"起头，这是必须遵守的规定\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否以\\\"喵～\\\"起头\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复首部可能存在的空白字符，然后判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处增加语气词\\\"汪～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否在结尾处增加了语气词\\\"汪～\\\"\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容末尾可能存在的空白字符（如换行符 \\\\n、空格等）\\n clean_response = response.rstrip()\\n \\n # 判断清理后的回复是否以指定语气词结尾\\n return clean_response.endswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"代码小助手\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查助手的回复中是否依然包含“我”字\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前端与相关代码文件扩展名\\n code_extensions = {\\\".html\\\", \\\".htm\\\", \\\".css\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".vue\\\", \\\".py\\\"}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查第一行中是否包含要求的声明（若文件为空，first_line为\\\"\\\"，也会直接判断为False）\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取、编码错误或其他异常情况，判断为不满足要求\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件扩展名，覆盖了指令中涉及到的前端开发及文档格式\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.json', '.md', '.py', '.txt', '.xml', '.java', '.c', \\n '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 空文件行数为0，0是偶数，不满足奇数值要求\\n if not content:\\n return False\\n \\n # 使用 '\\\\n' 分割可以精确统计出包含末尾空行的所有行数\\n line_count = len(content.split('\\\\n'))\\n \\n # 检查行数是否为奇数，如果是偶数则返回 False\\n if line_count % 2 == 0:\\n return False\\n \\n except UnicodeDecodeError:\\n # 如果不是普通的文本文件（可能编码有异或者存在不识别的二进制数据），则跳过\\n continue\\n except Exception:\\n # 跳过其它读取异常\\n continue\\n \\n # 如果项目中存在代码文件并且全都是奇数行，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件的缩进宽度调整为4个空格\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n \\n # 常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.js', '.css', '.py', '.java', '.c', '.cpp', \\n '.ts', '.vue', '.jsx', '.tsx', '.cs', '.php', '.go', \\n '.rb', '.rs', '.swift', '.kt'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的依赖或版本控制目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n # 过滤掉完全为空白字符的行\\n lines = [line for line in lines if line.strip()]\\n if not lines:\\n continue\\n \\n has_code_file = True\\n indents = []\\n \\n for line in lines:\\n # 如果某行以制表符开头，说明没有纯粹使用空格缩进，不满足要求\\n if line.startswith('\\\\t'):\\n return False\\n \\n # 计算前导空格数\\n leading_spaces = len(line) - len(line.lstrip(' '))\\n indents.append(leading_spaces)\\n \\n # 计算相邻两行的正向缩进差值\\n diffs = []\\n for i in range(1, len(indents)):\\n diff = indents[i] - indents[i-1]\\n if diff > 0:\\n diffs.append(diff)\\n \\n # 过滤掉可能是注释、对齐造成的1个空格的差值，保留块级缩进（>=2）\\n block_diffs = [d for d in diffs if d >= 2]\\n \\n if block_diffs:\\n c = Counter(block_diffs)\\n max_count = c.most_common(1)[0][1]\\n # 获取出现频率最高的所有差值\\n most_common_vals = [val for val, count in c.items() if count == max_count]\\n \\n # 4个空格必须是最常见的缩进步长之一，否则判定为失败\\n if 4 not in most_common_vals:\\n return False\\n else:\\n # 如果没有块级递增缩进，直接检查存在的非零前导空格是否满足是4的倍数\\n positive_indents = [ind for ind in indents if ind > 0]\\n if positive_indents:\\n if min(positive_indents) % 4 != 0:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须做到逐行注释，且所有注释均需强制置于同一行代码的右边，禁止写在代码行的上边或下边\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义不同代码文件扩展名对应的常见注释符号\\n # HTML和Vue等文件考虑到其中可能包含内部的script和style标签，所以加入//和/*进行兼容支持\\n code_exts = {\\n '.js': ['//', '/*'],\\n '.ts': ['//', '/*'],\\n '.jsx': ['//', '/*'],\\n '.tsx': ['//', '/*'],\\n '.html': ['<!--', '//', '/*'],\\n '.htm': ['<!--', '//', '/*'],\\n '.css': ['/*'],\\n '.scss': ['/*'],\\n '.less': ['/*'],\\n '.vue': ['<!--', '//', '/*'],\\n '.py': ['#'],\\n '.java': ['//', '/*'],\\n '.cpp': ['//', '/*'],\\n '.c': ['//', '/*'],\\n '.h': ['//', '/*'],\\n '.go': ['//', '/*'],\\n '.rs': ['//', '/*'],\\n '.php': ['//', '/*', '#', '<!--'],\\n '.sh': ['#']\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for idx, line in enumerate(lines):\\n s_line = line.strip()\\n \\n # 忽略空行\\n if not s_line:\\n continue\\n \\n # 针对系统提示中的第一行强制编码声明进行特殊放行\\n if \\\"-*- coding: utf-8 -*-\\\" in s_line and idx == 0:\\n continue\\n \\n # 剔除常见的可能引起//误判的网络协议前缀\\n s_line_clean = s_line.replace(\\\"http://\\\", \\\"\\\").replace(\\\"https://\\\", \\\"\\\")\\n \\n symbols = code_exts[ext]\\n has_comment = False\\n is_comment_start = False\\n \\n for sym in symbols:\\n if sym in s_line_clean:\\n has_comment = True\\n # 如果该行去除了首尾空格后仍然以注释符号开头，说明这是一个上方或下方的独立注释行\\n if s_line_clean.startswith(sym):\\n is_comment_start = True\\n break\\n \\n # 如果这行代码中完全没有包含注释符号，说明没有逐行注释\\n if not has_comment:\\n return False\\n \\n # 如果这行代码是以注释符号开头的纯注释，违反了“不能写在代码行的上边或下边”的规定\\n if is_comment_start:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，改为更新progress.md文件以记录本轮完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"务必保证你生成的每个代码文件的整体行数（算上空行）不是质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"判断一个整数是否为质数\\\"\\\"\\\"\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 定义常见的代码文件及相关文本文件的扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs',\\n '.md', '.txt', '.json', '.xml', '.yml', '.yaml',\\n '.sh', '.bat', '.ps1', '.sql'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的隐藏目录（如 .git, .idea 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用 \\\\n 分割，精确还原主流代码编辑器中的总行数（自动包含首尾和中间的空行）\\n num_lines = len(content.split('\\\\n'))\\n \\n # 如果有任何一个代码文件的行数是质数，判定为不满足要求\\n if is_prime(num_lines):\\n return False\\n except UnicodeDecodeError:\\n # 忽略无法以 utf-8 读取的非纯文本文件\\n continue\\n except Exception:\\n # 忽略其他读取异常\\n continue\\n\\n # 遍历完全部代码文件后，未发现任何文件的行数为质数，判定要求满足\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"务必保证你编写的代码文件中，引用文件时排斥绝对路径，统一使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，我们来增加后台分析追踪功能以及完成最终提交体验。请在代码里添加埋点逻辑：偷偷记录用户在每个步骤界面停留的具体时间（按秒计算），以及在整个填表过程中一共触发了几次表单验证错误。当用户在第三步点击\\\"提交\\\"时，不需要发真实的后端请求，而是弹出一个信息确认对话框，里面整齐地罗列出前三步填写的全部信息（包括根据身份不同填写的信息、图片文件名），并在最下方附加一个\\\"追踪数据汇总\\\"区域，显示总停留时长和错误触发总次数。关闭对话框后，整个多步表单自动重置回最初始的状态。另外，从本轮开始后续所有相关操作需始终遵循以下要求：第一，请确保你生成的每一个代码文件满足：所有代码行均需配备注释，并且注释仅能写在对应代码行的上方，不可出现在代码的右面或底端；第二，你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"所有的回复内容都应以语气词\\\"喵～\\\"起头，这是必须遵守的规定\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复内容是否以语气词\\\"喵～\\\"起头。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，代表是否满足要求\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容开头的可能存在的空白字符（如空格、换行符等）后进行匹配\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处增加语气词\\\"汪～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符\\n clean_response = response.strip()\\n \\n # 使用正则表达式检查回复结尾是否是\\\"汪～\\\"\\n # [\\\\W_]*$ 用于匹配结尾可能附带的非单词字符（如句号、感叹号等标点符号）\\n match = re.search(r'汪～[\\\\W_]*$', clean_response)\\n \\n return bool(match)\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"代码小助手\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否将自称\\\"我\\\"替换为了\\\"代码小助手\\\"。\\n 严格判断回复内容中是否包含字符\\\"我\\\"。\\n \\\"\\\"\\\"\\n if '我' in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有的代码文件在生成时，第一行一律要求写上给定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n \\n # 遍历项目工作区路径\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查给定的编码声明是否存在于第一行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果读取文件出现异常，直接视为验证失败\\n return False\\n \\n # 如果整个工作区内至少包含一个代码文件且全都满足要求，返回True\\n # 若没有生成任何代码文件，则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能产生的前端代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按换行符分割，真实计算包括空行在内的总行数\\n lines = content.split('\\\\n')\\n \\n # 判断行数是否为偶数\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 文件读取遇到异常时跳过（如非常规文本格式）\\n continue\\n \\n # 如果项目中至少包含一个代码文件且全部代码文件行数均为奇数，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"每次准备修改现有文件时，务必先在当前文件的同级目录下备份一份未曾修改的原文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次对生成代码的测试结束后，你的回复需严格按以下XML结构，分别输出执行结果和描述信息：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件的缩进宽度调整为4个空格\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义需要检查的代码文件的后缀名\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.html', \\n '.css', '.vue', '.json', '.java', '.cpp', '.c', \\n '.h', '.cs', '.go', '.rb', '.php'\\n }\\n \\n # 忽略掉常见的无关文件夹\\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'venv', 'env'}\\n file_checked = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以便 os.walk 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n count_4 = 0 # 前导空格为 4 的倍数的行数 (例如 4, 8, 12)\\n count_2_not_4 = 0 # 前导空格为非 4 的倍数的偶数行数 (例如 2, 6, 10)\\n count_tab = 0 # 包含 Tab 的行数\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 忽略空白或纯换行符的行\\n if not line.strip():\\n continue\\n \\n # 提取行首空白字符\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_whitespace += char\\n else:\\n break\\n \\n # 分析缩进行为\\n if '\\\\t' in leading_whitespace:\\n count_tab += 1\\n else:\\n spaces = len(leading_whitespace)\\n if spaces > 0:\\n if spaces % 4 == 0:\\n count_4 += 1\\n elif spaces % 2 == 0:\\n count_2_not_4 += 1\\n except Exception:\\n # 如果由于编码或其他问题无法读取文件，则跳过\\n continue\\n \\n total_even_spaces = count_4 + count_2_not_4\\n \\n # 判断当前代码文件是否满足4个空格缩进的要求\\n if total_even_spaces > 0:\\n file_checked = True\\n \\n if total_even_spaces < 3:\\n # 对于极其简短的文件，直接比较两者数量即可\\n if count_4 < count_2_not_4:\\n return False\\n else:\\n # 对于常规文件，若 4的倍数 行数占比少于 60%，意味着使用了大量的如 2空格 缩进\\n if count_4 / total_even_spaces < 0.6:\\n return False\\n \\n # 如果文件使用了 Tab 作为主要缩进方式，认为违反 4个空格要求\\n if count_tab > 0 and count_tab >= max(total_even_spaces, 1):\\n return False\\n\\n # 若至少检查过一个代码文件且全部满足要求，则返回 True\\n return file_checked\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的每一个代码文件满足：所有代码行均需配备注释，并且注释仅能写在对应代码行的上方，不可出现在代码的右面或底端\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n has_valid_files = False\\n \\n # 支持的前端及后端常见代码文件后缀\\n valid_extensions = ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.py']\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_extensions:\\n continue\\n \\n has_valid_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n line_is_comment_list = []\\n in_multiline_comment = False\\n \\n for i, line in enumerate(lines):\\n stripped = line.strip()\\n # 遇到空行直接记录并跳过，不将其算作任何逻辑判断层面的内容\\n if not stripped:\\n line_is_comment_list.append(False)\\n continue\\n \\n is_comment_line = False\\n \\n # 判断当前行是否属于注释（包含多行注释内部判断）\\n if in_multiline_comment:\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.vue'] and '*/' in stripped:\\n in_multiline_comment = False\\n # 确保注释闭合标签之后没有伴随同行代码\\n if stripped.rsplit('*/', 1)[1].strip():\\n return False \\n is_comment_line = True\\n elif ext in ['.html', '.vue'] and '-->' in stripped:\\n in_multiline_comment = False\\n # 确保注释闭合标签之后没有伴随同行代码\\n if stripped.rsplit('-->', 1)[1].strip():\\n return False\\n is_comment_line = True\\n else:\\n is_comment_line = True\\n else:\\n if stripped.startswith('//'):\\n is_comment_line = True\\n elif ext == '.py' and stripped.startswith('#'):\\n is_comment_line = True\\n elif stripped.startswith('/*'):\\n if '*/' not in stripped:\\n is_comment_line = True\\n in_multiline_comment = True\\n else:\\n if stripped.rsplit('*/', 1)[1].strip():\\n return False\\n is_comment_line = True\\n elif stripped.startswith('<!--'):\\n if '-->' not in stripped:\\n is_comment_line = True\\n in_multiline_comment = True\\n else:\\n if stripped.rsplit('-->', 1)[1].strip():\\n return False\\n is_comment_line = True\\n elif stripped.startswith('-*- coding') or stripped.startswith('# -*- coding'):\\n is_comment_line = True\\n \\n # 如果当前行为代码行，需验证\\\"不允许出现同行末尾注释\\\" 及 \\\"上一非空行必须是注释\\\"\\n if not is_comment_line:\\n # 使用正则过滤掉所有字符串常量内容以及常见的URL结构，避免把代码中合法的 `//` 或字符串里的 `/*` 误认为是行尾注释\\n temp_line = re.sub(r'\\\".*?\\\"', '\\\"\\\"', stripped)\\n temp_line = re.sub(r\\\"'.*?'\\\", \\\"''\\\", temp_line)\\n temp_line = re.sub(r'`.*?`', '``', temp_line)\\n temp_line = temp_line.replace('http://', '').replace('https://', '')\\n \\n if ext == '.py' and '#' in temp_line:\\n return False\\n if ext != '.py':\\n if '//' in temp_line or '/*' in temp_line or '<!--' in temp_line:\\n return False\\n \\n prev_is_comment = False\\n # 往回找距离当前最近的那个\\\"非空行\\\"，检查其是否为注释\\n for j in range(i-1, -1, -1):\\n if not lines[j].strip():\\n continue\\n prev_is_comment = line_is_comment_list[j]\\n break\\n \\n # 如果上方的有效行不是注释，则说明这行代码没有配备上方注释，即刻判定失败\\n if not prev_is_comment:\\n return False\\n \\n line_is_comment_list.append(is_comment_line)\\n \\n return has_valid_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，改为更新progress.md文件以记录本轮完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"务必保证你生成的每个代码文件的整体行数（算上空行）不是质数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n ** 0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n has_checked_any_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按 \\\\n 分割，能够准确计算出包含空行的总行数\\n num_lines = len(content.split('\\\\n'))\\n \\n # 如果有任何一个代码文件的行数是质数，则直接违反了要求\\n if is_prime(num_lines):\\n return False\\n \\n has_checked_any_file = True\\n except Exception:\\n # 忽略读取失败的文件（例如编码问题等）\\n continue\\n\\n # 如果没有任何代码文件，通常认为无法证实满足要求；如果有则说明全部非质数，满足条件\\n return has_checked_any_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"务必保证你编写的代码文件中，引用文件时排斥绝对路径，统一使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["打开页面观察初始状态。检查页面顶部是否有步骤指示器（如 Step 1, Step 2, Step 3），中间是否显示第一步的内容（包含'注册身份'单选框以及用户名、邮箱输入框），底部'上一步'按钮必须处于置灰禁用状态或被隐藏，同时存在'下一步'按钮。","在第一步界面直接点击'下一步'，或输入包含特殊字符的用户名、格式不规范的邮箱。检查界面是否在存在错误的输入框下方实时展示红色的错误提示信息，且此时'下一步'按钮被禁用，无法进入第二步。","在第一步将'注册身份'选择为'个人'，输入纯字母数字的用户名和正确格式的邮箱，点击'下一步'。检查页面是否成功进入第二步（顶部指示器相应高亮），'上一步'按钮变为正常可用状态，且中间表单内容显示为'联系方式'，仅包含手机号和详细地址输入框。","点击'上一步'回到第一步，将'注册身份'切换为'企业'，再次点击'下一步'。检查第二步表单内容是否动态替换为企业信息，即显示'企业名称'、'营业执照编号'的文本框以及'企业规模'的下拉菜单，不再显示手机号和地址输入框。","在第二步的企业表单中输入部分内容，然后点击'上一步'回到第一步，将身份重新切回'个人'并进入第二步。检查第二步此时是否清空了之前在'企业'身份下或更早前'个人'身份下遗留的填写内容，确认脏数据已被正确清理。","在身份为'个人'的第二步中，在手机号输入框填入非11位纯数字的内容（如包含字母或少于11位数字）。检查手机号输入框下方是否展示红色的错误提示，并且阻止用户通过点击'下一步'进入第三步。","在第二步填写正确规范的信息后点击'下一步'。观察页面切换瞬间，内容区域是否具备一定的过渡效果（如滑动或淡入淡出动画）。进入第三步后，检查页面标题是否显示'资质证明上传'，且底部'下一步'按钮已明确变为'提交'按钮。","在第三步的图片上传区域，选择上传一个非 jpg/png 格式的文件（如 pdf、txt 等）或者大于 2MB 的图片文件。检查界面是否弹出或显示醒目的错误提示信息，同时确认上传区域没有显示该错误文件的缩略图预览。","在第三步选择一张小于 2MB 的合法 jpg 或 png 图片上传。检查界面是否立刻渲染出该图片的缩略图预览，并且图片旁边提供了一个'删除'按钮。点击该'删除'按钮，确认图片预览彻底消失且上传区域文件被移除。","在第三步完成合规的图片上传后，点击底部的'提交'按钮。检查界面是否弹出一个信息确认对话框，对话框中需整齐罗列出前三步填写的全部有效信息（包括基本信息、根据当前身份填写的详细信息，以及上传的图片文件名）。","检查提交弹出的确认对话框最下方，是否存在'追踪数据汇总'区域。该区域必须明确显示用户在第一、二、三步页面各自停留的具体时长（按秒计算），并且显示的表单验证错误触发总次数需与填表过程中的实际报错次数相匹配。","点击关闭确认对话框（或点击对话框遮罩层关闭）。检查关闭后，整个多步骤表单是否自动重置回最原始的初始状态（包含步骤指示器回到第一步、所有输入框及单选框清空重置、文件清空、'上一步'按钮再次置灰禁用或隐藏）。","在提交确认对话框中，验证“追踪数据汇总”包含“总停留时长”，且该值与第一/二/三步停留时长之和在可接受误差范围内一致。"],"string":"[\n \"打开页面观察初始状态。检查页面顶部是否有步骤指示器（如 Step 1, Step 2, Step 3），中间是否显示第一步的内容（包含'注册身份'单选框以及用户名、邮箱输入框），底部'上一步'按钮必须处于置灰禁用状态或被隐藏，同时存在'下一步'按钮。\",\n \"在第一步界面直接点击'下一步'，或输入包含特殊字符的用户名、格式不规范的邮箱。检查界面是否在存在错误的输入框下方实时展示红色的错误提示信息，且此时'下一步'按钮被禁用，无法进入第二步。\",\n \"在第一步将'注册身份'选择为'个人'，输入纯字母数字的用户名和正确格式的邮箱，点击'下一步'。检查页面是否成功进入第二步（顶部指示器相应高亮），'上一步'按钮变为正常可用状态，且中间表单内容显示为'联系方式'，仅包含手机号和详细地址输入框。\",\n \"点击'上一步'回到第一步，将'注册身份'切换为'企业'，再次点击'下一步'。检查第二步表单内容是否动态替换为企业信息，即显示'企业名称'、'营业执照编号'的文本框以及'企业规模'的下拉菜单，不再显示手机号和地址输入框。\",\n \"在第二步的企业表单中输入部分内容，然后点击'上一步'回到第一步，将身份重新切回'个人'并进入第二步。检查第二步此时是否清空了之前在'企业'身份下或更早前'个人'身份下遗留的填写内容，确认脏数据已被正确清理。\",\n \"在身份为'个人'的第二步中，在手机号输入框填入非11位纯数字的内容（如包含字母或少于11位数字）。检查手机号输入框下方是否展示红色的错误提示，并且阻止用户通过点击'下一步'进入第三步。\",\n \"在第二步填写正确规范的信息后点击'下一步'。观察页面切换瞬间，内容区域是否具备一定的过渡效果（如滑动或淡入淡出动画）。进入第三步后，检查页面标题是否显示'资质证明上传'，且底部'下一步'按钮已明确变为'提交'按钮。\",\n \"在第三步的图片上传区域，选择上传一个非 jpg/png 格式的文件（如 pdf、txt 等）或者大于 2MB 的图片文件。检查界面是否弹出或显示醒目的错误提示信息，同时确认上传区域没有显示该错误文件的缩略图预览。\",\n \"在第三步选择一张小于 2MB 的合法 jpg 或 png 图片上传。检查界面是否立刻渲染出该图片的缩略图预览，并且图片旁边提供了一个'删除'按钮。点击该'删除'按钮，确认图片预览彻底消失且上传区域文件被移除。\",\n \"在第三步完成合规的图片上传后，点击底部的'提交'按钮。检查界面是否弹出一个信息确认对话框，对话框中需整齐罗列出前三步填写的全部有效信息（包括基本信息、根据当前身份填写的详细信息，以及上传的图片文件名）。\",\n \"检查提交弹出的确认对话框最下方，是否存在'追踪数据汇总'区域。该区域必须明确显示用户在第一、二、三步页面各自停留的具体时长（按秒计算），并且显示的表单验证错误触发总次数需与填表过程中的实际报错次数相匹配。\",\n \"点击关闭确认对话框（或点击对话框遮罩层关闭）。检查关闭后，整个多步骤表单是否自动重置回最原始的初始状态（包含步骤指示器回到第一步、所有输入框及单选框清空重置、文件清空、'上一步'按钮再次置灰禁用或隐藏）。\",\n \"在提交确认对话框中，验证“追踪数据汇总”包含“总停留时长”，且该值与第一/二/三步停留时长之和在可接受误差范围内一致。\"\n]","isConversational":false}}},{"rowIdx":50,"cells":{"task_id":{"kind":"number","value":51,"string":"51"},"repository_policy":{"kind":"string","value":"你作为具备代码生成、文件操作权限的代码智能体，后续所有交互轮次都必须严格遵循以下要求：\n1. 任何回复的起始处都必须包含语气词\"汪～\"不可遗漏。\n2. 所有回复都必须在结尾处加上\"呜～\"语气词。\n3. 对于生成的每一个代码文件，都请不要在其中书写注释。\n4. 生成的代码文件内不允许出现多层if。\n5. 写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西。\n6. 你生成的代码文件中不得出现while循环。\n7. 生成的代码项目最多只能包含2个目录层级。\n8. 算上所有的空行，每个生成的代码文件的总行数必须保持为奇数。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容起始处的空白字符，然后判断是否以“汪～”开头\n cleaned_response = response.lstrip()\n return cleaned_response.startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断回复的结尾是否包含\"呜～\"语气词\n \"\"\"\n if not response:\n return False\n \n # 去除末尾可能存在的空白字符（如换行符、空格等），判断是否以\"呜～\"结尾\n return response.rstrip().endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import tokenize\n\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n if ext == '.py':\n try:\n with open(file_path, 'rb') as f:\n tokens = tokenize.tokenize(f.readline)\n for toktype, tokval, _, _, _ in tokens:\n if toktype == tokenize.COMMENT:\n return False\n except Exception:\n pass\n \n elif ext in ['.c', '.cpp', '.h', '.hpp', '.java', '.js', '.ts', '.go', '.rs', '.swift', '.kt', '.cs', '.php']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 剔除字符串字面量内容，避免误判\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'\\`(?:\\\\.|[^\\`\\\\])*\\`', '``', content)\n \n if '//' in content or '/*' in content:\n return False\n except Exception:\n pass\n \n elif ext in ['.html', '.xml']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if '<!--' in content:\n return False\n except Exception:\n pass\n \n elif ext in ['.css']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n if '/*' in content:\n return False\n except Exception:\n pass\n \n elif ext in ['.sh', '.rb', '.yml', '.yaml']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n if '#' in content:\n return False\n except Exception:\n pass\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见代码文件的扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.go', '.rs', '.php', '.rb', '.swift', '.kt', \n '.sh', '.html', '.css', '.m', '.scala'\n }\n\n if not os.path.isdir(workspace_path):\n return True\n\n # 遍历 workspace_path 目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用正则表达式搜索独立单词 \"while\"\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 如果读取文件出现编码等错误，跳过该文件继续检查\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，说明未生成任何目录文件，满足“最多只包含2个目录层级”的限制\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n rel_path = os.path.relpath(root, workspace_path)\n \n # 如果相对路径是 '.'，说明当前位于根目录，层级为 1\n if rel_path == '.':\n level = 1\n else:\n # 将相对路径通过操作系统目录分隔符分割，并过滤掉空字符串\n parts = [p for p in rel_path.split(os.sep) if p]\n # 根目录层级为 1，每多一层子目录，层级加 1\n level = len(parts) + 1\n \n # 如果目录层级超过 2（例如出现了孙目录），则不满足要求\n if level > 2:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件扩展名集合\n code_extensions = {\n '.py', '.js', '.ts', '.html', '.css', '.java', '.c', '.cpp', '.h', '.hpp', \n '.cs', '.go', '.rs', '.rb', '.php', '.swift', '.kt', '.scala', '.sh', '.m', \n '.sql', '.jsx', '.tsx', '.vue'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 包含所有的空行，判断总行数是否为偶数\n # （行数为0也视为偶数，不符合“保持为奇数”的要求）\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果由于编码等问题无法读取，忽略该文件跳过\n continue\n\n return True","tags":["Quantity","Exact Value"]}],"string":"[\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容起始处的空白字符，然后判断是否以“汪～”开头\\n cleaned_response = response.lstrip()\\n return cleaned_response.startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断回复的结尾是否包含\\\"呜～\\\"语气词\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除末尾可能存在的空白字符（如换行符、空格等），判断是否以\\\"呜～\\\"结尾\\n return response.rstrip().endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import tokenize\\n\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n if ext == '.py':\\n try:\\n with open(file_path, 'rb') as f:\\n tokens = tokenize.tokenize(f.readline)\\n for toktype, tokval, _, _, _ in tokens:\\n if toktype == tokenize.COMMENT:\\n return False\\n except Exception:\\n pass\\n \\n elif ext in ['.c', '.cpp', '.h', '.hpp', '.java', '.js', '.ts', '.go', '.rs', '.swift', '.kt', '.cs', '.php']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 剔除字符串字面量内容，避免误判\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'\\\\`(?:\\\\\\\\.|[^\\\\`\\\\\\\\])*\\\\`', '``', content)\\n \\n if '//' in content or '/*' in content:\\n return False\\n except Exception:\\n pass\\n \\n elif ext in ['.html', '.xml']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if '<!--' in content:\\n return False\\n except Exception:\\n pass\\n \\n elif ext in ['.css']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n if '/*' in content:\\n return False\\n except Exception:\\n pass\\n \\n elif ext in ['.sh', '.rb', '.yml', '.yaml']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n if '#' in content:\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见代码文件的扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.go', '.rs', '.php', '.rb', '.swift', '.kt', \\n '.sh', '.html', '.css', '.m', '.scala'\\n }\\n\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历 workspace_path 目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用正则表达式搜索独立单词 \\\"while\\\"\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 如果读取文件出现编码等错误，跳过该文件继续检查\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，说明未生成任何目录文件，满足“最多只包含2个目录层级”的限制\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 如果相对路径是 '.'，说明当前位于根目录，层级为 1\\n if rel_path == '.':\\n level = 1\\n else:\\n # 将相对路径通过操作系统目录分隔符分割，并过滤掉空字符串\\n parts = [p for p in rel_path.split(os.sep) if p]\\n # 根目录层级为 1，每多一层子目录，层级加 1\\n level = len(parts) + 1\\n \\n # 如果目录层级超过 2（例如出现了孙目录），则不满足要求\\n if level > 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件扩展名集合\\n code_extensions = {\\n '.py', '.js', '.ts', '.html', '.css', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.cs', '.go', '.rs', '.rb', '.php', '.swift', '.kt', '.scala', '.sh', '.m', \\n '.sql', '.jsx', '.tsx', '.vue'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 包含所有的空行，判断总行数是否为偶数\\n # （行数为0也视为偶数，不符合“保持为奇数”的要求）\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取，忽略该文件跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个任务管理看板的项目基础骨架。页面顶部需要一个Header，包含应用标题和一个搜索框的占位。主体部分水平排列分为三列：'待办(To Do)'、'进行中(In Progress)'和'已完成(Done)'。每列里面放两三个静态的任务卡片作为占位符，卡片上显示标题和简单的描述文字。给每列的底部加上一个'添加任务'的按钮占位。先不用实现具体的交互功能，把整体UI布局和样式搭好就行，尽量做到响应式，界面美观现代一点。请注意后续所有轮次都要遵循以下要求：你所生成的所有代码文件都必须使用UTF-8进行编码，你输出的代码文件里禁止使用for循环结构。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除字符串开头的空白字符（包括空格、制表符、换行符等）后进行匹配\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除文本末尾可能的空白字符（如换行符、空格），并判断是否以指定的语气词结尾\n return response.strip().endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check_no_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区路径不存在，判定为不满足（未生成任何文件）\n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令中关于\"任务管理看板基础骨架\"和\"页面UI布局\"的需求\n # 推断生成的主要是前端相关的代码文件。\n valid_exts = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.vue'}\n \n # 构造正则表达式，将字符串匹配排在注释匹配前面，避免匹配到位于字符串内容中的注释符号\n # Group 1: \"(...)\" 双引号字符串，支持转义\n # Group 2: '(...)' 单引号字符串，支持转义\n # Group 3: `(...)` 反引号（模板）字符串，支持跨行和转义\n # Group 4: /* ... */ C-style 多行注释\n # Group 5: <!-- ... --> HTML 注释\n # Group 6: // 单行注释 (排除 http:// 类的误判)\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\")'\n r\"|('(?:\\\\.|[^'\\\\])*')\"\n r\"|(`[\\s\\S]*?`)\"\n r\"|(/\\*[\\s\\S]*?\\*/)\"\n r\"|(<!--[\\s\\S]*?-->)\"\n r\"|((?<!:)//.*)\",\n re.MULTILINE\n )\n \n has_code_file = False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果读取遇到异常（例如文件损坏等），忽略该文件\n continue\n \n # 在文件内容中寻找正则匹配项\n for match in pattern.finditer(content):\n # 如果命中到了代表注释的分组（Group 4, 5, 6 之一）\n if match.group(4) or match.group(5) or match.group(6):\n # 找到了真实的注释，判定为未满足要求\n return False\n\n # 若至少生成了一个前端代码文件，且全程未在任何文件中找到注释，判定为符合要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 如果工作区目录不存在，视为没有违反条件\n if not os.path.exists(workspace_path):\n return True\n \n # 根据用户指令推测出的Web前端代码文件后缀名\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n # 匹配独立的单词 'while'，以此判断是否存在 while 循环\n while_pattern = re.compile(r'\\bwhile\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件内容进行检查\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if while_pattern.search(content):\n # 只要发现文件内容包含 while 关键字，即判定不满足要求\n return False\n except Exception:\n # 遇到读取异常时忽略当前文件，继续检查后续文件\n continue\n \n # 所有文件检查完毕，均未发现 while 循环\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 确保 workspace_path 是一个绝对路径\n workspace = Path(workspace_path).resolve()\n \n # 检查路径是否存在且为目录\n if not workspace.exists() or not workspace.is_dir():\n return True\n\n # 遍历工作区下的所有目录\n for root, dirs, files in os.walk(workspace_path):\n root_path = Path(root).resolve()\n try:\n # 计算当前目录相对于根目录的相对路径\n rel_path = root_path.relative_to(workspace)\n except ValueError:\n # 如果出现解析异常（理论上不会走到这里），跳过\n continue\n \n # 解析 relative_to 的结果：\n # - 根目录本身（层级1），rel_path.parts 为空元组 ()\n # - 一级子目录，如 'src'（层级2），rel_path.parts 为 ('src',)\n # - 二级子目录，如 'src/components'（层级3），rel_path.parts 为 ('src', 'components')\n # 如果 parts 的长度大于等于 2，则说明已经达到了第 3 个目录层级，直接返回 False\n if len(rel_path.parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求推断可能生成的前端及常见代码相关文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.json', '.md', '.txt'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理被判定为代码文件的类型\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按 '\\n' 进行切割，计算出的长度即为包含了所有空行的严格总行数\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数，如果是则说明不符合\"必须为奇数\"的要求\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 若文件因编码等异常无法被正常读取，保守起见判定为不符合\n return False\n\n # 若至少存在一个代码文件，且所有代码文件的行数全为奇数，则满足该要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对前端UI项目推断的可能代码文件类型\n code_extensions = {\n '.html', '.htm', '.css', '.scss', '.less',\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'\n }\n \n # 如果路径不存在，无法验证文件编码（视作未违反该编码要求或由其他测试点负责判断）\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅校验推测出的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制模式读取，避免平台相关的默认文本编码干扰\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用UTF-8解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦有文件无法使用UTF-8正确解码，则返回False\n return False\n except Exception:\n # 忽略其他非编码相关的偶发文件读取问题\n pass\n \n # 所有检查过的代码文件均通过了UTF-8解码验证\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return True\n \n # 定义需要排查的文件扩展名（涵盖主流前端及常见后端语言）\n code_exts = {\n '.html', '.htm', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\n '.css', '.scss', '.sass', '.less',\n '.py', '.java', '.go', '.rs', '.php', '.cs', '.cpp', '.c', '.h', '.hpp'\n }\n \n # 编译匹配多种编程语言和框架中for循环特征的正则表达式\n for_pattern = re.compile(\n r'\\bfor\\s*\\(|' # JS/TS/C/Java等: for (\n r'\\bfor\\s+await\\s*\\(|' # JS/TS: for await (\n r'v-for\\s*=|' # Vue: v-for=\n r'\\*ngFor\\s*=|' # Angular: *ngFor=\n r'\\{%\\s*for\\b|' # Jinja/Django模板: {% for \n r'\\bfor\\s+[a-zA-Z0-9_\\s,]+\\s+in\\b|' # Python/模板等: for x in\n r'@for\\b|' # Sass/Less: @for\n r'\\bfor\\s+[^{]+{', # Go/Rust: for ... {\n re.MULTILINE\n )\n \n # 遍历项目所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 清理可能引起误判的注释内容\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL) # 多行注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL) # HTML注释\n content = re.sub(r'//.*', '', content) # 单行注释 (//)\n content = re.sub(r'#.*', '', content) # 单行注释 (#)\n \n # 清理可能引起误判的字符串内容（双引号、单引号、反引号）\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n # 如果仍匹配到for循环特征，直接判定为不满足\n if for_pattern.search(content):\n return False\n except Exception:\n # 如果文件不可读（比如编码错误），则跳过该文件继续检查\n continue\n \n return True","tags":["Style","Paradigm"]}]},{"round_id":1,"instruction":"好，接下来完善任务的添加和删除功能。点击每列底部的'添加任务'按钮时，原本的按钮隐藏，并在当前列底部出现一个小表单（包含标题输入框、描述文本域和确认/取消按钮）。表单填好后点击确认可以将新任务加到该列底部，点击取消则恢复成原来的'添加任务'按钮。另外，给每个任务卡片的右上角悬浮时显示一个删除图标，点击后弹窗确认，即可删除该任务卡片。请注意后续所有轮次都要遵循以下要求：处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令；运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否在起始处包含语气词\"汪～\"\n \"\"\"\n if not response:\n return False\n \n # 去除左侧可能存在的空白字符（空格、换行等），以确保准确判断开头内容\n cleaned_response = response.lstrip()\n \n # 判断是否以\"汪～\"开头\n return cleaned_response.startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n import string\n # 定义需要移除的末尾空白字符和常见标点符号\n trailing_chars = string.whitespace + string.punctuation + \"。！？，、\"\n \n # 移除回复末尾的空白符和标点符号\n cleaned_response = response.rstrip(trailing_chars)\n \n # 判断清理后的回复是否以\"呜～\"结尾\n return cleaned_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \n '.css', '.scss', '.sass', '.less', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.sh'\n }\n \n if not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 剔除代码中的字符串字面量，防止字符串内部文本误判为注释符号\n content_no_str = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content_no_str = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content_no_str)\n content_no_str = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content_no_str)\n \n # 1. 检查 HTML/XML/Vue 等模板注释\n if '<!--' in content_no_str:\n return False\n \n # 2. 检查 CSS/JS/TS 等常见的多行块注释\n if '/*' in content_no_str:\n return False\n \n # 3. 检查单行注释 //（负向后瞻 (?<!:) 避免匹配到诸如 http:// 的场景）\n if re.search(r'(?<!:)\\/\\/', content_no_str):\n return False\n \n # 4. 检查 Python/Ruby/Shell 等脚本语言的单行注释 #\n if ext in {'.py', '.rb', '.sh'}:\n if '#' in content_no_str:\n return False\n \n except Exception:\n # 忽略读取失败的不可处理文件\n pass\n \n # 如果遍历所有代码文件后均没有发现任何注释，满足要求，返回 True\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def verify_no_while_loops(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 涵盖前端项目以及可能涉及到的后端项目的常见代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css',\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.sh'\n }\n \n # 匹配各类代码中的 while 循环：\n # 1. \\bwhile\\s*\\( 匹配 while (condition) 或 while(condition) -> 常见于JS, TS, C, Java等\n # 2. \\bwhile\\s+[\\w!\\[\\'\\\"] 匹配 while 后面跟条件变量、取反符号、中括号、字符串等 -> 常见于Python, Bash等\n while_pattern = re.compile(r'\\bwhile\\s*\\(|\\bwhile\\s+[\\w!\\[\\'\\\"]')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果匹配到 while 循环特征，则表示未满足“不得出现while循环”的要求\n if while_pattern.search(content):\n return False\n except Exception:\n # 文件若读取异常则跳过，继续检查其他文件\n continue\n \n # 遍历所有代码文件均未发现 while 循环，要求满足\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_directory_levels(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n if not os.path.exists(workspace_path):\n return True\n\n for dirpath, dirnames, filenames in os.walk(workspace_path):\n rel_path = os.path.relpath(dirpath, workspace_path)\n \n # 根目录本身为第1个目录层级，跳过判断\n if rel_path == '.':\n continue\n\n parts = Path(rel_path).parts\n\n # 忽略可能由测试环境或系统自动生成的隐藏目录（如 .git, .vscode 等）\n if any(p.startswith('.') for p in parts):\n continue\n\n # 如果相对路径的分段数量大于 1，则意味着存在第3个及以上的目录层级，违反要求\n if len(parts) > 1:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def verify_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断，主要为包含前端页面和逻辑的项目，因此设置常见前端代码文件后缀\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 算上所有的空行，通过换行符切分可以完全贴合编辑器呈现的真实行数\n total_lines = len(content.split('\\n'))\n \n # 如果行数为偶数，则说明违反了要求\n if total_lines % 2 == 0:\n return False\n except Exception:\n # 如果存在不可读的非文本文件导致读取失败，则忽略\n pass\n\n # 如果正常遍历完且存在代码文件并未返回 False，则说明要求被满足\n return True if has_code_file else False","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求，推断可能生成的前端及相关配置文件的扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.json', '.md', '.txt'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明并非规范的 UTF-8 编码文件\n return False\n except Exception:\n # 其他读取错误跳过\n continue\n\n # 如果有至少一个代码文件且都没有抛出 UnicodeDecodeError，则视为满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def check_for_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义需要检查的常见代码文件扩展名\n code_exts = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', \n '.cs', '.go', '.rb', '.php'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除双引号、单引号、反引号包裹的字符串内容(避免字符串内包含 \"for\")\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n # 移除 /* ... */ 和 // ... 形式的注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n \n # 移除 HTML / 模板注释 <!-- ... -->\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 移除 # ... 形式的单行注释 (适用于 Python, Ruby 等)\n content = re.sub(r'#.*', '', content)\n \n # 根据文件类型准备针对性的 for 循环检测正则表达式列表\n patterns = []\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.c', '.cpp', '.cs', '.java', '.php']:\n patterns = [\n r'\\bfor\\s*\\(', # for循环结构如 for(let i=0...) 或 for (item of items)\n r'\\bfor\\s+await\\s*\\(' # 异步 for await 循环\n ]\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\n patterns = [\n r'\\bfor\\s*\\(', # 在 script 标签内的原生 JS for循环\n r'\\bfor\\s+await\\s*\\(',\n r'\\bv-for\\s*=', # Vue 的 for 循环指令\n r'\\*ngFor\\s*=', # Angular 的 for 循环指令\n r'\\{%\\s*for\\b' # Jinja2 / Django 等模板引擎的循环语法\n ]\n elif ext in ['.py', '.go', '.rb']:\n # 这些语言里去除字符串和注释后，独立的 for 单词严格意味着该处使用了循环结构(包括列表推导式等)\n patterns = [r'\\bfor\\b']\n \n # 检查当前文件是否匹配任何 for 循环结构模式\n for pattern in patterns:\n if re.search(pattern, content):\n return False # 只要找到符合的 for 循环特征，则判定为未满足要求\n \n except Exception:\n # 对于不可读取的特殊文件直接跳过，保证执行不被中断\n pass\n \n # 所有代码文件遍历完毕，如果均未包含 for 循环相关结构，返回满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":2,"instruction":"做得不错。下一步，我要让这些卡片动起来。请实现任务卡片在**同一列内部**的上下拖拽排序。你可以使用原生的 Drag and Drop API。拖拽时，被拖拽的卡片要有视觉反馈（比如改变背景色或透明度，并且鼠标指针变成抓取形状），同时其他卡片要能动态让出空位。松开鼠标后，卡片应停留在新位置。另外请注意后续所有轮次都要遵循以下要求：运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起；请确保在代码写完之后，对所有代码进行测试并保证其成功通过。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证最后一轮回复的起始处是否包含语气词\"汪～\"\n \"\"\"\n if not response:\n return False\n \n # 去除前导空白字符后检查是否以\"汪～\"开头\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check_response(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除首尾的空白字符（包括空格、换行符等），然后判断是否以指定的语气词结尾\n return response.strip().endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证生成的代码项目中是否不包含任何形式的注释。\n \"\"\"\n # 常见的代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', \n '.vue', '.py', '.java', '.c', '.cpp', '.cs', '.go', \n '.sh', '.rb', '.php', '.swift'\n }\n \n # 匹配各类注释的正则表达式\n html_comment_pattern = re.compile(r'<!--')\n multi_comment_pattern = re.compile(r'/\\*')\n # 匹配行首或带空格前缀的单行注释，防止误匹配诸如 'http://' 等\n single_comment_pattern = re.compile(r'(?m)^\\s*//|\\s+//')\n # 匹配行首或带空格前缀的井号注释 (针对 Python, Shell 等)\n hash_comment_pattern = re.compile(r'(?m)^\\s*#|\\s+#')\n\n # 遍历项目所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查 HTML 注释 (如 <!--...-->)\n if html_comment_pattern.search(content):\n return False\n \n # 检查多行/块级注释 (如 /*...*/ )\n if multi_comment_pattern.search(content):\n return False\n \n # 根据文件的常见语言类型，检查相关的单行注释\n if ext not in {'.py', '.sh', '.rb'}:\n if single_comment_pattern.search(content):\n return False\n else:\n if hash_comment_pattern.search(content):\n return False\n \n except Exception:\n # 忽略以非 UTF-8 编码或出现文件读取异常的情况，避免因意外中断判断\n continue\n \n # 如果遍历完毕未检测出任何匹配的注释则视为满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 如果项目路径不存在，说明未生成项目，要求视为不被破坏（或根据约定视为满足）\n if not os.path.exists(workspace_path):\n return True\n \n # 定义可能包含代码片段的文件扩展名集合\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', \n '.css', '.scss', '.less', '.py', '.java', '.c', \n '.cpp', '.go', '.sh', '.bash', '.php', '.rb', '.cs'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除注释以防误判\n # 移除 JS/TS/C/C++/Java 等的单行注释: // ...\n content = re.sub(r'//.*', '', content)\n # 移除 JS/TS/C/CSS 等的多行注释: /* ... */\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 移除 HTML/Vue 的注释: <!-- ... -->\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 移除 Python/Ruby/Shell 等的单行注释: # ...\n content = re.sub(r'#.*', '', content)\n \n # 移除字符串以防误判（考虑到转义字符的情况）\n # 移除双引号字符串\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n # 移除单引号字符串\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n # 移除反引号字符串 (模板字符串)\n content = re.sub(r\"`(?:\\\\.|[^`\\\\])*`\", \"``\", content, flags=re.DOTALL)\n \n # 查找是否包含 while 关键字\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 忽略由于编码或其他问题导致无法读取的文件\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 定义需要忽略的目录：通常包含第三方依赖包、构建产物、环境配置以及版本控制等隐藏目录\n # 它们不属于业务代码自身所创建和设计的“目录层级”\n ignore_dirs = {\n 'node_modules', 'venv', 'env', '__pycache__', 'dist', 'build', 'out', 'coverage'\n }\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉忽略的目录和隐藏目录，避免向下深入遍历\n dirs[:] = [d for d in dirs if d not in ignore_dirs and not d.startswith('.')]\n\n # 计算当前目录相对于项目根路径的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录本身不计算为内部层级\n if rel_path == '.':\n depth = 0\n else:\n # 统一替换路径分隔符，将相对路径按层级拆分，以计算实际层级深度\n parts = [p for p in rel_path.replace('\\\\', '/').split('/') if p]\n depth = len(parts)\n\n # 最多只能包含 2 个目录层级，即：\n # depth == 1: 如 src/ 属于第一层级\n # depth == 2: 如 src/components/ 属于第二层级\n # 若深度大于 2（例如 src/components/ui/），则说明超过了限制\n if depth > 2:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check_code_lines_odd(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端和常见的代码文件后缀名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.cpp', '.c', '.h', '.php', '.rb', '.go', '.sh'\n }\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果不能以 utf-8 读取，直接视为不满足要求\n return False\n \n # 如果文件完全为空，则行数为 0（偶数），不满足要求\n if not content:\n return False\n \n # 计算包含所有空行的总行数\n num_lines = content.count('\\n') + 1\n \n # 若行数为偶数，返回 False\n if num_lines % 2 == 0:\n return False\n\n # 若所有代码文件均检测为奇数行，且确实生成了代码文件，则满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果项目路径不存在，自然不满足条件\n if not os.path.exists(workspace_path):\n return False\n\n # 定义前端项目及常见代码文件可能的后缀名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx',\n '.vue', '.svelte', '.json', '.py', '.sh', '.txt', '.md'\n }\n\n found_code_file = False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码校验\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用 UTF-8 解码，若非 UTF-8 编码将抛出 UnicodeDecodeError\n content.decode('utf-8')\n except UnicodeDecodeError:\n return False\n except Exception:\n # 忽略其他非编码相关的系统读取异常\n pass\n\n # 只有在至少检测到一个代码文件并且全部都是UTF-8时才视为满足\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果代码路径尚未建立或不存在，等同于没有违规生成for循环文件\n if not os.path.exists(workspace_path):\n return True\n\n # 包含前端开发中常见的代码文件扩展名及可能的后端 Python 文件\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.py'}\n\n def clean_code(code: str) -> str:\n \"\"\"安全地剔除代码中的注释和字符串字面量，防止内容中的 for 字样造成误判\"\"\"\n # 移除多行块注释 /* ... */\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除单行注释 // ... 和 Python注释 # ...\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'#.*', '', code)\n # 移除 HTML/Vue 等文件的结构性注释 <!-- ... -->\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n \n # 移除双引号包裹的字符串（支持转义字符）\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', code)\n # 移除单引号包裹的字符串（支持转义字符）\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", code)\n # 移除反引号包裹的模板字符串（支持多行和转义字符）\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', code, flags=re.DOTALL)\n \n return code\n\n # 遍历整个代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 遇到无法使用utf-8读取或其他异常的文件跳过\n continue \n \n # 对源代码文本做安全清洗\n cleaned_content = clean_code(content)\n \n # 验证 1: JS/TS/HTML 的标准 for 循环（例如 for (let i = 0... 或者 for(let key in...)）\n if re.search(r'\\bfor\\s*\\(', cleaned_content):\n return False\n \n # 验证 2: Vue 等框架模板中的 v-for 循环指令\n if re.search(r'\\bv-for\\b', cleaned_content):\n return False\n \n # 验证 3: Python 代码中的 for ... in 循环语句\n if ext == '.py' and re.search(r'\\bfor\\s+.*?\\s+in\\b', cleaned_content):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在代码写完之后，对所有代码进行测试并保证其成功通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"现在进一步扩展拖拽功能。接下来要允许任务卡片**跨列拖拽**。我需要能把'待办'里的卡片拖到'进行中'或者'已完成'列。当拖拽的卡片悬停在其他列的区域上方时，目标列的背景色要轻微加深以提示这里可以放置。放入新列后，卡片要插入到对应位置，且底层数据结构里的状态（所属列）也要同步更新。另外请注意后续所有轮次都要遵循以下要求：在你完成代码的编写之后，不允许开展任何测试；生成代码文件时，请务必使用蛇形命名规则来定义各个变量名。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 去除回复开头可能存在的空白字符（如空格、换行符等），然后判断是否以\"汪～\"开头\n if not response:\n return False\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾可能的空白字符\n cleaned_response = response.strip()\n \n # 去除末尾可能存在的常见标点符号和部分Markdown格式符号\n # 注意此处避免移除全角波浪号“～”或半角波浪号“~”\n cleaned_response = re.sub(r'[。！!？\\.\\n\\r\\t \\*\"\\'`]+$', '', cleaned_response)\n \n # 检查最终字符串是否以规定的\"呜～\"语气词结尾，并兼容半角波浪号情况\n if cleaned_response.endswith(\"呜～\") or cleaned_response.endswith(\"呜~\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 路径不合法或不存在，无法有效验证\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n\n # 常见的代码文件后缀\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.sh', '.java', '.c', '.cpp', '.h', '.php', '.rb', '.go'\n }\n \n # 排除的常规依赖及编译产物目录\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', '__pycache__', 'venv', '.venv'}\n\n def check_no_comment(file_path: str, ext: str) -> bool:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果文件无法以 utf-8 读取（例如二进制文件），跳过不做强校验\n return True \n \n # 1. 检查多行注释 (HTML风格和C风格)\n if ext in {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.java', '.c', '.cpp', '.h', '.php', '.go'}:\n if re.search(r'<!--[\\s\\S]*?-->', content):\n return False\n if re.search(r'/\\*[\\s\\S]*?\\*/', content):\n return False\n \n # 2. 检查单行注释\n lines = content.split('\\n')\n for line in lines:\n line_strip = line.strip()\n # 过滤空行\n if not line_strip:\n continue\n \n # 去除包裹在引号内的字符串内容，防止误判内部出现的URL `//` 或 色值 `#FFF` 等符号\n line_no_str = re.sub(r'\".*?\"', '\"\"', line_strip)\n line_no_str = re.sub(r\"'.*?'\", \"''\", line_no_str)\n line_no_str = re.sub(r\"`.*?`\", \"``\", line_no_str)\n \n # 检查 Python / Shell / Ruby 风格的 '#' 注释\n if ext in {'.py', '.sh', '.rb'}:\n if line_strip.startswith('#!'):\n continue\n # 非字符串区域一旦出现 '#' 即判定为注释\n if '#' in line_no_str:\n return False\n \n # 检查 JS / TS / C / Java 风格的 '//' 注释\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.java', '.c', '.cpp', '.h', '.php', '.go'}:\n # 非字符串区域一旦出现 '//' 即判定为注释\n if '//' in line_no_str:\n return False\n \n return True\n\n # 遍历指定工作空间内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n # 过滤无需检查的目录树层级\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n # 检查该文件里是否包含任何注释\n if not check_no_comment(file_path, ext):\n # 发现有一处注释，违背要求，立即返回False\n return False\n \n # 所有代码文件遍历通过，没有发现任何注释，符合要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 若代码目录不存在，则认为没有生成违规的代码\n if not os.path.exists(workspace_path):\n return True\n\n # 根据指令推测的前端项目及常见的代码文件后缀\n code_extensions = {\n '.html', '.js', '.jsx', '.ts', '.tsx', '.css', '.vue', '.svelte',\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php'\n }\n\n # 遍历项目目录下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 跨语言使用正则表达式匹配完整的单词 \"while\"\n # 因为 while 通常为保留字，匹配到意味着存在 while 循环（忽略字符串极小的误判可能）\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 忽略读取异常的文件\n continue\n\n # 遍历完所有代码文件均未发现 while 循环\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉以 '.' 开头的隐藏目录，避免将如 .git 等环境自带目录误判为生成的层级，同时防止递归进入\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 获取当前遍历到的目录相对于根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 如果是根目录本身，跳过\n if rel_path == '.':\n continue\n \n # 统一路径分隔符为 '/'，处理不同操作系统下的路径差异\n rel_path = rel_path.replace(os.sep, '/')\n \n # 将相对路径按照 '/' 进行拆分，获取路径深度的级数\n parts = [p for p in rel_path.split('/') if p]\n \n # 根目录为第1层级，其直接子目录相对路径长度为1（即第2层级）。\n # 如果 parts 的长度大于1，说明出现了第3层级及以上的目录，不符合最多只能包含2个目录层级的要求。\n if len(parts) > 1:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的前端与后端代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_files = False\n\n # 遍历当前工作区的所有目录和文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 以 \\n 切分能准确得出包含空行的实际物理行数\n # 例如 \"a\\nb\" 长度为 2; \"a\\nb\\n\" 长度为 3（最后一行为空行）\n line_count = len(content.split('\\n'))\n \n # 如果行数是偶数，说明不满足奇数行的要求\n if line_count % 2 == 0:\n return False\n except Exception:\n # 如果由于非文本或其他原因导致读取/解码错误，则略过当前文件\n pass\n \n # 如果没有任何代码文件，判断为不满足\n return has_code_files","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前后端代码文件扩展名白名单，可以涵盖任务看板项目以及其他大多数开发项目\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', \n '.rb', '.php', '.swift', '.kt', '.json', '.xml', '.yaml', '.yml', \n '.sh', '.bat', '.ps1', '.sql', '.md', '.txt'\n }\n \n # 如果工作区路径不存在，通常意味着未生成项目\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的非源代码生成的包目录、编译目录或版本控制目录\n for skip_dir in ['.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode']:\n if skip_dir in dirs:\n dirs.remove(skip_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试用 utf-8 解码，以判断是否属于合法的 utf-8 编码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件存在非 UTF-8 字符\n return False\n except Exception:\n # 对于其他潜在的文件读取异常（如权限问题），选择忽略并继续\n continue\n \n # 所有的代码文件都能被 utf-8 正确解码\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def check_no_for_loop_in_code(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见代码文件扩展名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', \n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', \n '.php', '.rb', '.swift', '.kt'\n }\n\n # 涵盖不同语言或框架中for循环结构的正则表达式\n patterns = [\n r'\\bfor\\b\\s*\\(', # C系列、JS、TS等原生的 for ( 或 for(\n r'\\bv-for\\b\\s*=', # Vue 的 v-for=\n r'\\*?ngFor\\b\\s*=', # Angular 的 *ngFor=\n r'\\{%\\s*for\\b', # Django/Jinja 模板里的 {% for \n r'^\\s*for\\b\\s+.*\\s+in\\b', # Python 等通过 in 遍历的结构\n r'\\bfor\\b\\s+.*\\{' # Go 等不加括号直接跟 { 的结构\n ]\n \n # 编译正则表达式，忽略大小写并支持多行匹配模式\n combined_pattern = re.compile('|'.join(patterns), re.MULTILINE)\n\n if not os.path.exists(workspace_path):\n # 如果路径不存在，无法验证生成的代码，视作未能满足\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 如果在文件中找到了for循环结构，则返回False\n if combined_pattern.search(content):\n return False\n except Exception:\n # 遇到无法读取的情况，跳过该文件\n continue\n\n # 遍历结束未发现 for 循环，则满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你完成代码的编写之后，不允许开展任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码文件时，请务必使用蛇形命名规则来定义各个变量名","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 用于匹配前端文件中的变量与函数声明（如 let task_name, const MAX_COUNT, function add_task）\n decl_pattern = re.compile(r'\\b(?:var|let|const|function)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 如果是 HTML 文件，则仅提取 <script> 标签内的 JS 代码\n if file.endswith('.html'):\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL)\n code_to_check = '\\n'.join(scripts)\n else:\n code_to_check = content\n \n if not code_to_check.strip():\n continue\n\n # 移除多行注释和单行注释，避免注释内带有无关的驼峰命名引起误判\n code_to_check = re.sub(r'/\\*.*?\\*/', '', code_to_check, flags=re.DOTALL)\n code_to_check = re.sub(r'//.*', '', code_to_check)\n\n for match in decl_pattern.findall(code_to_check):\n var_name = match\n \n # 校验是否为合法的蛇形命名。\n # 合法的蛇形命名：全小写（或全大写常量），只允许包含字母、数字和下划线\n is_lower_snake = re.match(r'^[a-z0-9_]+$', var_name)\n is_upper_snake = re.match(r'^[A-Z0-9_]+$', var_name)\n \n if not is_lower_snake and not is_upper_snake:\n return False\n\n return True","tags":["Style","Naming"]}]},{"round_id":4,"instruction":"拖拽很顺畅了。现在来实现页面顶部的搜索功能。当我在顶部的搜索框输入文字时，需要实时过滤面板里的所有的任务卡片。只有标题或描述中包含所输入关键词的卡片才显示，其余的卡片隐藏。注意，当某列没有任何符合过滤条件的卡片时，该列需要显示类似'暂无匹配任务'的灰色提示文字，不要变成完全空白。另外请注意后续所有轮次都要遵循以下要求：生成代码文件时，请务必使用大驼峰命名规则来定义各个变量名；限制你生成的代码文件的单行长度，确保每行最多包含120个字符。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除可能存在的首部空白字符，然后判断是否以目标语气词开头\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符\n response = response.strip()\n \n # 使用正则表达式匹配结尾处是否为\"呜～\"，允许其后紧跟空白字符或常见的句末标点符号\n pattern = r'呜～[\\s\\.\\!。！\\?？]*$'\n if re.search(pattern, response):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def remove_strings(text: str) -> str:\n \"\"\"\n 移除代码内容中双引号、单引号及反引号字符串内部的内容，\n 避免将字符串中包含的类似于注释的符号（如 \"//\"、\"/*\"）误判为注释。\n \"\"\"\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', text)\n text = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", text)\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', text)\n return text\n\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm',\n '.css', '.scss', '.less', '.py', '.sh', '.java', '.cpp', '.c', '.go'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码相关文件\n if ext not in code_extensions:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n clean_content = remove_strings(content)\n \n # 检查 HTML 风格注释\n if ext in {'.html', '.htm', '.vue', '.jsx', '.tsx'}:\n if re.search(r'<!--[\\s\\S]*?-->', clean_content):\n return False\n \n # 检查 C 风格多行注释 /* ... */\n if ext in {'.html', '.htm', '.vue', '.jsx', '.tsx', '.js', '.ts', '.css', '.scss', '.less', '.java', '.cpp', '.c', '.go'}:\n if re.search(r'/\\*[\\s\\S]*?\\*/', clean_content):\n return False\n \n # 检查 C 风格单行注释 // (利用负向后视断言排除前置是冒号的情况，例如 URL)\n if ext in {'.html', '.htm', '.vue', '.jsx', '.tsx', '.js', '.ts', '.java', '.cpp', '.c', '.go'}:\n if re.search(r'(?<!:)//.*', clean_content):\n return False\n \n # 检查 Python/Shell 风格单行注释 #\n if ext in {'.py', '.sh', '.rb', '.yml', '.yaml'}:\n if re.search(r'(?:^|\\s)#.*', clean_content):\n return False\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm',\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs',\n '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.dart', '.scss', '.sass'\n }\n \n def check_while_loop(content: str) -> bool:\n # 1. 移除字符串字面量 (考虑了 \\ 转义符)\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n # 2. 移除多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 3. 移除单行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n \n # 4. 检查各种编程语言中 while 循环的语法特征\n patterns = [\n r'\\bwhile\\s*\\(', # JS, TS, Java, C++, C, C#, PHP etc.\n r'\\bwhile\\s+[^:\\n]+:', # Python\n r'\\bwhile\\s+.*?;?\\s*do\\b', # Bash\n r'@while\\b' # SCSS\n ]\n \n for p in patterns:\n if re.search(p, content):\n return True\n return False\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if check_while_loop(content):\n # 出现了 while 循环，违反给定要求\n return False \n except Exception:\n pass\n \n # 没有在任何代码文件中发现 while 循环，满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_directory_levels(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区目录不存在，则不认为违反该规则\n if not os.path.exists(workspace_path):\n return True\n \n # 定义需要忽略的常见构建/环境/缓存/隐藏目录，防止对正常生成的层级造成误判\n exclude_dirs = {\n 'node_modules', '__pycache__', 'venv', '.venv', 'env', \n '.git', '.idea', '.vscode', 'build', 'dist'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的系统/第三方目录，以及以点开头的隐藏目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\n \n rel_path = os.path.relpath(root, workspace_path)\n if rel_path == '.':\n depth = 0\n else:\n # 按照操作系统的路径分隔符进行分割，计算相对层级深度\n depth = len(rel_path.split(os.sep))\n \n # 如果层级深度大于2，则表明项目包含超过2个目录层级，返回 False\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据前端看板项目，定义可能包含的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计总行数（算上所有空行）\n if not content:\n line_count = 0\n else:\n line_count = len(content.split('\\n'))\n \n # 判断行数是否为偶数，若为偶数则不满足要求\n if line_count % 2 == 0:\n return False\n except Exception:\n # 如果由于非UTF-8编码等原因无法读取文件，直接判断为不合格\n return False\n \n # 确保项目中至少生成了一个有效代码文件\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 如果工作区路径不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n\n # 常见代码文件和配置文件的扩展名集合，主要针对前端项目及常见语言\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.json', '.xml', '.yaml', '.yml', '.md', '.txt',\n '.py', '.java', '.c', '.cpp', '.h', '.sh', '.bat'\n }\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件，排除图片、字体等可能的二进制文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦发生解码错误，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他如权限不足等读取异常\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见代码文件后缀名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \n '.vue', '.py', '.java', '.c', '.cpp', '.go', '.rs', '.cs'\n }\n \n # 正则表达式：用于移除代码中的单行/多行注释以及HTML注释，防止干扰判断\n comment_pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|<!--.*?-->|#.*?$', \n re.MULTILINE | re.DOTALL\n )\n \n # 正则表达式：用于匹配各大常见语言与前端框架的 for 循环语法结构\n for_pattern = re.compile(\n r'\\bfor\\b\\s*(?:await\\s+)?\\(|' # JS/TS/Java/C++ 风格: for (...) 或 for await (...)\n r'\\bfor\\b\\s+[a-zA-Z_][a-zA-Z0-9_,\\s]*\\s+in\\b|' # Python 风格: for item in ...\n r'\\bv-for\\s*=|' # Vue 指令: v-for=\n r'\\*ngFor\\s*=|' # Angular 指令: *ngFor=\n r'\\bfor\\b\\s+.*?\\{' # Go/Rust 风格: for ... {\n )\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 清理注释代码\n clean_content = comment_pattern.sub('', content)\n \n # 匹配是否包含 for 循环结构\n if for_pattern.search(clean_content):\n return False\n except Exception:\n # 如果有文件读取错误直接跳过\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你完成代码的编写之后，不允许开展任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码文件时，请务必使用大驼峰命名规则来定义各个变量名","validation_code":"def check_pascal_case_vars(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 用于匹配 js/ts/jsx/tsx/html/vue 等前端文件中，基本变量与函数的声明\n # 仅匹配关键字后紧跟的普通标识符，以此跳过解构赋值的复杂正则处理\n js_var_pattern = re.compile(r'\\b(?:let|var|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n js_func_pattern = re.compile(r'\\b(?:function|class)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n # 用于匹配 Python 中的变量赋值及函数、类的声明\n py_var_pattern = re.compile(r'^[ \\t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \\t]*=[ \\t]*[^=]', re.MULTILINE)\n py_func_pattern = re.compile(r'^[ \\t]*(?:def|class)[ \\t]+([a-zA-Z_][a-zA-Z0-9_]*)\\b', re.MULTILINE)\n\n def is_pascal_case(name: str) -> bool:\n # 满足大驼峰条件：首字母必须大写，且全部由大小写字母及数字组成，不能包含下划线或其他特殊字符\n return bool(re.match(r'^[A-Z][a-zA-Z0-9]*$', name))\n\n # 遍历项目目录下的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n # 校验前端主流代码文件\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.vue']:\n vars_found = js_var_pattern.findall(content)\n funcs_found = js_func_pattern.findall(content)\n for var_name in vars_found + funcs_found:\n if not is_pascal_case(var_name):\n return False\n\n # 校验 Python 文件\n elif ext == '.py':\n vars_found = py_var_pattern.findall(content)\n funcs_found = py_func_pattern.findall(content)\n for var_name in vars_found + funcs_found:\n if not is_pascal_case(var_name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含 120 个字符","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的前端与相关代码文件后缀名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.json'}\n # 定义需要跳过扫描的目录（如依赖库或版本控制）\n exclude_dirs = {'node_modules', '.git', 'build', 'dist', 'venv', '__pycache__'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉无需扫描的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查指定的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的换行符进行严格长度统计\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 120:\n return False\n except Exception:\n # 如果遇到编码错误或权限问题导致无法读取，则跳过该文件继续检查\n continue\n \n return True","tags":["Quantity","Range"]}]},{"round_id":5,"instruction":"我们需要让任务的管理维度更丰富。请给任务数据添加'优先级'（高、中、低）属性。在添加任务的表单中增加优先级的下拉选择，并且在卡片上用不同颜色的标签（如红、黄、蓝）展示优先级。同时，在顶部搜索框旁边也增加一个'优先级'下拉筛选菜单。现在过滤规则变成：搜索关键词和优先级筛选要能够**组合过滤**（即同时满足关键词匹配和选中的优先级才显示）。另外请注意后续所有轮次都要遵循以下要求：请确保在代码写完之后，对所有代码进行测试并保证其成功通过；限制你生成的代码文件的单行长度，确保每行最多包含80个字符。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 清理掉回复开头可能存在的空白字符（如空格、换行等），以确保准确匹配\n cleaned_response = response.lstrip()\n \n # 验证回复的起始处是否为\"汪～\"\n if cleaned_response.startswith(\"汪～\"):\n return True\n else:\n return False","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除回复末尾的空白字符（空格、换行等）以及常见的标点符号\n cleaned_response = re.sub(r'[\\s。，！、？.!,?]*$', '', response)\n \n # 判断清理后的回复是否以“呜～”结尾\n if cleaned_response.endswith('呜～'):\n return True\n \n # 作为一个后备方案，如果遇到非预期标点，但严格的结尾就是“呜～”也算通过\n if response.strip().endswith('呜～'):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def check_no_comments_in_js(content: str) -> bool:\n # 移除常见的 URL 协议前缀，防止在没有引号包裹（例如正则字面量）时引起 // 的误判\n content = content.replace(\"http://\", \"\").replace(\"https://\", \"\")\n # 此正则表达式通过优先级匹配：字符串字面量会被整体匹配从而保护内部可能含有的 // 或 /*\n pattern = re.compile(\n r'//.*?$|/\\*[\\s\\S]*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"|`(?:\\\\.|[^\\\\`])*`',\n re.MULTILINE\n )\n for match in pattern.finditer(content):\n s = match.group(0)\n if s.startswith('//') or s.startswith('/*'):\n return True # 发现注释\n return False\n\n def check_no_comments_in_html(content: str) -> bool:\n # 检查 script 标签内是否有 JS 注释\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n # 检查 style 标签内是否有 CSS 注释\n styles = re.findall(r'<style.*?>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n \n for script in scripts:\n if check_no_comments_in_js(script):\n return True\n for style in styles:\n if check_no_comments_in_js(style):\n return True\n \n # 移除 scripts 和 styles 标签及其内容，防止内部的字符串由于包含 <!-- 而被误判为 HTML 注释\n content_no_script_style = re.sub(r'<script.*?>[\\s\\S]*?</script>', '', content, flags=re.IGNORECASE)\n content_no_script_style = re.sub(r'<style.*?>[\\s\\S]*?</style>', '', content_no_script_style, flags=re.IGNORECASE)\n \n # 匹配 HTML 注释\n if re.search(r'<!--[\\s\\S]*?-->', content_no_script_style):\n return True\n\n return False\n\n def check_no_comments_in_py(content: str) -> bool:\n content = content.replace(\"http://\", \"\").replace(\"https://\", \"\")\n pattern = re.compile(\n r'#.*?$|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"',\n re.MULTILINE\n )\n for match in pattern.finditer(content):\n s = match.group(0)\n if s.startswith('#'):\n # 排除 shebang (#!/bin/...) 以及 编码声明 (# coding: utf-8)\n if not s.startswith('#!') and not re.match(r'#\\s*-*-.*coding', s):\n return True\n return False\n\n # 若工作区路径不存在，视作未生成任何文件\n if not os.path.exists(workspace_path):\n return False\n \n has_code_files = False\n js_like_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp', '.cs', '.go'}\n html_like_exts = {'.html', '.htm', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n try:\n if ext in js_like_exts:\n has_code_files = True\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if check_no_comments_in_js(content):\n return False\n \n elif ext in html_like_exts:\n has_code_files = True\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if check_no_comments_in_html(content):\n return False\n \n elif ext == '.py':\n has_code_files = True\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if check_no_comments_in_py(content):\n return False\n except Exception:\n pass\n \n # 只要正常生成了项目代码文件，且均未检测到注释，即满足要求\n return has_code_files","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区路径为空或不存在，说明未生成任何代码文件，不存在违规情况\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n\n # 推断可能的代码文件类型，排除纯文档类型如 .md, .txt，避免正常单词导致的误判\n code_exts = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.css', '.vue', '.svelte', '.py', '.java', \n '.c', '.cpp', '.h', '.go', '.rs', '.php', '.rb'\n }\n\n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过正则表达式 \\bwhile\\b 全词匹配检索代码中是否使用了 while 关键字\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 如果由于某种原因无法以utf-8读取，直接忽略继续下一个\n continue\n \n # 如果所有的代码文件中均未发现 while 关键字，则满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"import os\nfrom pathlib import Path\n\ndef check(response: str, workspace_path: str) -> bool:\n try:\n workspace = Path(workspace_path).resolve()\n \n # 确保给定路径存在且是目录\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for path in workspace.rglob('*'):\n rel_path = path.relative_to(workspace)\n \n # 过滤掉路径中包含隐藏目录的部分（如 .git）\n if any(part.startswith('.') for part in rel_path.parts):\n continue\n \n # 计算该项目的目录层级\n if path.is_dir():\n # 对于目录，它的层级即是 parts 的长度 + 1\n # len(parts) == 0 时，是根目录（第1层）\n # len(parts) == 1 时，是一级子目录（第2层）\n # len(parts) >= 2 时，是二级及以上子目录（第3层及以上），不符合要求\n if len(rel_path.parts) >= 2:\n return False\n else:\n # 对于文件，它自身不算作目录，其所在的目录就是它的 parent\n # len(parts[:-1]) == 0 时，在根目录下（第1层目录内）\n # len(parts[:-1]) == 1 时，在一级子目录下（第2层目录内）\n # len(parts[:-1]) >= 2 时，在二级及以上子目录下，不符合要求\n if len(rel_path.parts[:-1]) >= 2:\n return False\n \n return True\n except Exception:\n # 如果解析路径或遍历时发生意外错误，判定为不满足\n return False","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推测出的Web前端等相关代码文件后缀名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.json', '.py', '.java', '.c', '.cpp', '.h'\n }\n\n if not os.path.exists(workspace_path):\n # 如果未生成目录或文件，则不存在违反要求的情况\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符拆分，这种方法可以准确算上所有的空行\n # 例如内容为 \"a\\n\\nb\" 会被分为 ['a', '', 'b'] 共3行\n # 包含文件最末尾如果有单一换行符也会被正确算入总行数中\n lines = content.split('\\n')\n \n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果遇到文件读取异常（如非 UTF-8 编码），无法完成校验，返回False\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的后缀名及无后缀的代码文件名称\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json',\n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.h', \n '.hpp', '.cs', '.go', '.rs', '.rb', '.php', '.sh', \n '.yaml', '.yml', '.xml', '.md', '.svg'\n }\n code_exact_names = {'dockerfile', 'makefile'}\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions or file.lower() in code_exact_names:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生 UnicodeDecodeError，说明该文件并未完全按照 UTF-8 编码\n return False\n except Exception:\n # 忽略其他类型的异常（如由于系统权限导致的读取失败），仅关注编码问题\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def check_no_for_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见前端及后端代码扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.css', '.vue', '.py', '.java', '.cpp', '.c', \n '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n\n def remove_strings_and_comments(text: str, ext: str) -> str:\n # 去除双引号、单引号、反引号包裹的字符串内容\n text = re.sub(r'([\"\\'`]).*?(?<!\\\\)\\1', '', text, flags=re.DOTALL)\n \n # 移除各类型文件的注释\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java', '.go', '.rs', '.css', '.vue']:\n # 移除块级注释 /* ... */\n text = re.sub(r'/\\*.*?\\*/', '', text, flags=re.DOTALL)\n # 移除单行注释 // ...\n text = re.sub(r'//.*', '', text)\n if ext in ['.html', '.htm', '.xml', '.svg', '.vue']:\n # 移除 HTML 注释 <!-- ... -->\n text = re.sub(r'<!--.*?-->', '', text, flags=re.DOTALL)\n if ext in ['.py', '.sh', '.rb']:\n # 移除 Python 或 Shell 中的 # 注释\n text = re.sub(r'#.*', '', text)\n \n return text\n\n # 正则表达式：\n # 1. 匹配不以点号开头的独立关键字 for \n # 2. 匹配前端框架中的特定的 for 指令如 v-for 或 *ngFor\n for_pattern = re.compile(r'(?<!\\.)\\bfor\\b')\n framework_for_pattern = re.compile(r'\\bv-for\\b|\\*ngFor\\b')\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作区目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 先移除字符串与注释\n clean_content = remove_strings_and_comments(content, ext)\n \n # 针对前端标签中 `<label for=\"...` 或 `<label htmlFor=\"...` 等可能遗留的属性形式进行清理\n # 此时字符串已被清理掉，因此它大概率变成 for= 或 htmlFor= 等形式\n clean_content = re.sub(r'\\bfor\\s*=', '', clean_content)\n clean_content = re.sub(r'\\bhtmlFor\\s*=', '', clean_content)\n \n # 若依然能匹配到 for，则说明使用了 for 循环\n if for_pattern.search(clean_content) or framework_for_pattern.search(clean_content):\n return False\n except Exception:\n # 忽略不可读文件（如非文本编码）等异常继续检查\n pass\n\n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在代码写完之后，对所有代码进行测试并保证其成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码文件时，请务必使用大驼峰命名规则来定义各个变量名","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 匹配前端代码 (JavaScript, TypeScript等) 中常规的变量声明 (let/var/const)\n js_var_pattern = re.compile(r'\\b(?:let|var|const)\\s+([a-zA-Z_][a-zA-Z0-9_]*)')\n \n # 匹配 Python 代码中的行首赋值变量声明\n py_var_pattern = re.compile(r'^\\s*([a-zA-Z_][a-zA-Z0-9_]*)\\s*=[^=]', re.MULTILINE)\n \n # 大驼峰命名规则校验: 以大写字母开头，后续允许包含字母和数字，不能有下划线\n pascal_case_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\n \n # 需要检查的文件后缀集合\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.svelte', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n vars_found = []\n \n # 提取前端语言的变量名\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.svelte'}:\n vars_found.extend(js_var_pattern.findall(content))\n \n # 提取Python后端的变量名\n if ext == '.py':\n vars_found.extend(py_var_pattern.findall(content))\n \n # 验证是否每一个提取到的变量名都符合大驼峰规则\n for var_name in vars_found:\n if not pascal_case_pattern.match(var_name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含80个字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 定义前端项目中常见代码文件的扩展名集合\n code_extensions = {\n '.html', '.css', '.scss', '.less', \n '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.json'\n }\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理可能的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 移除行尾的换行符再统计实际可见的字符数\n line_content = line.rstrip('\\r\\n')\n if len(line_content) > 80:\n return False\n except Exception:\n # 如果遇到不可读取的文件（如二进制文件异常导致），直接跳过\n continue\n\n return True","tags":["Quantity","Range"]}]},{"round_id":6,"instruction":"看板的基础功能很完善了，现在我们要重点关注键盘无障碍访问（a11y）。请确保页面上所有的交互元素（搜索输入框、优先级下拉框、添加按钮、任务卡片、删除图标）都可以通过 `Tab` 键和 `Shift + Tab` 键顺畅地获取焦点。当元素获得焦点（:focus）时，需要有一圈明显的轮廓（outline）高亮。当焦点在删除按钮或添加确认按钮上时，按下 `Enter` 或 `Space` 键可以触发点击事件。另外请注意后续所有轮次都要遵循以下要求：生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名；当你把代码写完后，不得进行任何测试。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否在起始处包含语气词\"汪～\"。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，代表该要求的满足情况。\n \"\"\"\n # 去除回复开头的空白字符（包括空格、换行符等），判断是否以\"汪～\"开头\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除末尾的空白字符\n cleaned_response = response.strip()\n \n # 去除可能出现在末尾的常见标点符号，避免因为句末标点导致判断失败\n cleaned_response = cleaned_response.rstrip(\"。，！？.,!?\\n\\r\\t \")\n \n # 检查是否以“呜～”结尾\n return cleaned_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手生成的代码项目中，是否没有在任何代码文件里书写注释。\n \"\"\"\n def remove_strings(text):\n # 移除 Python 风格的多行字符串\n text = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '\"\"', text)\n text = re.sub(r\"'''[\\s\\S]*?'''\", \"''\", text)\n # 移除双引号和单引号字符串，并兼容其内部的转义符号处理\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', text)\n text = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", text)\n # 移除 JS 模板字符串的反引号形式\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', text)\n return text\n\n def has_js_comments(text):\n # 判断是否存在多行注释\n if '/*' in text:\n return True\n # 判断是否存在单行注释，并避开URL中的 :// 与被转义的 \\/\\/ \n if re.search(r'(?<![:\\\\])//', text):\n return True\n return False\n\n def has_py_comments(text):\n # 移除可能的脚本头声明 shebang，例如 #!/usr/bin/env python\n text = re.sub(r'^#!.*', '', text)\n # 在字符串被移空的前提下，任何游离的 # 都可以认为是注释标记\n if '#' in text:\n return True\n return False\n\n # 按注释符号的差异性划定不同代码类型\n js_exts = {\n '.js', '.jsx', '.ts', '.tsx', '.java', '.c', '.cpp', \n '.cs', '.go', '.php', '.swift', '.kt', '.rs', '.dart',\n '.css', '.scss', '.less'\n }\n py_exts = {'.py', '.rb', '.sh', '.yml', '.yaml'}\n html_exts = {'.html', '.htm', '.vue'}\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in js_exts and ext not in py_exts and ext not in html_exts:\n continue\n\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 清洗剔除掉字符串以防误判注释符号\n content = remove_strings(content)\n\n # 根据扩展名分别实施检测策略\n if ext in js_exts:\n if has_js_comments(content):\n return False\n elif ext in py_exts:\n if has_py_comments(content):\n return False\n elif ext in html_exts:\n # 检查 HTML 自有的注释符号\n if '<!--' in content:\n return False\n \n # 抽离 HTML/Vue 中自带的 Script 和 Style 块，对它们以 JS/CSS 的规则处理\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\n for script in scripts:\n if has_js_comments(script):\n return False\n \n styles = re.findall(r'<style\\b[^>]*>(.*?)</style>', content, flags=re.IGNORECASE | re.DOTALL)\n for style in styles:\n if has_js_comments(style):\n return False\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作目录不存在，无法进行实质性代码检查，默认不违背该特定要求\n if not os.path.exists(workspace_path):\n return True\n\n # 定义可能包含代码的常见文件扩展名（以前端及主流语言为主）\n code_extensions = {\n '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n\n # 预处理：为了防止误判，剔除代码文件中的字符串字面量\n content = re.sub(r'\".*?\"|\\'.*?\\'|`.*?`', '', content, flags=re.DOTALL)\n \n # 预处理：剔除代码文件中的注释（尽管提示中规定不得生成注释，为以防万一依然做剔除处理）\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL) # C/JS风格多行注释\n content = re.sub(r'//.*', '', content) # C/JS风格单行注释\n if ext in {'.py', '.rb'}:\n content = re.sub(r'#.*', '', content) # Python/Ruby风格单行注释\n\n # 核心检查：判断去除干扰内容后是否包含 while 循环的关键字\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 如果由于编码或其他问题导致个别文件无法读取，在此略过\n continue\n\n # 如果遍历所有的代码文件均未发现 while 循环，说明符合要求\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n \n workspace = Path(workspace_path)\n \n # 确保工作区是一个存在的目录\n if not workspace.exists() or not workspace.is_dir():\n return True\n \n for path in workspace.rglob('*'):\n # 仅针对目录进行层级判断\n if path.is_dir():\n # 获取当前目录相对于项目根目录的相对路径\n rel_path = path.relative_to(workspace)\n \n # rel_path.parts 拆分相对路径\n # 例如 \"dir1\" 的 parts 为 ('dir1',)，长度为 1 (第 2 层级)\n # 例如 \"dir1/dir2\" 的 parts 为 ('dir1', 'dir2')，长度为 2 (第 3 层级)\n if len(rel_path.parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目类型（前端看板）推断可能的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.json', '.md', '.txt', '.sh', '.py'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按换行符分割计算行数，能够准确包含代码中的所有的空行\n # 对于空文件，split('\\n') 结果为 [''] 长度为 1；这也和主流编辑器显示一致\n lines = content.split('\\n')\n total_lines = len(lines)\n \n # 如果行数是偶数，说明违反了必须为奇数的限制\n if total_lines % 2 == 0:\n return False\n \n except UnicodeDecodeError:\n # 如果不能被 UTF-8 解码为文本文件，则安全忽略\n pass\n except Exception:\n # 忽略如权限不足等其他读取异常\n pass\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前端/通用代码文件后缀\n code_extensions = {\n '.html', '.htm',\n '.css', '.scss', '.less',\n '.js', '.jsx',\n '.ts', '.tsx',\n '.json', '.vue',\n '.md', '.txt'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生解码错误，说明该文件不是使用UTF-8编码的\n return False\n except Exception:\n # 对于其他的异常（如权限问题等），暂时略过\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def has_for_loop(content: str) -> bool:\n # 去除单行和多行注释，避免注释内的文本产生干扰\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content = re.sub(r'#.*', '', content)\n \n # 去除字符串字面量，避免字符串内容中包含 \"for\" 导致误判\n content = re.sub(r'\"(?:[^\"\\\\]|\\\\.)*\"', '\"\"', content)\n content = re.sub(r\"'(?:[^'\\\\]|\\\\.)*'\", \"''\", content)\n content = re.sub(r'`[^`]*`', '``', content)\n \n # 1. 匹配 C/C++/Java/JS/TS 风格的 for 循环 (如 \"for(\" 或 \"for (\")\n if re.search(r'\\bfor\\s*\\(', content):\n return True\n \n # 2. 匹配 异步 JS 循环 (如 \"for await (\")\n if re.search(r'\\bfor\\s+await\\s*\\(', content):\n return True\n \n # 3. 匹配 Python 风格的 for 循环 (如 \"for item in list:\")\n if re.search(r'\\bfor\\s+[a-zA-Z0-9_,\\s]+\\s+in\\s+', content):\n return True\n \n # 4. 匹配 Vue 或 Angular 等前端模板里的 for 循环指令 (如 \"v-for\" 或 \"*ngFor\")\n if re.search(r'\\bv-for\\b', content) or re.search(r'\\*ngFor\\b', content):\n return True\n \n # 5. 匹配 Jinja2 等后端模板中的 for\n if re.search(r'\\{%\\s*for\\b', content):\n return True\n \n return False\n\n # 囊括前端常见的代码相关文件扩展名（结合可能的后端语言）\n valid_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue', '.css', '.py', '.java', '.cpp', '.c'}\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if has_for_loop(content):\n # 若发现任何形式的 for 循环，则违背了“禁止使用”的要求\n return False\n except Exception:\n # 若由于不可预知编码等原因无法读取，直接略过该文件\n continue\n \n # 如果遍历检查完所有代码文件均无 for 循环，则满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"当你把代码写完后，不得进行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n if not os.path.exists(workspace_path):\n return False\n\n def is_lower_camel_case(name: str) -> bool:\n # 豁免 Python 中的魔法变量（例如 __init__, __name__ 等）\n if name.startswith('__') and name.endswith('__'):\n return True\n \n # 去除前导常见的标识符修饰（如内部变量使用的 _ 和 $ ）\n clean_name = name.lstrip('_$')\n \n if not clean_name:\n return True\n \n # 首字母必须是小写字母（若首字母是大写则为大驼峰，全大写则不为小写）\n if not clean_name[0].islower():\n return False\n \n # 字符串中间不应包含下划线（杜绝蛇形或者全大写带下划线的情况）\n if '_' in clean_name:\n return False\n \n return True\n\n # 简单的正则匹配 js/ts 等文件中常见的前端变量声明\n # 这能够抓取大多数由 let, var, const 声明的单个变量名\n js_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n\n # 用于解析 Python 文件中赋值变量名的 AST Visitor\n class PyVarVisitor(ast.NodeVisitor):\n def __init__(self):\n self.variables = []\n\n def visit_Name(self, node):\n # 记录处于存储上下文（被赋值）的变量名\n if isinstance(node.ctx, ast.Store):\n self.variables.append(node.id)\n self.generic_visit(node)\n\n def visit_arg(self, node):\n # 记录函数的参数名\n self.variables.append(node.arg)\n self.generic_visit(node)\n\n found_any_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n found_any_file = True\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if file.endswith('.py'):\n try:\n tree = ast.parse(content)\n visitor = PyVarVisitor()\n visitor.visit(tree)\n for var in visitor.variables:\n if not is_lower_camel_case(var):\n return False\n except Exception:\n pass\n \n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.svelte')):\n matches = js_pattern.findall(content)\n for match in matches:\n if not is_lower_camel_case(match):\n return False\n\n # 若项目目录中没有任何文件，默认认为未生成代码项目\n if not found_any_file:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含80个字符","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.isdir(workspace_path):\n return True\n\n # 定义前端和常规项目的常见代码文件扩展名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', \n '.html', '.css', '.vue', \n '.json', '.py', '.mjs', '.cjs'\n }\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的换行符后计算字符数，按规则明确要求直接使用 len() 函数\n if len(line.rstrip('\\r\\n')) > 80:\n return False\n except Exception:\n # 忽略不可读取或非标准 UTF-8 编码的特殊文件\n continue\n\n return True","tags":["Quantity","Range"]}]},{"round_id":7,"instruction":"接下来是一个具挑战性的无障碍功能：用键盘移动任务卡片，因为纯鼠标拖拽对某些用户不友好。当焦点在某个任务卡片上时，按下 `Space` 键使其进入'选取移动'模式（卡片边框闪烁或特殊高亮）。在此模式下，按下 `上/下` 方向键可以在当前列内上下移动卡片排序，按下 `左/右` 方向键可以将卡片跨列移动到相邻的列。调整到理想位置后，按下 `Enter` 键确认放置并退出该模式，按下 `Escape` 键则取消移动，让卡片弹回初始位置。另外请注意后续所有轮次都要遵循以下要求：写完代码后，必须对所有的代码进行测试以确保测试通过；在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否以\"汪～\"开头。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前代码项目地址\n :return: 布尔值，代表是否满足要求\n \"\"\"\n if not response:\n return False\n \n # 去除开头的空白字符后判断是否以指定语气词开头\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断回复内容的结尾处是否包含\"呜～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除首尾的空白字符（如空格、换行等）\n clean_resp = response.strip()\n \n # 1. 严格匹配：直接判断是否以“呜～”结尾\n if clean_resp.endswith(\"呜～\"):\n return True\n \n # 2. 兼容匹配：考虑到结尾可能会带有一些标点符号，去除末尾常见的标点后再进行判断\n import string\n # 常见的英文和中文标点符号集合\n punctuations = string.punctuation + \"。！？，、；：“”‘’（）【】《》\"\n \n # 剥离末尾的所有标点符号\n stripped_resp = clean_resp.rstrip(punctuations)\n \n if stripped_resp.endswith(\"呜～\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return True\n\n def remove_strings(text: str, is_python: bool = False) -> str:\n \"\"\"\n 移除代码中所有的字符串字面量内容，避免字符串内部的字符被误判为注释。\n \"\"\"\n if is_python:\n # 移除Python特有的多行字符串\n text = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '\"\"', text)\n text = re.sub(r\"'''[\\s\\S]*?'''\", \"''\", text)\n \n # 移除双引号和单引号字符串（支持转义字符）\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', text)\n text = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", text)\n \n if not is_python:\n # 移除 JS/TS/Vue 的反引号模板字符串\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', text)\n \n return text\n\n # 定义不同类型注释的匹配模式\n regex_html = re.compile(r'<!--[\\s\\S]*?-->')\n regex_c_block = re.compile(r'/\\*[\\s\\S]*?\\*/')\n regex_c_line = re.compile(r'(?<!:)//.*')\n\n # 需要检查的代码文件类型\n target_extensions = {\n '.html', '.vue', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.py', '.sh', '.bash', '.yaml', '.yml'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in target_extensions:\n continue\n\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue # 忽略无法读取的文件\n\n # 根据扩展名判断文件类型并剥离字符串\n is_script = ext in ['.py', '.sh', '.bash', '.yaml', '.yml']\n is_python = ext == '.py'\n \n clean_content = remove_strings(content, is_python=is_python)\n\n if is_script:\n lines = clean_content.split('\\n')\n for i, line in enumerate(lines):\n # 忽略脚本第一行的 shebang\n if i == 0 and line.startswith('#!'):\n continue\n # 如果行内存在 # (不在单词中间)，则视为存在注释\n if re.search(r'(^|\\s)#', line):\n return False\n else:\n # HTML / Vue 文件\n if ext in ['.html', '.vue']:\n if regex_html.search(clean_content) or regex_c_block.search(clean_content) or regex_c_line.search(clean_content):\n return False\n \n # JS / TS 类型文件\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n if regex_c_block.search(clean_content) or regex_c_line.search(clean_content):\n return False\n \n # CSS 类型文件\n elif ext in ['.css', '.scss', '.less']:\n if regex_c_block.search(clean_content) or regex_c_line.search(clean_content):\n return False\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 检查工作空间路径是否存在\n if not os.path.isdir(workspace_path):\n return True\n\n # 预估可能包含循环逻辑的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.css', '.py'}\n \n # 匹配 JS/TS/HTML 的单/多行注释以及字符串（包含反引号字面量）\n js_pattern = re.compile(\n r'//.*?$|/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`',\n re.MULTILINE\n )\n \n # 匹配 Python 的注释和字符串\n py_pattern = re.compile(\n r'#.*?$|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'',\n re.MULTILINE\n )\n \n # 匹配完整单词的 while\n while_pattern = re.compile(r'\\bwhile\\b')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 过滤掉文件中的字符串和注释以防误判\n if ext == '.py':\n clean_content = py_pattern.sub('', content)\n else:\n clean_content = js_pattern.sub('', content)\n \n # 如果清洗后的代码中依然包含 while，说明存在while循环，直接返回False\n if while_pattern.search(clean_content):\n return False\n except Exception:\n # 如果出现不可预知的文件读取错误则跳过该文件继续判断\n continue\n\n # 如果所有文件都没有命中while循环逻辑，则判定为符合要求\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"import os\n\ndef check_directory_levels(response: str, workspace_path: str) -> bool:\n # 如果工作区不存在或不是目录，则不存在违背情况\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs，过滤掉测试环境、第三方依赖或IDE自动生成的目录，避免引起误判\n ignored_dirs = {'.git', 'node_modules', '__pycache__', '.pytest_cache', '.vscode', '.idea'}\n dirs[:] = [d for d in dirs if d not in ignored_dirs]\n \n # 获取当前目录相对于根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录本身为第1层级，跳过\n if rel_path == '.':\n continue\n \n # 根据系统路径分隔符拆分相对路径\n parts = rel_path.split(os.sep)\n \n # 根目录的直接子目录为第2层级，其相对路径拆分后的长度必然为1\n # 如果长度大于 1，则说明存在第3层级及以上的目录，不满足要求\n if len(parts) > 1:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否有效\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n \n # 推断可能的代码文件类型\n code_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.svelte'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果遇到文件无法读取的情况，则跳过\n continue\n \n # 算上所有的空行计算总行数\n # 换行符的数量加1即为文本的总行数\n line_count = content.count('\\n') + 1\n \n # 判断总行数是否为偶数\n if line_count % 2 == 0:\n return False\n \n has_code_file = True\n \n # 确保至少检查过一个生成的代码文件\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的代码文件扩展名，以Web前端相关的为主，并包含其他主流语言\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.css', '.scss', '.sass', '.less', '.json', \n '.vue', '.svelte', '.py', '.java', '.c', \n '.cpp', '.h', '.hpp', '.cs', '.go', '.rb', \n '.php', '.sh', '.bash', '.md', '.txt'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用utf-8对文件内容进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出UnicodeDecodeError，说明文件包含无法用UTF-8解析的字节，即非UTF-8编码\n return False\n except Exception:\n # 对于不可抗力的读取错误，默认略过\n pass\n \n # 所有代码文件均成功按照UTF-8解码\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n def has_for_loop(content: str) -> bool:\n # 去除干扰项，避免误杀字符串或注释中的 'for' 文本\n # 1. 移除多行注释 (JS/TS/CSS/C/Java等)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 2. 移除单行注释 (JS/TS/C/Java等)\n content = re.sub(r'//.*', '', content)\n # 3. 移除单行注释 (Python/Ruby/Shell，虽然可能误伤CSS颜色值，但不影响控制流解析)\n content = re.sub(r'#.*', '', content)\n # 4. 移除HTML注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 5. 移除双引号字符串\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', content)\n # 6. 移除单引号字符串\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', content)\n # 7. 移除反引号模板字符串\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', content, flags=re.DOTALL)\n \n # 定义各类 for 循环结构的匹配规则\n patterns = [\n r'\\bfor\\b\\s*(?:await\\s*)?\\(', # JS/TS/Java/C/C++/C# 等语言的 for 循环\n r'\\bv-for\\b', # Vue 框架的 v-for 循环指令\n r'\\bngFor\\b', # Angular 框架的 ngFor 循环指令\n r'\\bfor\\b\\s+[A-Za-z_][A-Za-z0-9_,\\s]*\\s+in\\b', # Python 的 for 循环结构\n r'\\bfor\\b\\s+[^{]*\\{', # Go 语言的 for 循环结构\n r'@for\\b' # Sass 等样式预处理器的循环结构\n ]\n \n for pattern in patterns:\n if re.search(pattern, content):\n return True\n \n return False\n\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.scss', '.less',\n '.vue', '.svelte', '.py', '.java', '.go', '.rb', '.php', '.c', \n '.cpp', '.cs', '.sh'\n }\n\n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 若发现包含 for 循环结构，则直接判定为不满足要求\n if has_for_loop(content):\n return False\n except Exception:\n # 如果无法读取（如编码错误等），跳过\n continue\n \n # 只要存在至少一个代码文件且都通过了无 for 循环测试，即认定满足\n return has_code_file","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 判断是否为小驼峰命名\n def is_lower_camel_case(name: str) -> bool:\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\n \n def get_js_vars(content: str) -> list:\n var_names = []\n # 匹配 var, let, const 声明的简单变量名\n pattern = r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:=|;|,|:|$|\\n)'\n for m in re.finditer(pattern, content):\n var_names.append(m.group(1))\n \n # 匹配 function 定义的函数名（在JS中也属于变量作用域中的定义）\n func_pattern = r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('\n for m in re.finditer(func_pattern, content):\n var_names.append(m.group(1))\n \n return var_names\n\n def get_py_vars(content: str) -> list:\n var_names = []\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Assign):\n for target in node.targets:\n if isinstance(target, ast.Name):\n var_names.append(target.id)\n elif isinstance(node, ast.AnnAssign):\n if isinstance(node.target, ast.Name):\n var_names.append(node.target.id)\n elif isinstance(node, ast.FunctionDef):\n var_names.append(node.name)\n except Exception:\n pass\n return var_names\n\n if not os.path.isdir(workspace_path):\n return False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n var_names = []\n if ext == '.py':\n var_names = get_py_vars(content)\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm']:\n var_names = get_js_vars(content)\n \n # 检查提取出的变量名是否均为小驼峰命名\n for name in var_names:\n if not is_lower_camel_case(name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含80个字符","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 如果工作区路径不存在或不是目录，则认为没有违反规则\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n # 定义常见的代码文件的扩展名集合\n code_exts = {\n '.html', '.css', '.scss', '.less',\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb',\n '.sh', '.bash'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 编码打开文件，避免 BOM 头被当成字符计算\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n for line in f:\n # 移除行中的回车符和换行符\n clean_line = line.replace('\\n', '').replace('\\r', '')\n # 根据要求明确指定的“字符数”，直接使用 len 函数统计\n if len(clean_line) > 80:\n return False\n except Exception:\n # 如果由于某种原因（如非文本编码等）导致无法正常读取，则安全地跳过\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":8,"instruction":"最后一步，来做数据持久化和体验增强。首先用浏览器的 `localStorage` 把所有的列数据和任务卡片状态保存下来，做到刷新页面不丢失数据。其次，实现一个操作撤销功能：当用户成功删除一个卡片，或者完成一次拖拽/键盘移动位置后，页面底部弹出一个全局 Toast 提示（3秒后自动消失），里面有一个'撤销'按钮。点击该按钮或者按下 `Ctrl + Z`（Mac下为 `Cmd + Z`）可以撤销刚刚的那一次操作，让卡片恢复到之前的状态和位置。另外请注意后续所有轮次都要遵循以下要求：限制你生成的代码文件的单行长度，确保每行最多包含70个字符；每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容。","instruction_following_checklist":[{"constraint":"任何回复的起始处都必须包含语气词\"汪～\"不可遗漏","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复起始处是否包含语气词\"汪～\"\n \"\"\"\n if not response:\n return False\n \n # 去除可能存在的前导空白字符，并判断是否以“汪～”开头\n return response.lstrip().startswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"所有回复都必须在结尾处加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 定义需要忽略的尾部字符（包括空白字符和常见的中英文标点符号，但不包括“～”）\n chars_to_strip = \" \\n\\r\\t.。!！?？,，;；\"\n \n # 去除尾部可能存在的空白符与标点\n clean_response = response.rstrip(chars_to_strip)\n \n # 判断是否以指定的语气词结尾\n return clean_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"对于生成的每一个代码文件，都请不要在其中书写注释","validation_code":"def check_no_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 前端项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.css'}\n \n # 构造正则表达式，按照优先级进行匹配：\n # 1. string: 匹配双引号、单引号及反引号字符串（支持跳过内部转义符，防止内部的斜杠被当做注释）\n # 2. url: 匹配常见的 URL scheme，如 http://, https://，避免将网址中的 // 误判为单行注释\n # 3. comment: 匹配多行注释 (/*...*/)、单行注释 (//...) 以及 HTML注释 (<!--...-->)\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<url>[a-zA-Z]+://)'\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/|//.*|<!--[\\s\\S]*?-->)'\n )\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 在文件内容中查找所有匹配项\n for match in pattern.finditer(content):\n # 如果匹配命中的分组是 'comment'，说明代码中存在注释\n if match.lastgroup == 'comment':\n return False\n except Exception:\n # 如果因不可预见的情况（如非文本编码）无法读取文件则跳过\n continue\n \n # 如果遍历所有代码文件都没有发现任何注释，则满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件内不允许出现多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的代码文件中不得出现while循环","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果代码目录不存在，则无法判定其违反了限制条件\n if not os.path.exists(workspace_path):\n return True\n\n # 针对看板项目，推断可能用到的前端代码文件后缀及其他常见后端后缀\n code_extensions = {'.js', '.ts', '.html', '.css', '.vue', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.cs'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除常见的块级/多行注释，避免对注释里的单词误判\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 按行处理内容，用来过滤单行注释和行内字符串\n lines = content.split('\\n')\n for line in lines:\n # 简单移除行内字符串字面量的文本，避免匹配到字符串里提到的 'while' 单词\n line = re.sub(r'\".*?\"', '\"\"', line)\n line = re.sub(r\"'.*?'\", \"''\", line)\n line = re.sub(r\"`.*?`\", \"``\", line)\n \n # 根据 '//' 或 '#' 截断当行的单行注释\n line = line.split('//')[0]\n line = line.split('#')[0]\n \n # 查找作为独立词汇出现的 while 关键字\n if re.search(r'\\bwhile\\b', line):\n return False\n except Exception:\n # 遇到异常（如非UTF-8的二进制文件、读取权限限制等）直接跳过\n continue\n\n return True","tags":["Style","Paradigm"]},{"constraint":"生成的代码项目最多只能包含2个目录层级","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 确保给定的项目路径确实存在\n if not os.path.exists(workspace_path):\n return False\n\n base_path = os.path.abspath(workspace_path)\n\n for root, dirs, files in os.walk(base_path):\n # 过滤掉以 '.' 开头的隐藏目录 (如 .git, .vscode 等)，防止这些系统/工具级目录干扰层级判定\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 计算当前目录相对于项目根目录的相对路径\n rel_path = os.path.relpath(root, base_path)\n \n # 判定目录层级\n if rel_path == '.':\n level = 1 # 根目录属于第1个目录层级\n else:\n # 通过路径分隔符拆分相对路径，计算层级深度\n parts = [p for p in rel_path.split(os.sep) if p]\n level = len(parts) + 1\n \n # 如果发现任何目录的层级超过 2，表示存在多于2个目录层级，不满足要求\n if level > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果项目目录不存在，默认不做阻拦\n if not os.path.exists(workspace_path):\n return True\n \n # 定义前端项目中涵盖代码文件的常见扩展名集\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.md'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和第三方依赖目录，避免对非自行生成的代码文件进行干扰测试\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果遇到无法读取的文件，则跳过\n continue\n \n # 按照换行符切分，以此准确计算算上所有空行的文件总行数\n lines = content.split('\\n')\n \n # 若行数为偶数，则说明未能满足要求\n if len(lines) % 2 == 0:\n return False\n \n # 所有检查完毕，未发现偶数行数的文件\n return True","tags":["Quantity","Exact Value"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 依据前端看板项目的特点推断，定义常见的代码相关文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.scss', '.less',\n '.js', '.jsx', '.ts', '.tsx', \n '.json', '.md', '.txt', '.sh', '.py'\n }\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n # 以二进制形式读取文件内容\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试将二进制数据作为 utf-8 进行解码\n raw_data.decode('utf-8')\n \n except UnicodeDecodeError:\n # 一旦发生解码错误，说明存在不是用 utf-8 编码的代码文件\n return False\n except Exception:\n # 处理诸如权限不足、文件损坏等其他无法预料的异常\n return False\n \n # 如果项目中没有任何代码文件，通常视为生成失败\n if not has_code_file:\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对看板开发项目，推测涉及的前端及相关代码文件类型\n target_exts = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte'}\n \n # 编译匹配 for 循环结构的正则表达式\n # 涵盖:\n # 1. JS/TS 原生: for(...) 或 for await(...)，利用 (?<!\\.) 排除 object.for 属性调用\n # 2. Vue 等前端框架: v-for, *ngFor, wx:for, @for\n # 3. 模板引擎或Python环境: for ... in\n for_pattern = re.compile(\n r'(?<!\\.)\\bfor\\b\\s*(?:await\\s*)?\\(|'\n r'\\bv-for\\b|'\n r'\\bngFor\\b|'\n r'\\bwx:for\\b|'\n r'@for\\b|'\n r'\\bfor\\s+[a-zA-Z0-9_,\\s]+\\s+in\\b'\n )\n \n def remove_strings_and_comments(text: str) -> str:\n # 清除常见的字符串字面量内容，防止误判包含关键字的字符串\n text = re.sub(r'\"(?:[^\"\\\\]|\\\\.)*\"', '\"\"', text)\n text = re.sub(r\"'(?:[^'\\\\]|\\\\.)*'\", \"''\", text)\n text = re.sub(r'`(?:[^`\\\\]|\\\\.)*`', '``', text)\n \n # 清除多行注释\n text = re.sub(r'/\\*[\\s\\S]*?\\*/', '', text)\n # 清除 HTML 注释\n text = re.sub(r'<!--[\\s\\S]*?-->', '', text)\n # 清除单行注释，利用负向先行断言排除如 https:// 格式的 URL 误判\n text = re.sub(r'(?<!:)\\/\\/.*', '', text)\n return text\n\n # 若该目录不存在，则说明未生成代码，默认没有违规使用\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历所有项目文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 消除字符串与注释带来的干扰\n cleaned_content = remove_strings_and_comments(content)\n \n # 若在实际执行代码结构中检索到了 for 循环，则直接返回不满足\n if for_pattern.search(cleaned_content):\n return False\n except Exception:\n # 忽略读取或编码错误的文件\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n camel_case_count = 0\n snake_case_count = 0\n pascal_case_count = 0\n \n # 匹配多单词命名的正则模式\n camel_pattern = re.compile(r'^[a-z][a-z0-9]*[A-Z][a-zA-Z0-9]*$')\n snake_pattern = re.compile(r'^[a-z0-9]+_[a-z0-9_]*[a-z0-9]+$')\n pascal_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\n \n def add_match(name):\n nonlocal camel_case_count, snake_case_count, pascal_case_count\n if camel_pattern.match(name):\n camel_case_count += 1\n elif snake_pattern.match(name):\n snake_case_count += 1\n elif pascal_pattern.match(name):\n pascal_case_count += 1\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html')):\n # 1. 匹配 let/const/var 声明的单个变量\n single_vars = re.findall(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n for v in single_vars:\n add_match(v)\n \n # 2. 匹配 function 定义的函数名\n funcs = re.findall(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n for f_name in funcs:\n add_match(f_name)\n \n # 3. 匹配使用花括号解构赋值时的变量名称 const { a, b } = obj\n dict_vars = re.findall(r'\\b(?:let|const|var)\\s*\\{([^}]+)\\}', content)\n for d in dict_vars:\n names = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', d)\n for n in names:\n add_match(n)\n \n # 4. 匹配使用方括号解构赋值时的变量名称 const [ a, b ] = arr\n array_vars = re.findall(r'\\b(?:let|const|var)\\s*\\[([^\\]]+)\\]', content)\n for a in array_vars:\n names = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', a)\n for n in names:\n add_match(n)\n \n elif file.endswith('.py'):\n # Python 变量赋值 (左侧)\n py_vars = re.findall(r'^[ \\t]*([a-zA-Z_][a-zA-Z0-9_]*)\\s*=', content, re.MULTILINE)\n for v in py_vars:\n add_match(v)\n # Python def 函数名\n py_funcs = re.findall(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(', content)\n for f_name in py_funcs:\n add_match(f_name)\n except Exception:\n continue\n \n total = camel_case_count + snake_case_count + pascal_case_count\n \n # 如果项目中没有任何复杂多词变量，无法判断直接视作容错通过\n if total == 0:\n return True \n\n # 判断条件：\n # 小驼峰的数量应当超越了蛇形的数量（说明摒弃了第4轮的蛇形要求）\n # 且不该被大驼峰完全压制（考虑到存在前端组件为大驼峰，通常小驼峰应该远超过组件数量，这里放宽为 ≥ 0.3倍即可）\n if camel_case_count > snake_case_count and camel_case_count >= pascal_case_count * 0.3:\n return True\n \n # 对于极小文件的容错补偿\n if total < 5 and camel_case_count > 0:\n return True\n \n return False","tags":["Style","Naming"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含70个字符","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 根据用户指令的任务看板项目推断可能的前端代码文件类型\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行末的换行符和回车符，保留行内原始空白符等\n clean_line = line.rstrip('\\r\\n')\n # 要求中明确使用了“字符”表述，因此直接使用 len() 统计字符数\n if len(clean_line) > 70:\n return False\n except Exception:\n # 忽略读取过程中的异常（如非UTF-8编码的特殊文件或无法读取的文件）\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个任务管理看板的项目基础骨架。页面顶部需要一个Header，包含应用标题和一个搜索框的占位。主体部分水平排列分为三列：'待办(To Do)'、'进行中(In Progress)'和'已完成(Done)'。每列里面放两三个静态的任务卡片作为占位符，卡片上显示标题和简单的描述文字。给每列的底部加上一个'添加任务'的按钮占位。先不用实现具体的交互功能，把整体UI布局和样式搭好就行，尽量做到响应式，界面美观现代一点。请注意后续所有轮次都要遵循以下要求：你所生成的所有代码文件都必须使用UTF-8进行编码，你输出的代码文件里禁止使用for循环结构。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除字符串开头的空白字符（包括空格、制表符、换行符等）后进行匹配\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除文本末尾可能的空白字符（如换行符、空格），并判断是否以指定的语气词结尾\\n return response.strip().endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check_no_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区路径不存在，判定为不满足（未生成任何文件）\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令中关于\\\"任务管理看板基础骨架\\\"和\\\"页面UI布局\\\"的需求\\n # 推断生成的主要是前端相关的代码文件。\\n valid_exts = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.vue'}\\n \\n # 构造正则表达式，将字符串匹配排在注释匹配前面，避免匹配到位于字符串内容中的注释符号\\n # Group 1: \\\"(...)\\\" 双引号字符串，支持转义\\n # Group 2: '(...)' 单引号字符串，支持转义\\n # Group 3: `(...)` 反引号（模板）字符串，支持跨行和转义\\n # Group 4: /* ... */ C-style 多行注释\\n # Group 5: <!-- ... --> HTML 注释\\n # Group 6: // 单行注释 (排除 http:// 类的误判)\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")'\\n r\\\"|('(?:\\\\\\\\.|[^'\\\\\\\\])*')\\\"\\n r\\\"|(`[\\\\s\\\\S]*?`)\\\"\\n r\\\"|(/\\\\*[\\\\s\\\\S]*?\\\\*/)\\\"\\n r\\\"|(<!--[\\\\s\\\\S]*?-->)\\\"\\n r\\\"|((?<!:)//.*)\\\",\\n re.MULTILINE\\n )\\n \\n has_code_file = False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果读取遇到异常（例如文件损坏等），忽略该文件\\n continue\\n \\n # 在文件内容中寻找正则匹配项\\n for match in pattern.finditer(content):\\n # 如果命中到了代表注释的分组（Group 4, 5, 6 之一）\\n if match.group(4) or match.group(5) or match.group(6):\\n # 找到了真实的注释，判定为未满足要求\\n return False\\n\\n # 若至少生成了一个前端代码文件，且全程未在任何文件中找到注释，判定为符合要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 如果工作区目录不存在，视为没有违反条件\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 根据用户指令推测出的Web前端代码文件后缀名\\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n # 匹配独立的单词 'while'，以此判断是否存在 while 循环\\n while_pattern = re.compile(r'\\\\bwhile\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件内容进行检查\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if while_pattern.search(content):\\n # 只要发现文件内容包含 while 关键字，即判定不满足要求\\n return False\\n except Exception:\\n # 遇到读取异常时忽略当前文件，继续检查后续文件\\n continue\\n \\n # 所有文件检查完毕，均未发现 while 循环\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 确保 workspace_path 是一个绝对路径\\n workspace = Path(workspace_path).resolve()\\n \\n # 检查路径是否存在且为目录\\n if not workspace.exists() or not workspace.is_dir():\\n return True\\n\\n # 遍历工作区下的所有目录\\n for root, dirs, files in os.walk(workspace_path):\\n root_path = Path(root).resolve()\\n try:\\n # 计算当前目录相对于根目录的相对路径\\n rel_path = root_path.relative_to(workspace)\\n except ValueError:\\n # 如果出现解析异常（理论上不会走到这里），跳过\\n continue\\n \\n # 解析 relative_to 的结果：\\n # - 根目录本身（层级1），rel_path.parts 为空元组 ()\\n # - 一级子目录，如 'src'（层级2），rel_path.parts 为 ('src',)\\n # - 二级子目录，如 'src/components'（层级3），rel_path.parts 为 ('src', 'components')\\n # 如果 parts 的长度大于等于 2，则说明已经达到了第 3 个目录层级，直接返回 False\\n if len(rel_path.parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求推断可能生成的前端及常见代码相关文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.json', '.md', '.txt'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理被判定为代码文件的类型\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按 '\\\\n' 进行切割，计算出的长度即为包含了所有空行的严格总行数\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数，如果是则说明不符合\\\"必须为奇数\\\"的要求\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 若文件因编码等异常无法被正常读取，保守起见判定为不符合\\n return False\\n\\n # 若至少存在一个代码文件，且所有代码文件的行数全为奇数，则满足该要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对前端UI项目推断的可能代码文件类型\\n code_extensions = {\\n '.html', '.htm', '.css', '.scss', '.less',\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'\\n }\\n \\n # 如果路径不存在，无法验证文件编码（视作未违反该编码要求或由其他测试点负责判断）\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅校验推测出的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制模式读取，避免平台相关的默认文本编码干扰\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用UTF-8解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦有文件无法使用UTF-8正确解码，则返回False\\n return False\\n except Exception:\\n # 忽略其他非编码相关的偶发文件读取问题\\n pass\\n \\n # 所有检查过的代码文件均通过了UTF-8解码验证\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义需要排查的文件扩展名（涵盖主流前端及常见后端语言）\\n code_exts = {\\n '.html', '.htm', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\\n '.css', '.scss', '.sass', '.less',\\n '.py', '.java', '.go', '.rs', '.php', '.cs', '.cpp', '.c', '.h', '.hpp'\\n }\\n \\n # 编译匹配多种编程语言和框架中for循环特征的正则表达式\\n for_pattern = re.compile(\\n r'\\\\bfor\\\\s*\\\\(|' # JS/TS/C/Java等: for (\\n r'\\\\bfor\\\\s+await\\\\s*\\\\(|' # JS/TS: for await (\\n r'v-for\\\\s*=|' # Vue: v-for=\\n r'\\\\*ngFor\\\\s*=|' # Angular: *ngFor=\\n r'\\\\{%\\\\s*for\\\\b|' # Jinja/Django模板: {% for \\n r'\\\\bfor\\\\s+[a-zA-Z0-9_\\\\s,]+\\\\s+in\\\\b|' # Python/模板等: for x in\\n r'@for\\\\b|' # Sass/Less: @for\\n r'\\\\bfor\\\\s+[^{]+{', # Go/Rust: for ... {\\n re.MULTILINE\\n )\\n \\n # 遍历项目所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 清理可能引起误判的注释内容\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL) # 多行注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL) # HTML注释\\n content = re.sub(r'//.*', '', content) # 单行注释 (//)\\n content = re.sub(r'#.*', '', content) # 单行注释 (#)\\n \\n # 清理可能引起误判的字符串内容（双引号、单引号、反引号）\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n # 如果仍匹配到for循环特征，直接判定为不满足\\n if for_pattern.search(content):\\n return False\\n except Exception:\\n # 如果文件不可读（比如编码错误），则跳过该文件继续检查\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来完善任务的添加和删除功能。点击每列底部的'添加任务'按钮时，原本的按钮隐藏，并在当前列底部出现一个小表单（包含标题输入框、描述文本域和确认/取消按钮）。表单填好后点击确认可以将新任务加到该列底部，点击取消则恢复成原来的'添加任务'按钮。另外，给每个任务卡片的右上角悬浮时显示一个删除图标，点击后弹窗确认，即可删除该任务卡片。请注意后续所有轮次都要遵循以下要求：处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令；运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否在起始处包含语气词\\\"汪～\\\"\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除左侧可能存在的空白字符（空格、换行等），以确保准确判断开头内容\\n cleaned_response = response.lstrip()\\n \\n # 判断是否以\\\"汪～\\\"开头\\n return cleaned_response.startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n import string\\n # 定义需要移除的末尾空白字符和常见标点符号\\n trailing_chars = string.whitespace + string.punctuation + \\\"。！？，、\\\"\\n \\n # 移除回复末尾的空白符和标点符号\\n cleaned_response = response.rstrip(trailing_chars)\\n \\n # 判断清理后的回复是否以\\\"呜～\\\"结尾\\n return cleaned_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \\n '.css', '.scss', '.sass', '.less', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.sh'\\n }\\n \\n if not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 剔除代码中的字符串字面量，防止字符串内部文本误判为注释符号\\n content_no_str = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content_no_str = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content_no_str)\\n content_no_str = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content_no_str)\\n \\n # 1. 检查 HTML/XML/Vue 等模板注释\\n if '<!--' in content_no_str:\\n return False\\n \\n # 2. 检查 CSS/JS/TS 等常见的多行块注释\\n if '/*' in content_no_str:\\n return False\\n \\n # 3. 检查单行注释 //（负向后瞻 (?<!:) 避免匹配到诸如 http:// 的场景）\\n if re.search(r'(?<!:)\\\\/\\\\/', content_no_str):\\n return False\\n \\n # 4. 检查 Python/Ruby/Shell 等脚本语言的单行注释 #\\n if ext in {'.py', '.rb', '.sh'}:\\n if '#' in content_no_str:\\n return False\\n \\n except Exception:\\n # 忽略读取失败的不可处理文件\\n pass\\n \\n # 如果遍历所有代码文件后均没有发现任何注释，满足要求，返回 True\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def verify_no_while_loops(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 涵盖前端项目以及可能涉及到的后端项目的常见代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css',\\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.sh'\\n }\\n \\n # 匹配各类代码中的 while 循环：\\n # 1. \\\\bwhile\\\\s*\\\\( 匹配 while (condition) 或 while(condition) -> 常见于JS, TS, C, Java等\\n # 2. \\\\bwhile\\\\s+[\\\\w!\\\\[\\\\'\\\\\\\"] 匹配 while 后面跟条件变量、取反符号、中括号、字符串等 -> 常见于Python, Bash等\\n while_pattern = re.compile(r'\\\\bwhile\\\\s*\\\\(|\\\\bwhile\\\\s+[\\\\w!\\\\[\\\\'\\\\\\\"]')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果匹配到 while 循环特征，则表示未满足“不得出现while循环”的要求\\n if while_pattern.search(content):\\n return False\\n except Exception:\\n # 文件若读取异常则跳过，继续检查其他文件\\n continue\\n \\n # 遍历所有代码文件均未发现 while 循环，要求满足\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_directory_levels(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for dirpath, dirnames, filenames in os.walk(workspace_path):\\n rel_path = os.path.relpath(dirpath, workspace_path)\\n \\n # 根目录本身为第1个目录层级，跳过判断\\n if rel_path == '.':\\n continue\\n\\n parts = Path(rel_path).parts\\n\\n # 忽略可能由测试环境或系统自动生成的隐藏目录（如 .git, .vscode 等）\\n if any(p.startswith('.') for p in parts):\\n continue\\n\\n # 如果相对路径的分段数量大于 1，则意味着存在第3个及以上的目录层级，违反要求\\n if len(parts) > 1:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def verify_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断，主要为包含前端页面和逻辑的项目，因此设置常见前端代码文件后缀\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 算上所有的空行，通过换行符切分可以完全贴合编辑器呈现的真实行数\\n total_lines = len(content.split('\\\\n'))\\n \\n # 如果行数为偶数，则说明违反了要求\\n if total_lines % 2 == 0:\\n return False\\n except Exception:\\n # 如果存在不可读的非文本文件导致读取失败，则忽略\\n pass\\n\\n # 如果正常遍历完且存在代码文件并未返回 False，则说明要求被满足\\n return True if has_code_file else False\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求，推断可能生成的前端及相关配置文件的扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.json', '.md', '.txt'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明并非规范的 UTF-8 编码文件\\n return False\\n except Exception:\\n # 其他读取错误跳过\\n continue\\n\\n # 如果有至少一个代码文件且都没有抛出 UnicodeDecodeError，则视为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def check_for_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义需要检查的常见代码文件扩展名\\n code_exts = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', \\n '.cs', '.go', '.rb', '.php'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除双引号、单引号、反引号包裹的字符串内容(避免字符串内包含 \\\"for\\\")\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n # 移除 /* ... */ 和 // ... 形式的注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n \\n # 移除 HTML / 模板注释 <!-- ... -->\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 移除 # ... 形式的单行注释 (适用于 Python, Ruby 等)\\n content = re.sub(r'#.*', '', content)\\n \\n # 根据文件类型准备针对性的 for 循环检测正则表达式列表\\n patterns = []\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.c', '.cpp', '.cs', '.java', '.php']:\\n patterns = [\\n r'\\\\bfor\\\\s*\\\\(', # for循环结构如 for(let i=0...) 或 for (item of items)\\n r'\\\\bfor\\\\s+await\\\\s*\\\\(' # 异步 for await 循环\\n ]\\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\\n patterns = [\\n r'\\\\bfor\\\\s*\\\\(', # 在 script 标签内的原生 JS for循环\\n r'\\\\bfor\\\\s+await\\\\s*\\\\(',\\n r'\\\\bv-for\\\\s*=', # Vue 的 for 循环指令\\n r'\\\\*ngFor\\\\s*=', # Angular 的 for 循环指令\\n r'\\\\{%\\\\s*for\\\\b' # Jinja2 / Django 等模板引擎的循环语法\\n ]\\n elif ext in ['.py', '.go', '.rb']:\\n # 这些语言里去除字符串和注释后，独立的 for 单词严格意味着该处使用了循环结构(包括列表推导式等)\\n patterns = [r'\\\\bfor\\\\b']\\n \\n # 检查当前文件是否匹配任何 for 循环结构模式\\n for pattern in patterns:\\n if re.search(pattern, content):\\n return False # 只要找到符合的 for 循环特征，则判定为未满足要求\\n \\n except Exception:\\n # 对于不可读取的特殊文件直接跳过，保证执行不被中断\\n pass\\n \\n # 所有代码文件遍历完毕，如果均未包含 for 循环相关结构，返回满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"做得不错。下一步，我要让这些卡片动起来。请实现任务卡片在**同一列内部**的上下拖拽排序。你可以使用原生的 Drag and Drop API。拖拽时，被拖拽的卡片要有视觉反馈（比如改变背景色或透明度，并且鼠标指针变成抓取形状），同时其他卡片要能动态让出空位。松开鼠标后，卡片应停留在新位置。另外请注意后续所有轮次都要遵循以下要求：运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起；请确保在代码写完之后，对所有代码进行测试并保证其成功通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证最后一轮回复的起始处是否包含语气词\\\"汪～\\\"\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除前导空白字符后检查是否以\\\"汪～\\\"开头\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_response(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（包括空格、换行符等），然后判断是否以指定的语气词结尾\\n return response.strip().endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证生成的代码项目中是否不包含任何形式的注释。\\n \\\"\\\"\\\"\\n # 常见的代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', \\n '.vue', '.py', '.java', '.c', '.cpp', '.cs', '.go', \\n '.sh', '.rb', '.php', '.swift'\\n }\\n \\n # 匹配各类注释的正则表达式\\n html_comment_pattern = re.compile(r'<!--')\\n multi_comment_pattern = re.compile(r'/\\\\*')\\n # 匹配行首或带空格前缀的单行注释，防止误匹配诸如 'http://' 等\\n single_comment_pattern = re.compile(r'(?m)^\\\\s*//|\\\\s+//')\\n # 匹配行首或带空格前缀的井号注释 (针对 Python, Shell 等)\\n hash_comment_pattern = re.compile(r'(?m)^\\\\s*#|\\\\s+#')\\n\\n # 遍历项目所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查 HTML 注释 (如 <!--...-->)\\n if html_comment_pattern.search(content):\\n return False\\n \\n # 检查多行/块级注释 (如 /*...*/ )\\n if multi_comment_pattern.search(content):\\n return False\\n \\n # 根据文件的常见语言类型，检查相关的单行注释\\n if ext not in {'.py', '.sh', '.rb'}:\\n if single_comment_pattern.search(content):\\n return False\\n else:\\n if hash_comment_pattern.search(content):\\n return False\\n \\n except Exception:\\n # 忽略以非 UTF-8 编码或出现文件读取异常的情况，避免因意外中断判断\\n continue\\n \\n # 如果遍历完毕未检测出任何匹配的注释则视为满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 如果项目路径不存在，说明未生成项目，要求视为不被破坏（或根据约定视为满足）\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义可能包含代码片段的文件扩展名集合\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', \\n '.css', '.scss', '.less', '.py', '.java', '.c', \\n '.cpp', '.go', '.sh', '.bash', '.php', '.rb', '.cs'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除注释以防误判\\n # 移除 JS/TS/C/C++/Java 等的单行注释: // ...\\n content = re.sub(r'//.*', '', content)\\n # 移除 JS/TS/C/CSS 等的多行注释: /* ... */\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 移除 HTML/Vue 的注释: <!-- ... -->\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 移除 Python/Ruby/Shell 等的单行注释: # ...\\n content = re.sub(r'#.*', '', content)\\n \\n # 移除字符串以防误判（考虑到转义字符的情况）\\n # 移除双引号字符串\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n # 移除单引号字符串\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n # 移除反引号字符串 (模板字符串)\\n content = re.sub(r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`\\\", \\\"``\\\", content, flags=re.DOTALL)\\n \\n # 查找是否包含 while 关键字\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 忽略由于编码或其他问题导致无法读取的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义需要忽略的目录：通常包含第三方依赖包、构建产物、环境配置以及版本控制等隐藏目录\\n # 它们不属于业务代码自身所创建和设计的“目录层级”\\n ignore_dirs = {\\n 'node_modules', 'venv', 'env', '__pycache__', 'dist', 'build', 'out', 'coverage'\\n }\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉忽略的目录和隐藏目录，避免向下深入遍历\\n dirs[:] = [d for d in dirs if d not in ignore_dirs and not d.startswith('.')]\\n\\n # 计算当前目录相对于项目根路径的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录本身不计算为内部层级\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 统一替换路径分隔符，将相对路径按层级拆分，以计算实际层级深度\\n parts = [p for p in rel_path.replace('\\\\\\\\', '/').split('/') if p]\\n depth = len(parts)\\n\\n # 最多只能包含 2 个目录层级，即：\\n # depth == 1: 如 src/ 属于第一层级\\n # depth == 2: 如 src/components/ 属于第二层级\\n # 若深度大于 2（例如 src/components/ui/），则说明超过了限制\\n if depth > 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check_code_lines_odd(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端和常见的代码文件后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.cpp', '.c', '.h', '.php', '.rb', '.go', '.sh'\\n }\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果不能以 utf-8 读取，直接视为不满足要求\\n return False\\n \\n # 如果文件完全为空，则行数为 0（偶数），不满足要求\\n if not content:\\n return False\\n \\n # 计算包含所有空行的总行数\\n num_lines = content.count('\\\\n') + 1\\n \\n # 若行数为偶数，返回 False\\n if num_lines % 2 == 0:\\n return False\\n\\n # 若所有代码文件均检测为奇数行，且确实生成了代码文件，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果项目路径不存在，自然不满足条件\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义前端项目及常见代码文件可能的后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx',\\n '.vue', '.svelte', '.json', '.py', '.sh', '.txt', '.md'\\n }\\n\\n found_code_file = False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码校验\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用 UTF-8 解码，若非 UTF-8 编码将抛出 UnicodeDecodeError\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n return False\\n except Exception:\\n # 忽略其他非编码相关的系统读取异常\\n pass\\n\\n # 只有在至少检测到一个代码文件并且全部都是UTF-8时才视为满足\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果代码路径尚未建立或不存在，等同于没有违规生成for循环文件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 包含前端开发中常见的代码文件扩展名及可能的后端 Python 文件\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.py'}\\n\\n def clean_code(code: str) -> str:\\n \\\"\\\"\\\"安全地剔除代码中的注释和字符串字面量，防止内容中的 for 字样造成误判\\\"\\\"\\\"\\n # 移除多行块注释 /* ... */\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除单行注释 // ... 和 Python注释 # ...\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'#.*', '', code)\\n # 移除 HTML/Vue 等文件的结构性注释 <!-- ... -->\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n \\n # 移除双引号包裹的字符串（支持转义字符）\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', code)\\n # 移除单引号包裹的字符串（支持转义字符）\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", code)\\n # 移除反引号包裹的模板字符串（支持多行和转义字符）\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', code, flags=re.DOTALL)\\n \\n return code\\n\\n # 遍历整个代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 遇到无法使用utf-8读取或其他异常的文件跳过\\n continue \\n \\n # 对源代码文本做安全清洗\\n cleaned_content = clean_code(content)\\n \\n # 验证 1: JS/TS/HTML 的标准 for 循环（例如 for (let i = 0... 或者 for(let key in...)）\\n if re.search(r'\\\\bfor\\\\s*\\\\(', cleaned_content):\\n return False\\n \\n # 验证 2: Vue 等框架模板中的 v-for 循环指令\\n if re.search(r'\\\\bv-for\\\\b', cleaned_content):\\n return False\\n \\n # 验证 3: Python 代码中的 for ... in 循环语句\\n if ext == '.py' and re.search(r'\\\\bfor\\\\s+.*?\\\\s+in\\\\b', cleaned_content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在代码写完之后，对所有代码进行测试并保证其成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在进一步扩展拖拽功能。接下来要允许任务卡片**跨列拖拽**。我需要能把'待办'里的卡片拖到'进行中'或者'已完成'列。当拖拽的卡片悬停在其他列的区域上方时，目标列的背景色要轻微加深以提示这里可以放置。放入新列后，卡片要插入到对应位置，且底层数据结构里的状态（所属列）也要同步更新。另外请注意后续所有轮次都要遵循以下要求：在你完成代码的编写之后，不允许开展任何测试；生成代码文件时，请务必使用蛇形命名规则来定义各个变量名。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 去除回复开头可能存在的空白字符（如空格、换行符等），然后判断是否以\\\"汪～\\\"开头\\n if not response:\\n return False\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾可能的空白字符\\n cleaned_response = response.strip()\\n \\n # 去除末尾可能存在的常见标点符号和部分Markdown格式符号\\n # 注意此处避免移除全角波浪号“～”或半角波浪号“~”\\n cleaned_response = re.sub(r'[。！!？\\\\.\\\\n\\\\r\\\\t \\\\*\\\"\\\\'`]+$', '', cleaned_response)\\n \\n # 检查最终字符串是否以规定的\\\"呜～\\\"语气词结尾，并兼容半角波浪号情况\\n if cleaned_response.endswith(\\\"呜～\\\") or cleaned_response.endswith(\\\"呜~\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 路径不合法或不存在，无法有效验证\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n\\n # 常见的代码文件后缀\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.sh', '.java', '.c', '.cpp', '.h', '.php', '.rb', '.go'\\n }\\n \\n # 排除的常规依赖及编译产物目录\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', '__pycache__', 'venv', '.venv'}\\n\\n def check_no_comment(file_path: str, ext: str) -> bool:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果文件无法以 utf-8 读取（例如二进制文件），跳过不做强校验\\n return True \\n \\n # 1. 检查多行注释 (HTML风格和C风格)\\n if ext in {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.java', '.c', '.cpp', '.h', '.php', '.go'}:\\n if re.search(r'<!--[\\\\s\\\\S]*?-->', content):\\n return False\\n if re.search(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content):\\n return False\\n \\n # 2. 检查单行注释\\n lines = content.split('\\\\n')\\n for line in lines:\\n line_strip = line.strip()\\n # 过滤空行\\n if not line_strip:\\n continue\\n \\n # 去除包裹在引号内的字符串内容，防止误判内部出现的URL `//` 或 色值 `#FFF` 等符号\\n line_no_str = re.sub(r'\\\".*?\\\"', '\\\"\\\"', line_strip)\\n line_no_str = re.sub(r\\\"'.*?'\\\", \\\"''\\\", line_no_str)\\n line_no_str = re.sub(r\\\"`.*?`\\\", \\\"``\\\", line_no_str)\\n \\n # 检查 Python / Shell / Ruby 风格的 '#' 注释\\n if ext in {'.py', '.sh', '.rb'}:\\n if line_strip.startswith('#!'):\\n continue\\n # 非字符串区域一旦出现 '#' 即判定为注释\\n if '#' in line_no_str:\\n return False\\n \\n # 检查 JS / TS / C / Java 风格的 '//' 注释\\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.java', '.c', '.cpp', '.h', '.php', '.go'}:\\n # 非字符串区域一旦出现 '//' 即判定为注释\\n if '//' in line_no_str:\\n return False\\n \\n return True\\n\\n # 遍历指定工作空间内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤无需检查的目录树层级\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n # 检查该文件里是否包含任何注释\\n if not check_no_comment(file_path, ext):\\n # 发现有一处注释，违背要求，立即返回False\\n return False\\n \\n # 所有代码文件遍历通过，没有发现任何注释，符合要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 若代码目录不存在，则认为没有生成违规的代码\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 根据指令推测的前端项目及常见的代码文件后缀\\n code_extensions = {\\n '.html', '.js', '.jsx', '.ts', '.tsx', '.css', '.vue', '.svelte',\\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php'\\n }\\n\\n # 遍历项目目录下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 跨语言使用正则表达式匹配完整的单词 \\\"while\\\"\\n # 因为 while 通常为保留字，匹配到意味着存在 while 循环（忽略字符串极小的误判可能）\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 忽略读取异常的文件\\n continue\\n\\n # 遍历完所有代码文件均未发现 while 循环\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉以 '.' 开头的隐藏目录，避免将如 .git 等环境自带目录误判为生成的层级，同时防止递归进入\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 获取当前遍历到的目录相对于根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 如果是根目录本身，跳过\\n if rel_path == '.':\\n continue\\n \\n # 统一路径分隔符为 '/'，处理不同操作系统下的路径差异\\n rel_path = rel_path.replace(os.sep, '/')\\n \\n # 将相对路径按照 '/' 进行拆分，获取路径深度的级数\\n parts = [p for p in rel_path.split('/') if p]\\n \\n # 根目录为第1层级，其直接子目录相对路径长度为1（即第2层级）。\\n # 如果 parts 的长度大于1，说明出现了第3层级及以上的目录，不符合最多只能包含2个目录层级的要求。\\n if len(parts) > 1:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的前端与后端代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_files = False\\n\\n # 遍历当前工作区的所有目录和文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 以 \\\\n 切分能准确得出包含空行的实际物理行数\\n # 例如 \\\"a\\\\nb\\\" 长度为 2; \\\"a\\\\nb\\\\n\\\" 长度为 3（最后一行为空行）\\n line_count = len(content.split('\\\\n'))\\n \\n # 如果行数是偶数，说明不满足奇数行的要求\\n if line_count % 2 == 0:\\n return False\\n except Exception:\\n # 如果由于非文本或其他原因导致读取/解码错误，则略过当前文件\\n pass\\n \\n # 如果没有任何代码文件，判断为不满足\\n return has_code_files\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前后端代码文件扩展名白名单，可以涵盖任务看板项目以及其他大多数开发项目\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', \\n '.rb', '.php', '.swift', '.kt', '.json', '.xml', '.yaml', '.yml', \\n '.sh', '.bat', '.ps1', '.sql', '.md', '.txt'\\n }\\n \\n # 如果工作区路径不存在，通常意味着未生成项目\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的非源代码生成的包目录、编译目录或版本控制目录\\n for skip_dir in ['.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode']:\\n if skip_dir in dirs:\\n dirs.remove(skip_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试用 utf-8 解码，以判断是否属于合法的 utf-8 编码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件存在非 UTF-8 字符\\n return False\\n except Exception:\\n # 对于其他潜在的文件读取异常（如权限问题），选择忽略并继续\\n continue\\n \\n # 所有的代码文件都能被 utf-8 正确解码\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def check_no_for_loop_in_code(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见代码文件扩展名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', \\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', \\n '.php', '.rb', '.swift', '.kt'\\n }\\n\\n # 涵盖不同语言或框架中for循环结构的正则表达式\\n patterns = [\\n r'\\\\bfor\\\\b\\\\s*\\\\(', # C系列、JS、TS等原生的 for ( 或 for(\\n r'\\\\bv-for\\\\b\\\\s*=', # Vue 的 v-for=\\n r'\\\\*?ngFor\\\\b\\\\s*=', # Angular 的 *ngFor=\\n r'\\\\{%\\\\s*for\\\\b', # Django/Jinja 模板里的 {% for \\n r'^\\\\s*for\\\\b\\\\s+.*\\\\s+in\\\\b', # Python 等通过 in 遍历的结构\\n r'\\\\bfor\\\\b\\\\s+.*\\\\{' # Go 等不加括号直接跟 { 的结构\\n ]\\n \\n # 编译正则表达式，忽略大小写并支持多行匹配模式\\n combined_pattern = re.compile('|'.join(patterns), re.MULTILINE)\\n\\n if not os.path.exists(workspace_path):\\n # 如果路径不存在，无法验证生成的代码，视作未能满足\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 如果在文件中找到了for循环结构，则返回False\\n if combined_pattern.search(content):\\n return False\\n except Exception:\\n # 遇到无法读取的情况，跳过该文件\\n continue\\n\\n # 遍历结束未发现 for 循环，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你完成代码的编写之后，不允许开展任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请务必使用蛇形命名规则来定义各个变量名\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 用于匹配前端文件中的变量与函数声明（如 let task_name, const MAX_COUNT, function add_task）\\n decl_pattern = re.compile(r'\\\\b(?:var|let|const|function)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 如果是 HTML 文件，则仅提取 <script> 标签内的 JS 代码\\n if file.endswith('.html'):\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL)\\n code_to_check = '\\\\n'.join(scripts)\\n else:\\n code_to_check = content\\n \\n if not code_to_check.strip():\\n continue\\n\\n # 移除多行注释和单行注释，避免注释内带有无关的驼峰命名引起误判\\n code_to_check = re.sub(r'/\\\\*.*?\\\\*/', '', code_to_check, flags=re.DOTALL)\\n code_to_check = re.sub(r'//.*', '', code_to_check)\\n\\n for match in decl_pattern.findall(code_to_check):\\n var_name = match\\n \\n # 校验是否为合法的蛇形命名。\\n # 合法的蛇形命名：全小写（或全大写常量），只允许包含字母、数字和下划线\\n is_lower_snake = re.match(r'^[a-z0-9_]+$', var_name)\\n is_upper_snake = re.match(r'^[A-Z0-9_]+$', var_name)\\n \\n if not is_lower_snake and not is_upper_snake:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"拖拽很顺畅了。现在来实现页面顶部的搜索功能。当我在顶部的搜索框输入文字时，需要实时过滤面板里的所有的任务卡片。只有标题或描述中包含所输入关键词的卡片才显示，其余的卡片隐藏。注意，当某列没有任何符合过滤条件的卡片时，该列需要显示类似'暂无匹配任务'的灰色提示文字，不要变成完全空白。另外请注意后续所有轮次都要遵循以下要求：生成代码文件时，请务必使用大驼峰命名规则来定义各个变量名；限制你生成的代码文件的单行长度，确保每行最多包含120个字符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除可能存在的首部空白字符，然后判断是否以目标语气词开头\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符\\n response = response.strip()\\n \\n # 使用正则表达式匹配结尾处是否为\\\"呜～\\\"，允许其后紧跟空白字符或常见的句末标点符号\\n pattern = r'呜～[\\\\s\\\\.\\\\!。！\\\\?？]*$'\\n if re.search(pattern, response):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def remove_strings(text: str) -> str:\\n \\\"\\\"\\\"\\n 移除代码内容中双引号、单引号及反引号字符串内部的内容，\\n 避免将字符串中包含的类似于注释的符号（如 \\\"//\\\"、\\\"/*\\\"）误判为注释。\\n \\\"\\\"\\\"\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", text)\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', text)\\n return text\\n\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm',\\n '.css', '.scss', '.less', '.py', '.sh', '.java', '.cpp', '.c', '.go'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码相关文件\\n if ext not in code_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n clean_content = remove_strings(content)\\n \\n # 检查 HTML 风格注释\\n if ext in {'.html', '.htm', '.vue', '.jsx', '.tsx'}:\\n if re.search(r'<!--[\\\\s\\\\S]*?-->', clean_content):\\n return False\\n \\n # 检查 C 风格多行注释 /* ... */\\n if ext in {'.html', '.htm', '.vue', '.jsx', '.tsx', '.js', '.ts', '.css', '.scss', '.less', '.java', '.cpp', '.c', '.go'}:\\n if re.search(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', clean_content):\\n return False\\n \\n # 检查 C 风格单行注释 // (利用负向后视断言排除前置是冒号的情况，例如 URL)\\n if ext in {'.html', '.htm', '.vue', '.jsx', '.tsx', '.js', '.ts', '.java', '.cpp', '.c', '.go'}:\\n if re.search(r'(?<!:)//.*', clean_content):\\n return False\\n \\n # 检查 Python/Shell 风格单行注释 #\\n if ext in {'.py', '.sh', '.rb', '.yml', '.yaml'}:\\n if re.search(r'(?:^|\\\\s)#.*', clean_content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm',\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs',\\n '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.dart', '.scss', '.sass'\\n }\\n \\n def check_while_loop(content: str) -> bool:\\n # 1. 移除字符串字面量 (考虑了 \\\\ 转义符)\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n # 2. 移除多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 3. 移除单行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n \\n # 4. 检查各种编程语言中 while 循环的语法特征\\n patterns = [\\n r'\\\\bwhile\\\\s*\\\\(', # JS, TS, Java, C++, C, C#, PHP etc.\\n r'\\\\bwhile\\\\s+[^:\\\\n]+:', # Python\\n r'\\\\bwhile\\\\s+.*?;?\\\\s*do\\\\b', # Bash\\n r'@while\\\\b' # SCSS\\n ]\\n \\n for p in patterns:\\n if re.search(p, content):\\n return True\\n return False\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if check_while_loop(content):\\n # 出现了 while 循环，违反给定要求\\n return False \\n except Exception:\\n pass\\n \\n # 没有在任何代码文件中发现 while 循环，满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_directory_levels(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区目录不存在，则不认为违反该规则\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义需要忽略的常见构建/环境/缓存/隐藏目录，防止对正常生成的层级造成误判\\n exclude_dirs = {\\n 'node_modules', '__pycache__', 'venv', '.venv', 'env', \\n '.git', '.idea', '.vscode', 'build', 'dist'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的系统/第三方目录，以及以点开头的隐藏目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\\n \\n rel_path = os.path.relpath(root, workspace_path)\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 按照操作系统的路径分隔符进行分割，计算相对层级深度\\n depth = len(rel_path.split(os.sep))\\n \\n # 如果层级深度大于2，则表明项目包含超过2个目录层级，返回 False\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据前端看板项目，定义可能包含的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计总行数（算上所有空行）\\n if not content:\\n line_count = 0\\n else:\\n line_count = len(content.split('\\\\n'))\\n \\n # 判断行数是否为偶数，若为偶数则不满足要求\\n if line_count % 2 == 0:\\n return False\\n except Exception:\\n # 如果由于非UTF-8编码等原因无法读取文件，直接判断为不合格\\n return False\\n \\n # 确保项目中至少生成了一个有效代码文件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 如果工作区路径不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 常见代码文件和配置文件的扩展名集合，主要针对前端项目及常见语言\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.json', '.xml', '.yaml', '.yml', '.md', '.txt',\\n '.py', '.java', '.c', '.cpp', '.h', '.sh', '.bat'\\n }\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件，排除图片、字体等可能的二进制文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦发生解码错误，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他如权限不足等读取异常\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见代码文件后缀名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \\n '.vue', '.py', '.java', '.c', '.cpp', '.go', '.rs', '.cs'\\n }\\n \\n # 正则表达式：用于移除代码中的单行/多行注释以及HTML注释，防止干扰判断\\n comment_pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|<!--.*?-->|#.*?$', \\n re.MULTILINE | re.DOTALL\\n )\\n \\n # 正则表达式：用于匹配各大常见语言与前端框架的 for 循环语法结构\\n for_pattern = re.compile(\\n r'\\\\bfor\\\\b\\\\s*(?:await\\\\s+)?\\\\(|' # JS/TS/Java/C++ 风格: for (...) 或 for await (...)\\n r'\\\\bfor\\\\b\\\\s+[a-zA-Z_][a-zA-Z0-9_,\\\\s]*\\\\s+in\\\\b|' # Python 风格: for item in ...\\n r'\\\\bv-for\\\\s*=|' # Vue 指令: v-for=\\n r'\\\\*ngFor\\\\s*=|' # Angular 指令: *ngFor=\\n r'\\\\bfor\\\\b\\\\s+.*?\\\\{' # Go/Rust 风格: for ... {\\n )\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 清理注释代码\\n clean_content = comment_pattern.sub('', content)\\n \\n # 匹配是否包含 for 循环结构\\n if for_pattern.search(clean_content):\\n return False\\n except Exception:\\n # 如果有文件读取错误直接跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你完成代码的编写之后，不允许开展任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请务必使用大驼峰命名规则来定义各个变量名\",\n \"validation_code\": \"def check_pascal_case_vars(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 用于匹配 js/ts/jsx/tsx/html/vue 等前端文件中，基本变量与函数的声明\\n # 仅匹配关键字后紧跟的普通标识符，以此跳过解构赋值的复杂正则处理\\n js_var_pattern = re.compile(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n js_func_pattern = re.compile(r'\\\\b(?:function|class)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n # 用于匹配 Python 中的变量赋值及函数、类的声明\\n py_var_pattern = re.compile(r'^[ \\\\t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \\\\t]*=[ \\\\t]*[^=]', re.MULTILINE)\\n py_func_pattern = re.compile(r'^[ \\\\t]*(?:def|class)[ \\\\t]+([a-zA-Z_][a-zA-Z0-9_]*)\\\\b', re.MULTILINE)\\n\\n def is_pascal_case(name: str) -> bool:\\n # 满足大驼峰条件：首字母必须大写，且全部由大小写字母及数字组成，不能包含下划线或其他特殊字符\\n return bool(re.match(r'^[A-Z][a-zA-Z0-9]*$', name))\\n\\n # 遍历项目目录下的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 校验前端主流代码文件\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.vue']:\\n vars_found = js_var_pattern.findall(content)\\n funcs_found = js_func_pattern.findall(content)\\n for var_name in vars_found + funcs_found:\\n if not is_pascal_case(var_name):\\n return False\\n\\n # 校验 Python 文件\\n elif ext == '.py':\\n vars_found = py_var_pattern.findall(content)\\n funcs_found = py_func_pattern.findall(content)\\n for var_name in vars_found + funcs_found:\\n if not is_pascal_case(var_name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含 120 个字符\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端与相关代码文件后缀名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.json'}\\n # 定义需要跳过扫描的目录（如依赖库或版本控制）\\n exclude_dirs = {'node_modules', '.git', 'build', 'dist', 'venv', '__pycache__'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉无需扫描的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查指定的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的换行符进行严格长度统计\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 120:\\n return False\\n except Exception:\\n # 如果遇到编码错误或权限问题导致无法读取，则跳过该文件继续检查\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"我们需要让任务的管理维度更丰富。请给任务数据添加'优先级'（高、中、低）属性。在添加任务的表单中增加优先级的下拉选择，并且在卡片上用不同颜色的标签（如红、黄、蓝）展示优先级。同时，在顶部搜索框旁边也增加一个'优先级'下拉筛选菜单。现在过滤规则变成：搜索关键词和优先级筛选要能够**组合过滤**（即同时满足关键词匹配和选中的优先级才显示）。另外请注意后续所有轮次都要遵循以下要求：请确保在代码写完之后，对所有代码进行测试并保证其成功通过；限制你生成的代码文件的单行长度，确保每行最多包含80个字符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 清理掉回复开头可能存在的空白字符（如空格、换行等），以确保准确匹配\\n cleaned_response = response.lstrip()\\n \\n # 验证回复的起始处是否为\\\"汪～\\\"\\n if cleaned_response.startswith(\\\"汪～\\\"):\\n return True\\n else:\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除回复末尾的空白字符（空格、换行等）以及常见的标点符号\\n cleaned_response = re.sub(r'[\\\\s。，！、？.!,?]*$', '', response)\\n \\n # 判断清理后的回复是否以“呜～”结尾\\n if cleaned_response.endswith('呜～'):\\n return True\\n \\n # 作为一个后备方案，如果遇到非预期标点，但严格的结尾就是“呜～”也算通过\\n if response.strip().endswith('呜～'):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def check_no_comments_in_js(content: str) -> bool:\\n # 移除常见的 URL 协议前缀，防止在没有引号包裹（例如正则字面量）时引起 // 的误判\\n content = content.replace(\\\"http://\\\", \\\"\\\").replace(\\\"https://\\\", \\\"\\\")\\n # 此正则表达式通过优先级匹配：字符串字面量会被整体匹配从而保护内部可能含有的 // 或 /*\\n pattern = re.compile(\\n r'//.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.MULTILINE\\n )\\n for match in pattern.finditer(content):\\n s = match.group(0)\\n if s.startswith('//') or s.startswith('/*'):\\n return True # 发现注释\\n return False\\n\\n def check_no_comments_in_html(content: str) -> bool:\\n # 检查 script 标签内是否有 JS 注释\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n # 检查 style 标签内是否有 CSS 注释\\n styles = re.findall(r'<style.*?>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n \\n for script in scripts:\\n if check_no_comments_in_js(script):\\n return True\\n for style in styles:\\n if check_no_comments_in_js(style):\\n return True\\n \\n # 移除 scripts 和 styles 标签及其内容，防止内部的字符串由于包含 <!-- 而被误判为 HTML 注释\\n content_no_script_style = re.sub(r'<script.*?>[\\\\s\\\\S]*?</script>', '', content, flags=re.IGNORECASE)\\n content_no_script_style = re.sub(r'<style.*?>[\\\\s\\\\S]*?</style>', '', content_no_script_style, flags=re.IGNORECASE)\\n \\n # 匹配 HTML 注释\\n if re.search(r'<!--[\\\\s\\\\S]*?-->', content_no_script_style):\\n return True\\n\\n return False\\n\\n def check_no_comments_in_py(content: str) -> bool:\\n content = content.replace(\\\"http://\\\", \\\"\\\").replace(\\\"https://\\\", \\\"\\\")\\n pattern = re.compile(\\n r'#.*?$|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"',\\n re.MULTILINE\\n )\\n for match in pattern.finditer(content):\\n s = match.group(0)\\n if s.startswith('#'):\\n # 排除 shebang (#!/bin/...) 以及 编码声明 (# coding: utf-8)\\n if not s.startswith('#!') and not re.match(r'#\\\\s*-*-.*coding', s):\\n return True\\n return False\\n\\n # 若工作区路径不存在，视作未生成任何文件\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_files = False\\n js_like_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp', '.cs', '.go'}\\n html_like_exts = {'.html', '.htm', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n try:\\n if ext in js_like_exts:\\n has_code_files = True\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if check_no_comments_in_js(content):\\n return False\\n \\n elif ext in html_like_exts:\\n has_code_files = True\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if check_no_comments_in_html(content):\\n return False\\n \\n elif ext == '.py':\\n has_code_files = True\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if check_no_comments_in_py(content):\\n return False\\n except Exception:\\n pass\\n \\n # 只要正常生成了项目代码文件，且均未检测到注释，即满足要求\\n return has_code_files\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区路径为空或不存在，说明未生成任何代码文件，不存在违规情况\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n\\n # 推断可能的代码文件类型，排除纯文档类型如 .md, .txt，避免正常单词导致的误判\\n code_exts = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.css', '.vue', '.svelte', '.py', '.java', \\n '.c', '.cpp', '.h', '.go', '.rs', '.php', '.rb'\\n }\\n\\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过正则表达式 \\\\bwhile\\\\b 全词匹配检索代码中是否使用了 while 关键字\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 如果由于某种原因无法以utf-8读取，直接忽略继续下一个\\n continue\\n \\n # 如果所有的代码文件中均未发现 while 关键字，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n try:\\n workspace = Path(workspace_path).resolve()\\n \\n # 确保给定路径存在且是目录\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for path in workspace.rglob('*'):\\n rel_path = path.relative_to(workspace)\\n \\n # 过滤掉路径中包含隐藏目录的部分（如 .git）\\n if any(part.startswith('.') for part in rel_path.parts):\\n continue\\n \\n # 计算该项目的目录层级\\n if path.is_dir():\\n # 对于目录，它的层级即是 parts 的长度 + 1\\n # len(parts) == 0 时，是根目录（第1层）\\n # len(parts) == 1 时，是一级子目录（第2层）\\n # len(parts) >= 2 时，是二级及以上子目录（第3层及以上），不符合要求\\n if len(rel_path.parts) >= 2:\\n return False\\n else:\\n # 对于文件，它自身不算作目录，其所在的目录就是它的 parent\\n # len(parts[:-1]) == 0 时，在根目录下（第1层目录内）\\n # len(parts[:-1]) == 1 时，在一级子目录下（第2层目录内）\\n # len(parts[:-1]) >= 2 时，在二级及以上子目录下，不符合要求\\n if len(rel_path.parts[:-1]) >= 2:\\n return False\\n \\n return True\\n except Exception:\\n # 如果解析路径或遍历时发生意外错误，判定为不满足\\n return False\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推测出的Web前端等相关代码文件后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.json', '.py', '.java', '.c', '.cpp', '.h'\\n }\\n\\n if not os.path.exists(workspace_path):\\n # 如果未生成目录或文件，则不存在违反要求的情况\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符拆分，这种方法可以准确算上所有的空行\\n # 例如内容为 \\\"a\\\\n\\\\nb\\\" 会被分为 ['a', '', 'b'] 共3行\\n # 包含文件最末尾如果有单一换行符也会被正确算入总行数中\\n lines = content.split('\\\\n')\\n \\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果遇到文件读取异常（如非 UTF-8 编码），无法完成校验，返回False\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的后缀名及无后缀的代码文件名称\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json',\\n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.h', \\n '.hpp', '.cs', '.go', '.rs', '.rb', '.php', '.sh', \\n '.yaml', '.yml', '.xml', '.md', '.svg'\\n }\\n code_exact_names = {'dockerfile', 'makefile'}\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions or file.lower() in code_exact_names:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生 UnicodeDecodeError，说明该文件并未完全按照 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他类型的异常（如由于系统权限导致的读取失败），仅关注编码问题\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def check_no_for_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见前端及后端代码扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.css', '.vue', '.py', '.java', '.cpp', '.c', \\n '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n\\n def remove_strings_and_comments(text: str, ext: str) -> str:\\n # 去除双引号、单引号、反引号包裹的字符串内容\\n text = re.sub(r'([\\\"\\\\'`]).*?(?<!\\\\\\\\)\\\\1', '', text, flags=re.DOTALL)\\n \\n # 移除各类型文件的注释\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java', '.go', '.rs', '.css', '.vue']:\\n # 移除块级注释 /* ... */\\n text = re.sub(r'/\\\\*.*?\\\\*/', '', text, flags=re.DOTALL)\\n # 移除单行注释 // ...\\n text = re.sub(r'//.*', '', text)\\n if ext in ['.html', '.htm', '.xml', '.svg', '.vue']:\\n # 移除 HTML 注释 <!-- ... -->\\n text = re.sub(r'<!--.*?-->', '', text, flags=re.DOTALL)\\n if ext in ['.py', '.sh', '.rb']:\\n # 移除 Python 或 Shell 中的 # 注释\\n text = re.sub(r'#.*', '', text)\\n \\n return text\\n\\n # 正则表达式：\\n # 1. 匹配不以点号开头的独立关键字 for \\n # 2. 匹配前端框架中的特定的 for 指令如 v-for 或 *ngFor\\n for_pattern = re.compile(r'(?<!\\\\.)\\\\bfor\\\\b')\\n framework_for_pattern = re.compile(r'\\\\bv-for\\\\b|\\\\*ngFor\\\\b')\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作区目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 先移除字符串与注释\\n clean_content = remove_strings_and_comments(content, ext)\\n \\n # 针对前端标签中 `<label for=\\\"...` 或 `<label htmlFor=\\\"...` 等可能遗留的属性形式进行清理\\n # 此时字符串已被清理掉，因此它大概率变成 for= 或 htmlFor= 等形式\\n clean_content = re.sub(r'\\\\bfor\\\\s*=', '', clean_content)\\n clean_content = re.sub(r'\\\\bhtmlFor\\\\s*=', '', clean_content)\\n \\n # 若依然能匹配到 for，则说明使用了 for 循环\\n if for_pattern.search(clean_content) or framework_for_pattern.search(clean_content):\\n return False\\n except Exception:\\n # 忽略不可读文件（如非文本编码）等异常继续检查\\n pass\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在代码写完之后，对所有代码进行测试并保证其成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请务必使用大驼峰命名规则来定义各个变量名\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 匹配前端代码 (JavaScript, TypeScript等) 中常规的变量声明 (let/var/const)\\n js_var_pattern = re.compile(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)')\\n \\n # 匹配 Python 代码中的行首赋值变量声明\\n py_var_pattern = re.compile(r'^\\\\s*([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=[^=]', re.MULTILINE)\\n \\n # 大驼峰命名规则校验: 以大写字母开头，后续允许包含字母和数字，不能有下划线\\n pascal_case_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\\n \\n # 需要检查的文件后缀集合\\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.svelte', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n vars_found = []\\n \\n # 提取前端语言的变量名\\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.svelte'}:\\n vars_found.extend(js_var_pattern.findall(content))\\n \\n # 提取Python后端的变量名\\n if ext == '.py':\\n vars_found.extend(py_var_pattern.findall(content))\\n \\n # 验证是否每一个提取到的变量名都符合大驼峰规则\\n for var_name in vars_found:\\n if not pascal_case_pattern.match(var_name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含80个字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义前端项目中常见代码文件的扩展名集合\\n code_extensions = {\\n '.html', '.css', '.scss', '.less', \\n '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.json'\\n }\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理可能的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 移除行尾的换行符再统计实际可见的字符数\\n line_content = line.rstrip('\\\\r\\\\n')\\n if len(line_content) > 80:\\n return False\\n except Exception:\\n # 如果遇到不可读取的文件（如二进制文件异常导致），直接跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"看板的基础功能很完善了，现在我们要重点关注键盘无障碍访问（a11y）。请确保页面上所有的交互元素（搜索输入框、优先级下拉框、添加按钮、任务卡片、删除图标）都可以通过 `Tab` 键和 `Shift + Tab` 键顺畅地获取焦点。当元素获得焦点（:focus）时，需要有一圈明显的轮廓（outline）高亮。当焦点在删除按钮或添加确认按钮上时，按下 `Enter` 或 `Space` 键可以触发点击事件。另外请注意后续所有轮次都要遵循以下要求：生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名；当你把代码写完后，不得进行任何测试。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否在起始处包含语气词\\\"汪～\\\"。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，代表该要求的满足情况。\\n \\\"\\\"\\\"\\n # 去除回复开头的空白字符（包括空格、换行符等），判断是否以\\\"汪～\\\"开头\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除末尾的空白字符\\n cleaned_response = response.strip()\\n \\n # 去除可能出现在末尾的常见标点符号，避免因为句末标点导致判断失败\\n cleaned_response = cleaned_response.rstrip(\\\"。，！？.,!?\\\\n\\\\r\\\\t \\\")\\n \\n # 检查是否以“呜～”结尾\\n return cleaned_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手生成的代码项目中，是否没有在任何代码文件里书写注释。\\n \\\"\\\"\\\"\\n def remove_strings(text):\\n # 移除 Python 风格的多行字符串\\n text = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", \\\"''\\\", text)\\n # 移除双引号和单引号字符串，并兼容其内部的转义符号处理\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", text)\\n # 移除 JS 模板字符串的反引号形式\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', text)\\n return text\\n\\n def has_js_comments(text):\\n # 判断是否存在多行注释\\n if '/*' in text:\\n return True\\n # 判断是否存在单行注释，并避开URL中的 :// 与被转义的 \\\\/\\\\/ \\n if re.search(r'(?<![:\\\\\\\\])//', text):\\n return True\\n return False\\n\\n def has_py_comments(text):\\n # 移除可能的脚本头声明 shebang，例如 #!/usr/bin/env python\\n text = re.sub(r'^#!.*', '', text)\\n # 在字符串被移空的前提下，任何游离的 # 都可以认为是注释标记\\n if '#' in text:\\n return True\\n return False\\n\\n # 按注释符号的差异性划定不同代码类型\\n js_exts = {\\n '.js', '.jsx', '.ts', '.tsx', '.java', '.c', '.cpp', \\n '.cs', '.go', '.php', '.swift', '.kt', '.rs', '.dart',\\n '.css', '.scss', '.less'\\n }\\n py_exts = {'.py', '.rb', '.sh', '.yml', '.yaml'}\\n html_exts = {'.html', '.htm', '.vue'}\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in js_exts and ext not in py_exts and ext not in html_exts:\\n continue\\n\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 清洗剔除掉字符串以防误判注释符号\\n content = remove_strings(content)\\n\\n # 根据扩展名分别实施检测策略\\n if ext in js_exts:\\n if has_js_comments(content):\\n return False\\n elif ext in py_exts:\\n if has_py_comments(content):\\n return False\\n elif ext in html_exts:\\n # 检查 HTML 自有的注释符号\\n if '<!--' in content:\\n return False\\n \\n # 抽离 HTML/Vue 中自带的 Script 和 Style 块，对它们以 JS/CSS 的规则处理\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n if has_js_comments(script):\\n return False\\n \\n styles = re.findall(r'<style\\\\b[^>]*>(.*?)</style>', content, flags=re.IGNORECASE | re.DOTALL)\\n for style in styles:\\n if has_js_comments(style):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作目录不存在，无法进行实质性代码检查，默认不违背该特定要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 定义可能包含代码的常见文件扩展名（以前端及主流语言为主）\\n code_extensions = {\\n '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n # 预处理：为了防止误判，剔除代码文件中的字符串字面量\\n content = re.sub(r'\\\".*?\\\"|\\\\'.*?\\\\'|`.*?`', '', content, flags=re.DOTALL)\\n \\n # 预处理：剔除代码文件中的注释（尽管提示中规定不得生成注释，为以防万一依然做剔除处理）\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL) # C/JS风格多行注释\\n content = re.sub(r'//.*', '', content) # C/JS风格单行注释\\n if ext in {'.py', '.rb'}:\\n content = re.sub(r'#.*', '', content) # Python/Ruby风格单行注释\\n\\n # 核心检查：判断去除干扰内容后是否包含 while 循环的关键字\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 如果由于编码或其他问题导致个别文件无法读取，在此略过\\n continue\\n\\n # 如果遍历所有的代码文件均未发现 while 循环，说明符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n \\n workspace = Path(workspace_path)\\n \\n # 确保工作区是一个存在的目录\\n if not workspace.exists() or not workspace.is_dir():\\n return True\\n \\n for path in workspace.rglob('*'):\\n # 仅针对目录进行层级判断\\n if path.is_dir():\\n # 获取当前目录相对于项目根目录的相对路径\\n rel_path = path.relative_to(workspace)\\n \\n # rel_path.parts 拆分相对路径\\n # 例如 \\\"dir1\\\" 的 parts 为 ('dir1',)，长度为 1 (第 2 层级)\\n # 例如 \\\"dir1/dir2\\\" 的 parts 为 ('dir1', 'dir2')，长度为 2 (第 3 层级)\\n if len(rel_path.parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目类型（前端看板）推断可能的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.json', '.md', '.txt', '.sh', '.py'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按换行符分割计算行数，能够准确包含代码中的所有的空行\\n # 对于空文件，split('\\\\n') 结果为 [''] 长度为 1；这也和主流编辑器显示一致\\n lines = content.split('\\\\n')\\n total_lines = len(lines)\\n \\n # 如果行数是偶数，说明违反了必须为奇数的限制\\n if total_lines % 2 == 0:\\n return False\\n \\n except UnicodeDecodeError:\\n # 如果不能被 UTF-8 解码为文本文件，则安全忽略\\n pass\\n except Exception:\\n # 忽略如权限不足等其他读取异常\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前端/通用代码文件后缀\\n code_extensions = {\\n '.html', '.htm',\\n '.css', '.scss', '.less',\\n '.js', '.jsx',\\n '.ts', '.tsx',\\n '.json', '.vue',\\n '.md', '.txt'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明该文件不是使用UTF-8编码的\\n return False\\n except Exception:\\n # 对于其他的异常（如权限问题等），暂时略过\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def has_for_loop(content: str) -> bool:\\n # 去除单行和多行注释，避免注释内的文本产生干扰\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content = re.sub(r'#.*', '', content)\\n \\n # 去除字符串字面量，避免字符串内容中包含 \\\"for\\\" 导致误判\\n content = re.sub(r'\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:[^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`[^`]*`', '``', content)\\n \\n # 1. 匹配 C/C++/Java/JS/TS 风格的 for 循环 (如 \\\"for(\\\" 或 \\\"for (\\\")\\n if re.search(r'\\\\bfor\\\\s*\\\\(', content):\\n return True\\n \\n # 2. 匹配 异步 JS 循环 (如 \\\"for await (\\\")\\n if re.search(r'\\\\bfor\\\\s+await\\\\s*\\\\(', content):\\n return True\\n \\n # 3. 匹配 Python 风格的 for 循环 (如 \\\"for item in list:\\\")\\n if re.search(r'\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s]+\\\\s+in\\\\s+', content):\\n return True\\n \\n # 4. 匹配 Vue 或 Angular 等前端模板里的 for 循环指令 (如 \\\"v-for\\\" 或 \\\"*ngFor\\\")\\n if re.search(r'\\\\bv-for\\\\b', content) or re.search(r'\\\\*ngFor\\\\b', content):\\n return True\\n \\n # 5. 匹配 Jinja2 等后端模板中的 for\\n if re.search(r'\\\\{%\\\\s*for\\\\b', content):\\n return True\\n \\n return False\\n\\n # 囊括前端常见的代码相关文件扩展名（结合可能的后端语言）\\n valid_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue', '.css', '.py', '.java', '.cpp', '.c'}\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if has_for_loop(content):\\n # 若发现任何形式的 for 循环，则违背了“禁止使用”的要求\\n return False\\n except Exception:\\n # 若由于不可预知编码等原因无法读取，直接略过该文件\\n continue\\n \\n # 如果遍历检查完所有代码文件均无 for 循环，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"当你把代码写完后，不得进行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n def is_lower_camel_case(name: str) -> bool:\\n # 豁免 Python 中的魔法变量（例如 __init__, __name__ 等）\\n if name.startswith('__') and name.endswith('__'):\\n return True\\n \\n # 去除前导常见的标识符修饰（如内部变量使用的 _ 和 $ ）\\n clean_name = name.lstrip('_$')\\n \\n if not clean_name:\\n return True\\n \\n # 首字母必须是小写字母（若首字母是大写则为大驼峰，全大写则不为小写）\\n if not clean_name[0].islower():\\n return False\\n \\n # 字符串中间不应包含下划线（杜绝蛇形或者全大写带下划线的情况）\\n if '_' in clean_name:\\n return False\\n \\n return True\\n\\n # 简单的正则匹配 js/ts 等文件中常见的前端变量声明\\n # 这能够抓取大多数由 let, var, const 声明的单个变量名\\n js_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n\\n # 用于解析 Python 文件中赋值变量名的 AST Visitor\\n class PyVarVisitor(ast.NodeVisitor):\\n def __init__(self):\\n self.variables = []\\n\\n def visit_Name(self, node):\\n # 记录处于存储上下文（被赋值）的变量名\\n if isinstance(node.ctx, ast.Store):\\n self.variables.append(node.id)\\n self.generic_visit(node)\\n\\n def visit_arg(self, node):\\n # 记录函数的参数名\\n self.variables.append(node.arg)\\n self.generic_visit(node)\\n\\n found_any_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n found_any_file = True\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if file.endswith('.py'):\\n try:\\n tree = ast.parse(content)\\n visitor = PyVarVisitor()\\n visitor.visit(tree)\\n for var in visitor.variables:\\n if not is_lower_camel_case(var):\\n return False\\n except Exception:\\n pass\\n \\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.vue', '.svelte')):\\n matches = js_pattern.findall(content)\\n for match in matches:\\n if not is_lower_camel_case(match):\\n return False\\n\\n # 若项目目录中没有任何文件，默认认为未生成代码项目\\n if not found_any_file:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含80个字符\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 定义前端和常规项目的常见代码文件扩展名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', \\n '.html', '.css', '.vue', \\n '.json', '.py', '.mjs', '.cjs'\\n }\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的换行符后计算字符数，按规则明确要求直接使用 len() 函数\\n if len(line.rstrip('\\\\r\\\\n')) > 80:\\n return False\\n except Exception:\\n # 忽略不可读取或非标准 UTF-8 编码的特殊文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"接下来是一个具挑战性的无障碍功能：用键盘移动任务卡片，因为纯鼠标拖拽对某些用户不友好。当焦点在某个任务卡片上时，按下 `Space` 键使其进入'选取移动'模式（卡片边框闪烁或特殊高亮）。在此模式下，按下 `上/下` 方向键可以在当前列内上下移动卡片排序，按下 `左/右` 方向键可以将卡片跨列移动到相邻的列。调整到理想位置后，按下 `Enter` 键确认放置并退出该模式，按下 `Escape` 键则取消移动，让卡片弹回初始位置。另外请注意后续所有轮次都要遵循以下要求：写完代码后，必须对所有的代码进行测试以确保测试通过；在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否以\\\"汪～\\\"开头。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前代码项目地址\\n :return: 布尔值，代表是否满足要求\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除开头的空白字符后判断是否以指定语气词开头\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断回复内容的结尾处是否包含\\\"呜～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（如空格、换行等）\\n clean_resp = response.strip()\\n \\n # 1. 严格匹配：直接判断是否以“呜～”结尾\\n if clean_resp.endswith(\\\"呜～\\\"):\\n return True\\n \\n # 2. 兼容匹配：考虑到结尾可能会带有一些标点符号，去除末尾常见的标点后再进行判断\\n import string\\n # 常见的英文和中文标点符号集合\\n punctuations = string.punctuation + \\\"。！？，、；：“”‘’（）【】《》\\\"\\n \\n # 剥离末尾的所有标点符号\\n stripped_resp = clean_resp.rstrip(punctuations)\\n \\n if stripped_resp.endswith(\\\"呜～\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n def remove_strings(text: str, is_python: bool = False) -> str:\\n \\\"\\\"\\\"\\n 移除代码中所有的字符串字面量内容，避免字符串内部的字符被误判为注释。\\n \\\"\\\"\\\"\\n if is_python:\\n # 移除Python特有的多行字符串\\n text = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", \\\"''\\\", text)\\n \\n # 移除双引号和单引号字符串（支持转义字符）\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", text)\\n \\n if not is_python:\\n # 移除 JS/TS/Vue 的反引号模板字符串\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', text)\\n \\n return text\\n\\n # 定义不同类型注释的匹配模式\\n regex_html = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n regex_c_block = re.compile(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n regex_c_line = re.compile(r'(?<!:)//.*')\\n\\n # 需要检查的代码文件类型\\n target_extensions = {\\n '.html', '.vue', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.py', '.sh', '.bash', '.yaml', '.yml'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in target_extensions:\\n continue\\n\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue # 忽略无法读取的文件\\n\\n # 根据扩展名判断文件类型并剥离字符串\\n is_script = ext in ['.py', '.sh', '.bash', '.yaml', '.yml']\\n is_python = ext == '.py'\\n \\n clean_content = remove_strings(content, is_python=is_python)\\n\\n if is_script:\\n lines = clean_content.split('\\\\n')\\n for i, line in enumerate(lines):\\n # 忽略脚本第一行的 shebang\\n if i == 0 and line.startswith('#!'):\\n continue\\n # 如果行内存在 # (不在单词中间)，则视为存在注释\\n if re.search(r'(^|\\\\s)#', line):\\n return False\\n else:\\n # HTML / Vue 文件\\n if ext in ['.html', '.vue']:\\n if regex_html.search(clean_content) or regex_c_block.search(clean_content) or regex_c_line.search(clean_content):\\n return False\\n \\n # JS / TS 类型文件\\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if regex_c_block.search(clean_content) or regex_c_line.search(clean_content):\\n return False\\n \\n # CSS 类型文件\\n elif ext in ['.css', '.scss', '.less']:\\n if regex_c_block.search(clean_content) or regex_c_line.search(clean_content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查工作空间路径是否存在\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 预估可能包含循环逻辑的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.css', '.py'}\\n \\n # 匹配 JS/TS/HTML 的单/多行注释以及字符串（包含反引号字面量）\\n js_pattern = re.compile(\\n r'//.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`',\\n re.MULTILINE\\n )\\n \\n # 匹配 Python 的注释和字符串\\n py_pattern = re.compile(\\n r'#.*?$|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'',\\n re.MULTILINE\\n )\\n \\n # 匹配完整单词的 while\\n while_pattern = re.compile(r'\\\\bwhile\\\\b')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 过滤掉文件中的字符串和注释以防误判\\n if ext == '.py':\\n clean_content = py_pattern.sub('', content)\\n else:\\n clean_content = js_pattern.sub('', content)\\n \\n # 如果清洗后的代码中依然包含 while，说明存在while循环，直接返回False\\n if while_pattern.search(clean_content):\\n return False\\n except Exception:\\n # 如果出现不可预知的文件读取错误则跳过该文件继续判断\\n continue\\n\\n # 如果所有文件都没有命中while循环逻辑，则判定为符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"import os\\n\\ndef check_directory_levels(response: str, workspace_path: str) -> bool:\\n # 如果工作区不存在或不是目录，则不存在违背情况\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs，过滤掉测试环境、第三方依赖或IDE自动生成的目录，避免引起误判\\n ignored_dirs = {'.git', 'node_modules', '__pycache__', '.pytest_cache', '.vscode', '.idea'}\\n dirs[:] = [d for d in dirs if d not in ignored_dirs]\\n \\n # 获取当前目录相对于根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录本身为第1层级，跳过\\n if rel_path == '.':\\n continue\\n \\n # 根据系统路径分隔符拆分相对路径\\n parts = rel_path.split(os.sep)\\n \\n # 根目录的直接子目录为第2层级，其相对路径拆分后的长度必然为1\\n # 如果长度大于 1，则说明存在第3层级及以上的目录，不满足要求\\n if len(parts) > 1:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否有效\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n \\n # 推断可能的代码文件类型\\n code_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.svelte'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果遇到文件无法读取的情况，则跳过\\n continue\\n \\n # 算上所有的空行计算总行数\\n # 换行符的数量加1即为文本的总行数\\n line_count = content.count('\\\\n') + 1\\n \\n # 判断总行数是否为偶数\\n if line_count % 2 == 0:\\n return False\\n \\n has_code_file = True\\n \\n # 确保至少检查过一个生成的代码文件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的代码文件扩展名，以Web前端相关的为主，并包含其他主流语言\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.css', '.scss', '.sass', '.less', '.json', \\n '.vue', '.svelte', '.py', '.java', '.c', \\n '.cpp', '.h', '.hpp', '.cs', '.go', '.rb', \\n '.php', '.sh', '.bash', '.md', '.txt'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用utf-8对文件内容进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出UnicodeDecodeError，说明文件包含无法用UTF-8解析的字节，即非UTF-8编码\\n return False\\n except Exception:\\n # 对于不可抗力的读取错误，默认略过\\n pass\\n \\n # 所有代码文件均成功按照UTF-8解码\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n def has_for_loop(content: str) -> bool:\\n # 去除干扰项，避免误杀字符串或注释中的 'for' 文本\\n # 1. 移除多行注释 (JS/TS/CSS/C/Java等)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 2. 移除单行注释 (JS/TS/C/Java等)\\n content = re.sub(r'//.*', '', content)\\n # 3. 移除单行注释 (Python/Ruby/Shell，虽然可能误伤CSS颜色值，但不影响控制流解析)\\n content = re.sub(r'#.*', '', content)\\n # 4. 移除HTML注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 5. 移除双引号字符串\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', content)\\n # 6. 移除单引号字符串\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', content)\\n # 7. 移除反引号模板字符串\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', content, flags=re.DOTALL)\\n \\n # 定义各类 for 循环结构的匹配规则\\n patterns = [\\n r'\\\\bfor\\\\b\\\\s*(?:await\\\\s*)?\\\\(', # JS/TS/Java/C/C++/C# 等语言的 for 循环\\n r'\\\\bv-for\\\\b', # Vue 框架的 v-for 循环指令\\n r'\\\\bngFor\\\\b', # Angular 框架的 ngFor 循环指令\\n r'\\\\bfor\\\\b\\\\s+[A-Za-z_][A-Za-z0-9_,\\\\s]*\\\\s+in\\\\b', # Python 的 for 循环结构\\n r'\\\\bfor\\\\b\\\\s+[^{]*\\\\{', # Go 语言的 for 循环结构\\n r'@for\\\\b' # Sass 等样式预处理器的循环结构\\n ]\\n \\n for pattern in patterns:\\n if re.search(pattern, content):\\n return True\\n \\n return False\\n\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.scss', '.less',\\n '.vue', '.svelte', '.py', '.java', '.go', '.rb', '.php', '.c', \\n '.cpp', '.cs', '.sh'\\n }\\n\\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 若发现包含 for 循环结构，则直接判定为不满足要求\\n if has_for_loop(content):\\n return False\\n except Exception:\\n # 如果无法读取（如编码错误等），跳过\\n continue\\n \\n # 只要存在至少一个代码文件且都通过了无 for 循环测试，即认定满足\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 判断是否为小驼峰命名\\n def is_lower_camel_case(name: str) -> bool:\\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\\n \\n def get_js_vars(content: str) -> list:\\n var_names = []\\n # 匹配 var, let, const 声明的简单变量名\\n pattern = r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:=|;|,|:|$|\\\\n)'\\n for m in re.finditer(pattern, content):\\n var_names.append(m.group(1))\\n \\n # 匹配 function 定义的函数名（在JS中也属于变量作用域中的定义）\\n func_pattern = r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('\\n for m in re.finditer(func_pattern, content):\\n var_names.append(m.group(1))\\n \\n return var_names\\n\\n def get_py_vars(content: str) -> list:\\n var_names = []\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Assign):\\n for target in node.targets:\\n if isinstance(target, ast.Name):\\n var_names.append(target.id)\\n elif isinstance(node, ast.AnnAssign):\\n if isinstance(node.target, ast.Name):\\n var_names.append(node.target.id)\\n elif isinstance(node, ast.FunctionDef):\\n var_names.append(node.name)\\n except Exception:\\n pass\\n return var_names\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n var_names = []\\n if ext == '.py':\\n var_names = get_py_vars(content)\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm']:\\n var_names = get_js_vars(content)\\n \\n # 检查提取出的变量名是否均为小驼峰命名\\n for name in var_names:\\n if not is_lower_camel_case(name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含80个字符\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 如果工作区路径不存在或不是目录，则认为没有违反规则\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n # 定义常见的代码文件的扩展名集合\\n code_exts = {\\n '.html', '.css', '.scss', '.less',\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb',\\n '.sh', '.bash'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 编码打开文件，避免 BOM 头被当成字符计算\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n for line in f:\\n # 移除行中的回车符和换行符\\n clean_line = line.replace('\\\\n', '').replace('\\\\r', '')\\n # 根据要求明确指定的“字符数”，直接使用 len 函数统计\\n if len(clean_line) > 80:\\n return False\\n except Exception:\\n # 如果由于某种原因（如非文本编码等）导致无法正常读取，则安全地跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"最后一步，来做数据持久化和体验增强。首先用浏览器的 `localStorage` 把所有的列数据和任务卡片状态保存下来，做到刷新页面不丢失数据。其次，实现一个操作撤销功能：当用户成功删除一个卡片，或者完成一次拖拽/键盘移动位置后，页面底部弹出一个全局 Toast 提示（3秒后自动消失），里面有一个'撤销'按钮。点击该按钮或者按下 `Ctrl + Z`（Mac下为 `Cmd + Z`）可以撤销刚刚的那一次操作，让卡片恢复到之前的状态和位置。另外请注意后续所有轮次都要遵循以下要求：限制你生成的代码文件的单行长度，确保每行最多包含70个字符；每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何回复的起始处都必须包含语气词\\\"汪～\\\"不可遗漏\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的回复起始处是否包含语气词\\\"汪～\\\"\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除可能存在的前导空白字符，并判断是否以“汪～”开头\\n return response.lstrip().startswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复都必须在结尾处加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 定义需要忽略的尾部字符（包括空白字符和常见的中英文标点符号，但不包括“～”）\\n chars_to_strip = \\\" \\\\n\\\\r\\\\t.。!！?？,，;；\\\"\\n \\n # 去除尾部可能存在的空白符与标点\\n clean_response = response.rstrip(chars_to_strip)\\n \\n # 判断是否以指定的语气词结尾\\n return clean_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于生成的每一个代码文件，都请不要在其中书写注释\",\n \"validation_code\": \"def check_no_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 前端项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.css'}\\n \\n # 构造正则表达式，按照优先级进行匹配：\\n # 1. string: 匹配双引号、单引号及反引号字符串（支持跳过内部转义符，防止内部的斜杠被当做注释）\\n # 2. url: 匹配常见的 URL scheme，如 http://, https://，避免将网址中的 // 误判为单行注释\\n # 3. comment: 匹配多行注释 (/*...*/)、单行注释 (//...) 以及 HTML注释 (<!--...-->)\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<url>[a-zA-Z]+://)'\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*|<!--[\\\\s\\\\S]*?-->)'\\n )\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 在文件内容中查找所有匹配项\\n for match in pattern.finditer(content):\\n # 如果匹配命中的分组是 'comment'，说明代码中存在注释\\n if match.lastgroup == 'comment':\\n return False\\n except Exception:\\n # 如果因不可预见的情况（如非文本编码）无法读取文件则跳过\\n continue\\n \\n # 如果遍历所有代码文件都没有发现任何注释，则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内不允许出现多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中不得出现while循环\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果代码目录不存在，则无法判定其违反了限制条件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 针对看板项目，推断可能用到的前端代码文件后缀及其他常见后端后缀\\n code_extensions = {'.js', '.ts', '.html', '.css', '.vue', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.cs'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除常见的块级/多行注释，避免对注释里的单词误判\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 按行处理内容，用来过滤单行注释和行内字符串\\n lines = content.split('\\\\n')\\n for line in lines:\\n # 简单移除行内字符串字面量的文本，避免匹配到字符串里提到的 'while' 单词\\n line = re.sub(r'\\\".*?\\\"', '\\\"\\\"', line)\\n line = re.sub(r\\\"'.*?'\\\", \\\"''\\\", line)\\n line = re.sub(r\\\"`.*?`\\\", \\\"``\\\", line)\\n \\n # 根据 '//' 或 '#' 截断当行的单行注释\\n line = line.split('//')[0]\\n line = line.split('#')[0]\\n \\n # 查找作为独立词汇出现的 while 关键字\\n if re.search(r'\\\\bwhile\\\\b', line):\\n return False\\n except Exception:\\n # 遇到异常（如非UTF-8的二进制文件、读取权限限制等）直接跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含2个目录层级\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 确保给定的项目路径确实存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n base_path = os.path.abspath(workspace_path)\\n\\n for root, dirs, files in os.walk(base_path):\\n # 过滤掉以 '.' 开头的隐藏目录 (如 .git, .vscode 等)，防止这些系统/工具级目录干扰层级判定\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 计算当前目录相对于项目根目录的相对路径\\n rel_path = os.path.relpath(root, base_path)\\n \\n # 判定目录层级\\n if rel_path == '.':\\n level = 1 # 根目录属于第1个目录层级\\n else:\\n # 通过路径分隔符拆分相对路径，计算层级深度\\n parts = [p for p in rel_path.split(os.sep) if p]\\n level = len(parts) + 1\\n \\n # 如果发现任何目录的层级超过 2，表示存在多于2个目录层级，不满足要求\\n if level > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"算上所有的空行，每个生成的代码文件的总行数必须保持为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果项目目录不存在，默认不做阻拦\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义前端项目中涵盖代码文件的常见扩展名集\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.md'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和第三方依赖目录，避免对非自行生成的代码文件进行干扰测试\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果遇到无法读取的文件，则跳过\\n continue\\n \\n # 按照换行符切分，以此准确计算算上所有空行的文件总行数\\n lines = content.split('\\\\n')\\n \\n # 若行数为偶数，则说明未能满足要求\\n if len(lines) % 2 == 0:\\n return False\\n \\n # 所有检查完毕，未发现偶数行数的文件\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 依据前端看板项目的特点推断，定义常见的代码相关文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.scss', '.less',\\n '.js', '.jsx', '.ts', '.tsx', \\n '.json', '.md', '.txt', '.sh', '.py'\\n }\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n # 以二进制形式读取文件内容\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试将二进制数据作为 utf-8 进行解码\\n raw_data.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 一旦发生解码错误，说明存在不是用 utf-8 编码的代码文件\\n return False\\n except Exception:\\n # 处理诸如权限不足、文件损坏等其他无法预料的异常\\n return False\\n \\n # 如果项目中没有任何代码文件，通常视为生成失败\\n if not has_code_file:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对看板开发项目，推测涉及的前端及相关代码文件类型\\n target_exts = {'.html', '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte'}\\n \\n # 编译匹配 for 循环结构的正则表达式\\n # 涵盖:\\n # 1. JS/TS 原生: for(...) 或 for await(...)，利用 (?<!\\\\.) 排除 object.for 属性调用\\n # 2. Vue 等前端框架: v-for, *ngFor, wx:for, @for\\n # 3. 模板引擎或Python环境: for ... in\\n for_pattern = re.compile(\\n r'(?<!\\\\.)\\\\bfor\\\\b\\\\s*(?:await\\\\s*)?\\\\(|'\\n r'\\\\bv-for\\\\b|'\\n r'\\\\bngFor\\\\b|'\\n r'\\\\bwx:for\\\\b|'\\n r'@for\\\\b|'\\n r'\\\\bfor\\\\s+[a-zA-Z0-9_,\\\\s]+\\\\s+in\\\\b'\\n )\\n \\n def remove_strings_and_comments(text: str) -> str:\\n # 清除常见的字符串字面量内容，防止误判包含关键字的字符串\\n text = re.sub(r'\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'(?:[^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", text)\\n text = re.sub(r'`(?:[^`\\\\\\\\]|\\\\\\\\.)*`', '``', text)\\n \\n # 清除多行注释\\n text = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', text)\\n # 清除 HTML 注释\\n text = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', text)\\n # 清除单行注释，利用负向先行断言排除如 https:// 格式的 URL 误判\\n text = re.sub(r'(?<!:)\\\\/\\\\/.*', '', text)\\n return text\\n\\n # 若该目录不存在，则说明未生成代码，默认没有违规使用\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历所有项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 消除字符串与注释带来的干扰\\n cleaned_content = remove_strings_and_comments(content)\\n \\n # 若在实际执行代码结构中检索到了 for 循环，则直接返回不满足\\n if for_pattern.search(cleaned_content):\\n return False\\n except Exception:\\n # 忽略读取或编码错误的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 60s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请务必使用小驼峰命名规则来定义各个变量名\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n camel_case_count = 0\\n snake_case_count = 0\\n pascal_case_count = 0\\n \\n # 匹配多单词命名的正则模式\\n camel_pattern = re.compile(r'^[a-z][a-z0-9]*[A-Z][a-zA-Z0-9]*$')\\n snake_pattern = re.compile(r'^[a-z0-9]+_[a-z0-9_]*[a-z0-9]+$')\\n pascal_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\\n \\n def add_match(name):\\n nonlocal camel_case_count, snake_case_count, pascal_case_count\\n if camel_pattern.match(name):\\n camel_case_count += 1\\n elif snake_pattern.match(name):\\n snake_case_count += 1\\n elif pascal_pattern.match(name):\\n pascal_case_count += 1\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html')):\\n # 1. 匹配 let/const/var 声明的单个变量\\n single_vars = re.findall(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n for v in single_vars:\\n add_match(v)\\n \\n # 2. 匹配 function 定义的函数名\\n funcs = re.findall(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n for f_name in funcs:\\n add_match(f_name)\\n \\n # 3. 匹配使用花括号解构赋值时的变量名称 const { a, b } = obj\\n dict_vars = re.findall(r'\\\\b(?:let|const|var)\\\\s*\\\\{([^}]+)\\\\}', content)\\n for d in dict_vars:\\n names = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', d)\\n for n in names:\\n add_match(n)\\n \\n # 4. 匹配使用方括号解构赋值时的变量名称 const [ a, b ] = arr\\n array_vars = re.findall(r'\\\\b(?:let|const|var)\\\\s*\\\\[([^\\\\]]+)\\\\]', content)\\n for a in array_vars:\\n names = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', a)\\n for n in names:\\n add_match(n)\\n \\n elif file.endswith('.py'):\\n # Python 变量赋值 (左侧)\\n py_vars = re.findall(r'^[ \\\\t]*([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=', content, re.MULTILINE)\\n for v in py_vars:\\n add_match(v)\\n # Python def 函数名\\n py_funcs = re.findall(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(', content)\\n for f_name in py_funcs:\\n add_match(f_name)\\n except Exception:\\n continue\\n \\n total = camel_case_count + snake_case_count + pascal_case_count\\n \\n # 如果项目中没有任何复杂多词变量，无法判断直接视作容错通过\\n if total == 0:\\n return True \\n\\n # 判断条件：\\n # 小驼峰的数量应当超越了蛇形的数量（说明摒弃了第4轮的蛇形要求）\\n # 且不该被大驼峰完全压制（考虑到存在前端组件为大驼峰，通常小驼峰应该远超过组件数量，这里放宽为 ≥ 0.3倍即可）\\n if camel_case_count > snake_case_count and camel_case_count >= pascal_case_count * 0.3:\\n return True\\n \\n # 对于极小文件的容错补偿\\n if total < 5 and camel_case_count > 0:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含70个字符\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 根据用户指令的任务看板项目推断可能的前端代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行末的换行符和回车符，保留行内原始空白符等\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 要求中明确使用了“字符”表述，因此直接使用 len() 统计字符数\\n if len(clean_line) > 70:\\n return False\\n except Exception:\\n # 忽略读取过程中的异常（如非UTF-8编码的特殊文件或无法读取的文件）\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["页面主体结构：查看页面主体是否水平划分为三列，且各列的顶部标题分别为'待办(To Do)'、'进行中(In Progress)'和'已完成(Done)'。","Header区域检查：查看页面顶部是否有一个 Header 区域，其中包含了应用标题、一个搜索输入框以及一个优先级下拉筛选菜单。","任务卡片基础信息展示：查看列表中的任务卡片，验证卡片上是否展示了任务标题、描述文字，以及用不同颜色标签（如红、黄、蓝）标示的优先级（高、中、低）。","添加任务交互-展开表单：点击任意一列底部的'添加任务'按钮，验证该按钮本身隐藏，并在原位置展开一个包含标题输入框、描述文本域、优先级下拉选择以及确认/取消按钮的表单。","添加任务交互-取消添加：在展开的添加任务表单中点击'取消'按钮，验证该表单消失，并在原位置恢复显示原本的'添加任务'按钮。","添加任务交互-确认添加：在展开的添加任务表单中填写标题、描述并选择一个优先级，点击'确认'按钮，验证表单变回'添加任务'按钮，且该列最底部成功新增了一张包含刚才填写内容的任务卡片。","删除功能-悬浮显示图标：模拟鼠标悬停在任意一个任务卡片上，验证该卡片的右上角出现删除图标；鼠标移开后图标消失。","删除功能-确认删除：点击任务卡片右上角的删除图标，验证弹出确认提示框；模拟点击确认后，验证该任务卡片从对应的列表中移除并消失。","同列拖拽-视觉反馈：鼠标按住某列中的任务卡片并稍微移动（不松开鼠标），验证被拖拽的卡片发生视觉改变（如变色或透明度改变，鼠标变成抓取形状），并且同列的其他卡片动态移动让出空位。","同列拖拽-位置排序：将按住的卡片拖拽到同列的另一个位置上方并松开鼠标，验证卡片成功停留在新位置，列内卡片的排列顺序发生对应的改变。","跨列拖拽-悬停反馈：按住'待办'列的一个卡片，将其拖拽并悬停至'进行中'列的区域上方（不松手），验证目标列（进行中列）的整体背景色发生轻微加深变化作为放置提示。","跨列拖拽-成功跨列放置：紧接上一步，在目标列上方松开鼠标，验证卡片从原先的列中消失，并成功插入到了目标列对应的位置显示，且底层数据结构中该任务的所属列/状态同步更新。","搜索过滤功能：在顶部的搜索框内输入某一个存在的任务卡片标题包含的关键词，验证看板中只显示标题或描述包含该关键词的卡片，其他不匹配的卡片被隐藏。","空列提示展示：在搜索框输入一串不存在的乱码字符使得某列没有任何卡片匹配，验证该列不会变成完全空白，而是显示出'暂无匹配任务'或类似的灰色提示文字。","组合过滤功能：在搜索框保留输入的关键词，同时在顶部的优先级下拉菜单中选择一个指定的优先级（如'高'），验证页面仅展示既包含搜索关键词，优先级又符合下拉选择的任务卡片。","无障碍焦点控制：连续按下键盘的 Tab 键和 Shift + Tab 键，验证焦点能够依次在搜索框、优先级筛选、添加按钮、任务卡片、删除图标等交互元素上顺畅正向和反向切换，且获得焦点的元素外部有一圈明显的轮廓（outline）高亮。","无障碍键盘点击：通过 Tab 键将焦点分别移动到某个任务卡片的删除图标和添加任务表单的确认按钮上，按下键盘上的 Enter 或 Space 键，验证能够成功唤起对应的删除确认弹窗或完成任务添加操作。","键盘移动-进入选取模式：通过 Tab 键将焦点移动到某个任务卡片整体上，按下 Space（空格）键，验证该卡片进入'选取移动'模式（表现为边框闪烁或特殊高亮包裹）。","键盘移动-同列上下排序：在卡片处于选取移动模式时，按下键盘的上/下方向键，验证该卡片在当前所在的列内上下移动以改变排列顺序。","键盘移动-跨列左右移动：在卡片处于选取移动模式时，按下键盘的左/右方向键，验证该卡片能从当前列跳跃移动到左侧或右侧的相邻列中。","键盘移动-回车确认放置：使用方向键将处于移动模式的卡片调整到一个新位置后，按下 Enter 键，验证卡片的闪烁或高亮消失，退出移动模式并固定在这个最终位置。","键盘移动-Esc取消操作：让某个卡片进入选取移动模式并通过方向键移动它，在未按Enter确认前按下 Escape 键，验证卡片立刻飞回最初所在的列和位置，并退出移动模式。","全局撤销提示(Toast)：通过鼠标拖拽、键盘移动或点击删除某张卡片后，验证页面底部弹出一个包含'撤销'按钮的全局 Toast 提示，且静置等待3秒钟后，该 Toast 自动消失。","点击撤销按钮恢复原状：拖拽移动一个卡片到新列触发 Toast 提示，在 Toast 提示自动消失前点击其中的'撤销'按钮，验证刚刚被移动的卡片立刻恢复到拖动之前的所属列和位置，且 Toast 关闭。","快捷键撤销功能：分别执行删除卡片、鼠标拖拽移动卡片、键盘移动卡片操作后，立即按下键盘的 Ctrl + Z（或 Mac 下的 Cmd + Z），验证刚才执行的操作被撤销，卡片恢复到操作前的状态和位置，且底部的 Toast 提示随之消失。","数据持久化存储：在看板中执行添加任务、删除任务、拖拽移动任务、键盘移动任务操作后，手动刷新浏览器页面，验证页面重新加载后，所有任务的内容、优先级、所属列和排列顺序均保持最新状态，没有发生数据丢失。","LocalStorage写入检查：打开浏览器开发者工具的Application面板，验证看板的列数据、任务内容、优先级、所属列和排序状态确实被保存到了localStorage中。"],"string":"[\n \"页面主体结构：查看页面主体是否水平划分为三列，且各列的顶部标题分别为'待办(To Do)'、'进行中(In Progress)'和'已完成(Done)'。\",\n \"Header区域检查：查看页面顶部是否有一个 Header 区域，其中包含了应用标题、一个搜索输入框以及一个优先级下拉筛选菜单。\",\n \"任务卡片基础信息展示：查看列表中的任务卡片，验证卡片上是否展示了任务标题、描述文字，以及用不同颜色标签（如红、黄、蓝）标示的优先级（高、中、低）。\",\n \"添加任务交互-展开表单：点击任意一列底部的'添加任务'按钮，验证该按钮本身隐藏，并在原位置展开一个包含标题输入框、描述文本域、优先级下拉选择以及确认/取消按钮的表单。\",\n \"添加任务交互-取消添加：在展开的添加任务表单中点击'取消'按钮，验证该表单消失，并在原位置恢复显示原本的'添加任务'按钮。\",\n \"添加任务交互-确认添加：在展开的添加任务表单中填写标题、描述并选择一个优先级，点击'确认'按钮，验证表单变回'添加任务'按钮，且该列最底部成功新增了一张包含刚才填写内容的任务卡片。\",\n \"删除功能-悬浮显示图标：模拟鼠标悬停在任意一个任务卡片上，验证该卡片的右上角出现删除图标；鼠标移开后图标消失。\",\n \"删除功能-确认删除：点击任务卡片右上角的删除图标，验证弹出确认提示框；模拟点击确认后，验证该任务卡片从对应的列表中移除并消失。\",\n \"同列拖拽-视觉反馈：鼠标按住某列中的任务卡片并稍微移动（不松开鼠标），验证被拖拽的卡片发生视觉改变（如变色或透明度改变，鼠标变成抓取形状），并且同列的其他卡片动态移动让出空位。\",\n \"同列拖拽-位置排序：将按住的卡片拖拽到同列的另一个位置上方并松开鼠标，验证卡片成功停留在新位置，列内卡片的排列顺序发生对应的改变。\",\n \"跨列拖拽-悬停反馈：按住'待办'列的一个卡片，将其拖拽并悬停至'进行中'列的区域上方（不松手），验证目标列（进行中列）的整体背景色发生轻微加深变化作为放置提示。\",\n \"跨列拖拽-成功跨列放置：紧接上一步，在目标列上方松开鼠标，验证卡片从原先的列中消失，并成功插入到了目标列对应的位置显示，且底层数据结构中该任务的所属列/状态同步更新。\",\n \"搜索过滤功能：在顶部的搜索框内输入某一个存在的任务卡片标题包含的关键词，验证看板中只显示标题或描述包含该关键词的卡片，其他不匹配的卡片被隐藏。\",\n \"空列提示展示：在搜索框输入一串不存在的乱码字符使得某列没有任何卡片匹配，验证该列不会变成完全空白，而是显示出'暂无匹配任务'或类似的灰色提示文字。\",\n \"组合过滤功能：在搜索框保留输入的关键词，同时在顶部的优先级下拉菜单中选择一个指定的优先级（如'高'），验证页面仅展示既包含搜索关键词，优先级又符合下拉选择的任务卡片。\",\n \"无障碍焦点控制：连续按下键盘的 Tab 键和 Shift + Tab 键，验证焦点能够依次在搜索框、优先级筛选、添加按钮、任务卡片、删除图标等交互元素上顺畅正向和反向切换，且获得焦点的元素外部有一圈明显的轮廓（outline）高亮。\",\n \"无障碍键盘点击：通过 Tab 键将焦点分别移动到某个任务卡片的删除图标和添加任务表单的确认按钮上，按下键盘上的 Enter 或 Space 键，验证能够成功唤起对应的删除确认弹窗或完成任务添加操作。\",\n \"键盘移动-进入选取模式：通过 Tab 键将焦点移动到某个任务卡片整体上，按下 Space（空格）键，验证该卡片进入'选取移动'模式（表现为边框闪烁或特殊高亮包裹）。\",\n \"键盘移动-同列上下排序：在卡片处于选取移动模式时，按下键盘的上/下方向键，验证该卡片在当前所在的列内上下移动以改变排列顺序。\",\n \"键盘移动-跨列左右移动：在卡片处于选取移动模式时，按下键盘的左/右方向键，验证该卡片能从当前列跳跃移动到左侧或右侧的相邻列中。\",\n \"键盘移动-回车确认放置：使用方向键将处于移动模式的卡片调整到一个新位置后，按下 Enter 键，验证卡片的闪烁或高亮消失，退出移动模式并固定在这个最终位置。\",\n \"键盘移动-Esc取消操作：让某个卡片进入选取移动模式并通过方向键移动它，在未按Enter确认前按下 Escape 键，验证卡片立刻飞回最初所在的列和位置，并退出移动模式。\",\n \"全局撤销提示(Toast)：通过鼠标拖拽、键盘移动或点击删除某张卡片后，验证页面底部弹出一个包含'撤销'按钮的全局 Toast 提示，且静置等待3秒钟后，该 Toast 自动消失。\",\n \"点击撤销按钮恢复原状：拖拽移动一个卡片到新列触发 Toast 提示，在 Toast 提示自动消失前点击其中的'撤销'按钮，验证刚刚被移动的卡片立刻恢复到拖动之前的所属列和位置，且 Toast 关闭。\",\n \"快捷键撤销功能：分别执行删除卡片、鼠标拖拽移动卡片、键盘移动卡片操作后，立即按下键盘的 Ctrl + Z（或 Mac 下的 Cmd + Z），验证刚才执行的操作被撤销，卡片恢复到操作前的状态和位置，且底部的 Toast 提示随之消失。\",\n \"数据持久化存储：在看板中执行添加任务、删除任务、拖拽移动任务、键盘移动任务操作后，手动刷新浏览器页面，验证页面重新加载后，所有任务的内容、优先级、所属列和排列顺序均保持最新状态，没有发生数据丢失。\",\n \"LocalStorage写入检查：打开浏览器开发者工具的Application面板，验证看板的列数据、任务内容、优先级、所属列和排序状态确实被保存到了localStorage中。\"\n]","isConversational":false}}},{"rowIdx":51,"cells":{"task_id":{"kind":"number","value":52,"string":"52"},"repository_policy":{"kind":"string","value":"请你根据后续我提出的各类代码开发相关需求，完成对应代码编写、文件生成、问题解答等相关任务，且需始终严格遵循以下约束要求，所有约束在后续所有交互轮次均必须遵守，不得违反：\n1. 你创建的代码文件无一例外，全都要用UTF-8进行编码。\n2. 生成的代码文件应全部使用 4 个空格来控制缩进。\n3. 生成的任何一个代码文件中，都必须通篇使用中文作为注释语言。\n4. 请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径。\n5. 在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距。\n6. 在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符。\n7. 对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹。\n8. 请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名集合，用于推断可能的代码文件类型\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.jsx',\n '.ts', '.tsx', '.html', '.css', '.go', '.rb', '.php', '.swift',\n '.kt', '.rs', '.sh', '.bat', '.sql', '.json', '.xml', '.yaml', \n '.yml', '.vue', '.dart', '.scala', '.m', '.mm'\n }\n\n # 如果工作区路径不存在，说明尚未生成文件，不判定为违背要求\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历工作目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并统一转换为小写进行匹配\n ext = os.path.splitext(file)[1].lower()\n \n # 判断当前文件是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明该文件包含了非 UTF-8 编码的字符\n return False\n except Exception:\n # 忽略诸如文件权限等其他读取异常，避免误判\n pass\n\n # 遍历结束，未发现非 UTF-8 编码的代码文件，满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"import os\nimport re\nfrom collections import Counter\n\ndef check_indentation(response: str, workspace_path: str) -> bool:\n # 常见的需要控制缩进的代码文件后缀集合\n code_extensions = {\n '.py', '.java', '.js', '.ts', '.c', '.cpp', '.cs', '.h', '.hpp', \n '.html', '.css', '.go', '.rs', '.rb', '.php', '.swift', '.kt', \n '.sh', '.xml', '.yml', '.yaml', '.json', '.vue', '.jsx', '.tsx'\n }\n \n diff_counter = Counter()\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n has_code_file = True\n filepath = os.path.join(root, file)\n \n try:\n # 忽略编码错误读取内容，防止个别非UTF-8生成异常直接导致崩溃\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n prev_spaces = 0\n for line in lines:\n # 提取行首所有的空白符号\n m = re.match(r'^([ \\t]+)', line)\n if m:\n leading_ws = m.group(1)\n # 发现行首包含制表符，直接判定违背了全部使用空格的规定\n if '\\t' in leading_ws:\n return False\n \n # 去除左侧所有空白符用于判断是否空行或者注释\n stripped = line.lstrip(' \\t\\n\\r')\n \n # 忽略完全为空的行，以及常见多行注释的持续行（如 \" * comment\"）\n if not stripped or stripped.startswith('*'):\n continue\n \n # 计算当前行前导空格数量\n curr_spaces = len(line) - len(line.lstrip(' \\t\\n\\r'))\n \n # 若相比于上一非空行产生了新的递增缩进，计入统计\n if curr_spaces > prev_spaces:\n diff_counter[curr_spaces - prev_spaces] += 1\n \n # 更新以备下一行对比\n prev_spaces = curr_spaces\n \n # 若项目中没有生成常规的代码文件，或者没有任何缩进行发生，视为没有违规\n if not has_code_file or not diff_counter:\n return True\n \n count_4 = diff_counter.get(4, 0)\n count_8 = diff_counter.get(8, 0)\n \n # 寻找非 4 或 8（如 2 或 3）的其他差值中，出现频率最高的那一个\n max_other = 0\n for k, v in diff_counter.items():\n if k not in (4, 8) and v > max_other:\n max_other = v\n \n # 如果基于 4 空格的递增次数（含跨多级的 8 空格）占据主导地位\n # 则认定确实采用了4个空格来控制主缩进\n if (count_4 + count_8) >= max_other:\n return True\n \n return False","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def check_equals_spacing(content: str) -> bool:\n # 遍历文本中每一个字符\n for i in range(len(content)):\n if content[i] == '=':\n is_single = True\n # 检查等号左侧，排除形如 !=, <=, >=, +=, -=, *=, /=, %=, &=, |=, ^=, ~=, :=, == 的情况\n if i > 0 and content[i-1] in '=><!+-*/%&|^~:':\n is_single = False\n # 检查等号右侧，排除形如 ==, => (部分语言中的箭头函数) 的情况\n if i < len(content) - 1 and content[i+1] in '=>':\n is_single = False\n \n # 如果是单个等号，则检查其左右两侧的空格\n if is_single:\n left_ok = False\n # 左侧需有恰好两个空格：i-1和i-2是空格，且i-3不是空格(或越界)\n if i >= 2 and content[i-1] == ' ' and content[i-2] == ' ':\n if i < 3 or content[i-3] != ' ':\n left_ok = True\n \n right_ok = False\n # 右侧需有恰好两个空格：i+1和i+2是空格，且i+3不是空格(或越界)\n if i + 2 < len(content) and content[i+1] == ' ' and content[i+2] == ' ':\n if i + 3 >= len(content) or content[i+3] != ' ':\n right_ok = True\n \n # 只要任何一边不符合恰好有两个空格的要求，直接判定不满足\n if not (left_ok and right_ok):\n return False\n return True\n\n # 常见代码文件后缀，用于推测生成项目中的代码文件类型\n code_extensions = {\n '.py', '.java', '.cpp', '.c', '.h', '.hpp', '.cs',\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue',\n '.go', '.php', '.rb', '.swift', '.kt', '.rs',\n '.sh', '.bat', '.sql', '.json', '.xml', '.yml', '.yaml'\n }\n \n # 验证生成的项目代码文件中的 \"=\" 空格\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_equals_spacing(content):\n return False\n except Exception:\n # 如无法以 UTF-8 读取（不是文本文件等异常），则忽略此文件\n pass\n \n # 验证响应文本中的 Markdown 代码块里的 \"=\" 空格\n # 提取形如 ```python ... ``` 等代码块的内容\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n if not check_equals_spacing(block):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"def check_req(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：检查给定文本中是否存在使用制表符缩进的行\n def has_tab_in_indentation(text: str) -> bool:\n lines = text.split('\\n')\n for line in lines:\n # 利用 lstrip 剥离左侧空白字符，从而提取行首的纯空白缩进部分\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n if '\\t' in leading_whitespace:\n return True\n return False\n\n # 1. 检查 response 中的 markdown 代码块\n code_blocks = re.findall(r'```[^\\n]*\\n([\\s\\S]*?)```', response)\n for block in code_blocks:\n if has_tab_in_indentation(block):\n return False\n\n # 2. 检查 workspace_path 下生成的代码文件\n if not os.path.exists(workspace_path):\n return True\n\n # 枚举常见的代码/配置文件类型后缀\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.html', '.css', '.rb', '.php', '.go', \n '.rs', '.swift', '.kt', '.sh', '.json', '.xml', '.yaml', \n '.yml', '.sql', '.vue', '.jsx', '.tsx', '.m', '.scala', '.dart'\n }\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if has_tab_in_indentation(content):\n return False\n except Exception:\n # 忽略读取异常（如非utf-8编码的特殊文件或二进制文件读入错误）\n pass\n\n return True","tags":["Style","Layout"]}],"string":"[\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名集合，用于推断可能的代码文件类型\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.jsx',\\n '.ts', '.tsx', '.html', '.css', '.go', '.rb', '.php', '.swift',\\n '.kt', '.rs', '.sh', '.bat', '.sql', '.json', '.xml', '.yaml', \\n '.yml', '.vue', '.dart', '.scala', '.m', '.mm'\\n }\\n\\n # 如果工作区路径不存在，说明尚未生成文件，不判定为违背要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历工作目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并统一转换为小写进行匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断当前文件是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明该文件包含了非 UTF-8 编码的字符\\n return False\\n except Exception:\\n # 忽略诸如文件权限等其他读取异常，避免误判\\n pass\\n\\n # 遍历结束，未发现非 UTF-8 编码的代码文件，满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"import os\\nimport re\\nfrom collections import Counter\\n\\ndef check_indentation(response: str, workspace_path: str) -> bool:\\n # 常见的需要控制缩进的代码文件后缀集合\\n code_extensions = {\\n '.py', '.java', '.js', '.ts', '.c', '.cpp', '.cs', '.h', '.hpp', \\n '.html', '.css', '.go', '.rs', '.rb', '.php', '.swift', '.kt', \\n '.sh', '.xml', '.yml', '.yaml', '.json', '.vue', '.jsx', '.tsx'\\n }\\n \\n diff_counter = Counter()\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n has_code_file = True\\n filepath = os.path.join(root, file)\\n \\n try:\\n # 忽略编码错误读取内容，防止个别非UTF-8生成异常直接导致崩溃\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n prev_spaces = 0\\n for line in lines:\\n # 提取行首所有的空白符号\\n m = re.match(r'^([ \\\\t]+)', line)\\n if m:\\n leading_ws = m.group(1)\\n # 发现行首包含制表符，直接判定违背了全部使用空格的规定\\n if '\\\\t' in leading_ws:\\n return False\\n \\n # 去除左侧所有空白符用于判断是否空行或者注释\\n stripped = line.lstrip(' \\\\t\\\\n\\\\r')\\n \\n # 忽略完全为空的行，以及常见多行注释的持续行（如 \\\" * comment\\\"）\\n if not stripped or stripped.startswith('*'):\\n continue\\n \\n # 计算当前行前导空格数量\\n curr_spaces = len(line) - len(line.lstrip(' \\\\t\\\\n\\\\r'))\\n \\n # 若相比于上一非空行产生了新的递增缩进，计入统计\\n if curr_spaces > prev_spaces:\\n diff_counter[curr_spaces - prev_spaces] += 1\\n \\n # 更新以备下一行对比\\n prev_spaces = curr_spaces\\n \\n # 若项目中没有生成常规的代码文件，或者没有任何缩进行发生，视为没有违规\\n if not has_code_file or not diff_counter:\\n return True\\n \\n count_4 = diff_counter.get(4, 0)\\n count_8 = diff_counter.get(8, 0)\\n \\n # 寻找非 4 或 8（如 2 或 3）的其他差值中，出现频率最高的那一个\\n max_other = 0\\n for k, v in diff_counter.items():\\n if k not in (4, 8) and v > max_other:\\n max_other = v\\n \\n # 如果基于 4 空格的递增次数（含跨多级的 8 空格）占据主导地位\\n # 则认定确实采用了4个空格来控制主缩进\\n if (count_4 + count_8) >= max_other:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def check_equals_spacing(content: str) -> bool:\\n # 遍历文本中每一个字符\\n for i in range(len(content)):\\n if content[i] == '=':\\n is_single = True\\n # 检查等号左侧，排除形如 !=, <=, >=, +=, -=, *=, /=, %=, &=, |=, ^=, ~=, :=, == 的情况\\n if i > 0 and content[i-1] in '=><!+-*/%&|^~:':\\n is_single = False\\n # 检查等号右侧，排除形如 ==, => (部分语言中的箭头函数) 的情况\\n if i < len(content) - 1 and content[i+1] in '=>':\\n is_single = False\\n \\n # 如果是单个等号，则检查其左右两侧的空格\\n if is_single:\\n left_ok = False\\n # 左侧需有恰好两个空格：i-1和i-2是空格，且i-3不是空格(或越界)\\n if i >= 2 and content[i-1] == ' ' and content[i-2] == ' ':\\n if i < 3 or content[i-3] != ' ':\\n left_ok = True\\n \\n right_ok = False\\n # 右侧需有恰好两个空格：i+1和i+2是空格，且i+3不是空格(或越界)\\n if i + 2 < len(content) and content[i+1] == ' ' and content[i+2] == ' ':\\n if i + 3 >= len(content) or content[i+3] != ' ':\\n right_ok = True\\n \\n # 只要任何一边不符合恰好有两个空格的要求，直接判定不满足\\n if not (left_ok and right_ok):\\n return False\\n return True\\n\\n # 常见代码文件后缀，用于推测生成项目中的代码文件类型\\n code_extensions = {\\n '.py', '.java', '.cpp', '.c', '.h', '.hpp', '.cs',\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue',\\n '.go', '.php', '.rb', '.swift', '.kt', '.rs',\\n '.sh', '.bat', '.sql', '.json', '.xml', '.yml', '.yaml'\\n }\\n \\n # 验证生成的项目代码文件中的 \\\"=\\\" 空格\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_equals_spacing(content):\\n return False\\n except Exception:\\n # 如无法以 UTF-8 读取（不是文本文件等异常），则忽略此文件\\n pass\\n \\n # 验证响应文本中的 Markdown 代码块里的 \\\"=\\\" 空格\\n # 提取形如 ```python ... ``` 等代码块的内容\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n if not check_equals_spacing(block):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"def check_req(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：检查给定文本中是否存在使用制表符缩进的行\\n def has_tab_in_indentation(text: str) -> bool:\\n lines = text.split('\\\\n')\\n for line in lines:\\n # 利用 lstrip 剥离左侧空白字符，从而提取行首的纯空白缩进部分\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n if '\\\\t' in leading_whitespace:\\n return True\\n return False\\n\\n # 1. 检查 response 中的 markdown 代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n([\\\\s\\\\S]*?)```', response)\\n for block in code_blocks:\\n if has_tab_in_indentation(block):\\n return False\\n\\n # 2. 检查 workspace_path 下生成的代码文件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 枚举常见的代码/配置文件类型后缀\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.html', '.css', '.rb', '.php', '.go', \\n '.rs', '.swift', '.kt', '.sh', '.json', '.xml', '.yaml', \\n '.yml', '.sql', '.vue', '.jsx', '.tsx', '.m', '.scala', '.dart'\\n }\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if has_tab_in_indentation(content):\\n return False\\n except Exception:\\n # 忽略读取异常（如非utf-8编码的特殊文件或二进制文件读入错误）\\n pass\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Machine Learning"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"我们要开发一个本地的电影数据库Web应用。首先，请帮我搭建出这个项目的基础骨架。页面主要分为顶部导航栏和下方的主内容区。导航栏左侧是应用的Logo标题，右侧是‘首页’和‘我的片单’两个导航Tab。内容区先居中放置一个‘欢迎来到本地电影库’的占位提示语和一个模拟用的空按钮。帮我完成这个基础结构的布局和样式，要求界面简洁现代，且具备基本的响应式适配，先不需要实现具体业务逻辑。另外请你在后续所有操作中始终遵循以下要求：第一，必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作；第二，为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用。","instruction_following_checklist":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义Web开发和常见编程可能涉及到的代码文件后缀\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.vue', '.json', '.py', '.java', '.c', '.cpp', '.h', '.md', '.txt'\n }\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明没有使用 UTF-8 进行编码\n return False\n except Exception:\n # 其他读取异常（如权限问题等）暂不作为违反编码规范处理\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证生成的代码文件是否全部使用 4 个空格来控制缩进。\n \"\"\"\n # 推断可能的代码文件类型（根据需求为 Web 前后端应用）\n code_extensions = ['.html', '.css', '.js', '.py', '.json']\n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤非目标类型的代码文件\n if not any(file.endswith(ext) for ext in code_extensions):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n files_checked += 1\n indent_counts = {}\n has_tab = False\n \n for line in lines:\n # 忽略完全为空的行\n if not line.strip():\n continue\n \n # 获取该行的前导空白字符\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 如果包含制表符，则判定违规\n if '\\t' in leading_whitespace:\n return False\n \n length = len(leading_whitespace)\n if length > 0:\n stripped = line.lstrip()\n # 忽略以常见的注释或文档标签开头的行，以防止注释对齐干扰判断\n if stripped.startswith(('*', '#', '//', '<!--', '/*', '*/')):\n continue\n \n indent_counts[length] = indent_counts.get(length, 0) + 1\n \n total_indents = sum(indent_counts.values())\n if total_indents > 0:\n # 统计前导空格数不是 4 的倍数的行数\n non_4_cnt = sum(count for length, count in indent_counts.items() if length % 4 != 0)\n \n # 如果所有的缩进都不是 4 的倍数，显然不符合 4 空格缩进\n if non_4_cnt == total_indents:\n return False\n \n # 如果非 4 倍数的缩进占比超过 25%，并且数量超过 5 行，认为违反了 4 空格缩进的规则\n if non_4_cnt > 5 and (non_4_cnt / total_indents) > 0.25:\n return False\n \n # 如果找到了至少一个代码文件且都通过了验证，则符合要求\n return files_checked > 0","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断文件是否为合法的UTF-8文本（代码）文件\n def is_text_file(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read(1024)\n if '\\0' in content:\n return False\n return True\n except UnicodeDecodeError:\n return False\n\n # 正则匹配独立的 \"=\"\n # 负向后视断言 (?<!...) 排除前缀含有 = ! < > + - * / % & | ^ : ~ ? 的情况（如 ==, !=, >=, +=, ??= 等）\n # 负向前视断言 (?![=>]) 排除后缀含有 = 或 > 的情况（如 ==, ===, => 等）\n pattern = re.compile(r'(?<![=!<>+\\-*/%&|^:~?])=(?![=>])')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # 过滤非文本文件\n if not is_text_file(file_path):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n matches = pattern.finditer(content)\n for match in matches:\n idx = match.start()\n \n # 1. 检查左侧是否恰好保留两个空格\n if idx < 2:\n return False # 左侧不足两个字符\n \n # 紧邻的左侧两个字符必须是空格\n if content[idx-2:idx] != ' ':\n return False\n \n # 左侧第三个字符不能还是空格（保证只有两个空格）\n if idx >= 3 and content[idx-3] == ' ':\n return False\n \n # 2. 检查右侧是否恰好保留两个空格\n if idx + 2 >= len(content):\n return False # 右侧不足两个字符\n \n # 紧邻的右侧两个字符必须是空格\n if content[idx+1:idx+3] != ' ':\n return False\n \n # 右侧第三个字符不能还是空格\n if idx + 3 < len(content) and content[idx+3] == ' ':\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"def check_indentation_rule(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的Web项目代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py', '.json'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n _, ext = os.path.splitext(file)\n # 仅验证代码类型的相关文件\n if ext.lower() in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 仅检查行首缩进部分的字符\n for char in line:\n if char == '\\t':\n # 行首缩进使用了制表符，违反要求\n return False\n elif char == ' ':\n # 空格缩进，继续检查该行下一个字符\n continue\n else:\n # 遇到非空白字符，结束当前行的缩进判断\n break\n except Exception:\n # 忽略读取异常的文件\n continue\n \n return True","tags":["Style","Layout"]},{"constraint":"必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":1,"instruction":"框架看起来不错。现在我们给‘首页’加上电影列表的展示功能。请在代码中构造一份包含大约15-20部电影的本地Mock数据数组（字段必须包含：ID、标题、海报图的占位色块、类型、上映年份、评分、导演、主演和剧情简介）。在首页主内容区，使用网格布局（CSS Grid）将这些电影以卡片的形式渲染出来。每张卡片展示封面、标题、年份和评分。鼠标悬浮在卡片上时，请加入轻微的上浮和阴影加深动效。另外请你在后续所有操作中始终遵循以下新增要求：第一，每当准备更改已有文件时，都必须先用相关工具阅读该文件；第二，在对已有文件作出修改时，必须以全量代码的形式输出并覆盖原文件，绝对不可以使用 Unified Diff 或者其他的补丁格式。","instruction_following_checklist":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 涵盖常见的Web应用代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.json', '.vue'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取整个文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生 Unicode 解码错误，说明文件包含了非法的 UTF-8 字符（可能使用其它编码）\n return False\n except Exception:\n # 忽略权限不足或其他无关的文件操作异常\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的前端与后端代码文件扩展名\n code_extensions = ('.html', '.js', '.css', '.py', '.ts', '.tsx', '.jsx', '.vue', '.json')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅针对代码文件进行检查\n if not file.endswith(code_extensions):\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 若文件无法读取，跳过\n continue\n\n valid_indents = 0\n invalid_indents = 0\n\n for line in lines:\n # 获取该行的前导空白字符\n leading_ws = line[:len(line) - len(line.lstrip())]\n \n # 检查是否混用了制表符\n if '\\t' in leading_ws:\n return False\n \n stripped = line.lstrip()\n # 忽略空行\n if not stripped:\n continue\n \n # 忽略常见的注释行，防止因为注释内容的排版引发误判\n if stripped.startswith(('*', '<!--', '//', '/*')):\n continue\n \n spaces = len(leading_ws)\n if spaces > 0:\n # 检查缩进是否为 4 的倍数\n if spaces % 4 == 0:\n valid_indents += 1\n else:\n invalid_indents += 1\n \n # 统计判定：如果存在未按 4 个空格控制的缩进行\n if invalid_indents > 0:\n # 如果完全没有合法的 4 空格缩进，或者非法缩进比例过高（大于等于 30%），则判定未满足要求\n if valid_indents == 0 or (invalid_indents / valid_indents) >= 0.3:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义目标代码文件的扩展名\n code_extensions = ('.html', '.htm', '.js', '.css', '.vue', '.jsx', '.ts', '.tsx', '.json', '.py')\n code_files = []\n \n # 遍历工作区目录找出所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(code_extensions):\n code_files.append(os.path.join(root, file))\n \n if not code_files:\n return False\n \n has_equals = False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n lines = content.split('\\n')\n \n for line in lines:\n stripped_line = line.strip()\n # 跳过空行或纯注释行\n if not stripped_line:\n continue\n if stripped_line.startswith(('//', '<!--', '*', '/*', '#')):\n continue\n \n # 简单去除行尾的注释内容，避免其内部的内容影响判定\n if file_path.endswith(('.js', '.ts', '.vue', '.css', '.jsx', '.tsx')):\n if '//' in line:\n line = line.split('//')[0]\n if '/*' in line:\n line = line.split('/*')[0]\n if file_path.endswith(('.html', '.htm', '.vue')):\n if '<!--' in line:\n line = line.split('<!--')[0]\n if file_path.endswith('.py'):\n if '#' in line:\n line = line.split('#')[0]\n \n # 匹配所有独立的单等号（排除 ==, ===, >=, <=, !=, =>, += 等复合符号）\n pattern = r'(?<![=!><+\\-*/%&|^~$])=(?![=>])'\n matches = list(re.finditer(pattern, line))\n \n for m in matches:\n has_equals = True\n start = m.start()\n end = m.end()\n \n # 左侧检查：确保刚好只有两个连续空格间距\n left_ok = False\n if start >= 2 and line[start-2:start] == ' ':\n # 再往前一个字符如果不是空格，说明“恰好”有两个空格\n if start == 2 or line[start-3] != ' ':\n left_ok = True\n \n # 右侧检查：确保刚好只有两个连续空格间距\n right_ok = False\n if len(line) - end >= 2 and line[end:end+2] == ' ':\n # 再往后一个字符如果不是空格，说明“恰好”有两个空格\n if len(line) - end == 2 or line[end+2] != ' ':\n right_ok = True\n \n # 只要有一个等号左右不严格遵循“分别保留两个空格间距”，则直接判断为未满足\n if not (left_ok and right_ok):\n return False\n except Exception:\n continue\n \n # 若整个项目的所有独立等号全部合规，并且确实包含了等号（排除空无内容的极端伪造场景），则返回True\n return has_equals","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能生成的代码文件后缀类型\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.py'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 逐行读取检查，防止大文件占用过多内存\n for line in f:\n # 严格限制：只要包含制表符即判定为不满足要求\n if '\\t' in line:\n return False\n except Exception:\n # 对于非UTF-8格式或无法读取的文件直接跳过（该情况由其他要求规范约束）\n continue\n \n # 如果项目中没有任何代码文件，也视作未满足要求\n return has_code_file","tags":["Style","Layout"]},{"constraint":"必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在对已有文件作出修改时，必须以全量代码的形式输出并覆盖原文件，绝对不可以使用 Unified Diff 或者其他的补丁格式","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":2,"instruction":"好，接下来完善搜索和筛选功能，这对模型是个小挑战。请在首页电影列表上方增加一个过滤区：包含一个文本输入框（支持按标题、导演或主演模糊搜索），以及一组电影类型（Genre）的多选按钮（如科幻、动作、悬疑等）。要求：文本输入必须有 500ms 的防抖（Debounce）处理；文本搜索与类型多选是互相叠加的‘且’逻辑；每当过滤条件变化时，下方的电影卡片列表必须动态更新，并且在没有匹配结果时显示‘未找到相关电影’的空状态提示。另外请你在后续所有操作中始终遵循两项更新后的要求：第一，修改已有文件时，强制要求输出严格的Unified Diff格式，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效；第二，进行任何文件处理之前，切勿使用`pwd`确定当前的目录，原有的操作文件前需先运行pwd确认目录的要求不再生效。","instruction_following_checklist":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能在 Web 开发或常见项目中用到的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.json', '.xml', '.yml', '.yaml', '.md', '.py', '.sh', '.bash',\n '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.php', '.rb', '.go', '.rs'\n }\n \n # 假如工作空间不存在，说明还未生成，不计为违反约束\n if not os.path.exists(workspace_path):\n return True\n \n # 遍历工作空间内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅校验被判定为代码类型的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 格式读取文件全部内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦发生 UnicodeDecodeError，证明其并非严格或兼容的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如权限不足等其他非编码相关的系统异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 针对前端项目和常见的代码格式，定义需要检查的文件扩展名\n code_exts = {'.html', '.css', '.js', '.py', '.java', '.cpp', '.c', '.ts', '.jsx', '.tsx', '.json', '.vue'}\n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n files_checked += 1\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n indent_counts = []\n for line in lines:\n # 忽略空行或完全没有内容的行\n if not line.strip():\n continue\n \n # 匹配行首的前导空白\n match = re.match(r'^([ \\t]+)', line)\n if match:\n spaces = match.group(1)\n # 只要有制表符即不符合“使用4个空格控制缩进”的规则\n if '\\t' in spaces:\n return False\n \n # 过滤掉一些常见的多行注释内部引发的非标准缩进\n stripped_line = line.strip()\n if stripped_line.startswith('*') or stripped_line.startswith('<!--') or stripped_line.startswith('/*'):\n continue\n \n # 记录有效缩进的空格长度\n indent_counts.append(len(spaces))\n \n if indent_counts:\n # 统计非 4 的倍数的缩进数量\n non_4_multiple = sum(1 for c in indent_counts if c % 4 != 0)\n \n # 容错判断：如果非 4 倍数缩进的行占总缩进行的 15% 以上\n # 说明大概率采用了 2 空格或 3 空格缩进，或者缩进极度不规范\n if non_4_multiple / len(indent_counts) > 0.15:\n return False\n\n # 若未找到或未检查任何代码文件，则无法证明满足了该要求\n if files_checked == 0:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"def check_equals_spacing(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见前端和代码文件的扩展名集合\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue', '.py', '.java'}\n has_files = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 正则表达式说明：\n # (?<![<>!+*/%&|^=-]) : 左边不包含组成复合运算符的字符\n # (=+) : 匹配一个或多个连续的 '='，当作等号整体\n # (?![=>]) : 右侧不能是 '=' 或 '>'（排除 '=>' 和防止将 '=' 截断）\n pattern = re.compile(r'(?<![<>!+*/%&|^=-])(=+)(?![=>])')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_files = True\n \n for line in lines:\n # 去除行末尾的换行符，以便准确判断右侧是否存在空格\n line_no_nl = line.rstrip('\\r\\n')\n matches = pattern.finditer(line_no_nl)\n \n for m in matches:\n start, end = m.span()\n left_text = line_no_nl[:start]\n right_text = line_no_nl[end:]\n \n # 检查左边：必须以恰好两个空格结尾\n if not left_text.endswith(' '):\n return False\n # 并且排查是否有三个及以上连续空格\n if len(left_text) >= 3 and left_text[-3] == ' ':\n return False\n \n # 检查右边：必须以恰好两个空格开头\n if not right_text.startswith(' '):\n return False\n # 并且排查是否有三个及以上连续空格\n if len(right_text) >= 3 and right_text[2] == ' ':\n return False\n\n # 如果工作区内检查了至少一个有效代码文件，且没有发现违规等号，则认为满足要求\n return has_files","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.json', '.vue', '.ts', \n '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h'\n }\n \n # 1. 检查工作区内生成的代码文件\n if workspace_path and os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 匹配行首的空白字符（包括空格和制表符）\n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent = match.group(1)\n # 如果行首缩进部分包含了制表符，则违反了要求\n if '\\t' in indent:\n return False\n except Exception:\n # 忽略可能出现编码错误或无法读取的文件\n pass\n \n # 2. 检查助手回复中的 Markdown 代码块\n # 使用正则表达式提取 ```代码语言...\\n ... ``` 中的代码内容\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n for line in block.split('\\n'):\n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent = match.group(1)\n # 同样检查代码块的行首缩进是否包含制表符\n if '\\t' in indent:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"修改已有文件时，强制要求输出严格的Unified Diff格式，严禁直接提供全量代码替换","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":3,"instruction":"现在的搜索筛选很好用。接着我们来做电影的详情和交互状态。当用户点击某张电影卡片时，弹出一个居中的模态框（Modal）展示该电影的全部详细信息。在详情模态框底部，增加一个‘加入我的片单’按钮。点击后，按钮文字变成‘已加入’，并且将该电影的 ID 存入浏览器的 localStorage 中进行持久化。同时，关闭模态框回到首页时，所有已经存在于 localStorage 里的电影卡片，其右上角需要常驻显示一个黄色的‘★’图标作为已收藏标记。另外请你在后续所有操作中始终遵循两项更新后的要求：第一，运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作，原有的操作文件前无需运行`pwd`确认目录的要求不再生效；第二，修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式，原有的修改文件需输出Unified Diff格式的要求不再生效。","instruction_following_checklist":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.vue', '.py', '.json', '.md', '.txt', '.xml', '.sh'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码格式验证\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用UTF-8进行解码，纯ASCII文件也能完美被UTF-8解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出UnicodeDecodeError，说明存在非UTF-8编码的文件\n return False\n except Exception:\n # 如果因为权限或其他问题无法读取，为严谨起见视作未通过\n return False\n\n # 要求必须至少存在一个创建出来的代码文件，且全部验证通过\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n # 定义需要检查的常见代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs', \n '.html', '.css', '.py', '.vue', '.json', \n '.java', '.c', '.cpp', '.h', '.hpp'\n }\n has_checked_any_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_checked_any_file = True\n \n non_zero_indents = []\n for line in lines:\n # 忽略空行\n if not line.strip():\n continue\n \n stripped = line.strip()\n # 过滤多行注释及常见可能影响缩进对齐的非代码内容行\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('//') or stripped.startswith('<!--'):\n continue\n \n # 提取行首的空白字符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent_str = match.group(1)\n # 如果存在制表符，违反了不使用制表符且要求4空格缩进的规则，直接判定失败\n if '\\t' in indent_str:\n return False\n \n spaces = len(indent_str)\n if spaces > 0:\n non_zero_indents.append(spaces)\n \n # 统计不符合 4 个空格基本倍数的缩进行数\n if non_zero_indents:\n non_4_mult_count = sum(1 for s in non_zero_indents if s % 4 != 0)\n # 容错机制：允许少量合理的非 4 倍数对齐（如多行字符串或特定长参折行对齐等）\n # 但如果是 2 空格或 3 空格缩进文件，此比例必定远超 30%，且绝对数量会较多\n if non_4_mult_count > max(3, len(non_zero_indents) * 0.3):\n return False\n\n return has_checked_any_file","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能的代码文件扩展名\n valid_exts = {'.js', '.html', '.htm', '.css', '.vue', '.ts', '.jsx', '.tsx', '.py'}\n \n has_code_files = False\n \n # 匹配独立的单等号\n # 排除复合运算符及比较运算符，如 !=, ==, ===, >=, <=, +=, -= 等，同时也排除 => 箭头函数\n pattern = r'(?<![=!<>\\+\\-\\*/%&\\|\\^~])=(?![=>])'\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n # 如果无法按utf-8读取，跳过该文件\n continue\n \n has_code_files = True\n \n for m in re.finditer(pattern, text):\n i = m.start()\n \n # 统计等号左侧连续的空格数\n left_spaces = 0\n j = i - 1\n while j >= 0 and text[j] == ' ':\n left_spaces += 1\n j -= 1\n \n # 统计等号右侧连续的空格数\n right_spaces = 0\n j = i + 1\n while j < len(text) and text[j] == ' ':\n right_spaces += 1\n j += 1\n \n # 如果任一侧的空格间距不是恰好为2个，则认为不符合要求\n if left_spaces != 2 or right_spaces != 2:\n return False\n \n # 如果没有找到任何相关的代码文件，无法验证，返回False\n if not has_code_files:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 如果工作区路径不存在，则无法验证代码，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.json', '.py', '.java', '.c', '.cpp', '.cs', '.go', '.php'\n }\n \n # 遍历工作区目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取代码文件\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n # lstrip(' \\t') 会去除左侧连续的空格和制表符\n # 借此计算行首空白部分的长度，并提取出真正的“缩进字符”\n leading_spaces = line[:len(line) - len(line.lstrip(' \\t'))]\n \n # 判断缩进中是否使用了制表符\n if '\\t' in leading_spaces:\n return False\n except Exception:\n # 如果个别文件读取出错则跳过，继续检查其他文件\n continue\n \n # 如果所有代码文件检查完毕且均无制表符缩进，则符合要求\n return True","tags":["Style","Layout"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":4,"instruction":"下一步是完善‘我的片单’页面的复杂交互。当用户点击顶部导航的‘我的片单’时，清空首页视图并展示已收藏的电影列表。在这个专属视图中，用户可以对每部电影进行两项高阶操作：一是通过下拉菜单给电影打上私人标签（‘想看’、‘正在看’或‘已看’），二是提供一个可交互的 5 星评分组件给电影打私人分数。标签和私人评分的状态变更必须立刻更新并持久化到 localStorage。此外，每张卡片需提供一个‘移出片单’按钮，点击后伴随确认提示将电影从该页面及本地存储中删除。另外请你在后续所有操作中始终遵循两项更新后的要求：第一，修改任何现有文件时，必须严格采用Unified Diff格式输出，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效；第二，在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。","instruction_following_checklist":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端Web项目常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.ts', '.tsx', '.jsx', '.md'}\n \n # 如果路径不存在，无法验证文件，直接返回 False 或是按照默认 True 处理。\n # 通常如果确实生成了文件，路径必定存在。这里加上基本的路径验证。\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件包含无法被 UTF-8 解码的字符（例如使用了 GBK 编码）\n return False\n except Exception:\n # 其他异常不属于编码判定范围，忽略处理\n pass\n \n # 如果所有代码文件都能成功以 UTF-8 读取（或者全为纯 ASCII），则视为满足条件\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n from collections import Counter\n \n # 限制只检查常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json', '.vue', '.java', '.c', '.cpp'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n diffs = Counter()\n prev_spaces = 0\n \n for line in lines:\n # 忽略空行和只包含空白符的行\n if not line.strip():\n continue\n \n # 匹配行首前导空白字符\n match = re.match(r'^([ \\t]+)\\S', line)\n if match:\n leading_ws = match.group(1)\n \n # 严禁使用制表符进行缩进\n if '\\t' in leading_ws:\n return False\n \n spaces = len(leading_ws)\n else:\n spaces = 0\n \n # 如果当前行缩进大于前一行，说明产生了新缩进层级，将其差值加入统计\n if spaces > prev_spaces:\n diffs[spaces - prev_spaces] += 1\n \n prev_spaces = spaces\n \n # 若文件包含缩进递进，最常见的主流缩进层级大小必须为4\n if diffs:\n most_common_diff = diffs.most_common(1)[0][0]\n if most_common_diff != 4:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 设定需要检查的常见代码和纯文本文件扩展名\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.json', '.md', '.txt'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n lines = content.split('\\n')\n for line in lines:\n i = 0\n while i < len(line):\n if line[i] == '=':\n # 判断该 \"=\" 是否属于复合运算符的一部分 (如 ==, !=, >=, +=, =>, ~= 等)\n is_compound = False\n if i > 0 and line[i-1] in ':=!<>+-*/%&|^?~$':\n is_compound = True\n if i < len(line) - 1 and line[i+1] in '=>':\n is_compound = True\n \n # 如果是孤立的 \"=\" (赋值或属性等号)，则严格检查两侧空格数量\n if not is_compound:\n # 统计左侧连续空格数\n left_spaces = 0\n j = i - 1\n while j >= 0 and line[j] == ' ':\n left_spaces += 1\n j -= 1\n \n # 统计右侧连续空格数\n right_spaces = 0\n k = i + 1\n while k < len(line) and line[k] == ' ':\n right_spaces += 1\n k += 1\n \n # 必须精确等于两个空格的间距，若非如此则违背该要求\n if left_spaces != 2 or right_spaces != 2:\n return False\n i += 1\n \n return True","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 辅助函数：逐行检查缩进是否包含制表符\n def check_indentation(lines) -> bool:\n for line in lines:\n # 匹配可能包含 Diff '+' 前缀的行，并捕获前面的所有空白字符\n # 避免匹配到 '-' 开始的删除行，只针对新生成/保留的代码进行规范约束\n match = re.match(r'^[+]?(\\s+)', line)\n if match:\n leading_spaces = match.group(1)\n # 如果前导空白中使用了制表符，判定违规\n if '\\t' in leading_spaces:\n return False\n return True\n\n # 1. 检查当前工作目录下的相关代码文件\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n # 根据用户指令中提到的“Web应用”，限定常见的代码与相关后缀\n valid_exts = {'.js', '.css', '.html', '.htm', '.jsx', '.tsx', '.ts', '.json', '.vue'}\n for root, dirs, files in os.walk(workspace_path):\n # 排除非相关且可能引起误判的隐藏目录及依赖目录\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file_name in files:\n if os.path.splitext(file_name)[1].lower() in valid_exts:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n if not check_indentation(lines):\n return False\n except Exception:\n continue\n\n # 2. 检查大模型回复内容中的 Markdown 代码块\n code_blocks = re.findall(r'```.*?\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n lines = block.split('\\n')\n if not check_indentation(lines):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"修改任何现有文件时，必须严格采用Unified Diff格式输出，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 针对前端和常见后端项目涉及的代码文件后缀进行校验\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.py', '.vue', '.php', '.java', '.go', '.rb', '.c', '.cpp', '.h', '.hpp'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于非文本或其他原因导致无法用 UTF-8 读取，则跳过\n continue\n \n # 提取非空行，忽略连续空行带来的干扰\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n if not non_empty_lines:\n # 若为一个没有任何实质内容的空代码文件，显然没有包含需要的注释，不符合要求\n return False\n \n # 检查该文件前 10 个非空行中是否包含目标字符串 (容忍 DOCTYPE 声明等特殊头部)\n found_in_head = False\n for line in non_empty_lines[:10]:\n if target_text in line:\n found_in_head = True\n break\n \n if not found_in_head:\n # 如果发现任何一个代码文件在开头未添加指定注释，则直接判定为不满足要求\n return False\n \n # 如果遍历了所有文件并且都符合要求，且至少存在一个代码文件，则判定为满足\n return has_code_file","tags":["Content","Format"]}]},{"round_id":5,"instruction":"最后，为了让应用更具个性化，我们需要在‘首页’的右侧（或顶部下方）加入一个‘猜你喜欢’的本地推荐算法模块。请实现一个纯前端的推荐逻辑：每次加载或切换回首页时，读取 localStorage 中‘我的片单’里的电影，统计用户收藏中最频繁出现的‘电影类型’和‘导演’的权重；然后遍历本地库中所有**尚未加入片单**的电影，按照计算出的特征匹配度（相似度得分）进行降序排序，选取前 3 部最高分的电影展示在推荐模块中。要求算法具有容错性，若片单为空，则随机推荐 3 部电影。另外请你在后续所有操作中始终遵循两项更新及新增要求：第一，在你生成的每一个代码文件中，请于结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”，原有的在代码开头添加该注释的要求不再生效；第二，限制每个生成的代码文件总行数（含空行）必须是偶数。","instruction_following_checklist":[{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前后端代码及配置文件类型扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.json', '.py', '.sh', '.md', '.txt'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅验证代码类型的文本文件，排除图片或其它二进制文件的干扰\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果发生解码错误，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 对于其它的不可预知读取异常也算作验证失败\n return False\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据“Web应用”的指令推断，设置常见的前端代码文件扩展名\n valid_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_exts):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 若存在编码读取等异常，暂且忽略跳过\n continue\n \n for line in lines:\n stripped_line = line.lstrip()\n # 跳过全空行\n if not stripped_line:\n continue\n \n # 排除多行注释的中间及结束行（例如 \" * comment\" 或 \" */\"）\n if stripped_line.startswith('*'):\n continue\n \n leading_len = len(line) - len(stripped_line)\n leading_str = line[:leading_len]\n \n # 若前导空白中包含制表符，说明未严格使用空格缩进，违规\n if '\\t' in leading_str:\n return False\n \n # 若前导空格的数量不是 4 的倍数，说明未使用 4 个空格控制缩进，违规\n if leading_len % 4 != 0:\n return False\n \n # 只要存在至少一个代码文件且全部通过了缩进检验，即算满足要求\n return has_code_file","tags":["Style","Layout"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n\n # 匹配独立的等号 \"=\"，排除常见的复合运算符如 +=, -=, ==, ===, <=, >=, !=, => 等\n pattern = re.compile(r'(?<![+\\-*/%&|^<>=!~$])=(?![=>])')\n found_any_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 忽略隐藏文件或.git等版本控制目录\n if '.git' in root or file.startswith('.'):\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n continue # 跳过二进制或非UTF-8编码文件\n\n found_any_file = True\n\n for match in pattern.finditer(content):\n start = match.start()\n end = match.end()\n\n # 检查等号前是否恰好有两个空格\n if start < 2:\n return False # 前方字符不足两个\n if content[start-2:start] != \" \":\n return False # 前方紧挨着的两个字符不是空格\n if start >= 3 and content[start-3] == \" \":\n return False # 前方第三个字符还是空格（超过两个）\n\n # 检查等号后是否恰好有两个空格\n if end + 2 > len(content):\n return False # 后方字符不足两个\n if content[end:end+2] != \" \":\n return False # 后方紧挨着的两个字符不是空格\n if end + 2 < len(content) and content[end+2] == \" \":\n return False # 后方第三个字符还是空格（超过两个）\n\n # 如果存在至少一个代码文件并且所有的 '=' 都满足要求，则返回 True\n return found_any_file","tags":["Style","Layout"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符","validation_code":"def check_indentation_rule(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n has_code = False\n\n # 1. 检查工作区实际生成的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n has_code = True\n except Exception:\n # 忽略非文本文件或无法读取的文件\n continue\n \n for line in lines:\n # 匹配行首所有的空格和制表符作为缩进部分\n match = re.match(r'^[ \\t]+', line)\n if match:\n # 如果缩进内容中包含制表符，则判定为不满足要求\n if '\\t' in match.group(0):\n return False\n\n # 2. 检查最后一轮回复中包含的代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n if code_blocks:\n has_code = True\n for block in code_blocks:\n for line in block.split('\\n'):\n line_to_check = line\n # 兼容 Unified Diff 格式：忽略行首表示变更的 '+' 或 '-'\n if line.startswith('+') or line.startswith('-'):\n line_to_check = line[1:]\n \n # 匹配行首缩进\n match = re.match(r'^[ \\t]+', line_to_check)\n if match:\n if '\\t' in match.group(0):\n return False\n\n # 必须至少包含相关的代码内容，且在检查中全部合格才返回 True\n return has_code","tags":["Style","Layout"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"修改任何现有文件时，必须严格采用Unified Diff格式输出，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的每一个代码文件中，请于结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 针对本地电影库Web前端应用，筛选常见的代码文件扩展名\n valid_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n \n files_found = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n files_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read().strip()\n except Exception:\n # 无法读取（不符合第一轮强制UTF-8的要求）等异常情况，视为未满足\n return False\n \n # 如果代码文件为空，则显然没有包含要求的注释\n if not content:\n return False\n \n # 必须包含目标字符串\n if target_text not in content:\n return False\n \n # 验证是否于\"结尾处\"添加了注释，截取文件的最后10个非空行进行校验\n lines = [line.strip() for line in content.split('\\n') if line.strip()]\n found_at_end = False\n for line in lines[-10:]:\n if target_text in line:\n found_at_end = True\n break\n \n if not found_at_end:\n return False\n \n # 如果找到了相应的代码文件且全部满足要求，则返回True；若无任何代码文件，返回False\n return files_found","tags":["Content","Format"]},{"constraint":"限制每个生成的代码文件总行数（含空行）必须是偶数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名，根据web应用的语境重点关注前端相关文件\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照 '\\n' 切分可以准确地统计总行数（包含末尾空行及连续空行）\n lines = content.split('\\n')\n \n # 如果行数不是偶数，则说明违反了要求\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果由于编码等问题无法读取，为保险起见可以直接视为未通过测试，或跳过\n # 根据上下文要求“无一例外用UTF-8编码”，无法使用UTF-8读取也可视为失败\n return False\n \n # 如果没有找到任何代码文件，说明没有生成有效项目，返回False；否则全满足条件返回True\n return has_code_file","tags":["Quantity","Exact Value"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"我们要开发一个本地的电影数据库Web应用。首先，请帮我搭建出这个项目的基础骨架。页面主要分为顶部导航栏和下方的主内容区。导航栏左侧是应用的Logo标题，右侧是‘首页’和‘我的片单’两个导航Tab。内容区先居中放置一个‘欢迎来到本地电影库’的占位提示语和一个模拟用的空按钮。帮我完成这个基础结构的布局和样式，要求界面简洁现代，且具备基本的响应式适配，先不需要实现具体业务逻辑。另外请你在后续所有操作中始终遵循以下要求：第一，必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作；第二，为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义Web开发和常见编程可能涉及到的代码文件后缀\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.vue', '.json', '.py', '.java', '.c', '.cpp', '.h', '.md', '.txt'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明没有使用 UTF-8 进行编码\\n return False\\n except Exception:\\n # 其他读取异常（如权限问题等）暂不作为违反编码规范处理\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证生成的代码文件是否全部使用 4 个空格来控制缩进。\\n \\\"\\\"\\\"\\n # 推断可能的代码文件类型（根据需求为 Web 前后端应用）\\n code_extensions = ['.html', '.css', '.js', '.py', '.json']\\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤非目标类型的代码文件\\n if not any(file.endswith(ext) for ext in code_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n files_checked += 1\\n indent_counts = {}\\n has_tab = False\\n \\n for line in lines:\\n # 忽略完全为空的行\\n if not line.strip():\\n continue\\n \\n # 获取该行的前导空白字符\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 如果包含制表符，则判定违规\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n length = len(leading_whitespace)\\n if length > 0:\\n stripped = line.lstrip()\\n # 忽略以常见的注释或文档标签开头的行，以防止注释对齐干扰判断\\n if stripped.startswith(('*', '#', '//', '<!--', '/*', '*/')):\\n continue\\n \\n indent_counts[length] = indent_counts.get(length, 0) + 1\\n \\n total_indents = sum(indent_counts.values())\\n if total_indents > 0:\\n # 统计前导空格数不是 4 的倍数的行数\\n non_4_cnt = sum(count for length, count in indent_counts.items() if length % 4 != 0)\\n \\n # 如果所有的缩进都不是 4 的倍数，显然不符合 4 空格缩进\\n if non_4_cnt == total_indents:\\n return False\\n \\n # 如果非 4 倍数的缩进占比超过 25%，并且数量超过 5 行，认为违反了 4 空格缩进的规则\\n if non_4_cnt > 5 and (non_4_cnt / total_indents) > 0.25:\\n return False\\n \\n # 如果找到了至少一个代码文件且都通过了验证，则符合要求\\n return files_checked > 0\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断文件是否为合法的UTF-8文本（代码）文件\\n def is_text_file(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read(1024)\\n if '\\\\0' in content:\\n return False\\n return True\\n except UnicodeDecodeError:\\n return False\\n\\n # 正则匹配独立的 \\\"=\\\"\\n # 负向后视断言 (?<!...) 排除前缀含有 = ! < > + - * / % & | ^ : ~ ? 的情况（如 ==, !=, >=, +=, ??= 等）\\n # 负向前视断言 (?![=>]) 排除后缀含有 = 或 > 的情况（如 ==, ===, => 等）\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%&|^:~?])=(?![=>])')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # 过滤非文本文件\\n if not is_text_file(file_path):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n matches = pattern.finditer(content)\\n for match in matches:\\n idx = match.start()\\n \\n # 1. 检查左侧是否恰好保留两个空格\\n if idx < 2:\\n return False # 左侧不足两个字符\\n \\n # 紧邻的左侧两个字符必须是空格\\n if content[idx-2:idx] != ' ':\\n return False\\n \\n # 左侧第三个字符不能还是空格（保证只有两个空格）\\n if idx >= 3 and content[idx-3] == ' ':\\n return False\\n \\n # 2. 检查右侧是否恰好保留两个空格\\n if idx + 2 >= len(content):\\n return False # 右侧不足两个字符\\n \\n # 紧邻的右侧两个字符必须是空格\\n if content[idx+1:idx+3] != ' ':\\n return False\\n \\n # 右侧第三个字符不能还是空格\\n if idx + 3 < len(content) and content[idx+3] == ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"def check_indentation_rule(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的Web项目代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py', '.json'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n _, ext = os.path.splitext(file)\\n # 仅验证代码类型的相关文件\\n if ext.lower() in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 仅检查行首缩进部分的字符\\n for char in line:\\n if char == '\\\\t':\\n # 行首缩进使用了制表符，违反要求\\n return False\\n elif char == ' ':\\n # 空格缩进，继续检查该行下一个字符\\n continue\\n else:\\n # 遇到非空白字符，结束当前行的缩进判断\\n break\\n except Exception:\\n # 忽略读取异常的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"框架看起来不错。现在我们给‘首页’加上电影列表的展示功能。请在代码中构造一份包含大约15-20部电影的本地Mock数据数组（字段必须包含：ID、标题、海报图的占位色块、类型、上映年份、评分、导演、主演和剧情简介）。在首页主内容区，使用网格布局（CSS Grid）将这些电影以卡片的形式渲染出来。每张卡片展示封面、标题、年份和评分。鼠标悬浮在卡片上时，请加入轻微的上浮和阴影加深动效。另外请你在后续所有操作中始终遵循以下新增要求：第一，每当准备更改已有文件时，都必须先用相关工具阅读该文件；第二，在对已有文件作出修改时，必须以全量代码的形式输出并覆盖原文件，绝对不可以使用 Unified Diff 或者其他的补丁格式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 涵盖常见的Web应用代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.json', '.vue'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取整个文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生 Unicode 解码错误，说明文件包含了非法的 UTF-8 字符（可能使用其它编码）\\n return False\\n except Exception:\\n # 忽略权限不足或其他无关的文件操作异常\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的前端与后端代码文件扩展名\\n code_extensions = ('.html', '.js', '.css', '.py', '.ts', '.tsx', '.jsx', '.vue', '.json')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅针对代码文件进行检查\\n if not file.endswith(code_extensions):\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 若文件无法读取，跳过\\n continue\\n\\n valid_indents = 0\\n invalid_indents = 0\\n\\n for line in lines:\\n # 获取该行的前导空白字符\\n leading_ws = line[:len(line) - len(line.lstrip())]\\n \\n # 检查是否混用了制表符\\n if '\\\\t' in leading_ws:\\n return False\\n \\n stripped = line.lstrip()\\n # 忽略空行\\n if not stripped:\\n continue\\n \\n # 忽略常见的注释行，防止因为注释内容的排版引发误判\\n if stripped.startswith(('*', '<!--', '//', '/*')):\\n continue\\n \\n spaces = len(leading_ws)\\n if spaces > 0:\\n # 检查缩进是否为 4 的倍数\\n if spaces % 4 == 0:\\n valid_indents += 1\\n else:\\n invalid_indents += 1\\n \\n # 统计判定：如果存在未按 4 个空格控制的缩进行\\n if invalid_indents > 0:\\n # 如果完全没有合法的 4 空格缩进，或者非法缩进比例过高（大于等于 30%），则判定未满足要求\\n if valid_indents == 0 or (invalid_indents / valid_indents) >= 0.3:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义目标代码文件的扩展名\\n code_extensions = ('.html', '.htm', '.js', '.css', '.vue', '.jsx', '.ts', '.tsx', '.json', '.py')\\n code_files = []\\n \\n # 遍历工作区目录找出所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n if not code_files:\\n return False\\n \\n has_equals = False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n lines = content.split('\\\\n')\\n \\n for line in lines:\\n stripped_line = line.strip()\\n # 跳过空行或纯注释行\\n if not stripped_line:\\n continue\\n if stripped_line.startswith(('//', '<!--', '*', '/*', '#')):\\n continue\\n \\n # 简单去除行尾的注释内容，避免其内部的内容影响判定\\n if file_path.endswith(('.js', '.ts', '.vue', '.css', '.jsx', '.tsx')):\\n if '//' in line:\\n line = line.split('//')[0]\\n if '/*' in line:\\n line = line.split('/*')[0]\\n if file_path.endswith(('.html', '.htm', '.vue')):\\n if '<!--' in line:\\n line = line.split('<!--')[0]\\n if file_path.endswith('.py'):\\n if '#' in line:\\n line = line.split('#')[0]\\n \\n # 匹配所有独立的单等号（排除 ==, ===, >=, <=, !=, =>, += 等复合符号）\\n pattern = r'(?<![=!><+\\\\-*/%&|^~$])=(?![=>])'\\n matches = list(re.finditer(pattern, line))\\n \\n for m in matches:\\n has_equals = True\\n start = m.start()\\n end = m.end()\\n \\n # 左侧检查：确保刚好只有两个连续空格间距\\n left_ok = False\\n if start >= 2 and line[start-2:start] == ' ':\\n # 再往前一个字符如果不是空格，说明“恰好”有两个空格\\n if start == 2 or line[start-3] != ' ':\\n left_ok = True\\n \\n # 右侧检查：确保刚好只有两个连续空格间距\\n right_ok = False\\n if len(line) - end >= 2 and line[end:end+2] == ' ':\\n # 再往后一个字符如果不是空格，说明“恰好”有两个空格\\n if len(line) - end == 2 or line[end+2] != ' ':\\n right_ok = True\\n \\n # 只要有一个等号左右不严格遵循“分别保留两个空格间距”，则直接判断为未满足\\n if not (left_ok and right_ok):\\n return False\\n except Exception:\\n continue\\n \\n # 若整个项目的所有独立等号全部合规，并且确实包含了等号（排除空无内容的极端伪造场景），则返回True\\n return has_equals\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能生成的代码文件后缀类型\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.py'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 逐行读取检查，防止大文件占用过多内存\\n for line in f:\\n # 严格限制：只要包含制表符即判定为不满足要求\\n if '\\\\t' in line:\\n return False\\n except Exception:\\n # 对于非UTF-8格式或无法读取的文件直接跳过（该情况由其他要求规范约束）\\n continue\\n \\n # 如果项目中没有任何代码文件，也视作未满足要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在对已有文件作出修改时，必须以全量代码的形式输出并覆盖原文件，绝对不可以使用 Unified Diff 或者其他的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，接下来完善搜索和筛选功能，这对模型是个小挑战。请在首页电影列表上方增加一个过滤区：包含一个文本输入框（支持按标题、导演或主演模糊搜索），以及一组电影类型（Genre）的多选按钮（如科幻、动作、悬疑等）。要求：文本输入必须有 500ms 的防抖（Debounce）处理；文本搜索与类型多选是互相叠加的‘且’逻辑；每当过滤条件变化时，下方的电影卡片列表必须动态更新，并且在没有匹配结果时显示‘未找到相关电影’的空状态提示。另外请你在后续所有操作中始终遵循两项更新后的要求：第一，修改已有文件时，强制要求输出严格的Unified Diff格式，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效；第二，进行任何文件处理之前，切勿使用`pwd`确定当前的目录，原有的操作文件前需先运行pwd确认目录的要求不再生效。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能在 Web 开发或常见项目中用到的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.json', '.xml', '.yml', '.yaml', '.md', '.py', '.sh', '.bash',\\n '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.php', '.rb', '.go', '.rs'\\n }\\n \\n # 假如工作空间不存在，说明还未生成，不计为违反约束\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 遍历工作空间内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅校验被判定为代码类型的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 格式读取文件全部内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦发生 UnicodeDecodeError，证明其并非严格或兼容的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如权限不足等其他非编码相关的系统异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 针对前端项目和常见的代码格式，定义需要检查的文件扩展名\\n code_exts = {'.html', '.css', '.js', '.py', '.java', '.cpp', '.c', '.ts', '.jsx', '.tsx', '.json', '.vue'}\\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n files_checked += 1\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n indent_counts = []\\n for line in lines:\\n # 忽略空行或完全没有内容的行\\n if not line.strip():\\n continue\\n \\n # 匹配行首的前导空白\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n spaces = match.group(1)\\n # 只要有制表符即不符合“使用4个空格控制缩进”的规则\\n if '\\\\t' in spaces:\\n return False\\n \\n # 过滤掉一些常见的多行注释内部引发的非标准缩进\\n stripped_line = line.strip()\\n if stripped_line.startswith('*') or stripped_line.startswith('<!--') or stripped_line.startswith('/*'):\\n continue\\n \\n # 记录有效缩进的空格长度\\n indent_counts.append(len(spaces))\\n \\n if indent_counts:\\n # 统计非 4 的倍数的缩进数量\\n non_4_multiple = sum(1 for c in indent_counts if c % 4 != 0)\\n \\n # 容错判断：如果非 4 倍数缩进的行占总缩进行的 15% 以上\\n # 说明大概率采用了 2 空格或 3 空格缩进，或者缩进极度不规范\\n if non_4_multiple / len(indent_counts) > 0.15:\\n return False\\n\\n # 若未找到或未检查任何代码文件，则无法证明满足了该要求\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"def check_equals_spacing(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见前端和代码文件的扩展名集合\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue', '.py', '.java'}\\n has_files = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 正则表达式说明：\\n # (?<![<>!+*/%&|^=-]) : 左边不包含组成复合运算符的字符\\n # (=+) : 匹配一个或多个连续的 '='，当作等号整体\\n # (?![=>]) : 右侧不能是 '=' 或 '>'（排除 '=>' 和防止将 '=' 截断）\\n pattern = re.compile(r'(?<![<>!+*/%&|^=-])(=+)(?![=>])')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_files = True\\n \\n for line in lines:\\n # 去除行末尾的换行符，以便准确判断右侧是否存在空格\\n line_no_nl = line.rstrip('\\\\r\\\\n')\\n matches = pattern.finditer(line_no_nl)\\n \\n for m in matches:\\n start, end = m.span()\\n left_text = line_no_nl[:start]\\n right_text = line_no_nl[end:]\\n \\n # 检查左边：必须以恰好两个空格结尾\\n if not left_text.endswith(' '):\\n return False\\n # 并且排查是否有三个及以上连续空格\\n if len(left_text) >= 3 and left_text[-3] == ' ':\\n return False\\n \\n # 检查右边：必须以恰好两个空格开头\\n if not right_text.startswith(' '):\\n return False\\n # 并且排查是否有三个及以上连续空格\\n if len(right_text) >= 3 and right_text[2] == ' ':\\n return False\\n\\n # 如果工作区内检查了至少一个有效代码文件，且没有发现违规等号，则认为满足要求\\n return has_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.json', '.vue', '.ts', \\n '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h'\\n }\\n \\n # 1. 检查工作区内生成的代码文件\\n if workspace_path and os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 匹配行首的空白字符（包括空格和制表符）\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent = match.group(1)\\n # 如果行首缩进部分包含了制表符，则违反了要求\\n if '\\\\t' in indent:\\n return False\\n except Exception:\\n # 忽略可能出现编码错误或无法读取的文件\\n pass\\n \\n # 2. 检查助手回复中的 Markdown 代码块\\n # 使用正则表达式提取 ```代码语言...\\\\n ... ``` 中的代码内容\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n for line in block.split('\\\\n'):\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent = match.group(1)\\n # 同样检查代码块的行首缩进是否包含制表符\\n if '\\\\t' in indent:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"修改已有文件时，强制要求输出严格的Unified Diff格式，严禁直接提供全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在的搜索筛选很好用。接着我们来做电影的详情和交互状态。当用户点击某张电影卡片时，弹出一个居中的模态框（Modal）展示该电影的全部详细信息。在详情模态框底部，增加一个‘加入我的片单’按钮。点击后，按钮文字变成‘已加入’，并且将该电影的 ID 存入浏览器的 localStorage 中进行持久化。同时，关闭模态框回到首页时，所有已经存在于 localStorage 里的电影卡片，其右上角需要常驻显示一个黄色的‘★’图标作为已收藏标记。另外请你在后续所有操作中始终遵循两项更新后的要求：第一，运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作，原有的操作文件前无需运行`pwd`确认目录的要求不再生效；第二，修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式，原有的修改文件需输出Unified Diff格式的要求不再生效。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.vue', '.py', '.json', '.md', '.txt', '.xml', '.sh'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码格式验证\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用UTF-8进行解码，纯ASCII文件也能完美被UTF-8解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出UnicodeDecodeError，说明存在非UTF-8编码的文件\\n return False\\n except Exception:\\n # 如果因为权限或其他问题无法读取，为严谨起见视作未通过\\n return False\\n\\n # 要求必须至少存在一个创建出来的代码文件，且全部验证通过\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs', \\n '.html', '.css', '.py', '.vue', '.json', \\n '.java', '.c', '.cpp', '.h', '.hpp'\\n }\\n has_checked_any_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_checked_any_file = True\\n \\n non_zero_indents = []\\n for line in lines:\\n # 忽略空行\\n if not line.strip():\\n continue\\n \\n stripped = line.strip()\\n # 过滤多行注释及常见可能影响缩进对齐的非代码内容行\\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('//') or stripped.startswith('<!--'):\\n continue\\n \\n # 提取行首的空白字符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent_str = match.group(1)\\n # 如果存在制表符，违反了不使用制表符且要求4空格缩进的规则，直接判定失败\\n if '\\\\t' in indent_str:\\n return False\\n \\n spaces = len(indent_str)\\n if spaces > 0:\\n non_zero_indents.append(spaces)\\n \\n # 统计不符合 4 个空格基本倍数的缩进行数\\n if non_zero_indents:\\n non_4_mult_count = sum(1 for s in non_zero_indents if s % 4 != 0)\\n # 容错机制：允许少量合理的非 4 倍数对齐（如多行字符串或特定长参折行对齐等）\\n # 但如果是 2 空格或 3 空格缩进文件，此比例必定远超 30%，且绝对数量会较多\\n if non_4_mult_count > max(3, len(non_zero_indents) * 0.3):\\n return False\\n\\n return has_checked_any_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能的代码文件扩展名\\n valid_exts = {'.js', '.html', '.htm', '.css', '.vue', '.ts', '.jsx', '.tsx', '.py'}\\n \\n has_code_files = False\\n \\n # 匹配独立的单等号\\n # 排除复合运算符及比较运算符，如 !=, ==, ===, >=, <=, +=, -= 等，同时也排除 => 箭头函数\\n pattern = r'(?<![=!<>\\\\+\\\\-\\\\*/%&\\\\|\\\\^~])=(?![=>])'\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n # 如果无法按utf-8读取，跳过该文件\\n continue\\n \\n has_code_files = True\\n \\n for m in re.finditer(pattern, text):\\n i = m.start()\\n \\n # 统计等号左侧连续的空格数\\n left_spaces = 0\\n j = i - 1\\n while j >= 0 and text[j] == ' ':\\n left_spaces += 1\\n j -= 1\\n \\n # 统计等号右侧连续的空格数\\n right_spaces = 0\\n j = i + 1\\n while j < len(text) and text[j] == ' ':\\n right_spaces += 1\\n j += 1\\n \\n # 如果任一侧的空格间距不是恰好为2个，则认为不符合要求\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n \\n # 如果没有找到任何相关的代码文件，无法验证，返回False\\n if not has_code_files:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 如果工作区路径不存在，则无法验证代码，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.json', '.py', '.java', '.c', '.cpp', '.cs', '.go', '.php'\\n }\\n \\n # 遍历工作区目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取代码文件\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n # lstrip(' \\\\t') 会去除左侧连续的空格和制表符\\n # 借此计算行首空白部分的长度，并提取出真正的“缩进字符”\\n leading_spaces = line[:len(line) - len(line.lstrip(' \\\\t'))]\\n \\n # 判断缩进中是否使用了制表符\\n if '\\\\t' in leading_spaces:\\n return False\\n except Exception:\\n # 如果个别文件读取出错则跳过，继续检查其他文件\\n continue\\n \\n # 如果所有代码文件检查完毕且均无制表符缩进，则符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"下一步是完善‘我的片单’页面的复杂交互。当用户点击顶部导航的‘我的片单’时，清空首页视图并展示已收藏的电影列表。在这个专属视图中，用户可以对每部电影进行两项高阶操作：一是通过下拉菜单给电影打上私人标签（‘想看’、‘正在看’或‘已看’），二是提供一个可交互的 5 星评分组件给电影打私人分数。标签和私人评分的状态变更必须立刻更新并持久化到 localStorage。此外，每张卡片需提供一个‘移出片单’按钮，点击后伴随确认提示将电影从该页面及本地存储中删除。另外请你在后续所有操作中始终遵循两项更新后的要求：第一，修改任何现有文件时，必须严格采用Unified Diff格式输出，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效；第二，在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端Web项目常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.ts', '.tsx', '.jsx', '.md'}\\n \\n # 如果路径不存在，无法验证文件，直接返回 False 或是按照默认 True 处理。\\n # 通常如果确实生成了文件，路径必定存在。这里加上基本的路径验证。\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件包含无法被 UTF-8 解码的字符（例如使用了 GBK 编码）\\n return False\\n except Exception:\\n # 其他异常不属于编码判定范围，忽略处理\\n pass\\n \\n # 如果所有代码文件都能成功以 UTF-8 读取（或者全为纯 ASCII），则视为满足条件\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n from collections import Counter\\n \\n # 限制只检查常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json', '.vue', '.java', '.c', '.cpp'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n diffs = Counter()\\n prev_spaces = 0\\n \\n for line in lines:\\n # 忽略空行和只包含空白符的行\\n if not line.strip():\\n continue\\n \\n # 匹配行首前导空白字符\\n match = re.match(r'^([ \\\\t]+)\\\\S', line)\\n if match:\\n leading_ws = match.group(1)\\n \\n # 严禁使用制表符进行缩进\\n if '\\\\t' in leading_ws:\\n return False\\n \\n spaces = len(leading_ws)\\n else:\\n spaces = 0\\n \\n # 如果当前行缩进大于前一行，说明产生了新缩进层级，将其差值加入统计\\n if spaces > prev_spaces:\\n diffs[spaces - prev_spaces] += 1\\n \\n prev_spaces = spaces\\n \\n # 若文件包含缩进递进，最常见的主流缩进层级大小必须为4\\n if diffs:\\n most_common_diff = diffs.most_common(1)[0][0]\\n if most_common_diff != 4:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 设定需要检查的常见代码和纯文本文件扩展名\\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.json', '.md', '.txt'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n lines = content.split('\\\\n')\\n for line in lines:\\n i = 0\\n while i < len(line):\\n if line[i] == '=':\\n # 判断该 \\\"=\\\" 是否属于复合运算符的一部分 (如 ==, !=, >=, +=, =>, ~= 等)\\n is_compound = False\\n if i > 0 and line[i-1] in ':=!<>+-*/%&|^?~$':\\n is_compound = True\\n if i < len(line) - 1 and line[i+1] in '=>':\\n is_compound = True\\n \\n # 如果是孤立的 \\\"=\\\" (赋值或属性等号)，则严格检查两侧空格数量\\n if not is_compound:\\n # 统计左侧连续空格数\\n left_spaces = 0\\n j = i - 1\\n while j >= 0 and line[j] == ' ':\\n left_spaces += 1\\n j -= 1\\n \\n # 统计右侧连续空格数\\n right_spaces = 0\\n k = i + 1\\n while k < len(line) and line[k] == ' ':\\n right_spaces += 1\\n k += 1\\n \\n # 必须精确等于两个空格的间距，若非如此则违背该要求\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n i += 1\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 辅助函数：逐行检查缩进是否包含制表符\\n def check_indentation(lines) -> bool:\\n for line in lines:\\n # 匹配可能包含 Diff '+' 前缀的行，并捕获前面的所有空白字符\\n # 避免匹配到 '-' 开始的删除行，只针对新生成/保留的代码进行规范约束\\n match = re.match(r'^[+]?(\\\\s+)', line)\\n if match:\\n leading_spaces = match.group(1)\\n # 如果前导空白中使用了制表符，判定违规\\n if '\\\\t' in leading_spaces:\\n return False\\n return True\\n\\n # 1. 检查当前工作目录下的相关代码文件\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n # 根据用户指令中提到的“Web应用”，限定常见的代码与相关后缀\\n valid_exts = {'.js', '.css', '.html', '.htm', '.jsx', '.tsx', '.ts', '.json', '.vue'}\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除非相关且可能引起误判的隐藏目录及依赖目录\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file_name in files:\\n if os.path.splitext(file_name)[1].lower() in valid_exts:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n if not check_indentation(lines):\\n return False\\n except Exception:\\n continue\\n\\n # 2. 检查大模型回复内容中的 Markdown 代码块\\n code_blocks = re.findall(r'```.*?\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n if not check_indentation(lines):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，必须严格采用Unified Diff格式输出，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 针对前端和常见后端项目涉及的代码文件后缀进行校验\\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.py', '.vue', '.php', '.java', '.go', '.rb', '.c', '.cpp', '.h', '.hpp'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于非文本或其他原因导致无法用 UTF-8 读取，则跳过\\n continue\\n \\n # 提取非空行，忽略连续空行带来的干扰\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n if not non_empty_lines:\\n # 若为一个没有任何实质内容的空代码文件，显然没有包含需要的注释，不符合要求\\n return False\\n \\n # 检查该文件前 10 个非空行中是否包含目标字符串 (容忍 DOCTYPE 声明等特殊头部)\\n found_in_head = False\\n for line in non_empty_lines[:10]:\\n if target_text in line:\\n found_in_head = True\\n break\\n \\n if not found_in_head:\\n # 如果发现任何一个代码文件在开头未添加指定注释，则直接判定为不满足要求\\n return False\\n \\n # 如果遍历了所有文件并且都符合要求，且至少存在一个代码文件，则判定为满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后，为了让应用更具个性化，我们需要在‘首页’的右侧（或顶部下方）加入一个‘猜你喜欢’的本地推荐算法模块。请实现一个纯前端的推荐逻辑：每次加载或切换回首页时，读取 localStorage 中‘我的片单’里的电影，统计用户收藏中最频繁出现的‘电影类型’和‘导演’的权重；然后遍历本地库中所有**尚未加入片单**的电影，按照计算出的特征匹配度（相似度得分）进行降序排序，选取前 3 部最高分的电影展示在推荐模块中。要求算法具有容错性，若片单为空，则随机推荐 3 部电影。另外请你在后续所有操作中始终遵循两项更新及新增要求：第一，在你生成的每一个代码文件中，请于结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”，原有的在代码开头添加该注释的要求不再生效；第二，限制每个生成的代码文件总行数（含空行）必须是偶数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前后端代码及配置文件类型扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.json', '.py', '.sh', '.md', '.txt'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅验证代码类型的文本文件，排除图片或其它二进制文件的干扰\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 对于其它的不可预知读取异常也算作验证失败\\n return False\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据“Web应用”的指令推断，设置常见的前端代码文件扩展名\\n valid_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_exts):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 若存在编码读取等异常，暂且忽略跳过\\n continue\\n \\n for line in lines:\\n stripped_line = line.lstrip()\\n # 跳过全空行\\n if not stripped_line:\\n continue\\n \\n # 排除多行注释的中间及结束行（例如 \\\" * comment\\\" 或 \\\" */\\\"）\\n if stripped_line.startswith('*'):\\n continue\\n \\n leading_len = len(line) - len(stripped_line)\\n leading_str = line[:leading_len]\\n \\n # 若前导空白中包含制表符，说明未严格使用空格缩进，违规\\n if '\\\\t' in leading_str:\\n return False\\n \\n # 若前导空格的数量不是 4 的倍数，说明未使用 4 个空格控制缩进，违规\\n if leading_len % 4 != 0:\\n return False\\n \\n # 只要存在至少一个代码文件且全部通过了缩进检验，即算满足要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件中，所有的“=”两边应分别保留两个空格的间距\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 匹配独立的等号 \\\"=\\\"，排除常见的复合运算符如 +=, -=, ==, ===, <=, >=, !=, => 等\\n pattern = re.compile(r'(?<![+\\\\-*/%&|^<>=!~$])=(?![=>])')\\n found_any_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 忽略隐藏文件或.git等版本控制目录\\n if '.git' in root or file.startswith('.'):\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n continue # 跳过二进制或非UTF-8编码文件\\n\\n found_any_file = True\\n\\n for match in pattern.finditer(content):\\n start = match.start()\\n end = match.end()\\n\\n # 检查等号前是否恰好有两个空格\\n if start < 2:\\n return False # 前方字符不足两个\\n if content[start-2:start] != \\\" \\\":\\n return False # 前方紧挨着的两个字符不是空格\\n if start >= 3 and content[start-3] == \\\" \\\":\\n return False # 前方第三个字符还是空格（超过两个）\\n\\n # 检查等号后是否恰好有两个空格\\n if end + 2 > len(content):\\n return False # 后方字符不足两个\\n if content[end:end+2] != \\\" \\\":\\n return False # 后方紧挨着的两个字符不是空格\\n if end + 2 < len(content) and content[end+2] == \\\" \\\":\\n return False # 后方第三个字符还是空格（超过两个）\\n\\n # 如果存在至少一个代码文件并且所有的 '=' 都满足要求，则返回 True\\n return found_any_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中的所有文件路径以及代码位置，请务必始终且仅使用双引号进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意缩进规范：生成的代码只允许使用空格缩进，不可使用制表符\",\n \"validation_code\": \"def check_indentation_rule(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n has_code = False\\n\\n # 1. 检查工作区实际生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n has_code = True\\n except Exception:\\n # 忽略非文本文件或无法读取的文件\\n continue\\n \\n for line in lines:\\n # 匹配行首所有的空格和制表符作为缩进部分\\n match = re.match(r'^[ \\\\t]+', line)\\n if match:\\n # 如果缩进内容中包含制表符，则判定为不满足要求\\n if '\\\\t' in match.group(0):\\n return False\\n\\n # 2. 检查最后一轮回复中包含的代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n if code_blocks:\\n has_code = True\\n for block in code_blocks:\\n for line in block.split('\\\\n'):\\n line_to_check = line\\n # 兼容 Unified Diff 格式：忽略行首表示变更的 '+' 或 '-'\\n if line.startswith('+') or line.startswith('-'):\\n line_to_check = line[1:]\\n \\n # 匹配行首缩进\\n match = re.match(r'^[ \\\\t]+', line_to_check)\\n if match:\\n if '\\\\t' in match.group(0):\\n return False\\n\\n # 必须至少包含相关的代码内容，且在检查中全部合格才返回 True\\n return has_code\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"为了实现高效率，必须尽可能地采取并行方式来执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，必须严格采用Unified Diff格式输出，严禁直接提供全量代码替换，原有的全量代码输出要求不再生效\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 针对本地电影库Web前端应用，筛选常见的代码文件扩展名\\n valid_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n \\n files_found = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read().strip()\\n except Exception:\\n # 无法读取（不符合第一轮强制UTF-8的要求）等异常情况，视为未满足\\n return False\\n \\n # 如果代码文件为空，则显然没有包含要求的注释\\n if not content:\\n return False\\n \\n # 必须包含目标字符串\\n if target_text not in content:\\n return False\\n \\n # 验证是否于\\\"结尾处\\\"添加了注释，截取文件的最后10个非空行进行校验\\n lines = [line.strip() for line in content.split('\\\\n') if line.strip()]\\n found_at_end = False\\n for line in lines[-10:]:\\n if target_text in line:\\n found_at_end = True\\n break\\n \\n if not found_at_end:\\n return False\\n \\n # 如果找到了相应的代码文件且全部满足要求，则返回True；若无任何代码文件，返回False\\n return files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"限制每个生成的代码文件总行数（含空行）必须是偶数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名，根据web应用的语境重点关注前端相关文件\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照 '\\\\n' 切分可以准确地统计总行数（包含末尾空行及连续空行）\\n lines = content.split('\\\\n')\\n \\n # 如果行数不是偶数，则说明违反了要求\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取，为保险起见可以直接视为未通过测试，或跳过\\n # 根据上下文要求“无一例外用UTF-8编码”，无法使用UTF-8读取也可视为失败\\n return False\\n \\n # 如果没有找到任何代码文件，说明没有生成有效项目，返回False；否则全满足条件返回True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["观察页面顶部，确认存在导航栏，且导航栏左侧包含应用Logo或标题，右侧包含'首页'和'我的片单'两个可点击的Tab。","在首页视图中，观察电影列表是否以网格（Grid）形式排列展示，并确认每张电影卡片上至少直观显示了封面图（占位色块）、标题、上映年份和评分信息。","将鼠标光标悬浮在首页任意一张电影卡片上，观察卡片是否触发了轻微的上浮动画以及阴影加深效果。","在首页上方的搜索文本框中输入任意已存在电影的主演或导演名字，等待约1秒，观察下方的电影列表是否已动态过滤并仅显示与该输入相关的电影卡片。","在首页上方的类型过滤区，点击选中一个电影类型（如'科幻'），观察下方的电影列表是否动态更新，仅显示包含该类型的电影卡片。","同时测试搜索和类型筛选：在搜索框输入某电影关键字，并选中一个不匹配该关键字的电影类型。观察页面是否执行了'且'逻辑，即如果没有同时满足这两个条件的电影，列表不应显示该电影。","在搜索框中输入一串明显不存在的乱码（如'xyz123'），等待页面更新，观察电影列表区是否清空并正确显示了'未找到相关电影'的提示文案。","点击首页列表中的任意一张电影卡片，验证页面是否弹出了一个居中的模态框（Modal），并且模态框内展示了该电影的详细信息（如剧情简介、导演、主演等完整字段）。","在详情模态框中找到并点击底部'加入我的片单'按钮，观察按钮的文字是否变更为'已加入'；随后关闭模态框返回首页，观察刚刚操作的电影卡片右上角是否长驻显示了一个黄色的'★'图标。","点击顶部导航栏的'我的片单'Tab，验证页面是否清空了首页的内容，并正确展示了一个只包含刚刚打上黄色星号的已收藏电影列表。","在'我的片单'页面中，点击某部电影卡片上的标签下拉菜单，将其选项修改为'正在看'。随后刷新浏览器页面，确认该电影卡片的标签选项依然保持选中为'正在看'，证明状态已持久化保存。","在'我的片单'页面中，找到某部电影卡片上的5星评分组件，点击将其评分为4颗星。随后刷新浏览器页面，确认该电影的评分依然保持为4颗星高亮，证明评分数据已持久化保存。","在'我的片单'页面中，点击某部电影卡片提供的'移出片单'按钮，验证是否出现操作确认提示；确认后该电影从当前列表消失。点击'首页'回到所有电影列表，观察该电影右上角的黄色星号标记是否已经被正确移除。","确保'我的片单'中没有任何电影（为空状态），切换到'首页'，观察右侧或顶部下方是否存在'猜你喜欢'模块，并确认里面是否随机展示了3部电影卡片。","在'首页'中挑选3部同一类型（如'科幻'）的电影并加入'我的片单'，刷新页面留在首页，观察'猜你喜欢'模块展示的3部电影，确认它们没有包含已经加入片单的电影，并且主要倾向于推荐你刚刚收藏的同类型（或同导演）的电影。"],"string":"[\n \"观察页面顶部，确认存在导航栏，且导航栏左侧包含应用Logo或标题，右侧包含'首页'和'我的片单'两个可点击的Tab。\",\n \"在首页视图中，观察电影列表是否以网格（Grid）形式排列展示，并确认每张电影卡片上至少直观显示了封面图（占位色块）、标题、上映年份和评分信息。\",\n \"将鼠标光标悬浮在首页任意一张电影卡片上，观察卡片是否触发了轻微的上浮动画以及阴影加深效果。\",\n \"在首页上方的搜索文本框中输入任意已存在电影的主演或导演名字，等待约1秒，观察下方的电影列表是否已动态过滤并仅显示与该输入相关的电影卡片。\",\n \"在首页上方的类型过滤区，点击选中一个电影类型（如'科幻'），观察下方的电影列表是否动态更新，仅显示包含该类型的电影卡片。\",\n \"同时测试搜索和类型筛选：在搜索框输入某电影关键字，并选中一个不匹配该关键字的电影类型。观察页面是否执行了'且'逻辑，即如果没有同时满足这两个条件的电影，列表不应显示该电影。\",\n \"在搜索框中输入一串明显不存在的乱码（如'xyz123'），等待页面更新，观察电影列表区是否清空并正确显示了'未找到相关电影'的提示文案。\",\n \"点击首页列表中的任意一张电影卡片，验证页面是否弹出了一个居中的模态框（Modal），并且模态框内展示了该电影的详细信息（如剧情简介、导演、主演等完整字段）。\",\n \"在详情模态框中找到并点击底部'加入我的片单'按钮，观察按钮的文字是否变更为'已加入'；随后关闭模态框返回首页，观察刚刚操作的电影卡片右上角是否长驻显示了一个黄色的'★'图标。\",\n \"点击顶部导航栏的'我的片单'Tab，验证页面是否清空了首页的内容，并正确展示了一个只包含刚刚打上黄色星号的已收藏电影列表。\",\n \"在'我的片单'页面中，点击某部电影卡片上的标签下拉菜单，将其选项修改为'正在看'。随后刷新浏览器页面，确认该电影卡片的标签选项依然保持选中为'正在看'，证明状态已持久化保存。\",\n \"在'我的片单'页面中，找到某部电影卡片上的5星评分组件，点击将其评分为4颗星。随后刷新浏览器页面，确认该电影的评分依然保持为4颗星高亮，证明评分数据已持久化保存。\",\n \"在'我的片单'页面中，点击某部电影卡片提供的'移出片单'按钮，验证是否出现操作确认提示；确认后该电影从当前列表消失。点击'首页'回到所有电影列表，观察该电影右上角的黄色星号标记是否已经被正确移除。\",\n \"确保'我的片单'中没有任何电影（为空状态），切换到'首页'，观察右侧或顶部下方是否存在'猜你喜欢'模块，并确认里面是否随机展示了3部电影卡片。\",\n \"在'首页'中挑选3部同一类型（如'科幻'）的电影并加入'我的片单'，刷新页面留在首页，观察'猜你喜欢'模块展示的3部电影，确认它们没有包含已经加入片单的电影，并且主要倾向于推荐你刚刚收藏的同类型（或同导演）的电影。\"\n]","isConversational":false}}},{"rowIdx":52,"cells":{"task_id":{"kind":"number","value":53,"string":"53"},"repository_policy":{"kind":"string","value":"你作为代码智能体，将为我提供代码开发、项目调整、问题排查等相关技术支持，后续所有交互轮次都必须严格遵循以下约束条件：\n1. 必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改。\n2. 请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容。\n3. 你创建的代码文件无一例外，全都要用UTF-8进行编码。\n4. 在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行。\n5. 在你生成的每个代码文件中，请将所有注释全用英文呈现。\n6. 你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n7. 无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词。\n8. 在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef verify_utf8_encoding(response: str, workspace_path: str) -> bool:\n # 确保给定的项目工作区路径存在\n if not os.path.exists(workspace_path):\n return False\n \n code_files_found = False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令中 -*- coding: utf-8 -*- 的特征推测，目标代码文件类型为 Python 文件\n if file.endswith('.py'):\n code_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制读取文件内容\n with open(file_path, 'rb') as f:\n raw_content = f.read()\n \n # 尝试用 utf-8 进行解码\n raw_content.decode('utf-8')\n \n except UnicodeDecodeError:\n # 如果抛出解码异常，说明文件不是用 UTF-8 编码的\n return False\n except Exception:\n # 忽略其他可能因文件权限导致无法读取的不可抗力异常\n pass\n \n # 如果找到了至少一个代码文件，并且全都没有触发UnicodeDecodeError异常，则验证通过\n return code_files_found","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts', \n '.go', '.rb', '.php', '.sh', '.html', '.css', '.rs', '.swift', '.kt', '.m'\n }\n\n # 如果工作区路径不存在，视作未生成任何违反规则的文件\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选出代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查最首行是否包含给定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若发生读取异常（如编码错误或权限问题），则视为不符合要求\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常见代码文件的扩展名集合，涵盖各类主流编程语言及标记语言\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \n '.cs', '.go', '.rs', '.js', '.jsx', '.ts', '.tsx', '.html', \n '.htm', '.css', '.scss', '.less', '.php', '.rb', '.swift', \n '.kt', '.kts', '.sh', '.bash', '.m', '.scala', '.pl', '.sql', '.vue'\n }\n \n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件，则进行结尾注释检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 去除文件首尾的空白字符和换行符\n content = f.read().strip()\n \n # 跳过空文件\n if not content:\n continue\n \n # 获取文件的最后一行内容（去除可能的首尾空白）\n lines = content.split('\\n')\n last_line = lines[-1].strip()\n \n # 判断最后一行是否包含了目标声明字符串\n if target_statement not in last_line:\n return False\n except Exception:\n # 遇到系统级别的文件读取异常保守处理，略过该文件\n pass\n \n return True","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复的最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除回复首部的空白字符，并判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_self_reference(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查AI助手的回复中是否将自称\"我\"替换为了\"本小姐\"。\n 通过检查回复文本中是否不再包含\"我\"字来进行严格判断。\n \"\"\"\n # 如果回复中出现了\"我\"，说明没有严格完成替换，返回False\n if \"我\" in response:\n return False\n \n # 如果没有出现\"我\"，则认为满足了要求\n return True","tags":["Content","Persona"]}],"string":"[\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify_utf8_encoding(response: str, workspace_path: str) -> bool:\\n # 确保给定的项目工作区路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files_found = False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令中 -*- coding: utf-8 -*- 的特征推测，目标代码文件类型为 Python 文件\\n if file.endswith('.py'):\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制读取文件内容\\n with open(file_path, 'rb') as f:\\n raw_content = f.read()\\n \\n # 尝试用 utf-8 进行解码\\n raw_content.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明文件不是用 UTF-8 编码的\\n return False\\n except Exception:\\n # 忽略其他可能因文件权限导致无法读取的不可抗力异常\\n pass\\n \\n # 如果找到了至少一个代码文件，并且全都没有触发UnicodeDecodeError异常，则验证通过\\n return code_files_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts', \\n '.go', '.rb', '.php', '.sh', '.html', '.css', '.rs', '.swift', '.kt', '.m'\\n }\\n\\n # 如果工作区路径不存在，视作未生成任何违反规则的文件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选出代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查最首行是否包含给定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若发生读取异常（如编码错误或权限问题），则视为不符合要求\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常见代码文件的扩展名集合，涵盖各类主流编程语言及标记语言\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \\n '.cs', '.go', '.rs', '.js', '.jsx', '.ts', '.tsx', '.html', \\n '.htm', '.css', '.scss', '.less', '.php', '.rb', '.swift', \\n '.kt', '.kts', '.sh', '.bash', '.m', '.scala', '.pl', '.sql', '.vue'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件，则进行结尾注释检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 去除文件首尾的空白字符和换行符\\n content = f.read().strip()\\n \\n # 跳过空文件\\n if not content:\\n continue\\n \\n # 获取文件的最后一行内容（去除可能的首尾空白）\\n lines = content.split('\\\\n')\\n last_line = lines[-1].strip()\\n \\n # 判断最后一行是否包含了目标声明字符串\\n if target_statement not in last_line:\\n return False\\n except Exception:\\n # 遇到系统级别的文件读取异常保守处理，略过该文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复的最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复首部的空白字符，并判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_self_reference(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查AI助手的回复中是否将自称\\\"我\\\"替换为了\\\"本小姐\\\"。\\n 通过检查回复文本中是否不再包含\\\"我\\\"字来进行严格判断。\\n \\\"\\\"\\\"\\n # 如果回复中出现了\\\"我\\\"，说明没有严格完成替换，返回False\\n if \\\"我\\\" in response:\\n return False\\n \\n # 如果没有出现\\\"我\\\"，则认为满足了要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"你好！我们需要开发一个包含前后端的活动预订系统前端原型。第一步，请帮我搭建项目的整体骨架。需求如下：1. 创建一个包含顶部导航栏的基础布局，导航栏左侧是Logo，右侧包含两个Tab：'活动预订' 和 '管理后台'。2. 实现简单的路由或视图切换逻辑：点击 Tab 可以切换下方主内容区的视图。3. 视图当前只需展示基础占位内容，比如在'活动预订'页面放一个标题和一个'立即探索'按钮，'管理后台'放一个欢迎标题。请给出结构清晰、带有现代感基础 CSS 样式的代码。另外请注意，后续所有轮次都请始终遵循以下要求：第一，生成的代码项目中，任何目录的层深都不允许大于 2 层；第二，必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 常见的代码文件和配置/说明文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.py', '.json', '.md', '.txt', '.sh'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用utf-8进行解码\n # 如果文件不是utf-8编码，则会抛出UnicodeDecodeError\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦发现无法用UTF-8解码的文件，说明不满足要求\n return False\n except Exception:\n # 忽略其他无关的异常（例如权限问题导致的读取失败）\n pass\n \n # 所有代码文件都能够正常被utf-8解码\n return True","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端和常见后端的代码文件扩展名\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.h', '.go', '.php', '.rb', '.sh'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查代码文件的最首行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果无法读取文件或编码不是 utf-8，则视作不满足要求\n return False\n\n # 若至少存在一个代码文件，且所有代码文件的最首行均满足条件，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见的主流编程语言与前端代码后缀名\n code_exts = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.cpp', '.c', '.cc', '.h', '.hpp', \n '.cs', '.go', '.php', '.rb', '.swift', '.kt', '.sh', '.bash', '.sql'\n }\n \n # 需要跳过的第三方包或环境目录，防止误判\n ignore_dirs = {'.git', 'node_modules', 'venv', 'env', '__pycache__', 'dist', 'build'}\n \n code_files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read().strip()\n except Exception:\n # 若因非UTF-8编码或其他读取异常，不符合文件编码要求，直接返回 False\n return False\n \n # 如果文件是空的，显然没包含必要的注释\n if not content:\n return False\n \n # 检查是否包含指定的声明\n if target_statement not in content:\n return False\n \n # 获取该声明最后出现的位置之后的所有尾部文本\n last_idx = content.rfind(target_statement)\n tail = content[last_idx + len(target_statement):].strip()\n \n # 检查尾部是否仅包含合法的注释关闭字符及空白字符\n # 例如： HTML: --> , CSS/JS: */ , Python: \"\"\" 或 空等\n allowed_chars = set(\"/*->\\\"'%# \\t\\r\\n\")\n if not all(c in allowed_chars for c in tail):\n return False\n \n code_files_checked += 1\n \n # 如果工作区中没找到任何代码文件，说明未生成相关代码\n if code_files_checked == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复的最开头是否带有\"喵～\"语气词。\n \"\"\"\n # 移除开头可能的空白字符，判断是否以\"喵～\"开头\n if not response:\n return False\n \n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证 AI 助手的回复中是否将自称的“我”全部替换为“本小姐”。\n 在当前代码交流的上下文中，只要回复文本中仍然包含“我”字，\n 即可认为其没有严格执行替换。\n \"\"\"\n # 只要存在“我”字，就说明替换未完全满足要求\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中，任何目录的层深都不允许大于 2 层","validation_code":"def check_directory_depth(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，则无法判断或视为未生成项目\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录（如 .git 等），防止由于环境带来的干扰导致误判\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 计算当前目录相对于 workspace_path 的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 如果是根目录本身，层级深度为 0\n if rel_path == '.':\n depth = 0\n else:\n # 根据系统的路径分隔符切分相对路径，列表的长度即为相对于根目录的层深\n depth = len(rel_path.split(os.sep))\n \n # 若发现任何目录层深大于 2，则不满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":1,"instruction":"太棒了，框架已经起来了。现在我们来细化“活动预订”页面。我们需要在这里实现一个日历与活动联动的组件：1. 不要使用原生的 input date，请用纯代码渲染一个完整的当前月份的日历视图。2. 在日历中挑选几个日期，用小圆点在日期下方标记为“有活动”。3. 当用户点击“有活动”的日期时，日历右侧（或下方）滑出一个活动列表，展示2-3个虚构的活动卡片（包含活动名称、时间段、票价起步）。如果点击没有活动的日期，则显示“当日暂无活动安排”的空状态提示。另外请注意，后续所有轮次都请遵循以下更新后的要求：第一，原有代码项目目录层深不大于2层的要求调整为生成的代码项目中任何目录的层深都不允许大于3层；第二，为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前端、后端及文档等代码文件类型后缀\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.sh', \n '.md', '.json', '.xml', '.txt'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件类型的编码\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦发现无法用UTF-8解码的代码文件，即判定为不满足要求\n return False\n except Exception:\n # 忽略诸如权限不足导致的读取错误，只针对编码异常做判定\n continue\n\n # 若所有筛选出的代码文件都能以UTF-8正常读取，则满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"import os\n\ndef check_encoding_declaration(response: str, workspace_path: str) -> bool:\n # 定义常见的前后端代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.h'}\n # 排除可能包含大量非AI直接生成代码的目录\n exclude_dirs = {'node_modules', '.git', '__pycache__', 'dist', 'build'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤需要排除的目录，避免深度遍历第三方库和构建产物\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 使用 utf-8 编码打开文件并读取最首行\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查指定的编码声明是否出现在第一行\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法打开或者编码错误，无法验证其是否满足要求，视为不通过\n return False\n \n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 常见的代码文件后缀集合，排除 Markdown 等非代码类文档以及不能随意写注释的 JSON 等\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', \n '.html', '.css', '.scss', '.less', \n '.vue', '.svelte', '.java', '.c', \n '.cpp', '.h', '.cs', '.go', '.rb', '.php'\n }\n \n if not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 过滤掉完全空白的行\n lines = [line for line in lines if line.strip()]\n \n if not lines:\n return False\n \n # 检查结尾处（最后 5 个非空行）是否包含目标声明\n # 使用多行检查以兼容将多行注释符单独成行的情况，如：\n # <!--\n # AI-Assisted: ...\n # -->\n found = False\n for line in lines[-5:]:\n if target_statement in line:\n found = True\n break\n \n if not found:\n return False\n \n except Exception:\n # 若文件无法读取或抛出异常，算作不符合要求\n return False\n \n # 如果找到了代码文件，并且所有代码文件都符合要求，则返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的最后一轮回复的最开头是否带有\"喵～\"语气词。\n \"\"\"\n # 移除最开头的空白字符后，判断是否以\"喵～\"开头\n if response.lstrip().startswith(\"喵～\"):\n return True\n return False","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n # 检查回复中是否还存在“我”字\n # 如果存在，说明助手存在自称未能严格替换为“本小姐”的违规情况\n if \"我\" in response:\n return False\n \n # 不存在“我”，视为满足要求\n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中任何目录的层深都不允许大于3层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果项目路径不存在，自然不存在大于3层的目录\n if not os.path.exists(workspace_path):\n return True\n \n for dirpath, dirnames, filenames in os.walk(workspace_path):\n # 过滤掉隐藏目录(如 .git, .vscode)以及 __pycache__ 缓存等非AI直接生成的目录\n dirnames[:] = [d for d in dirnames if not d.startswith('.') and d != '__pycache__']\n \n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(dirpath, workspace_path)\n \n if rel_path == '.':\n depth = 0\n else:\n # 统一路径分隔符并计算层级深度\n normalized_path = rel_path.replace('\\\\', '/')\n depth = len(normalized_path.split('/'))\n \n # 如果当前目录层深大于3，则不满足要求\n if depth > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":2,"instruction":"日历组件做得很漂亮！接下来是核心挑战，我们需要在用户点击某个活动卡片后，进入选座环节。1. 隐藏日历，在页面上渲染一个 SVG 构成的座位分布图，模拟一个小型剧院，顶部要有“舞台”标识。2. 座位需要分为前排的 VIP 区（金色或橙色）和后排的普通区（蓝色）。3. 用 SVG 的圆形或圆角矩形代表座位，并且座位需要支持三种状态：'可选'（默认颜色）、'已售'（灰色且不可点击）、'已选'（绿色）。4. 实现点击交互：点击'可选'座位使其状态变为'已选'，再次点击取消选中。另外请注意，后续所有轮次都请遵循以下更新后的要求：第一，原有代码项目目录层深不大于3层的要求调整为生成的代码项目中任何目录的层深都不允许大于4层；第二，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件后缀集合（包含了前端原型系统最常用的以及部分常见的后端文件格式）\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.json', '.svg', '.py', '.java', '.c', '.cpp', '.h', \n '.cs', '.go', '.rs', '.php', '.rb', '.sh'\n }\n \n has_code_file = False\n \n # 遍历代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅校验代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 以二进制形式读取文件\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试以 UTF-8 格式解码\n raw_data.decode('utf-8')\n \n except UnicodeDecodeError:\n # 如果发生 UnicodeDecodeError 异常，说明文件不是 UTF-8 编码\n return False\n except Exception:\n # 如果文件读取发生其他异常，视为未能满足验证，返回 False\n return False\n \n # 如果项目中没有生成任何代码文件，则无法被视为满足了“创建代码文件并用UTF-8编码”的前提\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义需要检查的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.cpp', '.c', '.h', '.go', '.rb', '.php', '.sh'\n }\n \n # 定义需要忽略的目录（依赖包、编译产物、隐藏目录等）\n ignore_dirs = {\n '.git', 'node_modules', '__pycache__', 'dist', 'build', \n '.idea', '.vscode', 'venv', 'env'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录和以点开头的隐藏目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs and not d.startswith('.')]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含目标编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法正常读取，视作未满足要求\n return False\n\n # 必须确保至少检查了一个代码文件，防止空项目误判为True\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 前端开发常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\n # 忽略第三方库等与 AI 生成源代码无关的目录\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'venv', 'env'}\n \n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果不是 UTF-8 编码或无法读取，违背了无一例外用 UTF-8 的系统提示要求\n return False\n \n # 如果文件为空，必然没有包含目标声明\n if not lines:\n return False\n \n # 从尾部开始遍历，找到第一个非空行（真实的最后一行）\n last_line = \"\"\n for line in reversed(lines):\n if line.strip():\n last_line = line.strip()\n break\n \n # 检查指定的声明文本是否在最后一行的注释中\n if target_statement not in last_line:\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容最开头的空白字符（如空格、换行符等）\n cleaned_response = response.lstrip()\n \n # 检查清理后的回复是否以\"喵～\"开头\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否满足：将所有自称“我”的地方严格换成“本小姐”。\n 在日常技术回复语境下，若出现“我”字，极大概率是自称遗漏了替换。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中任何目录的层深都不允许大于4层","validation_code":"import os\nfrom pathlib import Path\n\ndef check_directory_depth(response: str, workspace_path: str) -> bool:\n workspace = Path(workspace_path)\n \n # 如果工作区目录不存在，可以直接认为没有违反限制\n if not workspace.exists() or not workspace.is_dir():\n return True\n \n # 定义需要忽略的、非AI直接生成代码的常见依赖或隐藏目录\n ignored_dirs = {\n 'node_modules', '__pycache__', 'venv', 'env', \n '.venv', '.env', 'dist', 'build', 'out'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 裁剪不需要遍历的目录（隐藏目录如 .git 以及常见的依赖包目录）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ignored_dirs]\n \n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = Path(root).relative_to(workspace)\n \n # 计算目录深度：相对路径的组成部分数量\n # 若 rel_path 为 '.'，则 parts 为空元组 ()，长度即深度为 0\n depth = len(rel_path.parts)\n \n # 如果层深大于 4，则不满足要求\n if depth > 4:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"SVG选座的交互很流畅。现在来完善选座的业务逻辑和购物车功能：1. 在座位图右侧增加一个固定的“已选座位”面板。2. 当用户在 SVG 图中选中座位时，面板要实时更新，列出选中的座位排号（如 A排3座）及对应的单价（VIP区和普通区价格需有区分），并在底部计算总价。3. 增加一条限制规则：单次预订最多只能选择 4 个座位。当用户尝试选择第 5 个时，在页面中心弹出一条醒目的错误 Toast 提示并阻止选中。4. 面板最下方增加一个“确认选座”按钮，点击后弹出一个模态框，要求填写联系人姓名和手机号。\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有代码项目目录层深不大于4层的要求调整为生成的代码项目中任何目录的层深都不允许大于5层；第二，确保你生成的任何代码文件中都不包含\"TODO\"关键词。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\n # 若工作区路径不存在，则无法产生不满足条件的文件，视为空集满足\n if not os.path.exists(workspace_path):\n return True\n\n # 常见的前端、后端及文档的代码文件扩展名集合\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.css', '.scss', '.less', '.md', '.json', '.vue', \n '.sh', '.txt', '.svg', '.xml'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试用 utf-8 编码读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明并非合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略权限等其他导致文件无法读取的异常\n pass\n \n # 所有代码文件均成功使用 UTF-8 读取\n return True","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 识别常见的代码文件扩展名\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.svg'}\n \n if not os.path.exists(workspace_path):\n # 如果项目路径不存在，视作不满足要求\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 判断指定的编码声明是否出现在代码的最首行\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果打开或读取文件失败（例如编码不是UTF-8引发解码异常等），认为不满足要求\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义前端原型开发及其他常见代码文件后缀\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue',\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift'\n }\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏目录和常见的依赖环境目录，避免检查非AI生成的依赖文件\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__', 'venv', 'env']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于特殊原因无法读取（如非utf-8编码等），则跳过\n continue\n \n # 过滤掉文件最末尾的空行\n while lines and not lines[-1].strip():\n lines.pop()\n \n # 如果代码文件为空，显然不满足包含注释的要求\n if not lines:\n return False\n \n # 提取最后3行。这样可以兼容注释闭合符号如 '*/' 或 '-->' 在最末尾的情况\n last_few_lines = [line.strip() for line in lines[-3:]]\n \n # 判断最后几行中是否包含了特定的声明\n statement_found = any(target_statement in line for line in last_few_lines)\n \n if not statement_found:\n return False\n \n return found_code_file","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查AI助手的回复最开头是否带有\"喵～\"语气词。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，代表该要求是否被满足\n \"\"\"\n # 去除回复内容最前面的空白字符后，判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否将自称'我'都换成了'本小姐'。\n 只要回复中不包含'我'字，即可认为满足要求。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中任何目录的层深都不允许大于5层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区目录不存在，则说明尚未生成任何内容，不违背层深要求\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n # 需要忽略的常见非 AI 直接生成的依赖、环境或构建目录，避免深层结构引发误判\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'venv', '.venv', 'dist', 'build', 'coverage'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 列表，阻止 os.walk 进一步深入这些被忽略的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n # 获取当前遍历目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n if rel_path == '.':\n depth = 0\n else:\n # 兼容不同操作系统的路径分隔符，分割以计算层深\n parts = [p for p in rel_path.replace('\\\\', '/').split('/') if p]\n depth = len(parts)\n \n # 若发现任何目录层深大于 5，则验证失败\n if depth > 5:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts',\n '.jsx', '.tsx', '.vue', '.html', '.css', '.scss', '.less', '.go',\n '.rs', '.php', '.rb', '.swift', '.kt', '.sh', '.bat', '.ps1'\n }\n \n # 若工作区路径不存在或不是目录，说明未生成文件，自然不包含 TODO\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n # 遍历项目目录中的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理代码文件\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 检查是否包含 \"TODO\" 关键词\n if \"TODO\" in content:\n return False\n except Exception:\n # 如果由于权限等原因读取失败，则跳过\n continue\n \n # 如果所有的代码文件都没有包含 \"TODO\"，则满足要求\n return True","tags":["Content","Keyword"]}]},{"round_id":4,"instruction":"下一步是出票环节。当用户在模态框填完信息并点击“提交订单”后：1. 隐藏选座和表单界面，展示一个“预订成功”的门票详情视图。2. 页面中需要展示一张设计精美的电子门票卡片（带有一点阴影和圆角等拟物感），卡片上展示活动名称、时间、用户姓名和所选座位号。3. 在门票右侧或下半部分，使用前端代码（可以用一组排列紧密的黑色 SVG 小方块或 div）生成一个模拟的二维码图案。4. 门票下方提供一个“打印门票”按钮，点击后调用 window.print() 触发浏览器的打印功能。\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有编写完代码后务必对全部代码执行测试并确保测试通过的要求调整为代码写好以后，禁止运行任何测试；第二，确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n \n # 定义前端项目和相关配置、文档等常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py', '.json', '.md'}\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 以二进制模式读取文件内容\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用 utf-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码失败，说明该文件不是合法的UTF-8编码\n return False\n except Exception:\n # 其它读取异常亦视为验证未通过\n return False\n \n # 只要找到了至少一个代码文件并且没有发生解码错误，即视为满足要求\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义通常的前后端代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rb', '.php'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅校验代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查最首行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 若文件为空、读取失败或者不能使用 UTF-8 进行解码，说明不满足要求\n return False\n \n # 如果项目中没有任何代码文件，通常也不能算满足生成代码并添加声明的要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见的代码文件扩展名\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less',\n '.vue', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \n '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.svg'\n }\n \n # 定义需要忽略的常见无关或依赖目录\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode'}\n \n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤需要忽略的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n # 按照要求使用 utf-8 编码读取\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果读取失败（例如非 UTF-8 编码），说明不满足编码要求或文件损坏\n return False\n \n # 按行分割并去除空行\n lines = content.splitlines()\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n if not non_empty_lines:\n return False\n \n # 检查最后 5 个非空行是否包含指定的注释字符串\n # 考虑到某些语言的多行注释可能使得该文本不处于绝对的最后一行（如最后一行是 '*/' 或是 '-->'）\n found_in_tail = False\n for line in non_empty_lines[-5:]:\n if target_text in line:\n found_in_tail = True\n break\n \n if not found_in_tail:\n return False\n \n # 如果项目中没有任何代码文件，则不符合预期的代码生成任务\n return code_files_found","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复的最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容前面的空白字符，然后判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否严格避免了使用\"我\"，\n 因为所有的\"我\"都应该被换成\"本小姐\"。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中任何目录的层深都不允许大于5层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import pathlib\n\n # 如果工作区不存在，说明未生成项目\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏目录，避免将 .git 等非业务代码目录算进层深判断中\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 如果是根目录自身（相对路径为 '.'），跳过深度判断\n if rel_path == '.':\n continue\n \n # 使用 pathlib 拆分相对路径获取目录层级数量\n parts = pathlib.Path(rel_path).parts\n depth = len(parts)\n \n # 判断目录的层深是否大于 5 层\n if depth > 5:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"代码写好以后，禁止运行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径为空或不存在，说明未生成任何文件，默认符合要求\n if not workspace_path or not os.path.isdir(workspace_path):\n return True\n\n # 推断前端项目相关常见的代码与配置文件的扩展名\n code_extensions = {\n '.html', '.css', '.scss', '.less',\n '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.json', '.md', '.yaml', '.yml', '.py', '.sh'\n }\n\n # 需要排除的无关目录\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', '.idea', '.vscode'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 列表，过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取文件，忽略可能的编码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 检查文件内容中是否包含 \"TODO\" 关键词\n if 'TODO' in content:\n return False\n except Exception:\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def get_names(node):\n names = []\n if isinstance(node, ast.Name):\n names.append(node.id)\n elif isinstance(node, (ast.Tuple, ast.List)):\n for elt in node.elts:\n names.extend(get_names(elt))\n return names\n\n # 匹配前端 JS/TS 中的变量声明：let/const/var 后的第一个标识符\n js_pattern = re.compile(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 解析普通赋值：a = 1, a, b = 1, 2\n if isinstance(node, ast.Assign):\n for target in node.targets:\n for name in get_names(target):\n if not (3 <= len(name) <= 5):\n return False\n # 解析类型注解赋值：a: int = 1\n elif isinstance(node, ast.AnnAssign):\n for name in get_names(node.target):\n if not (3 <= len(name) <= 5):\n return False\n # 解析 for 循环变量：for i in range(10):\n elif isinstance(node, ast.For):\n for name in get_names(node.target):\n if not (3 <= len(name) <= 5):\n return False\n # 解析列表推导式等推导结构变量：[x for x in list]\n elif isinstance(node, ast.comprehension):\n for name in get_names(node.target):\n if not (3 <= len(name) <= 5):\n return False\n # 解析函数定义中的参数变量\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n args = []\n if hasattr(node.args, 'posonlyargs'):\n args.extend(node.args.posonlyargs)\n if hasattr(node.args, 'args'):\n args.extend(node.args.args)\n if hasattr(node.args, 'kwonlyargs'):\n args.extend(node.args.kwonlyargs)\n \n for arg in args:\n if not (3 <= len(arg.arg) <= 5):\n return False\n if getattr(node.args, 'vararg', None):\n if not (3 <= len(node.args.vararg.arg) <= 5):\n return False\n if getattr(node.args, 'kwarg', None):\n if not (3 <= len(node.args.kwarg.arg) <= 5):\n return False\n except Exception:\n pass\n \n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\n js_content = content\n if ext in ['.html', '.htm', '.vue']:\n # 提取 script 标签中的代码\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n js_content = '\\n'.join(scripts)\n \n # 剔除代码中的单行和多行注释，以防正则表达式误匹配到注释内书写的变量代码\n js_content = re.sub(r'//.*', '', js_content)\n js_content = re.sub(r'/\\*.*?\\*/', '', js_content, flags=re.DOTALL)\n \n matches = js_pattern.findall(js_content)\n for var_name in matches:\n if not (3 <= len(var_name) <= 5):\n return False\n \n return True","tags":["Style","Naming"]}]},{"round_id":5,"instruction":"前台流程基本闭环了，现在我们切换到顶部导航的“管理后台”Tab，来构建数据仪表盘。1. 后台顶部设计三个数据统计卡片：总销售额、已售票数、平均上座率。2. 在主体区域，请不依赖 Echarts 等第三方图表库，纯手工使用 SVG 或 HTML5 Canvas 绘制一个柱状图，展示过去7天的销售金额趋势。3. 柱状图需要有基础的交互：当鼠标悬停在某根柱子上时，上方浮现一个 Tooltip 显示具体的金额数值。4. 页面右侧添加一个“最新订单”的纵向滚动列表，用来展示几条模拟的订票记录。\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有确保生成的任何代码文件中都不包含\"TODO\"关键词的要求调整为你输出的所有代码文件都必须完全排除\"TODO\"和\"FIXME\"；第二，进行查找处理文件的任务时，请尽可能使用 Task 工具并配合 subagent_type=Explore 来执行查找处理文件的任务，以代替直接运行搜索或文件操作命令。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义常见的前后端及文档代码文件的后缀白名单\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.vue', '.py', '.json', \n '.md', '.svg'\n }\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的无关目录或第三方依赖，防止误判\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'dist', 'build'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 模式读取文件全部内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError 说明文件并不是以 utf-8 进行编码\n return False\n except Exception:\n # 其他错误 (如权限问题等) 忽略处理，尽量避免异常中断验证\n pass\n \n # 如果工作区中有代码文件并且全都未触发解码错误，证明文件均符合 UTF-8 编码\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的前后端代码文件后缀\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.svelte'}\n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found += 1\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含要求的声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于任何原因(如编码错误、权限等)读取失败，视为不满足\n return False\n \n # 如果项目中没有任何代码文件，则不符合生成代码的基础预期\n if code_files_found == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常见的代码文件后缀集合\n code_extensions = {\n '.html', '.css', '.scss', '.less', \n '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.svelte', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \n '.go', '.rs', '.php', '.rb', '.sh'\n }\n \n # 需要忽略的常见非源码目录\n ignored_dirs = {'node_modules', '.git', '__pycache__', '.venv', 'venv'}\n \n code_files_found = 0\n for root, dirs, files in os.walk(workspace_path):\n # 如果当前路径中存在需要忽略的目录名，则跳过\n if any(ignored in root.split(os.sep) for ignored in ignored_dirs):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理判断为代码文件的类型\n if ext in code_extensions:\n code_files_found += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果读取失败（如编码问题等），视作不符合要求\n return False\n \n # 过滤并提取所有非空行\n non_empty_lines = [line.strip() for line in lines if line.strip() != '']\n \n if not non_empty_lines:\n return False\n \n # 检查代码文件的结尾部分（由于多行注释可能会占多行，检查最后5个非空行）\n found = False\n for line in non_empty_lines[-5:]:\n if target_statement in line:\n found = True\n break\n \n # 如果当前文件没有声明，则直接返回 False\n if not found:\n return False\n\n # 若检测到了代码文件且均包含该注释，则满足要求\n return code_files_found > 0","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容最前面的空白字符后，判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证在助手的回复中，是否将自称\"我\"的地方严格换成了\"本小姐\"。\n \"\"\"\n # 剔除常见的包含“我”但非单数自称的词汇，以防止误判\n # 例如：“我们”作为复数通常不强制替换为“本小姐们”，“自我”、“你我”为固定词组\n clean_response = response.replace(\"我们\", \"\").replace(\"自我\", \"\").replace(\"你我\", \"\")\n \n # 剔除上述词汇后，如果回复中依然包含“我”字，说明有自称的情况未被替换\n if \"我\" in clean_response:\n return False\n \n # 如果不存在“我”，则视为满足“凡是需要自称'我'的地方均已替换”的条件\n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中任何目录的层深都不允许大于5层","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 如果工作区不存在或者不是目录，则认为没有触发超过深度的限制\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n workspace = Path(workspace_path).resolve()\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制系统自带目录的影响\n if '.git' in dirs:\n dirs.remove('.git')\n \n root_path = Path(root).resolve()\n \n try:\n # 计算当前目录相对于根目录的路径\n rel_path = root_path.relative_to(workspace)\n except ValueError:\n continue\n \n # rel_path.parts 是路径中的每一层。\n # 例如对于相对路径 \"dir1/dir2\" -> parts 为 ('dir1', 'dir2')，深度为 2\n # \".\" 的 parts 为空，深度为 0\n depth = len(rel_path.parts)\n \n if depth > 5:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"代码写好以后，禁止运行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你输出的所有代码文件都必须完全排除\"TODO\"和\"FIXME\"","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 检查助手回复的 Markdown 代码块中是否包含 TODO 或 FIXME\n code_blocks = re.findall(r'```(?:.*?)\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n if \"TODO\" in block or \"FIXME\" in block:\n return False\n\n # 常见代码/文本文件的后缀名集合\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', \n '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift', \n '.kt', '.md', '.txt', '.json', '.xml', '.yaml', '.yml', \n '.sh', '.bat', '.sql'\n }\n \n # 检查工作目录下的生成的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅筛选代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 如果代码文件中包含 TODO 或 FIXME，即判定为未满足要求\n if \"TODO\" in content or \"FIXME\" in content:\n return False\n except Exception:\n continue\n \n # 所有文件与代码块均未发现违规关键词，则满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n found_any_var = False\n \n # 针对前端JS/TS代码的正则表达式\n js_var_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n js_func_pattern = re.compile(r'\\bfunction\\s+(?:[a-zA-Z_$][a-zA-Z0-9_$]*\\s*)?\\(([^)]*)\\)')\n js_arg_pattern = re.compile(r'^\\s*(?:\\.\\.\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)')\n arrow_func_pattern = re.compile(r'\\(([^)]*)\\)\\s*=>')\n single_arg_arrow_pattern = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>')\n\n def clean_js_code(text: str, ext: str) -> str:\n # 如果是 HTML 或 Vue 文件，只提取 <script> 标签中的内容\n if ext in {'.html', '.htm', '.vue'}:\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', text, re.IGNORECASE)\n text = '\\n'.join(scripts)\n # 去除注释和字符串以防正则在其中发生误匹配\n text = re.sub(r'/\\*[\\s\\S]*?\\*/', '', text)\n text = re.sub(r'//.*', '', text)\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', text)\n text = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', text)\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', text)\n return text\n\n target_exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非源码依赖及隐藏目录\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'venv', '__pycache__', 'dist', 'build'}]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in target_exts:\n continue\n\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n var_name = node.id\n found_any_var = True\n if not (3 <= len(var_name) <= 5):\n return False\n elif isinstance(node, ast.arg):\n var_name = node.arg\n found_any_var = True\n if not (3 <= len(var_name) <= 5):\n return False\n except Exception:\n pass\n else:\n cleaned = clean_js_code(content, ext)\n \n # 检查通过 var / let / const 声明的基础变量\n for match in js_var_pattern.finditer(cleaned):\n var_name = match.group(1)\n found_any_var = True\n if not (3 <= len(var_name) <= 5):\n return False\n \n # 检查普通函数的参数\n for match in js_func_pattern.finditer(cleaned):\n args_str = match.group(1)\n if args_str and args_str.strip():\n args = args_str.split(',')\n for arg in args:\n arg_match = js_arg_pattern.search(arg)\n if arg_match:\n var_name = arg_match.group(1)\n found_any_var = True\n if not (3 <= len(var_name) <= 5):\n return False\n \n # 检查带括号的箭头函数参数\n for match in arrow_func_pattern.finditer(cleaned):\n args_str = match.group(1)\n if args_str and args_str.strip():\n args = args_str.split(',')\n for arg in args:\n arg_match = js_arg_pattern.search(arg)\n if arg_match:\n var_name = arg_match.group(1)\n found_any_var = True\n if not (3 <= len(var_name) <= 5):\n return False\n \n # 检查不带括号的单参数箭头函数\n for match in single_arg_arrow_pattern.finditer(cleaned):\n var_name = match.group(1)\n found_any_var = True\n if not (3 <= len(var_name) <= 5):\n return False\n\n return found_any_var","tags":["Style","Naming"]},{"constraint":"进行查找处理文件的任务时，请尽可能使用 Task 工具并配合 subagent_type=Explore 来执行查找处理文件的任务，以代替直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":6,"instruction":"最后一步，让我们为管理后台增加对座位的深度管控能力。1. 在仪表盘页面增加一个“场次座位管理”模块，将之前前台做的 SVG 座位图复用到这里，但作为管理员视角展示。2. 管理员可以点击任意“已售”或“可选”的座位，点击后弹出一个小型悬浮菜单，允许将其强制更改为“设备维修”状态。3. 变成“设备维修”状态的座位在 SVG 图中需要用带有红色叉号（X）的特殊样式表示。4. 在右侧的“最新订单”列表中，为每条订单增加一个“退票”按钮，点击退票后，不仅订单要从列表中移除，顶部的“总销售额”和“已售票数”统计卡片还要展示一个数字动态递减的动画效果。\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有你输出的所有代码文件都必须完全排除\"TODO\"和\"FIXME\"的要求调整为你输出的所有代码文件都必须完全排除\"TODO\"、\"FIXME\"和\"DEBUG\"；第二，你所生成的所有代码文件中，任意一行都不能超出 100 个字符的长度。","instruction_following_checklist":[{"constraint":"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前后端代码、脚本及文本文件扩展名\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', \n '.scss', '.less', '.json', '.md', '.java', '.c', '.cpp', '.h', \n '.go', '.rs', '.php', '.rb', '.sh', '.xml', '.yaml', '.yml'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码\n # 如果文件被编码为 GBK 等其他编码且包含非 ASCII 字符，这里将会引发 UnicodeDecodeError\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 解码失败，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能的文件读取系统错误\n pass\n \n # 所有被检查的代码文件均可无异常地以 UTF-8 解码，说明满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名（结合对话中的前端原型开发任务，涵盖前端及常见通用代码）\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.py', '.java', '.c', '.cpp', '.h', '.go', \n '.php', '.rb', '.sh', '.svg'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n code_file_count = 0\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录及常见依赖/构建产物目录，避免检查到无关文件\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'dist', 'build', 'out')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_file_count += 1\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 仅读取最首行\n first_line = f.readline()\n # 判断首行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件读取发生异常（如由于非文本内容导致等），视为不满足要求\n return False\n\n # 若工作区内没有找到任何代码文件，说明未有效生成代码项目，返回 False\n return code_file_count > 0","tags":["Content","Format"]},{"constraint":"在你生成的每个代码文件中，请将所有注释全用英文呈现","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 前端开发和一般项目中常见的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.sh'}\n \n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found += 1\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 无法按 utf-8 读取的情况，判断为未满足标准代码格式或意外文件错误\n return False\n \n # 剔除内容中的空行，以便准确判断文件“结尾处”的内容\n lines = [line.strip() for line in content.splitlines() if line.strip()]\n if not lines:\n return False\n \n found_in_new_line = False\n # 在代码文件的最后 5 个非空行中进行检索，验证\"结尾处\"要求\n for line in reversed(lines[-5:]):\n if target_str in line:\n # 验证\"使用新的一行书写注释\"的要求\n # 将目标字符串从该行中剔除，若该行是单纯的注释行，剩下的应该只有注释符号（如 //, <!-- -->, # 等）\n # 因此剩余内容中不应该再包含任何英文字母或数字\n remaining = line.replace(target_str, \"\").strip()\n if not re.search(r'[a-zA-Z0-9]', remaining):\n found_in_new_line = True\n break\n \n # 只要有任意一个代码文件不满足上述要求，则整体判负\n if not found_in_new_line:\n return False\n \n # 确保项目中确实存在被创建的代码文件，如果未发现任何代码文件则返回False\n return code_files_found > 0","tags":["Content","Format"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断AI助手的回复最开头是否带有\"喵～\"语气词。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，代表是否满足要求。\n \"\"\"\n if not response:\n return False\n \n # 去除回复可能存在的前导空白字符后，检查是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在整个对话的回复中，凡是需要自称\"我\"的地方，都必须严格换成\"本小姐\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复中是否严格将'我'替换为'本小姐'。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n # 如果回复内容中存在\"我\"字，说明并未严格替换，返回 False\n if \"我\" in response:\n return False\n \n # 否则满足要求（可能已替换为\"本小姐\"，或本轮回复中本身无需自称）\n return True","tags":["Content","Persona"]},{"constraint":"生成的代码项目中任何目录的层深都不允许大于5层","validation_code":"def check_directory_depth(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录自身的相对路径为 '.'，深度为 0\n if rel_path == '.':\n depth = 0\n else:\n # 统一路径分隔符，兼容不同操作系统\n normalized_path = rel_path.replace('\\\\', '/')\n # 分割后列表的长度即为层深\n depth = len(normalized_path.split('/'))\n \n # 如果层深大于 5，即视为不满足要求\n if depth > 5:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"代码写好以后，禁止运行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你输出的所有代码文件都必须完全排除\"TODO\"、\"FIXME\"和\"DEBUG\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义需要被禁用的目标关键词\n forbidden_words = [\"TODO\", \"FIXME\", \"DEBUG\"]\n \n # 定义常见的前端与后端代码文件的扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', \n '.html', '.htm', \n '.css', '.scss', '.less', \n '.vue', '.py', '.json'\n }\n \n # 如果工作区路径不存在，说明并没有生成代码，也就无从验证\n if not os.path.exists(workspace_path):\n return False\n \n # 遍历当前工作区的所有目录和文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为需要检查的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查是否包含任何禁用的关键词\n for word in forbidden_words:\n if word in content:\n return False\n except Exception:\n # 忽略不可读文件或编码无法解析的文件\n continue\n \n # 如果遍历完毕且没有发现任何禁用关键词，则代表满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）","validation_code":"def check_variable_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n \n # 定义各类代码文件的后缀\n py_extensions = {'.py'}\n js_ts_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}\n css_extensions = {'.css'}\n \n # 匹配 JS/TS 中变量声明的基础正则 (let, const, var 后面的标识符)\n var_pattern = re.compile(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 检查 Python 文件\n if ext in py_extensions:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 检查赋值变量\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if not (3 <= len(node.id) <= 5):\n return False\n # 检查函数参数\n elif isinstance(node, ast.arg):\n if not (3 <= len(node.arg) <= 5):\n return False\n except Exception:\n pass\n \n # 检查 JS/TS/HTML/VUE 等前端文件\n elif ext in js_ts_extensions:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除文件中的注释和字符串内容，防止正则误匹配\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n # 1. 检查所有的普通变量声明\n for v in var_pattern.findall(content):\n if not (3 <= len(v) <= 5):\n return False\n \n # 2. 检查普通函数(function)中定义的参数\n func_params = re.findall(r'\\bfunction\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', content)\n for params_str in func_params:\n params = [p.strip() for p in params_str.split(',')]\n for p in params:\n if not p:\n continue\n # 提取开头的标识符(防出现带默认值的复杂情况)\n p_match = re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\n if p_match:\n v_name = p_match.group(1)\n if not (3 <= len(v_name) <= 5):\n return False\n except Exception:\n pass\n \n # 检查 CSS 文件中的原生变量\n elif ext in css_extensions:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 匹配 --xxx: 的 CSS 变量格式\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', content)\n for v in css_vars:\n if not (3 <= len(v) <= 5):\n return False\n except Exception:\n pass\n \n # 如果遍历完毕且没有发现任何长度不在[3, 5]范围内的变量名，则满足要求\n return True","tags":["Style","Naming"]},{"constraint":"进行查找处理文件的任务时，请尽可能使用 Task 工具并配合 subagent_type=Explore 来执行查找处理文件的任务，以代替直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的所有代码文件中，任意一行都不能超出 100 个字符的长度","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 基于用户多轮指令推断可能的前端相关代码文件类型\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.svg', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏文件夹和常见的依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的换行符和回车符后，使用 len 函数检查字符数\n line_content = line.rstrip('\\r\\n')\n if len(line_content) > 100:\n return False\n except Exception:\n # 如果遇到无法读取或非正常编码的文件，安全跳过\n continue\n \n return True","tags":["Quantity","Range"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"你好！我们需要开发一个包含前后端的活动预订系统前端原型。第一步，请帮我搭建项目的整体骨架。需求如下：1. 创建一个包含顶部导航栏的基础布局，导航栏左侧是Logo，右侧包含两个Tab：'活动预订' 和 '管理后台'。2. 实现简单的路由或视图切换逻辑：点击 Tab 可以切换下方主内容区的视图。3. 视图当前只需展示基础占位内容，比如在'活动预订'页面放一个标题和一个'立即探索'按钮，'管理后台'放一个欢迎标题。请给出结构清晰、带有现代感基础 CSS 样式的代码。另外请注意，后续所有轮次都请始终遵循以下要求：第一，生成的代码项目中，任何目录的层深都不允许大于 2 层；第二，必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 常见的代码文件和配置/说明文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.py', '.json', '.md', '.txt', '.sh'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用utf-8进行解码\\n # 如果文件不是utf-8编码，则会抛出UnicodeDecodeError\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦发现无法用UTF-8解码的文件，说明不满足要求\\n return False\\n except Exception:\\n # 忽略其他无关的异常（例如权限问题导致的读取失败）\\n pass\\n \\n # 所有代码文件都能够正常被utf-8解码\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端和常见后端的代码文件扩展名\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.h', '.go', '.php', '.rb', '.sh'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查代码文件的最首行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取文件或编码不是 utf-8，则视作不满足要求\\n return False\\n\\n # 若至少存在一个代码文件，且所有代码文件的最首行均满足条件，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义常见的主流编程语言与前端代码后缀名\\n code_exts = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.cpp', '.c', '.cc', '.h', '.hpp', \\n '.cs', '.go', '.php', '.rb', '.swift', '.kt', '.sh', '.bash', '.sql'\\n }\\n \\n # 需要跳过的第三方包或环境目录，防止误判\\n ignore_dirs = {'.git', 'node_modules', 'venv', 'env', '__pycache__', 'dist', 'build'}\\n \\n code_files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read().strip()\\n except Exception:\\n # 若因非UTF-8编码或其他读取异常，不符合文件编码要求，直接返回 False\\n return False\\n \\n # 如果文件是空的，显然没包含必要的注释\\n if not content:\\n return False\\n \\n # 检查是否包含指定的声明\\n if target_statement not in content:\\n return False\\n \\n # 获取该声明最后出现的位置之后的所有尾部文本\\n last_idx = content.rfind(target_statement)\\n tail = content[last_idx + len(target_statement):].strip()\\n \\n # 检查尾部是否仅包含合法的注释关闭字符及空白字符\\n # 例如： HTML: --> , CSS/JS: */ , Python: \\\"\\\"\\\" 或 空等\\n allowed_chars = set(\\\"/*->\\\\\\\"'%# \\\\t\\\\r\\\\n\\\")\\n if not all(c in allowed_chars for c in tail):\\n return False\\n \\n code_files_checked += 1\\n \\n # 如果工作区中没找到任何代码文件，说明未生成相关代码\\n if code_files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复的最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n # 移除开头可能的空白字符，判断是否以\\\"喵～\\\"开头\\n if not response:\\n return False\\n \\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证 AI 助手的回复中是否将自称的“我”全部替换为“本小姐”。\\n 在当前代码交流的上下文中，只要回复文本中仍然包含“我”字，\\n 即可认为其没有严格执行替换。\\n \\\"\\\"\\\"\\n # 只要存在“我”字，就说明替换未完全满足要求\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中，任何目录的层深都不允许大于 2 层\",\n \"validation_code\": \"def check_directory_depth(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，则无法判断或视为未生成项目\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录（如 .git 等），防止由于环境带来的干扰导致误判\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 计算当前目录相对于 workspace_path 的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 如果是根目录本身，层级深度为 0\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 根据系统的路径分隔符切分相对路径，列表的长度即为相对于根目录的层深\\n depth = len(rel_path.split(os.sep))\\n \\n # 若发现任何目录层深大于 2，则不满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"太棒了，框架已经起来了。现在我们来细化“活动预订”页面。我们需要在这里实现一个日历与活动联动的组件：1. 不要使用原生的 input date，请用纯代码渲染一个完整的当前月份的日历视图。2. 在日历中挑选几个日期，用小圆点在日期下方标记为“有活动”。3. 当用户点击“有活动”的日期时，日历右侧（或下方）滑出一个活动列表，展示2-3个虚构的活动卡片（包含活动名称、时间段、票价起步）。如果点击没有活动的日期，则显示“当日暂无活动安排”的空状态提示。另外请注意，后续所有轮次都请遵循以下更新后的要求：第一，原有代码项目目录层深不大于2层的要求调整为生成的代码项目中任何目录的层深都不允许大于3层；第二，为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前端、后端及文档等代码文件类型后缀\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.sh', \\n '.md', '.json', '.xml', '.txt'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件类型的编码\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦发现无法用UTF-8解码的代码文件，即判定为不满足要求\\n return False\\n except Exception:\\n # 忽略诸如权限不足导致的读取错误，只针对编码异常做判定\\n continue\\n\\n # 若所有筛选出的代码文件都能以UTF-8正常读取，则满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"import os\\n\\ndef check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n # 定义常见的前后端代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.h'}\\n # 排除可能包含大量非AI直接生成代码的目录\\n exclude_dirs = {'node_modules', '.git', '__pycache__', 'dist', 'build'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤需要排除的目录，避免深度遍历第三方库和构建产物\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 使用 utf-8 编码打开文件并读取最首行\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查指定的编码声明是否出现在第一行\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法打开或者编码错误，无法验证其是否满足要求，视为不通过\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 常见的代码文件后缀集合，排除 Markdown 等非代码类文档以及不能随意写注释的 JSON 等\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', \\n '.html', '.css', '.scss', '.less', \\n '.vue', '.svelte', '.java', '.c', \\n '.cpp', '.h', '.cs', '.go', '.rb', '.php'\\n }\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 过滤掉完全空白的行\\n lines = [line for line in lines if line.strip()]\\n \\n if not lines:\\n return False\\n \\n # 检查结尾处（最后 5 个非空行）是否包含目标声明\\n # 使用多行检查以兼容将多行注释符单独成行的情况，如：\\n # <!--\\n # AI-Assisted: ...\\n # -->\\n found = False\\n for line in lines[-5:]:\\n if target_statement in line:\\n found = True\\n break\\n \\n if not found:\\n return False\\n \\n except Exception:\\n # 若文件无法读取或抛出异常，算作不符合要求\\n return False\\n \\n # 如果找到了代码文件，并且所有代码文件都符合要求，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的最后一轮回复的最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n # 移除最开头的空白字符后，判断是否以\\\"喵～\\\"开头\\n if response.lstrip().startswith(\\\"喵～\\\"):\\n return True\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否还存在“我”字\\n # 如果存在，说明助手存在自称未能严格替换为“本小姐”的违规情况\\n if \\\"我\\\" in response:\\n return False\\n \\n # 不存在“我”，视为满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中任何目录的层深都不允许大于3层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果项目路径不存在，自然不存在大于3层的目录\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for dirpath, dirnames, filenames in os.walk(workspace_path):\\n # 过滤掉隐藏目录(如 .git, .vscode)以及 __pycache__ 缓存等非AI直接生成的目录\\n dirnames[:] = [d for d in dirnames if not d.startswith('.') and d != '__pycache__']\\n \\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(dirpath, workspace_path)\\n \\n if rel_path == '.':\\n depth = 0\\n else:\\n # 统一路径分隔符并计算层级深度\\n normalized_path = rel_path.replace('\\\\\\\\', '/')\\n depth = len(normalized_path.split('/'))\\n \\n # 如果当前目录层深大于3，则不满足要求\\n if depth > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"日历组件做得很漂亮！接下来是核心挑战，我们需要在用户点击某个活动卡片后，进入选座环节。1. 隐藏日历，在页面上渲染一个 SVG 构成的座位分布图，模拟一个小型剧院，顶部要有“舞台”标识。2. 座位需要分为前排的 VIP 区（金色或橙色）和后排的普通区（蓝色）。3. 用 SVG 的圆形或圆角矩形代表座位，并且座位需要支持三种状态：'可选'（默认颜色）、'已售'（灰色且不可点击）、'已选'（绿色）。4. 实现点击交互：点击'可选'座位使其状态变为'已选'，再次点击取消选中。另外请注意，后续所有轮次都请遵循以下更新后的要求：第一，原有代码项目目录层深不大于3层的要求调整为生成的代码项目中任何目录的层深都不允许大于4层；第二，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件后缀集合（包含了前端原型系统最常用的以及部分常见的后端文件格式）\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.json', '.svg', '.py', '.java', '.c', '.cpp', '.h', \\n '.cs', '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n \\n has_code_file = False\\n \\n # 遍历代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅校验代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制形式读取文件\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试以 UTF-8 格式解码\\n raw_data.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 如果发生 UnicodeDecodeError 异常，说明文件不是 UTF-8 编码\\n return False\\n except Exception:\\n # 如果文件读取发生其他异常，视为未能满足验证，返回 False\\n return False\\n \\n # 如果项目中没有生成任何代码文件，则无法被视为满足了“创建代码文件并用UTF-8编码”的前提\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义需要检查的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.cpp', '.c', '.h', '.go', '.rb', '.php', '.sh'\\n }\\n \\n # 定义需要忽略的目录（依赖包、编译产物、隐藏目录等）\\n ignore_dirs = {\\n '.git', 'node_modules', '__pycache__', 'dist', 'build', \\n '.idea', '.vscode', 'venv', 'env'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录和以点开头的隐藏目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs and not d.startswith('.')]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含目标编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法正常读取，视作未满足要求\\n return False\\n\\n # 必须确保至少检查了一个代码文件，防止空项目误判为True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 前端开发常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\\n # 忽略第三方库等与 AI 生成源代码无关的目录\\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'venv', 'env'}\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果不是 UTF-8 编码或无法读取，违背了无一例外用 UTF-8 的系统提示要求\\n return False\\n \\n # 如果文件为空，必然没有包含目标声明\\n if not lines:\\n return False\\n \\n # 从尾部开始遍历，找到第一个非空行（真实的最后一行）\\n last_line = \\\"\\\"\\n for line in reversed(lines):\\n if line.strip():\\n last_line = line.strip()\\n break\\n \\n # 检查指定的声明文本是否在最后一行的注释中\\n if target_statement not in last_line:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容最开头的空白字符（如空格、换行符等）\\n cleaned_response = response.lstrip()\\n \\n # 检查清理后的回复是否以\\\"喵～\\\"开头\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否满足：将所有自称“我”的地方严格换成“本小姐”。\\n 在日常技术回复语境下，若出现“我”字，极大概率是自称遗漏了替换。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中任何目录的层深都不允许大于4层\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef check_directory_depth(response: str, workspace_path: str) -> bool:\\n workspace = Path(workspace_path)\\n \\n # 如果工作区目录不存在，可以直接认为没有违反限制\\n if not workspace.exists() or not workspace.is_dir():\\n return True\\n \\n # 定义需要忽略的、非AI直接生成代码的常见依赖或隐藏目录\\n ignored_dirs = {\\n 'node_modules', '__pycache__', 'venv', 'env', \\n '.venv', '.env', 'dist', 'build', 'out'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 裁剪不需要遍历的目录（隐藏目录如 .git 以及常见的依赖包目录）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ignored_dirs]\\n \\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = Path(root).relative_to(workspace)\\n \\n # 计算目录深度：相对路径的组成部分数量\\n # 若 rel_path 为 '.'，则 parts 为空元组 ()，长度即深度为 0\\n depth = len(rel_path.parts)\\n \\n # 如果层深大于 4，则不满足要求\\n if depth > 4:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"SVG选座的交互很流畅。现在来完善选座的业务逻辑和购物车功能：1. 在座位图右侧增加一个固定的“已选座位”面板。2. 当用户在 SVG 图中选中座位时，面板要实时更新，列出选中的座位排号（如 A排3座）及对应的单价（VIP区和普通区价格需有区分），并在底部计算总价。3. 增加一条限制规则：单次预订最多只能选择 4 个座位。当用户尝试选择第 5 个时，在页面中心弹出一条醒目的错误 Toast 提示并阻止选中。4. 面板最下方增加一个“确认选座”按钮，点击后弹出一个模态框，要求填写联系人姓名和手机号。\\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有代码项目目录层深不大于4层的要求调整为生成的代码项目中任何目录的层深都不允许大于5层；第二，确保你生成的任何代码文件中都不包含\\\"TODO\\\"关键词。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n # 若工作区路径不存在，则无法产生不满足条件的文件，视为空集满足\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 常见的前端、后端及文档的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.css', '.scss', '.less', '.md', '.json', '.vue', \\n '.sh', '.txt', '.svg', '.xml'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试用 utf-8 编码读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明并非合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略权限等其他导致文件无法读取的异常\\n pass\\n \\n # 所有代码文件均成功使用 UTF-8 读取\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 识别常见的代码文件扩展名\\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.svg'}\\n \\n if not os.path.exists(workspace_path):\\n # 如果项目路径不存在，视作不满足要求\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 判断指定的编码声明是否出现在代码的最首行\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果打开或读取文件失败（例如编码不是UTF-8引发解码异常等），认为不满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义前端原型开发及其他常见代码文件后缀\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue',\\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift'\\n }\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏目录和常见的依赖环境目录，避免检查非AI生成的依赖文件\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__', 'venv', 'env']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于特殊原因无法读取（如非utf-8编码等），则跳过\\n continue\\n \\n # 过滤掉文件最末尾的空行\\n while lines and not lines[-1].strip():\\n lines.pop()\\n \\n # 如果代码文件为空，显然不满足包含注释的要求\\n if not lines:\\n return False\\n \\n # 提取最后3行。这样可以兼容注释闭合符号如 '*/' 或 '-->' 在最末尾的情况\\n last_few_lines = [line.strip() for line in lines[-3:]]\\n \\n # 判断最后几行中是否包含了特定的声明\\n statement_found = any(target_statement in line for line in last_few_lines)\\n \\n if not statement_found:\\n return False\\n \\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查AI助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，代表该要求是否被满足\\n \\\"\\\"\\\"\\n # 去除回复内容最前面的空白字符后，判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否将自称'我'都换成了'本小姐'。\\n 只要回复中不包含'我'字，即可认为满足要求。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中任何目录的层深都不允许大于5层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区目录不存在，则说明尚未生成任何内容，不违背层深要求\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n # 需要忽略的常见非 AI 直接生成的依赖、环境或构建目录，避免深层结构引发误判\\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'venv', '.venv', 'dist', 'build', 'coverage'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 列表，阻止 os.walk 进一步深入这些被忽略的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n # 获取当前遍历目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n if rel_path == '.':\\n depth = 0\\n else:\\n # 兼容不同操作系统的路径分隔符，分割以计算层深\\n parts = [p for p in rel_path.replace('\\\\\\\\', '/').split('/') if p]\\n depth = len(parts)\\n \\n # 若发现任何目录层深大于 5，则验证失败\\n if depth > 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts',\\n '.jsx', '.tsx', '.vue', '.html', '.css', '.scss', '.less', '.go',\\n '.rs', '.php', '.rb', '.swift', '.kt', '.sh', '.bat', '.ps1'\\n }\\n \\n # 若工作区路径不存在或不是目录，说明未生成文件，自然不包含 TODO\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历项目目录中的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理代码文件\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 检查是否包含 \\\"TODO\\\" 关键词\\n if \\\"TODO\\\" in content:\\n return False\\n except Exception:\\n # 如果由于权限等原因读取失败，则跳过\\n continue\\n \\n # 如果所有的代码文件都没有包含 \\\"TODO\\\"，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"下一步是出票环节。当用户在模态框填完信息并点击“提交订单”后：1. 隐藏选座和表单界面，展示一个“预订成功”的门票详情视图。2. 页面中需要展示一张设计精美的电子门票卡片（带有一点阴影和圆角等拟物感），卡片上展示活动名称、时间、用户姓名和所选座位号。3. 在门票右侧或下半部分，使用前端代码（可以用一组排列紧密的黑色 SVG 小方块或 div）生成一个模拟的二维码图案。4. 门票下方提供一个“打印门票”按钮，点击后调用 window.print() 触发浏览器的打印功能。\\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有编写完代码后务必对全部代码执行测试并确保测试通过的要求调整为代码写好以后，禁止运行任何测试；第二，确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n \\n # 定义前端项目和相关配置、文档等常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py', '.json', '.md'}\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制模式读取文件内容\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用 utf-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码失败，说明该文件不是合法的UTF-8编码\\n return False\\n except Exception:\\n # 其它读取异常亦视为验证未通过\\n return False\\n \\n # 只要找到了至少一个代码文件并且没有发生解码错误，即视为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义通常的前后端代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rb', '.php'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅校验代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查最首行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 若文件为空、读取失败或者不能使用 UTF-8 进行解码，说明不满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，通常也不能算满足生成代码并添加声明的要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less',\\n '.vue', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \\n '.rs', '.php', '.rb', '.sh', '.swift', '.kt', '.svg'\\n }\\n \\n # 定义需要忽略的常见无关或依赖目录\\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode'}\\n \\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤需要忽略的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 按照要求使用 utf-8 编码读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果读取失败（例如非 UTF-8 编码），说明不满足编码要求或文件损坏\\n return False\\n \\n # 按行分割并去除空行\\n lines = content.splitlines()\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n if not non_empty_lines:\\n return False\\n \\n # 检查最后 5 个非空行是否包含指定的注释字符串\\n # 考虑到某些语言的多行注释可能使得该文本不处于绝对的最后一行（如最后一行是 '*/' 或是 '-->'）\\n found_in_tail = False\\n for line in non_empty_lines[-5:]:\\n if target_text in line:\\n found_in_tail = True\\n break\\n \\n if not found_in_tail:\\n return False\\n \\n # 如果项目中没有任何代码文件，则不符合预期的代码生成任务\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复的最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容前面的空白字符，然后判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否严格避免了使用\\\"我\\\"，\\n 因为所有的\\\"我\\\"都应该被换成\\\"本小姐\\\"。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中任何目录的层深都不允许大于5层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import pathlib\\n\\n # 如果工作区不存在，说明未生成项目\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏目录，避免将 .git 等非业务代码目录算进层深判断中\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 如果是根目录自身（相对路径为 '.'），跳过深度判断\\n if rel_path == '.':\\n continue\\n \\n # 使用 pathlib 拆分相对路径获取目录层级数量\\n parts = pathlib.Path(rel_path).parts\\n depth = len(parts)\\n \\n # 判断目录的层深是否大于 5 层\\n if depth > 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"代码写好以后，禁止运行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径为空或不存在，说明未生成任何文件，默认符合要求\\n if not workspace_path or not os.path.isdir(workspace_path):\\n return True\\n\\n # 推断前端项目相关常见的代码与配置文件的扩展名\\n code_extensions = {\\n '.html', '.css', '.scss', '.less',\\n '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.json', '.md', '.yaml', '.yml', '.py', '.sh'\\n }\\n\\n # 需要排除的无关目录\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', '.idea', '.vscode'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 列表，过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取文件，忽略可能的编码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 检查文件内容中是否包含 \\\"TODO\\\" 关键词\\n if 'TODO' in content:\\n return False\\n except Exception:\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def get_names(node):\\n names = []\\n if isinstance(node, ast.Name):\\n names.append(node.id)\\n elif isinstance(node, (ast.Tuple, ast.List)):\\n for elt in node.elts:\\n names.extend(get_names(elt))\\n return names\\n\\n # 匹配前端 JS/TS 中的变量声明：let/const/var 后的第一个标识符\\n js_pattern = re.compile(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 解析普通赋值：a = 1, a, b = 1, 2\\n if isinstance(node, ast.Assign):\\n for target in node.targets:\\n for name in get_names(target):\\n if not (3 <= len(name) <= 5):\\n return False\\n # 解析类型注解赋值：a: int = 1\\n elif isinstance(node, ast.AnnAssign):\\n for name in get_names(node.target):\\n if not (3 <= len(name) <= 5):\\n return False\\n # 解析 for 循环变量：for i in range(10):\\n elif isinstance(node, ast.For):\\n for name in get_names(node.target):\\n if not (3 <= len(name) <= 5):\\n return False\\n # 解析列表推导式等推导结构变量：[x for x in list]\\n elif isinstance(node, ast.comprehension):\\n for name in get_names(node.target):\\n if not (3 <= len(name) <= 5):\\n return False\\n # 解析函数定义中的参数变量\\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n args = []\\n if hasattr(node.args, 'posonlyargs'):\\n args.extend(node.args.posonlyargs)\\n if hasattr(node.args, 'args'):\\n args.extend(node.args.args)\\n if hasattr(node.args, 'kwonlyargs'):\\n args.extend(node.args.kwonlyargs)\\n \\n for arg in args:\\n if not (3 <= len(arg.arg) <= 5):\\n return False\\n if getattr(node.args, 'vararg', None):\\n if not (3 <= len(node.args.vararg.arg) <= 5):\\n return False\\n if getattr(node.args, 'kwarg', None):\\n if not (3 <= len(node.args.kwarg.arg) <= 5):\\n return False\\n except Exception:\\n pass\\n \\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\\n js_content = content\\n if ext in ['.html', '.htm', '.vue']:\\n # 提取 script 标签中的代码\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n js_content = '\\\\n'.join(scripts)\\n \\n # 剔除代码中的单行和多行注释，以防正则表达式误匹配到注释内书写的变量代码\\n js_content = re.sub(r'//.*', '', js_content)\\n js_content = re.sub(r'/\\\\*.*?\\\\*/', '', js_content, flags=re.DOTALL)\\n \\n matches = js_pattern.findall(js_content)\\n for var_name in matches:\\n if not (3 <= len(var_name) <= 5):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"前台流程基本闭环了，现在我们切换到顶部导航的“管理后台”Tab，来构建数据仪表盘。1. 后台顶部设计三个数据统计卡片：总销售额、已售票数、平均上座率。2. 在主体区域，请不依赖 Echarts 等第三方图表库，纯手工使用 SVG 或 HTML5 Canvas 绘制一个柱状图，展示过去7天的销售金额趋势。3. 柱状图需要有基础的交互：当鼠标悬停在某根柱子上时，上方浮现一个 Tooltip 显示具体的金额数值。4. 页面右侧添加一个“最新订单”的纵向滚动列表，用来展示几条模拟的订票记录。\\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有确保生成的任何代码文件中都不包含\\\"TODO\\\"关键词的要求调整为你输出的所有代码文件都必须完全排除\\\"TODO\\\"和\\\"FIXME\\\"；第二，进行查找处理文件的任务时，请尽可能使用 Task 工具并配合 subagent_type=Explore 来执行查找处理文件的任务，以代替直接运行搜索或文件操作命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义常见的前后端及文档代码文件的后缀白名单\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.vue', '.py', '.json', \\n '.md', '.svg'\\n }\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的无关目录或第三方依赖，防止误判\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'dist', 'build'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 模式读取文件全部内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError 说明文件并不是以 utf-8 进行编码\\n return False\\n except Exception:\\n # 其他错误 (如权限问题等) 忽略处理，尽量避免异常中断验证\\n pass\\n \\n # 如果工作区中有代码文件并且全都未触发解码错误，证明文件均符合 UTF-8 编码\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的前后端代码文件后缀\\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.svelte'}\\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found += 1\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含要求的声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何原因(如编码错误、权限等)读取失败，视为不满足\\n return False\\n \\n # 如果项目中没有任何代码文件，则不符合生成代码的基础预期\\n if code_files_found == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常见的代码文件后缀集合\\n code_extensions = {\\n '.html', '.css', '.scss', '.less', \\n '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.svelte', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n \\n # 需要忽略的常见非源码目录\\n ignored_dirs = {'node_modules', '.git', '__pycache__', '.venv', 'venv'}\\n \\n code_files_found = 0\\n for root, dirs, files in os.walk(workspace_path):\\n # 如果当前路径中存在需要忽略的目录名，则跳过\\n if any(ignored in root.split(os.sep) for ignored in ignored_dirs):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理判断为代码文件的类型\\n if ext in code_extensions:\\n code_files_found += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果读取失败（如编码问题等），视作不符合要求\\n return False\\n \\n # 过滤并提取所有非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip() != '']\\n \\n if not non_empty_lines:\\n return False\\n \\n # 检查代码文件的结尾部分（由于多行注释可能会占多行，检查最后5个非空行）\\n found = False\\n for line in non_empty_lines[-5:]:\\n if target_statement in line:\\n found = True\\n break\\n \\n # 如果当前文件没有声明，则直接返回 False\\n if not found:\\n return False\\n\\n # 若检测到了代码文件且均包含该注释，则满足要求\\n return code_files_found > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容最前面的空白字符后，判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证在助手的回复中，是否将自称\\\"我\\\"的地方严格换成了\\\"本小姐\\\"。\\n \\\"\\\"\\\"\\n # 剔除常见的包含“我”但非单数自称的词汇，以防止误判\\n # 例如：“我们”作为复数通常不强制替换为“本小姐们”，“自我”、“你我”为固定词组\\n clean_response = response.replace(\\\"我们\\\", \\\"\\\").replace(\\\"自我\\\", \\\"\\\").replace(\\\"你我\\\", \\\"\\\")\\n \\n # 剔除上述词汇后，如果回复中依然包含“我”字，说明有自称的情况未被替换\\n if \\\"我\\\" in clean_response:\\n return False\\n \\n # 如果不存在“我”，则视为满足“凡是需要自称'我'的地方均已替换”的条件\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中任何目录的层深都不允许大于5层\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 如果工作区不存在或者不是目录，则认为没有触发超过深度的限制\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n workspace = Path(workspace_path).resolve()\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制系统自带目录的影响\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n root_path = Path(root).resolve()\\n \\n try:\\n # 计算当前目录相对于根目录的路径\\n rel_path = root_path.relative_to(workspace)\\n except ValueError:\\n continue\\n \\n # rel_path.parts 是路径中的每一层。\\n # 例如对于相对路径 \\\"dir1/dir2\\\" -> parts 为 ('dir1', 'dir2')，深度为 2\\n # \\\".\\\" 的 parts 为空，深度为 0\\n depth = len(rel_path.parts)\\n \\n if depth > 5:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"代码写好以后，禁止运行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件都必须完全排除\\\"TODO\\\"和\\\"FIXME\\\"\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 检查助手回复的 Markdown 代码块中是否包含 TODO 或 FIXME\\n code_blocks = re.findall(r'```(?:.*?)\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n if \\\"TODO\\\" in block or \\\"FIXME\\\" in block:\\n return False\\n\\n # 常见代码/文本文件的后缀名集合\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \\n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', \\n '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift', \\n '.kt', '.md', '.txt', '.json', '.xml', '.yaml', '.yml', \\n '.sh', '.bat', '.sql'\\n }\\n \\n # 检查工作目录下的生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅筛选代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 如果代码文件中包含 TODO 或 FIXME，即判定为未满足要求\\n if \\\"TODO\\\" in content or \\\"FIXME\\\" in content:\\n return False\\n except Exception:\\n continue\\n \\n # 所有文件与代码块均未发现违规关键词，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n found_any_var = False\\n \\n # 针对前端JS/TS代码的正则表达式\\n js_var_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n js_func_pattern = re.compile(r'\\\\bfunction\\\\s+(?:[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*)?\\\\(([^)]*)\\\\)')\\n js_arg_pattern = re.compile(r'^\\\\s*(?:\\\\.\\\\.\\\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)')\\n arrow_func_pattern = re.compile(r'\\\\(([^)]*)\\\\)\\\\s*=>')\\n single_arg_arrow_pattern = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>')\\n\\n def clean_js_code(text: str, ext: str) -> str:\\n # 如果是 HTML 或 Vue 文件，只提取 <script> 标签中的内容\\n if ext in {'.html', '.htm', '.vue'}:\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', text, re.IGNORECASE)\\n text = '\\\\n'.join(scripts)\\n # 去除注释和字符串以防正则在其中发生误匹配\\n text = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', text)\\n text = re.sub(r'//.*', '', text)\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', text)\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', text)\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', text)\\n return text\\n\\n target_exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非源码依赖及隐藏目录\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'venv', '__pycache__', 'dist', 'build'}]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in target_exts:\\n continue\\n\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n var_name = node.id\\n found_any_var = True\\n if not (3 <= len(var_name) <= 5):\\n return False\\n elif isinstance(node, ast.arg):\\n var_name = node.arg\\n found_any_var = True\\n if not (3 <= len(var_name) <= 5):\\n return False\\n except Exception:\\n pass\\n else:\\n cleaned = clean_js_code(content, ext)\\n \\n # 检查通过 var / let / const 声明的基础变量\\n for match in js_var_pattern.finditer(cleaned):\\n var_name = match.group(1)\\n found_any_var = True\\n if not (3 <= len(var_name) <= 5):\\n return False\\n \\n # 检查普通函数的参数\\n for match in js_func_pattern.finditer(cleaned):\\n args_str = match.group(1)\\n if args_str and args_str.strip():\\n args = args_str.split(',')\\n for arg in args:\\n arg_match = js_arg_pattern.search(arg)\\n if arg_match:\\n var_name = arg_match.group(1)\\n found_any_var = True\\n if not (3 <= len(var_name) <= 5):\\n return False\\n \\n # 检查带括号的箭头函数参数\\n for match in arrow_func_pattern.finditer(cleaned):\\n args_str = match.group(1)\\n if args_str and args_str.strip():\\n args = args_str.split(',')\\n for arg in args:\\n arg_match = js_arg_pattern.search(arg)\\n if arg_match:\\n var_name = arg_match.group(1)\\n found_any_var = True\\n if not (3 <= len(var_name) <= 5):\\n return False\\n \\n # 检查不带括号的单参数箭头函数\\n for match in single_arg_arrow_pattern.finditer(cleaned):\\n var_name = match.group(1)\\n found_any_var = True\\n if not (3 <= len(var_name) <= 5):\\n return False\\n\\n return found_any_var\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行查找处理文件的任务时，请尽可能使用 Task 工具并配合 subagent_type=Explore 来执行查找处理文件的任务，以代替直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"最后一步，让我们为管理后台增加对座位的深度管控能力。1. 在仪表盘页面增加一个“场次座位管理”模块，将之前前台做的 SVG 座位图复用到这里，但作为管理员视角展示。2. 管理员可以点击任意“已售”或“可选”的座位，点击后弹出一个小型悬浮菜单，允许将其强制更改为“设备维修”状态。3. 变成“设备维修”状态的座位在 SVG 图中需要用带有红色叉号（X）的特殊样式表示。4. 在右侧的“最新订单”列表中，为每条订单增加一个“退票”按钮，点击退票后，不仅订单要从列表中移除，顶部的“总销售额”和“已售票数”统计卡片还要展示一个数字动态递减的动画效果。\\n另外请注意，后续所有轮次都请遵循以下更新及新增要求：第一，原有你输出的所有代码文件都必须完全排除\\\"TODO\\\"和\\\"FIXME\\\"的要求调整为你输出的所有代码文件都必须完全排除\\\"TODO\\\"、\\\"FIXME\\\"和\\\"DEBUG\\\"；第二，你所生成的所有代码文件中，任意一行都不能超出 100 个字符的长度。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先在task_plan.md中说明修改计划，然后才能对代码进行任何更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前后端代码、脚本及文本文件扩展名\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', \\n '.scss', '.less', '.json', '.md', '.java', '.c', '.cpp', '.h', \\n '.go', '.rs', '.php', '.rb', '.sh', '.xml', '.yaml', '.yml'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码\\n # 如果文件被编码为 GBK 等其他编码且包含非 ASCII 字符，这里将会引发 UnicodeDecodeError\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 解码失败，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能的文件读取系统错误\\n pass\\n \\n # 所有被检查的代码文件均可无异常地以 UTF-8 解码，说明满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在编写每个代码文件时，请切记把指定的编码声明 -*- coding: utf-8 -*- 放在代码的最首行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名（结合对话中的前端原型开发任务，涵盖前端及常见通用代码）\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.py', '.java', '.c', '.cpp', '.h', '.go', \\n '.php', '.rb', '.sh', '.svg'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n code_file_count = 0\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录及常见依赖/构建产物目录，避免检查到无关文件\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'dist', 'build', 'out')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_file_count += 1\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 仅读取最首行\\n first_line = f.readline()\\n # 判断首行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取发生异常（如由于非文本内容导致等），视为不满足要求\\n return False\\n\\n # 若工作区内没有找到任何代码文件，说明未有效生成代码项目，返回 False\\n return code_file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每个代码文件中，请将所有注释全用英文呈现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你创建的各个代码文件都必须在结尾处使用新的一行书写注释，清楚地声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 前端开发和一般项目中常见的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.sh'}\\n \\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found += 1\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 无法按 utf-8 读取的情况，判断为未满足标准代码格式或意外文件错误\\n return False\\n \\n # 剔除内容中的空行，以便准确判断文件“结尾处”的内容\\n lines = [line.strip() for line in content.splitlines() if line.strip()]\\n if not lines:\\n return False\\n \\n found_in_new_line = False\\n # 在代码文件的最后 5 个非空行中进行检索，验证\\\"结尾处\\\"要求\\n for line in reversed(lines[-5:]):\\n if target_str in line:\\n # 验证\\\"使用新的一行书写注释\\\"的要求\\n # 将目标字符串从该行中剔除，若该行是单纯的注释行，剩下的应该只有注释符号（如 //, <!-- -->, # 等）\\n # 因此剩余内容中不应该再包含任何英文字母或数字\\n remaining = line.replace(target_str, \\\"\\\").strip()\\n if not re.search(r'[a-zA-Z0-9]', remaining):\\n found_in_new_line = True\\n break\\n \\n # 只要有任意一个代码文件不满足上述要求，则整体判负\\n if not found_in_new_line:\\n return False\\n \\n # 确保项目中确实存在被创建的代码文件，如果未发现任何代码文件则返回False\\n return code_files_found > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断AI助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，代表是否满足要求。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复可能存在的前导空白字符后，检查是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在整个对话的回复中，凡是需要自称\\\"我\\\"的地方，都必须严格换成\\\"本小姐\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复中是否严格将'我'替换为'本小姐'。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n # 如果回复内容中存在\\\"我\\\"字，说明并未严格替换，返回 False\\n if \\\"我\\\" in response:\\n return False\\n \\n # 否则满足要求（可能已替换为\\\"本小姐\\\"，或本轮回复中本身无需自称）\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目中任何目录的层深都不允许大于5层\",\n \"validation_code\": \"def check_directory_depth(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录自身的相对路径为 '.'，深度为 0\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 统一路径分隔符，兼容不同操作系统\\n normalized_path = rel_path.replace('\\\\\\\\', '/')\\n # 分割后列表的长度即为层深\\n depth = len(normalized_path.split('/'))\\n \\n # 如果层深大于 5，即视为不满足要求\\n if depth > 5:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行bash相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"代码写好以后，禁止运行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件都必须完全排除\\\"TODO\\\"、\\\"FIXME\\\"和\\\"DEBUG\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义需要被禁用的目标关键词\\n forbidden_words = [\\\"TODO\\\", \\\"FIXME\\\", \\\"DEBUG\\\"]\\n \\n # 定义常见的前端与后端代码文件的扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', \\n '.html', '.htm', \\n '.css', '.scss', '.less', \\n '.vue', '.py', '.json'\\n }\\n \\n # 如果工作区路径不存在，说明并没有生成代码，也就无从验证\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 遍历当前工作区的所有目录和文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为需要检查的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查是否包含任何禁用的关键词\\n for word in forbidden_words:\\n if word in content:\\n return False\\n except Exception:\\n # 忽略不可读文件或编码无法解析的文件\\n continue\\n \\n # 如果遍历完毕且没有发现任何禁用关键词，则代表满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"确保在生成的代码文件里，所有的变量名长度都限制在3到5个字符之间（包含3和5）\",\n \"validation_code\": \"def check_variable_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n \\n # 定义各类代码文件的后缀\\n py_extensions = {'.py'}\\n js_ts_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}\\n css_extensions = {'.css'}\\n \\n # 匹配 JS/TS 中变量声明的基础正则 (let, const, var 后面的标识符)\\n var_pattern = re.compile(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 检查 Python 文件\\n if ext in py_extensions:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 检查赋值变量\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if not (3 <= len(node.id) <= 5):\\n return False\\n # 检查函数参数\\n elif isinstance(node, ast.arg):\\n if not (3 <= len(node.arg) <= 5):\\n return False\\n except Exception:\\n pass\\n \\n # 检查 JS/TS/HTML/VUE 等前端文件\\n elif ext in js_ts_extensions:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除文件中的注释和字符串内容，防止正则误匹配\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n # 1. 检查所有的普通变量声明\\n for v in var_pattern.findall(content):\\n if not (3 <= len(v) <= 5):\\n return False\\n \\n # 2. 检查普通函数(function)中定义的参数\\n func_params = re.findall(r'\\\\bfunction\\\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', content)\\n for params_str in func_params:\\n params = [p.strip() for p in params_str.split(',')]\\n for p in params:\\n if not p:\\n continue\\n # 提取开头的标识符(防出现带默认值的复杂情况)\\n p_match = re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\\n if p_match:\\n v_name = p_match.group(1)\\n if not (3 <= len(v_name) <= 5):\\n return False\\n except Exception:\\n pass\\n \\n # 检查 CSS 文件中的原生变量\\n elif ext in css_extensions:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 匹配 --xxx: 的 CSS 变量格式\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', content)\\n for v in css_vars:\\n if not (3 <= len(v) <= 5):\\n return False\\n except Exception:\\n pass\\n \\n # 如果遍历完毕且没有发现任何长度不在[3, 5]范围内的变量名，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"进行查找处理文件的任务时，请尽可能使用 Task 工具并配合 subagent_type=Explore 来执行查找处理文件的任务，以代替直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件中，任意一行都不能超出 100 个字符的长度\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 基于用户多轮指令推断可能的前端相关代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.svg', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏文件夹和常见的依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的换行符和回车符后，使用 len 函数检查字符数\\n line_content = line.rstrip('\\\\r\\\\n')\\n if len(line_content) > 100:\\n return False\\n except Exception:\\n # 如果遇到无法读取或非正常编码的文件，安全跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面顶部导航栏，检查左侧是否显示Logo标识，右侧是否包含'活动预订'和'管理后台'两个可点击的Tab页签。","点击导航栏的'管理后台'Tab，检查下方主内容区域是否切换并显示欢迎标题；接着点击'活动预订'Tab，检查主内容区是否切换为展示标题和'立即探索'按钮。","在'活动预订'视图中（部分设计可能需点击'立即探索'后展示），检查页面是否展示纯前端代码绘制的当前月份完整日历（不可是原生input弹窗），并且日历中部分特定日期的数字下方标有小圆点以表示'有活动'。","在日历中点击一个底部没有小圆点标记的日期，检查日历右侧或下方区域是否正确显示出'当日暂无活动安排'的空状态提示文本。","在日历中点击一个带有小圆点标记的日期，检查日历旁边是否滑出活动列表面板，面板中包含2-3个虚构的活动卡片，卡片需展示活动名称、时间段和起步票价。","点击活动列表中的任意一个活动卡片，检查日历区域是否被隐藏，页面主体是否随之渲染出一张SVG构成的座位分布图，且在图的顶部包含'舞台'字样的标识。","观察SVG座位分布图，检查座位形状是否为圆形或圆角矩形，前排VIP区座位是否使用金色或橙色标识，后排普通区座位是否使用蓝色标识；图中同时存在灰色的'已售'座位。 同时点击灰色‘已售’座位，确认其不可被选中，状态不发生变化，也不会加入已选座位面板。","在座位图中点击一个普通颜色的'可选'座位，检查该座位是否改变为绿色（'已选'状态）；再次点击该座位，检查它是否恢复为原来的默认颜色。","在座位图中分别点击选中一个前排VIP座位和一个后排普通座位，检查座位图右侧的'已选座位'面板是否实时出现这两个座位的排号信息（如A排3座），两者的单价需有区分，并且底部自动累加计算出正确的总价。","在座位图中连续点击选择5个'可选'座位，检查在试图选中第5个时，页面中心是否弹出醒目的错误Toast提示（最多只能选择4个座位），且第5个座位的状态被阻止变更，维持'可选'颜色不变。","在成功选中1-4个座位后，点击'已选座位'面板最下方的'确认选座'按钮，检查页面是否弹出一个模态框，要求用户填写'联系人姓名'和'手机号'。","在表单模态框中填入测试姓名和手机号并点击'提交订单'按钮，检查选座界面和表单模态框是否全部隐藏，页面切换至展示'预订成功'的门票详情视图。","观察预订成功的电子门票卡片，检查卡片外观是否带有拟物感的阴影和圆角效果，且文本中准确包含了此前选择的活动名称、时间、用户填写的联系人姓名以及正确的座位排号。","检查电子门票的右侧或下半部分，是否展示了一个使用排列紧密的黑色小方块（纯前端代码如SVG或div元素组合）模拟生成的二维码图案。","点击门票界面下方的'打印门票'按钮，检查浏览器是否成功被唤起并弹出系统原生的打印预览页面（验证window.print()的调用）。","切换顶部导航栏进入'管理后台'Tab页签，检查页面顶部是否并列展示了'总销售额'、'已售票数'、'平均上座率'三个数据统计卡片。","观察管理后台主体区域，检查是否展示了一张呈现过去7天销售金额趋势的柱状图（需纯手工绘制，审查代码/DOM不应存在Echarts等第三方图表库结构）。","使用鼠标悬停在销售趋势柱状图的某任意一根柱子上，检查该柱子上方是否即时浮现出一个Tooltip提示框，并在框中显示了对应的具体金额数值。","在管理后台页面右侧，检查是否包含了一个标有'最新订单'的纵向滚动列表区块，其中含有数条模拟的订票记录。","在管理后台的'场次座位管理'模块中找到复用的SVG座位分布图，点击任意一个'已售'或'可选'座位，检查点击后是否立即弹出一个允许更改为'设备维修'状态的小型悬浮菜单。","在上述悬浮菜单中点击设为'设备维修'选项，检查该座位在SVG图中的外观是否更新并显示带有红色叉号（X）的特殊样式标记。","在管理后台右侧'最新订单'列表中找到任意一条含有'退票'按钮的记录，点击该'退票'按钮，检查该条订单记录是否成功从纵向滚动列表中移除。","点击退票按钮并触发退票逻辑的同时，观察页面顶部的'总销售额'和'已售票数'统计卡片，检查这两个卡片上的数值是否展示了动态递减（数字平滑过渡变化）的动画效果，而非生硬突变。"],"string":"[\n \"查看页面顶部导航栏，检查左侧是否显示Logo标识，右侧是否包含'活动预订'和'管理后台'两个可点击的Tab页签。\",\n \"点击导航栏的'管理后台'Tab，检查下方主内容区域是否切换并显示欢迎标题；接着点击'活动预订'Tab，检查主内容区是否切换为展示标题和'立即探索'按钮。\",\n \"在'活动预订'视图中（部分设计可能需点击'立即探索'后展示），检查页面是否展示纯前端代码绘制的当前月份完整日历（不可是原生input弹窗），并且日历中部分特定日期的数字下方标有小圆点以表示'有活动'。\",\n \"在日历中点击一个底部没有小圆点标记的日期，检查日历右侧或下方区域是否正确显示出'当日暂无活动安排'的空状态提示文本。\",\n \"在日历中点击一个带有小圆点标记的日期，检查日历旁边是否滑出活动列表面板，面板中包含2-3个虚构的活动卡片，卡片需展示活动名称、时间段和起步票价。\",\n \"点击活动列表中的任意一个活动卡片，检查日历区域是否被隐藏，页面主体是否随之渲染出一张SVG构成的座位分布图，且在图的顶部包含'舞台'字样的标识。\",\n \"观察SVG座位分布图，检查座位形状是否为圆形或圆角矩形，前排VIP区座位是否使用金色或橙色标识，后排普通区座位是否使用蓝色标识；图中同时存在灰色的'已售'座位。 同时点击灰色‘已售’座位，确认其不可被选中，状态不发生变化，也不会加入已选座位面板。\",\n \"在座位图中点击一个普通颜色的'可选'座位，检查该座位是否改变为绿色（'已选'状态）；再次点击该座位，检查它是否恢复为原来的默认颜色。\",\n \"在座位图中分别点击选中一个前排VIP座位和一个后排普通座位，检查座位图右侧的'已选座位'面板是否实时出现这两个座位的排号信息（如A排3座），两者的单价需有区分，并且底部自动累加计算出正确的总价。\",\n \"在座位图中连续点击选择5个'可选'座位，检查在试图选中第5个时，页面中心是否弹出醒目的错误Toast提示（最多只能选择4个座位），且第5个座位的状态被阻止变更，维持'可选'颜色不变。\",\n \"在成功选中1-4个座位后，点击'已选座位'面板最下方的'确认选座'按钮，检查页面是否弹出一个模态框，要求用户填写'联系人姓名'和'手机号'。\",\n \"在表单模态框中填入测试姓名和手机号并点击'提交订单'按钮，检查选座界面和表单模态框是否全部隐藏，页面切换至展示'预订成功'的门票详情视图。\",\n \"观察预订成功的电子门票卡片，检查卡片外观是否带有拟物感的阴影和圆角效果，且文本中准确包含了此前选择的活动名称、时间、用户填写的联系人姓名以及正确的座位排号。\",\n \"检查电子门票的右侧或下半部分，是否展示了一个使用排列紧密的黑色小方块（纯前端代码如SVG或div元素组合）模拟生成的二维码图案。\",\n \"点击门票界面下方的'打印门票'按钮，检查浏览器是否成功被唤起并弹出系统原生的打印预览页面（验证window.print()的调用）。\",\n \"切换顶部导航栏进入'管理后台'Tab页签，检查页面顶部是否并列展示了'总销售额'、'已售票数'、'平均上座率'三个数据统计卡片。\",\n \"观察管理后台主体区域，检查是否展示了一张呈现过去7天销售金额趋势的柱状图（需纯手工绘制，审查代码/DOM不应存在Echarts等第三方图表库结构）。\",\n \"使用鼠标悬停在销售趋势柱状图的某任意一根柱子上，检查该柱子上方是否即时浮现出一个Tooltip提示框，并在框中显示了对应的具体金额数值。\",\n \"在管理后台页面右侧，检查是否包含了一个标有'最新订单'的纵向滚动列表区块，其中含有数条模拟的订票记录。\",\n \"在管理后台的'场次座位管理'模块中找到复用的SVG座位分布图，点击任意一个'已售'或'可选'座位，检查点击后是否立即弹出一个允许更改为'设备维修'状态的小型悬浮菜单。\",\n \"在上述悬浮菜单中点击设为'设备维修'选项，检查该座位在SVG图中的外观是否更新并显示带有红色叉号（X）的特殊样式标记。\",\n \"在管理后台右侧'最新订单'列表中找到任意一条含有'退票'按钮的记录，点击该'退票'按钮，检查该条订单记录是否成功从纵向滚动列表中移除。\",\n \"点击退票按钮并触发退票逻辑的同时，观察页面顶部的'总销售额'和'已售票数'统计卡片，检查这两个卡片上的数值是否展示了动态递减（数字平滑过渡变化）的动画效果，而非生硬突变。\"\n]","isConversational":false}}},{"rowIdx":53,"cells":{"task_id":{"kind":"number","value":54,"string":"54"},"repository_policy":{"kind":"string","value":"你在后续所有轮次生成内容时，必须严格遵守以下全部约束要求：\n1. 你所生成的所有代码文件都必须使用UTF-8进行编码。\n2. 生成代码文件中的各个函数，每一个都不能超出60行的限制。\n3. 限制你生成的代码文件中循环嵌套最多为2层。\n4. 你所生成的代码文件中绝对不能使用多层if。\n5. 代码文件中的命名要求是：每个变量名的字符长度必须正好是质数。\n6. 无论哪个代码文件，其注释块中都不能带有emoji表情符号。\n7. 你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.scss', \n '.less', '.json', '.sh', '.bash', '.go', '.rs', '.php', \n '.rb', '.swift', '.kt', '.kts', '.sql', '.xml', '.yaml', \n '.yml', '.vue', '.dart', '.m', '.mm'\n }\n\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选出符合常见代码后缀的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_content = f.read()\n \n # 尝试将字节流以 UTF-8 进行解码\n raw_content.decode('utf-8')\n \n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明存在不兼容 UTF-8 的编码字符\n return False\n except Exception:\n # 捕获其他不可预期的文件读取错误\n return False\n\n # 所有代码文件均成功以 UTF-8 解码\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"def check_function_length(response: str, workspace_path: str) -> bool:\n import os\n import ast\n\n # 遍历代码项目地址中的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 依据题目提示，推断可能的代码类型并进行验证，此处以Python文件为例\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n source_code = f.read()\n \n # 解析代码为抽象语法树\n tree = ast.parse(source_code)\n \n # 遍历语法树的每一个节点\n for node in ast.walk(tree):\n # 判断是否为普通函数定义或异步函数定义\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_line = node.lineno\n end_line = node.end_lineno\n \n # 存在结束行号时，计算函数的总行数\n if end_line is not None:\n total_lines = end_line - start_line + 1\n # 验证是否超过60行的限制\n if total_lines > 60:\n return False\n except Exception:\n # 如果读取文件失败或代码存在语法错误导致ast解析失败，则跳过并继续检查其他文件\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n def is_prime(n):\n if n < 2:\n return False\n for i in range(2, int(n ** 0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 推断可能的代码文件扩展名\n if not file.endswith(('.py', '.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go', '.rs')):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if file.endswith('.py'):\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 检查赋值变量\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if not is_prime(len(node.id)):\n return False\n # 检查函数参数变量\n elif isinstance(node, ast.arg):\n if not is_prime(len(node.arg)):\n return False\n except Exception:\n pass\n else:\n # 匹配常见类型声明的变量名，使用负向先行断言排除后面紧跟括号的函数定义\n pattern_decl = r'\\b(?:let|var|const|int|double|float|String|char|boolean|long|short|byte)\\s+([a-zA-Z_]\\w*)\\b(?!\\s*\\()'\n matches = re.findall(pattern_decl, content)\n for var_name in matches:\n if not is_prime(len(var_name)):\n return False\n \n # 匹配 Go 语言等使用的 := 短变量声明\n pattern_go = r'\\b([a-zA-Z_]\\w*)\\s*:='\n matches_go = re.findall(pattern_go, content)\n for var_name in matches_go:\n if not is_prime(len(var_name)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_comments_for_no_emoji(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的常见代码文件扩展名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs',\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', '.rs',\n '.rb', '.php', '.sh', '.swift', '.kt', '.m', '.scala'\n }\n\n # 定义匹配各类注释的正则表达式\n comment_patterns = [\n re.compile(r'//.*'), # C/C++/Java/JS/TS/Go 等的单行注释\n re.compile(r'#.*'), # Python/Ruby/Shell 等的单行注释\n re.compile(r'/\\*.*?\\*/', re.DOTALL), # C/C++/Java/JS/TS/CSS 等的多行注释\n re.compile(r'<!--.*?-->', re.DOTALL), # HTML 风格的多行注释\n re.compile(r'\\\"\\\"\\\"(?:.*?)\\\"\\\"\\\"', re.DOTALL), # Python 的双引号多行注释/文档字符串\n re.compile(r\"\\'\\'\\'(?:.*?)\\'\\'\\'\", re.DOTALL) # Python 的单引号多行注释/文档字符串\n ]\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file_name)\n try:\n # 要求1中规定生成代码文件使用UTF-8编码，因此以UTF-8尝试读取\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略无法正常解码或读取的文件\n continue\n\n # 遍历所有的注释匹配模式\n for pattern in comment_patterns:\n comments = pattern.findall(content)\n for comment in comments:\n # 遍历注释块中的每一个字符，检查是否为 emoji\n for char in comment:\n if is_emoji(char):\n return False # 一旦发现注释块中带有 emoji，即可判定不满足要求\n\n # 所有文件的注释中均未发现 emoji，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_code_lines_even(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', \n '.cs', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', \n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.sh', \n '.sql', '.dart', '.scala'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 计入空行的物理行数计算\n if not content:\n line_count = 0 # 空文件视为0行\n else:\n line_count = content.count('\\n') + 1\n \n # 如果存在任何一个代码文件的行数不是偶数，返回 False\n if line_count % 2 != 0:\n return False\n except Exception:\n # 无法读取的特殊文件直接跳过\n continue\n\n # 遍历检查了所有代码文件，没有发现行数为奇数的文件，即全部满足\n return True","tags":["Quantity","Exact Value"]}],"string":"[\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.scss', \\n '.less', '.json', '.sh', '.bash', '.go', '.rs', '.php', \\n '.rb', '.swift', '.kt', '.kts', '.sql', '.xml', '.yaml', \\n '.yml', '.vue', '.dart', '.m', '.mm'\\n }\\n\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选出符合常见代码后缀的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_content = f.read()\\n \\n # 尝试将字节流以 UTF-8 进行解码\\n raw_content.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明存在不兼容 UTF-8 的编码字符\\n return False\\n except Exception:\\n # 捕获其他不可预期的文件读取错误\\n return False\\n\\n # 所有代码文件均成功以 UTF-8 解码\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"def check_function_length(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n\\n # 遍历代码项目地址中的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 依据题目提示，推断可能的代码类型并进行验证，此处以Python文件为例\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n source_code = f.read()\\n \\n # 解析代码为抽象语法树\\n tree = ast.parse(source_code)\\n \\n # 遍历语法树的每一个节点\\n for node in ast.walk(tree):\\n # 判断是否为普通函数定义或异步函数定义\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_line = node.lineno\\n end_line = node.end_lineno\\n \\n # 存在结束行号时，计算函数的总行数\\n if end_line is not None:\\n total_lines = end_line - start_line + 1\\n # 验证是否超过60行的限制\\n if total_lines > 60:\\n return False\\n except Exception:\\n # 如果读取文件失败或代码存在语法错误导致ast解析失败，则跳过并继续检查其他文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n def is_prime(n):\\n if n < 2:\\n return False\\n for i in range(2, int(n ** 0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 推断可能的代码文件扩展名\\n if not file.endswith(('.py', '.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go', '.rs')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if file.endswith('.py'):\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 检查赋值变量\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if not is_prime(len(node.id)):\\n return False\\n # 检查函数参数变量\\n elif isinstance(node, ast.arg):\\n if not is_prime(len(node.arg)):\\n return False\\n except Exception:\\n pass\\n else:\\n # 匹配常见类型声明的变量名，使用负向先行断言排除后面紧跟括号的函数定义\\n pattern_decl = r'\\\\b(?:let|var|const|int|double|float|String|char|boolean|long|short|byte)\\\\s+([a-zA-Z_]\\\\w*)\\\\b(?!\\\\s*\\\\()'\\n matches = re.findall(pattern_decl, content)\\n for var_name in matches:\\n if not is_prime(len(var_name)):\\n return False\\n \\n # 匹配 Go 语言等使用的 := 短变量声明\\n pattern_go = r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*:='\\n matches_go = re.findall(pattern_go, content)\\n for var_name in matches_go:\\n if not is_prime(len(var_name)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_comments_for_no_emoji(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的常见代码文件扩展名\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs',\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', '.rs',\\n '.rb', '.php', '.sh', '.swift', '.kt', '.m', '.scala'\\n }\\n\\n # 定义匹配各类注释的正则表达式\\n comment_patterns = [\\n re.compile(r'//.*'), # C/C++/Java/JS/TS/Go 等的单行注释\\n re.compile(r'#.*'), # Python/Ruby/Shell 等的单行注释\\n re.compile(r'/\\\\*.*?\\\\*/', re.DOTALL), # C/C++/Java/JS/TS/CSS 等的多行注释\\n re.compile(r'<!--.*?-->', re.DOTALL), # HTML 风格的多行注释\\n re.compile(r'\\\\\\\"\\\\\\\"\\\\\\\"(?:.*?)\\\\\\\"\\\\\\\"\\\\\\\"', re.DOTALL), # Python 的双引号多行注释/文档字符串\\n re.compile(r\\\"\\\\'\\\\'\\\\'(?:.*?)\\\\'\\\\'\\\\'\\\", re.DOTALL) # Python 的单引号多行注释/文档字符串\\n ]\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file_name)\\n try:\\n # 要求1中规定生成代码文件使用UTF-8编码，因此以UTF-8尝试读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略无法正常解码或读取的文件\\n continue\\n\\n # 遍历所有的注释匹配模式\\n for pattern in comment_patterns:\\n comments = pattern.findall(content)\\n for comment in comments:\\n # 遍历注释块中的每一个字符，检查是否为 emoji\\n for char in comment:\\n if is_emoji(char):\\n return False # 一旦发现注释块中带有 emoji，即可判定不满足要求\\n\\n # 所有文件的注释中均未发现 emoji，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_code_lines_even(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', \\n '.cs', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', \\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.sh', \\n '.sql', '.dart', '.scala'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 计入空行的物理行数计算\\n if not content:\\n line_count = 0 # 空文件视为0行\\n else:\\n line_count = content.count('\\\\n') + 1\\n \\n # 如果存在任何一个代码文件的行数不是偶数，返回 False\\n if line_count % 2 != 0:\\n return False\\n except Exception:\\n # 无法读取的特殊文件直接跳过\\n continue\\n\\n # 遍历检查了所有代码文件，没有发现行数为奇数的文件，即全部满足\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我搭建一个新闻聚合器 PWA 应用的基础页面骨架。需要一个顶部的 App Bar（包含应用标题和一个菜单占位图标），中间的主体区域放几个简单的静态新闻卡片（暂时只需一个灰色的方块代表图片，旁边放两行假文字代表标题即可），底部放一个包含'新闻'和'书签'两个选项的标签导航栏（Tab Bar）。这是一个初版框架，不需要写任何核心业务逻辑，先把基本结构和CSS样式撑起来就好。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”字样；第二，在为用户生成回复时，你必须永远只用中文。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件后缀，可根据具体的生成内容（如PWA应用可能涉及的HTML, CSS, JS等）推断\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.json', '.xml',\n '.ts', '.jsx', '.tsx', '.vue', '.py', '.java',\n '.c', '.cpp', '.h', '.hpp', '.md', '.txt'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取整个代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明存在非 UTF-8 编码的字符\n return False\n except Exception:\n # 忽略其他非编码相关的异常（如权限问题等）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n # 如果工作区不存在，视作没有违背该限制\n return True\n\n # 替换函数：用于去除 JS/TS 代码中的注释和字符串字面量\n # 规则：若是换行符保留以维持行号，非换行符全替换为空格\n def replacer(match):\n s = match.group(0)\n return ''.join('\\n' if c == '\\n' else ' ' for c in s)\n\n # 匹配JS注释和字符串的正则表达式\n js_clean_pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"|`(?:\\\\.|[^\\\\`])*`',\n re.MULTILINE | re.DOTALL\n )\n\n def check_js_code(code_str: str) -> bool:\n # 去除注释与字符串字面量\n clean_code = js_clean_pattern.sub(replacer, code_str)\n \n matches = []\n # 1. 查找 function 关键字\n for match in re.finditer(r'\\bfunction\\b', clean_code):\n matches.append(('func', match.start(), match.end()))\n # 2. 查找箭头函数 =>\n for match in re.finditer(r'=>', clean_code):\n matches.append(('arrow', match.start(), match.end()))\n # 3. 查找类或对象的简写方法定义，排除各类控制流关键字\n method_pattern = r'\\b(?!(?:if|for|while|switch|catch|with|function|try|else|do|return|yield)\\b)[a-zA-Z_$][a-zA-Z0-9_$]*\\s*\\([^)]*\\)\\s*\\{'\n for match in re.finditer(method_pattern, clean_code):\n matches.append(('method', match.start(), match.end()))\n\n for m_type, m_start, m_end in matches:\n # 计算函数的起始行号\n start_line = clean_code.count('\\n', 0, m_start) + 1\n \n brace_idx = -1\n if m_type == 'method':\n # 简写方法的正则结尾本身就是 '{'\n brace_idx = m_end - 1\n else:\n # 寻找其后的第一个 '{'\n for i in range(m_end, len(clean_code)):\n if clean_code[i] == '{':\n brace_idx = i\n break\n elif clean_code[i] in ';}':\n # 如果在遇到 '{' 之前遇到 ';' 或 '}'，说明可能不是块级函数结构\n break\n \n # 若没找到 '{' 则不是包含代码块的函数（例如单行箭头函数），跳过\n if brace_idx == -1:\n continue\n \n depth = 0\n end_line = start_line\n in_brace = False\n \n # 使用花括号深度匹配找到函数体的结束位置\n for i in range(brace_idx, len(clean_code)):\n if clean_code[i] == '{':\n depth += 1\n in_brace = True\n elif clean_code[i] == '}':\n depth -= 1\n \n # 深度归零，说明函数体闭合\n if in_brace and depth == 0:\n end_line = clean_code.count('\\n', 0, i) + 1\n break\n \n if in_brace and depth == 0:\n # 检查函数总行数\n if end_line - start_line + 1 > 60:\n return False\n \n return True\n\n def check_py_code(code_str: str) -> bool:\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n func_length = node.end_lineno - node.lineno + 1\n if func_length > 60:\n return False\n except Exception:\n # 忽略语法解析错误的Python文件\n pass\n return True\n\n # HTML文件中的 script 块内容提取正则\n html_pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\n\n # 遍历工作目录下所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n if ext in ['.py', '.js', '.ts', '.html', '.htm', '.jsx', '.tsx']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_str = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_py_code(code_str):\n return False\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n if not check_js_code(code_str):\n return False\n elif ext in ['.html', '.htm']:\n # 针对HTML，校验所有内嵌的 script 标签内容\n for script_match in html_pattern.finditer(code_str):\n script_content = script_match.group(1)\n if not check_js_code(script_content):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断一个数是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 如果工作目录不存在，则无法验证，且由于不存在违规变量，视作没有违反该规则\n if not os.path.exists(workspace_path):\n return True\n\n # 编译匹配不同语言中变量声明的正则表达式\n # 匹配 JavaScript/TypeScript 变量：var, let, const 后面的第一个合法标识符\n js_var_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\b')\n # 匹配 CSS 变量：-- 开头的自定义属性名\n css_var_pattern = re.compile(r'(--[a-zA-Z0-9_-]+)\\s*:')\n # 匹配 Python 变量：行首紧跟的变量赋值\n py_var_pattern = re.compile(r'^\\s*([a-zA-Z_][a-zA-Z0-9_]*)\\s*=', re.MULTILINE)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略无法通过UTF-8解码或其他无法读取的文件\n continue\n\n # 提取当前文件内的变量列表\n vars_found = []\n\n # 针对 HTML/JavaScript 等前端代码文件\n if file.endswith(('.js', '.ts', '.html', '.htm', '.vue', '.jsx', '.tsx')):\n vars_found.extend(js_var_pattern.findall(content))\n \n # 针对 HTML/CSS 等样式文件\n if file.endswith(('.css', '.html', '.htm', '.vue')):\n vars_found.extend(css_var_pattern.findall(content))\n\n # 针对 Python 后端文件\n if file.endswith('.py'):\n vars_found.extend(py_var_pattern.findall(content))\n\n # 遍历检查提取出的每一个变量名\n for var_name in vars_found:\n var_name = var_name.strip()\n # 题目要求为字符长度（即字符数），按照指令提示可以直接使用len函数\n if not is_prime(len(var_name)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 遍历工作区下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 针对该 PWA 项目可能涉及的前端代码文件进行过滤\n if ext in ['.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue']:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n # 提取各类型代码的注释内容\n if ext == '.html':\n # 匹配 HTML 特有注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 匹配 HTML 中嵌入的 JS/CSS 注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'(?<!:)//(.*?)$', content, re.MULTILINE))\n elif ext == '.css':\n # 匹配 CSS 多行注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.vue']:\n # 匹配 JS 等脚本的多行和单行注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'(?<!:)//(.*?)$', content, re.MULTILINE))\n \n # 遍历提取出的注释内容，判断是否存在 emoji 表情符号\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对PWA应用推断常见的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.jsx', '.tsx'}\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按换行符精确切分，包含所有空行\n total_lines = len(content.split('\\n'))\n \n # 检查行数是否为偶数\n if total_lines % 2 != 0:\n return False\n except Exception:\n pass\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”字样","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # Regex pattern to match HTML comments, block comments (/* ... */), and line comments (// ...)\n comment_pattern = re.compile(r'(?:<!--[\\s\\S]*?-->)|(?:/\\*[\\s\\S]*?\\*/)|(?://[^\\r\\n]*)')\n \n valid_extensions = ('.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.py')\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n comments = comment_pattern.findall(content)\n for comment in comments:\n if '测试' in comment:\n return False\n except Exception:\n # Silently ignore files that cannot be read (e.g., unexpected binaries)\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]}]},{"round_id":1,"instruction":"第一版骨架不错。现在让新闻列表活起来，我们需要加载真实感强的数据。请用模拟的 JSON 数据加载新闻，并在主体区域渲染出更精致的新闻卡片。卡片左侧要展示新闻缩略图，右侧包含标题、新闻来源、发布时间（比如'2小时前'）和一段简短的摘要。文字过长要做截断处理（显示省略号），卡片之间留有适当的间距，排版需要显得清晰现代一些。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”字样；第二，你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断，PWA应用相关的常见代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.webmanifest', '.py'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试使用 utf-8 对字节序列进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明包含非 utf-8 编码的内容\n return False\n except Exception:\n # 其他读取异常保守跳过\n continue\n\n # 遍历完所有代码文件均未报错，则满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_js_functions(js_code: str) -> bool:\n # 用于匹配 JS 中函数的开头结构：\n # 1. 普通函数定义或匿名函数: function name() { 或 function() {\n # 2. 带有块的箭头函数: => {\n # 3. 类或对象方法: methodName(args) { (通过前向否定排除 if/for/while 等控制流关键字)\n pattern = re.compile(r'(\\bfunction\\b[^\\{]*\\{|=>\\s*\\{|\\b(?!(?:if|for|while|catch|switch)\\b)\\w+\\s*\\([^)]*\\)\\s*\\{)')\n idx = 0\n \n while idx < len(js_code):\n match = pattern.search(js_code, idx)\n if not match:\n break\n \n start_match = match.start()\n start_brace = match.end() - 1 # 找到开头结构中 `{` 的位置\n \n brace_count = 0\n in_string = False\n string_char = ''\n in_comment_line = False\n in_comment_block = False\n \n end_idx = start_brace\n while end_idx < len(js_code):\n char = js_code[end_idx]\n \n # 状态机：处理单行注释\n if in_comment_line:\n if char == '\\n':\n in_comment_line = False\n # 状态机：处理多行注释\n elif in_comment_block:\n if char == '*' and end_idx + 1 < len(js_code) and js_code[end_idx+1] == '/':\n in_comment_block = False\n end_idx += 1\n # 状态机：处理字符串\n elif in_string:\n if char == '\\\\':\n end_idx += 1 # 跳过转义符\n elif char == string_char:\n in_string = False\n # 正常代码匹配流程\n else:\n if char in '\"\\'`':\n in_string = True\n string_char = char\n elif char == '/' and end_idx + 1 < len(js_code):\n next_char = js_code[end_idx+1]\n if next_char == '/':\n in_comment_line = True\n end_idx += 1\n elif next_char == '*':\n in_comment_block = True\n end_idx += 1\n elif char == '{':\n brace_count += 1\n elif char == '}':\n brace_count -= 1\n if brace_count == 0: # 匹配到了当前函数结束的花括号\n break\n end_idx += 1\n \n # 若完整匹配到成对的花括号，截取代码段计算行数\n if brace_count == 0 and end_idx < len(js_code):\n func_str = js_code[start_match:end_idx+1]\n # 计算函数行数\n line_count = func_str.count('\\n') + 1\n if line_count > 60:\n return False\n \n # 继续往后扫描下一个潜在函数\n idx = start_brace + 1\n \n return True\n\n # 遍历当前工作区所有的生成文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 处理 Python 文件\n if file.endswith('.py'):\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 仅限制函数不能超过60行\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n # 兼容不同 Python 版本的行数获取\n if hasattr(node, 'end_lineno'):\n end = node.end_lineno\n else:\n max_line = node.lineno\n for child in ast.walk(node):\n if hasattr(child, 'lineno'):\n max_line = max(max_line, child.lineno)\n end = max_line\n \n if end - start + 1 > 60:\n return False\n except Exception:\n pass\n \n # 处理 JS / TS 文件\n elif file.endswith('.js') or file.endswith('.ts'):\n if not check_js_functions(content):\n return False\n \n # 处理 HTML 等内嵌了 script 标签的文件\n elif file.endswith('.html') or file.endswith('.htm'):\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_js_functions(script):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"import os\nimport re\nimport ast\n\ndef check(response: str, workspace_path: str) -> bool:\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n in (2, 3):\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n class PyVarVisitor(ast.NodeVisitor):\n def __init__(self):\n self.variables = set()\n \n def visit_Assign(self, node):\n for target in node.targets:\n if isinstance(target, ast.Name):\n self.variables.add(target.id)\n self.generic_visit(node)\n \n def visit_AnnAssign(self, node):\n if isinstance(node.target, ast.Name):\n self.variables.add(node.target.id)\n self.generic_visit(node)\n \n def visit_FunctionDef(self, node):\n for arg in node.args.args:\n self.variables.add(arg.arg)\n if hasattr(node.args, 'kwonlyargs'):\n for arg in node.args.kwonlyargs:\n self.variables.add(arg.arg)\n self.generic_visit(node)\n\n variables = set()\n # 匹配 JavaScript 简单变量定义 (var, let, const)\n js_var_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\b')\n # 匹配 JavaScript 函数参数\n js_func_pattern = re.compile(r'\\bfunction\\s+(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\s*\\(([^)]*)\\)')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 处理 JS 及 HTML\n if file.endswith(('.js', '.html')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 提取常规变量\n for match in js_var_pattern.findall(content):\n variables.add(match)\n \n # 提取函数参数\n for match in js_func_pattern.findall(content):\n params = match.split(',')\n for p in params:\n p_clean = p.strip()\n # 保证其是合法的标识符\n if re.fullmatch(r'[a-zA-Z_$][0-9a-zA-Z_$]*', p_clean):\n variables.add(p_clean)\n except Exception:\n pass\n # 处理可能存在的 Python 文件\n elif file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n visitor = PyVarVisitor()\n visitor.visit(tree)\n variables.update(visitor.variables)\n except Exception:\n pass\n\n # 验证是否每个提取的变量名的字符长度恰好为质数\n for var in variables:\n if not is_prime(len(var)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n\n code_extensions = {'.html', '.css', '.js', '.json', '.ts'}\n \n comment_patterns = [\n re.compile(r'<!--(.*?)-->', re.DOTALL),\n re.compile(r'/\\*(.*?)\\*/', re.DOTALL),\n re.compile(r'//(.*)$', re.MULTILINE)\n ]\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n for pattern in comment_patterns:\n comments = pattern.findall(content)\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n code_extensions = {'.html', '.css', '.js', '.json', '.ts', '.jsx', '.tsx'}\n file_checked = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n file_checked = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果文件无法读取，在此场景下忽略或视为未完全满足\n continue\n \n return file_checked","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”字样","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.vue', '.py', '.java', \n '.c', '.cpp', '.h', '.hpp', '.cs', '.php', '.go', '.rb'\n }\n \n # 匹配各类代码文件中的注释（开启多行模式以便匹配 $ 到行尾）\n # 分别对应：\n # 1. 块级注释: /* ... */\n # 2. 单行注释: // ... (用否定后视断言忽略如 http:// 中的 //)\n # 3. HTML注释: <!-- ... -->\n # 4. 单行注释: # ...\n comment_pattern = re.compile(\n r'(/\\*[\\s\\S]*?\\*/|(?<!:)//.*?$|<!--[\\s\\S]*?-->|#.*?$)',\n re.MULTILINE\n )\n \n # 需要被排查的违禁词（全部转换为小写，便于统一比较大小写不敏感的demo）\n forbidden_words = ['测试', 'demo']\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果读取文件出现编码等错误，则跳过\n continue\n \n # 提取出文件中的所有注释部分\n comments = comment_pattern.findall(content)\n for comment_text in comments:\n comment_text_lower = comment_text.lower()\n \n # 检查是否包含违禁词\n for word in forbidden_words:\n if word in comment_text_lower:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n\n # 匹配独立的等号，排除 ==, ===, >=, <=, !=, +=, -= 等复合赋值/比较运算符\n pattern = re.compile(r'(?<![<>=!\\+\\-\\*/%&\\|])=(?![=])')\n \n valid_extensions = ('.js', '.html', '.css', '.json', '.vue', '.jsx', '.ts', '.tsx')\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(valid_extensions):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n line = line.rstrip('\\n\\r')\n for match in pattern.finditer(line):\n idx = match.start()\n \n # 验证左侧是否恰好只有两个空格\n if idx < 2:\n return False\n if not (line[idx-1] == ' ' and line[idx-2] == ' '):\n return False\n if idx > 2 and line[idx-3] == ' ':\n return False\n \n # 验证右侧是否恰好只有两个空格\n if idx + 2 >= len(line):\n return False\n if not (line[idx+1] == ' ' and line[idx+2] == ' '):\n return False\n if idx + 3 < len(line) and line[idx+3] == ' ':\n return False\n\n return True","tags":["Style","Layout"]}]},{"round_id":2,"instruction":"好，接下来给顶部 App Bar 下方加一个可横向滑动的分类导航栏，包含如'推荐'、'科技'、'财经'、'娱乐'、'体育'等分类标签。用户点击不同分类时，要在下方的新闻列表加上淡入淡出的切换动画，并展示该分类对应的模拟新闻数据。记得当前选中的分类标签文字要加粗，并且底部带有一个高亮的下划线指示器。另外请你注意两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”字样，替换此前的注释关键词限制要求；第二，凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义PWA应用及常见可能生成的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.json', '.md', \n '.ts', '.jsx', '.tsx', '.vue', '.txt'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_content = f.read()\n # 尝试用 UTF-8 进行解码\n raw_content.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦发现无法用 UTF-8 解码的文件，直接返回 False\n return False\n except Exception:\n # 忽略其他可能因权限或文件系统引起的非编码相关错误\n pass\n\n # 只要检查的代码文件没有抛出 UnicodeDecodeError 且项目确实包含有效文件，即判定为符合要求\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证工作区下所有代码文件中的各个函数，每一个都不能超出 60 行的限制。\n \"\"\"\n def clean_js_code(code: str) -> str:\n # 将字符串和注释替换为等长的空格和换行符组合，保持原本的行数和行号一致\n pattern = r\"'(?:\\\\.|[^'\\\\])*'|\\\"(?:\\\\.|[^\\\"\\\\])*\\\"|`(?:\\\\.|[^`\\\\])*`|//.*|/\\*[\\s\\S]*?\\*/\"\n def replacer(match):\n text = match.group(0)\n return '\\n' * text.count('\\n') + ' ' * (len(text) - text.count('\\n'))\n return re.sub(pattern, replacer, code)\n\n def check_js_content(code: str) -> bool:\n # 如果代码内容整体行数就不超过 60 行，说明必然符合要求，直接放行\n if code.count('\\n') + 1 <= 60:\n return True\n \n cleaned = clean_js_code(code)\n stack = []\n for i, char in enumerate(cleaned):\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start_idx = stack.pop()\n start_line = cleaned.count('\\n', 0, start_idx) + 1\n end_line = cleaned.count('\\n', 0, i) + 1\n lines_count = end_line - start_line + 1\n \n if lines_count > 60:\n prefix = cleaned[:start_idx].rstrip()\n \n # 特征1：箭头函数\n if prefix.endswith('=>'):\n return False\n \n last_end = max(prefix.rfind('}'), prefix.rfind(';'), prefix.rfind('{'))\n segment = prefix[last_end+1:] if last_end != -1 else prefix\n \n # 特征2：包含 function 关键字声明的函数\n if re.search(r'\\bfunction\\b', segment):\n return False\n \n # 特征3：形式为 name(args) 的类成员方法或对象方法\n segment = segment.strip()\n match = re.search(r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*$', segment)\n if match:\n name = match.group(1)\n # 排除普通的控制流或异常处理作用域块\n if name not in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\n return False\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 再次整体过滤小于等于60行的文件\n if content.count('\\n') + 1 <= 60:\n continue\n\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n if not check_js_content(content):\n return False\n \n elif file.endswith(('.html', '.htm', '.vue')):\n scripts = re.finditer(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for match in scripts:\n script_content = match.group(1)\n if not check_js_content(script_content):\n return False\n \n elif file.endswith('.py'):\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n lines_count = node.end_lineno - node.lineno + 1\n if lines_count > 60:\n return False\n except Exception:\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2 or n == 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n def get_js_vars(code: str) -> set:\n vars_set = set()\n # 移除单行与多行注释，避免误匹配\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n # 提取 var, let, const 声明的变量名\n matches = re.findall(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\b', code)\n for m in matches:\n vars_set.add(m)\n \n # 提取函数参数作为变量名\n func_matches = re.findall(r'\\bfunction\\s+(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\s*\\(([^)]*)\\)', code)\n for args in func_matches:\n for arg in args.split(','):\n arg = arg.strip()\n if arg:\n arg_name = re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*', arg)\n if arg_name:\n vars_set.add(arg_name.group(0))\n return vars_set\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 验证 JavaScript / TypeScript 文件\n if ext in ['.js', '.ts']:\n variables = get_js_vars(content)\n for var in variables:\n if not is_prime(len(var)):\n return False\n \n # 验证 CSS 文件\n elif ext == '.css':\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', content)\n for var in css_vars:\n if not is_prime(len(var)):\n return False\n \n # 验证 HTML 文件内嵌的 script 和 style 标签\n elif ext == '.html':\n script_matches = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n for code in script_matches:\n variables = get_js_vars(code)\n for var in variables:\n if not is_prime(len(var)):\n return False\n \n style_matches = re.findall(r'<style\\b[^>]*>(.*?)</style>', content, flags=re.DOTALL | re.IGNORECASE)\n for style in style_matches:\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', style)\n for var in css_vars:\n if not is_prime(len(var)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n # 尝试以 utf-8 格式读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n comments = []\n \n # 匹配 C-style 注释（JS, TS, CSS 等）\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.css', '.c', '.cpp', '.h', '.hpp', '.java', '.go']:\n # 多行注释 /* ... */\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 单行注释 // ... （CSS不支持此写法）\n if ext != '.css':\n comments.extend(re.findall(r'//.*', content))\n \n # 匹配 HTML-style 注释（HTML, XML 等）\n if ext in ['.html', '.htm', '.vue', '.xml', '.svg']:\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n # 匹配 Python/Shell-style 注释（Python, Bash 等）\n if ext in ['.py', '.sh', '.yaml', '.yml', '.rb']:\n comments.extend(re.findall(r'#.*', content))\n \n # 遍历每一个提取出的注释字符串\n for comment in comments:\n # 遍历注释中的每个字符判断是否为 emoji\n for char in comment:\n if is_emoji(char):\n return False # 只要发现存在 emoji，便不满足要求\n \n return True # 遍历完所有代码文件均未在注释中发现 emoji","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_file_lines_are_even(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义当前PWA项目可能涉及的常见代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.jsx', '.tsx'}\n \n files_checked = 0\n \n # 遍历项目工作区路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符进行分割，可以准确统计出包含空行的文件总行数\n lines = content.split('\\n')\n \n # 如果行数是奇数，不满足要求，直接返回False\n if len(lines) % 2 != 0:\n return False\n \n files_checked += 1\n except Exception:\n # 读取出错直接忽略该文件（也可以视作失败，这里选择跳过异常文件）\n pass\n \n # 如果项目中没有任何对应的代码文件，说明没有生成有效项目，返回False\n # 只要成功验证了至少一个文件且都满足偶数行限制，则返回True\n return files_checked > 0","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”字样","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证代码项目中的任何代码文件注释部分是否不得含有“测试”、“demo”、“临时”、“调试”字样。\n \"\"\"\n if not os.path.exists(workspace_path):\n return True\n\n # 需要排除的关键词，统一使用小写以进行无视大小写的匹配\n forbidden_keywords = [\"测试\", \"demo\", \"临时\", \"调试\"]\n\n # 常见的前后端代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n\n comments = []\n \n # HTML 文件：提取 HTML 注释，并处理内部包含的 script 和 style 块\n if ext in ['.html', '.htm']:\n # HTML 格式注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n # 提取 script 块内的 JS 注释\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for script in scripts:\n comments.extend(re.findall(r'(?<!:)//.*', script)) # 避免匹配到 http://\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', script))\n \n # 提取 style 块内的 CSS 注释\n styles = re.findall(r'<style.*?>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n for style in styles:\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', style))\n \n # CSS 文件\n elif ext == '.css':\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n # JS / TS 文件\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n comments.extend(re.findall(r'(?<!:)//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n # Python 文件\n elif ext == '.py':\n comments.extend(re.findall(r'#.*', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n\n # 遍历并验证提取出的所有注释内容\n for comment in comments:\n comment_lower = comment.lower()\n for kw in forbidden_keywords:\n if kw in comment_lower:\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的前端与配置代码文件后缀\n valid_exts = {'.js', '.ts', '.html', '.css', '.vue', '.json'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n i = 0\n while i < len(content):\n if content[i] == '=':\n is_compound = False\n \n # 检查是否为复合运算符（如 !=, >=, <=, +=, -=, *=, /=, %=, &=, |=, ^=, ??=, :=）\n if i > 0 and content[i-1] in '!=><+-*/%&|^~?:':\n is_compound = True\n # 检查是否为箭头函数或者比较运算（如 =>, ==, ===）\n if i < len(content) - 1 and content[i+1] in '=>':\n is_compound = True\n \n if not is_compound:\n # 检查等号左边是否恰好有两个空格\n left_ok = False\n if i >= 2 and content[i-2:i] == ' ':\n # 左侧那两个空格的前一个字符不能是空格（保障恰好是两个空格而不是三个及以上）\n if i < 3 or content[i-3] != ' ':\n left_ok = True\n \n # 检查等号右边是否恰好有两个空格\n right_ok = False\n if i <= len(content) - 3 and content[i+1:i+3] == ' ':\n # 右侧那两个空格的后一个字符不能是空格\n if i > len(content) - 4 or content[i+3] != ' ':\n right_ok = True\n \n # 如果存在单独的等号，其左右两边不满足都有两个空格，则视为违反要求\n if not (left_ok and right_ok):\n return False\n i += 1\n \n return True","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":3,"instruction":"现在来完善新闻的阅读体验。点击任意一条新闻卡片后，不要直接跳走，而是从屏幕底部向上平滑弹出一个全屏覆盖的详情阅读面板。这个阅读面板里需要展示一张顶部的清晰大图、完整的长篇正文段落，右上角必须有一个明显的'关闭'按钮用来退回列表。面板的最底部要固定一个按钮'在原网站阅读'，点击时可以弹出一个'即将跳转到外部链接'的 Toast 提示。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”字样，替换此前的注释关键词限制要求；第二，后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 结合PWA项目特性推断可能的代码文件类型\n code_exts = {'.html', '.css', '.js', '.json', '.md'}\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果出现解码错误，说明存在未按UTF-8编码的文件\n return False\n except Exception:\n # 忽略诸如权限问题等非编码导致的文件读取错误\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n # 使用状态机将代码内的注释、字符串内容等扰乱文本替换为空格，以保持代码原结构的换行符和括号不受干扰\n def mask_strings_and_comments(content):\n masked = []\n state = 'NORMAL'\n i = 0\n n = len(content)\n while i < n:\n c = content[i]\n if state == 'NORMAL':\n if c == '/' and i + 1 < n:\n if content[i+1] == '/':\n state = 'LINE_COMMENT'\n masked.append(' ')\n masked.append(' ')\n i += 2\n continue\n elif content[i+1] == '*':\n state = 'BLOCK_COMMENT'\n masked.append(' ')\n masked.append(' ')\n i += 2\n continue\n if c == '\"':\n state = 'DQUOTE'\n masked.append(c)\n i += 1\n continue\n if c == \"'\":\n state = 'SQUOTE'\n masked.append(c)\n i += 1\n continue\n if c == '`':\n state = 'BQUOTE'\n masked.append(c)\n i += 1\n continue\n masked.append(c)\n i += 1\n elif state == 'LINE_COMMENT':\n if c == '\\n':\n state = 'NORMAL'\n masked.append(c)\n else:\n masked.append(' ')\n i += 1\n elif state == 'BLOCK_COMMENT':\n if c == '*' and i + 1 < n and content[i+1] == '/':\n state = 'NORMAL'\n masked.append(' ')\n masked.append(' ')\n i += 2\n else:\n if c == '\\n':\n masked.append(c)\n else:\n masked.append(' ')\n i += 1\n elif state in ('DQUOTE', 'SQUOTE', 'BQUOTE'):\n quote_char = '\"' if state == 'DQUOTE' else (\"'\" if state == 'SQUOTE' else '`')\n if c == '\\\\':\n masked.append(' ')\n masked.append(' ')\n i += 2\n elif c == quote_char:\n state = 'NORMAL'\n masked.append(c)\n i += 1\n else:\n if c == '\\n':\n masked.append(c)\n else:\n masked.append(' ')\n i += 1\n return \"\".join(masked)\n\n def get_previous_non_space(text, index):\n idx = index - 1\n while idx >= 0 and text[idx].isspace():\n idx -= 1\n return idx\n\n # 判断 JS/TS/HTML(script内) 中的代码块，校验是否含超过60行的函数\n def check_functions_in_js_text(text):\n masked = mask_strings_and_comments(text)\n \n blocks = {}\n stack = []\n # 通过匹配成对的 {} 锁定基础逻辑块\n for i, char in enumerate(masked):\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start_index = stack.pop()\n blocks[start_index] = i\n \n for start_index, end_index in blocks.items():\n prev_idx = get_previous_non_space(masked, start_index)\n if prev_idx < 0:\n continue\n \n is_func = False\n func_start_idx = start_index \n \n # 检测箭头函数结构 `=> {`\n if masked[prev_idx] == '>':\n if prev_idx - 1 >= 0 and masked[prev_idx - 1] == '=':\n is_func = True\n func_start_idx = prev_idx - 1 \n \n # 检测方法/函数常规结构 `)` 前配对 `(`\n elif masked[prev_idx] == ')':\n p_count = 0\n open_paren_idx = -1\n for j in range(prev_idx, -1, -1):\n if masked[j] == ')':\n p_count += 1\n elif masked[j] == '(':\n p_count -= 1\n if p_count == 0:\n open_paren_idx = j\n break\n \n if open_paren_idx != -1:\n # 查找对应的小括号前是否有关联标志\n word_end = get_previous_non_space(masked, open_paren_idx)\n if word_end >= 0 and (masked[word_end].isalnum() or masked[word_end] in '_$'):\n word_start = word_end\n while word_start >= 0 and (masked[word_start].isalnum() or masked[word_start] in '_$'):\n word_start -= 1\n word = masked[word_start+1 : word_end+1]\n \n # 排除掉 if / for 等控制流结构的大括号\n if word not in ['if', 'for', 'while', 'switch', 'catch']:\n is_func = True\n func_start_idx = word_start + 1\n \n prev_word_end = get_previous_non_space(masked, word_start)\n if prev_word_end >= 0 and masked[prev_word_end].isalpha():\n pw_start = prev_word_end\n while pw_start >= 0 and masked[pw_start].isalpha():\n pw_start -= 1\n prev_word = masked[pw_start+1 : prev_word_end+1]\n \n # 再往前回溯寻找可能的修饰关键字\n if prev_word == 'function':\n func_start_idx = pw_start + 1\n # 尝试捕获开头的 async\n pp_word_end = get_previous_non_space(masked, pw_start)\n if pp_word_end >= 0 and masked[pp_word_end].isalpha():\n ppw_start = pp_word_end\n while ppw_start >= 0 and masked[ppw_start].isalpha():\n ppw_start -= 1\n pp_word = masked[ppw_start+1 : pp_word_end+1]\n if pp_word == 'async':\n func_start_idx = ppw_start + 1\n \n elif prev_word in ['async', 'get', 'set']:\n func_start_idx = pw_start + 1\n\n if is_func:\n # 统计包含自身和内部块的总换行数，计算得出总行数\n lines_count = text.count('\\n', func_start_idx, end_index) + 1\n if lines_count > 60:\n return False\n \n return True\n\n if not os.path.exists(workspace_path):\n return True\n\n # 扫描当前生成的工作目录里的所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 支持 PWA 相关的 JS、TS 以及 Python\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.py']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n # Python 文件直接由原生的 ast 进行验证\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n lines_count = node.end_lineno - node.lineno + 1\n if lines_count > 60:\n return False\n except Exception:\n pass\n elif ext in ['.html', '.htm']:\n # 提取 HTML 内部带有的 <script> 代码进行校验\n script_blocks = re.finditer(r'<script\\b[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\n for block in script_blocks:\n script_content = block.group(1)\n if not check_functions_in_js_text(script_content):\n return False\n else:\n # 其它 JS/TS 文件直接校验\n if not check_functions_in_js_text(content):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def is_prime(n: int) -> bool:\n \"\"\"判断一个数是否为质数\"\"\"\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n def extract_js_variables(code: str) -> set:\n \"\"\"通过简单的有限状态机提取 JavaScript/HTML 中的变量名、函数名与参数\"\"\"\n # 移除单行与多行注释\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除字符串字面量（安全正则）\n code = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '\"\"', code)\n code = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", \"''\", code)\n code = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '``', code)\n \n vars_set = set()\n \n # 将代码分词：保留声明关键字、合法的标识符名以及控制流程的标点\n tokens = re.finditer(r'\\b(var|let|const|function|class|catch)\\b|([a-zA-Z_$][a-zA-Z0-9_$]*)|([=,;{}()[\\]])', code)\n \n state = 'normal'\n nest_level = 0\n base_level = 0\n \n for match in tokens:\n keyword = match.group(1)\n ident = match.group(2)\n punct = match.group(3)\n \n if punct in ('(', '[', '{'):\n nest_level += 1\n \n if keyword in ('var', 'let', 'const'):\n state = 'in_decl'\n base_level = nest_level\n elif keyword in ('function', 'class'):\n state = 'in_func_name'\n base_level = nest_level\n elif keyword == 'catch':\n state = 'in_catch'\n \n elif state == 'in_decl':\n if ident:\n vars_set.add(ident)\n state = 'after_ident'\n elif punct == ';' and nest_level == base_level:\n state = 'normal'\n \n elif state == 'after_ident':\n if punct == '=':\n state = 'after_eq'\n elif punct == ',':\n state = 'in_decl'\n elif punct == ';' and nest_level == base_level:\n state = 'normal'\n \n elif state == 'after_eq':\n if punct == ',' and nest_level == base_level:\n state = 'in_decl'\n elif punct == ';' and nest_level == base_level:\n state = 'normal'\n \n elif state == 'in_func_name':\n if ident:\n vars_set.add(ident)\n state = 'after_func_name'\n elif punct == '(':\n state = 'in_params'\n \n elif state == 'after_func_name':\n if punct == '(':\n state = 'in_params'\n \n elif state == 'in_catch':\n if punct == '(':\n state = 'in_params'\n \n elif state == 'in_params':\n if ident:\n vars_set.add(ident)\n elif punct == ')':\n state = 'normal'\n \n if punct in (')', ']', '}'):\n nest_level -= 1\n if nest_level < 0:\n nest_level = 0\n \n return vars_set\n\n def extract_py_variables(code: str) -> set:\n \"\"\"通过 AST 提取 Python 中的变量名、函数名、类名及参数\"\"\"\n vars_set = set()\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name):\n if isinstance(node.ctx, ast.Store):\n vars_set.add(node.id)\n elif isinstance(node, ast.FunctionDef):\n vars_set.add(node.name)\n for arg in node.args.args:\n vars_set.add(arg.arg)\n elif isinstance(node, ast.ClassDef):\n vars_set.add(node.name)\n except Exception:\n pass\n return vars_set\n\n all_vars = set()\n \n # 遍历项目文件提取变量\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if file.endswith('.js'):\n all_vars.update(extract_js_variables(content))\n elif file.endswith('.py'):\n all_vars.update(extract_py_variables(content))\n elif file.endswith('.html') or file.endswith('.htm') or file.endswith('.vue'):\n # 提取 <script> 标签内的 JS 代码\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n for script in scripts:\n all_vars.update(extract_js_variables(script))\n \n # 没有找到任何变量的情况默认算作满足\n if not all_vars:\n return True\n \n # 验证是否每个变量名的字符长度恰好是质数\n for var in all_vars:\n if not is_prime(len(var)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义可能用于项目的常见代码文件扩展名集合\n code_exts = {\n '.js', '.css', '.html', '.htm', '.ts', '.jsx', '.tsx', \n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.cs', \n '.go', '.php', '.rb', '.swift', '.kt', '.json'\n }\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 遇到无法读取或者编码不支持的文件则跳过\n continue\n \n comments = []\n \n # 提取 /* ... */ 风格块级注释\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 提取 <!-- ... --> 风格HTML注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # 提取 // 风格单行注释（利用负向后瞻排除 // 前为冒号或单双引号的情况，防误杀 URL）\n comments.extend(re.findall(r\"(?<![:'\\\"])\\/\\/.*\", content))\n \n # 针对特定后端/脚本语言提取 # 风格单行注释及多行注释\n if ext in {'.py', '.rb', '.php', '.sh', '.yaml', '.yml'}:\n comments.extend(re.findall(r'#.*', content))\n if ext == '.py':\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n \n # 逐个检查所有提取到的注释内容，判断是否包含 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端PWA项目可能涉及的常规代码及文档文件的扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.py', '.ts', '.jsx', '.tsx'}\n\n # 确保项目目录存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按换行符分割，能够精确还原编辑器中的行数分布，包括首尾及中间的空行\n lines_count = len(content.split('\\n'))\n \n # 判断行数是否为偶数，若存在奇数行的文件则返回 False\n if lines_count % 2 != 0:\n return False\n except Exception:\n # 忽略非UTF-8编码等无法以文本模式正常读取的文件\n continue\n\n # 所有代码文件都经过检查且行数全为偶数\n return True","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”字样","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n keywords = [\"测试\", \"demo\", \"临时\", \"调试\", \"草稿\", \"待完善\"]\n # 统一转换成小写，方便应对英文字母的不同大小写情况\n keywords_lower = [k.lower() for k in keywords]\n \n # 常见的代码文件后缀集合\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', \n '.html', '.htm', '.vue', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.cs', '.go', '.php', '.rb', '.swift', \n '.kt', '.rs', '.jsonc'\n }\n \n # 构建用于提取不同类型代码注释的正则表达式\n # 1. 匹配多行注释 /* ... */\n pattern_c_multi = re.compile(r'/\\*[\\s\\S]*?\\*/')\n # 2. 匹配 HTML多行注释 <!-- ... -->\n pattern_html = re.compile(r'<!--[\\s\\S]*?-->')\n # 3. 匹配单行注释 // ...，使用 (?<!:) 避免匹配到网址 http://\n pattern_c_single = re.compile(r'(?<!:)//.*')\n # 4. 匹配脚本语言单行注释 # ...\n pattern_hash_single = re.compile(r'#.*')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n comments = []\n # 粗略提取当前文件内的所有类型的注释片段\n comments.extend(pattern_c_multi.findall(content))\n comments.extend(pattern_html.findall(content))\n comments.extend(pattern_c_single.findall(content))\n comments.extend(pattern_hash_single.findall(content))\n \n # 对每段注释内容进行敏感关键词检测\n for comment in comments:\n comment_lower = comment.lower()\n for kw in keywords_lower:\n if kw in comment_lower:\n return False\n except Exception:\n # 遇到文件读取错误（如二进制内容读取异常）时主动跳过，继续检测其他文件\n pass\n \n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"def check_equal_spaces(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端与配置代码文件扩展名\n code_extensions = {'.js', '.css', '.html', '.htm', '.json', '.ts', '.jsx', '.tsx', '.vue', '.svelte'}\n \n # 匹配独立的等号，排除 ==, ===, !=, !==, >=, <=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~= 等组合情况\n pattern = re.compile(r'(?<![=!<>+\\-*/%&|^~])=(?![=>])')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 寻找代码文件内容中所有单独的 '='\n for match in pattern.finditer(content):\n idx = match.start()\n \n # 1. 检查左侧是否严格包含两个空格\n if idx < 2:\n return False\n if content[idx-2:idx] != ' ':\n return False\n # 确保左侧第三个字符（若存在）不是空格，保证“刚好”留出两个空格\n if idx >= 3 and content[idx-3] == ' ':\n return False\n \n # 2. 检查右侧是否严格包含两个空格\n if idx + 2 >= len(content):\n return False\n if content[idx+1:idx+3] != ' ':\n return False\n # 确保右侧第三个字符（若存在）不是空格，保证“刚好”留出两个空格\n if idx + 3 < len(content) and content[idx+3] == ' ':\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":4,"instruction":"非常棒。下一步是实现核心的'书签'（收藏）功能。在首页每一条新闻卡片的右下角，以及阅读详情面板的操作区，加上一个'空心星星'图标。点击它可以收藏新闻，星星会变成实心，并伴随一个短暂的放大缩小的弹性动画。被收藏的数据要通过 localStorage 保存在本地。同时，当我们在底部导航栏切换到'书签'页面时，要展示所有已被收藏的新闻列表，并在该列表中支持点击星星取消收藏，取消时该卡片要自动从列表中移除。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”字样，替换此前的注释关键词限制要求；第二，针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义 PWA 项目及指令中可能生成的代码和文本文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.txt', '.xml', '.svg'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对代码文件进行检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出解码错误，说明存在非 UTF-8 编码的文件，不满足要求\n return False\n except Exception:\n # 忽略文件权限等其他非编码相关的异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"import os\nimport re\nimport ast\n\ndef check_python_functions(code_str):\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n end = getattr(node, 'end_lineno', start)\n if end and start and (end - start + 1) > 60:\n return False\n except Exception:\n pass\n return True\n\ndef check_functions_in_code(code_str):\n def replacer(match):\n s = match.group(0)\n # 用相等数量的换行符和空格替换掉所有的注释与字符串主体，确保代码原有的行号和字符索引稳定不变。\n return ''.join(['\\n' if c == '\\n' else ' ' for c in s])\n\n # 匹配并抹除可能造成干扰的注释和多形态的字符串字面量（包含单行、多行及模板字符串）\n pattern = re.compile(\n r'//[^\\n]*|' \n r'/\\*.*?\\*/|' \n r'\"(?:\\\\.|[^\"\\\\])*\"|' \n r\"'(?:\\\\.|[^'\\\\])*'|\" \n r'`(?:\\\\.|[^`\\\\])*`', \n re.DOTALL\n )\n \n clean_code = pattern.sub(replacer, code_str)\n \n def is_function_block(clean_code, brace_index):\n i = brace_index - 1\n while i >= 0 and clean_code[i].isspace():\n i -= 1\n if i < 0: return False\n \n # 箭头函数\n if i >= 1 and clean_code[i-1:i+1] == '=>':\n return True\n \n if clean_code[i] == ')':\n paren_count = 1\n i -= 1\n while i >= 0 and paren_count > 0:\n if clean_code[i] == ')': paren_count += 1\n elif clean_code[i] == '(': paren_count -= 1\n i -= 1\n \n if paren_count > 0: return False\n \n while i >= 0 and clean_code[i].isspace():\n i -= 1\n \n word = \"\"\n while i >= 0 and (clean_code[i].isalnum() or clean_code[i] in '_$'):\n word = clean_code[i] + word\n i -= 1\n \n # 基础函数块 function(...) {\n if word == \"function\":\n return True\n \n # 方法或对象属性函数\n if word and word not in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\n j = i\n while j >= 0 and (clean_code[j].isspace() or clean_code[j] == '*'):\n j -= 1\n prev_word = \"\"\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in '_$'):\n prev_word = clean_code[j] + prev_word\n j -= 1\n if prev_word == \"function\":\n return True\n return True\n \n # 匿名函数\n if not word:\n j = i\n while j >= 0 and (clean_code[j].isspace() or clean_code[j] == '*'):\n j -= 1\n prev_word = \"\"\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in '_$'):\n prev_word = clean_code[j] + prev_word\n j -= 1\n if prev_word == \"function\":\n return True\n \n return False\n\n stack = []\n line_number = 1\n \n # 逐字遍历，借助堆栈准确匹配嵌套的函数大括号层级\n for i, char in enumerate(clean_code):\n if char == '\\n':\n line_number += 1\n elif char == '{':\n is_func = is_function_block(clean_code, i)\n stack.append((is_func, line_number))\n elif char == '}':\n if stack:\n is_func, start_line = stack.pop()\n if is_func:\n func_length = line_number - start_line + 1\n if func_length > 60:\n return False\n return True\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # Python后端文件处理\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_python_functions(content):\n return False\n except Exception:\n pass\n \n # 含有大括号语法的前端主业务文件及相关处理\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.java', '.cpp', '.c', '.cs', '.php', '.html', '.vue')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 剥离出含有 HTML 的网页组件中嵌有的 <script> 代码块\n if file.endswith(('.html', '.vue')):\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_functions_in_code(script):\n return False\n else:\n if not check_functions_in_code(content):\n return False\n except Exception:\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 辅助函数：判断一个整数是否是质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 辅助函数：根据文件扩展名提取文件中定义的变量名\n def extract_variables(ext: str, content: str) -> set:\n variables = set()\n \n if ext in ['.js', '.ts']:\n # 1. 匹配常规的 var / let / const 声明\n pattern = r'\\b(?:let|var|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b'\n for match in re.finditer(pattern, content):\n variables.add(match.group(1))\n \n # 2. 匹配数组解构赋值\n for match in re.finditer(r'\\b(?:let|var|const)\\s+\\[(.*?)\\]', content):\n for p in match.group(1).split(','):\n p = p.strip()\n if re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)$', p):\n variables.add(p)\n \n # 3. 匹配对象解构赋值\n for match in re.finditer(r'\\b(?:let|var|const)\\s+\\{(.*?)\\}', content):\n for p in match.group(1).split(','):\n p = p.strip()\n if ':' in p:\n p = p.split(':')[1].strip()\n if re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)$', p):\n variables.add(p)\n \n # 4. 匹配函数参数\n func_pattern = r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)'\n for match in re.finditer(func_pattern, content):\n for arg in match.group(1).split(','):\n arg = arg.strip()\n if re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)$', arg):\n variables.add(arg)\n \n elif ext in ['.py']:\n # 匹配 Python 简单赋值变量\n pattern = r'^[\\t ]*([a-zA-Z_][a-zA-Z0-9_]*)\\s*(?::[^=]+)?=\\s*'\n for match in re.finditer(pattern, content, re.MULTILINE):\n variables.add(match.group(1))\n \n elif ext in ['.html', '.htm']:\n # 提取 HTML 文件中 script 标签内的代码继续按 js 解析\n script_pattern = r'<script.*?>([\\s\\S]*?)</script>'\n for match in re.finditer(script_pattern, content, re.IGNORECASE):\n variables.update(extract_variables('.js', match.group(1)))\n \n return variables\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目文件提取并验证变量长度\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 只关心包含代码文件类型的扩展名\n if ext not in ['.js', '.ts', '.html', '.htm', '.py']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 获取该文件中所有的提取变量\n variables = extract_variables(ext, content)\n \n for var in variables:\n if not is_prime(len(var)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对不同语言类型的注释正则表达式\n js_css_pattern = re.compile(r'//.*|/\\*[\\s\\S]*?\\*/')\n html_pattern = re.compile(r'<!--[\\s\\S]*?-->')\n py_pattern = re.compile(r'#.*')\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 对文件类型进行分类\n is_js_css = ext in ['.js', '.ts', '.css', '.scss', '.sass', '.less']\n is_html = ext in ['.html', '.htm']\n is_mixed = ext in ['.vue', '.jsx', '.tsx']\n is_py = ext in ['.py']\n \n # 若并非可能包含代码的文件格式，则跳过\n if not (is_js_css or is_html or is_mixed or is_py):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略不可读或无法使用 utf-8 解码的文件\n continue\n \n comments = []\n if is_js_css:\n comments.extend(js_css_pattern.findall(content))\n elif is_html:\n comments.extend(html_pattern.findall(content))\n elif is_mixed:\n comments.extend(js_css_pattern.findall(content))\n comments.extend(html_pattern.findall(content))\n elif is_py:\n comments.extend(py_pattern.findall(content))\n \n # 遍历所有提取到的注释文本\n for comment in comments:\n for char in comment:\n # 如果任意注释的字符被判定为 emoji，则返回 False\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令上下文，推断出涉及到的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.jsx', '.tsx'}\n \n # 确保给定的工作区路径存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前项目文件夹下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 提取文件扩展名并转小写\n ext = os.path.splitext(file_name)[1].lower()\n \n # 如果是代码文件，则进行行数验证\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 计入空行的前提下，总行数如果为奇数，则不满足要求\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果遇到非文本内容或编码不正确的文件，安全地略过\n continue\n\n # 所有的代码文件检查均通过（或者是并未生成文件），则满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”字样","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 待检查的常见前端及后端代码文件扩展名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\n '.html', '.htm', '.css', '.scss', '.less',\n '.py', '.java', '.cpp', '.c', '.h', '.hpp', '.cs',\n '.php', '.rb', '.go', '.rs', '.swift', '.kt'\n }\n \n # 禁用关键词列表（均统一为小写以便不区分大小写匹配）\n forbidden_words = [\"测试\", \"demo\", \"临时\", \"调试\", \"草稿\", \"待完善\", \"todo\", \"待优化\"]\n \n # 构建正则表达式模式，利用捕获组分别匹配字符串和不同类型的注释\n # string组: 匹配双引号、单引号、反引号括起来的字符串（避免将字符串内的字词当作注释误判）\n # m_comment组: 匹配多行注释 /* ... */ 或 HTML注释 <!-- ... -->\n # s_comment组: 匹配单行注释 // ... 或 Python等脚本类注释 # ...\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<m_comment>/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->)'\n r'|(?P<s_comment>//.*|#.*)',\n re.MULTILINE\n )\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n for match in pattern.finditer(content):\n # 判断匹配项是否为多行或单行注释\n comment = match.group('m_comment') or match.group('s_comment')\n if comment:\n comment_lower = comment.lower()\n # 逐个检查违规词是否存在于当前注释中\n for word in forbidden_words:\n if word in comment_lower:\n return False\n except Exception:\n # 忽略不可读或无权限的文件\n continue\n \n # 所有文件验证完毕，均无违规关键词则返回True\n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 可能的代码文件后缀集合（针对 PWA 项目）\n code_extensions = {'.html', '.css', '.js', '.json', '.ts', '.jsx', '.tsx', '.webmanifest', '.vue', '.svelte'}\n has_checked_files = False\n \n # 复合等号左侧可能出现的关联字符，涵盖 JS、CSS 常见操作符\n left_chars = set('<>!=+-*/%&|^~?$')\n # 复合等号右侧可能出现的关联字符\n right_chars = set('=>')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_checked_files = True\n \n for line in lines:\n line = line.rstrip('\\n\\r')\n i = 0\n while i < len(line):\n if line[i] == '=':\n is_single_equal = True\n \n # 判定是否属于复合操作符的一部分\n if i > 0 and line[i-1] in left_chars:\n is_single_equal = False\n if i < len(line) - 1 and line[i+1] in right_chars:\n is_single_equal = False\n \n if is_single_equal:\n # 检查左侧是否正好有两个空格\n left_ok = False\n if i >= 2:\n if line[i-1] == ' ' and line[i-2] == ' ':\n # 确保第三个字符不是空格（或已到达行首）\n if i == 2 or line[i-3] != ' ':\n left_ok = True\n \n # 检查右侧是否正好有两个空格\n right_ok = False\n if i + 2 < len(line):\n if line[i+1] == ' ' and line[i+2] == ' ':\n # 确保第三个字符不是空格（或已到达行尾）\n if i + 3 == len(line) or line[i+3] != ' ':\n right_ok = True\n \n # 只要发现任何一个单独的等号前后不满足恰好留有两个空格的要求，直接返回 False\n if not (left_ok and right_ok):\n return False\n i += 1\n \n return has_checked_files","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":5,"instruction":"接下来给应用加点质感，引入'深色模式'。在顶部 App Bar 的右侧添加一个太阳/月亮的切换图标。点击该图标后，整个应用（包括主页列表、底栏、详情弹窗面板等所有元素）平滑过渡切换为深色主题。你需要精心调整深色背景下的文字颜色、卡片底色以及边框，确保对比度舒适。并且，必须把用户选择的主题模式保存到本地，页面刷新后依然能记住用户的选择。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json，替换此前的修改计划记录文件要求；第二，所有生成的代码文件中的注释必须全部使用中文。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断 PWA 前端项目及修改计划中可能涉及的代码和文本文件后缀名\n code_extensions = {'.html', '.htm', '.css', '.js', '.json', '.md'}\n \n if not os.path.exists(workspace_path):\n # 如果未生成任何文件，则没有违反“所有生成的文件必须是UTF-8”的条件\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出解码错误，说明存在非 UTF-8 编码的代码文件，不满足要求\n return False\n except Exception:\n # 如果由于其他异常（如权限问题）无法打开，这里选择忽略或继续\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_js_functions(script: str) -> bool:\n # 定义正则表达式匹配：双引号字符串、单引号字符串、反引号字符串、多行注释、单行注释\n pattern = r'(\"(?:\\\\\"|[^\"])*\")|(\\'(?:\\\\\\'|[^\\'])*\\')|(`(?:\\\\`|[^`])*`)|(/\\*[\\s\\S]*?\\*/)|(//[^\\n]*)'\n \n def replacer(match):\n if match.group(1) or match.group(2) or match.group(3):\n # 如果是字符串内容，用 X 替换所有的 {} 防止干扰匹配，同时不改变文本行数\n return match.group(0).replace('{', 'X').replace('}', 'X')\n elif match.group(4):\n # 如果是多行注释，替换为空换行符，确保总行数依然对齐\n return '\\n' * match.group(4).count('\\n')\n elif match.group(5):\n # 如果是单行注释，直接移除\n return ''\n return match.group(0)\n\n try:\n script = re.sub(pattern, replacer, script)\n except Exception:\n pass\n\n lines = script.split('\\n')\n stack = []\n\n for i, line in enumerate(lines):\n idx = 0\n while idx < len(line):\n char = line[idx]\n if char == '{':\n # 当遇到 { 时，获取它之前的上下文文本（追溯最多5行）\n start_line_idx = max(0, i - 5)\n prev_text_lines = lines[start_line_idx:i]\n prev_text_lines.append(line[:idx])\n prev_text = \" \".join(prev_text_lines).strip()\n\n is_func = False\n # 判断之前的字符串是否具备典型的JS函数声明或箭头函数特征\n if prev_text.endswith('=>') or prev_text.endswith('function'):\n is_func = True\n else:\n match = re.search(r'([\\w$]+)\\s*\\([^)]*\\)\\s*$', prev_text)\n if match:\n kw = match.group(1)\n # 过滤掉常见的控制流语句关键字\n if kw not in ('if', 'while', 'for', 'switch', 'catch'):\n is_func = True\n else:\n if re.search(r'\\bfunction\\s*$', prev_text):\n is_func = True\n \n # 兜底：如果以 ) 结尾并开启 {}，基本可以认为是一个逻辑代码块或者函数\n if not is_func and prev_text.endswith(')'):\n is_func = True\n\n stack.append((i, is_func))\n elif char == '}':\n if stack:\n start_idx, is_func = stack.pop()\n if is_func:\n # 判断该函数的行数是否大于 60\n if i - start_idx + 1 > 60:\n return False\n idx += 1\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 解析 Python AST，找出函数定义\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\n if node.end_lineno - node.lineno + 1 > 60:\n return False\n except Exception:\n pass\n elif file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_js_functions(content):\n return False\n except Exception:\n pass\n elif file.endswith('.html') or file.endswith('.htm'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 提取 HTML 中的所有 <script> 代码块\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for script in scripts:\n if not check_js_functions(script):\n return False\n except Exception:\n pass\n\n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"import os\nimport re\n\ndef check_variable_name_length(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 正则表达式：匹配 JS/TS/HTML 中的变量、函数和类声明\n # 仅提取使用 var, let, const, function, class 关键字进行直接声明的标识符\n # 这样能有效避免提取到系统的内置属性、已有的 API 及方法名，从而防止误判\n pattern = re.compile(r'\\b(?:var|let|const|function|class)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n\n # 需要检查的前端代码文件后缀名\n target_extensions = ('.js', '.ts', '.html', '.htm', '.vue', '.jsx', '.tsx')\n\n # 遍历工作空间目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(target_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找当前文件内所有声明匹配到的变量名\n matches = pattern.findall(content)\n for var_name in matches:\n # 验证该变量名的字符长度是否为质数\n if not is_prime(len(var_name)):\n return False\n except Exception:\n # 若文件由于编码等不可预见的原因无法读取，则跳过\n continue\n\n # 如果所有提取到的变量名的字符长度均正好是质数，或者未提取到变量名，则判定为满足要求\n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 穷举推断可能的代码相关文件后缀，适用于PWA及常见Web项目类型\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.md'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n # 按UTF-8要求进行解码读取\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略读取失败的非文本文件或错误编码文件\n continue\n \n comments = []\n \n # 1. 匹配 HTML 注释 (跨行)\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\n \n # 2. 匹配 多行/块注释 (CSS, JS, TS等，跨行)\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, flags=re.DOTALL))\n \n # 3. 匹配 JS/TS 单行注释，利用 (?<!:) 排除如 http:// 这类 URL 的干扰\n comments.extend(re.findall(r'(?<!:)//(.*?)$', content, flags=re.MULTILINE))\n \n # 4. 匹配 传统 # 的单行注释 (应对潜在的其他生成文件)\n comments.extend(re.findall(r'#(.*?)$', content, flags=re.MULTILINE))\n \n # 验证提取出的所有注释是否存在 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"import os\n\ndef check_even_line_counts(response: str, workspace_path: str) -> bool:\n # 如果工作区不存在，由于没有文件被创建，不会违反要求\n if not os.path.exists(workspace_path):\n return True\n \n # 定义常见的前端、配置及项目记录文件扩展名\n target_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.jsx', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除非用户主动创建的第三方依赖或系统目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 只针对代码文件及项目相关的文本文件\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 计入空行在内的总行数必须为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果文件不可读（比如存在非utf-8编码问题），跳过该文件继续验证\n pass\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”字样","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义禁止出现在注释中的词汇\n forbidden_words = [\"测试\", \"demo\", \"临时\", \"调试\", \"草稿\", \"待完善\", \"todo\", \"待优化\"]\n forbidden_words_lower = [w.lower() for w in forbidden_words]\n \n # 匹配各类代码文件中常见注释的正则表达式\n comment_patterns = [\n r'//.*', # JS/TS/CSS 等语言的单行注释\n r'/\\*[\\s\\S]*?\\*/', # JS/TS/CSS 等语言的多行注释\n r'<!--[\\s\\S]*?-->', # HTML 的注释\n r'#.*', # Shell/Python 的单行注释\n r\"'''[\\s\\S]*?'''\", # Python 的多行注释\n r'\"\"\"[\\s\\S]*?\"\"\"' # Python 的多行注释\n ]\n # 合并成一个正则匹配所有注释\n pattern = re.compile('|'.join(comment_patterns))\n \n # 需要检查的常见代码文件扩展名\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.json', '.md', '.py'}\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 提取该文件中的所有注释部分\n comments = pattern.findall(content)\n \n # 检查提取到的每段注释中是否包含禁用关键词\n for comment in comments:\n comment_lower = comment.lower()\n for word in forbidden_words_lower:\n if word in comment_lower:\n return False\n except Exception:\n # 如果遇到编码错误或不可读情况，直接忽略该文件并继续\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 确保工作区路径存在\n if not os.path.exists(workspace_path):\n return False\n\n # 定义常见的代码文件后缀\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.json', '.vue', '.svelte', '.py', '.java', '.c', '.cpp', \n '.h', '.cs', '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.dart'\n }\n \n has_code_file = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n file_content = f.read()\n except Exception:\n continue # 如果存在读取异常，跳过该文件\n \n n = len(file_content)\n for i in range(n):\n if file_content[i] == '=':\n # 检查是否为复合运算符（或箭头函数）的一部分\n is_compound = False\n \n # 检查左侧是否为常见复合操作符的前缀符号\n if i > 0 and file_content[i-1] in '=!<>-+*/%&|^~$:':\n is_compound = True\n \n # 检查右侧是否为相等判断或箭头函数后缀\n if i < n - 1 and file_content[i+1] in '=>':\n is_compound = True\n \n if not is_compound:\n # 检查左侧是否恰好有两个空格\n left_valid = False\n if i >= 2 and file_content[i-1] == ' ' and file_content[i-2] == ' ':\n if i >= 3:\n if file_content[i-3] != ' ':\n left_valid = True\n else:\n # 等号前仅有两个字符且均为空格（即文件开头）\n left_valid = True\n \n # 检查右侧是否恰好有两个空格\n right_valid = False\n if i <= n - 3 and file_content[i+1] == ' ' and file_content[i+2] == ' ':\n if i <= n - 4:\n if file_content[i+3] != ' ':\n right_valid = True\n else:\n # 等号后仅有两个字符且均为空格（即文件结尾）\n right_valid = True\n \n # 如果左侧或右侧任一不满足恰好留出两个空格，立刻返回 False\n if not (left_valid and right_valid):\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"所有生成的代码文件中的注释必须全部使用中文","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":6,"instruction":"交互体验还能继续优化。给主页的新闻列表加上'下拉刷新'的功能：用户在列表顶部继续向下滑动时，要出现一个带有旋转动画的刷新指示图标，松手后模拟网络加载2秒钟，然后更新顶部的第一条新闻数据。另外，增加'无限滚动'体验：拉到页面最底部时，要自动触发加载更多新闻的功能，底部要显示一个'正在加载更多...'的动画和文字提示，加载完成后把新数据追加到列表最后。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样，替换此前的注释关键词限制要求；第二，限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目背景，定义常见的代码及配置文本文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.txt', '.ts'}\n \n if not os.path.exists(workspace_path):\n # 如果路径不存在，视作没有违反约束（因没有生成代码文件）\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制模式读取文件\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试使用 utf-8 解码\n raw_data.decode('utf-8')\n \n except UnicodeDecodeError:\n # 如果解码报错，说明存在非 UTF-8 编码的代码文件\n return False\n except Exception:\n # 对于其他的异常（例如权限问题等），暂且忽略继续检查\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n import ast\n\n def is_function_context(context: str) -> bool:\n \"\"\"根据大括号前方的上下文，判断当前块是否属于一个函数定义\"\"\"\n context = context.strip()\n if not context:\n return False\n \n last_char = context[-1]\n \n # 明确不可能作为函数定义大括号前驱的字符（排除了传参对象字面量、数组等情况）\n if last_char in ('=', '(', ',', '[', ':', '?', '+', '-', '*', '/', '&', '|', '!', '<', '.', ';'):\n return False\n \n # 如果上下文中包含箭头符号，多为箭头函数\n if '=>' in context:\n return True\n \n # 包含 function 关键字\n if re.search(r'\\bfunction\\b', context):\n return True\n \n # 正则匹配常规的形如 methodName(...) 或者 methodName(...): ReturnType 的结构\n try:\n matches = list(re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:<[^>]+>\\s*)?\\(', context))\n if matches:\n # 排除可能带括号且紧接大括号的控制流等关键字\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with', 'return', 'typeof', 'instanceof'}\n last_match = matches[-1].group(1)\n if last_match not in keywords:\n return True\n except re.error:\n pass\n\n return False\n\n def check_js_functions_length(code: str, max_lines: int) -> bool:\n \"\"\"检查一段 JS/TS 代码内的函数长度是否在限制内\"\"\"\n in_string = False\n string_char = ''\n in_single_comment = False\n in_multi_comment = False\n in_template_string = False\n \n blocks = []\n stack = []\n \n i = 0\n length = len(code)\n \n # 建立行号映射表，便于快速从字符索引转化为行号\n lines_start = [0]\n for idx, char in enumerate(code):\n if char == '\\n':\n lines_start.append(idx + 1)\n \n def get_line_number(index: int) -> int:\n return bisect.bisect_right(lines_start, index)\n \n while i < length:\n char = code[i]\n \n # 单行注释处理\n if in_single_comment:\n if char == '\\n':\n in_single_comment = False\n i += 1\n continue\n \n # 多行注释处理\n if in_multi_comment:\n if char == '*' and i + 1 < length and code[i+1] == '/':\n in_multi_comment = False\n i += 2\n else:\n i += 1\n continue\n \n # 模板字符串处理\n if in_template_string:\n if char == '\\\\':\n i += 2\n continue\n if char == '`':\n in_template_string = False\n i += 1\n continue\n \n # 普通字符串处理\n if in_string:\n if char == '\\\\':\n i += 2\n continue\n if char == string_char:\n in_string = False\n i += 1\n continue\n \n # 检查注释开头\n if char == '/' and i + 1 < length:\n if code[i+1] == '/':\n in_single_comment = True\n i += 2\n continue\n if code[i+1] == '*':\n in_multi_comment = True\n i += 2\n continue\n \n # 检查字符串开头\n if char in (\"'\", '\"'):\n in_string = True\n string_char = char\n i += 1\n continue\n \n # 检查模板字符串开头\n if char == '`':\n in_template_string = True\n i += 1\n continue\n \n # 捕获代码块大括号层级\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start_i = stack.pop()\n blocks.append((start_i, i))\n \n i += 1\n \n # 遍历所有捕获到的代码块，对超出规定行数的进行函数特征验证\n for start_i, end_i in blocks:\n start_line = get_line_number(start_i)\n end_line = get_line_number(end_i)\n \n lines_count = end_line - start_line + 1\n if lines_count > max_lines:\n lookback_start = max(0, start_i - 300)\n pre_text = code[lookback_start:start_i]\n # 用分号和闭合花括号做切分，获取紧挨大括号前面的完整上下文特征\n parts = re.split(r'[;}{]', pre_text)\n context = parts[-1] if parts else ''\n \n # 若经判断确为函数，代表该函数超出行数限制\n if is_function_context(context):\n return False\n \n return True\n\n def check_python_functions_length(code: str, max_lines: int) -> bool:\n \"\"\"检查 Python 代码文件的函数长度（借助 AST）\"\"\"\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n end = node.end_lineno\n else:\n end = max((getattr(n, 'lineno', start) for n in ast.walk(node)), default=start)\n if end - start + 1 > max_lines:\n return False\n except SyntaxError:\n pass\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前代码项目文件，基于不同后缀作相应分析处理\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext in ['.js', '.ts', '.jsx', '.tsx']:\n if not check_js_functions_length(content, 60):\n return False\n elif ext in ['.html', '.htm', '.vue']:\n script_contents = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for sc in script_contents:\n if not check_js_functions_length(sc, 60):\n return False\n elif ext == '.py':\n if not check_python_functions_length(content, 60):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断数字是否为质数\n def is_prime(n):\n if n < 2:\n return False\n if n in (2, 3):\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 辅助函数：通过正则表达式提取前端 JS/TS 中的变量名\n def extract_js_vars(code):\n vars_set = set()\n # 清除注释避免正则误匹配\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n # 1. 匹配 var/let/const 声明的单变量\n for match in re.finditer(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', code):\n vars_set.add(match.group(1))\n \n # 2. 匹配 function 声明的函数名\n for match in re.finditer(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', code):\n vars_set.add(match.group(1))\n \n # 3. 匹配 function() 中的参数\n for match in re.finditer(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', code):\n params = match.group(1).split(',')\n for p in params:\n p = p.strip()\n if p:\n m = re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\n if m:\n vars_set.add(m.group(1))\n \n # 4. 匹配箭头函数多参数形式 (a, b) =>\n for match in re.finditer(r'\\(([^)]*)\\)\\s*=>', code):\n params = match.group(1).split(',')\n for p in params:\n p = p.strip()\n if p:\n m = re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\n if m:\n vars_set.add(m.group(1))\n \n # 5. 匹配箭头函数单参数形式 a =>\n for match in re.finditer(r'(?:^|[^a-zA-Z0-9_$])([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', code):\n vars_set.add(match.group(1))\n \n return vars_set\n\n # 辅助函数：通过 AST 提取 Python 中的变量名\n def extract_py_vars(code):\n vars_set = set()\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name):\n if isinstance(node.ctx, ast.Store):\n vars_set.add(node.id)\n elif isinstance(node, ast.arg):\n vars_set.add(node.arg)\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):\n vars_set.add(node.name)\n except Exception:\n pass\n return vars_set\n\n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n vars_in_file = set()\n if ext == '.py':\n vars_in_file = extract_py_vars(content)\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n vars_in_file = extract_js_vars(content)\n elif ext in ['.html', '.htm']:\n for script_match in re.finditer(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE):\n script_content = script_match.group(1)\n vars_in_file.update(extract_js_vars(script_content))\n \n # 校验每一个提取出来的变量名长度\n for v in vars_in_file:\n v_clean = v.strip()\n if not is_prime(len(v_clean)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def get_comments(content, ext):\n comments = []\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.json']:\n # 提取多行注释 /* ... */\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 提取单行注释 // ... ，排除了 http:// 这种常见的误报\n comments.extend(re.findall(r'(?<!:)//.*', content))\n elif ext in ['.html', '.htm', '.vue']:\n # 提取 HTML 注释 <!-- ... -->\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # HTML 文件内的 script 或 style 标签也可能包含 JS/CSS 注释\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n comments.extend(re.findall(r'(?<!:)//.*', content))\n elif ext in ['.py', '.rb', '.sh']:\n # 提取脚本语言的单行注释 # ...\n comments.extend(re.findall(r'#.*', content))\n # 简单兼容 Python 的多行注释字符串\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n return comments\n\n valid_exts = [\n '.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.json',\n '.html', '.htm', '.vue',\n '.py', '.rb', '.sh'\n ]\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 获取该文件内所有的注释内容列表\n comments = get_comments(content, ext)\n \n # 遍历每段注释，检查是否有字符被判断为 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n # 如果发现注释中含有 emoji，直接判定为不满足要求\n return False\n \n # 检查完所有文件，均未在注释中发现 emoji，返回 True 满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_even_line_count(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能的代码文件或文本数据类型后缀\n valid_extensions = {'.html', '.css', '.js', '.json', '.md'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 计算包含空行在内的总行数\n if not content:\n line_count = 0\n else:\n line_count = len(content.split('\\n'))\n \n # 如果有任何一个代码文件的行数不是偶数，则要求未被满足\n if line_count % 2 != 0:\n return False\n except Exception:\n # 对于不可读或非规范的文本文件跳过处理\n continue\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义禁止在注释中出现的关键词列表\n forbidden_keywords = [\n \"测试\", \"demo\", \"临时\", \"调试\", \"草稿\", \n \"待完善\", \"TODO\", \"待优化\", \"废弃\", \"占位\"\n ]\n # 转为全小写以便进行大小写不敏感匹配\n forbidden_lower = [kw.lower() for kw in forbidden_keywords]\n\n if not os.path.exists(workspace_path):\n return True\n\n # 正则表达式：用于提取 JS/TS/CSS 风格文件中的注释\n # 第 1 部分（不捕获）：匹配单双引号及反引号字符串，防止误伤字符串内的 // 或 /*\n # 第 2 部分（捕获组 1）：匹配单行注释 //，加上 (?<!:) 是为了避免匹配到 http:// 等 URL\n # 第 3 部分（捕获组 2）：匹配多行注释 /* ... */\n pattern_c_style = re.compile(\n r'(?:\"(?:\\\\\"|[^\"])*\"|\\'(?:\\\\\\'|[^\\'])*\\'|`(?:\\\\`|[\\s\\S])*?`)|((?<!:)//.*?$)|(/\\*[\\s\\S]*?\\*/)', \n re.MULTILINE\n )\n \n # 正则表达式：用于提取 HTML 风格的注释\n pattern_html = re.compile(r'<!--[\\s\\S]*?-->')\n \n # 正则表达式：用于提取 Python/Shell 风格文件中的注释\n pattern_py = re.compile(\n r'(?:\"(?:\\\\\"|[^\"])*\"|\\'(?:\\\\\\'|[^\\'])*\\'|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\')|(#.*?$)', \n re.MULTILINE\n )\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉无需检查的常见系统目录及依赖目录\n for ignore_dir in ['.git', 'node_modules', '__pycache__', 'dist', 'build']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n \n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue # 如果文件无法读取（例如编码问题或权限不足）则跳过\n\n comments = []\n\n # 对于前端及其他常见 C 风格注释文件\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm', '.vue']:\n # 若为 HTML 类文件，先提取出 <!-- --> 注释\n if ext in ['.html', '.htm', '.vue']:\n for m in pattern_html.finditer(content):\n comments.append(m.group(0))\n \n # HTML 中也可能内联了 <script> 和 <style>，因此统一再使用 C 风格匹配\n for m in pattern_c_style.finditer(content):\n if m.group(1): # 捕获到单行注释\n comments.append(m.group(1))\n elif m.group(2): # 捕获到多行注释\n comments.append(m.group(2))\n \n # 对于 Python 或 Shell 等脚本文件\n elif ext in ['.py', '.sh']:\n for m in pattern_py.finditer(content):\n if m.group(1): # 捕获到 # 号注释\n comments.append(m.group(1))\n\n # 检查提取到的所有注释文本\n for comment in comments:\n comment_lower = comment.lower()\n for kw in forbidden_lower:\n if kw in comment_lower:\n # 一旦在任意一条注释里发现任意一个限制关键词，立刻返回False\n return False\n\n # 若所有文件的全部注释均未查到限制关键词，则满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对PWA前端应用，定义常见的代码文件扩展名集合\n valid_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue', '.webmanifest'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录（例如 .git, .vscode 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 遇到无法使用UTF-8读取的文件直接跳过（该要求与空格无关，防止阻塞验证）\n continue\n \n for line in lines:\n # 仅去除尾部的回车及换行符，保留可能存在的尾随空格\n line = line.rstrip('\\n\\r')\n \n for idx, char in enumerate(line):\n if char == '=':\n is_combined = False\n \n # 检查左侧是否为组合符号的一部分 (涵盖 JS/CSS 等常见组合等号符)\n if idx > 0 and line[idx-1] in '=!<>+-*/%&|^~$':\n is_combined = True\n \n # 检查右侧是否为组合符号的一部分 (如 ==, ===, =>)\n if idx < len(line) - 1 and line[idx+1] in '=>':\n is_combined = True\n \n # 当它是一个独立的 '=' 时，验证两边是否恰好有两个空格\n if not is_combined:\n # 检查左侧\n left_ok = False\n if idx >= 2 and line[idx-1] == ' ' and line[idx-2] == ' ':\n # 确保前第三个字符不是空格，或者等号正好处于只有左边两个空格的行首\n if idx == 2 or line[idx-3] != ' ':\n left_ok = True\n \n # 检查右侧\n right_ok = False\n if idx <= len(line) - 3 and line[idx+1] == ' ' and line[idx+2] == ' ':\n # 确保后第三个字符不是空格，或者等号正好处于只有右边两个空格的行尾\n if idx == len(line) - 3 or line[idx+3] != ' ':\n right_ok = True\n \n # 如果有任何一端不满足要求，就视为不合规\n if not (left_ok and right_ok):\n return False\n \n # 只要包含了相关代码文件并且全部校验通过即返回True；如果是空目录则返回False\n return has_code_file","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"所有生成的代码文件中的注释必须全部使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，该PWA应用涉及的核心代码文件类型为主流前端文件格式\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果代码文件无法正常以 UTF-8 读取，或者出现其他IO异常，直接视为不满足\n return False\n \n # 收集文件中所有可能被识别为注释的内容\n comments = []\n \n # 1. HTML / SVG 风格注释 <!-- ... -->\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 2. C / JS / CSS 风格多行注释 /* ... */\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n # 3. JS / TS 风格单行注释 // ...\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\n \n # 将所有提取到的注释拼接在一起\n all_comments = \" \".join(comments)\n \n # 检查注释部分是否包含指定的字样\n if \"PWA新闻聚合器\" not in all_comments:\n return False\n \n return has_code_file","tags":["Content","Keyword"]}]},{"round_id":7,"instruction":"为了更方便找新闻，需要增加搜索功能。在顶部 App Bar 标题旁边放一个搜索图标（放大镜），点击后应用标题变成一个输入框。用户输入关键字时，下方实时过滤当前的新闻列表数据进行展示。重点要求：在展现的搜索结果卡片中，无论是新闻标题还是摘要里，只要匹配到用户输入的搜索关键词，就必须用醒目的背景色（比如亮黄色）将其高亮显示出来。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md，替换此前的修改计划记录文件要求；第二，要求输出代码时，统一使用4个空格作为缩进标准。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 定义前端项目中可能生成的代码文件后缀名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.json', '.md', \n '.ts', '.jsx', '.tsx', '.vue', '.txt', '.xml', \n '.svg', '.webmanifest', '.mjs', '.cjs'\n }\n \n # 如果工作区路径不存在，从严谨角度认为没有违反编码要求\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅判断代码和文本相关文件，排除图片等二进制文件的干扰\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 格式读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件并非合法的 UTF-8 编码\n return False\n except Exception:\n # 其他读取异常（如权限等非编码导致的问题）忽略不计\n pass\n \n # 所有被找到的代码文件均成功以 UTF-8 读取，返回 True\n return True","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n \n def clean_js_code(code: str) -> str:\n \"\"\"\n 利用状态机清洗JS代码，将注释、字符串字面量均替换为空格，但保留换行符，\n 保证返回的代码在字符和行数位置上与原代码一一对应，同时排除大括号误判。\n \"\"\"\n state_stack = ['NORMAL']\n result = []\n i = 0\n n = len(code)\n while i < n:\n current_state = state_stack[-1]\n \n if current_state in ('NORMAL', 'TEMPLATE_EXPR'):\n if i + 1 < n and code[i:i+2] == '//':\n state_stack.append('LINE_COMMENT')\n result.append(' ')\n i += 2\n elif i + 1 < n and code[i:i+2] == '/*':\n state_stack.append('BLOCK_COMMENT')\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n state_stack.append('STRING_D')\n result.append(' ')\n i += 1\n elif code[i] == \"'\":\n state_stack.append('STRING_S')\n result.append(' ')\n i += 1\n elif code[i] == '`':\n state_stack.append('STRING_T')\n result.append(' ')\n i += 1\n elif current_state == 'TEMPLATE_EXPR' and code[i] == '}':\n state_stack.pop()\n result.append(' ')\n i += 1\n elif current_state == 'TEMPLATE_EXPR' and code[i] == '{':\n state_stack.append('BLOCK')\n result.append(code[i])\n i += 1\n elif current_state == 'NORMAL' and code[i] == '{':\n result.append(code[i])\n i += 1\n elif current_state == 'NORMAL' and code[i] == '}':\n result.append(code[i])\n i += 1\n else:\n result.append(code[i])\n i += 1\n \n elif current_state == 'BLOCK':\n if code[i] == '}':\n state_stack.pop()\n result.append(code[i])\n i += 1\n elif code[i] == '{':\n state_stack.append('BLOCK')\n result.append(code[i])\n i += 1\n elif i + 1 < n and code[i:i+2] == '//':\n state_stack.append('LINE_COMMENT')\n result.append(' ')\n i += 2\n elif i + 1 < n and code[i:i+2] == '/*':\n state_stack.append('BLOCK_COMMENT')\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n state_stack.append('STRING_D')\n result.append(' ')\n i += 1\n elif code[i] == \"'\":\n state_stack.append('STRING_S')\n result.append(' ')\n i += 1\n elif code[i] == '`':\n state_stack.append('STRING_T')\n result.append(' ')\n i += 1\n else:\n result.append(code[i])\n i += 1\n \n elif current_state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state_stack.pop()\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n \n elif current_state == 'BLOCK_COMMENT':\n if i + 1 < n and code[i:i+2] == '*/':\n state_stack.pop()\n result.append(' ')\n i += 2\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n \n elif current_state == 'STRING_D':\n if code[i] == '\\\\':\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n state_stack.pop()\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n \n elif current_state == 'STRING_S':\n if code[i] == '\\\\':\n result.append(' ')\n i += 2\n elif code[i] == \"'\":\n state_stack.pop()\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n \n elif current_state == 'STRING_T':\n if code[i] == '\\\\':\n result.append(' ')\n i += 2\n elif i + 1 < n and code[i:i+2] == '${':\n state_stack.append('TEMPLATE_EXPR')\n result.append(' ')\n i += 2\n elif code[i] == '`':\n state_stack.pop()\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n\n def is_js_function_block(cleaned_code: str, start_idx: int) -> bool:\n \"\"\"\n 判断匹配到的 { } 块是否为函数\n \"\"\"\n idx = start_idx - 1\n while idx >= 0 and cleaned_code[idx].isspace():\n idx -= 1\n if idx < 0:\n return False\n \n char = cleaned_code[idx]\n \n # 处理箭头函数如: () => { ... } \n if char == '>':\n if idx - 1 >= 0 and cleaned_code[idx-1] == '=':\n return True\n return False\n \n # 处理普通函数/方法如: function foo() { ... } 或者 method() { ... }\n if char == ')':\n paren_count = 1\n p_idx = idx - 1\n while p_idx >= 0:\n if cleaned_code[p_idx] == ')':\n paren_count += 1\n elif cleaned_code[p_idx] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n p_idx -= 1\n \n if p_idx < 0:\n return False\n \n w_idx = p_idx - 1\n while w_idx >= 0 and cleaned_code[w_idx].isspace():\n w_idx -= 1\n \n if w_idx < 0:\n return True\n \n word_end = w_idx\n while w_idx >= 0 and (cleaned_code[w_idx].isalnum() or cleaned_code[w_idx] in '_$'):\n w_idx -= 1\n \n word = cleaned_code[w_idx+1 : word_end+1]\n \n # 排除控制流关键字\n if word in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\n return False\n else:\n return True\n \n return False\n\n def check_function_length_js(content: str) -> bool:\n cleaned = clean_js_code(content)\n stack = []\n \n for i, char in enumerate(cleaned):\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start = stack.pop()\n if is_js_function_block(cleaned, start):\n lines = content.count('\\n', start, i) + 1\n if lines > 60:\n return False\n return True\n\n def check_html_file(content: str) -> bool:\n # 寻找HTML中的内联<script>代码\n pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\n for match in pattern.finditer(content):\n script_content = match.group(1)\n if not check_function_length_js(script_content):\n return False\n return True\n\n def check_function_length_py(content: str) -> bool:\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n length = getattr(node, 'end_lineno', node.lineno) - node.lineno + 1\n if length > 60:\n return False\n return True\n except Exception:\n # 如遭遇语法解析错误则跳过该文件处理\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext in {'.js', '.ts', '.jsx', '.tsx'}:\n if not check_function_length_js(content):\n return False\n elif ext in {'.html', '.htm', '.vue'}:\n if not check_html_file(content):\n return False\n elif ext == '.py':\n if not check_function_length_py(content):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"def check_variable_names_length(response, workspace_path):\n import os\n import re\n \n def is_prime(n):\n \"\"\"判断一个数是否为质数\"\"\"\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n \n def clean_code(code):\n \"\"\"清洗代码，去除注释和字符串以防止误提取\"\"\"\n # 移除单行注释\n code = re.sub(r'//.*', '', code)\n # 移除多行注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除双引号字符串\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', code)\n # 移除单引号字符串\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", code)\n # 移除反引号字符串模板\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', code)\n return code\n\n def extract_variables(code):\n \"\"\"从JavaScript/TypeScript代码中提取变量、函数和类名\"\"\"\n variables = set()\n \n # 匹配 var, let, const 声明的常规变量名（排除部分复杂解构赋值的情况以免误伤）\n pattern = r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b'\n for match in re.finditer(pattern, code):\n variables.add(match.group(1))\n \n # 匹配具名函数定义中的函数名和参数\n func_pattern = r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(([^)]*)\\)'\n for match in re.finditer(func_pattern, code):\n variables.add(match.group(1)) # 提取函数名\n params_str = match.group(2)\n if params_str.strip():\n for p in params_str.split(','):\n p_name = p.split('=')[0].strip() # 剥除默认值部分\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\n variables.add(p_name)\n \n # 匹配匿名函数中的参数\n anon_func_pattern = r'\\bfunction\\s*\\(([^)]*)\\)'\n for match in re.finditer(anon_func_pattern, code):\n params_str = match.group(1)\n if params_str.strip():\n for p in params_str.split(','):\n p_name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\n variables.add(p_name)\n \n # 匹配类名\n class_pattern = r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b'\n for match in re.finditer(class_pattern, code):\n variables.add(match.group(1))\n \n return variables\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n code_blocks = []\n if file.endswith(('.js', '.ts')):\n code_blocks.append(content)\n elif file.endswith(('.html', '.htm', '.vue', '.jsx', '.tsx')):\n # 提取 HTML 类型文件中的脚本标签内容\n script_pattern = re.compile(r'<script[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\n scripts = script_pattern.findall(content)\n code_blocks.extend(scripts)\n \n for code in code_blocks:\n cleaned_code = clean_code(code)\n vars_in_code = extract_variables(cleaned_code)\n for v in vars_in_code:\n # 获取长度，严格判断是否为质数\n if not is_prime(len(v)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义各类代码文件可能出现的注释正则表达式\n # 注意：(?<!:)//.* 用于避免匹配到 http:// 或 https:// 这样的URL片段\n comment_patterns = {\n '.html': [r'<!--[\\s\\S]*?-->'],\n '.htm': [r'<!--[\\s\\S]*?-->'],\n '.css': [r'/\\*[\\s\\S]*?\\*/'],\n '.js': [r'(?<!:)//.*', r'/\\*[\\s\\S]*?\\*/'],\n '.ts': [r'(?<!:)//.*', r'/\\*[\\s\\S]*?\\*/'],\n '.jsx': [r'(?<!:)//.*', r'/\\*[\\s\\S]*?\\*/', r'<!--[\\s\\S]*?-->'],\n '.tsx': [r'(?<!:)//.*', r'/\\*[\\s\\S]*?\\*/', r'<!--[\\s\\S]*?-->'],\n '.vue': [r'(?<!:)//.*', r'/\\*[\\s\\S]*?\\*/', r'<!--[\\s\\S]*?-->'],\n '.py': [r'#.*', r\"'''[\\s\\S]*?'''\", r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"']\n }\n\n if not os.path.isdir(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理可能包含注释的已知代码文件\n if ext in comment_patterns:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于编码或其他原因无法读取文件，则跳过\n continue\n \n # 获取该文件对应的注释匹配规则\n patterns = comment_patterns[ext]\n for pattern in patterns:\n # 匹配出所有的注释块\n comments = re.findall(pattern, content)\n for comment in comments:\n # 遍历注释内容中的每个字符，检查是否为emoji\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"def check_file_lines_even(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据对话上下文中PWA项目以及要求的计划文件，推断出相关的代码文件扩展名\n target_extensions = ('.html', '.css', '.js', '.json', '.md')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(target_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 按照换行符分割，这样即使包含纯空行（无论是中间还是文件末尾）也能被准确计入行数\n lines_count = len(content.split('\\n'))\n \n if lines_count % 2 != 0:\n return False\n except Exception:\n # 忽略读取异常的文件，防止因某些特殊的编码问题导致验证程序崩溃\n pass\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 待检查的禁用词汇（全小写，便于后续统一转小写匹配）\n keywords = [\"测试\", \"demo\", \"临时\", \"调试\", \"草稿\", \"待完善\", \"todo\", \"待优化\", \"废弃\", \"占位\"]\n \n # 构造正则表达式\n # (?P<string>...) 组：匹配由于转义和字符闭合所包裹的字符串，防止其内部出现的 `//` 等被误认作注释\n # (?P<comment>...) 组：匹配常见单行及多行注释格式，包括 HTML、JS、CSS、Python 等语言规范的注释\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)' \n r'|(?P<comment>//.*|/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->|#.*)'\n )\n \n def check_content(content: str) -> bool:\n \"\"\"检查特定文本内容中的注释是否包含禁用字样\"\"\"\n for match in pattern.finditer(content):\n comment = match.group('comment')\n # 只有当捕获到注释时才进行匹配\n if comment:\n comment_lower = comment.lower()\n for kw in keywords:\n if kw in comment_lower:\n return False\n return True\n\n # 1. 检查 workspace_path 目录下的所有相关代码文件\n if os.path.isdir(workspace_path):\n # 针对该 PWA 聚合新闻项目的可能文件格式\n valid_exts = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py', '.json'}\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_content(content):\n return False\n except Exception:\n # 忽略无法读取的文件\n continue\n \n # 2. 检查 response 回复体中的 markdown 代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9]*\\n([\\s\\S]*?)```', response)\n for code in code_blocks:\n if not check_content(code):\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 匹配独立的等号 \"=\"，排除诸如 ==, ===, !=, !==, >=, <=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~=, :=, ?= 等符号\n pattern = re.compile(r'(?<![=!<>+\\-*/%&|^?:~])=(?![=>])')\n\n def check_content(content: str) -> bool:\n matches = list(pattern.finditer(content))\n for match in matches:\n idx = match.start()\n \n # 检查等号左侧是否恰好有两个空格\n if idx < 2:\n return False\n if content[idx-2:idx] != ' ':\n return False\n # 确保左侧第3个字符不是空格，保证“只有”两个空格\n if idx >= 3 and content[idx-3] == ' ':\n return False\n \n # 检查等号右侧是否恰好有两个空格\n if idx + 3 > len(content):\n return False\n if content[idx+1:idx+3] != ' ':\n return False\n # 确保右侧第3个字符不是空格，保证“只有”两个空格\n if idx + 3 < len(content) and content[idx+3] == ' ':\n return False\n \n return True\n\n # 1. 检查工作区中的所有代码文件\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤常见的代码及相关配置文件类型\n if file.endswith(('.py', '.js', '.html', '.css', '.json', '.md', '.vue', '.ts', '.jsx', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_content(content):\n return False\n except Exception:\n pass\n \n # 2. 检查回复中包含的代码块\n code_blocks = re.findall(r'```[a-zA-Z]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n if not check_content(block):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"所有生成的代码文件中的注释必须全部使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断每个生成的代码文件的注释部分是否都呈现了“PWA新闻聚合器”字样。\n \"\"\"\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n \n # 定义需要检查的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.cs'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果无法按utf-8读取，可以跳过该文件或者视为不满足。\n # 这里选择稳妥方式，跳过异常文件。\n continue\n \n # 提取不同语言的各类注释\n comments = []\n \n # HTML 格式注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # C风格块注释 (JS/CSS/TS/C/Java等)\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # C风格单行注释 (JS/TS/C/Java等)\n comments.extend(re.findall(r'//.*', content))\n # Python/Shell 单行注释\n comments.extend(re.findall(r'#.*', content))\n # Python 多行注释 (文档字符串)\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n \n # 合并当前文件所有提取出的注释\n all_comments_text = '\\n'.join(comments)\n \n # 如果该代码文件的所有注释中未包含指定字样，则直接判定为不满足要求\n if \"PWA新闻聚合器\" not in all_comments_text:\n return False\n\n # 若遍历完毕都没有发现任何代码文件，则不符合\"生成的每个代码文件\"的语境前提\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"import os\n\ndef check_indentation_requirement(response: str, workspace_path: str) -> bool:\n # 设定需要进行缩进检查的常见代码文件后缀\n code_extensions = {'.html', '.js', '.css', '.json', '.py', '.ts', '.jsx', '.tsx', '.vue'}\n \n if not os.path.isdir(workspace_path):\n return False\n \n has_code_files = False\n \n # 遍历当前工作区的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 忽略无法用utf-8读取的文件\n continue\n \n has_code_files = True\n \n tab_count = 0\n mod4_0_count = 0\n mod4_not_0_count = 0\n \n for line in lines:\n # 忽略完全无内容的空行\n if not line.strip():\n continue\n \n # 截取前导空白字符\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 忽略无缩进的行\n if not leading_whitespace:\n continue\n \n # 检查是否混用了制表符\n if '\\t' in leading_whitespace:\n tab_count += 1\n continue\n \n indent_len = len(leading_whitespace)\n \n # 统计前导空格是否为 4 的倍数\n if indent_len % 4 == 0:\n mod4_0_count += 1\n else:\n stripped_line = line.lstrip()\n # 排除 JSDoc 等多行注释中常见的使用 '*' 对齐的情况\n if stripped_line.startswith('*'):\n continue\n mod4_not_0_count += 1\n \n # 规则1：如果使用了 tab 缩进，则未“统一使用 4 个空格”，不满足要求\n if tab_count > 0:\n return False\n \n total_indents = mod4_0_count + mod4_not_0_count\n if total_indents > 0:\n # 规则2：如果文件存在缩进行，但没有一个是4的倍数，证明没有使用 4 空格标准\n if mod4_0_count == 0 and mod4_not_0_count > 0:\n return False\n \n # 规则3：如果非 4 个空格倍数的缩进行数占比较高（超过30%），且行数达到至少3行。\n # 这种情况通常说明代码错误地使用了如 2 个空格作为缩进标准。\n if (mod4_not_0_count / total_indents) > 0.3 and mod4_not_0_count >= 3:\n return False\n \n # 如果找到了代码文件且均通过了检查，才视作满足要求\n return has_code_files","tags":["Style","Layout"]}]},{"round_id":8,"instruction":"最后一步，我们要强化它的 PWA 特性和离线可用性体验。首先在 UI 层面，编写 JavaScript 监听设备的 online 和 offline 网络状态事件。如果检测到处于离线状态，要在页面顶部向下滑出一个红色的警示条，提示'当前处于离线模式，只能查看已缓存内容'，并在离线期间禁用下拉刷新功能。同时，模拟 Service Worker 的离线缓存机制，使得即使用户在断网状态下，依然能正常打开底部'书签'页面，且能点击阅读之前收藏过的新闻详情页面。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json，替换此前的修改计划记录文件要求；第二，运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 验证工作区目录是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令（PWA应用），推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts'}\n has_code_file = False\n \n # 遍历工作区内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤筛选代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试以 UTF-8 编码读取该文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生 UnicodeDecodeError，说明该文件不是使用合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如文件权限等导致的读取错误\n pass\n \n # 如果项目中存在代码文件并且全部通过验证，则认为满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 60 行的限制","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 清理JS/TS代码，去除会干扰括号匹配的字符串和注释，同时保留换行符保持行号一致\n def clean_js_code_preserve_lines(js_content):\n # 替换多行注释为相应数量的换行符\n def multiline_repl(match):\n return '\\n' * match.group(0).count('\\n')\n js_content = re.sub(r'/\\*.*?\\*/', multiline_repl, js_content, flags=re.DOTALL)\n \n # 移除单行注释\n js_content = re.sub(r'//.*', '', js_content)\n \n # 替换字符串并保留内部的换行符\n def string_repl(match):\n return '\"\"' + '\\n' * match.group(0).count('\\n')\n js_content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', string_repl, js_content)\n js_content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', js_content)\n js_content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", js_content)\n \n return js_content\n\n # 通过大括号特征匹配JS函数及其行数\n def check_js_functions(js_content):\n cleaned = clean_js_code_preserve_lines(js_content)\n current_line = 1\n stack = []\n \n for idx, char in enumerate(cleaned):\n if char == '\\n':\n current_line += 1\n continue\n \n if char == '{':\n # 获取该大括号之前的上下文，直到遇到上一个截断符号\n context = \"\"\n for j in range(idx - 1, -1, -1):\n if cleaned[j] in ['{', '}', ';']:\n break\n context = cleaned[j] + context\n \n context = context.replace('\\n', ' ').strip()\n \n # 判断当前大括号是否属于函数体\n is_func = False\n # 排除典型的控制流语句和类定义\n if re.search(r'\\b(?:catch|if|for|while|switch|with|else|try|finally|do)\\b', context):\n is_func = False\n elif re.search(r'\\bclass\\b', context):\n is_func = False\n elif 'function' in context:\n is_func = True\n elif '=>' in context:\n is_func = True\n # 方法声明（往往在大括号前具有成对或闭合的圆括号结构）\n elif re.search(r'\\)\\s*$', context):\n is_func = True\n \n stack.append((current_line, is_func))\n \n elif char == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n # 发现超长函数，即刻返回False\n if (current_line - start_line + 1) > 60:\n return False\n return True\n\n # 使用AST解析检测Python文件中的函数\n def check_py_functions(py_content):\n try:\n tree = ast.parse(py_content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n end = getattr(node, 'end_lineno', start)\n if (end - start + 1) > 60:\n return False\n return True\n except SyntaxError:\n return True\n\n # 遍历代码区所有可解析文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n if file.endswith(('.js', '.ts')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_js_functions(content):\n return False\n except:\n pass\n \n elif file.endswith(('.html', '.htm', '.vue')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 提取内嵌的脚本块进行判断\n script_blocks = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, flags=re.IGNORECASE | re.DOTALL)\n for block in script_blocks:\n if not check_js_functions(block):\n return False\n except:\n pass\n \n elif file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_py_functions(content):\n return False\n except:\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数","validation_code":"def check_variable_name_prime_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断一个数字是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n if n in (2, 3):\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n for i in range(5, int(n ** 0.5) + 1, 6):\n if n % i == 0 or n % (i + 2) == 0:\n return False\n return True\n\n # 匹配常规的 var/let/const 声明的变量\n js_var_pattern = re.compile(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n # 匹配对象解构赋值中的变量，例如：const { a, b } = obj;\n js_destruct_obj_pattern = re.compile(r'\\b(?:let|const|var)\\s*\\{([^}]+)\\}')\n # 匹配数组解构赋值中的变量，例如：let [x, y] = arr;\n js_destruct_arr_pattern = re.compile(r'\\b(?:let|const|var)\\s*\\[([^\\]]+)\\]')\n \n # 匹配可能的 Python 变量声明赋值（补充兼容）\n py_var_pattern = re.compile(r'^[ \\t]*([a-zA-Z_][a-zA-Z0-9_]*)\\s*=[^=]', re.MULTILINE)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查前端相关代码文件\n if ext in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue', '.css']:\n # 清除代码中的多行与单行注释，避免误匹配\n content_clean = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content_clean = re.sub(r'//.*', '', content_clean)\n \n # 1. 验证基础变量声明\n for var_name in js_var_pattern.findall(content_clean):\n if not is_prime(len(var_name)):\n return False\n \n # 2. 验证对象解构赋值出的变量\n for group in js_destruct_obj_pattern.findall(content_clean):\n for var_name in re.findall(r'([a-zA-Z_$][a-zA-Z0-9_$]*)', group):\n if not is_prime(len(var_name)):\n return False\n \n # 3. 验证数组解构赋值出的变量\n for group in js_destruct_arr_pattern.findall(content_clean):\n for var_name in re.findall(r'([a-zA-Z_$][a-zA-Z0-9_$]*)', group):\n if not is_prime(len(var_name)):\n return False\n \n # 4. 验证 CSS 变量 (形如 --main-color)\n if ext in ['.css', '.html', '.vue']:\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', content_clean)\n for var_name in css_vars:\n # 拼接为完整的 CSS 变量名检查长度\n full_name = f\"--{var_name}\"\n if not is_prime(len(full_name)):\n return False\n\n # 检查可能的 Python 脚本文件\n elif ext == '.py':\n # 清除 Python 的注释\n content_clean = re.sub(r'#.*', '', content)\n for var_name in py_var_pattern.findall(content_clean):\n if not is_prime(len(var_name)):\n return False\n \n except Exception:\n # 忽略读取异常的文件或非文本文件\n continue\n\n return True","tags":["Style","Naming"]},{"constraint":"无论哪个代码文件，其注释块中都不能带有 emoji表情符号","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 构造能够匹配字符串字面量和各类代码注释的正则表达式\n # Group 1: 双引号字符串 (避免内部的双斜杠等被误判为注释)\n # Group 2: 单引号字符串\n # Group 3: 反引号 (ES6 模板字符串)\n # Group 4: // 单行注释 (使用负向先行断言排除常见 http://, https://, file:// 协议)\n # Group 5: /* ... */ 块注释\n # Group 6: <!-- ... --> HTML 注释\n # Group 7: # 单行注释 (Python, Shell 等)\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\")|'\n r\"('(?:\\\\.|[^'\\\\])*')|\"\n r'(`(?:\\\\.|[^`\\\\])*`)|'\n r'((?<!https:)(?<!http:)(?<!file:)//.*$)|'\n r'(/\\*[\\s\\S]*?\\*/)|'\n r'(<!--[\\s\\S]*?-->)|'\n r'(#.*$)',\n re.MULTILINE\n )\n\n def has_emoji_in_comments(text: str) -> bool:\n for match in pattern.finditer(text):\n # 判断命中结果是否属于注释相关的 Group (即 4, 5, 6, 7 组)\n if any(match.group(i) for i in range(4, 8)):\n comment_text = match.group(0)\n # 遍历注释内容中的每一个字符\n for char in comment_text:\n # 使用专家实现的 is_emoji 函数\n if is_emoji(char):\n return True\n return False\n\n # 1. 首先检查回复文本中包含的所有 markdown 代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for code in code_blocks:\n if has_emoji_in_comments(code):\n return False\n\n # 2. 检查工作空间路径下的具体代码文件\n # 针对可能出现的各类前后端代码文件进行约束检查\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', \n '.css', '.scss', '.less', '.py', '.java', \n '.c', '.cpp', '.cs', '.php', '.go', '.rs', '.swift'\n }\n\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查文件内容中的注释是否含有 emoji\n if has_emoji_in_comments(content):\n return False\n except Exception:\n # 如遇到非 utf-8 编码或无法读取的文件，则跳过\n continue\n\n # 如果均未发现违规 emoji，则说明满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义PWA项目中常见的生成的代码及配置文件后缀\n valid_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts'}\n file_checked = False\n \n # 优先在工作区路径下检查实际生成的文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符明确切分来统计行数，能够确保所有空行均被计入\n lines = content.split('\\n')\n if len(lines) % 2 != 0:\n return False\n \n file_checked = True\n except Exception:\n # 忽略可能出现的文件读取异常（如编码不支持或权限问题）\n continue\n \n # 如果工作区中实际检查到了代码文件并且所有文件均通过验证\n if file_checked:\n return True\n \n # 降级检查：如果工作区中没有任何文件，退回到 response 字符串中提取 Markdown 代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9]*\\n(.*?)```', response, re.DOTALL)\n if code_blocks:\n for block in code_blocks:\n # Markdown代码块捕获的内容通常末尾会带一个不属于代码本身的换行符，需剔除\n if block.endswith('\\n'):\n block = block[:-1]\n \n lines = block.split('\\n')\n if len(lines) % 2 != 0:\n return False\n return True\n \n # 如果既没有创建文件，也没有在回复中输出代码块，算作不满足\n return False","tags":["Quantity","Exact Value"]},{"constraint":"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n forbidden_words = [\"测试\", \"demo\", \"临时\", \"调试\", \"草稿\", \"待完善\", \"TODO\", \"待优化\", \"废弃\", \"占位\"]\n \n def contains_forbidden(text: str) -> bool:\n lower_text = text.lower()\n for fw in forbidden_words:\n if fw.lower() in lower_text:\n return True\n return False\n\n def extract_js_css_comments(code: str) -> list:\n comments = []\n state = 'NORMAL'\n i = 0\n n = len(code)\n current_comment = []\n quote_char = ''\n while i < n:\n if state == 'NORMAL':\n if code[i:i+2] == '//':\n # 防止匹配到 URL 中的 http:// 或 https:// 结构\n if i > 0 and code[i-1] == ':':\n i += 1\n else:\n state = 'LINE_COMMENT'\n current_comment = ['//']\n i += 2\n elif code[i:i+2] == '/*':\n state = 'BLOCK_COMMENT'\n current_comment = ['/*']\n i += 2\n elif code[i] in \"\\\"'`\":\n state = 'STRING'\n quote_char = code[i]\n i += 1\n else:\n i += 1\n elif state == 'STRING':\n if code[i] == '\\\\':\n i += 2\n elif code[i] == quote_char:\n state = 'NORMAL'\n i += 1\n else:\n i += 1\n elif state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state = 'NORMAL'\n comments.append(''.join(current_comment))\n current_comment = []\n i += 1\n else:\n current_comment.append(code[i])\n i += 1\n elif state == 'BLOCK_COMMENT':\n if code[i:i+2] == '*/':\n current_comment.append('*/')\n comments.append(''.join(current_comment))\n current_comment = []\n state = 'NORMAL'\n i += 2\n else:\n current_comment.append(code[i])\n i += 1\n if current_comment:\n comments.append(''.join(current_comment))\n return comments\n\n def extract_html_comments(code: str) -> list:\n return re.findall(r'<!--[\\s\\S]*?-->', code)\n\n def extract_python_comments(code: str) -> list:\n comments = []\n state = 'NORMAL'\n i = 0\n n = len(code)\n current_comment = []\n quote_char = ''\n while i < n:\n if state == 'NORMAL':\n if code[i] == '#':\n state = 'LINE_COMMENT'\n current_comment = ['#']\n i += 1\n elif code[i:i+3] in ['\"\"\"', \"'''\"]:\n state = 'MULTI_STRING'\n quote_char = code[i:i+3]\n i += 3\n elif code[i] in \"\\\"'\":\n state = 'STRING'\n quote_char = code[i]\n i += 1\n else:\n i += 1\n elif state == 'STRING':\n if code[i] == '\\\\':\n i += 2\n elif code[i] == quote_char:\n state = 'NORMAL'\n i += 1\n else:\n i += 1\n elif state == 'MULTI_STRING':\n if code[i] == '\\\\':\n i += 2\n elif code[i:i+3] == quote_char:\n state = 'NORMAL'\n i += 3\n else:\n i += 1\n elif state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state = 'NORMAL'\n comments.append(''.join(current_comment))\n current_comment = []\n i += 1\n else:\n current_comment.append(code[i])\n i += 1\n if current_comment:\n comments.append(''.join(current_comment))\n return comments\n\n # 定义目标代码文件的后缀集合\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', \n '.css', '.scss', '.less', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n\n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.swift', '.kt'}:\n comments.extend(extract_js_css_comments(content))\n elif ext in {'.html', '.htm', '.vue'}:\n comments.extend(extract_html_comments(content))\n # 提取 HTML 内部可能包含在 <script> 或 <style> 的 JS / CSS 注释\n comments.extend(extract_js_css_comments(content))\n elif ext in {'.py', '.rb'}:\n comments.extend(extract_python_comments(content))\n\n # 遍历所有提取出的注释内容，验证是否存在禁用关键字\n for comment in comments:\n if contains_forbidden(comment):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格","validation_code":"import os\nimport re\n\ndef check_equals_spacing(response: str, workspace_path: str) -> bool:\n # 常见前端/通用代码文件拓展名\n code_extensions = {'.js', '.html', '.css', '.json', '.ts', '.jsx', '.tsx', '.vue'}\n found_code_files = False\n \n # 正则：匹配不属于组合运算符一部分的独立等号 '='\n # 左侧不能有 = < > ! + - * / % & | ^ ~ ?\n # 右侧不能有 = >\n pattern = re.compile(r\"(?<![=<>!+\\-*/%&|^~?])=(?![=>])\")\n \n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n continue\n \n # 遍历文件中所有匹配到的独立等号\n for match in pattern.finditer(text):\n start = match.start()\n end = match.end()\n \n left = text[:start]\n right = text[end:]\n \n # 检查等号左侧是否是以且仅以两个连续空格结尾\n if not left.endswith(\" \") or left.endswith(\" \"):\n return False\n \n # 检查等号右侧是否是以且仅以两个连续空格开头\n if not right.startswith(\" \") or right.startswith(\" \"):\n return False\n\n return found_code_files","tags":["Style","Layout"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"所有生成的代码文件中的注释必须全部使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 如果工作目录不存在，则返回不满足\n if not os.path.exists(workspace_path):\n return False\n\n # 根据 PWA 应用特点，推断可能生成的代码文件扩展名\n code_extensions = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.py', '.json'}\n \n # 构造正则表达式，匹配常见的各类代码注释\n # 使用非捕获组的写法直接提取出完整的匹配文本\n comment_pattern = re.compile(\n r'//.*|' # JS/CSS/TS 单行注释\n r'/\\*[\\s\\S]*?\\*/|' # JS/CSS/TS 多行注释\n r'<!--[\\s\\S]*?-->|' # HTML 注释\n r'#.*|' # Python/Shell 单行注释\n r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"|' # Python 多行注释\n r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'', # Python 多行注释\n )\n\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n # json 文件在标准规范下不能包含注释，但如果存在可跳过注释检测或是被视为无注释\n # 为了防止误判为失败，我们将单纯的 JSON 数据文件排除在严格代码注释检查之外\n if ext == '.json':\n continue\n \n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于编码等问题无法读取，按照指令中的UTF-8强制约束属于异常状态，判为False\n return False\n \n # 提取出文件内所有的注释内容\n comments = comment_pattern.findall(content)\n \n # 检查这些提取出来的注释中，是否包含指定的字样\n target_found = False\n for comment_text in comments:\n if 'PWA新闻聚合器' in comment_text:\n target_found = True\n break\n \n # 若当前代码文件内没有任何注释包含该字样，则不符合\"每个代码文件\"的要求\n if not target_found:\n return False\n\n # 若遍历完毕均符合条件，并且至少存在一个需要检查的代码文件，则视为满足\n return has_code_files","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n lines_to_check = []\n \n # 1. 提取 workspace 中的代码文件行\n valid_extensions = {'.js', '.html', '.css', '.json', '.py', '.vue', '.ts', '.jsx', '.tsx'}\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines_to_check.extend(f.readlines())\n except Exception:\n pass\n \n # 2. 提取 response 中的代码块行\n valid_languages = {\n 'js', 'javascript', 'html', 'css', 'json', \n 'py', 'python', 'vue', 'ts', 'typescript', 'jsx', 'tsx'\n }\n code_blocks = re.findall(r'```([a-zA-Z0-9_+-]*)\\n(.*?)```', response, re.DOTALL)\n for lang, block in code_blocks:\n lang = lang.strip().lower()\n if lang in valid_languages or not lang:\n # 如果没有语言标记，简单判断其是否包含常见的代码结构字符，以防将文本列表当作代码\n if not lang and not re.search(r'[{<};=]', block):\n continue\n lines_to_check.extend(block.split('\\n'))\n \n total_indents = 0\n non_four_indents = 0\n tab_indents = 0\n \n # 3. 统计并分析所有行的缩进模式\n for line in lines_to_check:\n if not line.strip():\n continue\n \n match = re.match(r'^([ \\t]+)(.*)', line)\n if match:\n indent = match.group(1)\n content = match.group(2)\n \n # 忽略以 '*' 开头的多行注释内容（如 JSDoc 注释），因为它们的格式化可能会包含多余的单个空格\n if content.startswith('*'):\n continue\n \n if '\\t' in indent:\n tab_indents += 1\n continue\n \n indent_len = len(indent)\n total_indents += 1\n if indent_len % 4 != 0:\n non_four_indents += 1\n \n # 只要存在制表符缩进，即违背了“仅使用空格”的要求\n if tab_indents > 0:\n return False\n \n # 如果存在缩进行，且不能被4整除的行占比超过了15%，说明AI大概率没有采用4空格缩进（例如混用了2空格）\n if total_indents > 0:\n if (non_four_indents / total_indents) > 0.15:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作","validation_code":"","tags":["Workflow","Orchestration"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我搭建一个新闻聚合器 PWA 应用的基础页面骨架。需要一个顶部的 App Bar（包含应用标题和一个菜单占位图标），中间的主体区域放几个简单的静态新闻卡片（暂时只需一个灰色的方块代表图片，旁边放两行假文字代表标题即可），底部放一个包含'新闻'和'书签'两个选项的标签导航栏（Tab Bar）。这是一个初版框架，不需要写任何核心业务逻辑，先把基本结构和CSS样式撑起来就好。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”字样；第二，在为用户生成回复时，你必须永远只用中文。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件后缀，可根据具体的生成内容（如PWA应用可能涉及的HTML, CSS, JS等）推断\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.json', '.xml',\\n '.ts', '.jsx', '.tsx', '.vue', '.py', '.java',\\n '.c', '.cpp', '.h', '.hpp', '.md', '.txt'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取整个代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明存在非 UTF-8 编码的字符\\n return False\\n except Exception:\\n # 忽略其他非编码相关的异常（如权限问题等）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n # 如果工作区不存在，视作没有违背该限制\\n return True\\n\\n # 替换函数：用于去除 JS/TS 代码中的注释和字符串字面量\\n # 规则：若是换行符保留以维持行号，非换行符全替换为空格\\n def replacer(match):\\n s = match.group(0)\\n return ''.join('\\\\n' if c == '\\\\n' else ' ' for c in s)\\n\\n # 匹配JS注释和字符串的正则表达式\\n js_clean_pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.MULTILINE | re.DOTALL\\n )\\n\\n def check_js_code(code_str: str) -> bool:\\n # 去除注释与字符串字面量\\n clean_code = js_clean_pattern.sub(replacer, code_str)\\n \\n matches = []\\n # 1. 查找 function 关键字\\n for match in re.finditer(r'\\\\bfunction\\\\b', clean_code):\\n matches.append(('func', match.start(), match.end()))\\n # 2. 查找箭头函数 =>\\n for match in re.finditer(r'=>', clean_code):\\n matches.append(('arrow', match.start(), match.end()))\\n # 3. 查找类或对象的简写方法定义，排除各类控制流关键字\\n method_pattern = r'\\\\b(?!(?:if|for|while|switch|catch|with|function|try|else|do|return|yield)\\\\b)[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n for match in re.finditer(method_pattern, clean_code):\\n matches.append(('method', match.start(), match.end()))\\n\\n for m_type, m_start, m_end in matches:\\n # 计算函数的起始行号\\n start_line = clean_code.count('\\\\n', 0, m_start) + 1\\n \\n brace_idx = -1\\n if m_type == 'method':\\n # 简写方法的正则结尾本身就是 '{'\\n brace_idx = m_end - 1\\n else:\\n # 寻找其后的第一个 '{'\\n for i in range(m_end, len(clean_code)):\\n if clean_code[i] == '{':\\n brace_idx = i\\n break\\n elif clean_code[i] in ';}':\\n # 如果在遇到 '{' 之前遇到 ';' 或 '}'，说明可能不是块级函数结构\\n break\\n \\n # 若没找到 '{' 则不是包含代码块的函数（例如单行箭头函数），跳过\\n if brace_idx == -1:\\n continue\\n \\n depth = 0\\n end_line = start_line\\n in_brace = False\\n \\n # 使用花括号深度匹配找到函数体的结束位置\\n for i in range(brace_idx, len(clean_code)):\\n if clean_code[i] == '{':\\n depth += 1\\n in_brace = True\\n elif clean_code[i] == '}':\\n depth -= 1\\n \\n # 深度归零，说明函数体闭合\\n if in_brace and depth == 0:\\n end_line = clean_code.count('\\\\n', 0, i) + 1\\n break\\n \\n if in_brace and depth == 0:\\n # 检查函数总行数\\n if end_line - start_line + 1 > 60:\\n return False\\n \\n return True\\n\\n def check_py_code(code_str: str) -> bool:\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n func_length = node.end_lineno - node.lineno + 1\\n if func_length > 60:\\n return False\\n except Exception:\\n # 忽略语法解析错误的Python文件\\n pass\\n return True\\n\\n # HTML文件中的 script 块内容提取正则\\n html_pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\\n\\n # 遍历工作目录下所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in ['.py', '.js', '.ts', '.html', '.htm', '.jsx', '.tsx']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_py_code(code_str):\\n return False\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n if not check_js_code(code_str):\\n return False\\n elif ext in ['.html', '.htm']:\\n # 针对HTML，校验所有内嵌的 script 标签内容\\n for script_match in html_pattern.finditer(code_str):\\n script_content = script_match.group(1)\\n if not check_js_code(script_content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断一个数是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 如果工作目录不存在，则无法验证，且由于不存在违规变量，视作没有违反该规则\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 编译匹配不同语言中变量声明的正则表达式\\n # 匹配 JavaScript/TypeScript 变量：var, let, const 后面的第一个合法标识符\\n js_var_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\b')\\n # 匹配 CSS 变量：-- 开头的自定义属性名\\n css_var_pattern = re.compile(r'(--[a-zA-Z0-9_-]+)\\\\s*:')\\n # 匹配 Python 变量：行首紧跟的变量赋值\\n py_var_pattern = re.compile(r'^\\\\s*([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=', re.MULTILINE)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略无法通过UTF-8解码或其他无法读取的文件\\n continue\\n\\n # 提取当前文件内的变量列表\\n vars_found = []\\n\\n # 针对 HTML/JavaScript 等前端代码文件\\n if file.endswith(('.js', '.ts', '.html', '.htm', '.vue', '.jsx', '.tsx')):\\n vars_found.extend(js_var_pattern.findall(content))\\n \\n # 针对 HTML/CSS 等样式文件\\n if file.endswith(('.css', '.html', '.htm', '.vue')):\\n vars_found.extend(css_var_pattern.findall(content))\\n\\n # 针对 Python 后端文件\\n if file.endswith('.py'):\\n vars_found.extend(py_var_pattern.findall(content))\\n\\n # 遍历检查提取出的每一个变量名\\n for var_name in vars_found:\\n var_name = var_name.strip()\\n # 题目要求为字符长度（即字符数），按照指令提示可以直接使用len函数\\n if not is_prime(len(var_name)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 遍历工作区下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 针对该 PWA 项目可能涉及的前端代码文件进行过滤\\n if ext in ['.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue']:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n # 提取各类型代码的注释内容\\n if ext == '.html':\\n # 匹配 HTML 特有注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 匹配 HTML 中嵌入的 JS/CSS 注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'(?<!:)//(.*?)$', content, re.MULTILINE))\\n elif ext == '.css':\\n # 匹配 CSS 多行注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.vue']:\\n # 匹配 JS 等脚本的多行和单行注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'(?<!:)//(.*?)$', content, re.MULTILINE))\\n \\n # 遍历提取出的注释内容，判断是否存在 emoji 表情符号\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对PWA应用推断常见的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.jsx', '.tsx'}\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按换行符精确切分，包含所有空行\\n total_lines = len(content.split('\\\\n'))\\n \\n # 检查行数是否为偶数\\n if total_lines % 2 != 0:\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”字样\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # Regex pattern to match HTML comments, block comments (/* ... */), and line comments (// ...)\\n comment_pattern = re.compile(r'(?:<!--[\\\\s\\\\S]*?-->)|(?:/\\\\*[\\\\s\\\\S]*?\\\\*/)|(?://[^\\\\r\\\\n]*)')\\n \\n valid_extensions = ('.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.py')\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n comments = comment_pattern.findall(content)\\n for comment in comments:\\n if '测试' in comment:\\n return False\\n except Exception:\\n # Silently ignore files that cannot be read (e.g., unexpected binaries)\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"第一版骨架不错。现在让新闻列表活起来，我们需要加载真实感强的数据。请用模拟的 JSON 数据加载新闻，并在主体区域渲染出更精致的新闻卡片。卡片左侧要展示新闻缩略图，右侧包含标题、新闻来源、发布时间（比如'2小时前'）和一段简短的摘要。文字过长要做截断处理（显示省略号），卡片之间留有适当的间距，排版需要显得清晰现代一些。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”字样；第二，你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断，PWA应用相关的常见代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.webmanifest', '.py'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试使用 utf-8 对字节序列进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明包含非 utf-8 编码的内容\\n return False\\n except Exception:\\n # 其他读取异常保守跳过\\n continue\\n\\n # 遍历完所有代码文件均未报错，则满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_js_functions(js_code: str) -> bool:\\n # 用于匹配 JS 中函数的开头结构：\\n # 1. 普通函数定义或匿名函数: function name() { 或 function() {\\n # 2. 带有块的箭头函数: => {\\n # 3. 类或对象方法: methodName(args) { (通过前向否定排除 if/for/while 等控制流关键字)\\n pattern = re.compile(r'(\\\\bfunction\\\\b[^\\\\{]*\\\\{|=>\\\\s*\\\\{|\\\\b(?!(?:if|for|while|catch|switch)\\\\b)\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{)')\\n idx = 0\\n \\n while idx < len(js_code):\\n match = pattern.search(js_code, idx)\\n if not match:\\n break\\n \\n start_match = match.start()\\n start_brace = match.end() - 1 # 找到开头结构中 `{` 的位置\\n \\n brace_count = 0\\n in_string = False\\n string_char = ''\\n in_comment_line = False\\n in_comment_block = False\\n \\n end_idx = start_brace\\n while end_idx < len(js_code):\\n char = js_code[end_idx]\\n \\n # 状态机：处理单行注释\\n if in_comment_line:\\n if char == '\\\\n':\\n in_comment_line = False\\n # 状态机：处理多行注释\\n elif in_comment_block:\\n if char == '*' and end_idx + 1 < len(js_code) and js_code[end_idx+1] == '/':\\n in_comment_block = False\\n end_idx += 1\\n # 状态机：处理字符串\\n elif in_string:\\n if char == '\\\\\\\\':\\n end_idx += 1 # 跳过转义符\\n elif char == string_char:\\n in_string = False\\n # 正常代码匹配流程\\n else:\\n if char in '\\\"\\\\'`':\\n in_string = True\\n string_char = char\\n elif char == '/' and end_idx + 1 < len(js_code):\\n next_char = js_code[end_idx+1]\\n if next_char == '/':\\n in_comment_line = True\\n end_idx += 1\\n elif next_char == '*':\\n in_comment_block = True\\n end_idx += 1\\n elif char == '{':\\n brace_count += 1\\n elif char == '}':\\n brace_count -= 1\\n if brace_count == 0: # 匹配到了当前函数结束的花括号\\n break\\n end_idx += 1\\n \\n # 若完整匹配到成对的花括号，截取代码段计算行数\\n if brace_count == 0 and end_idx < len(js_code):\\n func_str = js_code[start_match:end_idx+1]\\n # 计算函数行数\\n line_count = func_str.count('\\\\n') + 1\\n if line_count > 60:\\n return False\\n \\n # 继续往后扫描下一个潜在函数\\n idx = start_brace + 1\\n \\n return True\\n\\n # 遍历当前工作区所有的生成文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 处理 Python 文件\\n if file.endswith('.py'):\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 仅限制函数不能超过60行\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n # 兼容不同 Python 版本的行数获取\\n if hasattr(node, 'end_lineno'):\\n end = node.end_lineno\\n else:\\n max_line = node.lineno\\n for child in ast.walk(node):\\n if hasattr(child, 'lineno'):\\n max_line = max(max_line, child.lineno)\\n end = max_line\\n \\n if end - start + 1 > 60:\\n return False\\n except Exception:\\n pass\\n \\n # 处理 JS / TS 文件\\n elif file.endswith('.js') or file.endswith('.ts'):\\n if not check_js_functions(content):\\n return False\\n \\n # 处理 HTML 等内嵌了 script 标签的文件\\n elif file.endswith('.html') or file.endswith('.htm'):\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_js_functions(script):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n in (2, 3):\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n class PyVarVisitor(ast.NodeVisitor):\\n def __init__(self):\\n self.variables = set()\\n \\n def visit_Assign(self, node):\\n for target in node.targets:\\n if isinstance(target, ast.Name):\\n self.variables.add(target.id)\\n self.generic_visit(node)\\n \\n def visit_AnnAssign(self, node):\\n if isinstance(node.target, ast.Name):\\n self.variables.add(node.target.id)\\n self.generic_visit(node)\\n \\n def visit_FunctionDef(self, node):\\n for arg in node.args.args:\\n self.variables.add(arg.arg)\\n if hasattr(node.args, 'kwonlyargs'):\\n for arg in node.args.kwonlyargs:\\n self.variables.add(arg.arg)\\n self.generic_visit(node)\\n\\n variables = set()\\n # 匹配 JavaScript 简单变量定义 (var, let, const)\\n js_var_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\b')\\n # 匹配 JavaScript 函数参数\\n js_func_pattern = re.compile(r'\\\\bfunction\\\\s+(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\\\s*\\\\(([^)]*)\\\\)')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 处理 JS 及 HTML\\n if file.endswith(('.js', '.html')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 提取常规变量\\n for match in js_var_pattern.findall(content):\\n variables.add(match)\\n \\n # 提取函数参数\\n for match in js_func_pattern.findall(content):\\n params = match.split(',')\\n for p in params:\\n p_clean = p.strip()\\n # 保证其是合法的标识符\\n if re.fullmatch(r'[a-zA-Z_$][0-9a-zA-Z_$]*', p_clean):\\n variables.add(p_clean)\\n except Exception:\\n pass\\n # 处理可能存在的 Python 文件\\n elif file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n visitor = PyVarVisitor()\\n visitor.visit(tree)\\n variables.update(visitor.variables)\\n except Exception:\\n pass\\n\\n # 验证是否每个提取的变量名的字符长度恰好为质数\\n for var in variables:\\n if not is_prime(len(var)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n code_extensions = {'.html', '.css', '.js', '.json', '.ts'}\\n \\n comment_patterns = [\\n re.compile(r'<!--(.*?)-->', re.DOTALL),\\n re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL),\\n re.compile(r'//(.*)$', re.MULTILINE)\\n ]\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n for pattern in comment_patterns:\\n comments = pattern.findall(content)\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n code_extensions = {'.html', '.css', '.js', '.json', '.ts', '.jsx', '.tsx'}\\n file_checked = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n file_checked = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果文件无法读取，在此场景下忽略或视为未完全满足\\n continue\\n \\n return file_checked\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”字样\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.vue', '.py', '.java', \\n '.c', '.cpp', '.h', '.hpp', '.cs', '.php', '.go', '.rb'\\n }\\n \\n # 匹配各类代码文件中的注释（开启多行模式以便匹配 $ 到行尾）\\n # 分别对应：\\n # 1. 块级注释: /* ... */\\n # 2. 单行注释: // ... (用否定后视断言忽略如 http:// 中的 //)\\n # 3. HTML注释: <!-- ... -->\\n # 4. 单行注释: # ...\\n comment_pattern = re.compile(\\n r'(/\\\\*[\\\\s\\\\S]*?\\\\*/|(?<!:)//.*?$|<!--[\\\\s\\\\S]*?-->|#.*?$)',\\n re.MULTILINE\\n )\\n \\n # 需要被排查的违禁词（全部转换为小写，便于统一比较大小写不敏感的demo）\\n forbidden_words = ['测试', 'demo']\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果读取文件出现编码等错误，则跳过\\n continue\\n \\n # 提取出文件中的所有注释部分\\n comments = comment_pattern.findall(content)\\n for comment_text in comments:\\n comment_text_lower = comment_text.lower()\\n \\n # 检查是否包含违禁词\\n for word in forbidden_words:\\n if word in comment_text_lower:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 匹配独立的等号，排除 ==, ===, >=, <=, !=, +=, -= 等复合赋值/比较运算符\\n pattern = re.compile(r'(?<![<>=!\\\\+\\\\-\\\\*/%&\\\\|])=(?![=])')\\n \\n valid_extensions = ('.js', '.html', '.css', '.json', '.vue', '.jsx', '.ts', '.tsx')\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(valid_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n line = line.rstrip('\\\\n\\\\r')\\n for match in pattern.finditer(line):\\n idx = match.start()\\n \\n # 验证左侧是否恰好只有两个空格\\n if idx < 2:\\n return False\\n if not (line[idx-1] == ' ' and line[idx-2] == ' '):\\n return False\\n if idx > 2 and line[idx-3] == ' ':\\n return False\\n \\n # 验证右侧是否恰好只有两个空格\\n if idx + 2 >= len(line):\\n return False\\n if not (line[idx+1] == ' ' and line[idx+2] == ' '):\\n return False\\n if idx + 3 < len(line) and line[idx+3] == ' ':\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，接下来给顶部 App Bar 下方加一个可横向滑动的分类导航栏，包含如'推荐'、'科技'、'财经'、'娱乐'、'体育'等分类标签。用户点击不同分类时，要在下方的新闻列表加上淡入淡出的切换动画，并展示该分类对应的模拟新闻数据。记得当前选中的分类标签文字要加粗，并且底部带有一个高亮的下划线指示器。另外请你注意两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”字样，替换此前的注释关键词限制要求；第二，凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义PWA应用及常见可能生成的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.json', '.md', \\n '.ts', '.jsx', '.tsx', '.vue', '.txt'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_content = f.read()\\n # 尝试用 UTF-8 进行解码\\n raw_content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦发现无法用 UTF-8 解码的文件，直接返回 False\\n return False\\n except Exception:\\n # 忽略其他可能因权限或文件系统引起的非编码相关错误\\n pass\\n\\n # 只要检查的代码文件没有抛出 UnicodeDecodeError 且项目确实包含有效文件，即判定为符合要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证工作区下所有代码文件中的各个函数，每一个都不能超出 60 行的限制。\\n \\\"\\\"\\\"\\n def clean_js_code(code: str) -> str:\\n # 将字符串和注释替换为等长的空格和换行符组合，保持原本的行数和行号一致\\n pattern = r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\\\\\"(?:\\\\\\\\.|[^\\\\\\\"\\\\\\\\])*\\\\\\\"|`(?:\\\\\\\\.|[^`\\\\\\\\])*`|//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/\\\"\\n def replacer(match):\\n text = match.group(0)\\n return '\\\\n' * text.count('\\\\n') + ' ' * (len(text) - text.count('\\\\n'))\\n return re.sub(pattern, replacer, code)\\n\\n def check_js_content(code: str) -> bool:\\n # 如果代码内容整体行数就不超过 60 行，说明必然符合要求，直接放行\\n if code.count('\\\\n') + 1 <= 60:\\n return True\\n \\n cleaned = clean_js_code(code)\\n stack = []\\n for i, char in enumerate(cleaned):\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start_idx = stack.pop()\\n start_line = cleaned.count('\\\\n', 0, start_idx) + 1\\n end_line = cleaned.count('\\\\n', 0, i) + 1\\n lines_count = end_line - start_line + 1\\n \\n if lines_count > 60:\\n prefix = cleaned[:start_idx].rstrip()\\n \\n # 特征1：箭头函数\\n if prefix.endswith('=>'):\\n return False\\n \\n last_end = max(prefix.rfind('}'), prefix.rfind(';'), prefix.rfind('{'))\\n segment = prefix[last_end+1:] if last_end != -1 else prefix\\n \\n # 特征2：包含 function 关键字声明的函数\\n if re.search(r'\\\\bfunction\\\\b', segment):\\n return False\\n \\n # 特征3：形式为 name(args) 的类成员方法或对象方法\\n segment = segment.strip()\\n match = re.search(r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*$', segment)\\n if match:\\n name = match.group(1)\\n # 排除普通的控制流或异常处理作用域块\\n if name not in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\\n return False\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 再次整体过滤小于等于60行的文件\\n if content.count('\\\\n') + 1 <= 60:\\n continue\\n\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n if not check_js_content(content):\\n return False\\n \\n elif file.endswith(('.html', '.htm', '.vue')):\\n scripts = re.finditer(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for match in scripts:\\n script_content = match.group(1)\\n if not check_js_content(script_content):\\n return False\\n \\n elif file.endswith('.py'):\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n lines_count = node.end_lineno - node.lineno + 1\\n if lines_count > 60:\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2 or n == 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n def get_js_vars(code: str) -> set:\\n vars_set = set()\\n # 移除单行与多行注释，避免误匹配\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n # 提取 var, let, const 声明的变量名\\n matches = re.findall(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\b', code)\\n for m in matches:\\n vars_set.add(m)\\n \\n # 提取函数参数作为变量名\\n func_matches = re.findall(r'\\\\bfunction\\\\s+(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\\\s*\\\\(([^)]*)\\\\)', code)\\n for args in func_matches:\\n for arg in args.split(','):\\n arg = arg.strip()\\n if arg:\\n arg_name = re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*', arg)\\n if arg_name:\\n vars_set.add(arg_name.group(0))\\n return vars_set\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 验证 JavaScript / TypeScript 文件\\n if ext in ['.js', '.ts']:\\n variables = get_js_vars(content)\\n for var in variables:\\n if not is_prime(len(var)):\\n return False\\n \\n # 验证 CSS 文件\\n elif ext == '.css':\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', content)\\n for var in css_vars:\\n if not is_prime(len(var)):\\n return False\\n \\n # 验证 HTML 文件内嵌的 script 和 style 标签\\n elif ext == '.html':\\n script_matches = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n for code in script_matches:\\n variables = get_js_vars(code)\\n for var in variables:\\n if not is_prime(len(var)):\\n return False\\n \\n style_matches = re.findall(r'<style\\\\b[^>]*>(.*?)</style>', content, flags=re.DOTALL | re.IGNORECASE)\\n for style in style_matches:\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', style)\\n for var in css_vars:\\n if not is_prime(len(var)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n # 尝试以 utf-8 格式读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n comments = []\\n \\n # 匹配 C-style 注释（JS, TS, CSS 等）\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.css', '.c', '.cpp', '.h', '.hpp', '.java', '.go']:\\n # 多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 单行注释 // ... （CSS不支持此写法）\\n if ext != '.css':\\n comments.extend(re.findall(r'//.*', content))\\n \\n # 匹配 HTML-style 注释（HTML, XML 等）\\n if ext in ['.html', '.htm', '.vue', '.xml', '.svg']:\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n # 匹配 Python/Shell-style 注释（Python, Bash 等）\\n if ext in ['.py', '.sh', '.yaml', '.yml', '.rb']:\\n comments.extend(re.findall(r'#.*', content))\\n \\n # 遍历每一个提取出的注释字符串\\n for comment in comments:\\n # 遍历注释中的每个字符判断是否为 emoji\\n for char in comment:\\n if is_emoji(char):\\n return False # 只要发现存在 emoji，便不满足要求\\n \\n return True # 遍历完所有代码文件均未在注释中发现 emoji\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_file_lines_are_even(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义当前PWA项目可能涉及的常见代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.jsx', '.tsx'}\\n \\n files_checked = 0\\n \\n # 遍历项目工作区路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符进行分割，可以准确统计出包含空行的文件总行数\\n lines = content.split('\\\\n')\\n \\n # 如果行数是奇数，不满足要求，直接返回False\\n if len(lines) % 2 != 0:\\n return False\\n \\n files_checked += 1\\n except Exception:\\n # 读取出错直接忽略该文件（也可以视作失败，这里选择跳过异常文件）\\n pass\\n \\n # 如果项目中没有任何对应的代码文件，说明没有生成有效项目，返回False\\n # 只要成功验证了至少一个文件且都满足偶数行限制，则返回True\\n return files_checked > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证代码项目中的任何代码文件注释部分是否不得含有“测试”、“demo”、“临时”、“调试”字样。\\n \\\"\\\"\\\"\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 需要排除的关键词，统一使用小写以进行无视大小写的匹配\\n forbidden_keywords = [\\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\"]\\n\\n # 常见的前后端代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n comments = []\\n \\n # HTML 文件：提取 HTML 注释，并处理内部包含的 script 和 style 块\\n if ext in ['.html', '.htm']:\\n # HTML 格式注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n # 提取 script 块内的 JS 注释\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for script in scripts:\\n comments.extend(re.findall(r'(?<!:)//.*', script)) # 避免匹配到 http://\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', script))\\n \\n # 提取 style 块内的 CSS 注释\\n styles = re.findall(r'<style.*?>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n for style in styles:\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', style))\\n \\n # CSS 文件\\n elif ext == '.css':\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n # JS / TS 文件\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n # Python 文件\\n elif ext == '.py':\\n comments.extend(re.findall(r'#.*', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n\\n # 遍历并验证提取出的所有注释内容\\n for comment in comments:\\n comment_lower = comment.lower()\\n for kw in forbidden_keywords:\\n if kw in comment_lower:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的前端与配置代码文件后缀\\n valid_exts = {'.js', '.ts', '.html', '.css', '.vue', '.json'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n i = 0\\n while i < len(content):\\n if content[i] == '=':\\n is_compound = False\\n \\n # 检查是否为复合运算符（如 !=, >=, <=, +=, -=, *=, /=, %=, &=, |=, ^=, ??=, :=）\\n if i > 0 and content[i-1] in '!=><+-*/%&|^~?:':\\n is_compound = True\\n # 检查是否为箭头函数或者比较运算（如 =>, ==, ===）\\n if i < len(content) - 1 and content[i+1] in '=>':\\n is_compound = True\\n \\n if not is_compound:\\n # 检查等号左边是否恰好有两个空格\\n left_ok = False\\n if i >= 2 and content[i-2:i] == ' ':\\n # 左侧那两个空格的前一个字符不能是空格（保障恰好是两个空格而不是三个及以上）\\n if i < 3 or content[i-3] != ' ':\\n left_ok = True\\n \\n # 检查等号右边是否恰好有两个空格\\n right_ok = False\\n if i <= len(content) - 3 and content[i+1:i+3] == ' ':\\n # 右侧那两个空格的后一个字符不能是空格\\n if i > len(content) - 4 or content[i+3] != ' ':\\n right_ok = True\\n \\n # 如果存在单独的等号，其左右两边不满足都有两个空格，则视为违反要求\\n if not (left_ok and right_ok):\\n return False\\n i += 1\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在来完善新闻的阅读体验。点击任意一条新闻卡片后，不要直接跳走，而是从屏幕底部向上平滑弹出一个全屏覆盖的详情阅读面板。这个阅读面板里需要展示一张顶部的清晰大图、完整的长篇正文段落，右上角必须有一个明显的'关闭'按钮用来退回列表。面板的最底部要固定一个按钮'在原网站阅读'，点击时可以弹出一个'即将跳转到外部链接'的 Toast 提示。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”字样，替换此前的注释关键词限制要求；第二，后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 结合PWA项目特性推断可能的代码文件类型\\n code_exts = {'.html', '.css', '.js', '.json', '.md'}\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果出现解码错误，说明存在未按UTF-8编码的文件\\n return False\\n except Exception:\\n # 忽略诸如权限问题等非编码导致的文件读取错误\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n # 使用状态机将代码内的注释、字符串内容等扰乱文本替换为空格，以保持代码原结构的换行符和括号不受干扰\\n def mask_strings_and_comments(content):\\n masked = []\\n state = 'NORMAL'\\n i = 0\\n n = len(content)\\n while i < n:\\n c = content[i]\\n if state == 'NORMAL':\\n if c == '/' and i + 1 < n:\\n if content[i+1] == '/':\\n state = 'LINE_COMMENT'\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n continue\\n elif content[i+1] == '*':\\n state = 'BLOCK_COMMENT'\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n continue\\n if c == '\\\"':\\n state = 'DQUOTE'\\n masked.append(c)\\n i += 1\\n continue\\n if c == \\\"'\\\":\\n state = 'SQUOTE'\\n masked.append(c)\\n i += 1\\n continue\\n if c == '`':\\n state = 'BQUOTE'\\n masked.append(c)\\n i += 1\\n continue\\n masked.append(c)\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if c == '\\\\n':\\n state = 'NORMAL'\\n masked.append(c)\\n else:\\n masked.append(' ')\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if c == '*' and i + 1 < n and content[i+1] == '/':\\n state = 'NORMAL'\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n else:\\n if c == '\\\\n':\\n masked.append(c)\\n else:\\n masked.append(' ')\\n i += 1\\n elif state in ('DQUOTE', 'SQUOTE', 'BQUOTE'):\\n quote_char = '\\\"' if state == 'DQUOTE' else (\\\"'\\\" if state == 'SQUOTE' else '`')\\n if c == '\\\\\\\\':\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n elif c == quote_char:\\n state = 'NORMAL'\\n masked.append(c)\\n i += 1\\n else:\\n if c == '\\\\n':\\n masked.append(c)\\n else:\\n masked.append(' ')\\n i += 1\\n return \\\"\\\".join(masked)\\n\\n def get_previous_non_space(text, index):\\n idx = index - 1\\n while idx >= 0 and text[idx].isspace():\\n idx -= 1\\n return idx\\n\\n # 判断 JS/TS/HTML(script内) 中的代码块，校验是否含超过60行的函数\\n def check_functions_in_js_text(text):\\n masked = mask_strings_and_comments(text)\\n \\n blocks = {}\\n stack = []\\n # 通过匹配成对的 {} 锁定基础逻辑块\\n for i, char in enumerate(masked):\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start_index = stack.pop()\\n blocks[start_index] = i\\n \\n for start_index, end_index in blocks.items():\\n prev_idx = get_previous_non_space(masked, start_index)\\n if prev_idx < 0:\\n continue\\n \\n is_func = False\\n func_start_idx = start_index \\n \\n # 检测箭头函数结构 `=> {`\\n if masked[prev_idx] == '>':\\n if prev_idx - 1 >= 0 and masked[prev_idx - 1] == '=':\\n is_func = True\\n func_start_idx = prev_idx - 1 \\n \\n # 检测方法/函数常规结构 `)` 前配对 `(`\\n elif masked[prev_idx] == ')':\\n p_count = 0\\n open_paren_idx = -1\\n for j in range(prev_idx, -1, -1):\\n if masked[j] == ')':\\n p_count += 1\\n elif masked[j] == '(':\\n p_count -= 1\\n if p_count == 0:\\n open_paren_idx = j\\n break\\n \\n if open_paren_idx != -1:\\n # 查找对应的小括号前是否有关联标志\\n word_end = get_previous_non_space(masked, open_paren_idx)\\n if word_end >= 0 and (masked[word_end].isalnum() or masked[word_end] in '_$'):\\n word_start = word_end\\n while word_start >= 0 and (masked[word_start].isalnum() or masked[word_start] in '_$'):\\n word_start -= 1\\n word = masked[word_start+1 : word_end+1]\\n \\n # 排除掉 if / for 等控制流结构的大括号\\n if word not in ['if', 'for', 'while', 'switch', 'catch']:\\n is_func = True\\n func_start_idx = word_start + 1\\n \\n prev_word_end = get_previous_non_space(masked, word_start)\\n if prev_word_end >= 0 and masked[prev_word_end].isalpha():\\n pw_start = prev_word_end\\n while pw_start >= 0 and masked[pw_start].isalpha():\\n pw_start -= 1\\n prev_word = masked[pw_start+1 : prev_word_end+1]\\n \\n # 再往前回溯寻找可能的修饰关键字\\n if prev_word == 'function':\\n func_start_idx = pw_start + 1\\n # 尝试捕获开头的 async\\n pp_word_end = get_previous_non_space(masked, pw_start)\\n if pp_word_end >= 0 and masked[pp_word_end].isalpha():\\n ppw_start = pp_word_end\\n while ppw_start >= 0 and masked[ppw_start].isalpha():\\n ppw_start -= 1\\n pp_word = masked[ppw_start+1 : pp_word_end+1]\\n if pp_word == 'async':\\n func_start_idx = ppw_start + 1\\n \\n elif prev_word in ['async', 'get', 'set']:\\n func_start_idx = pw_start + 1\\n\\n if is_func:\\n # 统计包含自身和内部块的总换行数，计算得出总行数\\n lines_count = text.count('\\\\n', func_start_idx, end_index) + 1\\n if lines_count > 60:\\n return False\\n \\n return True\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 扫描当前生成的工作目录里的所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 支持 PWA 相关的 JS、TS 以及 Python\\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.py']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n # Python 文件直接由原生的 ast 进行验证\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n lines_count = node.end_lineno - node.lineno + 1\\n if lines_count > 60:\\n return False\\n except Exception:\\n pass\\n elif ext in ['.html', '.htm']:\\n # 提取 HTML 内部带有的 <script> 代码进行校验\\n script_blocks = re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\\n for block in script_blocks:\\n script_content = block.group(1)\\n if not check_functions_in_js_text(script_content):\\n return False\\n else:\\n # 其它 JS/TS 文件直接校验\\n if not check_functions_in_js_text(content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"判断一个数是否为质数\\\"\\\"\\\"\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n def extract_js_variables(code: str) -> set:\\n \\\"\\\"\\\"通过简单的有限状态机提取 JavaScript/HTML 中的变量名、函数名与参数\\\"\\\"\\\"\\n # 移除单行与多行注释\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除字符串字面量（安全正则）\\n code = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", \\\"''\\\", code)\\n code = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '``', code)\\n \\n vars_set = set()\\n \\n # 将代码分词：保留声明关键字、合法的标识符名以及控制流程的标点\\n tokens = re.finditer(r'\\\\b(var|let|const|function|class|catch)\\\\b|([a-zA-Z_$][a-zA-Z0-9_$]*)|([=,;{}()[\\\\]])', code)\\n \\n state = 'normal'\\n nest_level = 0\\n base_level = 0\\n \\n for match in tokens:\\n keyword = match.group(1)\\n ident = match.group(2)\\n punct = match.group(3)\\n \\n if punct in ('(', '[', '{'):\\n nest_level += 1\\n \\n if keyword in ('var', 'let', 'const'):\\n state = 'in_decl'\\n base_level = nest_level\\n elif keyword in ('function', 'class'):\\n state = 'in_func_name'\\n base_level = nest_level\\n elif keyword == 'catch':\\n state = 'in_catch'\\n \\n elif state == 'in_decl':\\n if ident:\\n vars_set.add(ident)\\n state = 'after_ident'\\n elif punct == ';' and nest_level == base_level:\\n state = 'normal'\\n \\n elif state == 'after_ident':\\n if punct == '=':\\n state = 'after_eq'\\n elif punct == ',':\\n state = 'in_decl'\\n elif punct == ';' and nest_level == base_level:\\n state = 'normal'\\n \\n elif state == 'after_eq':\\n if punct == ',' and nest_level == base_level:\\n state = 'in_decl'\\n elif punct == ';' and nest_level == base_level:\\n state = 'normal'\\n \\n elif state == 'in_func_name':\\n if ident:\\n vars_set.add(ident)\\n state = 'after_func_name'\\n elif punct == '(':\\n state = 'in_params'\\n \\n elif state == 'after_func_name':\\n if punct == '(':\\n state = 'in_params'\\n \\n elif state == 'in_catch':\\n if punct == '(':\\n state = 'in_params'\\n \\n elif state == 'in_params':\\n if ident:\\n vars_set.add(ident)\\n elif punct == ')':\\n state = 'normal'\\n \\n if punct in (')', ']', '}'):\\n nest_level -= 1\\n if nest_level < 0:\\n nest_level = 0\\n \\n return vars_set\\n\\n def extract_py_variables(code: str) -> set:\\n \\\"\\\"\\\"通过 AST 提取 Python 中的变量名、函数名、类名及参数\\\"\\\"\\\"\\n vars_set = set()\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name):\\n if isinstance(node.ctx, ast.Store):\\n vars_set.add(node.id)\\n elif isinstance(node, ast.FunctionDef):\\n vars_set.add(node.name)\\n for arg in node.args.args:\\n vars_set.add(arg.arg)\\n elif isinstance(node, ast.ClassDef):\\n vars_set.add(node.name)\\n except Exception:\\n pass\\n return vars_set\\n\\n all_vars = set()\\n \\n # 遍历项目文件提取变量\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if file.endswith('.js'):\\n all_vars.update(extract_js_variables(content))\\n elif file.endswith('.py'):\\n all_vars.update(extract_py_variables(content))\\n elif file.endswith('.html') or file.endswith('.htm') or file.endswith('.vue'):\\n # 提取 <script> 标签内的 JS 代码\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n all_vars.update(extract_js_variables(script))\\n \\n # 没有找到任何变量的情况默认算作满足\\n if not all_vars:\\n return True\\n \\n # 验证是否每个变量名的字符长度恰好是质数\\n for var in all_vars:\\n if not is_prime(len(var)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义可能用于项目的常见代码文件扩展名集合\\n code_exts = {\\n '.js', '.css', '.html', '.htm', '.ts', '.jsx', '.tsx', \\n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.cs', \\n '.go', '.php', '.rb', '.swift', '.kt', '.json'\\n }\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 遇到无法读取或者编码不支持的文件则跳过\\n continue\\n \\n comments = []\\n \\n # 提取 /* ... */ 风格块级注释\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 提取 <!-- ... --> 风格HTML注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # 提取 // 风格单行注释（利用负向后瞻排除 // 前为冒号或单双引号的情况，防误杀 URL）\\n comments.extend(re.findall(r\\\"(?<![:'\\\\\\\"])\\\\/\\\\/.*\\\", content))\\n \\n # 针对特定后端/脚本语言提取 # 风格单行注释及多行注释\\n if ext in {'.py', '.rb', '.php', '.sh', '.yaml', '.yml'}:\\n comments.extend(re.findall(r'#.*', content))\\n if ext == '.py':\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n \\n # 逐个检查所有提取到的注释内容，判断是否包含 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端PWA项目可能涉及的常规代码及文档文件的扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.py', '.ts', '.jsx', '.tsx'}\\n\\n # 确保项目目录存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按换行符分割，能够精确还原编辑器中的行数分布，包括首尾及中间的空行\\n lines_count = len(content.split('\\\\n'))\\n \\n # 判断行数是否为偶数，若存在奇数行的文件则返回 False\\n if lines_count % 2 != 0:\\n return False\\n except Exception:\\n # 忽略非UTF-8编码等无法以文本模式正常读取的文件\\n continue\\n\\n # 所有代码文件都经过检查且行数全为偶数\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n keywords = [\\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\", \\\"草稿\\\", \\\"待完善\\\"]\\n # 统一转换成小写，方便应对英文字母的不同大小写情况\\n keywords_lower = [k.lower() for k in keywords]\\n \\n # 常见的代码文件后缀集合\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', \\n '.html', '.htm', '.vue', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.cs', '.go', '.php', '.rb', '.swift', \\n '.kt', '.rs', '.jsonc'\\n }\\n \\n # 构建用于提取不同类型代码注释的正则表达式\\n # 1. 匹配多行注释 /* ... */\\n pattern_c_multi = re.compile(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n # 2. 匹配 HTML多行注释 <!-- ... -->\\n pattern_html = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n # 3. 匹配单行注释 // ...，使用 (?<!:) 避免匹配到网址 http://\\n pattern_c_single = re.compile(r'(?<!:)//.*')\\n # 4. 匹配脚本语言单行注释 # ...\\n pattern_hash_single = re.compile(r'#.*')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n comments = []\\n # 粗略提取当前文件内的所有类型的注释片段\\n comments.extend(pattern_c_multi.findall(content))\\n comments.extend(pattern_html.findall(content))\\n comments.extend(pattern_c_single.findall(content))\\n comments.extend(pattern_hash_single.findall(content))\\n \\n # 对每段注释内容进行敏感关键词检测\\n for comment in comments:\\n comment_lower = comment.lower()\\n for kw in keywords_lower:\\n if kw in comment_lower:\\n return False\\n except Exception:\\n # 遇到文件读取错误（如二进制内容读取异常）时主动跳过，继续检测其他文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"def check_equal_spaces(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端与配置代码文件扩展名\\n code_extensions = {'.js', '.css', '.html', '.htm', '.json', '.ts', '.jsx', '.tsx', '.vue', '.svelte'}\\n \\n # 匹配独立的等号，排除 ==, ===, !=, !==, >=, <=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~= 等组合情况\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%&|^~])=(?![=>])')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 寻找代码文件内容中所有单独的 '='\\n for match in pattern.finditer(content):\\n idx = match.start()\\n \\n # 1. 检查左侧是否严格包含两个空格\\n if idx < 2:\\n return False\\n if content[idx-2:idx] != ' ':\\n return False\\n # 确保左侧第三个字符（若存在）不是空格，保证“刚好”留出两个空格\\n if idx >= 3 and content[idx-3] == ' ':\\n return False\\n \\n # 2. 检查右侧是否严格包含两个空格\\n if idx + 2 >= len(content):\\n return False\\n if content[idx+1:idx+3] != ' ':\\n return False\\n # 确保右侧第三个字符（若存在）不是空格，保证“刚好”留出两个空格\\n if idx + 3 < len(content) and content[idx+3] == ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"非常棒。下一步是实现核心的'书签'（收藏）功能。在首页每一条新闻卡片的右下角，以及阅读详情面板的操作区，加上一个'空心星星'图标。点击它可以收藏新闻，星星会变成实心，并伴随一个短暂的放大缩小的弹性动画。被收藏的数据要通过 localStorage 保存在本地。同时，当我们在底部导航栏切换到'书签'页面时，要展示所有已被收藏的新闻列表，并在该列表中支持点击星星取消收藏，取消时该卡片要自动从列表中移除。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”字样，替换此前的注释关键词限制要求；第二，针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义 PWA 项目及指令中可能生成的代码和文本文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.txt', '.xml', '.svg'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对代码文件进行检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出解码错误，说明存在非 UTF-8 编码的文件，不满足要求\\n return False\\n except Exception:\\n # 忽略文件权限等其他非编码相关的异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_python_functions(code_str):\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n end = getattr(node, 'end_lineno', start)\\n if end and start and (end - start + 1) > 60:\\n return False\\n except Exception:\\n pass\\n return True\\n\\ndef check_functions_in_code(code_str):\\n def replacer(match):\\n s = match.group(0)\\n # 用相等数量的换行符和空格替换掉所有的注释与字符串主体，确保代码原有的行号和字符索引稳定不变。\\n return ''.join(['\\\\n' if c == '\\\\n' else ' ' for c in s])\\n\\n # 匹配并抹除可能造成干扰的注释和多形态的字符串字面量（包含单行、多行及模板字符串）\\n pattern = re.compile(\\n r'//[^\\\\n]*|' \\n r'/\\\\*.*?\\\\*/|' \\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|' \\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\" \\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', \\n re.DOTALL\\n )\\n \\n clean_code = pattern.sub(replacer, code_str)\\n \\n def is_function_block(clean_code, brace_index):\\n i = brace_index - 1\\n while i >= 0 and clean_code[i].isspace():\\n i -= 1\\n if i < 0: return False\\n \\n # 箭头函数\\n if i >= 1 and clean_code[i-1:i+1] == '=>':\\n return True\\n \\n if clean_code[i] == ')':\\n paren_count = 1\\n i -= 1\\n while i >= 0 and paren_count > 0:\\n if clean_code[i] == ')': paren_count += 1\\n elif clean_code[i] == '(': paren_count -= 1\\n i -= 1\\n \\n if paren_count > 0: return False\\n \\n while i >= 0 and clean_code[i].isspace():\\n i -= 1\\n \\n word = \\\"\\\"\\n while i >= 0 and (clean_code[i].isalnum() or clean_code[i] in '_$'):\\n word = clean_code[i] + word\\n i -= 1\\n \\n # 基础函数块 function(...) {\\n if word == \\\"function\\\":\\n return True\\n \\n # 方法或对象属性函数\\n if word and word not in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\\n j = i\\n while j >= 0 and (clean_code[j].isspace() or clean_code[j] == '*'):\\n j -= 1\\n prev_word = \\\"\\\"\\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in '_$'):\\n prev_word = clean_code[j] + prev_word\\n j -= 1\\n if prev_word == \\\"function\\\":\\n return True\\n return True\\n \\n # 匿名函数\\n if not word:\\n j = i\\n while j >= 0 and (clean_code[j].isspace() or clean_code[j] == '*'):\\n j -= 1\\n prev_word = \\\"\\\"\\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in '_$'):\\n prev_word = clean_code[j] + prev_word\\n j -= 1\\n if prev_word == \\\"function\\\":\\n return True\\n \\n return False\\n\\n stack = []\\n line_number = 1\\n \\n # 逐字遍历，借助堆栈准确匹配嵌套的函数大括号层级\\n for i, char in enumerate(clean_code):\\n if char == '\\\\n':\\n line_number += 1\\n elif char == '{':\\n is_func = is_function_block(clean_code, i)\\n stack.append((is_func, line_number))\\n elif char == '}':\\n if stack:\\n is_func, start_line = stack.pop()\\n if is_func:\\n func_length = line_number - start_line + 1\\n if func_length > 60:\\n return False\\n return True\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # Python后端文件处理\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_python_functions(content):\\n return False\\n except Exception:\\n pass\\n \\n # 含有大括号语法的前端主业务文件及相关处理\\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.java', '.cpp', '.c', '.cs', '.php', '.html', '.vue')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 剥离出含有 HTML 的网页组件中嵌有的 <script> 代码块\\n if file.endswith(('.html', '.vue')):\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_functions_in_code(script):\\n return False\\n else:\\n if not check_functions_in_code(content):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 辅助函数：判断一个整数是否是质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 辅助函数：根据文件扩展名提取文件中定义的变量名\\n def extract_variables(ext: str, content: str) -> set:\\n variables = set()\\n \\n if ext in ['.js', '.ts']:\\n # 1. 匹配常规的 var / let / const 声明\\n pattern = r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b'\\n for match in re.finditer(pattern, content):\\n variables.add(match.group(1))\\n \\n # 2. 匹配数组解构赋值\\n for match in re.finditer(r'\\\\b(?:let|var|const)\\\\s+\\\\[(.*?)\\\\]', content):\\n for p in match.group(1).split(','):\\n p = p.strip()\\n if re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)$', p):\\n variables.add(p)\\n \\n # 3. 匹配对象解构赋值\\n for match in re.finditer(r'\\\\b(?:let|var|const)\\\\s+\\\\{(.*?)\\\\}', content):\\n for p in match.group(1).split(','):\\n p = p.strip()\\n if ':' in p:\\n p = p.split(':')[1].strip()\\n if re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)$', p):\\n variables.add(p)\\n \\n # 4. 匹配函数参数\\n func_pattern = r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)'\\n for match in re.finditer(func_pattern, content):\\n for arg in match.group(1).split(','):\\n arg = arg.strip()\\n if re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)$', arg):\\n variables.add(arg)\\n \\n elif ext in ['.py']:\\n # 匹配 Python 简单赋值变量\\n pattern = r'^[\\\\t ]*([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*(?::[^=]+)?=\\\\s*'\\n for match in re.finditer(pattern, content, re.MULTILINE):\\n variables.add(match.group(1))\\n \\n elif ext in ['.html', '.htm']:\\n # 提取 HTML 文件中 script 标签内的代码继续按 js 解析\\n script_pattern = r'<script.*?>([\\\\s\\\\S]*?)</script>'\\n for match in re.finditer(script_pattern, content, re.IGNORECASE):\\n variables.update(extract_variables('.js', match.group(1)))\\n \\n return variables\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目文件提取并验证变量长度\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 只关心包含代码文件类型的扩展名\\n if ext not in ['.js', '.ts', '.html', '.htm', '.py']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 获取该文件中所有的提取变量\\n variables = extract_variables(ext, content)\\n \\n for var in variables:\\n if not is_prime(len(var)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对不同语言类型的注释正则表达式\\n js_css_pattern = re.compile(r'//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n html_pattern = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n py_pattern = re.compile(r'#.*')\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 对文件类型进行分类\\n is_js_css = ext in ['.js', '.ts', '.css', '.scss', '.sass', '.less']\\n is_html = ext in ['.html', '.htm']\\n is_mixed = ext in ['.vue', '.jsx', '.tsx']\\n is_py = ext in ['.py']\\n \\n # 若并非可能包含代码的文件格式，则跳过\\n if not (is_js_css or is_html or is_mixed or is_py):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略不可读或无法使用 utf-8 解码的文件\\n continue\\n \\n comments = []\\n if is_js_css:\\n comments.extend(js_css_pattern.findall(content))\\n elif is_html:\\n comments.extend(html_pattern.findall(content))\\n elif is_mixed:\\n comments.extend(js_css_pattern.findall(content))\\n comments.extend(html_pattern.findall(content))\\n elif is_py:\\n comments.extend(py_pattern.findall(content))\\n \\n # 遍历所有提取到的注释文本\\n for comment in comments:\\n for char in comment:\\n # 如果任意注释的字符被判定为 emoji，则返回 False\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令上下文，推断出涉及到的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.jsx', '.tsx'}\\n \\n # 确保给定的工作区路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前项目文件夹下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 提取文件扩展名并转小写\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 如果是代码文件，则进行行数验证\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 计入空行的前提下，总行数如果为奇数，则不满足要求\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果遇到非文本内容或编码不正确的文件，安全地略过\\n continue\\n\\n # 所有的代码文件检查均通过（或者是并未生成文件），则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 待检查的常见前端及后端代码文件扩展名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\\n '.html', '.htm', '.css', '.scss', '.less',\\n '.py', '.java', '.cpp', '.c', '.h', '.hpp', '.cs',\\n '.php', '.rb', '.go', '.rs', '.swift', '.kt'\\n }\\n \\n # 禁用关键词列表（均统一为小写以便不区分大小写匹配）\\n forbidden_words = [\\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\", \\\"草稿\\\", \\\"待完善\\\", \\\"todo\\\", \\\"待优化\\\"]\\n \\n # 构建正则表达式模式，利用捕获组分别匹配字符串和不同类型的注释\\n # string组: 匹配双引号、单引号、反引号括起来的字符串（避免将字符串内的字词当作注释误判）\\n # m_comment组: 匹配多行注释 /* ... */ 或 HTML注释 <!-- ... -->\\n # s_comment组: 匹配单行注释 // ... 或 Python等脚本类注释 # ...\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<m_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->)'\\n r'|(?P<s_comment>//.*|#.*)',\\n re.MULTILINE\\n )\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n for match in pattern.finditer(content):\\n # 判断匹配项是否为多行或单行注释\\n comment = match.group('m_comment') or match.group('s_comment')\\n if comment:\\n comment_lower = comment.lower()\\n # 逐个检查违规词是否存在于当前注释中\\n for word in forbidden_words:\\n if word in comment_lower:\\n return False\\n except Exception:\\n # 忽略不可读或无权限的文件\\n continue\\n \\n # 所有文件验证完毕，均无违规关键词则返回True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 可能的代码文件后缀集合（针对 PWA 项目）\\n code_extensions = {'.html', '.css', '.js', '.json', '.ts', '.jsx', '.tsx', '.webmanifest', '.vue', '.svelte'}\\n has_checked_files = False\\n \\n # 复合等号左侧可能出现的关联字符，涵盖 JS、CSS 常见操作符\\n left_chars = set('<>!=+-*/%&|^~?$')\\n # 复合等号右侧可能出现的关联字符\\n right_chars = set('=>')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_checked_files = True\\n \\n for line in lines:\\n line = line.rstrip('\\\\n\\\\r')\\n i = 0\\n while i < len(line):\\n if line[i] == '=':\\n is_single_equal = True\\n \\n # 判定是否属于复合操作符的一部分\\n if i > 0 and line[i-1] in left_chars:\\n is_single_equal = False\\n if i < len(line) - 1 and line[i+1] in right_chars:\\n is_single_equal = False\\n \\n if is_single_equal:\\n # 检查左侧是否正好有两个空格\\n left_ok = False\\n if i >= 2:\\n if line[i-1] == ' ' and line[i-2] == ' ':\\n # 确保第三个字符不是空格（或已到达行首）\\n if i == 2 or line[i-3] != ' ':\\n left_ok = True\\n \\n # 检查右侧是否正好有两个空格\\n right_ok = False\\n if i + 2 < len(line):\\n if line[i+1] == ' ' and line[i+2] == ' ':\\n # 确保第三个字符不是空格（或已到达行尾）\\n if i + 3 == len(line) or line[i+3] != ' ':\\n right_ok = True\\n \\n # 只要发现任何一个单独的等号前后不满足恰好留有两个空格的要求，直接返回 False\\n if not (left_ok and right_ok):\\n return False\\n i += 1\\n \\n return has_checked_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"接下来给应用加点质感，引入'深色模式'。在顶部 App Bar 的右侧添加一个太阳/月亮的切换图标。点击该图标后，整个应用（包括主页列表、底栏、详情弹窗面板等所有元素）平滑过渡切换为深色主题。你需要精心调整深色背景下的文字颜色、卡片底色以及边框，确保对比度舒适。并且，必须把用户选择的主题模式保存到本地，页面刷新后依然能记住用户的选择。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json，替换此前的修改计划记录文件要求；第二，所有生成的代码文件中的注释必须全部使用中文。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断 PWA 前端项目及修改计划中可能涉及的代码和文本文件后缀名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.json', '.md'}\\n \\n if not os.path.exists(workspace_path):\\n # 如果未生成任何文件，则没有违反“所有生成的文件必须是UTF-8”的条件\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出解码错误，说明存在非 UTF-8 编码的代码文件，不满足要求\\n return False\\n except Exception:\\n # 如果由于其他异常（如权限问题）无法打开，这里选择忽略或继续\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_js_functions(script: str) -> bool:\\n # 定义正则表达式匹配：双引号字符串、单引号字符串、反引号字符串、多行注释、单行注释\\n pattern = r'(\\\"(?:\\\\\\\\\\\"|[^\\\"])*\\\")|(\\\\'(?:\\\\\\\\\\\\'|[^\\\\'])*\\\\')|(`(?:\\\\\\\\`|[^`])*`)|(/\\\\*[\\\\s\\\\S]*?\\\\*/)|(//[^\\\\n]*)'\\n \\n def replacer(match):\\n if match.group(1) or match.group(2) or match.group(3):\\n # 如果是字符串内容，用 X 替换所有的 {} 防止干扰匹配，同时不改变文本行数\\n return match.group(0).replace('{', 'X').replace('}', 'X')\\n elif match.group(4):\\n # 如果是多行注释，替换为空换行符，确保总行数依然对齐\\n return '\\\\n' * match.group(4).count('\\\\n')\\n elif match.group(5):\\n # 如果是单行注释，直接移除\\n return ''\\n return match.group(0)\\n\\n try:\\n script = re.sub(pattern, replacer, script)\\n except Exception:\\n pass\\n\\n lines = script.split('\\\\n')\\n stack = []\\n\\n for i, line in enumerate(lines):\\n idx = 0\\n while idx < len(line):\\n char = line[idx]\\n if char == '{':\\n # 当遇到 { 时，获取它之前的上下文文本（追溯最多5行）\\n start_line_idx = max(0, i - 5)\\n prev_text_lines = lines[start_line_idx:i]\\n prev_text_lines.append(line[:idx])\\n prev_text = \\\" \\\".join(prev_text_lines).strip()\\n\\n is_func = False\\n # 判断之前的字符串是否具备典型的JS函数声明或箭头函数特征\\n if prev_text.endswith('=>') or prev_text.endswith('function'):\\n is_func = True\\n else:\\n match = re.search(r'([\\\\w$]+)\\\\s*\\\\([^)]*\\\\)\\\\s*$', prev_text)\\n if match:\\n kw = match.group(1)\\n # 过滤掉常见的控制流语句关键字\\n if kw not in ('if', 'while', 'for', 'switch', 'catch'):\\n is_func = True\\n else:\\n if re.search(r'\\\\bfunction\\\\s*$', prev_text):\\n is_func = True\\n \\n # 兜底：如果以 ) 结尾并开启 {}，基本可以认为是一个逻辑代码块或者函数\\n if not is_func and prev_text.endswith(')'):\\n is_func = True\\n\\n stack.append((i, is_func))\\n elif char == '}':\\n if stack:\\n start_idx, is_func = stack.pop()\\n if is_func:\\n # 判断该函数的行数是否大于 60\\n if i - start_idx + 1 > 60:\\n return False\\n idx += 1\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 解析 Python AST，找出函数定义\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\\n if node.end_lineno - node.lineno + 1 > 60:\\n return False\\n except Exception:\\n pass\\n elif file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_js_functions(content):\\n return False\\n except Exception:\\n pass\\n elif file.endswith('.html') or file.endswith('.htm'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 提取 HTML 中的所有 <script> 代码块\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for script in scripts:\\n if not check_js_functions(script):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_variable_name_length(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 正则表达式：匹配 JS/TS/HTML 中的变量、函数和类声明\\n # 仅提取使用 var, let, const, function, class 关键字进行直接声明的标识符\\n # 这样能有效避免提取到系统的内置属性、已有的 API 及方法名，从而防止误判\\n pattern = re.compile(r'\\\\b(?:var|let|const|function|class)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n\\n # 需要检查的前端代码文件后缀名\\n target_extensions = ('.js', '.ts', '.html', '.htm', '.vue', '.jsx', '.tsx')\\n\\n # 遍历工作空间目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(target_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找当前文件内所有声明匹配到的变量名\\n matches = pattern.findall(content)\\n for var_name in matches:\\n # 验证该变量名的字符长度是否为质数\\n if not is_prime(len(var_name)):\\n return False\\n except Exception:\\n # 若文件由于编码等不可预见的原因无法读取，则跳过\\n continue\\n\\n # 如果所有提取到的变量名的字符长度均正好是质数，或者未提取到变量名，则判定为满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 穷举推断可能的代码相关文件后缀，适用于PWA及常见Web项目类型\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.md'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n # 按UTF-8要求进行解码读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略读取失败的非文本文件或错误编码文件\\n continue\\n \\n comments = []\\n \\n # 1. 匹配 HTML 注释 (跨行)\\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\\n \\n # 2. 匹配 多行/块注释 (CSS, JS, TS等，跨行)\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, flags=re.DOTALL))\\n \\n # 3. 匹配 JS/TS 单行注释，利用 (?<!:) 排除如 http:// 这类 URL 的干扰\\n comments.extend(re.findall(r'(?<!:)//(.*?)$', content, flags=re.MULTILINE))\\n \\n # 4. 匹配 传统 # 的单行注释 (应对潜在的其他生成文件)\\n comments.extend(re.findall(r'#(.*?)$', content, flags=re.MULTILINE))\\n \\n # 验证提取出的所有注释是否存在 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"import os\\n\\ndef check_even_line_counts(response: str, workspace_path: str) -> bool:\\n # 如果工作区不存在，由于没有文件被创建，不会违反要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义常见的前端、配置及项目记录文件扩展名\\n target_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.jsx', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除非用户主动创建的第三方依赖或系统目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 只针对代码文件及项目相关的文本文件\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 计入空行在内的总行数必须为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果文件不可读（比如存在非utf-8编码问题），跳过该文件继续验证\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”字样\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义禁止出现在注释中的词汇\\n forbidden_words = [\\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\", \\\"草稿\\\", \\\"待完善\\\", \\\"todo\\\", \\\"待优化\\\"]\\n forbidden_words_lower = [w.lower() for w in forbidden_words]\\n \\n # 匹配各类代码文件中常见注释的正则表达式\\n comment_patterns = [\\n r'//.*', # JS/TS/CSS 等语言的单行注释\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/', # JS/TS/CSS 等语言的多行注释\\n r'<!--[\\\\s\\\\S]*?-->', # HTML 的注释\\n r'#.*', # Shell/Python 的单行注释\\n r\\\"'''[\\\\s\\\\S]*?'''\\\", # Python 的多行注释\\n r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"' # Python 的多行注释\\n ]\\n # 合并成一个正则匹配所有注释\\n pattern = re.compile('|'.join(comment_patterns))\\n \\n # 需要检查的常见代码文件扩展名\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.json', '.md', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 提取该文件中的所有注释部分\\n comments = pattern.findall(content)\\n \\n # 检查提取到的每段注释中是否包含禁用关键词\\n for comment in comments:\\n comment_lower = comment.lower()\\n for word in forbidden_words_lower:\\n if word in comment_lower:\\n return False\\n except Exception:\\n # 如果遇到编码错误或不可读情况，直接忽略该文件并继续\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 确保工作区路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义常见的代码文件后缀\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.json', '.vue', '.svelte', '.py', '.java', '.c', '.cpp', \\n '.h', '.cs', '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.dart'\\n }\\n \\n has_code_file = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n file_content = f.read()\\n except Exception:\\n continue # 如果存在读取异常，跳过该文件\\n \\n n = len(file_content)\\n for i in range(n):\\n if file_content[i] == '=':\\n # 检查是否为复合运算符（或箭头函数）的一部分\\n is_compound = False\\n \\n # 检查左侧是否为常见复合操作符的前缀符号\\n if i > 0 and file_content[i-1] in '=!<>-+*/%&|^~$:':\\n is_compound = True\\n \\n # 检查右侧是否为相等判断或箭头函数后缀\\n if i < n - 1 and file_content[i+1] in '=>':\\n is_compound = True\\n \\n if not is_compound:\\n # 检查左侧是否恰好有两个空格\\n left_valid = False\\n if i >= 2 and file_content[i-1] == ' ' and file_content[i-2] == ' ':\\n if i >= 3:\\n if file_content[i-3] != ' ':\\n left_valid = True\\n else:\\n # 等号前仅有两个字符且均为空格（即文件开头）\\n left_valid = True\\n \\n # 检查右侧是否恰好有两个空格\\n right_valid = False\\n if i <= n - 3 and file_content[i+1] == ' ' and file_content[i+2] == ' ':\\n if i <= n - 4:\\n if file_content[i+3] != ' ':\\n right_valid = True\\n else:\\n # 等号后仅有两个字符且均为空格（即文件结尾）\\n right_valid = True\\n \\n # 如果左侧或右侧任一不满足恰好留出两个空格，立刻返回 False\\n if not (left_valid and right_valid):\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的注释必须全部使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"交互体验还能继续优化。给主页的新闻列表加上'下拉刷新'的功能：用户在列表顶部继续向下滑动时，要出现一个带有旋转动画的刷新指示图标，松手后模拟网络加载2秒钟，然后更新顶部的第一条新闻数据。另外，增加'无限滚动'体验：拉到页面最底部时，要自动触发加载更多新闻的功能，底部要显示一个'正在加载更多...'的动画和文字提示，加载完成后把新数据追加到列表最后。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样，替换此前的注释关键词限制要求；第二，限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目背景，定义常见的代码及配置文本文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.txt', '.ts'}\\n \\n if not os.path.exists(workspace_path):\\n # 如果路径不存在，视作没有违反约束（因没有生成代码文件）\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制模式读取文件\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试使用 utf-8 解码\\n raw_data.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 如果解码报错，说明存在非 UTF-8 编码的代码文件\\n return False\\n except Exception:\\n # 对于其他的异常（例如权限问题等），暂且忽略继续检查\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n import ast\\n\\n def is_function_context(context: str) -> bool:\\n \\\"\\\"\\\"根据大括号前方的上下文，判断当前块是否属于一个函数定义\\\"\\\"\\\"\\n context = context.strip()\\n if not context:\\n return False\\n \\n last_char = context[-1]\\n \\n # 明确不可能作为函数定义大括号前驱的字符（排除了传参对象字面量、数组等情况）\\n if last_char in ('=', '(', ',', '[', ':', '?', '+', '-', '*', '/', '&', '|', '!', '<', '.', ';'):\\n return False\\n \\n # 如果上下文中包含箭头符号，多为箭头函数\\n if '=>' in context:\\n return True\\n \\n # 包含 function 关键字\\n if re.search(r'\\\\bfunction\\\\b', context):\\n return True\\n \\n # 正则匹配常规的形如 methodName(...) 或者 methodName(...): ReturnType 的结构\\n try:\\n matches = list(re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:<[^>]+>\\\\s*)?\\\\(', context))\\n if matches:\\n # 排除可能带括号且紧接大括号的控制流等关键字\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with', 'return', 'typeof', 'instanceof'}\\n last_match = matches[-1].group(1)\\n if last_match not in keywords:\\n return True\\n except re.error:\\n pass\\n\\n return False\\n\\n def check_js_functions_length(code: str, max_lines: int) -> bool:\\n \\\"\\\"\\\"检查一段 JS/TS 代码内的函数长度是否在限制内\\\"\\\"\\\"\\n in_string = False\\n string_char = ''\\n in_single_comment = False\\n in_multi_comment = False\\n in_template_string = False\\n \\n blocks = []\\n stack = []\\n \\n i = 0\\n length = len(code)\\n \\n # 建立行号映射表，便于快速从字符索引转化为行号\\n lines_start = [0]\\n for idx, char in enumerate(code):\\n if char == '\\\\n':\\n lines_start.append(idx + 1)\\n \\n def get_line_number(index: int) -> int:\\n return bisect.bisect_right(lines_start, index)\\n \\n while i < length:\\n char = code[i]\\n \\n # 单行注释处理\\n if in_single_comment:\\n if char == '\\\\n':\\n in_single_comment = False\\n i += 1\\n continue\\n \\n # 多行注释处理\\n if in_multi_comment:\\n if char == '*' and i + 1 < length and code[i+1] == '/':\\n in_multi_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n # 模板字符串处理\\n if in_template_string:\\n if char == '\\\\\\\\':\\n i += 2\\n continue\\n if char == '`':\\n in_template_string = False\\n i += 1\\n continue\\n \\n # 普通字符串处理\\n if in_string:\\n if char == '\\\\\\\\':\\n i += 2\\n continue\\n if char == string_char:\\n in_string = False\\n i += 1\\n continue\\n \\n # 检查注释开头\\n if char == '/' and i + 1 < length:\\n if code[i+1] == '/':\\n in_single_comment = True\\n i += 2\\n continue\\n if code[i+1] == '*':\\n in_multi_comment = True\\n i += 2\\n continue\\n \\n # 检查字符串开头\\n if char in (\\\"'\\\", '\\\"'):\\n in_string = True\\n string_char = char\\n i += 1\\n continue\\n \\n # 检查模板字符串开头\\n if char == '`':\\n in_template_string = True\\n i += 1\\n continue\\n \\n # 捕获代码块大括号层级\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start_i = stack.pop()\\n blocks.append((start_i, i))\\n \\n i += 1\\n \\n # 遍历所有捕获到的代码块，对超出规定行数的进行函数特征验证\\n for start_i, end_i in blocks:\\n start_line = get_line_number(start_i)\\n end_line = get_line_number(end_i)\\n \\n lines_count = end_line - start_line + 1\\n if lines_count > max_lines:\\n lookback_start = max(0, start_i - 300)\\n pre_text = code[lookback_start:start_i]\\n # 用分号和闭合花括号做切分，获取紧挨大括号前面的完整上下文特征\\n parts = re.split(r'[;}{]', pre_text)\\n context = parts[-1] if parts else ''\\n \\n # 若经判断确为函数，代表该函数超出行数限制\\n if is_function_context(context):\\n return False\\n \\n return True\\n\\n def check_python_functions_length(code: str, max_lines: int) -> bool:\\n \\\"\\\"\\\"检查 Python 代码文件的函数长度（借助 AST）\\\"\\\"\\\"\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n end = node.end_lineno\\n else:\\n end = max((getattr(n, 'lineno', start) for n in ast.walk(node)), default=start)\\n if end - start + 1 > max_lines:\\n return False\\n except SyntaxError:\\n pass\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前代码项目文件，基于不同后缀作相应分析处理\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\\n if not check_js_functions_length(content, 60):\\n return False\\n elif ext in ['.html', '.htm', '.vue']:\\n script_contents = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for sc in script_contents:\\n if not check_js_functions_length(sc, 60):\\n return False\\n elif ext == '.py':\\n if not check_python_functions_length(content, 60):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断数字是否为质数\\n def is_prime(n):\\n if n < 2:\\n return False\\n if n in (2, 3):\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 辅助函数：通过正则表达式提取前端 JS/TS 中的变量名\\n def extract_js_vars(code):\\n vars_set = set()\\n # 清除注释避免正则误匹配\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n # 1. 匹配 var/let/const 声明的单变量\\n for match in re.finditer(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', code):\\n vars_set.add(match.group(1))\\n \\n # 2. 匹配 function 声明的函数名\\n for match in re.finditer(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', code):\\n vars_set.add(match.group(1))\\n \\n # 3. 匹配 function() 中的参数\\n for match in re.finditer(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', code):\\n params = match.group(1).split(',')\\n for p in params:\\n p = p.strip()\\n if p:\\n m = re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\\n if m:\\n vars_set.add(m.group(1))\\n \\n # 4. 匹配箭头函数多参数形式 (a, b) =>\\n for match in re.finditer(r'\\\\(([^)]*)\\\\)\\\\s*=>', code):\\n params = match.group(1).split(',')\\n for p in params:\\n p = p.strip()\\n if p:\\n m = re.match(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\\n if m:\\n vars_set.add(m.group(1))\\n \\n # 5. 匹配箭头函数单参数形式 a =>\\n for match in re.finditer(r'(?:^|[^a-zA-Z0-9_$])([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', code):\\n vars_set.add(match.group(1))\\n \\n return vars_set\\n\\n # 辅助函数：通过 AST 提取 Python 中的变量名\\n def extract_py_vars(code):\\n vars_set = set()\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name):\\n if isinstance(node.ctx, ast.Store):\\n vars_set.add(node.id)\\n elif isinstance(node, ast.arg):\\n vars_set.add(node.arg)\\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):\\n vars_set.add(node.name)\\n except Exception:\\n pass\\n return vars_set\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n vars_in_file = set()\\n if ext == '.py':\\n vars_in_file = extract_py_vars(content)\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n vars_in_file = extract_js_vars(content)\\n elif ext in ['.html', '.htm']:\\n for script_match in re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE):\\n script_content = script_match.group(1)\\n vars_in_file.update(extract_js_vars(script_content))\\n \\n # 校验每一个提取出来的变量名长度\\n for v in vars_in_file:\\n v_clean = v.strip()\\n if not is_prime(len(v_clean)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def get_comments(content, ext):\\n comments = []\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.json']:\\n # 提取多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 提取单行注释 // ... ，排除了 http:// 这种常见的误报\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n elif ext in ['.html', '.htm', '.vue']:\\n # 提取 HTML 注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # HTML 文件内的 script 或 style 标签也可能包含 JS/CSS 注释\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n elif ext in ['.py', '.rb', '.sh']:\\n # 提取脚本语言的单行注释 # ...\\n comments.extend(re.findall(r'#.*', content))\\n # 简单兼容 Python 的多行注释字符串\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n return comments\\n\\n valid_exts = [\\n '.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.json',\\n '.html', '.htm', '.vue',\\n '.py', '.rb', '.sh'\\n ]\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 获取该文件内所有的注释内容列表\\n comments = get_comments(content, ext)\\n \\n # 遍历每段注释，检查是否有字符被判断为 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n # 如果发现注释中含有 emoji，直接判定为不满足要求\\n return False\\n \\n # 检查完所有文件，均未在注释中发现 emoji，返回 True 满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_even_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能的代码文件或文本数据类型后缀\\n valid_extensions = {'.html', '.css', '.js', '.json', '.md'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 计算包含空行在内的总行数\\n if not content:\\n line_count = 0\\n else:\\n line_count = len(content.split('\\\\n'))\\n \\n # 如果有任何一个代码文件的行数不是偶数，则要求未被满足\\n if line_count % 2 != 0:\\n return False\\n except Exception:\\n # 对于不可读或非规范的文本文件跳过处理\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义禁止在注释中出现的关键词列表\\n forbidden_keywords = [\\n \\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\", \\\"草稿\\\", \\n \\\"待完善\\\", \\\"TODO\\\", \\\"待优化\\\", \\\"废弃\\\", \\\"占位\\\"\\n ]\\n # 转为全小写以便进行大小写不敏感匹配\\n forbidden_lower = [kw.lower() for kw in forbidden_keywords]\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 正则表达式：用于提取 JS/TS/CSS 风格文件中的注释\\n # 第 1 部分（不捕获）：匹配单双引号及反引号字符串，防止误伤字符串内的 // 或 /*\\n # 第 2 部分（捕获组 1）：匹配单行注释 //，加上 (?<!:) 是为了避免匹配到 http:// 等 URL\\n # 第 3 部分（捕获组 2）：匹配多行注释 /* ... */\\n pattern_c_style = re.compile(\\n r'(?:\\\"(?:\\\\\\\\\\\"|[^\\\"])*\\\"|\\\\'(?:\\\\\\\\\\\\'|[^\\\\'])*\\\\'|`(?:\\\\\\\\`|[\\\\s\\\\S])*?`)|((?<!:)//.*?$)|(/\\\\*[\\\\s\\\\S]*?\\\\*/)', \\n re.MULTILINE\\n )\\n \\n # 正则表达式：用于提取 HTML 风格的注释\\n pattern_html = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n \\n # 正则表达式：用于提取 Python/Shell 风格文件中的注释\\n pattern_py = re.compile(\\n r'(?:\\\"(?:\\\\\\\\\\\"|[^\\\"])*\\\"|\\\\'(?:\\\\\\\\\\\\'|[^\\\\'])*\\\\'|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\')|(#.*?$)', \\n re.MULTILINE\\n )\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉无需检查的常见系统目录及依赖目录\\n for ignore_dir in ['.git', 'node_modules', '__pycache__', 'dist', 'build']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue # 如果文件无法读取（例如编码问题或权限不足）则跳过\\n\\n comments = []\\n\\n # 对于前端及其他常见 C 风格注释文件\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm', '.vue']:\\n # 若为 HTML 类文件，先提取出 <!-- --> 注释\\n if ext in ['.html', '.htm', '.vue']:\\n for m in pattern_html.finditer(content):\\n comments.append(m.group(0))\\n \\n # HTML 中也可能内联了 <script> 和 <style>，因此统一再使用 C 风格匹配\\n for m in pattern_c_style.finditer(content):\\n if m.group(1): # 捕获到单行注释\\n comments.append(m.group(1))\\n elif m.group(2): # 捕获到多行注释\\n comments.append(m.group(2))\\n \\n # 对于 Python 或 Shell 等脚本文件\\n elif ext in ['.py', '.sh']:\\n for m in pattern_py.finditer(content):\\n if m.group(1): # 捕获到 # 号注释\\n comments.append(m.group(1))\\n\\n # 检查提取到的所有注释文本\\n for comment in comments:\\n comment_lower = comment.lower()\\n for kw in forbidden_lower:\\n if kw in comment_lower:\\n # 一旦在任意一条注释里发现任意一个限制关键词，立刻返回False\\n return False\\n\\n # 若所有文件的全部注释均未查到限制关键词，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对PWA前端应用，定义常见的代码文件扩展名集合\\n valid_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue', '.webmanifest'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录（例如 .git, .vscode 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 遇到无法使用UTF-8读取的文件直接跳过（该要求与空格无关，防止阻塞验证）\\n continue\\n \\n for line in lines:\\n # 仅去除尾部的回车及换行符，保留可能存在的尾随空格\\n line = line.rstrip('\\\\n\\\\r')\\n \\n for idx, char in enumerate(line):\\n if char == '=':\\n is_combined = False\\n \\n # 检查左侧是否为组合符号的一部分 (涵盖 JS/CSS 等常见组合等号符)\\n if idx > 0 and line[idx-1] in '=!<>+-*/%&|^~$':\\n is_combined = True\\n \\n # 检查右侧是否为组合符号的一部分 (如 ==, ===, =>)\\n if idx < len(line) - 1 and line[idx+1] in '=>':\\n is_combined = True\\n \\n # 当它是一个独立的 '=' 时，验证两边是否恰好有两个空格\\n if not is_combined:\\n # 检查左侧\\n left_ok = False\\n if idx >= 2 and line[idx-1] == ' ' and line[idx-2] == ' ':\\n # 确保前第三个字符不是空格，或者等号正好处于只有左边两个空格的行首\\n if idx == 2 or line[idx-3] != ' ':\\n left_ok = True\\n \\n # 检查右侧\\n right_ok = False\\n if idx <= len(line) - 3 and line[idx+1] == ' ' and line[idx+2] == ' ':\\n # 确保后第三个字符不是空格，或者等号正好处于只有右边两个空格的行尾\\n if idx == len(line) - 3 or line[idx+3] != ' ':\\n right_ok = True\\n \\n # 如果有任何一端不满足要求，就视为不合规\\n if not (left_ok and right_ok):\\n return False\\n \\n # 只要包含了相关代码文件并且全部校验通过即返回True；如果是空目录则返回False\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的注释必须全部使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，该PWA应用涉及的核心代码文件类型为主流前端文件格式\\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果代码文件无法正常以 UTF-8 读取，或者出现其他IO异常，直接视为不满足\\n return False\\n \\n # 收集文件中所有可能被识别为注释的内容\\n comments = []\\n \\n # 1. HTML / SVG 风格注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 2. C / JS / CSS 风格多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n # 3. JS / TS 风格单行注释 // ...\\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\\n \\n # 将所有提取到的注释拼接在一起\\n all_comments = \\\" \\\".join(comments)\\n \\n # 检查注释部分是否包含指定的字样\\n if \\\"PWA新闻聚合器\\\" not in all_comments:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"为了更方便找新闻，需要增加搜索功能。在顶部 App Bar 标题旁边放一个搜索图标（放大镜），点击后应用标题变成一个输入框。用户输入关键字时，下方实时过滤当前的新闻列表数据进行展示。重点要求：在展现的搜索结果卡片中，无论是新闻标题还是摘要里，只要匹配到用户输入的搜索关键词，就必须用醒目的背景色（比如亮黄色）将其高亮显示出来。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md，替换此前的修改计划记录文件要求；第二，要求输出代码时，统一使用4个空格作为缩进标准。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 定义前端项目中可能生成的代码文件后缀名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.json', '.md', \\n '.ts', '.jsx', '.tsx', '.vue', '.txt', '.xml', \\n '.svg', '.webmanifest', '.mjs', '.cjs'\\n }\\n \\n # 如果工作区路径不存在，从严谨角度认为没有违反编码要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅判断代码和文本相关文件，排除图片等二进制文件的干扰\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 格式读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件并非合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 其他读取异常（如权限等非编码导致的问题）忽略不计\\n pass\\n \\n # 所有被找到的代码文件均成功以 UTF-8 读取，返回 True\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\n def clean_js_code(code: str) -> str:\\n \\\"\\\"\\\"\\n 利用状态机清洗JS代码，将注释、字符串字面量均替换为空格，但保留换行符，\\n 保证返回的代码在字符和行数位置上与原代码一一对应，同时排除大括号误判。\\n \\\"\\\"\\\"\\n state_stack = ['NORMAL']\\n result = []\\n i = 0\\n n = len(code)\\n while i < n:\\n current_state = state_stack[-1]\\n \\n if current_state in ('NORMAL', 'TEMPLATE_EXPR'):\\n if i + 1 < n and code[i:i+2] == '//':\\n state_stack.append('LINE_COMMENT')\\n result.append(' ')\\n i += 2\\n elif i + 1 < n and code[i:i+2] == '/*':\\n state_stack.append('BLOCK_COMMENT')\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n state_stack.append('STRING_D')\\n result.append(' ')\\n i += 1\\n elif code[i] == \\\"'\\\":\\n state_stack.append('STRING_S')\\n result.append(' ')\\n i += 1\\n elif code[i] == '`':\\n state_stack.append('STRING_T')\\n result.append(' ')\\n i += 1\\n elif current_state == 'TEMPLATE_EXPR' and code[i] == '}':\\n state_stack.pop()\\n result.append(' ')\\n i += 1\\n elif current_state == 'TEMPLATE_EXPR' and code[i] == '{':\\n state_stack.append('BLOCK')\\n result.append(code[i])\\n i += 1\\n elif current_state == 'NORMAL' and code[i] == '{':\\n result.append(code[i])\\n i += 1\\n elif current_state == 'NORMAL' and code[i] == '}':\\n result.append(code[i])\\n i += 1\\n else:\\n result.append(code[i])\\n i += 1\\n \\n elif current_state == 'BLOCK':\\n if code[i] == '}':\\n state_stack.pop()\\n result.append(code[i])\\n i += 1\\n elif code[i] == '{':\\n state_stack.append('BLOCK')\\n result.append(code[i])\\n i += 1\\n elif i + 1 < n and code[i:i+2] == '//':\\n state_stack.append('LINE_COMMENT')\\n result.append(' ')\\n i += 2\\n elif i + 1 < n and code[i:i+2] == '/*':\\n state_stack.append('BLOCK_COMMENT')\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n state_stack.append('STRING_D')\\n result.append(' ')\\n i += 1\\n elif code[i] == \\\"'\\\":\\n state_stack.append('STRING_S')\\n result.append(' ')\\n i += 1\\n elif code[i] == '`':\\n state_stack.append('STRING_T')\\n result.append(' ')\\n i += 1\\n else:\\n result.append(code[i])\\n i += 1\\n \\n elif current_state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state_stack.pop()\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n \\n elif current_state == 'BLOCK_COMMENT':\\n if i + 1 < n and code[i:i+2] == '*/':\\n state_stack.pop()\\n result.append(' ')\\n i += 2\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n \\n elif current_state == 'STRING_D':\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n state_stack.pop()\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n \\n elif current_state == 'STRING_S':\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif code[i] == \\\"'\\\":\\n state_stack.pop()\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n \\n elif current_state == 'STRING_T':\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif i + 1 < n and code[i:i+2] == '${':\\n state_stack.append('TEMPLATE_EXPR')\\n result.append(' ')\\n i += 2\\n elif code[i] == '`':\\n state_stack.pop()\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def is_js_function_block(cleaned_code: str, start_idx: int) -> bool:\\n \\\"\\\"\\\"\\n 判断匹配到的 { } 块是否为函数\\n \\\"\\\"\\\"\\n idx = start_idx - 1\\n while idx >= 0 and cleaned_code[idx].isspace():\\n idx -= 1\\n if idx < 0:\\n return False\\n \\n char = cleaned_code[idx]\\n \\n # 处理箭头函数如: () => { ... } \\n if char == '>':\\n if idx - 1 >= 0 and cleaned_code[idx-1] == '=':\\n return True\\n return False\\n \\n # 处理普通函数/方法如: function foo() { ... } 或者 method() { ... }\\n if char == ')':\\n paren_count = 1\\n p_idx = idx - 1\\n while p_idx >= 0:\\n if cleaned_code[p_idx] == ')':\\n paren_count += 1\\n elif cleaned_code[p_idx] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n p_idx -= 1\\n \\n if p_idx < 0:\\n return False\\n \\n w_idx = p_idx - 1\\n while w_idx >= 0 and cleaned_code[w_idx].isspace():\\n w_idx -= 1\\n \\n if w_idx < 0:\\n return True\\n \\n word_end = w_idx\\n while w_idx >= 0 and (cleaned_code[w_idx].isalnum() or cleaned_code[w_idx] in '_$'):\\n w_idx -= 1\\n \\n word = cleaned_code[w_idx+1 : word_end+1]\\n \\n # 排除控制流关键字\\n if word in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\\n return False\\n else:\\n return True\\n \\n return False\\n\\n def check_function_length_js(content: str) -> bool:\\n cleaned = clean_js_code(content)\\n stack = []\\n \\n for i, char in enumerate(cleaned):\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start = stack.pop()\\n if is_js_function_block(cleaned, start):\\n lines = content.count('\\\\n', start, i) + 1\\n if lines > 60:\\n return False\\n return True\\n\\n def check_html_file(content: str) -> bool:\\n # 寻找HTML中的内联<script>代码\\n pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\\n for match in pattern.finditer(content):\\n script_content = match.group(1)\\n if not check_function_length_js(script_content):\\n return False\\n return True\\n\\n def check_function_length_py(content: str) -> bool:\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n length = getattr(node, 'end_lineno', node.lineno) - node.lineno + 1\\n if length > 60:\\n return False\\n return True\\n except Exception:\\n # 如遭遇语法解析错误则跳过该文件处理\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext in {'.js', '.ts', '.jsx', '.tsx'}:\\n if not check_function_length_js(content):\\n return False\\n elif ext in {'.html', '.htm', '.vue'}:\\n if not check_html_file(content):\\n return False\\n elif ext == '.py':\\n if not check_function_length_py(content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"def check_variable_names_length(response, workspace_path):\\n import os\\n import re\\n \\n def is_prime(n):\\n \\\"\\\"\\\"判断一个数是否为质数\\\"\\\"\\\"\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n \\n def clean_code(code):\\n \\\"\\\"\\\"清洗代码，去除注释和字符串以防止误提取\\\"\\\"\\\"\\n # 移除单行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除多行注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除双引号字符串\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', code)\\n # 移除单引号字符串\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", code)\\n # 移除反引号字符串模板\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', code)\\n return code\\n\\n def extract_variables(code):\\n \\\"\\\"\\\"从JavaScript/TypeScript代码中提取变量、函数和类名\\\"\\\"\\\"\\n variables = set()\\n \\n # 匹配 var, let, const 声明的常规变量名（排除部分复杂解构赋值的情况以免误伤）\\n pattern = r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b'\\n for match in re.finditer(pattern, code):\\n variables.add(match.group(1))\\n \\n # 匹配具名函数定义中的函数名和参数\\n func_pattern = r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(([^)]*)\\\\)'\\n for match in re.finditer(func_pattern, code):\\n variables.add(match.group(1)) # 提取函数名\\n params_str = match.group(2)\\n if params_str.strip():\\n for p in params_str.split(','):\\n p_name = p.split('=')[0].strip() # 剥除默认值部分\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\\n variables.add(p_name)\\n \\n # 匹配匿名函数中的参数\\n anon_func_pattern = r'\\\\bfunction\\\\s*\\\\(([^)]*)\\\\)'\\n for match in re.finditer(anon_func_pattern, code):\\n params_str = match.group(1)\\n if params_str.strip():\\n for p in params_str.split(','):\\n p_name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\\n variables.add(p_name)\\n \\n # 匹配类名\\n class_pattern = r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b'\\n for match in re.finditer(class_pattern, code):\\n variables.add(match.group(1))\\n \\n return variables\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n code_blocks = []\\n if file.endswith(('.js', '.ts')):\\n code_blocks.append(content)\\n elif file.endswith(('.html', '.htm', '.vue', '.jsx', '.tsx')):\\n # 提取 HTML 类型文件中的脚本标签内容\\n script_pattern = re.compile(r'<script[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\\n scripts = script_pattern.findall(content)\\n code_blocks.extend(scripts)\\n \\n for code in code_blocks:\\n cleaned_code = clean_code(code)\\n vars_in_code = extract_variables(cleaned_code)\\n for v in vars_in_code:\\n # 获取长度，严格判断是否为质数\\n if not is_prime(len(v)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义各类代码文件可能出现的注释正则表达式\\n # 注意：(?<!:)//.* 用于避免匹配到 http:// 或 https:// 这样的URL片段\\n comment_patterns = {\\n '.html': [r'<!--[\\\\s\\\\S]*?-->'],\\n '.htm': [r'<!--[\\\\s\\\\S]*?-->'],\\n '.css': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/'],\\n '.js': [r'(?<!:)//.*', r'/\\\\*[\\\\s\\\\S]*?\\\\*/'],\\n '.ts': [r'(?<!:)//.*', r'/\\\\*[\\\\s\\\\S]*?\\\\*/'],\\n '.jsx': [r'(?<!:)//.*', r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'<!--[\\\\s\\\\S]*?-->'],\\n '.tsx': [r'(?<!:)//.*', r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'<!--[\\\\s\\\\S]*?-->'],\\n '.vue': [r'(?<!:)//.*', r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'<!--[\\\\s\\\\S]*?-->'],\\n '.py': [r'#.*', r\\\"'''[\\\\s\\\\S]*?'''\\\", r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"']\\n }\\n\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理可能包含注释的已知代码文件\\n if ext in comment_patterns:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于编码或其他原因无法读取文件，则跳过\\n continue\\n \\n # 获取该文件对应的注释匹配规则\\n patterns = comment_patterns[ext]\\n for pattern in patterns:\\n # 匹配出所有的注释块\\n comments = re.findall(pattern, content)\\n for comment in comments:\\n # 遍历注释内容中的每个字符，检查是否为emoji\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"def check_file_lines_even(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据对话上下文中PWA项目以及要求的计划文件，推断出相关的代码文件扩展名\\n target_extensions = ('.html', '.css', '.js', '.json', '.md')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(target_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 按照换行符分割，这样即使包含纯空行（无论是中间还是文件末尾）也能被准确计入行数\\n lines_count = len(content.split('\\\\n'))\\n \\n if lines_count % 2 != 0:\\n return False\\n except Exception:\\n # 忽略读取异常的文件，防止因某些特殊的编码问题导致验证程序崩溃\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 待检查的禁用词汇（全小写，便于后续统一转小写匹配）\\n keywords = [\\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\", \\\"草稿\\\", \\\"待完善\\\", \\\"todo\\\", \\\"待优化\\\", \\\"废弃\\\", \\\"占位\\\"]\\n \\n # 构造正则表达式\\n # (?P<string>...) 组：匹配由于转义和字符闭合所包裹的字符串，防止其内部出现的 `//` 等被误认作注释\\n # (?P<comment>...) 组：匹配常见单行及多行注释格式，包括 HTML、JS、CSS、Python 等语言规范的注释\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' \\n r'|(?P<comment>//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->|#.*)'\\n )\\n \\n def check_content(content: str) -> bool:\\n \\\"\\\"\\\"检查特定文本内容中的注释是否包含禁用字样\\\"\\\"\\\"\\n for match in pattern.finditer(content):\\n comment = match.group('comment')\\n # 只有当捕获到注释时才进行匹配\\n if comment:\\n comment_lower = comment.lower()\\n for kw in keywords:\\n if kw in comment_lower:\\n return False\\n return True\\n\\n # 1. 检查 workspace_path 目录下的所有相关代码文件\\n if os.path.isdir(workspace_path):\\n # 针对该 PWA 聚合新闻项目的可能文件格式\\n valid_exts = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py', '.json'}\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_content(content):\\n return False\\n except Exception:\\n # 忽略无法读取的文件\\n continue\\n \\n # 2. 检查 response 回复体中的 markdown 代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9]*\\\\n([\\\\s\\\\S]*?)```', response)\\n for code in code_blocks:\\n if not check_content(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 匹配独立的等号 \\\"=\\\"，排除诸如 ==, ===, !=, !==, >=, <=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~=, :=, ?= 等符号\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%&|^?:~])=(?![=>])')\\n\\n def check_content(content: str) -> bool:\\n matches = list(pattern.finditer(content))\\n for match in matches:\\n idx = match.start()\\n \\n # 检查等号左侧是否恰好有两个空格\\n if idx < 2:\\n return False\\n if content[idx-2:idx] != ' ':\\n return False\\n # 确保左侧第3个字符不是空格，保证“只有”两个空格\\n if idx >= 3 and content[idx-3] == ' ':\\n return False\\n \\n # 检查等号右侧是否恰好有两个空格\\n if idx + 3 > len(content):\\n return False\\n if content[idx+1:idx+3] != ' ':\\n return False\\n # 确保右侧第3个字符不是空格，保证“只有”两个空格\\n if idx + 3 < len(content) and content[idx+3] == ' ':\\n return False\\n \\n return True\\n\\n # 1. 检查工作区中的所有代码文件\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤常见的代码及相关配置文件类型\\n if file.endswith(('.py', '.js', '.html', '.css', '.json', '.md', '.vue', '.ts', '.jsx', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_content(content):\\n return False\\n except Exception:\\n pass\\n \\n # 2. 检查回复中包含的代码块\\n code_blocks = re.findall(r'```[a-zA-Z]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n if not check_content(block):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的注释必须全部使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断每个生成的代码文件的注释部分是否都呈现了“PWA新闻聚合器”字样。\\n \\\"\\\"\\\"\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n \\n # 定义需要检查的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.cs'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果无法按utf-8读取，可以跳过该文件或者视为不满足。\\n # 这里选择稳妥方式，跳过异常文件。\\n continue\\n \\n # 提取不同语言的各类注释\\n comments = []\\n \\n # HTML 格式注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # C风格块注释 (JS/CSS/TS/C/Java等)\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # C风格单行注释 (JS/TS/C/Java等)\\n comments.extend(re.findall(r'//.*', content))\\n # Python/Shell 单行注释\\n comments.extend(re.findall(r'#.*', content))\\n # Python 多行注释 (文档字符串)\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n \\n # 合并当前文件所有提取出的注释\\n all_comments_text = '\\\\n'.join(comments)\\n \\n # 如果该代码文件的所有注释中未包含指定字样，则直接判定为不满足要求\\n if \\\"PWA新闻聚合器\\\" not in all_comments_text:\\n return False\\n\\n # 若遍历完毕都没有发现任何代码文件，则不符合\\\"生成的每个代码文件\\\"的语境前提\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"import os\\n\\ndef check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n # 设定需要进行缩进检查的常见代码文件后缀\\n code_extensions = {'.html', '.js', '.css', '.json', '.py', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_files = False\\n \\n # 遍历当前工作区的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 忽略无法用utf-8读取的文件\\n continue\\n \\n has_code_files = True\\n \\n tab_count = 0\\n mod4_0_count = 0\\n mod4_not_0_count = 0\\n \\n for line in lines:\\n # 忽略完全无内容的空行\\n if not line.strip():\\n continue\\n \\n # 截取前导空白字符\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 忽略无缩进的行\\n if not leading_whitespace:\\n continue\\n \\n # 检查是否混用了制表符\\n if '\\\\t' in leading_whitespace:\\n tab_count += 1\\n continue\\n \\n indent_len = len(leading_whitespace)\\n \\n # 统计前导空格是否为 4 的倍数\\n if indent_len % 4 == 0:\\n mod4_0_count += 1\\n else:\\n stripped_line = line.lstrip()\\n # 排除 JSDoc 等多行注释中常见的使用 '*' 对齐的情况\\n if stripped_line.startswith('*'):\\n continue\\n mod4_not_0_count += 1\\n \\n # 规则1：如果使用了 tab 缩进，则未“统一使用 4 个空格”，不满足要求\\n if tab_count > 0:\\n return False\\n \\n total_indents = mod4_0_count + mod4_not_0_count\\n if total_indents > 0:\\n # 规则2：如果文件存在缩进行，但没有一个是4的倍数，证明没有使用 4 空格标准\\n if mod4_0_count == 0 and mod4_not_0_count > 0:\\n return False\\n \\n # 规则3：如果非 4 个空格倍数的缩进行数占比较高（超过30%），且行数达到至少3行。\\n # 这种情况通常说明代码错误地使用了如 2 个空格作为缩进标准。\\n if (mod4_not_0_count / total_indents) > 0.3 and mod4_not_0_count >= 3:\\n return False\\n \\n # 如果找到了代码文件且均通过了检查，才视作满足要求\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"最后一步，我们要强化它的 PWA 特性和离线可用性体验。首先在 UI 层面，编写 JavaScript 监听设备的 online 和 offline 网络状态事件。如果检测到处于离线状态，要在页面顶部向下滑出一个红色的警示条，提示'当前处于离线模式，只能查看已缓存内容'，并在离线期间禁用下拉刷新功能。同时，模拟 Service Worker 的离线缓存机制，使得即使用户在断网状态下，依然能正常打开底部'书签'页面，且能点击阅读之前收藏过的新闻详情页面。另外请你注意以下两项后续所有轮次都需要严格遵守的要求：第一，凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json，替换此前的修改计划记录文件要求；第二，运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 验证工作区目录是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令（PWA应用），推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts'}\\n has_code_file = False\\n \\n # 遍历工作区内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤筛选代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试以 UTF-8 编码读取该文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生 UnicodeDecodeError，说明该文件不是使用合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如文件权限等导致的读取错误\\n pass\\n \\n # 如果项目中存在代码文件并且全部通过验证，则认为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 60 行的限制\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 清理JS/TS代码，去除会干扰括号匹配的字符串和注释，同时保留换行符保持行号一致\\n def clean_js_code_preserve_lines(js_content):\\n # 替换多行注释为相应数量的换行符\\n def multiline_repl(match):\\n return '\\\\n' * match.group(0).count('\\\\n')\\n js_content = re.sub(r'/\\\\*.*?\\\\*/', multiline_repl, js_content, flags=re.DOTALL)\\n \\n # 移除单行注释\\n js_content = re.sub(r'//.*', '', js_content)\\n \\n # 替换字符串并保留内部的换行符\\n def string_repl(match):\\n return '\\\"\\\"' + '\\\\n' * match.group(0).count('\\\\n')\\n js_content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', string_repl, js_content)\\n js_content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', js_content)\\n js_content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", js_content)\\n \\n return js_content\\n\\n # 通过大括号特征匹配JS函数及其行数\\n def check_js_functions(js_content):\\n cleaned = clean_js_code_preserve_lines(js_content)\\n current_line = 1\\n stack = []\\n \\n for idx, char in enumerate(cleaned):\\n if char == '\\\\n':\\n current_line += 1\\n continue\\n \\n if char == '{':\\n # 获取该大括号之前的上下文，直到遇到上一个截断符号\\n context = \\\"\\\"\\n for j in range(idx - 1, -1, -1):\\n if cleaned[j] in ['{', '}', ';']:\\n break\\n context = cleaned[j] + context\\n \\n context = context.replace('\\\\n', ' ').strip()\\n \\n # 判断当前大括号是否属于函数体\\n is_func = False\\n # 排除典型的控制流语句和类定义\\n if re.search(r'\\\\b(?:catch|if|for|while|switch|with|else|try|finally|do)\\\\b', context):\\n is_func = False\\n elif re.search(r'\\\\bclass\\\\b', context):\\n is_func = False\\n elif 'function' in context:\\n is_func = True\\n elif '=>' in context:\\n is_func = True\\n # 方法声明（往往在大括号前具有成对或闭合的圆括号结构）\\n elif re.search(r'\\\\)\\\\s*$', context):\\n is_func = True\\n \\n stack.append((current_line, is_func))\\n \\n elif char == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n # 发现超长函数，即刻返回False\\n if (current_line - start_line + 1) > 60:\\n return False\\n return True\\n\\n # 使用AST解析检测Python文件中的函数\\n def check_py_functions(py_content):\\n try:\\n tree = ast.parse(py_content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n end = getattr(node, 'end_lineno', start)\\n if (end - start + 1) > 60:\\n return False\\n return True\\n except SyntaxError:\\n return True\\n\\n # 遍历代码区所有可解析文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n if file.endswith(('.js', '.ts')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_js_functions(content):\\n return False\\n except:\\n pass\\n \\n elif file.endswith(('.html', '.htm', '.vue')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 提取内嵌的脚本块进行判断\\n script_blocks = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, flags=re.IGNORECASE | re.DOTALL)\\n for block in script_blocks:\\n if not check_js_functions(block):\\n return False\\n except:\\n pass\\n \\n elif file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_py_functions(content):\\n return False\\n except:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"代码文件中的命名要求是：每个变量名的字符长度必须正好是质数\",\n \"validation_code\": \"def check_variable_name_prime_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断一个数字是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n if n in (2, 3):\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n for i in range(5, int(n ** 0.5) + 1, 6):\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n return True\\n\\n # 匹配常规的 var/let/const 声明的变量\\n js_var_pattern = re.compile(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n # 匹配对象解构赋值中的变量，例如：const { a, b } = obj;\\n js_destruct_obj_pattern = re.compile(r'\\\\b(?:let|const|var)\\\\s*\\\\{([^}]+)\\\\}')\\n # 匹配数组解构赋值中的变量，例如：let [x, y] = arr;\\n js_destruct_arr_pattern = re.compile(r'\\\\b(?:let|const|var)\\\\s*\\\\[([^\\\\]]+)\\\\]')\\n \\n # 匹配可能的 Python 变量声明赋值（补充兼容）\\n py_var_pattern = re.compile(r'^[ \\\\t]*([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=[^=]', re.MULTILINE)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查前端相关代码文件\\n if ext in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue', '.css']:\\n # 清除代码中的多行与单行注释，避免误匹配\\n content_clean = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content_clean = re.sub(r'//.*', '', content_clean)\\n \\n # 1. 验证基础变量声明\\n for var_name in js_var_pattern.findall(content_clean):\\n if not is_prime(len(var_name)):\\n return False\\n \\n # 2. 验证对象解构赋值出的变量\\n for group in js_destruct_obj_pattern.findall(content_clean):\\n for var_name in re.findall(r'([a-zA-Z_$][a-zA-Z0-9_$]*)', group):\\n if not is_prime(len(var_name)):\\n return False\\n \\n # 3. 验证数组解构赋值出的变量\\n for group in js_destruct_arr_pattern.findall(content_clean):\\n for var_name in re.findall(r'([a-zA-Z_$][a-zA-Z0-9_$]*)', group):\\n if not is_prime(len(var_name)):\\n return False\\n \\n # 4. 验证 CSS 变量 (形如 --main-color)\\n if ext in ['.css', '.html', '.vue']:\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', content_clean)\\n for var_name in css_vars:\\n # 拼接为完整的 CSS 变量名检查长度\\n full_name = f\\\"--{var_name}\\\"\\n if not is_prime(len(full_name)):\\n return False\\n\\n # 检查可能的 Python 脚本文件\\n elif ext == '.py':\\n # 清除 Python 的注释\\n content_clean = re.sub(r'#.*', '', content)\\n for var_name in py_var_pattern.findall(content_clean):\\n if not is_prime(len(var_name)):\\n return False\\n \\n except Exception:\\n # 忽略读取异常的文件或非文本文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论哪个代码文件，其注释块中都不能带有 emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 构造能够匹配字符串字面量和各类代码注释的正则表达式\\n # Group 1: 双引号字符串 (避免内部的双斜杠等被误判为注释)\\n # Group 2: 单引号字符串\\n # Group 3: 反引号 (ES6 模板字符串)\\n # Group 4: // 单行注释 (使用负向先行断言排除常见 http://, https://, file:// 协议)\\n # Group 5: /* ... */ 块注释\\n # Group 6: <!-- ... --> HTML 注释\\n # Group 7: # 单行注释 (Python, Shell 等)\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|'\\n r\\\"('(?:\\\\\\\\.|[^'\\\\\\\\])*')|\\\"\\n r'(`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'((?<!https:)(?<!http:)(?<!file:)//.*$)|'\\n r'(/\\\\*[\\\\s\\\\S]*?\\\\*/)|'\\n r'(<!--[\\\\s\\\\S]*?-->)|'\\n r'(#.*$)',\\n re.MULTILINE\\n )\\n\\n def has_emoji_in_comments(text: str) -> bool:\\n for match in pattern.finditer(text):\\n # 判断命中结果是否属于注释相关的 Group (即 4, 5, 6, 7 组)\\n if any(match.group(i) for i in range(4, 8)):\\n comment_text = match.group(0)\\n # 遍历注释内容中的每一个字符\\n for char in comment_text:\\n # 使用专家实现的 is_emoji 函数\\n if is_emoji(char):\\n return True\\n return False\\n\\n # 1. 首先检查回复文本中包含的所有 markdown 代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for code in code_blocks:\\n if has_emoji_in_comments(code):\\n return False\\n\\n # 2. 检查工作空间路径下的具体代码文件\\n # 针对可能出现的各类前后端代码文件进行约束检查\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', \\n '.css', '.scss', '.less', '.py', '.java', \\n '.c', '.cpp', '.cs', '.php', '.go', '.rs', '.swift'\\n }\\n\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查文件内容中的注释是否含有 emoji\\n if has_emoji_in_comments(content):\\n return False\\n except Exception:\\n # 如遇到非 utf-8 编码或无法读取的文件，则跳过\\n continue\\n\\n # 如果均未发现违规 emoji，则说明满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，在计入空行的前提下，其总行数都得是偶数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义PWA项目中常见的生成的代码及配置文件后缀\\n valid_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts'}\\n file_checked = False\\n \\n # 优先在工作区路径下检查实际生成的文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符明确切分来统计行数，能够确保所有空行均被计入\\n lines = content.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n \\n file_checked = True\\n except Exception:\\n # 忽略可能出现的文件读取异常（如编码不支持或权限问题）\\n continue\\n \\n # 如果工作区中实际检查到了代码文件并且所有文件均通过验证\\n if file_checked:\\n return True\\n \\n # 降级检查：如果工作区中没有任何文件，退回到 response 字符串中提取 Markdown 代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9]*\\\\n(.*?)```', response, re.DOTALL)\\n if code_blocks:\\n for block in code_blocks:\\n # Markdown代码块捕获的内容通常末尾会带一个不属于代码本身的换行符，需剔除\\n if block.endswith('\\\\n'):\\n block = block[:-1]\\n \\n lines = block.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n return True\\n \\n # 如果既没有创建文件，也没有在回复中输出代码块，算作不满足\\n return False\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"任何代码文件的注释部分都不得含有“测试”、“demo”、“临时”、“调试”、“草稿”、“待完善”、“TODO”、“待优化”、“废弃”、“占位”字样\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n forbidden_words = [\\\"测试\\\", \\\"demo\\\", \\\"临时\\\", \\\"调试\\\", \\\"草稿\\\", \\\"待完善\\\", \\\"TODO\\\", \\\"待优化\\\", \\\"废弃\\\", \\\"占位\\\"]\\n \\n def contains_forbidden(text: str) -> bool:\\n lower_text = text.lower()\\n for fw in forbidden_words:\\n if fw.lower() in lower_text:\\n return True\\n return False\\n\\n def extract_js_css_comments(code: str) -> list:\\n comments = []\\n state = 'NORMAL'\\n i = 0\\n n = len(code)\\n current_comment = []\\n quote_char = ''\\n while i < n:\\n if state == 'NORMAL':\\n if code[i:i+2] == '//':\\n # 防止匹配到 URL 中的 http:// 或 https:// 结构\\n if i > 0 and code[i-1] == ':':\\n i += 1\\n else:\\n state = 'LINE_COMMENT'\\n current_comment = ['//']\\n i += 2\\n elif code[i:i+2] == '/*':\\n state = 'BLOCK_COMMENT'\\n current_comment = ['/*']\\n i += 2\\n elif code[i] in \\\"\\\\\\\"'`\\\":\\n state = 'STRING'\\n quote_char = code[i]\\n i += 1\\n else:\\n i += 1\\n elif state == 'STRING':\\n if code[i] == '\\\\\\\\':\\n i += 2\\n elif code[i] == quote_char:\\n state = 'NORMAL'\\n i += 1\\n else:\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state = 'NORMAL'\\n comments.append(''.join(current_comment))\\n current_comment = []\\n i += 1\\n else:\\n current_comment.append(code[i])\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if code[i:i+2] == '*/':\\n current_comment.append('*/')\\n comments.append(''.join(current_comment))\\n current_comment = []\\n state = 'NORMAL'\\n i += 2\\n else:\\n current_comment.append(code[i])\\n i += 1\\n if current_comment:\\n comments.append(''.join(current_comment))\\n return comments\\n\\n def extract_html_comments(code: str) -> list:\\n return re.findall(r'<!--[\\\\s\\\\S]*?-->', code)\\n\\n def extract_python_comments(code: str) -> list:\\n comments = []\\n state = 'NORMAL'\\n i = 0\\n n = len(code)\\n current_comment = []\\n quote_char = ''\\n while i < n:\\n if state == 'NORMAL':\\n if code[i] == '#':\\n state = 'LINE_COMMENT'\\n current_comment = ['#']\\n i += 1\\n elif code[i:i+3] in ['\\\"\\\"\\\"', \\\"'''\\\"]:\\n state = 'MULTI_STRING'\\n quote_char = code[i:i+3]\\n i += 3\\n elif code[i] in \\\"\\\\\\\"'\\\":\\n state = 'STRING'\\n quote_char = code[i]\\n i += 1\\n else:\\n i += 1\\n elif state == 'STRING':\\n if code[i] == '\\\\\\\\':\\n i += 2\\n elif code[i] == quote_char:\\n state = 'NORMAL'\\n i += 1\\n else:\\n i += 1\\n elif state == 'MULTI_STRING':\\n if code[i] == '\\\\\\\\':\\n i += 2\\n elif code[i:i+3] == quote_char:\\n state = 'NORMAL'\\n i += 3\\n else:\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state = 'NORMAL'\\n comments.append(''.join(current_comment))\\n current_comment = []\\n i += 1\\n else:\\n current_comment.append(code[i])\\n i += 1\\n if current_comment:\\n comments.append(''.join(current_comment))\\n return comments\\n\\n # 定义目标代码文件的后缀集合\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', \\n '.css', '.scss', '.less', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.swift', '.kt'}:\\n comments.extend(extract_js_css_comments(content))\\n elif ext in {'.html', '.htm', '.vue'}:\\n comments.extend(extract_html_comments(content))\\n # 提取 HTML 内部可能包含在 <script> 或 <style> 的 JS / CSS 注释\\n comments.extend(extract_js_css_comments(content))\\n elif ext in {'.py', '.rb'}:\\n comments.extend(extract_python_comments(content))\\n\\n # 遍历所有提取出的注释内容，验证是否存在禁用关键字\\n for comment in comments:\\n if contains_forbidden(comment):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，“=”的左侧和右侧都要保持留出两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_equals_spacing(response: str, workspace_path: str) -> bool:\\n # 常见前端/通用代码文件拓展名\\n code_extensions = {'.js', '.html', '.css', '.json', '.ts', '.jsx', '.tsx', '.vue'}\\n found_code_files = False\\n \\n # 正则：匹配不属于组合运算符一部分的独立等号 '='\\n # 左侧不能有 = < > ! + - * / % & | ^ ~ ?\\n # 右侧不能有 = >\\n pattern = re.compile(r\\\"(?<![=<>!+\\\\-*/%&|^~?])=(?![=>])\\\")\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n # 遍历文件中所有匹配到的独立等号\\n for match in pattern.finditer(text):\\n start = match.start()\\n end = match.end()\\n \\n left = text[:start]\\n right = text[end:]\\n \\n # 检查等号左侧是否是以且仅以两个连续空格结尾\\n if not left.endswith(\\\" \\\") or left.endswith(\\\" \\\"):\\n return False\\n \\n # 检查等号右侧是否是以且仅以两个连续空格开头\\n if not right.startswith(\\\" \\\") or right.startswith(\\\" \\\"):\\n return False\\n\\n return found_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"针对外部 I/O 操作、网络请求和数据格式转换，生成的代码必须加入异常捕获功能，若遇异常情况需返回相应的错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的注释必须全部使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"限制生成的每个代码文件：必须在注释部分呈现“PWA新闻聚合器”字样\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 如果工作目录不存在，则返回不满足\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据 PWA 应用特点，推断可能生成的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.py', '.json'}\\n \\n # 构造正则表达式，匹配常见的各类代码注释\\n # 使用非捕获组的写法直接提取出完整的匹配文本\\n comment_pattern = re.compile(\\n r'//.*|' # JS/CSS/TS 单行注释\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|' # JS/CSS/TS 多行注释\\n r'<!--[\\\\s\\\\S]*?-->|' # HTML 注释\\n r'#.*|' # Python/Shell 单行注释\\n r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"|' # Python 多行注释\\n r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'', # Python 多行注释\\n )\\n\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n # json 文件在标准规范下不能包含注释，但如果存在可跳过注释检测或是被视为无注释\\n # 为了防止误判为失败，我们将单纯的 JSON 数据文件排除在严格代码注释检查之外\\n if ext == '.json':\\n continue\\n \\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于编码等问题无法读取，按照指令中的UTF-8强制约束属于异常状态，判为False\\n return False\\n \\n # 提取出文件内所有的注释内容\\n comments = comment_pattern.findall(content)\\n \\n # 检查这些提取出来的注释中，是否包含指定的字样\\n target_found = False\\n for comment_text in comments:\\n if 'PWA新闻聚合器' in comment_text:\\n target_found = True\\n break\\n \\n # 若当前代码文件内没有任何注释包含该字样，则不符合\\\"每个代码文件\\\"的要求\\n if not target_found:\\n return False\\n\\n # 若遍历完毕均符合条件，并且至少存在一个需要检查的代码文件，则视为满足\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n lines_to_check = []\\n \\n # 1. 提取 workspace 中的代码文件行\\n valid_extensions = {'.js', '.html', '.css', '.json', '.py', '.vue', '.ts', '.jsx', '.tsx'}\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines_to_check.extend(f.readlines())\\n except Exception:\\n pass\\n \\n # 2. 提取 response 中的代码块行\\n valid_languages = {\\n 'js', 'javascript', 'html', 'css', 'json', \\n 'py', 'python', 'vue', 'ts', 'typescript', 'jsx', 'tsx'\\n }\\n code_blocks = re.findall(r'```([a-zA-Z0-9_+-]*)\\\\n(.*?)```', response, re.DOTALL)\\n for lang, block in code_blocks:\\n lang = lang.strip().lower()\\n if lang in valid_languages or not lang:\\n # 如果没有语言标记，简单判断其是否包含常见的代码结构字符，以防将文本列表当作代码\\n if not lang and not re.search(r'[{<};=]', block):\\n continue\\n lines_to_check.extend(block.split('\\\\n'))\\n \\n total_indents = 0\\n non_four_indents = 0\\n tab_indents = 0\\n \\n # 3. 统计并分析所有行的缩进模式\\n for line in lines_to_check:\\n if not line.strip():\\n continue\\n \\n match = re.match(r'^([ \\\\t]+)(.*)', line)\\n if match:\\n indent = match.group(1)\\n content = match.group(2)\\n \\n # 忽略以 '*' 开头的多行注释内容（如 JSDoc 注释），因为它们的格式化可能会包含多余的单个空格\\n if content.startswith('*'):\\n continue\\n \\n if '\\\\t' in indent:\\n tab_indents += 1\\n continue\\n \\n indent_len = len(indent)\\n total_indents += 1\\n if indent_len % 4 != 0:\\n non_four_indents += 1\\n \\n # 只要存在制表符缩进，即违背了“仅使用空格”的要求\\n if tab_indents > 0:\\n return False\\n \\n # 如果存在缩进行，且不能被4整除的行占比超过了15%，说明AI大概率没有采用4空格缩进（例如混用了2空格）\\n if total_indents > 0:\\n if (non_four_indents / total_indents) > 0.15:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"运行`pwd`确定当前所在目录，是进行所有文件操作前必须完成的前提动作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查页面基础骨架布局：查看页面结构，顶部包含一个带有应用标题和菜单占位图标的 App Bar，中间是新闻列表区域，底部固定一个包含'新闻'和'书签'两个选项卡的 Tab Bar。截图并验证这三个区域的存在和相对位置。","检查新闻列表卡片排版：查看列表中的单条新闻卡片，左侧有图片缩略图，右侧有标题、来源、发布时间（如'2小时前'）及摘要文本；如果标题或摘要过长，结尾应显示省略号。截屏验证卡片元素构成及长文本的截断效果。","检查分类导航栏UI与选中态：查看 App Bar 下方是否有一个可横向滑动的分类导航条（包含如推荐、科技等标签）；点击其中一个分类，该分类的文本变为加粗样式，并且文字底部带有高亮的下划线指示器。截图验证选中态的样式改变。","检查分类切换的新闻列表更新及动画：在分类导航栏点击不同的分类标签，下方的新闻列表内容发生改变，并且在内容切换时伴有明显的淡入淡出动画效果。录屏或连续截图验证数据变化及动画过渡。","检查详情面板的平滑弹出：点击任意一条新闻卡片，一个全屏的详情阅读面板从屏幕底部平滑向上滑动弹出覆盖原页面，面板内包含顶部清晰大图和完整的正文段落。通过连续截图验证面板滑动弹出的方向和内部展示元素。","检查详情面板关闭功能：在已弹出的详情面板中，点击右上角的'关闭'按钮，详情面板向下收回消失，页面恢复显示之前的新闻列表。截图验证点击后是否成功退回原列表。","检查外部链接跳转的 Toast 提示：在新闻详情阅读面板中，确认最底部固定显示'在原网站阅读'按钮；点击该按钮后，页面上出现一个包含文字'即将跳转到外部链接'的 Toast 提示框。截图验证按钮固定位置、Toast 的出现与文字内容。","检查卡片收藏交互及动画：在首页找到任意新闻卡片右下角的空心星星图标，点击它，星星变成实心状态，并且在变化时伴随一个放大后再缩小的短暂弹性动画。截图或录屏验证图标状态改变和动画效果。","检查书签页的数据同步展示：点击某几条新闻的星星图标进行收藏后，点击底部 Tab Bar 的'书签'标签切换到书签页面，此时书签页面中展现了刚刚点击收藏的新闻卡片列表。截图对比首页收藏内容和书签页展示内容是否一致。","检查书签页取消收藏功能：在'书签'页面中，点击任意新闻卡片上的实心星星图标，星星图标状态取消，并且该条新闻卡片自动从当前页面列表中消失移除。截图验证点击前后的卡片列表数量变化。","检查深色模式切换功能：点击顶部 App Bar 右侧的太阳/月亮图标，整个应用的界面（包括背景、文字、卡片底色）平滑过渡为深色的暗黑主题。截图验证主题色的改变以及对比度是否合适阅读。","检查主题模式的本地持久化保存：在切换应用为深色模式后，手动刷新整个网页（F5），页面重新加载完成后，背景和界面依然保持为深色模式，而不会回退到浅色模式。截图验证刷新后的主题状态。","检查列表下拉刷新交互：在主页新闻列表的最顶部位置继续向下滑动（模拟下拉动作），顶部会显示出带有旋转动画的刷新指示器；松手后模拟网络加载约2秒，随后指示器消失，并且顶部的第一条新闻卡片内容更新为新的内容。截图验证下拉图标与数据更新情况。","检查列表无限滚动加载更多：向下滑动新闻列表直至滚动条触底，页面最底部出现包含'正在加载更多...'的文字和动画提示，几秒后提示消失，并在列表尾部自动追加了新的新闻卡片。截图验证底部加载提示及新增卡片元素。","检查搜索框的激活：点击顶部 App Bar 中应用标题旁边的放大镜图标，原先显示应用标题的位置变为一个文本输入框。截图验证搜索输入框的出现机制。","检查搜索实时过滤与关键词高亮显示：在搜索输入框中输入某个可见新闻的词语，下方新闻列表实时过滤仅展示包含该词语的卡片，并且卡片中的标题或摘要里匹配到的关键词必须带有醒目的背景色（如亮黄色）高亮。截图验证过滤结果和关键词背景高亮样式。","检查网络离线状态警示提示：通过浏览器开发者工具将网络模拟切换为'离线(Offline)'状态，页面顶部会立即向下滑出一个红色的警示条，并包含提示文案'当前处于离线模式，只能查看已缓存内容'。截图验证警示条的出现及其文本颜色。","检查离线状态下禁用下拉刷新：在处于'离线(Offline)'状态且已出现离线警示条的情况下，尝试在新闻列表顶部下拉，此时不会出现下拉刷新的旋转指示动画。验证下拉被阻断。","检查离线缓存页面访问能力：在'离线(Offline)'状态下，点击底部导航的'书签'按钮，能够正常切换并显示之前已收藏的新闻列表；进一步点击列表中的新闻卡片，依然可以正常打开并查看该新闻的详情阅读面板。截图验证离线状态下书签和详情页的可见性。","检查详情面板收藏功能：打开任意新闻详情阅读面板，在操作区找到空心星星图标；点击后星星变为实心，并伴随短暂的放大缩小弹性动画，同时该新闻被加入收藏列表。","检查书签数据的本地持久化：收藏若干新闻后刷新页面，再切换到底部'书签'页面，确认之前收藏的新闻仍然存在，说明收藏数据已通过 localStorage 保存在本地。"],"string":"[\n \"检查页面基础骨架布局：查看页面结构，顶部包含一个带有应用标题和菜单占位图标的 App Bar，中间是新闻列表区域，底部固定一个包含'新闻'和'书签'两个选项卡的 Tab Bar。截图并验证这三个区域的存在和相对位置。\",\n \"检查新闻列表卡片排版：查看列表中的单条新闻卡片，左侧有图片缩略图，右侧有标题、来源、发布时间（如'2小时前'）及摘要文本；如果标题或摘要过长，结尾应显示省略号。截屏验证卡片元素构成及长文本的截断效果。\",\n \"检查分类导航栏UI与选中态：查看 App Bar 下方是否有一个可横向滑动的分类导航条（包含如推荐、科技等标签）；点击其中一个分类，该分类的文本变为加粗样式，并且文字底部带有高亮的下划线指示器。截图验证选中态的样式改变。\",\n \"检查分类切换的新闻列表更新及动画：在分类导航栏点击不同的分类标签，下方的新闻列表内容发生改变，并且在内容切换时伴有明显的淡入淡出动画效果。录屏或连续截图验证数据变化及动画过渡。\",\n \"检查详情面板的平滑弹出：点击任意一条新闻卡片，一个全屏的详情阅读面板从屏幕底部平滑向上滑动弹出覆盖原页面，面板内包含顶部清晰大图和完整的正文段落。通过连续截图验证面板滑动弹出的方向和内部展示元素。\",\n \"检查详情面板关闭功能：在已弹出的详情面板中，点击右上角的'关闭'按钮，详情面板向下收回消失，页面恢复显示之前的新闻列表。截图验证点击后是否成功退回原列表。\",\n \"检查外部链接跳转的 Toast 提示：在新闻详情阅读面板中，确认最底部固定显示'在原网站阅读'按钮；点击该按钮后，页面上出现一个包含文字'即将跳转到外部链接'的 Toast 提示框。截图验证按钮固定位置、Toast 的出现与文字内容。\",\n \"检查卡片收藏交互及动画：在首页找到任意新闻卡片右下角的空心星星图标，点击它，星星变成实心状态，并且在变化时伴随一个放大后再缩小的短暂弹性动画。截图或录屏验证图标状态改变和动画效果。\",\n \"检查书签页的数据同步展示：点击某几条新闻的星星图标进行收藏后，点击底部 Tab Bar 的'书签'标签切换到书签页面，此时书签页面中展现了刚刚点击收藏的新闻卡片列表。截图对比首页收藏内容和书签页展示内容是否一致。\",\n \"检查书签页取消收藏功能：在'书签'页面中，点击任意新闻卡片上的实心星星图标，星星图标状态取消，并且该条新闻卡片自动从当前页面列表中消失移除。截图验证点击前后的卡片列表数量变化。\",\n \"检查深色模式切换功能：点击顶部 App Bar 右侧的太阳/月亮图标，整个应用的界面（包括背景、文字、卡片底色）平滑过渡为深色的暗黑主题。截图验证主题色的改变以及对比度是否合适阅读。\",\n \"检查主题模式的本地持久化保存：在切换应用为深色模式后，手动刷新整个网页（F5），页面重新加载完成后，背景和界面依然保持为深色模式，而不会回退到浅色模式。截图验证刷新后的主题状态。\",\n \"检查列表下拉刷新交互：在主页新闻列表的最顶部位置继续向下滑动（模拟下拉动作），顶部会显示出带有旋转动画的刷新指示器；松手后模拟网络加载约2秒，随后指示器消失，并且顶部的第一条新闻卡片内容更新为新的内容。截图验证下拉图标与数据更新情况。\",\n \"检查列表无限滚动加载更多：向下滑动新闻列表直至滚动条触底，页面最底部出现包含'正在加载更多...'的文字和动画提示，几秒后提示消失，并在列表尾部自动追加了新的新闻卡片。截图验证底部加载提示及新增卡片元素。\",\n \"检查搜索框的激活：点击顶部 App Bar 中应用标题旁边的放大镜图标，原先显示应用标题的位置变为一个文本输入框。截图验证搜索输入框的出现机制。\",\n \"检查搜索实时过滤与关键词高亮显示：在搜索输入框中输入某个可见新闻的词语，下方新闻列表实时过滤仅展示包含该词语的卡片，并且卡片中的标题或摘要里匹配到的关键词必须带有醒目的背景色（如亮黄色）高亮。截图验证过滤结果和关键词背景高亮样式。\",\n \"检查网络离线状态警示提示：通过浏览器开发者工具将网络模拟切换为'离线(Offline)'状态，页面顶部会立即向下滑出一个红色的警示条，并包含提示文案'当前处于离线模式，只能查看已缓存内容'。截图验证警示条的出现及其文本颜色。\",\n \"检查离线状态下禁用下拉刷新：在处于'离线(Offline)'状态且已出现离线警示条的情况下，尝试在新闻列表顶部下拉，此时不会出现下拉刷新的旋转指示动画。验证下拉被阻断。\",\n \"检查离线缓存页面访问能力：在'离线(Offline)'状态下，点击底部导航的'书签'按钮，能够正常切换并显示之前已收藏的新闻列表；进一步点击列表中的新闻卡片，依然可以正常打开并查看该新闻的详情阅读面板。截图验证离线状态下书签和详情页的可见性。\",\n \"检查详情面板收藏功能：打开任意新闻详情阅读面板，在操作区找到空心星星图标；点击后星星变为实心，并伴随短暂的放大缩小弹性动画，同时该新闻被加入收藏列表。\",\n \"检查书签数据的本地持久化：收藏若干新闻后刷新页面，再切换到底部'书签'页面，确认之前收藏的新闻仍然存在，说明收藏数据已通过 localStorage 保存在本地。\"\n]","isConversational":false}}},{"rowIdx":54,"cells":{"task_id":{"kind":"number","value":55,"string":"55"},"repository_policy":{"kind":"string","value":"请你在后续所有工作轮次中严格遵循以下所有约束条件：\n1. 必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改。\n2. 凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换。\n3. 为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装。\n4. 请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读。\n5. 无论生成多少个代码文件，其内部的注释内容均要求全是中文。\n6. 优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段。\n7. 在任何文件操作前，都不允许执行`pwd`来获取当前目录信息。\n8. 用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]}],"string":"[\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我实现一个基于 Web 的本地表格编辑器的雏形。首先，在页面上方固定一个工具栏区域，里面放几个占位的按钮（比如新增行列、删除、合并等，先不用实现具体功能）。工具栏下方居中展示一个 5x5 的初始静态表格。要求表格有明显的边框，鼠标点击任意单元格时，该单元格会有浅蓝色的背景高亮，表示被选中。先不用实现数据编辑功能，只要把页面骨架、表格渲染和基础的选中交互效果做出来就好。\n请注意后续所有开发操作都必须始终遵循以下两项要求：第一，务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限。第二，每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":1,"instruction":"不错，基础布局有了。接下来我们让单元格能够输入内容。请实现：双击被选中的单元格时，它会变成一个文本输入框，光标聚焦，我可以往里面打字。当我按下回车键或者鼠标点击表格外的地方使其失去焦点时，输入框消失，刚才打的文字就留在那个单元格里。注意保持之前点击高亮选中的逻辑仍然能够正常工作。另外后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在task_plan.md中写入本次修改的规划；第二，要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论何时修改代码，首要步骤都必须是在task_plan.md中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":2,"instruction":"好，现在我们需要引入 LocalStorage 让数据持久化。要求每一次单元格内容的修改完成时，都会自动将当前表格的数据存入浏览器的 LocalStorage 中。当我刷新网页或重新打开页面时，表格能直接加载上次保存的数据并展示出来。另外，请在工具栏最右侧加一个红色的「重置表格」按钮，点击后会清空 LocalStorage 中的相关数据，并把表格强制恢复成初始的 5x5 空白状态。后续所有开发操作都必须始终遵循以下新要求：第一，无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划，此前要求写入task_plan.md的规则不再生效；第二，你所生成的所有代码文件都必须使用UTF-8进行编码。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，主要涉及Web前端文件以及特定的配置/日志文件\n target_extensions = {'.html', '.css', '.js', '.json', '.md', '.log'}\n has_target_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录，如 .git\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_target_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n file_bytes = f.read()\n # 尝试用 UTF-8 解码，若失败将抛出 UnicodeDecodeError\n file_bytes.decode('utf-8')\n except UnicodeDecodeError:\n # 遇到任何一个不能被 UTF-8 解码的文件，即不满足要求\n return False\n except Exception:\n # 忽略权限等其他读取错误\n pass\n \n # 如果找到了目标文件并且都没有触发解码错误，则视为满足要求\n return has_target_files","tags":["Language","File Encoding"]}]},{"round_id":3,"instruction":"接下来增加对行列的操作，这对你底层数据结构的维护是个考验。请激活工具栏里的相关按钮：要求当我选中某个单元格后，可以点击工具栏的按钮分别执行『在上方插入行』、『在下方插入行』、『在左侧插入列』、『在右侧插入列』。同时还要实现『删除当前行』和『删除当前列』的功能。所有这些行列结构的动态改变，都必须实时同步保存到 LocalStorage 中，保证页面刷新后表格行列数量和内容都不错乱。后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在dev_plan.md中写入本次修改的规划，此前要求写入implementation_plan.json的规则不再生效；第二，请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度，该要求后续所有开发过程均需遵守。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证生成的代码文件是否都使用了UTF-8编码。\n \"\"\"\n # 根据用户指令推断，基于Web的本地表格编辑器可能包含的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的读取错误（如权限问题等）\n pass\n \n # 如果所有代码文件均能成功用 UTF-8 读取，则认为符合要求\n return True","tags":["Language","File Encoding"]},{"constraint":"无论何时修改代码，首要步骤都必须是在dev_plan.md中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度","validation_code":"def check_function_name_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 提取 JS / TS / HTML 中不同风格的函数名称定义\n js_patterns = [\n re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('), # function foo()\n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*function\\b'), # foo = function()\n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*function\\b'), # foo: function()\n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>'), # foo = () => \n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{') # foo() { (类方法/对象简写方法)\n ]\n \n # 排除类方法/简写方法中容易因为大括号与括号误判的 JavaScript 关键字\n js_keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', 'finally',\n 'return', 'with', 'function', 'super', 'this', 'typeof', 'instanceof', 'new',\n 'import', 'export', 'await', 'yield'\n }\n\n # 当路径不存在时，说明没有代码产生，默认符合要求\n if not os.path.exists(workspace_path):\n return True\n\n # Web项目可能的前端文件后缀\n valid_exts = ('.js', '.mjs', '.cjs', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue', '.svelte')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n path = os.path.join(root, file)\n \n # 使用 AST 精确解析 Python 文件\n if file.endswith('.py'):\n try:\n with open(path, 'r', encoding='utf-8') as f:\n tree = ast.parse(f.read())\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if len(node.name) < 4:\n return False\n except Exception:\n pass\n \n # 使用正则表达式解析前端文件\n elif file.endswith(valid_exts):\n try:\n with open(path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n for p in js_patterns:\n names = p.findall(content)\n for name in names:\n if name not in js_keywords:\n if len(name) < 4:\n return False\n\n return True","tags":["Style","Naming"]}]},{"round_id":4,"instruction":"现在核心挑战来了，我们来实现单元格的合并与拆分。首先，支持用户按住鼠标左键在表格上滑动拖拽，能够一次性框选出一个矩形区域的多个连续单元格（给它们统一加上高亮背景色）。然后，在工具栏增加一个『合并单元格』按钮，点击后将当前框选的区域合并成一个大单元格（内容保留左上角第一个单元格的文本）。再加一个『拆分单元格』按钮，当选中已经合并的大单元格时，点击它可以恢复成原来的多个小格子。这部分你需要正确处理 HTML 表格的 rowspan 和 colspan 属性，并妥善把这些合并状态的数据结构存入 LocalStorage。\n后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在cell_op_plan.md中写入本次修改的规划，此前要求写入dev_plan.md的规则不再生效；第二，对于你生成的代码文件，任何单个函数的行数均不得多于60行，该要求后续所有开发过程均需遵守。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断的可能的代码及相关文本文件类型\n target_extensions = {'.html', '.css', '.js', '.md', '.json', '.log'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码格式读取该文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 只要有一个文件引发解码错误，说明存在非 UTF-8 编码的文件\n return False\n except Exception:\n # 忽略由于文件权限不存在等其他非编码问题引发的异常\n continue\n \n return True","tags":["Language","File Encoding"]},{"constraint":"无论何时修改代码，首要步骤都必须是在cell_op_plan.md中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度","validation_code":"def check_function_name_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对 JavaScript/TypeScript 的常见函数定义方式构造的正则模式\n js_patterns = [\n # 模式1: 传统函数声明 function foo()\n re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b'),\n \n # 模式2: 赋值函数 foo = function() 或对象属性 foo: function()\n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*[:=]\\s*(?:async\\s*)?function\\b'),\n \n # 模式3: 箭头函数赋值 foo = () => {} 或 foo: async x => {}\n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*[:=]\\s*(?:async\\s*)?(?:\\([^()]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>'),\n \n # 模式4: 类方法或对象简写方法 foo() { ... }\n re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^()]*\\)\\s*\\{')\n ]\n \n # 针对 Python 函数定义\n py_patterns = [\n re.compile(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n ]\n \n # 在 JS 中容易被模式4误伤的常见关键字\n ignore_keywords = {\n 'if', 'for', 'while', 'catch', 'switch', 'with', 'function', 'return', \n 'constructor', 'else', 'import', 'export', 'typeof', 'instanceof', \n 'await', 'yield', 'super', 'get', 'set', 'async', 'const', 'let', 'var',\n 'true', 'false', 'null', 'undefined', 'new', 'delete', 'in', 'of', 'class',\n 'default', 'try', 'finally', 'do', 'break', 'continue', 'debugger', 'void'\n }\n \n # 遍历当前工作空间内的所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.py']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 对前端文件进行正则提取和判断\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\n for pattern in js_patterns:\n for match in pattern.findall(content):\n func_name = match.strip()\n # 跳过被误伤的语言关键字\n if func_name in ignore_keywords:\n continue\n # 如果发现函数的名称长度小于 4 个字符，则直接返回不满足要求\n if len(func_name) < 4:\n return False\n \n # 对 Python 文件进行正则提取和判断\n elif ext == '.py':\n for pattern in py_patterns:\n for match in pattern.findall(content):\n func_name = match.strip()\n if len(func_name) < 4:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何单个函数的行数均不得多于60行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_py_functions(code: str) -> bool:\n try:\n tree = ast.parse(code)\n except Exception:\n return True\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n end_line = node.end_lineno\n else:\n lines = [getattr(n, 'lineno', node.lineno) for n in ast.walk(node) if hasattr(n, 'lineno')]\n end_line = max(lines) if lines else node.lineno\n start_line = node.lineno\n \n if end_line - start_line + 1 > 60:\n return False\n return True\n\n def check_js_functions(code: str) -> bool:\n # 清除转义花括号以防干扰大括号栈匹配\n code = code.replace('\\\\{', ' ').replace('\\\\}', ' ')\n \n # 匹配字符串（双引号/单引号/反引号）和注释（多行/单行）\n pattern = r'(?P<string>\"([^\"\\\\]|\\\\.)*\"|\\'([^\\'\\\\]|\\\\.)*\\'|`([^`\\\\]|\\\\.)*`)' + \\\n r'|(?P<mcomment>/\\*[\\s\\S]*?\\*/)' + \\\n r'|(?P<scomment>//[^\\n]*)'\n\n def replacer(match):\n # 将匹配到的字符串或注释替换为相同数量的换行符，保证后续行数计算依旧准确\n return '\\n' * match.group(0).count('\\n')\n \n code = re.sub(pattern, replacer, code)\n \n left_braces = []\n pair = {}\n stack = []\n for i, c in enumerate(code):\n if c == '{':\n stack.append(i)\n left_braces.append(i)\n elif c == '}':\n if stack:\n pair[stack.pop()] = i\n \n for i in left_braces:\n prefix = code[:i].rstrip()\n is_func = False\n start_pos = i\n \n # 校验大括号前的内容特征是否符合 JS 的函数定义规则\n if prefix.endswith('=>'):\n is_func = True\n start_pos = prefix.rfind('=>')\n elif prefix.endswith(')'):\n paren_count = 0\n idx = len(prefix) - 1\n while idx >= 0:\n if prefix[idx] == ')':\n paren_count += 1\n elif prefix[idx] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n idx -= 1\n if paren_count == 0 and idx >= 0:\n before_paren = prefix[:idx].rstrip()\n match = re.search(r'([A-Za-z0-9_$]+)$', before_paren)\n if match:\n word = match.group(1)\n # 排除常规控制流和普通代码块\n if word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n is_func = True\n start_pos = len(before_paren) - len(match.group(1))\n before_word = before_paren[:start_pos].rstrip()\n if before_word.endswith('function'):\n start_pos = before_word.rfind('function')\n elif before_word.endswith('get') or before_word.endswith('set'):\n start_pos = before_word.rfind('get') if before_word.endswith('get') else before_word.rfind('set')\n else:\n if before_paren.endswith('function'):\n is_func = True\n start_pos = before_paren.rfind('function')\n else:\n is_func = True\n start_pos = idx\n \n if is_func:\n end_idx = pair.get(i, -1)\n if end_idx != -1:\n # 按照大括号匹配区域计算函数的起始行和终止行\n start_line = code[:start_pos].count('\\n') + 1\n end_line = code[:end_idx+1].count('\\n') + 1\n if end_line - start_line + 1 > 60:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(('.js', '.html', '.py', '.ts', '.jsx', '.tsx', '.vue', '.htm')):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if file.endswith('.py'):\n if not check_py_functions(content):\n return False\n else:\n if file.endswith(('.html', '.htm', '.vue')):\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n js_content = \"\\n\".join(scripts)\n else:\n js_content = content\n \n if not check_js_functions(js_content):\n return False\n \n return True","tags":["Quantity","Range"]}]},{"round_id":5,"instruction":"做得非常好！现在为了提升编辑器的容错率，我们加入撤销（Undo）和重做（Redo）功能。在工具栏左侧增加这两个按钮。不管是修改文字、增删行列还是复杂的合并拆分单元格，每一次会改变表格状态的操作都要记录进历史记录栈。点击『撤销』可以完整回退到上一步的状态，点击『重做』可以恢复撤销的操作。同时要求支持使用快捷键 `Ctrl+Z` 和 `Ctrl+Y`（Mac上是 `Cmd+Z` / `Cmd+Shift+Z`）来快速触发。为了性能考虑，历史栈长度限制最多保存 50 步即可。\n后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在undo_redo_plan.md中写入本次修改的规划，此前要求写入cell_op_plan.md的规则不再生效；第二，确保在你生成的所有代码文件中，等号（=）的左右两侧都留有两个空格，该要求后续所有开发过程均需遵守。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 验证工作区目录是否存在\n if not os.path.isdir(workspace_path):\n return False\n\n # 根据项目推断常见的代码和配置文件的扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.json', '.md', '.py', '.java', '.c', '.cpp', '.h'\n }\n\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git 等版本控制和配置目录）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n # 尝试用 utf-8 编码读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果解码发生异常，说明不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的系统级读取错误（如权限问题等）\n pass\n\n # 如果有代码文件被扫描到并且都没有引发解码异常，则满足要求\n return has_code_files","tags":["Language","File Encoding"]},{"constraint":"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_js_functions(js_code: str) -> bool:\n # 去除注释内容，防止注释内的示例代码影响判断\n js_code = re.sub(r'//.*', '', js_code)\n js_code = re.sub(r'/\\*.*?\\*/', '', js_code, flags=re.DOTALL)\n \n # 去除字符串内容，防止字符串中包含类似函数的结构误导判断\n js_code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', js_code)\n js_code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", js_code)\n js_code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', js_code, flags=re.DOTALL)\n \n # 正则匹配不同类型的 JS/TS 函数定义\n # 1. 声明式函数: function foo(...)\n p1 = re.compile(r'\\bfunction\\s+([a-zA-Z_$][\\w$]*)\\s*\\(')\n # 2. 赋值型函数: foo = function(...) 或 foo = (...) =>\n p2 = re.compile(r'\\b([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n # 3. 对象属性函数: foo: function(...) 或 foo: (...) =>\n p3 = re.compile(r'\\b([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n # 4. 类方法/对象简写方法: foo(...) {\n p4 = re.compile(r'(?:^|[\\{\\s;])(?:async\\s+)?(?:get\\s+|set\\s+)?(?:static\\s+)?([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{')\n \n funcs = []\n funcs.extend(p1.findall(js_code))\n funcs.extend(p2.findall(js_code))\n funcs.extend(p3.findall(js_code))\n funcs.extend(p4.findall(js_code))\n \n # 排除JS关键字和常见控制流语句等可能被作为类方法正则误捕获的内容\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', \n 'else', 'function', 'import', 'export', 'require', 'super', 'get', 'set',\n 'async', 'await', 'with', 'typeof', 'instanceof', 'in', 'of', 'new',\n 'class', 'extends', 'yield', 'let', 'const', 'var', 'default', 'finally',\n 'break', 'continue', 'debugger', 'do', 'throw', 'try'\n }\n \n for f_name in funcs:\n if f_name in keywords:\n continue\n # 找到任何长度小于4的函数名则直接判定不符合要求\n if len(f_name) < 4:\n return False\n \n return True\n\n def check_py_functions(py_code: str) -> bool:\n try:\n tree = ast.parse(py_code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if len(node.name) < 4:\n return False\n except Exception:\n pass\n return True\n\n # 遍历项目文件\n for root, dirs, files in os.walk(workspace_path):\n # 忽略包依赖和版本控制目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n filepath = os.path.join(root, file)\n \n # Python 文件验证\n if ext == '.py':\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_py_functions(content):\n return False\n except Exception:\n pass\n \n # HTML 或 JS/TS 文件验证\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.htm']:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if ext in ['.html', '.htm']:\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for js_code in scripts:\n if not check_js_functions(js_code):\n return False\n else:\n if not check_js_functions(content):\n return False\n except Exception:\n pass\n \n # 没有发现任何不符合规定的情况\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何单个函数的行数均不得多于60行","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n def clean_js(code: str) -> str:\n \"\"\"移除JS代码中的注释和字符串内容，保留换行和基础结构，不干扰行号计算\"\"\"\n i = 0\n n = len(code)\n cleaned = []\n in_single_comment = False\n in_multi_comment = False\n in_string = False\n string_char = ''\n \n while i < n:\n char = code[i]\n if in_single_comment:\n if char == '\\n':\n in_single_comment = False\n cleaned.append(char)\n else:\n cleaned.append(' ')\n i += 1\n continue\n \n if in_multi_comment:\n if char == '*' and i + 1 < n and code[i+1] == '/':\n in_multi_comment = False\n cleaned.append(' ')\n i += 2\n else:\n cleaned.append('\\n' if char == '\\n' else ' ')\n i += 1\n continue\n \n if in_string:\n if char == '\\\\':\n cleaned.append(' ')\n i += 2\n continue\n if char == string_char:\n in_string = False\n cleaned.append(' ')\n else:\n cleaned.append('\\n' if char == '\\n' else ' ')\n i += 1\n continue\n \n if char == '/' and i + 1 < n and code[i+1] == '/':\n in_single_comment = True\n cleaned.append(' ')\n i += 2\n continue\n \n if char == '/' and i + 1 < n and code[i+1] == '*':\n in_multi_comment = True\n cleaned.append(' ')\n i += 2\n continue\n \n if char in (\"'\", '\"', \"`\"):\n in_string = True\n string_char = char\n cleaned.append(' ')\n i += 1\n continue\n \n cleaned.append(char)\n i += 1\n \n return \"\".join(cleaned)\n\n def extract_js_from_html(html_code: str) -> str:\n \"\"\"抽取HTML/Vue中<script>标签内容，通过将非JS部分替换为空白的方式保持行数对齐\"\"\"\n js_code = \"\"\n i = 0\n n = len(html_code)\n in_script = False\n \n while i < n:\n if not in_script:\n script_start = html_code.find('<script', i)\n if script_start == -1:\n js_code += re.sub(r'[^\\n]', ' ', html_code[i:])\n break\n else:\n tag_end = html_code.find('>', script_start)\n if tag_end == -1:\n js_code += re.sub(r'[^\\n]', ' ', html_code[i:])\n break\n js_code += re.sub(r'[^\\n]', ' ', html_code[i:tag_end+1])\n i = tag_end + 1\n in_script = True\n else:\n script_end = html_code.find('</script>', i)\n if script_end == -1:\n js_code += html_code[i:]\n break\n else:\n js_code += html_code[i:script_end]\n js_code += re.sub(r'[^\\n]', ' ', html_code[script_end:script_end+9])\n i = script_end + 9\n in_script = False\n return js_code\n\n def get_prefix_for_brace(code_str: str, brace_index: int) -> str:\n \"\"\"从某一个 '{' 向前搜索，获取属于该代码块的声明语句，处理了嵌套圆括号折叠的规避\"\"\"\n i = brace_index - 1\n parens = 0\n braces = 0\n brackets = 0\n collected_tokens = []\n \n while i >= 0:\n char = code_str[i]\n if char == ')':\n parens += 1\n collected_tokens.append(char)\n elif char == '(':\n if parens > 0:\n parens -= 1\n collected_tokens.append(char)\n elif char == '}':\n braces += 1\n collected_tokens.append(char)\n elif char == '{':\n if braces > 0:\n braces -= 1\n collected_tokens.append(char)\n else:\n if parens == 0 and brackets == 0:\n break\n else:\n collected_tokens.append(char)\n elif char == ']':\n brackets += 1\n collected_tokens.append(char)\n elif char == '[':\n if brackets > 0:\n brackets -= 1\n collected_tokens.append(char)\n elif char == ';':\n if parens == 0 and braces == 0 and brackets == 0:\n break\n else:\n collected_tokens.append(char)\n else:\n collected_tokens.append(char)\n i -= 1\n \n collected_tokens.reverse()\n return \"\".join(collected_tokens)\n\n def is_function_prefix(prefix: str) -> bool:\n \"\"\"根据大括号前缀来确定该块是不是一个函数（常规函数、箭头函数、类方法等）\"\"\"\n prefix = prefix.strip()\n if not prefix:\n return False\n if prefix.endswith('=>'):\n return True\n \n keywords_exclude = {'if', 'for', 'while', 'switch', 'catch', 'with', 'return'}\n \n if prefix.endswith(')'):\n parens = 0\n idx = len(prefix) - 1\n while idx >= 0:\n if prefix[idx] == ')':\n parens += 1\n elif prefix[idx] == '(':\n parens -= 1\n if parens == 0:\n break\n idx -= 1\n \n if idx >= 0:\n before_paren = prefix[:idx]\n match = re.search(r'\\b(\\w+)\\s*$', before_paren)\n if match:\n word = match.group(1)\n if word not in keywords_exclude:\n return True\n elif prefix.endswith('function'):\n return True\n \n return False\n\n def check_functions_length_js(code: str) -> bool:\n cleaned = clean_js(code)\n stack = []\n line_number = 1\n \n for i, char in enumerate(cleaned):\n if char == '\\n':\n line_number += 1\n elif char == '{':\n prefix = get_prefix_for_brace(cleaned, i)\n is_func = is_function_prefix(prefix)\n stack.append((line_number, is_func))\n elif char == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n if line_number - start_line + 1 > 60:\n return False\n return True\n\n def check_functions_length_py(code: str) -> bool:\n try:\n tree = ast.parse(code)\n except Exception:\n return True\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_line = node.lineno\n end_line = getattr(node, 'end_lineno', start_line)\n if end_line - start_line + 1 > 60:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 过滤隐藏文件及文件夹\n if '/.' in file_path or '\\\\.' in file_path:\n continue\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext in ['.js', '.jsx', '.ts', '.tsx']:\n if not check_functions_length_js(content):\n return False\n elif ext in ['.html', '.htm', '.vue']:\n js_content = extract_js_from_html(content)\n if not check_functions_length_js(js_content):\n return False\n elif ext == '.py':\n if not check_functions_length_py(content):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"无论何时修改代码，首要步骤都必须是在undo_redo_plan.md中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"确保在你生成的所有代码文件中，等号（=）的左右两侧都留有两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对前端及常见代码文件的扩展名进行限制，过滤掉 .md、.txt 等非代码类型的文档\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 仅移除行末的换行符，保留行末可能存在的空格以便精确验证\n line = line.rstrip('\\n\\r')\n i = 0\n while i < len(line):\n if line[i] == '=':\n # 判定该等号是否属于复合操作符的一部分（如 ==, >=, !=, +=, => 等）\n is_compound = False\n if i > 0 and line[i-1] in '=<>!+-*/%&|^':\n is_compound = True\n if i < len(line) - 1 and line[i+1] in '=>':\n is_compound = True\n \n if not is_compound:\n # 检查左侧是否有且仅有两个空格\n left_ok = (i >= 2 and line[i-1] == ' ' and line[i-2] == ' ' and (i == 2 or line[i-3] != ' '))\n \n # 检查右侧是否有且仅有两个空格\n right_ok = (i <= len(line) - 3 and line[i+1] == ' ' and line[i+2] == ' ' and (i == len(line) - 3 or line[i+3] != ' '))\n \n # 如果有任意一侧不满足留有两个空格的条件，则判定为失败\n if not (left_ok and right_ok):\n return False\n i += 1\n \n return True","tags":["Style","Layout"]}]},{"round_id":6,"instruction":"最后一步，我们给表格增加一些基础的富文本样式定制能力。在工具栏补充一个样式编辑面板：包括文本的『加粗』、『斜体』开关按钮，以及『文字颜色』和『单元格背景色』的原生取色器（color picker）。当我选中单个单元格或者框选多个单元格后，操作这些样式工具，选中的单元格样式会立刻发生改变。理所当然地，这些关于文字和背景的样式元数据也必须深度整合进你的数据结构中，存入 LocalStorage，并在撤销/重做以及页面刷新时完美恢复。\n后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在style_edit_plan.md中写入本次修改的规划，此前要求写入undo_redo_plan.md的规则不再生效；第二，你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径，该要求后续所有开发过程均需遵守。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限","validation_code":"","tags":["Environment","Logging"]},{"constraint":"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断，前端Web编辑器项目主要涉及的代码及相关文本文件类型\n target_extensions = {'.html', '.js', '.css', '.md', '.json', '.log'}\n \n if not os.path.exists(workspace_path):\n # 如果项目路径不存在，则无法判断或视为未违背此规则\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 检查是否为相关的代码或文本文件\n if any(file.lower().endswith(ext) for ext in target_extensions):\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码进行读取\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明该文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如权限不足等其他与编码无关的读取异常\n continue\n \n # 所有文件都可以成功按照utf-8解析，满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度","validation_code":"import os\nimport re\n\ndef check_function_name_length(response: str, workspace_path: str) -> bool:\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 定义正则表达式匹配各类语言中的函数定义\n \n # 1. Python 函数定义\n py_func_pattern = re.compile(r'def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n py_lambda_pattern = re.compile(r'([a-zA-Z_][a-zA-Z0-9_]*)\\s*=\\s*lambda\\b')\n \n # 2. JS/TS 关键字 function 声明 (支持 function foo( 以及 function * foo( 等)\n js_func_pattern = re.compile(r'function(?:\\s+|\\s*\\*\\s*)([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\(')\n \n # 3. JS/TS function 赋值给变量或属性，如 foo = function() 或 foo: function()\n js_assign_func_pattern = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*(?:=|:)\\s*function\\b')\n \n # 4. JS/TS 箭头函数赋值给变量或属性，如 foo = () => 或 foo: a =>\n js_arrow_func_pattern = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*(?:=|:)\\s*(?:\\([^)]*\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=>')\n \n # 5. JS/TS 对象或类方法定义，形如 methodName(params) {\n # 确保前导不是小数点、单词字符或 $ 符号等，以免匹配到对象属性点调用\n js_method_pattern = re.compile(r'(?:^|[^\\w.$])([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\(([^()]*)\\)\\s*\\{')\n \n # HTML 中 <script> 标签的提取正则，避免直接在 HTML 层面正则引起误判（如 CSS 伪类）\n script_pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\n \n # 排除容易被类方法正则误判的控制流关键字、原生类型或 CSS 函数\n exclude_keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'with', \n 'function', 'return', 'constructor', 'typeof',\n 'try', 'finally', 'else', 'elif', 'except',\n 'is', 'has', 'not', 'var', 'url', 'let', 'get', 'set',\n 'new'\n }\n \n # 指定需要检查的代码文件扩展名\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.py'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 文件不可读时跳过\n continue\n \n # HTML 文件只检查 <script> 标签中的内容\n if ext in {'.html', '.htm', '.vue'}:\n scripts = script_pattern.findall(content)\n content_to_check = '\\n'.join(scripts)\n else:\n content_to_check = content\n \n func_names = []\n \n if ext == '.py':\n func_names.extend(py_func_pattern.findall(content_to_check))\n func_names.extend(py_lambda_pattern.findall(content_to_check))\n else:\n func_names.extend(js_func_pattern.findall(content_to_check))\n func_names.extend(js_assign_func_pattern.findall(content_to_check))\n func_names.extend(js_arrow_func_pattern.findall(content_to_check))\n \n # 对象或类的方法需特殊过滤\n methods = js_method_pattern.findall(content_to_check)\n for name, params in methods:\n # 如果匹配到的方法参数中包含 '=>'，那通常是作为参数传递进去的匿名箭头函数被误匹配，剔除\n if '=>' in params:\n continue\n func_names.append(name)\n \n # 过滤并检查所有识别到的函数名称长度\n for name in func_names:\n if name in exclude_keywords:\n continue\n # 一旦找到名称长度小于 4 个字符的函数，即不满足要求\n if len(name) < 4:\n return False\n \n # 没有发现任何违背规则的情况，满足要求\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何单个函数的行数均不得多于60行","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \n def extract_js_from_html(html_code: str) -> str:\n # 使用正则表达式查找 script 标签的内容\n pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\n result = list(html_code)\n # 默认全部替换为空格，遇到换行保留换行符（不改变行号和行结构）\n for i in range(len(result)):\n if result[i] != '\\n':\n result[i] = ' '\n # 只将被匹配的 script 内部代码恢复出来\n for match in pattern.finditer(html_code):\n start = match.start(1)\n end = match.end(1)\n result[start:end] = list(html_code[start:end])\n return \"\".join(result)\n\n def strip_comments_and_strings(code: str) -> str:\n # 将 JS 中的字符串和注释清理为空格，防止内部的括号干扰匹配\n result = list(code)\n i = 0\n n = len(code)\n \n def is_regex_start(idx):\n j = idx - 1\n while j >= 0 and code[j].isspace():\n j -= 1\n if j < 0: return True\n c = code[j]\n # 如果正则表达式前面是这些符号，大概率是合法的正则开始\n if c in '=({[<>,:;!&|?+-*/%~^':\n return True\n if c.isalpha() or c in '_$':\n k = j\n while k >= 0 and (code[k].isalnum() or code[k] in '_$'):\n k -= 1\n word = code[k+1:j+1]\n # 特殊的关键字后面可能跟着正则\n if word in ['return', 'typeof', 'yield', 'await', 'case', 'throw', 'delete', 'in', 'instanceof', 'new', 'do', 'else']:\n return True\n return False\n\n while i < n:\n c = code[i]\n if c == '/' and i + 1 < n:\n nc = code[i+1]\n if nc == '/': # 单行注释\n while i < n and code[i] != '\\n':\n result[i] = ' '\n i += 1\n continue\n elif nc == '*': # 多行注释\n result[i] = ' '\n result[i+1] = ' '\n i += 2\n while i < n and not (code[i] == '*' and i + 1 < n and code[i+1] == '/'):\n if result[i] != '\\n':\n result[i] = ' '\n i += 1\n if i < n:\n result[i] = ' '\n if i + 1 < n:\n result[i+1] = ' '\n i += 2\n continue\n elif is_regex_start(i): # 简单的正则字面量探测\n result[i] = ' '\n i += 1\n while i < n:\n if code[i] == '\\\\':\n result[i] = ' '\n if i + 1 < n:\n if result[i+1] != '\\n':\n result[i+1] = ' '\n i += 1\n elif code[i] == '/':\n result[i] = ' '\n i += 1\n break\n elif code[i] == '\\n':\n break\n else:\n result[i] = ' '\n i += 1\n continue\n \n elif c in ['\\'', '\"', '`']: # 字符串探测\n quote = c\n result[i] = ' '\n i += 1\n while i < n:\n if code[i] == '\\\\':\n result[i] = ' '\n if i + 1 < n:\n if result[i+1] != '\\n':\n result[i+1] = ' '\n i += 1\n elif code[i] == quote:\n result[i] = ' '\n i += 1\n break\n else:\n if result[i] != '\\n':\n result[i] = ' '\n i += 1\n continue\n i += 1\n return \"\".join(result)\n\n def find_function_blocks(clean_code: str):\n # 通过寻找成对的大括号，并根据大括号前的关键词智能判定是否是函数\n blocks = []\n stack_curly = []\n stack_paren = []\n \n pairs_curly = {}\n pairs_paren = {}\n \n for idx, char in enumerate(clean_code):\n if char == '{':\n stack_curly.append(idx)\n elif char == '}':\n if stack_curly:\n start = stack_curly.pop()\n pairs_curly[start] = idx\n elif char == '(':\n stack_paren.append(idx)\n elif char == ')':\n if stack_paren:\n start = stack_paren.pop()\n pairs_paren[idx] = start\n\n for start_idx, end_idx in pairs_curly.items():\n j = start_idx - 1\n while j >= 0 and clean_code[j].isspace():\n j -= 1\n if j < 0:\n continue\n \n is_func = False\n prev_char = clean_code[j]\n \n # 判断箭头函数\n if prev_char == '>':\n if j - 1 >= 0 and clean_code[j-1] == '=':\n is_func = True\n # 判断常规方法或者函数\n elif prev_char == ')':\n if j in pairs_paren:\n paren_start = pairs_paren[j]\n k = paren_start - 1\n while k >= 0 and clean_code[k].isspace():\n k -= 1\n \n if k >= 0:\n if clean_code[k].isalnum() or clean_code[k] in '_$':\n m = k\n while m >= 0 and (clean_code[m].isalnum() or clean_code[m] in '_$'):\n m -= 1\n word = clean_code[m+1:k+1]\n control_keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\n if word not in control_keywords:\n is_func = True\n elif clean_code[k] == ']': # 可能为动态属性名称的方法，如 `[methodName]() {`\n is_func = True\n \n if is_func:\n start_line = clean_code[:start_idx].count('\\n') + 1\n end_line = clean_code[:end_idx].count('\\n') + 1\n blocks.append((start_line, end_line))\n \n return blocks\n\n valid_extensions = ('.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n code = f.read()\n except Exception:\n continue\n \n if file.endswith(('.html', '.htm', '.vue')):\n code = extract_js_from_html(code)\n \n clean_code = strip_comments_and_strings(code)\n funcs = find_function_blocks(clean_code)\n \n # 检查每一个提取出来的函数行数\n for start, end in funcs:\n func_lines = end - start + 1\n if func_lines > 60:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"确保在你生成的所有代码文件中，等号（=）的左右两侧都留有两个空格","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的前端项目代码文件扩展名\n code_extensions = {'.js', '.ts', '.html', '.css', '.vue', '.jsx', '.tsx'}\n has_code_file = False\n \n # 正则：匹配独立的 '='，通过否定环视排除前后的复合运算符部分（如 ==, ===, >=, <=, !=, +=, => 等）\n pattern = re.compile(r'(?<![=!<>+\\-*/%^&|~?])=(?![=>])')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_code_file = True\n \n for line in lines:\n # 去除行尾换行符，以免影响右侧空格数量的判断\n line = line.rstrip('\\n').rstrip('\\r')\n \n for match in pattern.finditer(line):\n idx = match.start()\n \n left_part = line[:idx]\n right_part = line[idx+1:]\n \n # 统计左侧末尾的连续空格数量\n left_spaces = len(left_part) - len(left_part.rstrip(' '))\n # 统计右侧开头的连续空格数量\n right_spaces = len(right_part) - len(right_part.lstrip(' '))\n \n # 如果有一侧连续空格不严格为2，则不满足要求\n if left_spaces != 2 or right_spaces != 2:\n return False\n \n # 只要检查到了至少一个代码文件且都通过了校验，就返回True\n return has_code_file","tags":["Style","Layout"]},{"constraint":"无论何时修改代码，首要步骤都必须是在style_edit_plan.md中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我实现一个基于 Web 的本地表格编辑器的雏形。首先，在页面上方固定一个工具栏区域，里面放几个占位的按钮（比如新增行列、删除、合并等，先不用实现具体功能）。工具栏下方居中展示一个 5x5 的初始静态表格。要求表格有明显的边框，鼠标点击任意单元格时，该单元格会有浅蓝色的背景高亮，表示被选中。先不用实现数据编辑功能，只要把页面骨架、表格渲染和基础的选中交互效果做出来就好。\\n请注意后续所有开发操作都必须始终遵循以下两项要求：第一，务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限。第二，每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，基础布局有了。接下来我们让单元格能够输入内容。请实现：双击被选中的单元格时，它会变成一个文本输入框，光标聚焦，我可以往里面打字。当我按下回车键或者鼠标点击表格外的地方使其失去焦点时，输入框消失，刚才打的文字就留在那个单元格里。注意保持之前点击高亮选中的逻辑仍然能够正常工作。另外后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在task_plan.md中写入本次修改的规划；第二，要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在task_plan.md中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，现在我们需要引入 LocalStorage 让数据持久化。要求每一次单元格内容的修改完成时，都会自动将当前表格的数据存入浏览器的 LocalStorage 中。当我刷新网页或重新打开页面时，表格能直接加载上次保存的数据并展示出来。另外，请在工具栏最右侧加一个红色的「重置表格」按钮，点击后会清空 LocalStorage 中的相关数据，并把表格强制恢复成初始的 5x5 空白状态。后续所有开发操作都必须始终遵循以下新要求：第一，无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划，此前要求写入task_plan.md的规则不再生效；第二，你所生成的所有代码文件都必须使用UTF-8进行编码。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，主要涉及Web前端文件以及特定的配置/日志文件\\n target_extensions = {'.html', '.css', '.js', '.json', '.md', '.log'}\\n has_target_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录，如 .git\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_target_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n file_bytes = f.read()\\n # 尝试用 UTF-8 解码，若失败将抛出 UnicodeDecodeError\\n file_bytes.decode('utf-8')\\n except UnicodeDecodeError:\\n # 遇到任何一个不能被 UTF-8 解码的文件，即不满足要求\\n return False\\n except Exception:\\n # 忽略权限等其他读取错误\\n pass\\n \\n # 如果找到了目标文件并且都没有触发解码错误，则视为满足要求\\n return has_target_files\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"接下来增加对行列的操作，这对你底层数据结构的维护是个考验。请激活工具栏里的相关按钮：要求当我选中某个单元格后，可以点击工具栏的按钮分别执行『在上方插入行』、『在下方插入行』、『在左侧插入列』、『在右侧插入列』。同时还要实现『删除当前行』和『删除当前列』的功能。所有这些行列结构的动态改变，都必须实时同步保存到 LocalStorage 中，保证页面刷新后表格行列数量和内容都不错乱。后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在dev_plan.md中写入本次修改的规划，此前要求写入implementation_plan.json的规则不再生效；第二，请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度，该要求后续所有开发过程均需遵守。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证生成的代码文件是否都使用了UTF-8编码。\\n \\\"\\\"\\\"\\n # 根据用户指令推断，基于Web的本地表格编辑器可能包含的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的读取错误（如权限问题等）\\n pass\\n \\n # 如果所有代码文件均能成功用 UTF-8 读取，则认为符合要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在dev_plan.md中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度\",\n \"validation_code\": \"def check_function_name_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 提取 JS / TS / HTML 中不同风格的函数名称定义\\n js_patterns = [\\n re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('), # function foo()\\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*function\\\\b'), # foo = function()\\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*function\\\\b'), # foo: function()\\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>'), # foo = () => \\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{') # foo() { (类方法/对象简写方法)\\n ]\\n \\n # 排除类方法/简写方法中容易因为大括号与括号误判的 JavaScript 关键字\\n js_keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', 'finally',\\n 'return', 'with', 'function', 'super', 'this', 'typeof', 'instanceof', 'new',\\n 'import', 'export', 'await', 'yield'\\n }\\n\\n # 当路径不存在时，说明没有代码产生，默认符合要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # Web项目可能的前端文件后缀\\n valid_exts = ('.js', '.mjs', '.cjs', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue', '.svelte')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n path = os.path.join(root, file)\\n \\n # 使用 AST 精确解析 Python 文件\\n if file.endswith('.py'):\\n try:\\n with open(path, 'r', encoding='utf-8') as f:\\n tree = ast.parse(f.read())\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if len(node.name) < 4:\\n return False\\n except Exception:\\n pass\\n \\n # 使用正则表达式解析前端文件\\n elif file.endswith(valid_exts):\\n try:\\n with open(path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n for p in js_patterns:\\n names = p.findall(content)\\n for name in names:\\n if name not in js_keywords:\\n if len(name) < 4:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在核心挑战来了，我们来实现单元格的合并与拆分。首先，支持用户按住鼠标左键在表格上滑动拖拽，能够一次性框选出一个矩形区域的多个连续单元格（给它们统一加上高亮背景色）。然后，在工具栏增加一个『合并单元格』按钮，点击后将当前框选的区域合并成一个大单元格（内容保留左上角第一个单元格的文本）。再加一个『拆分单元格』按钮，当选中已经合并的大单元格时，点击它可以恢复成原来的多个小格子。这部分你需要正确处理 HTML 表格的 rowspan 和 colspan 属性，并妥善把这些合并状态的数据结构存入 LocalStorage。\\n后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在cell_op_plan.md中写入本次修改的规划，此前要求写入dev_plan.md的规则不再生效；第二，对于你生成的代码文件，任何单个函数的行数均不得多于60行，该要求后续所有开发过程均需遵守。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断的可能的代码及相关文本文件类型\\n target_extensions = {'.html', '.css', '.js', '.md', '.json', '.log'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码格式读取该文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 只要有一个文件引发解码错误，说明存在非 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 忽略由于文件权限不存在等其他非编码问题引发的异常\\n continue\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在cell_op_plan.md中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度\",\n \"validation_code\": \"def check_function_name_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对 JavaScript/TypeScript 的常见函数定义方式构造的正则模式\\n js_patterns = [\\n # 模式1: 传统函数声明 function foo()\\n re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b'),\\n \\n # 模式2: 赋值函数 foo = function() 或对象属性 foo: function()\\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*[:=]\\\\s*(?:async\\\\s*)?function\\\\b'),\\n \\n # 模式3: 箭头函数赋值 foo = () => {} 或 foo: async x => {}\\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*[:=]\\\\s*(?:async\\\\s*)?(?:\\\\([^()]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>'),\\n \\n # 模式4: 类方法或对象简写方法 foo() { ... }\\n re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^()]*\\\\)\\\\s*\\\\{')\\n ]\\n \\n # 针对 Python 函数定义\\n py_patterns = [\\n re.compile(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n ]\\n \\n # 在 JS 中容易被模式4误伤的常见关键字\\n ignore_keywords = {\\n 'if', 'for', 'while', 'catch', 'switch', 'with', 'function', 'return', \\n 'constructor', 'else', 'import', 'export', 'typeof', 'instanceof', \\n 'await', 'yield', 'super', 'get', 'set', 'async', 'const', 'let', 'var',\\n 'true', 'false', 'null', 'undefined', 'new', 'delete', 'in', 'of', 'class',\\n 'default', 'try', 'finally', 'do', 'break', 'continue', 'debugger', 'void'\\n }\\n \\n # 遍历当前工作空间内的所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.py']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 对前端文件进行正则提取和判断\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\\n for pattern in js_patterns:\\n for match in pattern.findall(content):\\n func_name = match.strip()\\n # 跳过被误伤的语言关键字\\n if func_name in ignore_keywords:\\n continue\\n # 如果发现函数的名称长度小于 4 个字符，则直接返回不满足要求\\n if len(func_name) < 4:\\n return False\\n \\n # 对 Python 文件进行正则提取和判断\\n elif ext == '.py':\\n for pattern in py_patterns:\\n for match in pattern.findall(content):\\n func_name = match.strip()\\n if len(func_name) < 4:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何单个函数的行数均不得多于60行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_py_functions(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return True\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n end_line = node.end_lineno\\n else:\\n lines = [getattr(n, 'lineno', node.lineno) for n in ast.walk(node) if hasattr(n, 'lineno')]\\n end_line = max(lines) if lines else node.lineno\\n start_line = node.lineno\\n \\n if end_line - start_line + 1 > 60:\\n return False\\n return True\\n\\n def check_js_functions(code: str) -> bool:\\n # 清除转义花括号以防干扰大括号栈匹配\\n code = code.replace('\\\\\\\\{', ' ').replace('\\\\\\\\}', ' ')\\n \\n # 匹配字符串（双引号/单引号/反引号）和注释（多行/单行）\\n pattern = r'(?P<string>\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'([^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`([^`\\\\\\\\]|\\\\\\\\.)*`)' + \\\\\\n r'|(?P<mcomment>/\\\\*[\\\\s\\\\S]*?\\\\*/)' + \\\\\\n r'|(?P<scomment>//[^\\\\n]*)'\\n\\n def replacer(match):\\n # 将匹配到的字符串或注释替换为相同数量的换行符，保证后续行数计算依旧准确\\n return '\\\\n' * match.group(0).count('\\\\n')\\n \\n code = re.sub(pattern, replacer, code)\\n \\n left_braces = []\\n pair = {}\\n stack = []\\n for i, c in enumerate(code):\\n if c == '{':\\n stack.append(i)\\n left_braces.append(i)\\n elif c == '}':\\n if stack:\\n pair[stack.pop()] = i\\n \\n for i in left_braces:\\n prefix = code[:i].rstrip()\\n is_func = False\\n start_pos = i\\n \\n # 校验大括号前的内容特征是否符合 JS 的函数定义规则\\n if prefix.endswith('=>'):\\n is_func = True\\n start_pos = prefix.rfind('=>')\\n elif prefix.endswith(')'):\\n paren_count = 0\\n idx = len(prefix) - 1\\n while idx >= 0:\\n if prefix[idx] == ')':\\n paren_count += 1\\n elif prefix[idx] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n idx -= 1\\n if paren_count == 0 and idx >= 0:\\n before_paren = prefix[:idx].rstrip()\\n match = re.search(r'([A-Za-z0-9_$]+)$', before_paren)\\n if match:\\n word = match.group(1)\\n # 排除常规控制流和普通代码块\\n if word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n is_func = True\\n start_pos = len(before_paren) - len(match.group(1))\\n before_word = before_paren[:start_pos].rstrip()\\n if before_word.endswith('function'):\\n start_pos = before_word.rfind('function')\\n elif before_word.endswith('get') or before_word.endswith('set'):\\n start_pos = before_word.rfind('get') if before_word.endswith('get') else before_word.rfind('set')\\n else:\\n if before_paren.endswith('function'):\\n is_func = True\\n start_pos = before_paren.rfind('function')\\n else:\\n is_func = True\\n start_pos = idx\\n \\n if is_func:\\n end_idx = pair.get(i, -1)\\n if end_idx != -1:\\n # 按照大括号匹配区域计算函数的起始行和终止行\\n start_line = code[:start_pos].count('\\\\n') + 1\\n end_line = code[:end_idx+1].count('\\\\n') + 1\\n if end_line - start_line + 1 > 60:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(('.js', '.html', '.py', '.ts', '.jsx', '.tsx', '.vue', '.htm')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if file.endswith('.py'):\\n if not check_py_functions(content):\\n return False\\n else:\\n if file.endswith(('.html', '.htm', '.vue')):\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n js_content = \\\"\\\\n\\\".join(scripts)\\n else:\\n js_content = content\\n \\n if not check_js_functions(js_content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"做得非常好！现在为了提升编辑器的容错率，我们加入撤销（Undo）和重做（Redo）功能。在工具栏左侧增加这两个按钮。不管是修改文字、增删行列还是复杂的合并拆分单元格，每一次会改变表格状态的操作都要记录进历史记录栈。点击『撤销』可以完整回退到上一步的状态，点击『重做』可以恢复撤销的操作。同时要求支持使用快捷键 `Ctrl+Z` 和 `Ctrl+Y`（Mac上是 `Cmd+Z` / `Cmd+Shift+Z`）来快速触发。为了性能考虑，历史栈长度限制最多保存 50 步即可。\\n后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在undo_redo_plan.md中写入本次修改的规划，此前要求写入cell_op_plan.md的规则不再生效；第二，确保在你生成的所有代码文件中，等号（=）的左右两侧都留有两个空格，该要求后续所有开发过程均需遵守。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 验证工作区目录是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据项目推断常见的代码和配置文件的扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.json', '.md', '.py', '.java', '.c', '.cpp', '.h'\\n }\\n\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git 等版本控制和配置目录）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试用 utf-8 编码读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果解码发生异常，说明不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的系统级读取错误（如权限问题等）\\n pass\\n\\n # 如果有代码文件被扫描到并且都没有引发解码异常，则满足要求\\n return has_code_files\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_js_functions(js_code: str) -> bool:\\n # 去除注释内容，防止注释内的示例代码影响判断\\n js_code = re.sub(r'//.*', '', js_code)\\n js_code = re.sub(r'/\\\\*.*?\\\\*/', '', js_code, flags=re.DOTALL)\\n \\n # 去除字符串内容，防止字符串中包含类似函数的结构误导判断\\n js_code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', js_code)\\n js_code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", js_code)\\n js_code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', js_code, flags=re.DOTALL)\\n \\n # 正则匹配不同类型的 JS/TS 函数定义\\n # 1. 声明式函数: function foo(...)\\n p1 = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\(')\\n # 2. 赋值型函数: foo = function(...) 或 foo = (...) =>\\n p2 = re.compile(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n # 3. 对象属性函数: foo: function(...) 或 foo: (...) =>\\n p3 = re.compile(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n # 4. 类方法/对象简写方法: foo(...) {\\n p4 = re.compile(r'(?:^|[\\\\{\\\\s;])(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+)?(?:static\\\\s+)?([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n funcs = []\\n funcs.extend(p1.findall(js_code))\\n funcs.extend(p2.findall(js_code))\\n funcs.extend(p3.findall(js_code))\\n funcs.extend(p4.findall(js_code))\\n \\n # 排除JS关键字和常见控制流语句等可能被作为类方法正则误捕获的内容\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', \\n 'else', 'function', 'import', 'export', 'require', 'super', 'get', 'set',\\n 'async', 'await', 'with', 'typeof', 'instanceof', 'in', 'of', 'new',\\n 'class', 'extends', 'yield', 'let', 'const', 'var', 'default', 'finally',\\n 'break', 'continue', 'debugger', 'do', 'throw', 'try'\\n }\\n \\n for f_name in funcs:\\n if f_name in keywords:\\n continue\\n # 找到任何长度小于4的函数名则直接判定不符合要求\\n if len(f_name) < 4:\\n return False\\n \\n return True\\n\\n def check_py_functions(py_code: str) -> bool:\\n try:\\n tree = ast.parse(py_code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if len(node.name) < 4:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n # 遍历项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略包依赖和版本控制目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n filepath = os.path.join(root, file)\\n \\n # Python 文件验证\\n if ext == '.py':\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_py_functions(content):\\n return False\\n except Exception:\\n pass\\n \\n # HTML 或 JS/TS 文件验证\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.htm']:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if ext in ['.html', '.htm']:\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for js_code in scripts:\\n if not check_js_functions(js_code):\\n return False\\n else:\\n if not check_js_functions(content):\\n return False\\n except Exception:\\n pass\\n \\n # 没有发现任何不符合规定的情况\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何单个函数的行数均不得多于60行\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n def clean_js(code: str) -> str:\\n \\\"\\\"\\\"移除JS代码中的注释和字符串内容，保留换行和基础结构，不干扰行号计算\\\"\\\"\\\"\\n i = 0\\n n = len(code)\\n cleaned = []\\n in_single_comment = False\\n in_multi_comment = False\\n in_string = False\\n string_char = ''\\n \\n while i < n:\\n char = code[i]\\n if in_single_comment:\\n if char == '\\\\n':\\n in_single_comment = False\\n cleaned.append(char)\\n else:\\n cleaned.append(' ')\\n i += 1\\n continue\\n \\n if in_multi_comment:\\n if char == '*' and i + 1 < n and code[i+1] == '/':\\n in_multi_comment = False\\n cleaned.append(' ')\\n i += 2\\n else:\\n cleaned.append('\\\\n' if char == '\\\\n' else ' ')\\n i += 1\\n continue\\n \\n if in_string:\\n if char == '\\\\\\\\':\\n cleaned.append(' ')\\n i += 2\\n continue\\n if char == string_char:\\n in_string = False\\n cleaned.append(' ')\\n else:\\n cleaned.append('\\\\n' if char == '\\\\n' else ' ')\\n i += 1\\n continue\\n \\n if char == '/' and i + 1 < n and code[i+1] == '/':\\n in_single_comment = True\\n cleaned.append(' ')\\n i += 2\\n continue\\n \\n if char == '/' and i + 1 < n and code[i+1] == '*':\\n in_multi_comment = True\\n cleaned.append(' ')\\n i += 2\\n continue\\n \\n if char in (\\\"'\\\", '\\\"', \\\"`\\\"):\\n in_string = True\\n string_char = char\\n cleaned.append(' ')\\n i += 1\\n continue\\n \\n cleaned.append(char)\\n i += 1\\n \\n return \\\"\\\".join(cleaned)\\n\\n def extract_js_from_html(html_code: str) -> str:\\n \\\"\\\"\\\"抽取HTML/Vue中<script>标签内容，通过将非JS部分替换为空白的方式保持行数对齐\\\"\\\"\\\"\\n js_code = \\\"\\\"\\n i = 0\\n n = len(html_code)\\n in_script = False\\n \\n while i < n:\\n if not in_script:\\n script_start = html_code.find('<script', i)\\n if script_start == -1:\\n js_code += re.sub(r'[^\\\\n]', ' ', html_code[i:])\\n break\\n else:\\n tag_end = html_code.find('>', script_start)\\n if tag_end == -1:\\n js_code += re.sub(r'[^\\\\n]', ' ', html_code[i:])\\n break\\n js_code += re.sub(r'[^\\\\n]', ' ', html_code[i:tag_end+1])\\n i = tag_end + 1\\n in_script = True\\n else:\\n script_end = html_code.find('</script>', i)\\n if script_end == -1:\\n js_code += html_code[i:]\\n break\\n else:\\n js_code += html_code[i:script_end]\\n js_code += re.sub(r'[^\\\\n]', ' ', html_code[script_end:script_end+9])\\n i = script_end + 9\\n in_script = False\\n return js_code\\n\\n def get_prefix_for_brace(code_str: str, brace_index: int) -> str:\\n \\\"\\\"\\\"从某一个 '{' 向前搜索，获取属于该代码块的声明语句，处理了嵌套圆括号折叠的规避\\\"\\\"\\\"\\n i = brace_index - 1\\n parens = 0\\n braces = 0\\n brackets = 0\\n collected_tokens = []\\n \\n while i >= 0:\\n char = code_str[i]\\n if char == ')':\\n parens += 1\\n collected_tokens.append(char)\\n elif char == '(':\\n if parens > 0:\\n parens -= 1\\n collected_tokens.append(char)\\n elif char == '}':\\n braces += 1\\n collected_tokens.append(char)\\n elif char == '{':\\n if braces > 0:\\n braces -= 1\\n collected_tokens.append(char)\\n else:\\n if parens == 0 and brackets == 0:\\n break\\n else:\\n collected_tokens.append(char)\\n elif char == ']':\\n brackets += 1\\n collected_tokens.append(char)\\n elif char == '[':\\n if brackets > 0:\\n brackets -= 1\\n collected_tokens.append(char)\\n elif char == ';':\\n if parens == 0 and braces == 0 and brackets == 0:\\n break\\n else:\\n collected_tokens.append(char)\\n else:\\n collected_tokens.append(char)\\n i -= 1\\n \\n collected_tokens.reverse()\\n return \\\"\\\".join(collected_tokens)\\n\\n def is_function_prefix(prefix: str) -> bool:\\n \\\"\\\"\\\"根据大括号前缀来确定该块是不是一个函数（常规函数、箭头函数、类方法等）\\\"\\\"\\\"\\n prefix = prefix.strip()\\n if not prefix:\\n return False\\n if prefix.endswith('=>'):\\n return True\\n \\n keywords_exclude = {'if', 'for', 'while', 'switch', 'catch', 'with', 'return'}\\n \\n if prefix.endswith(')'):\\n parens = 0\\n idx = len(prefix) - 1\\n while idx >= 0:\\n if prefix[idx] == ')':\\n parens += 1\\n elif prefix[idx] == '(':\\n parens -= 1\\n if parens == 0:\\n break\\n idx -= 1\\n \\n if idx >= 0:\\n before_paren = prefix[:idx]\\n match = re.search(r'\\\\b(\\\\w+)\\\\s*$', before_paren)\\n if match:\\n word = match.group(1)\\n if word not in keywords_exclude:\\n return True\\n elif prefix.endswith('function'):\\n return True\\n \\n return False\\n\\n def check_functions_length_js(code: str) -> bool:\\n cleaned = clean_js(code)\\n stack = []\\n line_number = 1\\n \\n for i, char in enumerate(cleaned):\\n if char == '\\\\n':\\n line_number += 1\\n elif char == '{':\\n prefix = get_prefix_for_brace(cleaned, i)\\n is_func = is_function_prefix(prefix)\\n stack.append((line_number, is_func))\\n elif char == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n if line_number - start_line + 1 > 60:\\n return False\\n return True\\n\\n def check_functions_length_py(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return True\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_line = node.lineno\\n end_line = getattr(node, 'end_lineno', start_line)\\n if end_line - start_line + 1 > 60:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 过滤隐藏文件及文件夹\\n if '/.' in file_path or '\\\\\\\\.' in file_path:\\n continue\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if not check_functions_length_js(content):\\n return False\\n elif ext in ['.html', '.htm', '.vue']:\\n js_content = extract_js_from_html(content)\\n if not check_functions_length_js(js_content):\\n return False\\n elif ext == '.py':\\n if not check_functions_length_py(content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在undo_redo_plan.md中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，等号（=）的左右两侧都留有两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对前端及常见代码文件的扩展名进行限制，过滤掉 .md、.txt 等非代码类型的文档\\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 仅移除行末的换行符，保留行末可能存在的空格以便精确验证\\n line = line.rstrip('\\\\n\\\\r')\\n i = 0\\n while i < len(line):\\n if line[i] == '=':\\n # 判定该等号是否属于复合操作符的一部分（如 ==, >=, !=, +=, => 等）\\n is_compound = False\\n if i > 0 and line[i-1] in '=<>!+-*/%&|^':\\n is_compound = True\\n if i < len(line) - 1 and line[i+1] in '=>':\\n is_compound = True\\n \\n if not is_compound:\\n # 检查左侧是否有且仅有两个空格\\n left_ok = (i >= 2 and line[i-1] == ' ' and line[i-2] == ' ' and (i == 2 or line[i-3] != ' '))\\n \\n # 检查右侧是否有且仅有两个空格\\n right_ok = (i <= len(line) - 3 and line[i+1] == ' ' and line[i+2] == ' ' and (i == len(line) - 3 or line[i+3] != ' '))\\n \\n # 如果有任意一侧不满足留有两个空格的条件，则判定为失败\\n if not (left_ok and right_ok):\\n return False\\n i += 1\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"最后一步，我们给表格增加一些基础的富文本样式定制能力。在工具栏补充一个样式编辑面板：包括文本的『加粗』、『斜体』开关按钮，以及『文字颜色』和『单元格背景色』的原生取色器（color picker）。当我选中单个单元格或者框选多个单元格后，操作这些样式工具，选中的单元格样式会立刻发生改变。理所当然地，这些关于文字和背景的样式元数据也必须深度整合进你的数据结构中，存入 LocalStorage，并在撤销/重做以及页面刷新时完美恢复。\\n后续所有开发操作都必须始终遵循以下两项要求：第一，无论何时修改代码，首要步骤都必须是在style_edit_plan.md中写入本次修改的规划，此前要求写入undo_redo_plan.md的规则不再生效；第二，你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径，该要求后续所有开发过程均需遵守。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡涉及对现有文件的更改，请务必以严格的Unified Diff格式呈现，切勿直接返回全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了避免进程挂起，请在执行任意 bash 测试、安装或构建命令的最前面，强制加上一层 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保在改动已有文件前，已经使用相应工具对该文件进行了阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在任何文件操作前，都不允许执行`pwd`来获取当前目录信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"用以记录本次改动的新文件 change_log_X.log 必须在每次项目改动后生成，X 需要替换为从 0 开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必保证项目内包含一个名为 me.md 的结构与功能说明文档，且每次有内容更新时都要对其进行同步更新。每次开始编写新功能前，切记先用 tail -500 me.md 命令读取该文件的最后500行，以免触发上下文超限\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"要求代码实现走最短路径，杜绝过度设计，并且不得擅自增添兜底、降级方案等任何脱离原需求的额外逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断，前端Web编辑器项目主要涉及的代码及相关文本文件类型\\n target_extensions = {'.html', '.js', '.css', '.md', '.json', '.log'}\\n \\n if not os.path.exists(workspace_path):\\n # 如果项目路径不存在，则无法判断或视为未违背此规则\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 检查是否为相关的代码或文本文件\\n if any(file.lower().endswith(ext) for ext in target_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码进行读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明该文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如权限不足等其他与编码无关的读取异常\\n continue\\n \\n # 所有文件都可以成功按照utf-8解析，满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中每个函数的名称都具有不少于4个字符的长度\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_function_name_length(response: str, workspace_path: str) -> bool:\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义正则表达式匹配各类语言中的函数定义\\n \\n # 1. Python 函数定义\\n py_func_pattern = re.compile(r'def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n py_lambda_pattern = re.compile(r'([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=\\\\s*lambda\\\\b')\\n \\n # 2. JS/TS 关键字 function 声明 (支持 function foo( 以及 function * foo( 等)\\n js_func_pattern = re.compile(r'function(?:\\\\s+|\\\\s*\\\\*\\\\s*)([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\(')\\n \\n # 3. JS/TS function 赋值给变量或属性，如 foo = function() 或 foo: function()\\n js_assign_func_pattern = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*(?:=|:)\\\\s*function\\\\b')\\n \\n # 4. JS/TS 箭头函数赋值给变量或属性，如 foo = () => 或 foo: a =>\\n js_arrow_func_pattern = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*(?:=|:)\\\\s*(?:\\\\([^)]*\\\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=>')\\n \\n # 5. JS/TS 对象或类方法定义，形如 methodName(params) {\\n # 确保前导不是小数点、单词字符或 $ 符号等，以免匹配到对象属性点调用\\n js_method_pattern = re.compile(r'(?:^|[^\\\\w.$])([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\(([^()]*)\\\\)\\\\s*\\\\{')\\n \\n # HTML 中 <script> 标签的提取正则，避免直接在 HTML 层面正则引起误判（如 CSS 伪类）\\n script_pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\\n \\n # 排除容易被类方法正则误判的控制流关键字、原生类型或 CSS 函数\\n exclude_keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'with', \\n 'function', 'return', 'constructor', 'typeof',\\n 'try', 'finally', 'else', 'elif', 'except',\\n 'is', 'has', 'not', 'var', 'url', 'let', 'get', 'set',\\n 'new'\\n }\\n \\n # 指定需要检查的代码文件扩展名\\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.py'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 文件不可读时跳过\\n continue\\n \\n # HTML 文件只检查 <script> 标签中的内容\\n if ext in {'.html', '.htm', '.vue'}:\\n scripts = script_pattern.findall(content)\\n content_to_check = '\\\\n'.join(scripts)\\n else:\\n content_to_check = content\\n \\n func_names = []\\n \\n if ext == '.py':\\n func_names.extend(py_func_pattern.findall(content_to_check))\\n func_names.extend(py_lambda_pattern.findall(content_to_check))\\n else:\\n func_names.extend(js_func_pattern.findall(content_to_check))\\n func_names.extend(js_assign_func_pattern.findall(content_to_check))\\n func_names.extend(js_arrow_func_pattern.findall(content_to_check))\\n \\n # 对象或类的方法需特殊过滤\\n methods = js_method_pattern.findall(content_to_check)\\n for name, params in methods:\\n # 如果匹配到的方法参数中包含 '=>'，那通常是作为参数传递进去的匿名箭头函数被误匹配，剔除\\n if '=>' in params:\\n continue\\n func_names.append(name)\\n \\n # 过滤并检查所有识别到的函数名称长度\\n for name in func_names:\\n if name in exclude_keywords:\\n continue\\n # 一旦找到名称长度小于 4 个字符的函数，即不满足要求\\n if len(name) < 4:\\n return False\\n \\n # 没有发现任何违背规则的情况，满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何单个函数的行数均不得多于60行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\n def extract_js_from_html(html_code: str) -> str:\\n # 使用正则表达式查找 script 标签的内容\\n pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\\n result = list(html_code)\\n # 默认全部替换为空格，遇到换行保留换行符（不改变行号和行结构）\\n for i in range(len(result)):\\n if result[i] != '\\\\n':\\n result[i] = ' '\\n # 只将被匹配的 script 内部代码恢复出来\\n for match in pattern.finditer(html_code):\\n start = match.start(1)\\n end = match.end(1)\\n result[start:end] = list(html_code[start:end])\\n return \\\"\\\".join(result)\\n\\n def strip_comments_and_strings(code: str) -> str:\\n # 将 JS 中的字符串和注释清理为空格，防止内部的括号干扰匹配\\n result = list(code)\\n i = 0\\n n = len(code)\\n \\n def is_regex_start(idx):\\n j = idx - 1\\n while j >= 0 and code[j].isspace():\\n j -= 1\\n if j < 0: return True\\n c = code[j]\\n # 如果正则表达式前面是这些符号，大概率是合法的正则开始\\n if c in '=({[<>,:;!&|?+-*/%~^':\\n return True\\n if c.isalpha() or c in '_$':\\n k = j\\n while k >= 0 and (code[k].isalnum() or code[k] in '_$'):\\n k -= 1\\n word = code[k+1:j+1]\\n # 特殊的关键字后面可能跟着正则\\n if word in ['return', 'typeof', 'yield', 'await', 'case', 'throw', 'delete', 'in', 'instanceof', 'new', 'do', 'else']:\\n return True\\n return False\\n\\n while i < n:\\n c = code[i]\\n if c == '/' and i + 1 < n:\\n nc = code[i+1]\\n if nc == '/': # 单行注释\\n while i < n and code[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n continue\\n elif nc == '*': # 多行注释\\n result[i] = ' '\\n result[i+1] = ' '\\n i += 2\\n while i < n and not (code[i] == '*' and i + 1 < n and code[i+1] == '/'):\\n if result[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n if i < n:\\n result[i] = ' '\\n if i + 1 < n:\\n result[i+1] = ' '\\n i += 2\\n continue\\n elif is_regex_start(i): # 简单的正则字面量探测\\n result[i] = ' '\\n i += 1\\n while i < n:\\n if code[i] == '\\\\\\\\':\\n result[i] = ' '\\n if i + 1 < n:\\n if result[i+1] != '\\\\n':\\n result[i+1] = ' '\\n i += 1\\n elif code[i] == '/':\\n result[i] = ' '\\n i += 1\\n break\\n elif code[i] == '\\\\n':\\n break\\n else:\\n result[i] = ' '\\n i += 1\\n continue\\n \\n elif c in ['\\\\'', '\\\"', '`']: # 字符串探测\\n quote = c\\n result[i] = ' '\\n i += 1\\n while i < n:\\n if code[i] == '\\\\\\\\':\\n result[i] = ' '\\n if i + 1 < n:\\n if result[i+1] != '\\\\n':\\n result[i+1] = ' '\\n i += 1\\n elif code[i] == quote:\\n result[i] = ' '\\n i += 1\\n break\\n else:\\n if result[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n continue\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def find_function_blocks(clean_code: str):\\n # 通过寻找成对的大括号，并根据大括号前的关键词智能判定是否是函数\\n blocks = []\\n stack_curly = []\\n stack_paren = []\\n \\n pairs_curly = {}\\n pairs_paren = {}\\n \\n for idx, char in enumerate(clean_code):\\n if char == '{':\\n stack_curly.append(idx)\\n elif char == '}':\\n if stack_curly:\\n start = stack_curly.pop()\\n pairs_curly[start] = idx\\n elif char == '(':\\n stack_paren.append(idx)\\n elif char == ')':\\n if stack_paren:\\n start = stack_paren.pop()\\n pairs_paren[idx] = start\\n\\n for start_idx, end_idx in pairs_curly.items():\\n j = start_idx - 1\\n while j >= 0 and clean_code[j].isspace():\\n j -= 1\\n if j < 0:\\n continue\\n \\n is_func = False\\n prev_char = clean_code[j]\\n \\n # 判断箭头函数\\n if prev_char == '>':\\n if j - 1 >= 0 and clean_code[j-1] == '=':\\n is_func = True\\n # 判断常规方法或者函数\\n elif prev_char == ')':\\n if j in pairs_paren:\\n paren_start = pairs_paren[j]\\n k = paren_start - 1\\n while k >= 0 and clean_code[k].isspace():\\n k -= 1\\n \\n if k >= 0:\\n if clean_code[k].isalnum() or clean_code[k] in '_$':\\n m = k\\n while m >= 0 and (clean_code[m].isalnum() or clean_code[m] in '_$'):\\n m -= 1\\n word = clean_code[m+1:k+1]\\n control_keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\\n if word not in control_keywords:\\n is_func = True\\n elif clean_code[k] == ']': # 可能为动态属性名称的方法，如 `[methodName]() {`\\n is_func = True\\n \\n if is_func:\\n start_line = clean_code[:start_idx].count('\\\\n') + 1\\n end_line = clean_code[:end_idx].count('\\\\n') + 1\\n blocks.append((start_line, end_line))\\n \\n return blocks\\n\\n valid_extensions = ('.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n if file.endswith(('.html', '.htm', '.vue')):\\n code = extract_js_from_html(code)\\n \\n clean_code = strip_comments_and_strings(code)\\n funcs = find_function_blocks(clean_code)\\n \\n # 检查每一个提取出来的函数行数\\n for start, end in funcs:\\n func_lines = end - start + 1\\n if func_lines > 60:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，等号（=）的左右两侧都留有两个空格\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的前端项目代码文件扩展名\\n code_extensions = {'.js', '.ts', '.html', '.css', '.vue', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n # 正则：匹配独立的 '='，通过否定环视排除前后的复合运算符部分（如 ==, ===, >=, <=, !=, +=, => 等）\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%^&|~?])=(?![=>])')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_code_file = True\\n \\n for line in lines:\\n # 去除行尾换行符，以免影响右侧空格数量的判断\\n line = line.rstrip('\\\\n').rstrip('\\\\r')\\n \\n for match in pattern.finditer(line):\\n idx = match.start()\\n \\n left_part = line[:idx]\\n right_part = line[idx+1:]\\n \\n # 统计左侧末尾的连续空格数量\\n left_spaces = len(left_part) - len(left_part.rstrip(' '))\\n # 统计右侧开头的连续空格数量\\n right_spaces = len(right_part) - len(right_part.lstrip(' '))\\n \\n # 如果有一侧连续空格不严格为2，则不满足要求\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n \\n # 只要检查到了至少一个代码文件且都通过了校验，就返回True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在style_edit_plan.md中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["打开页面检查初始布局：页面上方存在固定显示的工具栏区域（包含按钮占位），工具栏下方居中显示一个带有明显边框的初始状态的 5行 x 5列 空白表格。","鼠标单击表格中任意一个未被选中的空白单元格，该单元格背景立刻变为浅蓝色，表示该单元格被高亮选中。","双击一个已经被单击选中（高亮）的单元格，验证该单元格会变成一个激活聚焦的文本输入框，允许在其中输入文字。","在双击变出的输入框中输入测试文本，随后使用鼠标点击表格外部任意非输入框区域使其失去焦点，验证输入框消失，输入的文字留在该单元格中，且该单元格依旧保持浅蓝色选中状态。","在另一个单元格双击输入文本，完成输入后按下键盘回车键（Enter），验证输入框立刻消失，刚才输入的文本正确保留显示在该单元格内。","对表格输入内容后直接刷新当前网页，页面加载完成后，验证此前输入并保存的文本内容依然存在于对应的单元格中，实现 LocalStorage 基础持久化。","找到工具栏最右侧的呈现红色的“重置表格”按钮并点击，验证原有的数据和行列更改全部被清空，表格强制立刻恢复到初始的 5x5 空白表格状态。","选中特定单元格并依次点击工具栏中的“在上方插入行”和“在下方插入行”按钮，验证表格总行数增加 2 行，并在所选单元格的上方与下方分别出现一整行的全新空白单元格。","选中特定单元格并依次点击工具栏中的“在左侧插入列”和“在右侧插入列”按钮，验证表格总列数增加 2 列，并在所选单元格的左右两边分别出现一整列的全新空白单元格。","选中存在测试文字的单元格，依次点击“删除当前行”和“删除当前列”按钮，验证所在的整行与整列立刻从表格中消失，对应的行列数量随之减少。","在原 5x5 的基础上增加若干行列，并删除一列后（如改变为 7行4列），刷新当前网页，验证加载后的表格依然完整维持改变后的行列数和位置布局。","按住鼠标左键在某单元格处按下，向右下角连续滑动到其它单元格后松开鼠标，验证这两个单元格组成的连续矩形区域内的所有单元格都被添加上浅蓝色的高亮背景。","确保被框选高亮的多个矩形区域单元格内有不同的文本内容，点击工具栏中的“合并单元格”按钮，验证该框选区域变为一个大单元格，并且大单元格中仅保留并显示原左上角第一个单元格的文本，其余文本被丢弃。","鼠标单击选中上一步生成的合并大单元格，点击工具栏中的“拆分单元格”按钮，验证该大单元格恢复拆分为最初的多个独立的小格子。","重新框选合并出大单元格并刷新网页，验证页面加载完成后，之前合并的大单元格结构及其保留的文本仍然被正确渲染展现。","在某一单元格输入新文本并保存，随后点击工具栏左侧的“撤销(Undo)”按钮，验证该单元格刚输入的文本消失，恢复到修改前的内容状态。","完成上一步“撤销”操作后，点击工具栏的“重做(Redo)”按钮，验证刚才消失的文本再次出现并恢复在该单元格内。","对表格内容进行修改后，不点击按钮，直接在键盘上按下快捷键组合 Ctrl+Z (或 Cmd+Z) ，验证发生撤销操作；再按下组合 Ctrl+Y (或 Cmd+Shift+Z) ，验证发生重做恢复操作。","点击“在下方插入行”增加一行后，立刻触发“撤销”操作（点击按钮或快捷键），验证刚才新增的一整行空白单元格立刻被删除，表格回到插入前的结构和状态,再触发“重做”操作（点击按钮或快捷键），验证新增了一整行空白单元格。","选中含有文字的某个单元格，点击工具栏中的“加粗”和“斜体”开关按钮，验证所选单元格内的文本立即变为了加粗和斜体样式展现。","拖拽框选一组含有文本的连续单元格，在工具栏的颜色取色器控件中选取一种文字颜色（如红色）和单元格背景色（如黄色），验证被选中的这一组单元格字体颜色及背景颜色立即同步发生相应的改变。","在单元格上应用加粗样式和特定的背景颜色后，刷新整个网页，验证加载后的表格依然保留并正确显示刚刚设置的文本加粗和背景色样式。","给原本没有样式的单元格文字设置为红色后，执行“撤销”操作，验证文字颜色立刻恢复成默认颜色；随后执行“重做”操作，文字颜色再次变为红色。","框选一组单元格并合并后，点击'撤销'按钮，验证合并的大单元格恢复为合并前的多个独立小格子；随后点击'重做'按钮，验证小格子再次合并为一个大单元格。"],"string":"[\n \"打开页面检查初始布局：页面上方存在固定显示的工具栏区域（包含按钮占位），工具栏下方居中显示一个带有明显边框的初始状态的 5行 x 5列 空白表格。\",\n \"鼠标单击表格中任意一个未被选中的空白单元格，该单元格背景立刻变为浅蓝色，表示该单元格被高亮选中。\",\n \"双击一个已经被单击选中（高亮）的单元格，验证该单元格会变成一个激活聚焦的文本输入框，允许在其中输入文字。\",\n \"在双击变出的输入框中输入测试文本，随后使用鼠标点击表格外部任意非输入框区域使其失去焦点，验证输入框消失，输入的文字留在该单元格中，且该单元格依旧保持浅蓝色选中状态。\",\n \"在另一个单元格双击输入文本，完成输入后按下键盘回车键（Enter），验证输入框立刻消失，刚才输入的文本正确保留显示在该单元格内。\",\n \"对表格输入内容后直接刷新当前网页，页面加载完成后，验证此前输入并保存的文本内容依然存在于对应的单元格中，实现 LocalStorage 基础持久化。\",\n \"找到工具栏最右侧的呈现红色的“重置表格”按钮并点击，验证原有的数据和行列更改全部被清空，表格强制立刻恢复到初始的 5x5 空白表格状态。\",\n \"选中特定单元格并依次点击工具栏中的“在上方插入行”和“在下方插入行”按钮，验证表格总行数增加 2 行，并在所选单元格的上方与下方分别出现一整行的全新空白单元格。\",\n \"选中特定单元格并依次点击工具栏中的“在左侧插入列”和“在右侧插入列”按钮，验证表格总列数增加 2 列，并在所选单元格的左右两边分别出现一整列的全新空白单元格。\",\n \"选中存在测试文字的单元格，依次点击“删除当前行”和“删除当前列”按钮，验证所在的整行与整列立刻从表格中消失，对应的行列数量随之减少。\",\n \"在原 5x5 的基础上增加若干行列，并删除一列后（如改变为 7行4列），刷新当前网页，验证加载后的表格依然完整维持改变后的行列数和位置布局。\",\n \"按住鼠标左键在某单元格处按下，向右下角连续滑动到其它单元格后松开鼠标，验证这两个单元格组成的连续矩形区域内的所有单元格都被添加上浅蓝色的高亮背景。\",\n \"确保被框选高亮的多个矩形区域单元格内有不同的文本内容，点击工具栏中的“合并单元格”按钮，验证该框选区域变为一个大单元格，并且大单元格中仅保留并显示原左上角第一个单元格的文本，其余文本被丢弃。\",\n \"鼠标单击选中上一步生成的合并大单元格，点击工具栏中的“拆分单元格”按钮，验证该大单元格恢复拆分为最初的多个独立的小格子。\",\n \"重新框选合并出大单元格并刷新网页，验证页面加载完成后，之前合并的大单元格结构及其保留的文本仍然被正确渲染展现。\",\n \"在某一单元格输入新文本并保存，随后点击工具栏左侧的“撤销(Undo)”按钮，验证该单元格刚输入的文本消失，恢复到修改前的内容状态。\",\n \"完成上一步“撤销”操作后，点击工具栏的“重做(Redo)”按钮，验证刚才消失的文本再次出现并恢复在该单元格内。\",\n \"对表格内容进行修改后，不点击按钮，直接在键盘上按下快捷键组合 Ctrl+Z (或 Cmd+Z) ，验证发生撤销操作；再按下组合 Ctrl+Y (或 Cmd+Shift+Z) ，验证发生重做恢复操作。\",\n \"点击“在下方插入行”增加一行后，立刻触发“撤销”操作（点击按钮或快捷键），验证刚才新增的一整行空白单元格立刻被删除，表格回到插入前的结构和状态,再触发“重做”操作（点击按钮或快捷键），验证新增了一整行空白单元格。\",\n \"选中含有文字的某个单元格，点击工具栏中的“加粗”和“斜体”开关按钮，验证所选单元格内的文本立即变为了加粗和斜体样式展现。\",\n \"拖拽框选一组含有文本的连续单元格，在工具栏的颜色取色器控件中选取一种文字颜色（如红色）和单元格背景色（如黄色），验证被选中的这一组单元格字体颜色及背景颜色立即同步发生相应的改变。\",\n \"在单元格上应用加粗样式和特定的背景颜色后，刷新整个网页，验证加载后的表格依然保留并正确显示刚刚设置的文本加粗和背景色样式。\",\n \"给原本没有样式的单元格文字设置为红色后，执行“撤销”操作，验证文字颜色立刻恢复成默认颜色；随后执行“重做”操作，文字颜色再次变为红色。\",\n \"框选一组单元格并合并后，点击'撤销'按钮，验证合并的大单元格恢复为合并前的多个独立小格子；随后点击'重做'按钮，验证小格子再次合并为一个大单元格。\"\n]","isConversational":false}}},{"rowIdx":55,"cells":{"task_id":{"kind":"number","value":56,"string":"56"},"repository_policy":{"kind":"string","value":"你在后续所有轮次的回复和操作都必须严格遵循以下约束：\n1. 你的任何用户回复都必须全程使用中文进行生成。\n2. 确保每一个生成的代码文件里，注释部分完全使用中文。\n3. 目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%。\n4. 你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中。\n5. 生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回。\n6. 所有由你创建的代码文件，其编码必须统一为UTF-8。\n7. 所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线。\n8. 若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码和配置文件后缀名，包含了推断出的JS/TS项目可能涉及的文件类型\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.json', \n '.html', '.css', '.md', '.txt'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除诸如 node_modules 或 .git 等第三方或版本控制相关的冗余目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦抛出解码错误，说明存在非UTF-8（如GBK等）编码的文件\n return False\n except Exception:\n # 忽略其他非编码相关的偶发文件读取异常\n pass\n \n # 如果扫描了代码文件并且均无解码报错，则满足条件\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 匹配 Python 函数定义: def func_name(\n python_func_pattern = re.compile(r'^\\s*def\\s+([a-zA-Z0-9_]+)\\s*\\(', re.MULTILINE)\n \n # 匹配 JS/TS 常见函数声明关键字，用于过滤控制流保留字，避免误触发类方法匹配\n js_keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'constructor', 'with'}\n \n # 匹配 JS/TS 普通函数: function funcName(\n js_pattern1 = re.compile(r'\\bfunction\\s+([a-zA-Z0-9_]+)\\s*\\(')\n # 匹配 JS/TS 赋值形式和箭头函数: const/let/var funcName = (...) => 或者 = function(...)\n js_pattern2 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:function\\s*\\(|\\([^)]*\\)\\s*=>|[a-zA-Z0-9_]+\\s*=>)')\n # 匹配 JS/TS 类方法或对象内的方法: [async/get/set] methodName( ... ) {\n js_pattern3 = re.compile(r'\\b(?:async\\s+|get\\s+|set\\s+)?([a-zA-Z0-9_]+)\\s*\\([^)]*\\)\\s*\\{')\n \n # 校验规则：只允许小写字母和下划线\n valid_name_pattern = re.compile(r'^[a-z_]+$')\n \n def remove_comments(code_str: str, ext: str) -> str:\n \"\"\"移除代码中的注释，防止正则表达式在注释中发生误匹配\"\"\"\n if ext == '.py':\n code_str = re.sub(r'#.*', '', code_str)\n # 移除Python多行注释/字符串\n code_str = re.sub(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', '', code_str, flags=re.DOTALL)\n code_str = re.sub(r'\\'\\'\\'(.*?)\\'\\'\\'', '', code_str, flags=re.DOTALL)\n else:\n code_str = re.sub(r'//.*', '', code_str)\n # 移除JS/TS多行注释\n code_str = re.sub(r'/\\*.*?\\*/', '', code_str, flags=re.DOTALL)\n return code_str\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤主流的Python和JS/TS相关代码文件\n if ext in {'.py', '.js', '.jsx', '.ts', '.tsx'}:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 预处理移除注释\n content = remove_comments(content, ext)\n names = []\n \n # 依据后缀进行正则提取\n if ext == '.py':\n names.extend(python_func_pattern.findall(content))\n else:\n names.extend(js_pattern1.findall(content))\n names.extend(js_pattern2.findall(content))\n \n methods = js_pattern3.findall(content)\n for method in methods:\n # 剔除常见的控制流关键字等，避免误报\n if method not in js_keywords:\n names.append(method)\n \n # 检验提取出的函数名\n for name in names:\n if not valid_name_pattern.match(name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]}],"string":"[\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码和配置文件后缀名，包含了推断出的JS/TS项目可能涉及的文件类型\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.json', \\n '.html', '.css', '.md', '.txt'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除诸如 node_modules 或 .git 等第三方或版本控制相关的冗余目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦抛出解码错误，说明存在非UTF-8（如GBK等）编码的文件\\n return False\\n except Exception:\\n # 忽略其他非编码相关的偶发文件读取异常\\n pass\\n \\n # 如果扫描了代码文件并且均无解码报错，则满足条件\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 匹配 Python 函数定义: def func_name(\\n python_func_pattern = re.compile(r'^\\\\s*def\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(', re.MULTILINE)\\n \\n # 匹配 JS/TS 常见函数声明关键字，用于过滤控制流保留字，避免误触发类方法匹配\\n js_keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'constructor', 'with'}\\n \\n # 匹配 JS/TS 普通函数: function funcName(\\n js_pattern1 = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(')\\n # 匹配 JS/TS 赋值形式和箭头函数: const/let/var funcName = (...) => 或者 = function(...)\\n js_pattern2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\s*\\\\(|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z0-9_]+\\\\s*=>)')\\n # 匹配 JS/TS 类方法或对象内的方法: [async/get/set] methodName( ... ) {\\n js_pattern3 = re.compile(r'\\\\b(?:async\\\\s+|get\\\\s+|set\\\\s+)?([a-zA-Z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # 校验规则：只允许小写字母和下划线\\n valid_name_pattern = re.compile(r'^[a-z_]+$')\\n \\n def remove_comments(code_str: str, ext: str) -> str:\\n \\\"\\\"\\\"移除代码中的注释，防止正则表达式在注释中发生误匹配\\\"\\\"\\\"\\n if ext == '.py':\\n code_str = re.sub(r'#.*', '', code_str)\\n # 移除Python多行注释/字符串\\n code_str = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\"(.*?)\\\\\\\"\\\\\\\"\\\\\\\"', '', code_str, flags=re.DOTALL)\\n code_str = re.sub(r'\\\\'\\\\'\\\\'(.*?)\\\\'\\\\'\\\\'', '', code_str, flags=re.DOTALL)\\n else:\\n code_str = re.sub(r'//.*', '', code_str)\\n # 移除JS/TS多行注释\\n code_str = re.sub(r'/\\\\*.*?\\\\*/', '', code_str, flags=re.DOTALL)\\n return code_str\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤主流的Python和JS/TS相关代码文件\\n if ext in {'.py', '.js', '.jsx', '.ts', '.tsx'}:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 预处理移除注释\\n content = remove_comments(content, ext)\\n names = []\\n \\n # 依据后缀进行正则提取\\n if ext == '.py':\\n names.extend(python_func_pattern.findall(content))\\n else:\\n names.extend(js_pattern1.findall(content))\\n names.extend(js_pattern2.findall(content))\\n \\n methods = js_pattern3.findall(content)\\n for method in methods:\\n # 剔除常见的控制流关键字等，避免误报\\n if method not in js_keywords:\\n names.append(method)\\n \\n # 检验提取出的函数名\\n for name in names:\\n if not valid_name_pattern.match(name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"UI/UX Optimization"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我搭建一个简单的HTML页面，里面包含一个SVG画布，用于设计一个Z字母的Logo。页面背景请设为深灰色（#1a1a1a），SVG画布居中对齐，宽高设置为 500x500。在这个SVG里，先用白色的粗线条或一个简单的多边形（polygon）勾勒出一个占满核心区域的大的“Z”字母草图骨架，设定点半透明度即可。目前只需要搭好页面框架、定好Z字母的比例和位置，不需要其他复杂的设计。后续所有开发过程请始终遵循以下要求：每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划；代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 定义可能由AI创建的代码文件的扩展名集合\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.css', '.json', '.svg'}\n \n # 用于记录是否扫描到了代码文件，虽然空文件夹不直接违背“所有...编码必须为UTF-8”逻辑，\n # 但作为常规操作，没有扫描到异常文件即认为满足当前要求。\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非AI生成的常见依赖/环境目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n # 尝试用 utf-8 编码格式读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生 UnicodeDecodeError 解码错误，说明文件不是纯合法的 UTF-8 编码\n return False\n except Exception:\n # 其他非编码层面的IO异常则忽略\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线","validation_code":"def check_function_name_format(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 匹配 JavaScript/TypeScript 常见函数定义的正则表达式\n js_patterns = [\n re.compile(r'function\\s+([a-zA-Z0-9_]+)\\s*\\('),\n re.compile(r'(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?function\\b'),\n re.compile(r'(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z0-9_]+)\\s*=>')\n ]\n \n # 匹配 Python 常见函数定义的正则表达式\n py_patterns = [\n re.compile(r'\\bdef\\s+([a-zA-Z0-9_]+)\\s*\\(')\n ]\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖包和Git管理目录\n path_str = root.replace('\\\\', '/')\n if 'node_modules' in path_str or '.git' in path_str:\n continue\n \n for file in files:\n file_path = os.path.join(root, file)\n is_js_ts_html = file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm'))\n is_py = file.endswith('.py')\n \n if not (is_js_ts_html or is_py):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略编码问题或无法读取的文件\n continue\n \n names = []\n if is_js_ts_html:\n for pattern in js_patterns:\n names.extend(pattern.findall(content))\n if is_py:\n for pattern in py_patterns:\n names.extend(pattern.findall(content))\n \n # 校验每一个提取出来的函数名是否只有小写字母和下划线\n for name in names:\n name = name.strip()\n if not name:\n continue\n if not re.fullmatch(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":1,"instruction":"非常好，基础骨架有了。现在我们要赋予它核心的“几何切割”特征。请把原来单一的Z字母骨架删掉，替换为由大约5到7个独立的锐利几何多边形（如锐角三角形、梯形、平行四边形）拼装而成的结构，拼合后的整体轮廓仍是一个完美的“Z”。要求：这些多边形之间必须留出大约3个像素的均匀间隙，以形成凌厉的切割感。给每个多边形分别设置不同角度的线性渐变（linearGradient），渐变色使用具有科技感的青蓝色（Cyan）过渡到深紫色（Purple）。让整个Logo看起来像是由几块发光的切片拼成的立体形态。后续所有开发过程请始终遵循以下更新要求：在你真正动手更改代码之前，务必把准备进行的修改计划写入task_plan.md；生成代码文件时，路径分隔符应当优先采用正斜杠 /。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能由大模型生成的代码文件扩展名\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.json', '.css'}\n found_code_file = False\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 遍历工作区中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅针对代码文件进行编码检测\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制格式读取文件内容\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 utf-8 对字节进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略读取文件可能产生的其他系统级异常\n pass\n\n # 至少存在一个代码文件，并且所有代码文件都可以被 UTF-8 成功解码\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证生成的代码文件内的每个函数名是否仅限于小写字母和下划线。\n \"\"\"\n if not os.path.exists(workspace_path):\n return True\n \n # 定义用于匹配各种常见函数定义的正则表达式\n func_patterns = [\n # Python: def func_name(\n re.compile(r'\\bdef\\s+([a-zA-Z0-9_]+)\\s*\\('),\n # JS/TS/PHP等: function funcName(\n re.compile(r'\\bfunction\\s+([a-zA-Z0-9_]+)\\s*\\('),\n # JS/TS 箭头函数: const funcName = (...) => { ... } \n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z0-9_]+)\\s*=>'),\n # JS/TS 匿名函数赋值: const funcName = async function(\n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?function\\b'),\n # JS/TS 类方法或对象方法简写: methodName() {\n re.compile(r'^[ \\t]*(?:async\\s+)?([a-zA-Z0-9_]+)\\s*\\([^)]*\\)\\s*\\{', re.MULTILINE)\n ]\n \n # 限制分析的文件扩展名\n valid_exts = {'.js', '.ts', '.py', '.html', '.htm', '.jsx', '.tsx', '.vue'}\n # 排除可能被正则误伤的控制流关键字\n reserved_words = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 文件读取失败时跳过，继续检查其他文件\n continue\n \n # 在文件内容中匹配所有的函数声明\n for pattern in func_patterns:\n matches = pattern.findall(content)\n for match in matches:\n func_name = match.strip()\n # 跳过空匹配和保留关键字\n if not func_name or func_name in reserved_words:\n continue\n \n # 检查提取出的函数名是否完全由小写字母和下划线组成\n if not re.match(r'^[a-z_]+$', func_name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入task_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":2,"instruction":"接下来，我们需要进一步增强这个Logo的立体感和视觉层次。请使用SVG的 `<filter>` 标签（结合 `feGaussianBlur` 和 `feOffset` 等），为其中交叠或者边缘的几何碎块添加阴影效果，制造出有的切片在上、有的切片在下的错落高低差错觉。此外，在Z字母的最底部（所有几何色块层级之下），绘制一个扁平的椭圆，并添加较大范围的高斯模糊滤镜，用幽蓝色作为Logo在“地面”上投射的环境光晕或倒影。后续所有开发过程请始终遵循以下更新要求：每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划；限制你生成的代码文件的单行长度，确保每行最多包含100个字符。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目指令，推断可能的代码文件或配置文件的扩展名集合\n code_extensions = {\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx',\n '.css', '.scss', '.less', '.svg',\n '.json', '.md', '.py', '.txt', '.sh'\n }\n \n # 定义需要忽略的无关或第三方依赖目录\n ignore_dirs = {'.git', 'node_modules', '__pycache__', '.venv', 'venv', '.idea', '.vscode'}\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 修改 dirs 以便 os.walk 自动跳过忽略的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理被判定为代码类型的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试将文件内容严格按 utf-8 编码进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码报错，意味着文件使用了非 utf-8 编码（如 GBK 保存了中文字符）\n return False\n except Exception:\n # 如果由于权限等问题无法打开或读取，保守返回 False\n return False\n \n # 所有的代码文件都能够正常使用 UTF-8 进行解码，说明满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n\n # 匹配 JS/TS 等语言中常见的函数命名方式\n # 形式1: function my_func(\n js_ts_func_pattern1 = re.compile(r'function\\s+([a-zA-Z0-9_]+)\\s*\\(')\n # 形式2: const/let/var my_func = function( 或 const/let/var my_func = async function(\n js_ts_func_pattern2 = re.compile(r'(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?function\\s*\\(')\n # 形式3: const/let/var my_func = (...) => 或 const/let/var my_func = async (...) =>\n js_ts_func_pattern3 = re.compile(r'(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z0-9_]+)\\s*=>')\n \n # 匹配 HTML 文件中的 script 标签内容\n html_script_pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n \n # 校验函数名：要求只能限于小写字母和下划线\n valid_name_pattern = re.compile(r'^[a-z_]+$')\n\n def check_js_ts_content(content: str) -> bool:\n \"\"\"检查一段 JS/TS 代码字符串中的所有匹配到的函数名\"\"\"\n funcs = js_ts_func_pattern1.findall(content) + \\\n js_ts_func_pattern2.findall(content) + \\\n js_ts_func_pattern3.findall(content)\n \n for func_name in funcs:\n if not valid_name_pattern.match(func_name):\n return False\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方依赖或版本控制等不需要检查的目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if 'venv' in dirs:\n dirs.remove('venv')\n \n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 1. 检查 Python 文件\n if ext == '.py':\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n source = f.read()\n tree = ast.parse(source)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not valid_name_pattern.match(node.name):\n return False\n except Exception:\n pass\n \n # 2. 检查 JavaScript / TypeScript 文件\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_js_ts_content(content):\n return False\n except Exception:\n pass\n \n # 3. 检查 HTML 文件中内联的脚本\n elif ext in ['.html', '.htm']:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n scripts = html_script_pattern.findall(content)\n for script in scripts:\n if not check_js_ts_content(script):\n return False\n except Exception:\n pass\n\n return True","tags":["Style","Naming"]},{"constraint":"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含100个字符","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据上下文推测可能出现的前端及项目相关代码文件后缀\n code_extensions = {\n '.html', '.htm', '.js', '.jsx', \n '.ts', '.tsx', '.css', '.scss', \n '.json', '.svg'\n }\n \n # 检查工作空间路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除末尾的换行符计算一行的实际纯字符长度\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 100:\n return False\n except Exception:\n # 对于非文本格式导致读取错误的文件（极为罕见）选择跳过\n continue\n \n # 如果遍历所有代码文件都没有发现单行超过100个字符的情况，即为满足要求\n return True","tags":["Quantity","Range"]}]},{"round_id":3,"instruction":"视觉效果看起来很酷了。现在我们给Logo加上互动感，让它“活”起来。请使用CSS给这些切割碎块添加过渡动画（transition）。要求实现：当鼠标悬停（hover）在这个Z字Logo（或整个SVG区域）上时，这些几何切片会沿着各自的中心向四周微微“解体”或“炸开”，产生一种磁力排斥的效果，位移大概在15到30像素之间，个别碎片可以带有微小的旋转（rotate）。当鼠标移开时，所有碎片需平滑、流畅地吸附合拢回原先完美的Z字形状。后续所有开发过程请始终遵循以下更新要求：第一，限制你生成的代码文件的单行长度，确保每行最多包含70个字符；第二，进行文件操作时，必须最大程度地运用专用工具（如Read, Write, Edit）来替代 bash 命令。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能生成的代码文件及相关文件格式\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.json', '.md', '.svg', '.py'\n }\n \n if not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉一些不需要检查的常见无关或第三方库目录\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他的常规 IO 异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线","validation_code":"import os\nimport re\n\ndef check_function_names(response: str, workspace_path: str) -> bool:\n # 辅助函数：移除代码文本中的所有主流注释类型\n def remove_comments(text: str) -> str:\n # 移除 /* ... */ 形式的多行注释\n text = re.sub(r'/\\*[\\s\\S]*?\\*/', '', text)\n # 移除 // 形式的单行注释\n text = re.sub(r'//.*', '', text)\n # 移除 # 形式的单行注释\n text = re.sub(r'#.*', '', text)\n # 移除 <!-- ... --> 形式的 HTML 注释\n text = re.sub(r'<!--[\\s\\S]*?-->', '', text)\n return text\n\n # 正则表达式集合：用于匹配不同语法结构下的函数名定义\n # 模式1: function foo()\n p1 = re.compile(r'\\bfunction\\s+([A-Za-z0-9_]+)\\s*\\(')\n # 模式2: const/let/var foo = function 或者 () => 箭头函数\n p2 = re.compile(r'\\b(?:const|let|var)\\s+([A-Za-z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[A-Za-z0-9_]+\\s*=>)')\n # 模式3: Python形式的 def foo()\n p3 = re.compile(r'\\bdef\\s+([A-Za-z0-9_]+)\\s*\\(')\n # 模式4: ES6 类方法定义（匹配行首可选的修饰符以及其方法名）\n p4 = re.compile(r'^[ \\t]*(?:(?:public|private|protected|static|async|get|set)[ \\t]+)*([A-Za-z0-9_]+)\\s*\\([^)]*\\)\\s*\\{', re.MULTILINE)\n # 模式5: 对象字面量内的方法定义 foo: function 或者箭头函数\n p5 = re.compile(r'\\b([A-Za-z0-9_]+)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[A-Za-z0-9_]+\\s*=>)')\n # 模式6: 对对象属性赋值的函数声明 obj.foo = function 或者箭头函数\n p6 = re.compile(r'\\.\\s*([A-Za-z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[A-Za-z0-9_]+\\s*=>)')\n\n # JavaScript控制流关键字等，避免在匹配类或对象方法时被误认为函数\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', \n 'return', 'yield', 'constructor', 'super', 'import', 'export'\n }\n \n # 遍历当前工作空间中的相关代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.py', '.html', '.vue')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 如果是 HTML / Vue 类型文件，仅提取并合并其 <script> 标签内包含的 JavaScript 代码逻辑\n if file.endswith(('.html', '.vue')):\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n text_to_check = '\\n'.join(scripts)\n else:\n text_to_check = content\n \n # 先行移除代码文本中的注释，避免产生误报\n text_to_check = remove_comments(text_to_check)\n \n if not text_to_check.strip():\n continue\n \n # 收集所有匹配到的函数名称\n names = []\n names.extend(p1.findall(text_to_check))\n names.extend(p2.findall(text_to_check))\n names.extend(p3.findall(text_to_check))\n names.extend(p4.findall(text_to_check))\n names.extend(p5.findall(text_to_check))\n names.extend(p6.findall(text_to_check))\n \n # 开始逐个验证所定义的函数名称是否满足约束\n for name in names:\n if name in keywords:\n continue\n # 检查其内容是否完全被限定为 小写字母 和 下划线 (严禁出现大写字母和数字)\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含70个字符","validation_code":"def verify_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 结合用户指令推断可能生成的代码文件类型\n target_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.svg'}\n has_code_file = False\n \n # 确保路径有效\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行末的换行回车符\n clean_line = line.rstrip('\\r\\n')\n # 要求中明确说明“字符”数，因此使用内置的 len() 函数进行判断\n if len(clean_line) > 70:\n return False\n except Exception:\n # 跳过因特殊情况（如编码错误或权限问题）无法读取的文件\n continue\n \n # 如果项目中没有任何代码文件，则不符合基本产出预期\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"进行文件操作时，必须最大程度地运用专用工具（如Read, Write, Edit）来替代 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":4,"instruction":"最后一步，我们把它升级为一个可定制的动态Logo展示台。请在页面底部（SVG画布下方）增加一个极简风格的控制面板，包含两个控件：\n1. 一个颜色主题切换按钮组（或者下拉菜单），除了当前的“青蓝-深紫”科技主题外，至少再提供一种“粉红-亮黄”的赛博朋克主题，和一种“黑-金”的奢华主题。切换后能实时改变SVG内渐变色的定义。\n2. 一个滑动条（Range Slider），标题为“爆炸张力”，可以让用户手动控制鼠标悬停时碎块散开的最大距离系数。\n请用JavaScript将这些控件与SVG的属性、CSS变量结合起来，要求控件样式也要匹配目前的深色科技感页面风格。\n后续所有开发过程请始终遵循以下更新要求：第一，限制你生成的代码文件的单行长度，确保每行最多包含60个字符；第二，每次对项目进行改动后，都必须更新progress.md文件以记录本轮已完成的内容。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能是代码文件和文本配置文件的扩展名白名单\n code_extensions = {'.html', '.js', '.css', '.json', '.md', '.svg', '.py', '.txt', '.sh'}\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能存在的代码仓库隐藏目录，避免不必要的干扰\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码以检验编码合法性\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明包含非 UTF-8 编码格式的字节\n return False\n except Exception:\n # 遇到其他无法正常读取的问题，保守返回 False\n return False\n\n return True","tags":["Language","File Encoding"]},{"constraint":"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义各类函数的正则表达式模式以捕捉函数名\n py_func_pattern = r'def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\('\n \n # JavaScript / TypeScript / HTML中的<script>函数定义模式\n # 模式1: function foo_bar(\n js_pattern1 = r'function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('\n # 模式2: const foo_bar = function(...) 或 let foo_bar = (...) => 或 var foo_bar = async e =>\n js_pattern2 = r'(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'\n # 模式3: 对象属性定义 foo_bar: function() 或者 foo_bar: () =>\n js_pattern3 = r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?(?:function|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'\n # 模式4: ES6 对象/类方法简写 foo_bar(args) {\n js_pattern4 = r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{'\n # 模式5: 对象/类方法赋值定义 this.foo_bar = () => 或者 obj.foo_bar = function()\n js_pattern5 = r'\\.\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'\n \n # 需忽略的目录（不包含 __tests__，因为指令指明测试代码为生成代码）\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'venv', 'env'}\n # ES6方法正则中需要排除的小写关键词\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'class', 'import', 'export'}\n \n all_functions = []\n \n # 遍历当前工作区路径解析源文件\n for root, dirs, files in os.walk(workspace_path):\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n for file in files:\n if file.endswith(('.py', '.js', '.html', '.vue', '.jsx', '.ts', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 抽取 Python 函数名\n if file.endswith('.py'):\n all_functions.extend(re.findall(py_func_pattern, content))\n \n # 抽取 Web/JS 类文件中的函数名\n if file.endswith(('.js', '.html', '.vue', '.jsx', '.ts', '.tsx')):\n all_functions.extend(re.findall(js_pattern1, content))\n all_functions.extend(re.findall(js_pattern2, content))\n all_functions.extend(re.findall(js_pattern3, content))\n all_functions.extend(re.findall(js_pattern5, content))\n \n raw_pattern4_funcs = re.findall(js_pattern4, content)\n for func in raw_pattern4_funcs:\n if func not in keywords:\n all_functions.append(func)\n except Exception:\n # 屏蔽个别非标准编码文件的读取错误\n pass\n\n # 验证提取出的所有函数名：是否严格由小写字母和下划线组成\n unique_functions = set(all_functions)\n for func in unique_functions:\n if not re.match(r'^[a-z_]+$', func):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"限制你生成的代码文件的单行长度，确保每行最多包含60个字符","validation_code":"import os\n\ndef check_line_length(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，本项目相关的代码文件类型主要为 html, css, js, json\n code_extensions = {'.html', '.js', '.css', '.json'}\n \n # 确保给定的工作区路径存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为推断的代码文件类型\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 按照用户指令要求的UTF-8编码打开文件\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的回车符(\\r)和换行符(\\n)，不影响单行有效字符的计算\n line_content = line.rstrip('\\r\\n')\n \n # 判断该行字符长度是否超过了60个字符\n if len(line_content) > 60:\n return False\n except Exception:\n # 忽略读取过程中的意外错误（如非UTF-8的残留二进制文件等）\n continue\n\n # 遍历完所有相关文件的全部行，没有任何一行超过60个字符\n return True","tags":["Quantity","Range"]},{"constraint":"进行文件操作时，必须最大程度地运用专用工具（如Read, Write, Edit）来替代 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对项目进行改动后，都必须更新progress.md文件以记录本轮已完成的内容","validation_code":"","tags":["Environment","Logging"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我搭建一个简单的HTML页面，里面包含一个SVG画布，用于设计一个Z字母的Logo。页面背景请设为深灰色（#1a1a1a），SVG画布居中对齐，宽高设置为 500x500。在这个SVG里，先用白色的粗线条或一个简单的多边形（polygon）勾勒出一个占满核心区域的大的“Z”字母草图骨架，设定点半透明度即可。目前只需要搭好页面框架、定好Z字母的比例和位置，不需要其他复杂的设计。后续所有开发过程请始终遵循以下要求：每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划；代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 定义可能由AI创建的代码文件的扩展名集合\\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.css', '.json', '.svg'}\\n \\n # 用于记录是否扫描到了代码文件，虽然空文件夹不直接违背“所有...编码必须为UTF-8”逻辑，\\n # 但作为常规操作，没有扫描到异常文件即认为满足当前要求。\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非AI生成的常见依赖/环境目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试用 utf-8 编码格式读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生 UnicodeDecodeError 解码错误，说明文件不是纯合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 其他非编码层面的IO异常则忽略\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线\",\n \"validation_code\": \"def check_function_name_format(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 匹配 JavaScript/TypeScript 常见函数定义的正则表达式\\n js_patterns = [\\n re.compile(r'function\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\('),\\n re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\b'),\\n re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z0-9_]+)\\\\s*=>')\\n ]\\n \\n # 匹配 Python 常见函数定义的正则表达式\\n py_patterns = [\\n re.compile(r'\\\\bdef\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(')\\n ]\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖包和Git管理目录\\n path_str = root.replace('\\\\\\\\', '/')\\n if 'node_modules' in path_str or '.git' in path_str:\\n continue\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n is_js_ts_html = file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm'))\\n is_py = file.endswith('.py')\\n \\n if not (is_js_ts_html or is_py):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略编码问题或无法读取的文件\\n continue\\n \\n names = []\\n if is_js_ts_html:\\n for pattern in js_patterns:\\n names.extend(pattern.findall(content))\\n if is_py:\\n for pattern in py_patterns:\\n names.extend(pattern.findall(content))\\n \\n # 校验每一个提取出来的函数名是否只有小写字母和下划线\\n for name in names:\\n name = name.strip()\\n if not name:\\n continue\\n if not re.fullmatch(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"非常好，基础骨架有了。现在我们要赋予它核心的“几何切割”特征。请把原来单一的Z字母骨架删掉，替换为由大约5到7个独立的锐利几何多边形（如锐角三角形、梯形、平行四边形）拼装而成的结构，拼合后的整体轮廓仍是一个完美的“Z”。要求：这些多边形之间必须留出大约3个像素的均匀间隙，以形成凌厉的切割感。给每个多边形分别设置不同角度的线性渐变（linearGradient），渐变色使用具有科技感的青蓝色（Cyan）过渡到深紫色（Purple）。让整个Logo看起来像是由几块发光的切片拼成的立体形态。后续所有开发过程请始终遵循以下更新要求：在你真正动手更改代码之前，务必把准备进行的修改计划写入task_plan.md；生成代码文件时，路径分隔符应当优先采用正斜杠 /。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能由大模型生成的代码文件扩展名\\n code_extensions = {'.html', '.js', '.ts', '.jsx', '.tsx', '.json', '.css'}\\n found_code_file = False\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历工作区中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅针对代码文件进行编码检测\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制格式读取文件内容\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 utf-8 对字节进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略读取文件可能产生的其他系统级异常\\n pass\\n\\n # 至少存在一个代码文件，并且所有代码文件都可以被 UTF-8 成功解码\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证生成的代码文件内的每个函数名是否仅限于小写字母和下划线。\\n \\\"\\\"\\\"\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义用于匹配各种常见函数定义的正则表达式\\n func_patterns = [\\n # Python: def func_name(\\n re.compile(r'\\\\bdef\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\('),\\n # JS/TS/PHP等: function funcName(\\n re.compile(r'\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\('),\\n # JS/TS 箭头函数: const funcName = (...) => { ... } \\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z0-9_]+)\\\\s*=>'),\\n # JS/TS 匿名函数赋值: const funcName = async function(\\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?function\\\\b'),\\n # JS/TS 类方法或对象方法简写: methodName() {\\n re.compile(r'^[ \\\\t]*(?:async\\\\s+)?([a-zA-Z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', re.MULTILINE)\\n ]\\n \\n # 限制分析的文件扩展名\\n valid_exts = {'.js', '.ts', '.py', '.html', '.htm', '.jsx', '.tsx', '.vue'}\\n # 排除可能被正则误伤的控制流关键字\\n reserved_words = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 文件读取失败时跳过，继续检查其他文件\\n continue\\n \\n # 在文件内容中匹配所有的函数声明\\n for pattern in func_patterns:\\n matches = pattern.findall(content)\\n for match in matches:\\n func_name = match.strip()\\n # 跳过空匹配和保留关键字\\n if not func_name or func_name in reserved_words:\\n continue\\n \\n # 检查提取出的函数名是否完全由小写字母和下划线组成\\n if not re.match(r'^[a-z_]+$', func_name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入task_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来，我们需要进一步增强这个Logo的立体感和视觉层次。请使用SVG的 `<filter>` 标签（结合 `feGaussianBlur` 和 `feOffset` 等），为其中交叠或者边缘的几何碎块添加阴影效果，制造出有的切片在上、有的切片在下的错落高低差错觉。此外，在Z字母的最底部（所有几何色块层级之下），绘制一个扁平的椭圆，并添加较大范围的高斯模糊滤镜，用幽蓝色作为Logo在“地面”上投射的环境光晕或倒影。后续所有开发过程请始终遵循以下更新要求：每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划；限制你生成的代码文件的单行长度，确保每行最多包含100个字符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目指令，推断可能的代码文件或配置文件的扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx',\\n '.css', '.scss', '.less', '.svg',\\n '.json', '.md', '.py', '.txt', '.sh'\\n }\\n \\n # 定义需要忽略的无关或第三方依赖目录\\n ignore_dirs = {'.git', 'node_modules', '__pycache__', '.venv', 'venv', '.idea', '.vscode'}\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 修改 dirs 以便 os.walk 自动跳过忽略的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理被判定为代码类型的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试将文件内容严格按 utf-8 编码进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码报错，意味着文件使用了非 utf-8 编码（如 GBK 保存了中文字符）\\n return False\\n except Exception:\\n # 如果由于权限等问题无法打开或读取，保守返回 False\\n return False\\n \\n # 所有的代码文件都能够正常使用 UTF-8 进行解码，说明满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 匹配 JS/TS 等语言中常见的函数命名方式\\n # 形式1: function my_func(\\n js_ts_func_pattern1 = re.compile(r'function\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(')\\n # 形式2: const/let/var my_func = function( 或 const/let/var my_func = async function(\\n js_ts_func_pattern2 = re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?function\\\\s*\\\\(')\\n # 形式3: const/let/var my_func = (...) => 或 const/let/var my_func = async (...) =>\\n js_ts_func_pattern3 = re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z0-9_]+)\\\\s*=>')\\n \\n # 匹配 HTML 文件中的 script 标签内容\\n html_script_pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n \\n # 校验函数名：要求只能限于小写字母和下划线\\n valid_name_pattern = re.compile(r'^[a-z_]+$')\\n\\n def check_js_ts_content(content: str) -> bool:\\n \\\"\\\"\\\"检查一段 JS/TS 代码字符串中的所有匹配到的函数名\\\"\\\"\\\"\\n funcs = js_ts_func_pattern1.findall(content) + \\\\\\n js_ts_func_pattern2.findall(content) + \\\\\\n js_ts_func_pattern3.findall(content)\\n \\n for func_name in funcs:\\n if not valid_name_pattern.match(func_name):\\n return False\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方依赖或版本控制等不需要检查的目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'venv' in dirs:\\n dirs.remove('venv')\\n \\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 1. 检查 Python 文件\\n if ext == '.py':\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n source = f.read()\\n tree = ast.parse(source)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not valid_name_pattern.match(node.name):\\n return False\\n except Exception:\\n pass\\n \\n # 2. 检查 JavaScript / TypeScript 文件\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_js_ts_content(content):\\n return False\\n except Exception:\\n pass\\n \\n # 3. 检查 HTML 文件中内联的脚本\\n elif ext in ['.html', '.htm']:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n scripts = html_script_pattern.findall(content)\\n for script in scripts:\\n if not check_js_ts_content(script):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含100个字符\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据上下文推测可能出现的前端及项目相关代码文件后缀\\n code_extensions = {\\n '.html', '.htm', '.js', '.jsx', \\n '.ts', '.tsx', '.css', '.scss', \\n '.json', '.svg'\\n }\\n \\n # 检查工作空间路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除末尾的换行符计算一行的实际纯字符长度\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 100:\\n return False\\n except Exception:\\n # 对于非文本格式导致读取错误的文件（极为罕见）选择跳过\\n continue\\n \\n # 如果遍历所有代码文件都没有发现单行超过100个字符的情况，即为满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"视觉效果看起来很酷了。现在我们给Logo加上互动感，让它“活”起来。请使用CSS给这些切割碎块添加过渡动画（transition）。要求实现：当鼠标悬停（hover）在这个Z字Logo（或整个SVG区域）上时，这些几何切片会沿着各自的中心向四周微微“解体”或“炸开”，产生一种磁力排斥的效果，位移大概在15到30像素之间，个别碎片可以带有微小的旋转（rotate）。当鼠标移开时，所有碎片需平滑、流畅地吸附合拢回原先完美的Z字形状。后续所有开发过程请始终遵循以下更新要求：第一，限制你生成的代码文件的单行长度，确保每行最多包含70个字符；第二，进行文件操作时，必须最大程度地运用专用工具（如Read, Write, Edit）来替代 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能生成的代码文件及相关文件格式\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.json', '.md', '.svg', '.py'\\n }\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉一些不需要检查的常见无关或第三方库目录\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他的常规 IO 异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_function_names(response: str, workspace_path: str) -> bool:\\n # 辅助函数：移除代码文本中的所有主流注释类型\\n def remove_comments(text: str) -> str:\\n # 移除 /* ... */ 形式的多行注释\\n text = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', text)\\n # 移除 // 形式的单行注释\\n text = re.sub(r'//.*', '', text)\\n # 移除 # 形式的单行注释\\n text = re.sub(r'#.*', '', text)\\n # 移除 <!-- ... --> 形式的 HTML 注释\\n text = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', text)\\n return text\\n\\n # 正则表达式集合：用于匹配不同语法结构下的函数名定义\\n # 模式1: function foo()\\n p1 = re.compile(r'\\\\bfunction\\\\s+([A-Za-z0-9_]+)\\\\s*\\\\(')\\n # 模式2: const/let/var foo = function 或者 () => 箭头函数\\n p2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([A-Za-z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[A-Za-z0-9_]+\\\\s*=>)')\\n # 模式3: Python形式的 def foo()\\n p3 = re.compile(r'\\\\bdef\\\\s+([A-Za-z0-9_]+)\\\\s*\\\\(')\\n # 模式4: ES6 类方法定义（匹配行首可选的修饰符以及其方法名）\\n p4 = re.compile(r'^[ \\\\t]*(?:(?:public|private|protected|static|async|get|set)[ \\\\t]+)*([A-Za-z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', re.MULTILINE)\\n # 模式5: 对象字面量内的方法定义 foo: function 或者箭头函数\\n p5 = re.compile(r'\\\\b([A-Za-z0-9_]+)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[A-Za-z0-9_]+\\\\s*=>)')\\n # 模式6: 对对象属性赋值的函数声明 obj.foo = function 或者箭头函数\\n p6 = re.compile(r'\\\\.\\\\s*([A-Za-z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[A-Za-z0-9_]+\\\\s*=>)')\\n\\n # JavaScript控制流关键字等，避免在匹配类或对象方法时被误认为函数\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', \\n 'return', 'yield', 'constructor', 'super', 'import', 'export'\\n }\\n \\n # 遍历当前工作空间中的相关代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.py', '.html', '.vue')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 如果是 HTML / Vue 类型文件，仅提取并合并其 <script> 标签内包含的 JavaScript 代码逻辑\\n if file.endswith(('.html', '.vue')):\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n text_to_check = '\\\\n'.join(scripts)\\n else:\\n text_to_check = content\\n \\n # 先行移除代码文本中的注释，避免产生误报\\n text_to_check = remove_comments(text_to_check)\\n \\n if not text_to_check.strip():\\n continue\\n \\n # 收集所有匹配到的函数名称\\n names = []\\n names.extend(p1.findall(text_to_check))\\n names.extend(p2.findall(text_to_check))\\n names.extend(p3.findall(text_to_check))\\n names.extend(p4.findall(text_to_check))\\n names.extend(p5.findall(text_to_check))\\n names.extend(p6.findall(text_to_check))\\n \\n # 开始逐个验证所定义的函数名称是否满足约束\\n for name in names:\\n if name in keywords:\\n continue\\n # 检查其内容是否完全被限定为 小写字母 和 下划线 (严禁出现大写字母和数字)\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含70个字符\",\n \"validation_code\": \"def verify_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 结合用户指令推断可能生成的代码文件类型\\n target_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.svg'}\\n has_code_file = False\\n \\n # 确保路径有效\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行末的换行回车符\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 要求中明确说明“字符”数，因此使用内置的 len() 函数进行判断\\n if len(clean_line) > 70:\\n return False\\n except Exception:\\n # 跳过因特殊情况（如编码错误或权限问题）无法读取的文件\\n continue\\n \\n # 如果项目中没有任何代码文件，则不符合基本产出预期\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"进行文件操作时，必须最大程度地运用专用工具（如Read, Write, Edit）来替代 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后一步，我们把它升级为一个可定制的动态Logo展示台。请在页面底部（SVG画布下方）增加一个极简风格的控制面板，包含两个控件：\\n1. 一个颜色主题切换按钮组（或者下拉菜单），除了当前的“青蓝-深紫”科技主题外，至少再提供一种“粉红-亮黄”的赛博朋克主题，和一种“黑-金”的奢华主题。切换后能实时改变SVG内渐变色的定义。\\n2. 一个滑动条（Range Slider），标题为“爆炸张力”，可以让用户手动控制鼠标悬停时碎块散开的最大距离系数。\\n请用JavaScript将这些控件与SVG的属性、CSS变量结合起来，要求控件样式也要匹配目前的深色科技感页面风格。\\n后续所有开发过程请始终遵循以下更新要求：第一，限制你生成的代码文件的单行长度，确保每行最多包含60个字符；第二，每次对项目进行改动后，都必须更新progress.md文件以记录本轮已完成的内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"目标代码的语句覆盖率必须通过运行你生成的单元测试文件达到下限95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你在生成代码时产生的所有中间测试代码必须被保留，且需放置于项目目录下的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能是代码文件和文本配置文件的扩展名白名单\\n code_extensions = {'.html', '.js', '.css', '.json', '.md', '.svg', '.py', '.txt', '.sh'}\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能存在的代码仓库隐藏目录，避免不必要的干扰\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码以检验编码合法性\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明包含非 UTF-8 编码格式的字节\\n return False\\n except Exception:\\n # 遇到其他无法正常读取的问题，保守返回 False\\n return False\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"所生成代码文件内的每个函数名，其字符内容只能限于小写字母和下划线\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义各类函数的正则表达式模式以捕捉函数名\\n py_func_pattern = r'def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\('\\n \\n # JavaScript / TypeScript / HTML中的<script>函数定义模式\\n # 模式1: function foo_bar(\\n js_pattern1 = r'function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('\\n # 模式2: const foo_bar = function(...) 或 let foo_bar = (...) => 或 var foo_bar = async e =>\\n js_pattern2 = r'(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'\\n # 模式3: 对象属性定义 foo_bar: function() 或者 foo_bar: () =>\\n js_pattern3 = r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'\\n # 模式4: ES6 对象/类方法简写 foo_bar(args) {\\n js_pattern4 = r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n # 模式5: 对象/类方法赋值定义 this.foo_bar = () => 或者 obj.foo_bar = function()\\n js_pattern5 = r'\\\\.\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'\\n \\n # 需忽略的目录（不包含 __tests__，因为指令指明测试代码为生成代码）\\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'venv', 'env'}\\n # ES6方法正则中需要排除的小写关键词\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'class', 'import', 'export'}\\n \\n all_functions = []\\n \\n # 遍历当前工作区路径解析源文件\\n for root, dirs, files in os.walk(workspace_path):\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n for file in files:\\n if file.endswith(('.py', '.js', '.html', '.vue', '.jsx', '.ts', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 抽取 Python 函数名\\n if file.endswith('.py'):\\n all_functions.extend(re.findall(py_func_pattern, content))\\n \\n # 抽取 Web/JS 类文件中的函数名\\n if file.endswith(('.js', '.html', '.vue', '.jsx', '.ts', '.tsx')):\\n all_functions.extend(re.findall(js_pattern1, content))\\n all_functions.extend(re.findall(js_pattern2, content))\\n all_functions.extend(re.findall(js_pattern3, content))\\n all_functions.extend(re.findall(js_pattern5, content))\\n \\n raw_pattern4_funcs = re.findall(js_pattern4, content)\\n for func in raw_pattern4_funcs:\\n if func not in keywords:\\n all_functions.append(func)\\n except Exception:\\n # 屏蔽个别非标准编码文件的读取错误\\n pass\\n\\n # 验证提取出的所有函数名：是否严格由小写字母和下划线组成\\n unique_functions = set(all_functions)\\n for func in unique_functions:\\n if not re.match(r'^[a-z_]+$', func):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"若对生成的代码完成了测试，就要去更新test_results.json文件。要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次调整代码前必须执行的步骤是：先在implementation_plan.json内写下接下来的修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"代码实现应当遵循最短路径原则，禁止过度设计及超纲实现（如引入未经要求的兜底机制或降级方案）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件的单行长度，确保每行最多包含60个字符\",\n \"validation_code\": \"import os\\n\\ndef check_line_length(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，本项目相关的代码文件类型主要为 html, css, js, json\\n code_extensions = {'.html', '.js', '.css', '.json'}\\n \\n # 确保给定的工作区路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为推断的代码文件类型\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 按照用户指令要求的UTF-8编码打开文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的回车符(\\\\r)和换行符(\\\\n)，不影响单行有效字符的计算\\n line_content = line.rstrip('\\\\r\\\\n')\\n \\n # 判断该行字符长度是否超过了60个字符\\n if len(line_content) > 60:\\n return False\\n except Exception:\\n # 忽略读取过程中的意外错误（如非UTF-8的残留二进制文件等）\\n continue\\n\\n # 遍历完所有相关文件的全部行，没有任何一行超过60个字符\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"进行文件操作时，必须最大程度地运用专用工具（如Read, Write, Edit）来替代 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对项目进行改动后，都必须更新progress.md文件以记录本轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查页面整体布局与背景：截取整个网页，确认页面背景色为深灰色，并且中间包含一个居中对齐的SVG画布区域（尺寸约为500x500）。","检查Logo基础形态与切割特征：观察画面中心的Logo，确认其整体轮廓是字母'Z'，且由约5到7个独立的锐利多边形（如三角形、梯形、平行四边形等几何切片）拼装而成，切片之间留有约3像素的均匀缝隙。","检查默认渐变颜色：检查页面初始加载时，拼成Z字母的各个几何切片是否都使用线性渐变填充，渐变色为青蓝色到深紫色，并且不同切片的渐变方向/角度存在差异。","检查切片立体阴影效果：观察切片交叠处或边缘，确认切片带有阴影效果（产生于SVG滤镜），视觉上呈现出切片前后错落的高低层次感。","检查底部环境光晕：观察Z字母的最底部正下方区域，确认是否存在一个幽蓝色的扁平椭圆，且带有明显的高斯模糊效果，充当环境光晕或倒影。","检查Hover解体与恢复动画：模拟鼠标悬停在Z字母Logo区域，确认几何切片沿各自中心向四周平滑散开，位移大致在15到30像素范围内，部分碎片可带有轻微旋转，产生解体/磁力排斥的效果；随后将鼠标移开，确认切片能平滑地聚拢吸附，恢复完美的'Z'字形状。","检查控制面板UI展示：观察页面底部，确认存在一个控制面板，其中包含一个颜色主题切换控件（包含青蓝-深紫、粉红-亮黄、黑-金等选项）和一个标题包含'爆炸张力'的滑动条，且控件样式与深色背景协调。","验证'粉红-亮黄'主题切换交互：操作主题切换控件选中'粉红-亮黄'主题（或赛博朋克主题），确认Logo内所有切片的渐变色立刻变为粉红到亮黄的过渡。","验证'黑-金'主题切换交互：操作主题切换控件选中'黑-金'主题（或奢华主题），确认Logo内所有切片的渐变色立刻变为黑色到金色的过渡。","验证'爆炸张力'滑动条交互：将'爆炸张力'滑动条向右拖动调大数值，再次模拟鼠标悬停在Logo上，通过对比确认此时切片向四周炸开/散开的位移距离比默认状态下更远。"],"string":"[\n \"检查页面整体布局与背景：截取整个网页，确认页面背景色为深灰色，并且中间包含一个居中对齐的SVG画布区域（尺寸约为500x500）。\",\n \"检查Logo基础形态与切割特征：观察画面中心的Logo，确认其整体轮廓是字母'Z'，且由约5到7个独立的锐利多边形（如三角形、梯形、平行四边形等几何切片）拼装而成，切片之间留有约3像素的均匀缝隙。\",\n \"检查默认渐变颜色：检查页面初始加载时，拼成Z字母的各个几何切片是否都使用线性渐变填充，渐变色为青蓝色到深紫色，并且不同切片的渐变方向/角度存在差异。\",\n \"检查切片立体阴影效果：观察切片交叠处或边缘，确认切片带有阴影效果（产生于SVG滤镜），视觉上呈现出切片前后错落的高低层次感。\",\n \"检查底部环境光晕：观察Z字母的最底部正下方区域，确认是否存在一个幽蓝色的扁平椭圆，且带有明显的高斯模糊效果，充当环境光晕或倒影。\",\n \"检查Hover解体与恢复动画：模拟鼠标悬停在Z字母Logo区域，确认几何切片沿各自中心向四周平滑散开，位移大致在15到30像素范围内，部分碎片可带有轻微旋转，产生解体/磁力排斥的效果；随后将鼠标移开，确认切片能平滑地聚拢吸附，恢复完美的'Z'字形状。\",\n \"检查控制面板UI展示：观察页面底部，确认存在一个控制面板，其中包含一个颜色主题切换控件（包含青蓝-深紫、粉红-亮黄、黑-金等选项）和一个标题包含'爆炸张力'的滑动条，且控件样式与深色背景协调。\",\n \"验证'粉红-亮黄'主题切换交互：操作主题切换控件选中'粉红-亮黄'主题（或赛博朋克主题），确认Logo内所有切片的渐变色立刻变为粉红到亮黄的过渡。\",\n \"验证'黑-金'主题切换交互：操作主题切换控件选中'黑-金'主题（或奢华主题），确认Logo内所有切片的渐变色立刻变为黑色到金色的过渡。\",\n \"验证'爆炸张力'滑动条交互：将'爆炸张力'滑动条向右拖动调大数值，再次模拟鼠标悬停在Logo上，通过对比确认此时切片向四周炸开/散开的位移距离比默认状态下更远。\"\n]","isConversational":false}}},{"rowIdx":56,"cells":{"task_id":{"kind":"number","value":57,"string":"57"},"repository_policy":{"kind":"string","value":"你在后续所有轮次的操作都需要严格遵循以下约束：\n1. 进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤。\n2. 凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读。\n3. 只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换。\n4. 务必使用UTF-8编码你所创建的全部代码文件。\n5. 确保每一个生成的代码文件里，注释部分完全使用中文。\n6. 请确保在所有情况下都使用中文来生成用户回复。\n7. 代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改。\n8. 在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码相关文件的后缀集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts', \n '.html', '.css', '.php', '.rb', '.go', '.rs', '.swift', '.kt', \n '.sh', '.bat', '.json', '.xml', '.yaml', '.yml', '.sql', '.m', \n '.vue', '.jsx', '.tsx', '.dart', '.scala', '.lua', '.r', '.jl'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和常见的非源码依赖/编译目录\n dirs[:] = [\n d for d in dirs \n if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env', 'dist', 'build')\n ]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 格式读取文件全部内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 只要有任何文件不能被 UTF-8 解码，说明包含了非 UTF-8 编码字符\n return False\n except Exception:\n # 忽略权限等其他读取问题\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]}],"string":"[\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码相关文件的后缀集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts', \\n '.html', '.css', '.php', '.rb', '.go', '.rs', '.swift', '.kt', \\n '.sh', '.bat', '.json', '.xml', '.yaml', '.yml', '.sql', '.m', \\n '.vue', '.jsx', '.tsx', '.dart', '.scala', '.lua', '.r', '.jl'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和常见的非源码依赖/编译目录\\n dirs[:] = [\\n d for d in dirs \\n if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env', 'dist', 'build')\\n ]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 格式读取文件全部内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 只要有任何文件不能被 UTF-8 解码，说明包含了非 UTF-8 编码字符\\n return False\\n except Exception:\\n # 忽略权限等其他读取问题\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用React开发一个运动计时器和间歇训练器的网页应用。首先，我们需要搭建基本的页面骨架。请创建一个包含底部导航栏的应用框架，导航栏有三个Tab：'训练'、'数据'和'社交'。默认显示'训练'页面，在这个页面里先只放一个最基础的数字秒表（显示00:00.00），以及'开始'和'暂停'两个按钮就行。暂时不需要真正的路由配置，用组件状态来切换底部Tab即可。帮我把这个骨架搭好。另外请你在后续所有开发过程中始终遵循以下要求：一是你生成的每一个代码文件都精确包含3行注释，二是确保你输出的代码项目文件夹嵌套深度不超过2层。","instruction_following_checklist":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令开发React应用的要求，推断出常见的代码文件后缀\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n code_files = []\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除掉可能存在的依赖或版本控制等非直接生成的目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n code_files.append(os.path.join(root, file))\n \n # 如果没有找到任何代码文件，则算作未完成代码生成任务，无法满足该编码要求\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦出现无法以 UTF-8 解码的文件，说明未满足要求\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每一个代码文件都精确包含3行注释","validation_code":"import os\nimport re\nimport bisect\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的前端相关代码文件类型\n EXTENSIONS = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.htm', '.scss', '.less'}\n \n # 构建正则表达式：\n # 第一部分：匹配字符串内容（避免将字符串中的 // 等当成注释）\n # 第二部分：匹配单行注释 // ...\n # 第三部分：匹配多行块注释 /* ... */\n # 第四部分：匹配 HTML 注释 <!-- ... -->\n pattern = re.compile(\n r'(?P<string>\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\'|`(?:[^`\\\\]|\\\\.)*`)'\n r'|(?P<sline>//[^\\n]*)'\n r'|(?P<mline>/\\*[\\s\\S]*?\\*/)'\n r'|(?P<html_comment><!--[\\s\\S]*?-->)'\n )\n \n code_files = []\n # 遍历工作区目录，收集所有生成的代码文件\n for root, dirs, files in os.walk(workspace_path):\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in EXTENSIONS:\n code_files.append(os.path.join(root, file))\n \n # 如果没有找到任何代码文件，说明并未生成，要求未满足\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 记录所有换行符的索引位置，便于后续计算绝对字符位置所属的行号\n nl_positions = [m.start() for m in re.finditer(r'\\n', content)]\n \n def get_line_number(index):\n # 利用二分查找获取当前字符所在的行号 (从 0 开始)\n return bisect.bisect_right(nl_positions, index)\n \n comment_line_indices = set()\n \n # 匹配文本中的注释\n for m in pattern.finditer(content):\n group = m.lastgroup\n # 只统计属于注释的分组\n if group in ('sline', 'mline', 'html_comment'):\n start_index = m.start(group)\n # m.end(group) 是结束后的索引，因此需减去1来获取真实的最后包含的字符位置\n end_index = m.end(group) - 1 if m.end(group) > m.start(group) else start_index\n \n start_line = get_line_number(start_index)\n end_line = get_line_number(end_index)\n \n # 将该注释跨越的所有行号加入集合\n for line in range(start_line, end_line + 1):\n comment_line_indices.add(line)\n \n # 每一个代码文件的注释行数必须精确等于3行\n if len(comment_line_indices) != 3:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过2层","validation_code":"def check_folder_depth(response: str, workspace_path: str) -> bool:\n import os\n \n # 若工作区路径不存在，可视为未创建任何多余嵌套文件夹，符合条件\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非AI直接编写的代码目录，如隐藏文件夹、编译产物和第三方依赖包目录\n # 修改 dirs[:] 能够直接影响 os.walk 今后的遍历行为，避免深入不必要的目录\n dirs[:] = [d for d in dirs if d not in ['node_modules', 'build', 'dist', '__pycache__'] and not d.startswith('.')]\n \n # 计算当前所在文件夹相对于 workspace_path 的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 项目根目录相对路径为 '.'，深度记为0\n if rel_path == '.':\n depth = 0\n else:\n # 将路径标准化，并通过 os.sep 分割以计算当前目录的实际嵌套层级\n rel_path = os.path.normpath(rel_path)\n depth = len(rel_path.split(os.sep))\n \n # 判定：若检测到文件夹的嵌套深度大于2层，则说明不满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]}]},{"round_id":1,"instruction":"好，接下来重点完善核心的'训练'页面。我要能在三种模式之间自由切换：'普通秒表'、'倒计时'和'间歇训练(HIIT)'。对于最复杂的'间歇训练'模式，页面上必须包含一个设置表单，让我可以自定义调整：准备时间、单次工作时长、单次休息时长，以及循环的组数。点击开始后，倒计时数字下方要清晰地显示当前处于'准备'、'运动中'还是'休息'状态，并且界面主背景色要根据状态平滑切换（比如准备时黄色，运动时红色，休息时绿色），方便运动时瞟一眼就能确认状态。另外从本轮开始后续所有开发过程中请遵循两项调整后的要求：一是针对每一个生成的代码文件，注释的行数需精准控制在5行，二是生成的代码项目最多只能包含3个目录层级，原有的对应旧要求作废，请严格执行。","instruction_following_checklist":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据React应用推断常见代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名并转为小写进行匹配\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制形式读取文件内容\n with open(file_path, 'rb') as f:\n file_data = f.read()\n # 尝试以UTF-8解码，若成功则编码无误\n file_data.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦捕获到解码错误，说明存在非UTF-8编码的代码文件\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在5行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见前端业务和样式代码的扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.scss', '.less'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 跳过常见的无关目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n comment_lines = 0\n \n # HTML文件注释解析\n if ext in ['.html']:\n in_comment = False\n for line in code.split('\\n'):\n i = 0\n is_comment_line = in_comment\n while i < len(line):\n if in_comment:\n if line[i:i+3] == '-->':\n in_comment = False\n i += 2\n else:\n if line[i:i+4] == '<!--':\n in_comment = True\n is_comment_line = True\n i += 3\n i += 1\n if is_comment_line:\n comment_lines += 1\n # JS/TS/CSS 类文件的注释解析\n else:\n in_string = False\n string_char = ''\n in_block_comment = False\n \n for line in code.split('\\n'):\n i = 0\n is_comment_line = in_block_comment\n \n while i < len(line):\n if in_block_comment:\n if line[i:i+2] == '*/':\n in_block_comment = False\n i += 1\n elif in_string:\n if line[i] == '\\\\':\n i += 1 # 忽略转义字符\n elif line[i] == string_char:\n in_string = False\n else:\n if line[i:i+2] == '//':\n is_comment_line = True\n break\n elif line[i:i+2] == '/*':\n in_block_comment = True\n is_comment_line = True\n i += 1\n elif line[i] in '\"\\'`':\n in_string = True\n string_char = line[i]\n i += 1\n if is_comment_line:\n comment_lines += 1\n \n # 若发现任意代码文件的注释行数不为 5，则未满足要求\n if comment_lines != 5:\n return False\n \n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码项目最多只能包含3个目录层级","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 在原地修改dirs以控制os.walk的后续遍历行为\n # 忽略隐藏目录以及可能由前端环境和系统自动生成的较深且非代码的目录\n dirs[:] = [\n d for d in dirs \n if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'build', 'dist')\n ]\n \n # 计算相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n if rel_path == '.':\n # 工作空间根目录算作第1层级\n depth = 1\n else:\n # 统一替换路径分隔符以兼容不同操作系统，随后进行拆分计算\n normalized_rel_path = rel_path.replace('\\\\', '/')\n parts = [p for p in normalized_rel_path.split('/') if p]\n # 根目录的1层级 + 相对路径包含的子层级数\n depth = len(parts) + 1\n \n # 目录层级超过了3层\n if depth > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]}]},{"round_id":2,"instruction":"现在的计时功能有了，来优化一下间歇训练时的视觉反馈。除了背景变色，我希望加上一个充满科技感的SVG环形进度条，把时间倒计时的数字包围在中间。这个环形进度条要能根据当前阶段（工作或休息）的总时长和剩余时间实时动态递减。另外，为了增加紧迫感，当任意阶段倒计时只剩最后3秒时，中间的数字要出现明显的放大缩小脉冲动画（类似心跳效果），以此来模拟倒数冲刺的紧迫感。从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是针对每一个生成的代码文件，注释的行数需精准控制在6行，原有的对应注释行数旧要求作废；二是你输出的每一个代码文件，连同空行在内的总行数必须是奇数，请严格执行。","instruction_following_checklist":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React等网页开发常见代码文件的后缀名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试以UTF-8编码解码文件内容\n try:\n with open(file_path, 'rb') as f:\n raw_content = f.read()\n raw_content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码失败，说明不是有效的UTF-8编码\n return False\n except Exception:\n # 读取遇到其他异常情况视为未满足\n return False\n\n # 如果存在代码文件且全都通过了UTF-8验证则返回True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在6行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n def count_comment_lines(filepath: str) -> int:\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n text = f.read()\n\n ext = filepath.split('.')[-1].lower()\n \n in_single_quote = False\n in_double_quote = False\n in_backtick = False\n in_multi_line_comment = False\n \n comment_lines = set()\n line_num = 1\n \n i = 0\n n = len(text)\n \n while i < n:\n if text[i] == '\\n':\n line_num += 1\n # 容错：遇到换行后如果处于普通字符串中，重置以防不规范代码导致误判\n if in_single_quote: in_single_quote = False\n if in_double_quote: in_double_quote = False\n i += 1\n continue\n \n if ext in ['html', 'htm']:\n if in_multi_line_comment:\n comment_lines.add(line_num)\n if text[i:i+3] == '-->':\n in_multi_line_comment = False\n i += 3\n else:\n i += 1\n else:\n if text[i:i+4] == '<!--':\n in_multi_line_comment = True\n comment_lines.add(line_num)\n i += 4\n else:\n i += 1\n else:\n if in_multi_line_comment:\n comment_lines.add(line_num)\n if text[i:i+2] == '*/':\n in_multi_line_comment = False\n i += 2\n else:\n i += 1\n continue\n \n if in_single_quote:\n if text[i] == '\\\\':\n i += 2\n elif text[i] == \"'\":\n in_single_quote = False\n i += 1\n else:\n i += 1\n continue\n \n if in_double_quote:\n if text[i] == '\\\\':\n i += 2\n elif text[i] == '\"':\n in_double_quote = False\n i += 1\n else:\n i += 1\n continue\n \n if in_backtick:\n if text[i] == '\\\\':\n i += 2\n elif text[i] == '`':\n in_backtick = False\n i += 1\n else:\n i += 1\n continue\n \n if text[i] == \"'\":\n in_single_quote = True\n i += 1\n continue\n \n if text[i] == '\"':\n in_double_quote = True\n i += 1\n continue\n \n if text[i] == '`':\n in_backtick = True\n i += 1\n continue\n \n if text[i:i+2] == '//':\n comment_lines.add(line_num)\n # 遇到单行注释直接往后跳过直到行尾，不包含 \\n 以便让上层处理递增行号\n while i < n and text[i] != '\\n':\n i += 1\n continue\n \n if text[i:i+2] == '/*':\n in_multi_line_comment = True\n comment_lines.add(line_num)\n i += 2\n continue\n \n i += 1\n\n return len(comment_lines)\n\n if not os.path.isdir(workspace_path):\n return False\n \n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.htm'}\n code_files = []\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的目录\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', 'build', 'dist']]\n for f in files:\n if any(f.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, f))\n \n if not code_files:\n return False\n \n # 验证是否每一个代码文件的注释行数都精确等于6\n for file_path in code_files:\n if count_comment_lines(file_path) != 6:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码项目最多只能包含3个目录层级","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果项目路径不存在，无法判断，返回False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏文件夹（如.git等），这些通常不计入生成的代码项目的层级中\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n rel_path = os.path.relpath(root, workspace_path)\n if rel_path == '.':\n depth = 0\n else:\n # 通过操作系统的路径分隔符拆分，计算当前目录相对于根目录的深度\n depth = len(rel_path.split(os.sep))\n \n # 如果目录层级大于3层，即违反了“最多只能包含3个目录层级”的约束\n if depth > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"你输出的每一个代码文件，连同空行在内的总行数必须是奇数","validation_code":"def check_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端React项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的第三方依赖或版本控制目录，避免检查非生成的代码文件\n path_parts = root.split(os.sep)\n if 'node_modules' in path_parts or '.git' in path_parts:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 使用 split('\\n') 能够严格、准确地统计出包含所有空行在内的行数\n lines = content.split('\\n')\n \n # 如果行数是偶数，说明不符合“总行数必须是奇数”的要求\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果由于编码等原因无法读取，则跳过\n continue\n \n # 所有被成功读取的代码文件的行数均为奇数\n return True","tags":["Quantity","Exact Value"]}]},{"round_id":3,"instruction":"下一步是做数据追踪。当我在任意模式下点击'结束训练'时，将这次的训练数据（模式名称、总时长、完成组数等）保存到 localStorage 中。然后帮我完善底部的'数据'Tab页面：上半部分做一个数据看板，显示'累计运动时长'、'总完成组数'和'预估消耗千卡'；下半部分用简易的柱状图（可以直接用不同高度带颜色的div元素模拟）展示最近7天每天的运动时长变化趋势。最下方再列出最近5条的详细训练历史记录卡片。另外从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是你输出的每一个代码文件，连同空行在内的总行数必须是偶数，原有的对应总行数旧要求作废；二是请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号，请严格执行。","instruction_following_checklist":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React项目开发中常见的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.json', '.md'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖库和版本控制等不需要检查的目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅针对推断出的可能代码文件进行UTF-8编码检查\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(filepath, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦捕获到解码错误，说明存在非UTF-8编码的代码文件\n return False\n except Exception:\n # 忽略读取过程中的其他无关异常（如权限问题等）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在6行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义属于React项目前端代码文件的扩展名集合\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n file_count = 0\n all_meet = True\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非代码源码目录\n for ignore_dir in ['node_modules', '.git', 'build', 'dist']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_count += 1\n file_path = os.path.join(root, file)\n \n comment_lines = 0\n in_block_comment = False\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n stripped = line.strip()\n \n # 处理空行\n if not stripped:\n # 只有在多行注释块内的空行才计作注释行\n if in_block_comment:\n comment_lines += 1\n continue\n \n # 处理仍处于多行注释块内的行\n if in_block_comment:\n comment_lines += 1\n if ext == '.html':\n if '-->' in stripped:\n in_block_comment = False\n else:\n if '*/' in stripped:\n in_block_comment = False\n continue\n \n # 处理尚未处于多行注释块内的行\n if ext == '.html':\n if '<!--' in stripped:\n comment_lines += 1\n if '-->' not in stripped:\n in_block_comment = True\n else:\n if '/*' in stripped:\n comment_lines += 1\n if '*/' not in stripped:\n in_block_comment = True\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n # 剔除常见的URL前缀以防误判单行注释\n tmp = stripped.replace('http://', '').replace('https://', '')\n # 判断是否包含单行注释\n if tmp.startswith('//') or ' //' in tmp or '\\t//' in tmp:\n comment_lines += 1\n \n if comment_lines != 6:\n all_meet = False\n break\n if not all_meet:\n break\n \n # 如果项目中没有任何指定的代码文件，则不满足要求\n if file_count == 0:\n return False\n \n return all_meet","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码项目最多只能包含3个目录层级","validation_code":"import os\n\ndef check_directory_depth(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和特定的非代码生成目录，避免误判并且提高遍历效率\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n rel_path = os.path.relpath(root, workspace_path)\n if rel_path == '.':\n depth = 1\n else:\n # 将路径分隔符统一替换为 '/' 以防跨平台差异，然后计算层级深度\n normalized_rel_path = rel_path.replace('\\\\', '/')\n depth = 1 + len(normalized_rel_path.split('/'))\n \n # 如果发现任何深度大于3的目录，说明违反了限制要求\n if depth > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"你输出的每一个代码文件，连同空行在内的总行数必须是偶数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件后缀类型\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css'}\n # 忽略扫描的目录\n exclude_dirs = {'node_modules', '.git', 'build', 'dist'}\n \n found_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以便 os.walk 忽略指定文件夹\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按换行符分割以精确计算编辑器中展示的总行数(包含空行)\n lines = content.split('\\n')\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 对于不可预期的文件读取异常不应直接让函数崩溃，但可视情况跳过\n continue\n \n # 如果找到了至少一个代码文件且全部符合偶数行要求，则判定通过\n return found_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 预定义运动、健身、计时相关的emoji集合（包含基础字符形式，兼容复合emoji匹配）\n SPORTS_EMOJIS = {\n \"⚽\", \"🏀\", \"🏈\", \"⚾\", \"🥎\", \"🎾\", \"🏐\", \"🏉\", \"🥏\", \"🎱\",\n \"🪀\", \"🏓\", \"🏸\", \"🏒\", \"🏑\", \"🥍\", \"🏏\", \"🪃\", \"🥅\", \"⛳\",\n \"🪁\", \"🏹\", \"🎣\", \"🤿\", \"🥊\", \"🥋\", \"🎽\", \"🛹\", \"🛼\", \"🛷\",\n \"⛸\", \"🥌\", \"🎿\", \"⛷\", \"🏂\", \"🏋\", \"🤼\", \"🤸\", \"⛹\", \"🤺\",\n \"🤾\", \"🏌\", \"🏇\", \"🧘\", \"🏄\", \"🏊\", \"🤽\", \"🚣\", \"🧗\", \"🚵\",\n \"🚴\", \"🏃\", \"🚶\", \"🏅\", \"🏆\", \"🥇\", \"🥈\", \"🥉\",\n \"⏱\", \"⏲\", \"⌚\", \"💪\", \"🦵\", \"🦶\", \"👟\", \"💦\", \"💧\", \"🔥\", \n \"❤️\", \"💓\", \"💗\", \"💖\", \"⚡\", \"🏁\", \"🎌\", \"🎯\"\n }\n\n def extract_comments(code, ext):\n comments = []\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css']:\n # 匹配单行注释 // 和多行注释 /* ... */\n pattern = r'(//.*?$)|(/\\*[\\s\\S]*?\\*/)'\n matches = re.finditer(pattern, code, re.MULTILINE)\n for match in matches:\n comments.append(match.group(0))\n elif ext in ['.html']:\n # 匹配 HTML 注释 <!-- ... -->\n pattern = r'<!--[\\s\\S]*?-->'\n matches = re.finditer(pattern, code)\n for match in matches:\n comments.append(match.group(0))\n return \"\".join(comments)\n\n code_file_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n found_any_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录及常见的构建、依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'build', 'dist')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_file_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n\n found_any_code_file = True\n comments_text = extract_comments(code, ext)\n\n has_sports_emoji = False\n for char in comments_text:\n # 使用题目提供的 is_emoji 函数，并校验是否属于运动相关集合\n if is_emoji(char) and char in SPORTS_EMOJIS:\n has_sports_emoji = True\n break\n\n # 只要有任何一个代码文件的注释里不包含运动相关emoji，即为不满足要求\n if not has_sports_emoji:\n return False\n\n # 若未找到任何代码文件，通常视为未满足开发要求\n return found_any_code_file","tags":["Content","Keyword"]}]},{"round_id":4,"instruction":"不错，现在我们来构建'社交'页面的UI。在这个Tab里顶部设计一个切换菜单，分为'排行榜'和'运动动态'。'排行榜'是一个列表，展示排名前5的好友头像、昵称和本周运动总时长，前三名的序号要分别用金银铜牌的颜色区分。'运动动态'是一个信息流列表，展示其他人分享的训练卡片，每条动态要包含发布者的基本信息、完成的具体项目（比如'完成了4组TABATA'），以及一个可交互的'点赞'按钮，点击后爱心图标会有个放大的微交互并变成红色，同时点赞数+1。另外从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是针对每一个生成的代码文件，注释的行数需精准控制在7行，原有的对应注释行数旧要求作废；二是处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令，请严格执行。","instruction_following_checklist":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，则说明该文件不是使用 UTF-8 编码的\n return False\n except Exception:\n # 对于其他异常（如权限问题等），安全起见也视为验证失败\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在7行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React相关常见代码文件后缀\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less'}\n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 剔除无关紧要的目录，加快遍历并避免扫描构建产物与依赖\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'build', 'dist', '.next', 'coverage'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n code_files_found += 1\n \n in_block_star = False\n in_block_html = False\n comment_lines_count = 0\n \n lines = content.split('\\n')\n for line in lines:\n stripped = line.strip()\n \n # 处于块级 /* ... */ 注释块内\n if in_block_star:\n comment_lines_count += 1\n if '*/' in line:\n in_block_star = False\n \n # 处于HTML <!-- ... --> 注释块内\n elif in_block_html:\n comment_lines_count += 1\n if '-->' in line:\n in_block_html = False\n \n else:\n if not stripped: \n continue\n \n # 查找当前行各类注释标记的最早出现位置\n idx_star = stripped.find('/*')\n idx_html = stripped.find('<!--')\n \n # 特殊处理 //：通过长度相等的下划线替换 http(s) 协议符，避免改变原字符串索引并排除干扰\n temp = stripped.replace('http://', '_______').replace('https://', '________')\n idx_slash = temp.find('//')\n \n valid_indices = []\n if idx_star != -1: valid_indices.append(('star', idx_star))\n if idx_html != -1: valid_indices.append(('html', idx_html))\n if idx_slash != -1: valid_indices.append(('slash', idx_slash))\n \n if not valid_indices:\n continue\n \n # 找出同一行中最先出现的注释符号类型\n first_marker = min(valid_indices, key=lambda x: x[1])[0]\n \n if first_marker == 'star':\n comment_lines_count += 1\n # 确认该行是否包含闭合符号，不包含则进入多行块注释状态\n if '*/' not in stripped[idx_star+2:]:\n in_block_star = True\n elif first_marker == 'html':\n comment_lines_count += 1\n # 同理，处理 HTML 注释的闭合情况\n if '-->' not in stripped[idx_html+4:]:\n in_block_html = True\n elif first_marker == 'slash':\n comment_lines_count += 1\n \n # 如果任意一个生成的代码文件的注释不为7行，则不满足要求\n if comment_lines_count != 7:\n return False\n \n # 如果未找到任何符合条件的代码文件，也视为不满足要求\n if code_files_found == 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码项目最多只能包含3个目录层级","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n for dirpath, dirnames, filenames in os.walk(workspace_path):\n # 排除非AI自动生成的系统、环境依赖等文件夹，避免因为这些文件夹过深产生误判\n for exclude_dir in ['.git', 'node_modules', '__pycache__', '.idea', '.vscode', 'build', 'dist']:\n if exclude_dir in dirnames:\n dirnames.remove(exclude_dir)\n \n rel_path = os.path.relpath(dirpath, workspace_path)\n \n # 跳过项目根目录自身\n if rel_path == '.':\n continue\n \n # Path.parts 可以将相对路径拆解为由目录层级名称组成的元组\n # 例如 'A/B/C' 会被拆分为 ('A', 'B', 'C')，长度为 3\n parts = Path(rel_path).parts\n \n # 发现超过3个目录层级的子目录，则不满足要求\n if len(parts) > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"你输出的每一个代码文件，连同空行在内的总行数必须是偶数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React和前端开发中常见的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.json'}\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的第三方依赖或系统隐藏目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用 split('\\n') 能够准确统计包含所有空行（包括文件结尾空行）的物理行数\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n \n has_code_files = True\n except Exception:\n # 如果遇到文件无法读取等异常，忽略该文件继续检查\n pass\n \n # 确保至少检查过一个代码文件且全部符合要求\n return has_code_files","tags":["Quantity","Exact Value"]},{"constraint":"请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号","validation_code":"def verify_sport_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 涵盖React项目可能用到的代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\n # 排除不需要检查的工程/依赖目录\n exclude_dirs = {'node_modules', 'build', 'dist', '.git'}\n \n # 预设运动相关基础 emoji 列表（囊括主要运动类别）\n sport_emojis = [\n \"⚽\", \"⚾\", \"🏀\", \"🏐\", \"🏈\", \"🏉\", \"🎾\", \"🎳\", \"🏏\", \"🏑\", \n \"🏒\", \"🏓\", \"🏸\", \"🥊\", \"🥋\", \"🥅\", \"⛳\", \"⛸\", \"🎣\", \"🎽\", \n \"🎿\", \"🏂\", \"🏋\", \"🤼\", \"🤸\", \"⛹\", \"🤺\", \"🤾\", \"🏌\", \"🏇\", \n \"🧘\", \"🏄\", \"🏊\", \"🤽\", \"🚣\", \"🧗\", \"🚵\", \"🚴\", \"🏆\", \"🏅\", \n \"🎖\", \"🥇\", \"🥈\", \"🥉\", \"🏃\", \"🚶\", \"🥏\", \"🥎\", \"🥍\", \"🪀\", \n \"🪁\", \"🪃\", \"🛼\", \"🛹\", \"🛷\", \"🥌\", \"🤿\", \"🩰\"\n ]\n \n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤排除特定目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取代码中的注释部分\n comments = []\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less'}:\n # 匹配单行注释 //\n comments.extend(re.findall(r'//.*', content))\n # 匹配多行注释 /* */ (涵盖JSX注释)\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n elif ext == '.html':\n # 匹配 HTML 注释 <!-- -->\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n comments_text = \"\".join(comments)\n \n # 若文件没有任何注释内容，则直接不符合要求\n if not comments_text:\n return False\n \n has_sport_emoji = False\n \n # 方式1：按照字符级别，通过 is_emoji 预制函数辅助验证\n for char in comments_text:\n if is_emoji(char) and char in sport_emojis:\n has_sport_emoji = True\n break\n \n # 方式2：按子串判断，预防带修饰符或ZWJ拼接的复合型 emoji 未被单字符识别\n if not has_sport_emoji:\n for em in sport_emojis:\n if em in comments_text:\n has_sport_emoji = True\n break\n \n # 若当前这一个代码文件注释中未找到任何运动表情，即返回False\n if not has_sport_emoji:\n return False\n \n files_checked += 1\n \n # 必须有解析到的代码文件，才算通过测试\n if files_checked == 0:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":5,"instruction":"最后，让这个应用更有分享欲。回到'训练'页面，当间歇训练全部组数彻底结束时，自动弹出一个'恭喜完成'的全屏模态框。在这个弹窗里不仅要有本次训练的总结数据，还要提供一个'生成分享海报'的按钮。点击该按钮后，在当前页面渲染出一张精美的海报预览区域，包含炫酷的渐变背景、用户的训练成绩排版、一句随机生成的运动激励语（比如'汗水是脂肪的眼泪'），以及右下角一个模拟的应用二维码。这样用户就可以很自豪地截图分享了。另外从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是针对每一个生成的代码文件，注释的行数需精准控制在8行，原有的对应注释行数旧要求作废；二是任何时刻都不允许出现一个以上的agent处于同时运行的状态，请严格执行。","instruction_following_checklist":[{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必使用UTF-8编码你所创建的全部代码文件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回未满足\n if not os.path.exists(workspace_path):\n return False\n\n # 根据 React 网页应用推断可能的代码文件扩展名\n target_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.json'}\n code_file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略无需检查的依赖目录和打包目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if 'build' in dirs:\n dirs.remove('build')\n if 'dist' in dirs:\n dirs.remove('dist')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n try:\n # 尝试强制以 utf-8 编码读取内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若发生 Unicode 解码错误，说明含有非 UTF-8 编码格式的内容（如其他编码保存的中文字符）\n return False\n except Exception:\n # 对于其他潜在的无权限读取等非编码异常，选择安全忽略\n pass\n \n # 至少要能找到相关联的代码文件，且其全部都能被 UTF-8 成功解码，才算满足条件\n return code_file_found","tags":["Language","File Encoding"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在8行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 匹配JS/TS及其React版本的字符串字面量、单行注释、多行注释\n # 使用负向先行断言 (?<!:) 防止将 http:// 或 https:// 中的 // 误识别为注释\n js_pattern = re.compile(\n r'(?P<string>\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\'|`(?:[^`\\\\]|\\\\.)*`)'\n r'|(?P<single>(?<!:)//[^\\n]*)'\n r'|(?P<multi>/\\*[\\s\\S]*?\\*/)'\n )\n \n # CSS系列只有字符串和多行注释\n css_pattern = re.compile(\n r'(?P<string>\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\')'\n r'|(?P<multi>/\\*[\\s\\S]*?\\*/)'\n )\n \n # HTML系列包括字符串和HTML注释\n html_pattern = re.compile(\n r'(?P<string>\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\')'\n r'|(?P<multi><!--[\\s\\S]*?-->)'\n )\n \n # 代码文件拓展名及对应的正则模式\n code_extensions = {\n '.js': js_pattern,\n '.jsx': js_pattern,\n '.ts': js_pattern,\n '.tsx': js_pattern,\n '.css': css_pattern,\n '.scss': js_pattern, # SCSS和LESS支持单行 // 注释\n '.less': js_pattern,\n '.html': html_pattern,\n '.htm': html_pattern\n }\n \n ignored_dirs = {'node_modules', '.git', 'build', 'dist', 'out', 'coverage'}\n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改dirs列表，跳过不需要的目录（如依赖库或构建产物）\n dirs[:] = [d for d in dirs if d not in ignored_dirs]\n \n for file in files:\n lower_name = file.lower()\n # 跳过历史备份文件，否则会因为早前轮次的备份文件不符合最新要求而引起误判\n if (lower_name.endswith('.bak') or \n lower_name.endswith('.old') or \n lower_name.endswith('.prev') or\n '_backup' in lower_name or \n '-backup' in lower_name or\n '_copy' in lower_name or\n '-copy' in lower_name or\n '_prev' in lower_name or\n '-prev' in lower_name):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n pattern = code_extensions[ext]\n comment_lines = set()\n \n for match in pattern.finditer(content):\n # 如果匹配到的是单行或多行注释（排除匹配到字符串的情况）\n if match.lastgroup in ['single', 'multi']:\n start_pos = match.start()\n end_pos = match.end()\n \n # 计算注释所在的起止行号 (通过统计该位置前出现换行符的次数得出，0-indexed)\n start_line = content.count('\\n', 0, start_pos)\n end_line = content.count('\\n', 0, end_pos)\n \n # 将占据的行号全部加入集合中，自动天然去重\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n \n # 判断当前代码文件注释覆盖的行数是否精准为8行\n if len(comment_lines) != 8:\n return False\n \n files_checked += 1\n \n # 如果没有找到任何有效的代码文件，则视作不满足要求\n if files_checked == 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码项目最多只能包含3个目录层级","validation_code":"def check_directory_depth(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n \n workspace = Path(workspace_path)\n \n # 检查工作区路径是否存在且为目录\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for root, dirs, files in os.walk(workspace):\n # 过滤隐藏目录（如 .git, .idea 等），避免错误地计算隐藏配置层级\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n try:\n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = Path(root).relative_to(workspace)\n \n # rel_path.parts 的长度代表嵌套的目录层级数\n # 根目录 '.' 的 parts 长度为 0\n # 'src' 的 parts 长度为 1\n # 'src/components/button' 的 parts 长度为 3\n if len(rel_path.parts) > 3:\n return False\n \n except ValueError:\n # 防止发生路径解析错误\n continue\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"你输出的每一个代码文件，连同空行在内的总行数必须是偶数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对React网页应用，推断可能生成的常见代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\n \n # 如果工作区不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n \n has_checked_any_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非AI生成的常见依赖目录及隐藏版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_checked_any_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按 \\n 分割文件内容，该列表长度即为包含空行的准确总行数\n lines_count = len(content.split('\\n'))\n \n # 检查总行数是否为偶数，若有奇数则违反要求\n if lines_count % 2 != 0:\n return False\n \n except Exception:\n # 若文件读取发生异常（如编码非UTF-8等），保守判定为不满足要求\n return False\n \n # 如果至少检查过一个代码文件且都通过了行数测试，才算真正满足条件\n return has_checked_any_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 预定义常见的运动相关emoji集合\n SPORT_EMOJIS = {\n \"🏃\", \"🏃♂️\", \"🏃♀️\", \"🏃♂\", \"🏃♀\", \"💪\", \"🏋️\", \"🏋️♂️\", \"🏋️♀️\", \"🏋\", \n \"🚴\", \"🚴♂️\", \"🚴♀️\", \"🚴♂\", \"🚴♀\", \"🚵\", \"🚵♂️\", \"🚵♀️\", \"🚵♂\", \"🚵♀\",\n \"🤸\", \"🤸♂️\", \"🤸♀️\", \"🤸♂\", \"🤸♀\", \"🤼\", \"🤼♂️\", \"🤼♀️\", \"🤼♂\", \"🤼♀\",\n \"🤽\", \"🤽♂️\", \"🤽♀️\", \"🤽♂\", \"🤽♀\", \"🤾\", \"🤾♂️\", \"🤾♀️\", \"🤾♂\", \"🤾♀\",\n \"🤺\", \"🏌️\", \"🏌\", \"🏇\", \"🧘\", \"🧘♂️\", \"🧘♀️\", \"🧘♂\", \"🧘♀\",\n \"🏄\", \"🏄♂️\", \"🏄♀️\", \"🏄♂\", \"🏄♀\", \"🏊\", \"🏊♂️\", \"🏊♀️\", \"🏊♂\", \"🏊♀\",\n \"🚣\", \"🚣♂️\", \"🚣♀️\", \"🚣♂\", \"🚣♀\", \"🧗\", \"🧗♂️\", \"🧗♀️\", \"🧗♂\", \"🧗♀\",\n \"⚽\", \"🏀\", \"🏈\", \"⚾\", \"🥎\", \"🎾\", \"🏐\", \"🏉\", \"🥏\", \"🎱\", \"🪀\", \"🏓\", \"🏸\", \"🏒\",\n \"🏑\", \"🥍\", \"🏏\", \"🪃\", \"🥅\", \"⛳\", \"🪁\", \"🏹\", \"🎣\", \"🤿\", \"🥊\", \"🥋\", \"🎽\", \"🛹\",\n \"🛼\", \"🛷\", \"⛸\", \"🥌\", \"🎿\", \"⛷\", \"🏂\", \"🪂\", \"👟\", \"🚲\", \"🏆\", \"🏅\", \"🥇\", \"🥈\", \"🥉\"\n }\n \n # 针对React开发项目，定义可能的代码文件后缀列表\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历项目文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件读取失败，则视为无法验证该文件是否满足要求\n return False\n \n # 提取多种格式的注释内容\n comments = []\n # 匹配 /* ... */ 类型的多行注释\n comments.extend(re.findall(r'/\\*.*?\\*/', content, flags=re.DOTALL))\n # 匹配 <!-- ... --> 类型的HTML多行注释\n comments.extend(re.findall(r'<!--.*?-->', content, flags=re.DOTALL))\n # 匹配 // ... 类型的单行注释\n comments.extend(re.findall(r'//.*', content))\n \n comments_text = \" \".join(comments)\n \n # 检查该文件的所有注释拼接起来后是否包含任一运动相关的emoji\n found = any(emoji in comments_text for emoji in SPORT_EMOJIS)\n \n # 一旦发现某一个代码文件未包含运动emoji，直接判断为未满足要求\n if not found:\n return False\n \n # 如果找到了代码文件且均通过检验，则满足要求；若没有任何代码文件，则也不算满足。\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"任何时刻都不允许出现一个以上的agent处于同时运行的状态","validation_code":"","tags":["Workflow","Orchestration"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用React开发一个运动计时器和间歇训练器的网页应用。首先，我们需要搭建基本的页面骨架。请创建一个包含底部导航栏的应用框架，导航栏有三个Tab：'训练'、'数据'和'社交'。默认显示'训练'页面，在这个页面里先只放一个最基础的数字秒表（显示00:00.00），以及'开始'和'暂停'两个按钮就行。暂时不需要真正的路由配置，用组件状态来切换底部Tab即可。帮我把这个骨架搭好。另外请你在后续所有开发过程中始终遵循以下要求：一是你生成的每一个代码文件都精确包含3行注释，二是确保你输出的代码项目文件夹嵌套深度不超过2层。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令开发React应用的要求，推断出常见的代码文件后缀\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n code_files = []\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉可能存在的依赖或版本控制等非直接生成的目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何代码文件，则算作未完成代码生成任务，无法满足该编码要求\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦出现无法以 UTF-8 解码的文件，说明未满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每一个代码文件都精确包含3行注释\",\n \"validation_code\": \"import os\\nimport re\\nimport bisect\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端相关代码文件类型\\n EXTENSIONS = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.htm', '.scss', '.less'}\\n \\n # 构建正则表达式：\\n # 第一部分：匹配字符串内容（避免将字符串中的 // 等当成注释）\\n # 第二部分：匹配单行注释 // ...\\n # 第三部分：匹配多行块注释 /* ... */\\n # 第四部分：匹配 HTML 注释 <!-- ... -->\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`(?:[^`\\\\\\\\]|\\\\\\\\.)*`)'\\n r'|(?P<sline>//[^\\\\n]*)'\\n r'|(?P<mline>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n r'|(?P<html_comment><!--[\\\\s\\\\S]*?-->)'\\n )\\n \\n code_files = []\\n # 遍历工作区目录，收集所有生成的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in EXTENSIONS:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何代码文件，说明并未生成，要求未满足\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 记录所有换行符的索引位置，便于后续计算绝对字符位置所属的行号\\n nl_positions = [m.start() for m in re.finditer(r'\\\\n', content)]\\n \\n def get_line_number(index):\\n # 利用二分查找获取当前字符所在的行号 (从 0 开始)\\n return bisect.bisect_right(nl_positions, index)\\n \\n comment_line_indices = set()\\n \\n # 匹配文本中的注释\\n for m in pattern.finditer(content):\\n group = m.lastgroup\\n # 只统计属于注释的分组\\n if group in ('sline', 'mline', 'html_comment'):\\n start_index = m.start(group)\\n # m.end(group) 是结束后的索引，因此需减去1来获取真实的最后包含的字符位置\\n end_index = m.end(group) - 1 if m.end(group) > m.start(group) else start_index\\n \\n start_line = get_line_number(start_index)\\n end_line = get_line_number(end_index)\\n \\n # 将该注释跨越的所有行号加入集合\\n for line in range(start_line, end_line + 1):\\n comment_line_indices.add(line)\\n \\n # 每一个代码文件的注释行数必须精确等于3行\\n if len(comment_line_indices) != 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过2层\",\n \"validation_code\": \"def check_folder_depth(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 若工作区路径不存在，可视为未创建任何多余嵌套文件夹，符合条件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非AI直接编写的代码目录，如隐藏文件夹、编译产物和第三方依赖包目录\\n # 修改 dirs[:] 能够直接影响 os.walk 今后的遍历行为，避免深入不必要的目录\\n dirs[:] = [d for d in dirs if d not in ['node_modules', 'build', 'dist', '__pycache__'] and not d.startswith('.')]\\n \\n # 计算当前所在文件夹相对于 workspace_path 的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 项目根目录相对路径为 '.'，深度记为0\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 将路径标准化，并通过 os.sep 分割以计算当前目录的实际嵌套层级\\n rel_path = os.path.normpath(rel_path)\\n depth = len(rel_path.split(os.sep))\\n \\n # 判定：若检测到文件夹的嵌套深度大于2层，则说明不满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来重点完善核心的'训练'页面。我要能在三种模式之间自由切换：'普通秒表'、'倒计时'和'间歇训练(HIIT)'。对于最复杂的'间歇训练'模式，页面上必须包含一个设置表单，让我可以自定义调整：准备时间、单次工作时长、单次休息时长，以及循环的组数。点击开始后，倒计时数字下方要清晰地显示当前处于'准备'、'运动中'还是'休息'状态，并且界面主背景色要根据状态平滑切换（比如准备时黄色，运动时红色，休息时绿色），方便运动时瞟一眼就能确认状态。另外从本轮开始后续所有开发过程中请遵循两项调整后的要求：一是针对每一个生成的代码文件，注释的行数需精准控制在5行，二是生成的代码项目最多只能包含3个目录层级，原有的对应旧要求作废，请严格执行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据React应用推断常见代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名并转为小写进行匹配\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制形式读取文件内容\\n with open(file_path, 'rb') as f:\\n file_data = f.read()\\n # 尝试以UTF-8解码，若成功则编码无误\\n file_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦捕获到解码错误，说明存在非UTF-8编码的代码文件\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在5行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见前端业务和样式代码的扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.scss', '.less'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 跳过常见的无关目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n comment_lines = 0\\n \\n # HTML文件注释解析\\n if ext in ['.html']:\\n in_comment = False\\n for line in code.split('\\\\n'):\\n i = 0\\n is_comment_line = in_comment\\n while i < len(line):\\n if in_comment:\\n if line[i:i+3] == '-->':\\n in_comment = False\\n i += 2\\n else:\\n if line[i:i+4] == '<!--':\\n in_comment = True\\n is_comment_line = True\\n i += 3\\n i += 1\\n if is_comment_line:\\n comment_lines += 1\\n # JS/TS/CSS 类文件的注释解析\\n else:\\n in_string = False\\n string_char = ''\\n in_block_comment = False\\n \\n for line in code.split('\\\\n'):\\n i = 0\\n is_comment_line = in_block_comment\\n \\n while i < len(line):\\n if in_block_comment:\\n if line[i:i+2] == '*/':\\n in_block_comment = False\\n i += 1\\n elif in_string:\\n if line[i] == '\\\\\\\\':\\n i += 1 # 忽略转义字符\\n elif line[i] == string_char:\\n in_string = False\\n else:\\n if line[i:i+2] == '//':\\n is_comment_line = True\\n break\\n elif line[i:i+2] == '/*':\\n in_block_comment = True\\n is_comment_line = True\\n i += 1\\n elif line[i] in '\\\"\\\\'`':\\n in_string = True\\n string_char = line[i]\\n i += 1\\n if is_comment_line:\\n comment_lines += 1\\n \\n # 若发现任意代码文件的注释行数不为 5，则未满足要求\\n if comment_lines != 5:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含3个目录层级\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 在原地修改dirs以控制os.walk的后续遍历行为\\n # 忽略隐藏目录以及可能由前端环境和系统自动生成的较深且非代码的目录\\n dirs[:] = [\\n d for d in dirs \\n if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'build', 'dist')\\n ]\\n \\n # 计算相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n if rel_path == '.':\\n # 工作空间根目录算作第1层级\\n depth = 1\\n else:\\n # 统一替换路径分隔符以兼容不同操作系统，随后进行拆分计算\\n normalized_rel_path = rel_path.replace('\\\\\\\\', '/')\\n parts = [p for p in normalized_rel_path.split('/') if p]\\n # 根目录的1层级 + 相对路径包含的子层级数\\n depth = len(parts) + 1\\n \\n # 目录层级超过了3层\\n if depth > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在的计时功能有了，来优化一下间歇训练时的视觉反馈。除了背景变色，我希望加上一个充满科技感的SVG环形进度条，把时间倒计时的数字包围在中间。这个环形进度条要能根据当前阶段（工作或休息）的总时长和剩余时间实时动态递减。另外，为了增加紧迫感，当任意阶段倒计时只剩最后3秒时，中间的数字要出现明显的放大缩小脉冲动画（类似心跳效果），以此来模拟倒数冲刺的紧迫感。从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是针对每一个生成的代码文件，注释的行数需精准控制在6行，原有的对应注释行数旧要求作废；二是你输出的每一个代码文件，连同空行在内的总行数必须是奇数，请严格执行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React等网页开发常见代码文件的后缀名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试以UTF-8编码解码文件内容\\n try:\\n with open(file_path, 'rb') as f:\\n raw_content = f.read()\\n raw_content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码失败，说明不是有效的UTF-8编码\\n return False\\n except Exception:\\n # 读取遇到其他异常情况视为未满足\\n return False\\n\\n # 如果存在代码文件且全都通过了UTF-8验证则返回True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在6行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def count_comment_lines(filepath: str) -> int:\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n text = f.read()\\n\\n ext = filepath.split('.')[-1].lower()\\n \\n in_single_quote = False\\n in_double_quote = False\\n in_backtick = False\\n in_multi_line_comment = False\\n \\n comment_lines = set()\\n line_num = 1\\n \\n i = 0\\n n = len(text)\\n \\n while i < n:\\n if text[i] == '\\\\n':\\n line_num += 1\\n # 容错：遇到换行后如果处于普通字符串中，重置以防不规范代码导致误判\\n if in_single_quote: in_single_quote = False\\n if in_double_quote: in_double_quote = False\\n i += 1\\n continue\\n \\n if ext in ['html', 'htm']:\\n if in_multi_line_comment:\\n comment_lines.add(line_num)\\n if text[i:i+3] == '-->':\\n in_multi_line_comment = False\\n i += 3\\n else:\\n i += 1\\n else:\\n if text[i:i+4] == '<!--':\\n in_multi_line_comment = True\\n comment_lines.add(line_num)\\n i += 4\\n else:\\n i += 1\\n else:\\n if in_multi_line_comment:\\n comment_lines.add(line_num)\\n if text[i:i+2] == '*/':\\n in_multi_line_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n if in_single_quote:\\n if text[i] == '\\\\\\\\':\\n i += 2\\n elif text[i] == \\\"'\\\":\\n in_single_quote = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n if in_double_quote:\\n if text[i] == '\\\\\\\\':\\n i += 2\\n elif text[i] == '\\\"':\\n in_double_quote = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n if in_backtick:\\n if text[i] == '\\\\\\\\':\\n i += 2\\n elif text[i] == '`':\\n in_backtick = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n if text[i] == \\\"'\\\":\\n in_single_quote = True\\n i += 1\\n continue\\n \\n if text[i] == '\\\"':\\n in_double_quote = True\\n i += 1\\n continue\\n \\n if text[i] == '`':\\n in_backtick = True\\n i += 1\\n continue\\n \\n if text[i:i+2] == '//':\\n comment_lines.add(line_num)\\n # 遇到单行注释直接往后跳过直到行尾，不包含 \\\\n 以便让上层处理递增行号\\n while i < n and text[i] != '\\\\n':\\n i += 1\\n continue\\n \\n if text[i:i+2] == '/*':\\n in_multi_line_comment = True\\n comment_lines.add(line_num)\\n i += 2\\n continue\\n \\n i += 1\\n\\n return len(comment_lines)\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.htm'}\\n code_files = []\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的目录\\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', 'build', 'dist']]\\n for f in files:\\n if any(f.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, f))\\n \\n if not code_files:\\n return False\\n \\n # 验证是否每一个代码文件的注释行数都精确等于6\\n for file_path in code_files:\\n if count_comment_lines(file_path) != 6:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含3个目录层级\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果项目路径不存在，无法判断，返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏文件夹（如.git等），这些通常不计入生成的代码项目的层级中\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n rel_path = os.path.relpath(root, workspace_path)\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 通过操作系统的路径分隔符拆分，计算当前目录相对于根目录的深度\\n depth = len(rel_path.split(os.sep))\\n \\n # 如果目录层级大于3层，即违反了“最多只能包含3个目录层级”的约束\\n if depth > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你输出的每一个代码文件，连同空行在内的总行数必须是奇数\",\n \"validation_code\": \"def check_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端React项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的第三方依赖或版本控制目录，避免检查非生成的代码文件\\n path_parts = root.split(os.sep)\\n if 'node_modules' in path_parts or '.git' in path_parts:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 使用 split('\\\\n') 能够严格、准确地统计出包含所有空行在内的行数\\n lines = content.split('\\\\n')\\n \\n # 如果行数是偶数，说明不符合“总行数必须是奇数”的要求\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果由于编码等原因无法读取，则跳过\\n continue\\n \\n # 所有被成功读取的代码文件的行数均为奇数\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步是做数据追踪。当我在任意模式下点击'结束训练'时，将这次的训练数据（模式名称、总时长、完成组数等）保存到 localStorage 中。然后帮我完善底部的'数据'Tab页面：上半部分做一个数据看板，显示'累计运动时长'、'总完成组数'和'预估消耗千卡'；下半部分用简易的柱状图（可以直接用不同高度带颜色的div元素模拟）展示最近7天每天的运动时长变化趋势。最下方再列出最近5条的详细训练历史记录卡片。另外从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是你输出的每一个代码文件，连同空行在内的总行数必须是偶数，原有的对应总行数旧要求作废；二是请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号，请严格执行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React项目开发中常见的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.json', '.md'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖库和版本控制等不需要检查的目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅针对推断出的可能代码文件进行UTF-8编码检查\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(filepath, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦捕获到解码错误，说明存在非UTF-8编码的代码文件\\n return False\\n except Exception:\\n # 忽略读取过程中的其他无关异常（如权限问题等）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在6行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义属于React项目前端代码文件的扩展名集合\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n file_count = 0\\n all_meet = True\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非代码源码目录\\n for ignore_dir in ['node_modules', '.git', 'build', 'dist']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_count += 1\\n file_path = os.path.join(root, file)\\n \\n comment_lines = 0\\n in_block_comment = False\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n stripped = line.strip()\\n \\n # 处理空行\\n if not stripped:\\n # 只有在多行注释块内的空行才计作注释行\\n if in_block_comment:\\n comment_lines += 1\\n continue\\n \\n # 处理仍处于多行注释块内的行\\n if in_block_comment:\\n comment_lines += 1\\n if ext == '.html':\\n if '-->' in stripped:\\n in_block_comment = False\\n else:\\n if '*/' in stripped:\\n in_block_comment = False\\n continue\\n \\n # 处理尚未处于多行注释块内的行\\n if ext == '.html':\\n if '<!--' in stripped:\\n comment_lines += 1\\n if '-->' not in stripped:\\n in_block_comment = True\\n else:\\n if '/*' in stripped:\\n comment_lines += 1\\n if '*/' not in stripped:\\n in_block_comment = True\\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n # 剔除常见的URL前缀以防误判单行注释\\n tmp = stripped.replace('http://', '').replace('https://', '')\\n # 判断是否包含单行注释\\n if tmp.startswith('//') or ' //' in tmp or '\\\\t//' in tmp:\\n comment_lines += 1\\n \\n if comment_lines != 6:\\n all_meet = False\\n break\\n if not all_meet:\\n break\\n \\n # 如果项目中没有任何指定的代码文件，则不满足要求\\n if file_count == 0:\\n return False\\n \\n return all_meet\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含3个目录层级\",\n \"validation_code\": \"import os\\n\\ndef check_directory_depth(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和特定的非代码生成目录，避免误判并且提高遍历效率\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n rel_path = os.path.relpath(root, workspace_path)\\n if rel_path == '.':\\n depth = 1\\n else:\\n # 将路径分隔符统一替换为 '/' 以防跨平台差异，然后计算层级深度\\n normalized_rel_path = rel_path.replace('\\\\\\\\', '/')\\n depth = 1 + len(normalized_rel_path.split('/'))\\n \\n # 如果发现任何深度大于3的目录，说明违反了限制要求\\n if depth > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你输出的每一个代码文件，连同空行在内的总行数必须是偶数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件后缀类型\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css'}\\n # 忽略扫描的目录\\n exclude_dirs = {'node_modules', '.git', 'build', 'dist'}\\n \\n found_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以便 os.walk 忽略指定文件夹\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按换行符分割以精确计算编辑器中展示的总行数(包含空行)\\n lines = content.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 对于不可预期的文件读取异常不应直接让函数崩溃，但可视情况跳过\\n continue\\n \\n # 如果找到了至少一个代码文件且全部符合偶数行要求，则判定通过\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 预定义运动、健身、计时相关的emoji集合（包含基础字符形式，兼容复合emoji匹配）\\n SPORTS_EMOJIS = {\\n \\\"⚽\\\", \\\"🏀\\\", \\\"🏈\\\", \\\"⚾\\\", \\\"🥎\\\", \\\"🎾\\\", \\\"🏐\\\", \\\"🏉\\\", \\\"🥏\\\", \\\"🎱\\\",\\n \\\"🪀\\\", \\\"🏓\\\", \\\"🏸\\\", \\\"🏒\\\", \\\"🏑\\\", \\\"🥍\\\", \\\"🏏\\\", \\\"🪃\\\", \\\"🥅\\\", \\\"⛳\\\",\\n \\\"🪁\\\", \\\"🏹\\\", \\\"🎣\\\", \\\"🤿\\\", \\\"🥊\\\", \\\"🥋\\\", \\\"🎽\\\", \\\"🛹\\\", \\\"🛼\\\", \\\"🛷\\\",\\n \\\"⛸\\\", \\\"🥌\\\", \\\"🎿\\\", \\\"⛷\\\", \\\"🏂\\\", \\\"🏋\\\", \\\"🤼\\\", \\\"🤸\\\", \\\"⛹\\\", \\\"🤺\\\",\\n \\\"🤾\\\", \\\"🏌\\\", \\\"🏇\\\", \\\"🧘\\\", \\\"🏄\\\", \\\"🏊\\\", \\\"🤽\\\", \\\"🚣\\\", \\\"🧗\\\", \\\"🚵\\\",\\n \\\"🚴\\\", \\\"🏃\\\", \\\"🚶\\\", \\\"🏅\\\", \\\"🏆\\\", \\\"🥇\\\", \\\"🥈\\\", \\\"🥉\\\",\\n \\\"⏱\\\", \\\"⏲\\\", \\\"⌚\\\", \\\"💪\\\", \\\"🦵\\\", \\\"🦶\\\", \\\"👟\\\", \\\"💦\\\", \\\"💧\\\", \\\"🔥\\\", \\n \\\"❤️\\\", \\\"💓\\\", \\\"💗\\\", \\\"💖\\\", \\\"⚡\\\", \\\"🏁\\\", \\\"🎌\\\", \\\"🎯\\\"\\n }\\n\\n def extract_comments(code, ext):\\n comments = []\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css']:\\n # 匹配单行注释 // 和多行注释 /* ... */\\n pattern = r'(//.*?$)|(/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n matches = re.finditer(pattern, code, re.MULTILINE)\\n for match in matches:\\n comments.append(match.group(0))\\n elif ext in ['.html']:\\n # 匹配 HTML 注释 <!-- ... -->\\n pattern = r'<!--[\\\\s\\\\S]*?-->'\\n matches = re.finditer(pattern, code)\\n for match in matches:\\n comments.append(match.group(0))\\n return \\\"\\\".join(comments)\\n\\n code_file_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n found_any_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录及常见的构建、依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'build', 'dist')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_file_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n\\n found_any_code_file = True\\n comments_text = extract_comments(code, ext)\\n\\n has_sports_emoji = False\\n for char in comments_text:\\n # 使用题目提供的 is_emoji 函数，并校验是否属于运动相关集合\\n if is_emoji(char) and char in SPORTS_EMOJIS:\\n has_sports_emoji = True\\n break\\n\\n # 只要有任何一个代码文件的注释里不包含运动相关emoji，即为不满足要求\\n if not has_sports_emoji:\\n return False\\n\\n # 若未找到任何代码文件，通常视为未满足开发要求\\n return found_any_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"不错，现在我们来构建'社交'页面的UI。在这个Tab里顶部设计一个切换菜单，分为'排行榜'和'运动动态'。'排行榜'是一个列表，展示排名前5的好友头像、昵称和本周运动总时长，前三名的序号要分别用金银铜牌的颜色区分。'运动动态'是一个信息流列表，展示其他人分享的训练卡片，每条动态要包含发布者的基本信息、完成的具体项目（比如'完成了4组TABATA'），以及一个可交互的'点赞'按钮，点击后爱心图标会有个放大的微交互并变成红色，同时点赞数+1。另外从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是针对每一个生成的代码文件，注释的行数需精准控制在7行，原有的对应注释行数旧要求作废；二是处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令，请严格执行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，则说明该文件不是使用 UTF-8 编码的\\n return False\\n except Exception:\\n # 对于其他异常（如权限问题等），安全起见也视为验证失败\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在7行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React相关常见代码文件后缀\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less'}\\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 剔除无关紧要的目录，加快遍历并避免扫描构建产物与依赖\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'build', 'dist', '.next', 'coverage'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n code_files_found += 1\\n \\n in_block_star = False\\n in_block_html = False\\n comment_lines_count = 0\\n \\n lines = content.split('\\\\n')\\n for line in lines:\\n stripped = line.strip()\\n \\n # 处于块级 /* ... */ 注释块内\\n if in_block_star:\\n comment_lines_count += 1\\n if '*/' in line:\\n in_block_star = False\\n \\n # 处于HTML <!-- ... --> 注释块内\\n elif in_block_html:\\n comment_lines_count += 1\\n if '-->' in line:\\n in_block_html = False\\n \\n else:\\n if not stripped: \\n continue\\n \\n # 查找当前行各类注释标记的最早出现位置\\n idx_star = stripped.find('/*')\\n idx_html = stripped.find('<!--')\\n \\n # 特殊处理 //：通过长度相等的下划线替换 http(s) 协议符，避免改变原字符串索引并排除干扰\\n temp = stripped.replace('http://', '_______').replace('https://', '________')\\n idx_slash = temp.find('//')\\n \\n valid_indices = []\\n if idx_star != -1: valid_indices.append(('star', idx_star))\\n if idx_html != -1: valid_indices.append(('html', idx_html))\\n if idx_slash != -1: valid_indices.append(('slash', idx_slash))\\n \\n if not valid_indices:\\n continue\\n \\n # 找出同一行中最先出现的注释符号类型\\n first_marker = min(valid_indices, key=lambda x: x[1])[0]\\n \\n if first_marker == 'star':\\n comment_lines_count += 1\\n # 确认该行是否包含闭合符号，不包含则进入多行块注释状态\\n if '*/' not in stripped[idx_star+2:]:\\n in_block_star = True\\n elif first_marker == 'html':\\n comment_lines_count += 1\\n # 同理，处理 HTML 注释的闭合情况\\n if '-->' not in stripped[idx_html+4:]:\\n in_block_html = True\\n elif first_marker == 'slash':\\n comment_lines_count += 1\\n \\n # 如果任意一个生成的代码文件的注释不为7行，则不满足要求\\n if comment_lines_count != 7:\\n return False\\n \\n # 如果未找到任何符合条件的代码文件，也视为不满足要求\\n if code_files_found == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含3个目录层级\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n for dirpath, dirnames, filenames in os.walk(workspace_path):\\n # 排除非AI自动生成的系统、环境依赖等文件夹，避免因为这些文件夹过深产生误判\\n for exclude_dir in ['.git', 'node_modules', '__pycache__', '.idea', '.vscode', 'build', 'dist']:\\n if exclude_dir in dirnames:\\n dirnames.remove(exclude_dir)\\n \\n rel_path = os.path.relpath(dirpath, workspace_path)\\n \\n # 跳过项目根目录自身\\n if rel_path == '.':\\n continue\\n \\n # Path.parts 可以将相对路径拆解为由目录层级名称组成的元组\\n # 例如 'A/B/C' 会被拆分为 ('A', 'B', 'C')，长度为 3\\n parts = Path(rel_path).parts\\n \\n # 发现超过3个目录层级的子目录，则不满足要求\\n if len(parts) > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你输出的每一个代码文件，连同空行在内的总行数必须是偶数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React和前端开发中常见的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.json'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的第三方依赖或系统隐藏目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用 split('\\\\n') 能够准确统计包含所有空行（包括文件结尾空行）的物理行数\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n \\n has_code_files = True\\n except Exception:\\n # 如果遇到文件无法读取等异常，忽略该文件继续检查\\n pass\\n \\n # 确保至少检查过一个代码文件且全部符合要求\\n return has_code_files\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号\",\n \"validation_code\": \"def verify_sport_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 涵盖React项目可能用到的代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\\n # 排除不需要检查的工程/依赖目录\\n exclude_dirs = {'node_modules', 'build', 'dist', '.git'}\\n \\n # 预设运动相关基础 emoji 列表（囊括主要运动类别）\\n sport_emojis = [\\n \\\"⚽\\\", \\\"⚾\\\", \\\"🏀\\\", \\\"🏐\\\", \\\"🏈\\\", \\\"🏉\\\", \\\"🎾\\\", \\\"🎳\\\", \\\"🏏\\\", \\\"🏑\\\", \\n \\\"🏒\\\", \\\"🏓\\\", \\\"🏸\\\", \\\"🥊\\\", \\\"🥋\\\", \\\"🥅\\\", \\\"⛳\\\", \\\"⛸\\\", \\\"🎣\\\", \\\"🎽\\\", \\n \\\"🎿\\\", \\\"🏂\\\", \\\"🏋\\\", \\\"🤼\\\", \\\"🤸\\\", \\\"⛹\\\", \\\"🤺\\\", \\\"🤾\\\", \\\"🏌\\\", \\\"🏇\\\", \\n \\\"🧘\\\", \\\"🏄\\\", \\\"🏊\\\", \\\"🤽\\\", \\\"🚣\\\", \\\"🧗\\\", \\\"🚵\\\", \\\"🚴\\\", \\\"🏆\\\", \\\"🏅\\\", \\n \\\"🎖\\\", \\\"🥇\\\", \\\"🥈\\\", \\\"🥉\\\", \\\"🏃\\\", \\\"🚶\\\", \\\"🥏\\\", \\\"🥎\\\", \\\"🥍\\\", \\\"🪀\\\", \\n \\\"🪁\\\", \\\"🪃\\\", \\\"🛼\\\", \\\"🛹\\\", \\\"🛷\\\", \\\"🥌\\\", \\\"🤿\\\", \\\"🩰\\\"\\n ]\\n \\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤排除特定目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取代码中的注释部分\\n comments = []\\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less'}:\\n # 匹配单行注释 //\\n comments.extend(re.findall(r'//.*', content))\\n # 匹配多行注释 /* */ (涵盖JSX注释)\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n elif ext == '.html':\\n # 匹配 HTML 注释 <!-- -->\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n comments_text = \\\"\\\".join(comments)\\n \\n # 若文件没有任何注释内容，则直接不符合要求\\n if not comments_text:\\n return False\\n \\n has_sport_emoji = False\\n \\n # 方式1：按照字符级别，通过 is_emoji 预制函数辅助验证\\n for char in comments_text:\\n if is_emoji(char) and char in sport_emojis:\\n has_sport_emoji = True\\n break\\n \\n # 方式2：按子串判断，预防带修饰符或ZWJ拼接的复合型 emoji 未被单字符识别\\n if not has_sport_emoji:\\n for em in sport_emojis:\\n if em in comments_text:\\n has_sport_emoji = True\\n break\\n \\n # 若当前这一个代码文件注释中未找到任何运动表情，即返回False\\n if not has_sport_emoji:\\n return False\\n \\n files_checked += 1\\n \\n # 必须有解析到的代码文件，才算通过测试\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后，让这个应用更有分享欲。回到'训练'页面，当间歇训练全部组数彻底结束时，自动弹出一个'恭喜完成'的全屏模态框。在这个弹窗里不仅要有本次训练的总结数据，还要提供一个'生成分享海报'的按钮。点击该按钮后，在当前页面渲染出一张精美的海报预览区域，包含炫酷的渐变背景、用户的训练成绩排版、一句随机生成的运动激励语（比如'汗水是脂肪的眼泪'），以及右下角一个模拟的应用二维码。这样用户就可以很自豪地截图分享了。另外从本轮开始后续所有开发过程中请遵循两项调整及新增要求：一是针对每一个生成的代码文件，注释的行数需精准控制在8行，原有的对应注释行数旧要求作废；二是任何时刻都不允许出现一个以上的agent处于同时运行的状态，请严格执行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"凡需修改已有文件，均须提前使用相关工具完成对该文件的阅读\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必使用UTF-8编码你所创建的全部代码文件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回未满足\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据 React 网页应用推断可能的代码文件扩展名\\n target_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.json'}\\n code_file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略无需检查的依赖目录和打包目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'build' in dirs:\\n dirs.remove('build')\\n if 'dist' in dirs:\\n dirs.remove('dist')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试强制以 utf-8 编码读取内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若发生 Unicode 解码错误，说明含有非 UTF-8 编码格式的内容（如其他编码保存的中文字符）\\n return False\\n except Exception:\\n # 对于其他潜在的无权限读取等非编码异常，选择安全忽略\\n pass\\n \\n # 至少要能找到相关联的代码文件，且其全部都能被 UTF-8 成功解码，才算满足条件\\n return code_file_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"请确保在所有情况下都使用中文来生成用户回复\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"代码修改的先决条件是必须使用 TodoWrite 工具规划任务。当某个任务的代码修改（Write/Edit）步骤一经完成，必须毫不延迟地首先调用 TodoComplete 标记该任务完成，期间绝不能穿插处理其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对已有文件进行任何更改之前，你必须先在同级目录里生成一份对应的修改前备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在8行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 匹配JS/TS及其React版本的字符串字面量、单行注释、多行注释\\n # 使用负向先行断言 (?<!:) 防止将 http:// 或 https:// 中的 // 误识别为注释\\n js_pattern = re.compile(\\n r'(?P<string>\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`(?:[^`\\\\\\\\]|\\\\\\\\.)*`)'\\n r'|(?P<single>(?<!:)//[^\\\\n]*)'\\n r'|(?P<multi>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n \\n # CSS系列只有字符串和多行注释\\n css_pattern = re.compile(\\n r'(?P<string>\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\')'\\n r'|(?P<multi>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n \\n # HTML系列包括字符串和HTML注释\\n html_pattern = re.compile(\\n r'(?P<string>\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\')'\\n r'|(?P<multi><!--[\\\\s\\\\S]*?-->)'\\n )\\n \\n # 代码文件拓展名及对应的正则模式\\n code_extensions = {\\n '.js': js_pattern,\\n '.jsx': js_pattern,\\n '.ts': js_pattern,\\n '.tsx': js_pattern,\\n '.css': css_pattern,\\n '.scss': js_pattern, # SCSS和LESS支持单行 // 注释\\n '.less': js_pattern,\\n '.html': html_pattern,\\n '.htm': html_pattern\\n }\\n \\n ignored_dirs = {'node_modules', '.git', 'build', 'dist', 'out', 'coverage'}\\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改dirs列表，跳过不需要的目录（如依赖库或构建产物）\\n dirs[:] = [d for d in dirs if d not in ignored_dirs]\\n \\n for file in files:\\n lower_name = file.lower()\\n # 跳过历史备份文件，否则会因为早前轮次的备份文件不符合最新要求而引起误判\\n if (lower_name.endswith('.bak') or \\n lower_name.endswith('.old') or \\n lower_name.endswith('.prev') or\\n '_backup' in lower_name or \\n '-backup' in lower_name or\\n '_copy' in lower_name or\\n '-copy' in lower_name or\\n '_prev' in lower_name or\\n '-prev' in lower_name):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n pattern = code_extensions[ext]\\n comment_lines = set()\\n \\n for match in pattern.finditer(content):\\n # 如果匹配到的是单行或多行注释（排除匹配到字符串的情况）\\n if match.lastgroup in ['single', 'multi']:\\n start_pos = match.start()\\n end_pos = match.end()\\n \\n # 计算注释所在的起止行号 (通过统计该位置前出现换行符的次数得出，0-indexed)\\n start_line = content.count('\\\\n', 0, start_pos)\\n end_line = content.count('\\\\n', 0, end_pos)\\n \\n # 将占据的行号全部加入集合中，自动天然去重\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n \\n # 判断当前代码文件注释覆盖的行数是否精准为8行\\n if len(comment_lines) != 8:\\n return False\\n \\n files_checked += 1\\n \\n # 如果没有找到任何有效的代码文件，则视作不满足要求\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码项目最多只能包含3个目录层级\",\n \"validation_code\": \"def check_directory_depth(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n \\n workspace = Path(workspace_path)\\n \\n # 检查工作区路径是否存在且为目录\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for root, dirs, files in os.walk(workspace):\\n # 过滤隐藏目录（如 .git, .idea 等），避免错误地计算隐藏配置层级\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n try:\\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = Path(root).relative_to(workspace)\\n \\n # rel_path.parts 的长度代表嵌套的目录层级数\\n # 根目录 '.' 的 parts 长度为 0\\n # 'src' 的 parts 长度为 1\\n # 'src/components/button' 的 parts 长度为 3\\n if len(rel_path.parts) > 3:\\n return False\\n \\n except ValueError:\\n # 防止发生路径解析错误\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你输出的每一个代码文件，连同空行在内的总行数必须是偶数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对React网页应用，推断可能生成的常见代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\\n \\n # 如果工作区不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_checked_any_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非AI生成的常见依赖目录及隐藏版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_checked_any_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按 \\\\n 分割文件内容，该列表长度即为包含空行的准确总行数\\n lines_count = len(content.split('\\\\n'))\\n \\n # 检查总行数是否为偶数，若有奇数则违反要求\\n if lines_count % 2 != 0:\\n return False\\n \\n except Exception:\\n # 若文件读取发生异常（如编码非UTF-8等），保守判定为不满足要求\\n return False\\n \\n # 如果至少检查过一个代码文件且都通过了行数测试，才算真正满足条件\\n return has_checked_any_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的每一个代码文件的注释里都包含运动相关的emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 预定义常见的运动相关emoji集合\\n SPORT_EMOJIS = {\\n \\\"🏃\\\", \\\"🏃♂️\\\", \\\"🏃♀️\\\", \\\"🏃♂\\\", \\\"🏃♀\\\", \\\"💪\\\", \\\"🏋️\\\", \\\"🏋️♂️\\\", \\\"🏋️♀️\\\", \\\"🏋\\\", \\n \\\"🚴\\\", \\\"🚴♂️\\\", \\\"🚴♀️\\\", \\\"🚴♂\\\", \\\"🚴♀\\\", \\\"🚵\\\", \\\"🚵♂️\\\", \\\"🚵♀️\\\", \\\"🚵♂\\\", \\\"🚵♀\\\",\\n \\\"🤸\\\", \\\"🤸♂️\\\", \\\"🤸♀️\\\", \\\"🤸♂\\\", \\\"🤸♀\\\", \\\"🤼\\\", \\\"🤼♂️\\\", \\\"🤼♀️\\\", \\\"🤼♂\\\", \\\"🤼♀\\\",\\n \\\"🤽\\\", \\\"🤽♂️\\\", \\\"🤽♀️\\\", \\\"🤽♂\\\", \\\"🤽♀\\\", \\\"🤾\\\", \\\"🤾♂️\\\", \\\"🤾♀️\\\", \\\"🤾♂\\\", \\\"🤾♀\\\",\\n \\\"🤺\\\", \\\"🏌️\\\", \\\"🏌\\\", \\\"🏇\\\", \\\"🧘\\\", \\\"🧘♂️\\\", \\\"🧘♀️\\\", \\\"🧘♂\\\", \\\"🧘♀\\\",\\n \\\"🏄\\\", \\\"🏄♂️\\\", \\\"🏄♀️\\\", \\\"🏄♂\\\", \\\"🏄♀\\\", \\\"🏊\\\", \\\"🏊♂️\\\", \\\"🏊♀️\\\", \\\"🏊♂\\\", \\\"🏊♀\\\",\\n \\\"🚣\\\", \\\"🚣♂️\\\", \\\"🚣♀️\\\", \\\"🚣♂\\\", \\\"🚣♀\\\", \\\"🧗\\\", \\\"🧗♂️\\\", \\\"🧗♀️\\\", \\\"🧗♂\\\", \\\"🧗♀\\\",\\n \\\"⚽\\\", \\\"🏀\\\", \\\"🏈\\\", \\\"⚾\\\", \\\"🥎\\\", \\\"🎾\\\", \\\"🏐\\\", \\\"🏉\\\", \\\"🥏\\\", \\\"🎱\\\", \\\"🪀\\\", \\\"🏓\\\", \\\"🏸\\\", \\\"🏒\\\",\\n \\\"🏑\\\", \\\"🥍\\\", \\\"🏏\\\", \\\"🪃\\\", \\\"🥅\\\", \\\"⛳\\\", \\\"🪁\\\", \\\"🏹\\\", \\\"🎣\\\", \\\"🤿\\\", \\\"🥊\\\", \\\"🥋\\\", \\\"🎽\\\", \\\"🛹\\\",\\n \\\"🛼\\\", \\\"🛷\\\", \\\"⛸\\\", \\\"🥌\\\", \\\"🎿\\\", \\\"⛷\\\", \\\"🏂\\\", \\\"🪂\\\", \\\"👟\\\", \\\"🚲\\\", \\\"🏆\\\", \\\"🏅\\\", \\\"🥇\\\", \\\"🥈\\\", \\\"🥉\\\"\\n }\\n \\n # 针对React开发项目，定义可能的代码文件后缀列表\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件读取失败，则视为无法验证该文件是否满足要求\\n return False\\n \\n # 提取多种格式的注释内容\\n comments = []\\n # 匹配 /* ... */ 类型的多行注释\\n comments.extend(re.findall(r'/\\\\*.*?\\\\*/', content, flags=re.DOTALL))\\n # 匹配 <!-- ... --> 类型的HTML多行注释\\n comments.extend(re.findall(r'<!--.*?-->', content, flags=re.DOTALL))\\n # 匹配 // ... 类型的单行注释\\n comments.extend(re.findall(r'//.*', content))\\n \\n comments_text = \\\" \\\".join(comments)\\n \\n # 检查该文件的所有注释拼接起来后是否包含任一运动相关的emoji\\n found = any(emoji in comments_text for emoji in SPORT_EMOJIS)\\n \\n # 一旦发现某一个代码文件未包含运动emoji，直接判断为未满足要求\\n if not found:\\n return False\\n \\n # 如果找到了代码文件且均通过检验，则满足要求；若没有任何代码文件，则也不算满足。\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"任何时刻都不允许出现一个以上的agent处于同时运行的状态\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["启动应用，截屏验证底部包含'训练'、'数据'、'社交'三个Tab，并且应用默认显示'训练'页面的内容。分别点击'数据'Tab和'社交'Tab并截屏，验证页面主体内容能正确跟随底部导航切换。","在'训练'页面的'普通秒表'模式下，截屏验证初始时间显示格式为'00:00.00'，且包含'开始'和'暂停'按钮。点击'开始'按钮等待几秒后截屏验证数字递增，点击'暂停'按钮等待几秒后再次截屏，验证数字停止递增且保持不变。","在'训练'页面，点击并确认能在'普通秒表'、'倒计时'、'间歇训练(HIIT)'三个模式间进行切换。切换至'间歇训练'模式后截屏，验证页面中出现一个设置表单，里面包含可以调节'准备时间'、'单次工作时长'、'单次休息时长'和'循环组数'的表单项。","在'间歇训练'模式下设置短时间参数并点击开始，在'准备'、'运动中'和'休息'这三个不同阶段分别进行截屏。验证倒计时数字下方清晰显示了对应的状态文字，并且对比这三张截图，验证页面的主背景色随着状态不同发生了明显的变化（如不同颜色切换）。","在间歇训练倒计时的同一阶段内，于阶段早期和阶段晚期分别截屏。验证倒计时数字外围包裹着一个SVG环形进度条，且晚期截图中的环状进度条长度明显短于早期截图中的长度，证实进度条能随时间动态递减。","在任意倒计时阶段到达最后3秒时进行高频连续截屏。对比截取的多张图片，验证中间的时间数字在屏幕上呈现出不同大小的缩放状态，证实实现了明显的放大缩小脉冲动画效果。","完成一段任意模式的训练并点击'结束训练'，切换到'数据'Tab。截屏验证页面上半部分的数据看板显示了'累计运动时长'、'总完成组数'和'预估消耗千卡'的数据，并在页面最下方列出了训练历史记录卡片。当累计完成5次以上的训练后，截屏验证历史记录列表仅显示最近的5条。","在完成训练并查看'数据'Tab确认数据已显示后，刷新浏览器并再次进入'数据'Tab。截屏验证刷新前的数据看板数值与历史记录卡片依然存在且未丢失，以此确认训练数据已成功保存至 localStorage 中。","在'数据'Tab页面截屏，验证页面下半部分使用带有不同颜色和高度的 div 元素，模拟渲染出了展示最近7天每天运动时长变化趋势的简易柱状图。","切换至'社交'Tab，在顶部菜单点击切换到'排行榜'并截屏。验证列表展示了排名前5的好友头像、昵称和本周运动总时长，并且第1名到第3名的排名序号分别使用了金色、银色、铜色进行区分。","在'社交'Tab切换到'运动动态'信息流列表，记录某一条动态点赞前的状态。点击该动态上的'点赞'按钮，通过连续截屏或录屏观察点击瞬间，验证爱心图标出现了放大的微交互动画效果，随后验证爱心图标变为红色，且旁边的点赞数量比点击前增加了1次。","设置1组且单次时长极短的间歇训练，等待计时彻底归零结束时截屏。验证当前页面自动弹出了一个全屏的'恭喜完成'模态框，弹窗内部展示了本次训练的总结数据以及一个'生成分享海报'的按钮。","在'恭喜完成'的全屏模态框中，点击'生成分享海报'按钮并截屏。验证当前页面渲染出一个精美的海报预览区域，海报中包含渐变背景、用户训练成绩排版、一句随机生成的运动激励语，以及位于右下角的模拟应用二维码。","切换到'倒计时'模式，设置一个倒计时时长（如60秒），点击'开始'按钮，等待数秒后截屏，验证计时器数字从设定时长开始向下递减；待计时归零后，验证计时器停止运行。"],"string":"[\n \"启动应用，截屏验证底部包含'训练'、'数据'、'社交'三个Tab，并且应用默认显示'训练'页面的内容。分别点击'数据'Tab和'社交'Tab并截屏，验证页面主体内容能正确跟随底部导航切换。\",\n \"在'训练'页面的'普通秒表'模式下，截屏验证初始时间显示格式为'00:00.00'，且包含'开始'和'暂停'按钮。点击'开始'按钮等待几秒后截屏验证数字递增，点击'暂停'按钮等待几秒后再次截屏，验证数字停止递增且保持不变。\",\n \"在'训练'页面，点击并确认能在'普通秒表'、'倒计时'、'间歇训练(HIIT)'三个模式间进行切换。切换至'间歇训练'模式后截屏，验证页面中出现一个设置表单，里面包含可以调节'准备时间'、'单次工作时长'、'单次休息时长'和'循环组数'的表单项。\",\n \"在'间歇训练'模式下设置短时间参数并点击开始，在'准备'、'运动中'和'休息'这三个不同阶段分别进行截屏。验证倒计时数字下方清晰显示了对应的状态文字，并且对比这三张截图，验证页面的主背景色随着状态不同发生了明显的变化（如不同颜色切换）。\",\n \"在间歇训练倒计时的同一阶段内，于阶段早期和阶段晚期分别截屏。验证倒计时数字外围包裹着一个SVG环形进度条，且晚期截图中的环状进度条长度明显短于早期截图中的长度，证实进度条能随时间动态递减。\",\n \"在任意倒计时阶段到达最后3秒时进行高频连续截屏。对比截取的多张图片，验证中间的时间数字在屏幕上呈现出不同大小的缩放状态，证实实现了明显的放大缩小脉冲动画效果。\",\n \"完成一段任意模式的训练并点击'结束训练'，切换到'数据'Tab。截屏验证页面上半部分的数据看板显示了'累计运动时长'、'总完成组数'和'预估消耗千卡'的数据，并在页面最下方列出了训练历史记录卡片。当累计完成5次以上的训练后，截屏验证历史记录列表仅显示最近的5条。\",\n \"在完成训练并查看'数据'Tab确认数据已显示后，刷新浏览器并再次进入'数据'Tab。截屏验证刷新前的数据看板数值与历史记录卡片依然存在且未丢失，以此确认训练数据已成功保存至 localStorage 中。\",\n \"在'数据'Tab页面截屏，验证页面下半部分使用带有不同颜色和高度的 div 元素，模拟渲染出了展示最近7天每天运动时长变化趋势的简易柱状图。\",\n \"切换至'社交'Tab，在顶部菜单点击切换到'排行榜'并截屏。验证列表展示了排名前5的好友头像、昵称和本周运动总时长，并且第1名到第3名的排名序号分别使用了金色、银色、铜色进行区分。\",\n \"在'社交'Tab切换到'运动动态'信息流列表，记录某一条动态点赞前的状态。点击该动态上的'点赞'按钮，通过连续截屏或录屏观察点击瞬间，验证爱心图标出现了放大的微交互动画效果，随后验证爱心图标变为红色，且旁边的点赞数量比点击前增加了1次。\",\n \"设置1组且单次时长极短的间歇训练，等待计时彻底归零结束时截屏。验证当前页面自动弹出了一个全屏的'恭喜完成'模态框，弹窗内部展示了本次训练的总结数据以及一个'生成分享海报'的按钮。\",\n \"在'恭喜完成'的全屏模态框中，点击'生成分享海报'按钮并截屏。验证当前页面渲染出一个精美的海报预览区域，海报中包含渐变背景、用户训练成绩排版、一句随机生成的运动激励语，以及位于右下角的模拟应用二维码。\",\n \"切换到'倒计时'模式，设置一个倒计时时长（如60秒），点击'开始'按钮，等待数秒后截屏，验证计时器数字从设定时长开始向下递减；待计时归零后，验证计时器停止运行。\"\n]","isConversational":false}}},{"rowIdx":57,"cells":{"task_id":{"kind":"number","value":58,"string":"58"},"repository_policy":{"kind":"string","value":"你作为代码智能体，在后续所有操作轮次中必须严格遵循以下全部约束：\n1. 你所生成的所有代码文件都必须使用UTF-8进行编码。\n2. 在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起。\n3. 你给出的所有代码文件，在命名时必须加上 dev_ 作为开头。\n4. 规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配。\n5. 在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件。\n6. 务必始终采用中文作为生成用户回复的语言。\n7. 在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少。\n8. 对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见代码、标记和配置文件后缀名白名单\n code_extensions = {\n '.py', '.java', '.js', '.ts', '.jsx', '.tsx', \n '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.rb', '.php', '.swift', '.kt', \n '.sh', '.bash', '.html', '.css', '.sql', \n '.m', '.mm', '.scala', '.dart', '.json', \n '.xml', '.yaml', '.yml', '.ini', '.toml'\n }\n # 特定名称的代码或配置文件（无后缀名）\n code_files_no_ext = {'dockerfile', 'makefile', 'cmakelists.txt'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_name = file.lower()\n ext = os.path.splitext(file_name)[1]\n \n # 判断是否属于代码文件\n if ext in code_extensions or file_name in code_files_no_ext:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦抛出 UnicodeDecodeError，说明不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能非编码相关的异常，如系统权限问题等\n pass\n \n # 如果所有代码文件都能被正常通过 UTF-8 解码读取，满足条件\n return True","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cxx', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', \n '.swift', '.kt', '.rs', '.sh', '.html', '.css', '.sql',\n '.m', '.scala', '.dart', '.lua'\n }\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录及常见的依赖/编译输出目录，避免受到非开发者直接编写的文件干扰\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\n \n for file in files:\n # 获取文件扩展名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断当前文件是否为代码文件\n if ext in code_extensions:\n # 检查文件名是否严格以 \"dev_\" 开头\n if not file.startswith('dev_'):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n \n def get_py_vars(filepath: str) -> set:\n vars_set = set()\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 匹配常见赋值的变量\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n vars_set.add(node.id)\n # 匹配函数的参数\n elif isinstance(node, ast.arg):\n vars_set.add(node.arg)\n # 匹配对象属性赋值的属性名\n elif isinstance(node, ast.Attribute) and isinstance(node.ctx, ast.Store):\n vars_set.add(node.attr)\n except Exception:\n pass\n return vars_set\n\n def get_general_vars(filepath: str) -> set:\n vars_set = set()\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n # 匹配 JS / TS variables\n js_matches = re.findall(r'\\b(?:var|let|const)\\s+([a-zA-Z_]\\w*)', content)\n vars_set.update(js_matches)\n # 匹配 C / C++ / Java / C# 强类型 variables (考虑到可能的指针前缀)\n c_matches = re.findall(r'\\b(?:int|char|float|double|long|short|String|boolean|bool)\\s+(?:\\*\\s*)?([a-zA-Z_]\\w*)', content)\n vars_set.update(c_matches)\n # 匹配 Bash variables\n bash_matches = re.findall(r'(?:^|\\n)\\s*([a-zA-Z_]\\w*)=', content)\n vars_set.update(bash_matches)\n # 匹配 Go 语言短变量声明\n go_matches = re.findall(r'\\b([a-zA-Z_]\\w*)\\s*:=', content)\n vars_set.update(go_matches)\n except Exception:\n pass\n return vars_set\n\n if not os.path.exists(workspace_path):\n return False\n \n found_code_file = False\n \n # 常见的代码文件扩展名\n code_exts = {'.py', '.js', '.ts', '.c', '.cpp', '.java', '.cs', '.go', '.sh', '.php', '.rb', '.swift'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n filepath = os.path.join(root, file)\n \n # 分别对 Python 和其他代码提取变量\n if ext == '.py':\n file_vars = get_py_vars(filepath)\n else:\n file_vars = get_general_vars(filepath)\n \n for var_name in file_vars:\n # 如果任一提取出的变量名长度为偶数，则不满足奇数条件\n if len(var_name) % 2 == 0:\n return False\n\n # 只要成功扫描到代码文件，并且没有出现过偶数长度的变量名，即视为满足要求\n return found_code_file","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"def check_todo_marks(response: str, workspace_path: str) -> bool:\n import os\n\n # 涵盖绝大多数常见编程语言的代码文件后缀\n code_extensions = {\n '.py', '.java', '.js', '.ts', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.rb', '.php', '.swift', '.kt', '.m', '.sh', \n '.html', '.css', '.jsx', '.tsx', '.sql', '.vue'\n }\n \n found_code_file = False\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git, .idea 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件：依据常见扩展名，或前序约束中的 \"dev_\" 文件名前缀\n if ext in code_extensions or file.startswith('dev_'):\n file_path = os.path.join(root, file)\n \n if not os.path.isfile(file_path):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n found_code_file = True\n \n # 检查 \"TODO(dev)\" 的数量是否恰好等于 3\n if content.count('TODO(dev)') != 3:\n return False\n \n except UnicodeDecodeError:\n # 如果不能以UTF-8读取为文本文件（可能为二进制或其他格式），则跳过\n continue\n except Exception:\n # 处理权限不足等其他读取异常\n continue\n \n # 如果找到了至少一个合法的代码文件，并且全都没有触发 False，证明满足要求\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}],"string":"[\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见代码、标记和配置文件后缀名白名单\\n code_extensions = {\\n '.py', '.java', '.js', '.ts', '.jsx', '.tsx', \\n '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.rb', '.php', '.swift', '.kt', \\n '.sh', '.bash', '.html', '.css', '.sql', \\n '.m', '.mm', '.scala', '.dart', '.json', \\n '.xml', '.yaml', '.yml', '.ini', '.toml'\\n }\\n # 特定名称的代码或配置文件（无后缀名）\\n code_files_no_ext = {'dockerfile', 'makefile', 'cmakelists.txt'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_name = file.lower()\\n ext = os.path.splitext(file_name)[1]\\n \\n # 判断是否属于代码文件\\n if ext in code_extensions or file_name in code_files_no_ext:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦抛出 UnicodeDecodeError，说明不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能非编码相关的异常，如系统权限问题等\\n pass\\n \\n # 如果所有代码文件都能被正常通过 UTF-8 解码读取，满足条件\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cxx', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', \\n '.swift', '.kt', '.rs', '.sh', '.html', '.css', '.sql',\\n '.m', '.scala', '.dart', '.lua'\\n }\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录及常见的依赖/编译输出目录，避免受到非开发者直接编写的文件干扰\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\\n \\n for file in files:\\n # 获取文件扩展名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断当前文件是否为代码文件\\n if ext in code_extensions:\\n # 检查文件名是否严格以 \\\"dev_\\\" 开头\\n if not file.startswith('dev_'):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n \\n def get_py_vars(filepath: str) -> set:\\n vars_set = set()\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 匹配常见赋值的变量\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n vars_set.add(node.id)\\n # 匹配函数的参数\\n elif isinstance(node, ast.arg):\\n vars_set.add(node.arg)\\n # 匹配对象属性赋值的属性名\\n elif isinstance(node, ast.Attribute) and isinstance(node.ctx, ast.Store):\\n vars_set.add(node.attr)\\n except Exception:\\n pass\\n return vars_set\\n\\n def get_general_vars(filepath: str) -> set:\\n vars_set = set()\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 匹配 JS / TS variables\\n js_matches = re.findall(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_]\\\\w*)', content)\\n vars_set.update(js_matches)\\n # 匹配 C / C++ / Java / C# 强类型 variables (考虑到可能的指针前缀)\\n c_matches = re.findall(r'\\\\b(?:int|char|float|double|long|short|String|boolean|bool)\\\\s+(?:\\\\*\\\\s*)?([a-zA-Z_]\\\\w*)', content)\\n vars_set.update(c_matches)\\n # 匹配 Bash variables\\n bash_matches = re.findall(r'(?:^|\\\\n)\\\\s*([a-zA-Z_]\\\\w*)=', content)\\n vars_set.update(bash_matches)\\n # 匹配 Go 语言短变量声明\\n go_matches = re.findall(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*:=', content)\\n vars_set.update(go_matches)\\n except Exception:\\n pass\\n return vars_set\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n found_code_file = False\\n \\n # 常见的代码文件扩展名\\n code_exts = {'.py', '.js', '.ts', '.c', '.cpp', '.java', '.cs', '.go', '.sh', '.php', '.rb', '.swift'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n filepath = os.path.join(root, file)\\n \\n # 分别对 Python 和其他代码提取变量\\n if ext == '.py':\\n file_vars = get_py_vars(filepath)\\n else:\\n file_vars = get_general_vars(filepath)\\n \\n for var_name in file_vars:\\n # 如果任一提取出的变量名长度为偶数，则不满足奇数条件\\n if len(var_name) % 2 == 0:\\n return False\\n\\n # 只要成功扫描到代码文件，并且没有出现过偶数长度的变量名，即视为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def check_todo_marks(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 涵盖绝大多数常见编程语言的代码文件后缀\\n code_extensions = {\\n '.py', '.java', '.js', '.ts', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.rb', '.php', '.swift', '.kt', '.m', '.sh', \\n '.html', '.css', '.jsx', '.tsx', '.sql', '.vue'\\n }\\n \\n found_code_file = False\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git, .idea 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件：依据常见扩展名，或前序约束中的 \\\"dev_\\\" 文件名前缀\\n if ext in code_extensions or file.startswith('dev_'):\\n file_path = os.path.join(root, file)\\n \\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n found_code_file = True\\n \\n # 检查 \\\"TODO(dev)\\\" 的数量是否恰好等于 3\\n if content.count('TODO(dev)') != 3:\\n return False\\n \\n except UnicodeDecodeError:\\n # 如果不能以UTF-8读取为文本文件（可能为二进制或其他格式），则跳过\\n continue\\n except Exception:\\n # 处理权限不足等其他读取异常\\n continue\\n \\n # 如果找到了至少一个合法的代码文件，并且全都没有触发 False，证明满足要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我搭建一个基于React的音乐播放列表创建与分析工具的整体骨架。需要一个全局布局：顶部是一个包含应用Logo和当前用户头像的导航栏；左侧是一个侧边栏，包含“我的歌单”、“数据分析”、“发现社区”三个导航菜单。主内容区域默认显示“我的歌单”的占位页面。在主区域居中放置一个“新建歌单”的按钮，点击后能弹出一个简单的Modal弹窗，里面包含一个“歌单名称”输入框和一个“确认”按钮即可，目前暂时不需要实现数据的实际保存逻辑。此外，后续所有轮次的输出都请始终遵循以下要求：你所提供的代码文件，其缩进宽度需设置为2个空格；生成的代码文件中，“=”的两边均需存在两个空格。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据React项目推断常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html', '.json', '.md'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略无关的依赖目录或版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试用 utf-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明文件未按合法的 UTF-8 进行编码\n return False\n except Exception:\n # 其他读取相关的异常忽略（如权限问题等）\n pass\n\n # 只要存在至少一个代码文件且全部都能成功用utf-8读取，即满足条件\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令提及的“基于React的音乐播放列表创建与分析工具”，推断可能的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json', '.scss', '.less'}\n has_code_file = False\n \n # 若工作区路径不存在，自然不满足\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉可能的无关目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n # 提取文件后缀并转换为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为推断的代码文件类型\n if ext in code_extensions:\n has_code_file = True\n \n # 如果代码文件名不是以 \"dev_\" 开头，则判定为不满足要求\n if not file.startswith('dev_'):\n return False\n \n # 如果找到了至少一个代码文件，且均满足要求，则返回 True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义正则表达式，提取代码中可能的变量名\n patterns = [\n # 匹配普通变量、函数及类的声明，如 let num, function handle(), class Component\n re.compile(r'\\b(?:let|var|const|function|class)\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)'),\n # 匹配常见的 React 数组解构赋值，如 const [isOpen, setIsOpen] = useState()\n re.compile(r'\\b(?:let|var|const)\\s*\\[\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*,\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\]'),\n # 匹配单变量对象解构，如 const { data } = ...\n re.compile(r'\\b(?:let|var|const)\\s*\\{\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\}')\n ]\n\n found_any = False\n \n # 辅助验证函数，扫描代码字符串文本中的变量声明\n def check_content(content: str) -> bool:\n nonlocal found_any\n # 首先移除注释（单行及多行），以免将注释中提及的代码作为检验对象\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n for pattern in patterns:\n for match in pattern.findall(content):\n # 统一为列表结构方便遍历\n vars_to_check = [match] if isinstance(match, str) else list(match)\n for v in vars_to_check:\n v = v.strip()\n if v:\n found_any = True\n # 发现任何一个偶数长度的变量名即返回 False\n if len(v) % 2 == 0:\n return False\n return True\n\n # 扫描当前项目的代码文件\n js_files = []\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 针对 React 推断所需的文件类型\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n js_files.append(os.path.join(root, file))\n \n # 如果项目中存在文件，则从文件中校验\n for file_path in js_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n if not check_content(f.read()):\n return False\n except Exception:\n continue\n \n # 如果项目路径没有对应的代码文件生成，则从回复的代码块中提取验证\n if not js_files:\n code_blocks = re.findall(r'```(?:javascript|js|jsx|typescript|ts|tsx)?(.*?)```', response, re.DOTALL | re.IGNORECASE)\n for block in code_blocks:\n if not check_content(block):\n return False\n\n # 若所有匹配变量均为奇数且至少找到了一个变量，才认为条件达成\n return found_any","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n # 定义React骨架项目可能存在的常见代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的依赖和编译目录\n for ignore_dir in ['node_modules', '.git', 'build', 'dist']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件无法以utf-8读取，保守视为不满足要求\n return False\n \n # 验证每个文件中的 \"TODO(dev)\" 标记是否正好为 3 个\n if content.count(\"TODO(dev)\") != 3:\n return False\n \n # 如果项目中找到了至少一个代码文件，且每个代码文件都完全符合条件，则返回True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所提供的代码文件，其缩进宽度需设置为2个空格","validation_code":"def check_indentation_width_2(response: str, workspace_path: str) -> bool:\n import os\n import re\n from collections import defaultdict\n\n indent_diff_counts = defaultdict(int)\n tab_count = 0\n\n def process_text(text: str):\n nonlocal tab_count\n lines = text.split('\\n')\n prev_spaces = 0\n for line in lines:\n # 忽略空行\n if not line.strip():\n continue\n \n # 匹配行首前导空白\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_whitespace = match.group(1)\n # 检查是否掺杂使用了制表符缩进\n if '\\t' in leading_whitespace:\n tab_count += 1\n # 无法确定具体的空格等价量，重置prev_spaces继续后续检查\n prev_spaces = 0\n continue\n curr_spaces = len(leading_whitespace)\n else:\n curr_spaces = 0\n\n # 去除前导空白后检查是否是多行注释中的内容\n stripped_line = line.lstrip()\n if stripped_line.startswith('*'):\n continue\n \n diff = curr_spaces - prev_spaces\n # 我们只关心正的缩进增量（缩进层级的增加）\n if diff > 0:\n indent_diff_counts[diff] += 1\n \n prev_spaces = curr_spaces\n\n # 1. 扫描工作目录中的代码文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n # 排除 Git 目录和 Node.js 依赖包目录的干扰\n if '.git' in root or 'node_modules' in root:\n continue\n for file in files:\n # 针对 React 相关的常见前端与通用代码文件类型进行检测\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json', '.py', '.java', '.c', '.cpp', '.h')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n process_text(f.read())\n except Exception:\n pass\n\n # 2. 扫描回复内容中可能的 Markdown 代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n process_text(block)\n\n total_space_indents = sum(indent_diff_counts.values())\n \n # 如果完全没有任何有效的缩进或者缩进完全被 tab 主导，则无法认定满足要求\n if total_space_indents == 0:\n return False\n \n if tab_count > total_space_indents:\n return False\n\n # 找到出现频次最高（众数）的缩进增量，即被设定的缩进宽度\n most_common_diff = max(indent_diff_counts.items(), key=lambda x: x[1])[0]\n \n return most_common_diff == 2","tags":["Style","Layout"]},{"constraint":"生成的代码文件中，“=”的两边均需存在两个空格","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码和文本文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html', '.htm', '.json', '.md', '.txt', '.py', '.sh'}\n \n # 匹配所有的单独等号，排除 ==, ===, !=, !==, <=, >=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~=, ??= 等\n pattern = r'(?<![=!<>+\\-*/%&|^~?])=(?![=>])'\n has_checked_any_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略依赖和版本控制等目录\n for skip_dir in ['.git', 'node_modules', 'dist', 'build']:\n if skip_dir in dirs:\n dirs.remove(skip_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n has_checked_any_file = True\n matches = list(re.finditer(pattern, content))\n \n for match in matches:\n start, end = match.span()\n left_str = content[:start]\n right_str = content[end:]\n \n # 检查左边是否恰好有两个空格（末尾是以2个空格结尾，且不是以3个空格结尾）\n if not left_str.endswith(' ') or left_str.endswith(' '):\n return False\n \n # 检查右边是否恰好有两个空格（开头是以2个空格开头，且不是以3个空格开头）\n if not right_str.startswith(' ') or right_str.startswith(' '):\n return False\n\n # 至少检查过一个文件，并且全部文件中的等号都符合要求才能算成功\n return has_checked_any_file","tags":["Style","Layout"]}]},{"round_id":1,"instruction":"好，接下来我们在“我的歌单”页面实现核心的歌曲搜索与添加逻辑。请在页面上方增加一个音乐搜索框，支持输入关键字模拟搜索歌曲（请在代码里内置一批Mock数据，包含歌名、歌手、时长、流派、BPM和封面图片）。搜索结果需要以列表形式展示，并且每条结果右侧有一个“+”按钮。点击“+”可以将该歌曲添加至下方的“当前歌单”表格中。在表格底部，需要实时计算并显示当前歌单的总歌曲数和总播放时长。另外，后续所有轮次的输出都请遵循以下更新要求：切记在生成的代码文件中，“=”的左右两侧不能包含空格；在你生成的所有代码文件中，注释内容必须统一使用中文编写。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n\n # React项目常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明文件不是 UTF-8 编码\n return False\n except Exception:\n # 其它异常暂时忽略，或视作读取失败\n pass\n\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"def check_file_name_prefix(response: str, workspace_path: str) -> bool:\n import os\n \n # 验证工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 基于React项目推测可能的代码文件扩展名\n code_extensions = {\".js\", \".jsx\", \".ts\", \".tsx\", \".css\", \".html\"}\n \n # 忽略第三方库、构建产物或版本控制等非模型直接手写的目录\n exclude_dirs = {\".git\", \"node_modules\", \"build\", \"dist\", \"public\", \".next\", \"coverage\"}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n # 若存在任何一个代码文件不以 'dev_' 开头，则判定不满足\n if not file.startswith(\"dev_\"):\n return False\n \n # 如果项目下完全没有代码文件，也不能视作满足\n return has_code_file","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_files = []\n # 检查项目目录是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 获取所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.css', '.html')):\n code_files.append(os.path.join(root, file))\n \n if not code_files:\n return False\n\n has_vars = False\n \n # 排除常见的保留关键字\n reserved_words = {\n 'const', 'let', 'var', 'function', 'class', 'import', 'export', 'default', \n 'return', 'if', 'else', 'for', 'while', 'switch', 'case', 'break', 'continue', \n 'true', 'false', 'null', 'undefined', 'new', 'this', 'super', 'yield', 'await', \n 'async', 'typeof', 'instanceof', 'in', 'void', 'delete', 'catch', 'try', 'finally',\n 'do', 'with', 'debugger', 'extends', 'implements', 'interface', 'package',\n 'private', 'protected', 'public', 'static', 'enum', 'type'\n }\n\n for file in code_files:\n try:\n with open(file, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 移除单行和多行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n # 移除字符串内的内容（避免误将字符串内的文字识别为变量）\n content = re.sub(r'\".*?\"', '\"\"', content)\n content = re.sub(r\"'.*?'\", \"''\", content)\n content = re.sub(r'`.*?`', '``', content, flags=re.DOTALL)\n \n variables = set()\n \n # 1. 提取普通变量声明\n simple_vars = re.findall(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n variables.update(simple_vars)\n \n # 2. 提取解构声明（如数组和对象的解构）\n destruct_blocks = re.findall(r'\\b(?:const|let|var)\\s*([\\[\\{][^;]*?[\\]\\}])\\s*=', content)\n for block in destruct_blocks:\n words = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', block)\n variables.update(words)\n \n # 3. 提取函数和类声明\n func_vars = re.findall(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n variables.update(func_vars)\n \n class_vars = re.findall(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n variables.update(class_vars)\n \n # 4. 提取函数参数（普通函数参数）\n func_args_blocks = re.findall(r'\\bfunction\\s+[a-zA-Z_$][a-zA-Z0-9_$]*\\s*\\(([^;\\{]*?)\\)', content)\n for block in func_args_blocks:\n words = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', block)\n variables.update(words)\n \n # 5. 提取箭头函数的多参数\n arrow_args_blocks = re.findall(r'\\(([^;\\{]*?)\\)\\s*=>', content)\n for block in arrow_args_blocks:\n words = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', block)\n variables.update(words)\n \n # 6. 提取箭头函数的单参数\n arrow_single_args = re.findall(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', content)\n variables.update(arrow_single_args)\n \n # 过滤掉保留字与系统关键词\n variables = variables - reserved_words\n \n if variables:\n has_vars = True\n for var in variables:\n # 若发现存在任意变量名长度为偶数的情况，直接判定未满足要求\n if len(var) % 2 == 0:\n return False\n \n # 所有变量皆经过检验且文件内确实定义了变量\n return has_vars","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 根据指令中React应用的上下文，推测并定义可能的代码文件后缀名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的依赖目录和隐藏目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对可能的代码文件进行检查\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计 \"TODO(dev)\" 标记出现的次数\n if content.count('TODO(dev)') != 3:\n return False\n except Exception:\n # 如果由于任何原因（如编码问题或权限不足）导致文件无法读取，也视为未满足条件\n return False\n\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所提供的代码文件，其缩进宽度需设置为2个空格","validation_code":"def check_indentation(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n all_lines = []\n\n # 1. 从 response 中提取代码块内容\n code_blocks = re.findall(r'```[a-zA-Z]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n all_lines.extend(block.split('\\n'))\n\n # 2. 从 workspace_path 中提取常见代码文件内容\n if os.path.exists(workspace_path):\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json', '.py'}\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n all_lines.extend(f.readlines())\n except Exception:\n pass\n \n # 如果完全没有获取到任何代码行，视为未违背规则\n if not all_lines:\n return True\n\n indents = []\n \n # 3. 逐行检查缩进规则\n for line in all_lines:\n stripped = line.lstrip()\n # 跳过空行和以 '*' 开头的多行注释格式行\n if not stripped or stripped.startswith('*'):\n continue\n \n # 提取前导空白字符\n leading_whitespace = line[:len(line) - len(stripped)]\n \n # 严禁使用制表符\n if '\\t' in leading_whitespace:\n return False\n \n num_spaces = len(leading_whitespace)\n if num_spaces > 0:\n # 如果存在非偶数的空格数（如 1, 3, 5），显然不是 2 个空格缩进\n if num_spaces % 2 != 0:\n return False\n indents.append(num_spaces)\n\n # 如果代码全部都是零缩进（即顶格写），不违反原则\n if not indents:\n return True\n\n # 4. 判断是否可能错用了 4 个空格的缩进\n has_two_indent = any(i % 2 == 0 and i % 4 != 0 for i in indents)\n has_four_indent = any(i % 4 == 0 for i in indents)\n\n # 如果全局只存在 4, 8, 12 等 4 的倍数的缩进，却毫无 2, 6, 10 等缩进，则实质上使用的是 4 空格缩进\n if has_four_indent and not has_two_indent:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"切记在生成的代码文件中，“=”的左右两侧不能包含空格","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对React项目的常见代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\n \n if not os.path.exists(workspace_path):\n # 若目录不存在，无法证明违反了要求\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉 node_modules 等干扰目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 判断等号的左侧或右侧是否出现了空格\n if \" =\" in content or \"= \" in content:\n return False\n except Exception:\n # 对于不可读的文件或编码问题，直接跳过\n continue\n \n return True","tags":["Style","Layout"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":2,"instruction":"现在来完善播放列表的管理体验和底部播放器UI。首先，请让“当前歌单”表格里的歌曲支持拖拽排序（Drag and Drop），并且可以在列表里直接点击删除某首歌曲。其次，在整个页面的最底部固定一个“正在播放”的控制栏，包含歌曲封面、歌名、播放/暂停、上一首/下一首按钮，以及一个可拖动的进度条。当用户点击“当前歌单”中的某首歌曲时，底部控制栏要同步更新为该歌曲的信息，并自动切换为播放状态（UI表现即可，无需真实音频播放）。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的代码注释统一使用中文的规则作废，改为生成的各个代码文件内，注释部分需完全使用英文撰写；其次，在你生成的各个代码文件中，必须包含正好 2 个 箭头运算符(=>)，不可多也不可少。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 定义React项目及前端项目中常见代码文件的扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n has_code_file = False\n \n # 遍历生成代码所在的项目目录\n for root, dirs, files in os.walk(workspace_path):\n # 忽略可能的依赖包目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n _, ext = os.path.splitext(file)\n # 仅验证代码文件\n if ext.lower() in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 以二进制格式读取文件，以检查原始字节\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试强制用utf-8进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出UnicodeDecodeError解码异常，说明不符合UTF-8编码要求\n return False\n except Exception:\n # 发生其他无法读取的问题暂不视为编码错误\n pass\n \n # 如果找到了至少一个代码文件并且没有发生解码错误，则返回True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"def check_code_file_prefix(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端和常用相关的“代码文件”扩展名集合\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', \n '.less', '.html', '.htm', '.py', '.sh'\n }\n \n has_code_file = False\n \n # 若工作区路径不存在，说明没有生成代码项目\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖、构建输出和版本控制目录，避免误判内部原本就存在的文件\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', 'build', 'dist', '__pycache__', '.next')]\n \n for file in files:\n # 忽略隐藏文件（例如 .gitignore, .eslintrc 等）\n if file.startswith('.'):\n continue\n \n # 获取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否属于代码文件\n if ext in code_extensions:\n has_code_file = True\n # 检查文件名是否以 'dev_' 作为开头\n if not file.startswith('dev_'):\n return False\n \n # 如果项目中包含至少一个代码文件，并且所有代码文件都以 dev_ 开头，则返回 True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"import os\nimport re\n\ndef check_odd_length_variables(response: str, workspace_path: str) -> bool:\n # JavaScript/TypeScript 中常见的关键字，需要排除被误判为变量\n JS_KEYWORDS = {\n 'break', 'case', 'catch', 'class', 'const', 'continue', 'debugger', 'default', \n 'delete', 'do', 'else', 'export', 'extends', 'finally', 'for', 'function', 'if', \n 'import', 'in', 'instanceof', 'new', 'return', 'super', 'switch', 'this', 'throw', \n 'try', 'typeof', 'var', 'void', 'while', 'with', 'yield', 'let', 'static', 'enum', \n 'await', 'implements', 'package', 'protected', 'interface', 'private', 'public', \n 'null', 'true', 'false', 'undefined'\n }\n\n # 各类变量及函数定义的正则表达式\n pattern_direct = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n pattern_array = re.compile(r'\\b(?:var|let|const)\\s*\\[(.*?)\\]')\n pattern_object = re.compile(r'\\b(?:var|let|const)\\s*\\{(.*?)\\}')\n pattern_func = re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\((.*?)\\)')\n pattern_class = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n pattern_arrow = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>')\n pattern_arrow_paren = re.compile(r'\\((.*?)\\)\\s*=>')\n\n var_names = set()\n has_files = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目文件，寻找 React 相关的前端代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n has_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 1. 匹配直接定义的变量 (const a = 1)\n for match in pattern_direct.finditer(content):\n var_names.add(match.group(1))\n\n # 2. 匹配数组解构中的变量 (const [a, b] = arr)\n for match in pattern_array.finditer(content):\n inner = match.group(1)\n idents = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', inner)\n var_names.update(idents)\n\n # 3. 匹配对象解构中的变量 (const { a, b } = obj)\n for match in pattern_object.finditer(content):\n inner = match.group(1)\n idents = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', inner)\n var_names.update(idents)\n\n # 4. 匹配函数名和函数参数 (function foo(a, b) {})\n for match in pattern_func.finditer(content):\n var_names.add(match.group(1))\n inner = match.group(2)\n idents = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', inner)\n var_names.update(idents)\n\n # 5. 匹配类名 (class MyClass {})\n for match in pattern_class.finditer(content):\n var_names.add(match.group(1))\n\n # 6. 匹配无括号箭头函数参数 (a => {})\n for match in pattern_arrow.finditer(content):\n var_names.add(match.group(1))\n\n # 7. 匹配有括号箭头函数参数 ((a, b) => {})\n for match in pattern_arrow_paren.finditer(content):\n inner = match.group(1)\n idents = re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', inner)\n var_names.update(idents)\n\n valid_vars = set()\n # 过滤出合法的用户自定义变量名（剔除关键字与纯数字常量等误匹配）\n for v in var_names:\n if v not in JS_KEYWORDS and not v.isdigit():\n valid_vars.add(v)\n\n # 如果代码文件不存在或者没有发现任何变量，无法判断为满足条件\n if not has_files or not valid_vars:\n return False\n\n # 验证是否每一个提取出的变量名长度均为奇数\n for v in valid_vars:\n if len(v) % 2 == 0:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，则认为不满足\n if not os.path.exists(workspace_path):\n return False\n \n # 定义React项目中常见的代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n has_code_file = False\n \n # 遍历项目目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计 \"TODO(dev)\" 标记在该代码文件中出现的次数\n if content.count(\"TODO(dev)\") != 3:\n return False\n except Exception:\n # 如果文件读取出现异常（如编码错误等），视作未能正确验证，返回 False\n return False\n \n # 必须至少验证过一个代码文件，且全部符合要求才算满足\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所提供的代码文件，其缩进宽度需设置为2个空格","validation_code":"def check_indent_two_spaces(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的React项目相关代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n found_any_target_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤以 dev_ 开头的规定代码文件\n if not file.startswith('dev_'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n found_any_target_file = True\n\n indent_2_6_10 = 0 # 不能被 4 整除的偶数缩进（如 2, 6, 10）\n indent_4_8_12 = 0 # 能被 4 整除的偶数缩进（如 4, 8, 12）\n indent_odd = 0 # 奇数缩进\n tab_lines = 0\n total_indented_lines = 0\n\n for line in lines:\n stripped = line.lstrip()\n # 忽略完全空行\n if not stripped:\n continue\n # 忽略多行注释中以 '*' 开头的中间对齐行\n if stripped.startswith('*'):\n continue\n\n # 匹配行首的前导空白\n match = re.match(r'^([ \\t]+)', line)\n if match:\n spaces = match.group(1)\n if '\\t' in spaces:\n tab_lines += 1\n else:\n length = len(spaces)\n if length % 2 != 0:\n indent_odd += 1\n elif length % 4 != 0:\n indent_2_6_10 += 1\n else:\n indent_4_8_12 += 1\n total_indented_lines += 1\n\n if total_indented_lines > 0:\n # 1. 存在使用 Tab 进行的缩进\n if tab_lines > 0:\n return False\n \n # 2. 奇数前导空格的行占据主导地位（容忍少量因代码对齐产生的奇数空格）\n if indent_odd > total_indented_lines * 0.5:\n return False\n \n # 3. 只有 4 空格及其倍数的缩进，而完全不存在 2 空格及其特征倍数的缩进\n if indent_4_8_12 > 0 and indent_2_6_10 == 0:\n return False\n\n return found_any_target_file","tags":["Style","Layout"]},{"constraint":"切记在生成的代码文件中，“=”的左右两侧不能包含空格","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对React项目的常见代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less'}\n # 匹配等号左侧或右侧存在空格（含普通空格、制表符、全角空格）的情况\n pattern = re.compile(r'[ \\t\\u3000]=|=[ \\t\\u3000]')\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若发现等号左右两侧存在空格，则返回False\n if pattern.search(content):\n return False\n except Exception:\n # 忽略读取异常的文件\n continue\n \n # 如果没有找到任何代码文件，视作未满足要求\n if not found_code_file:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 2 个 箭头运算符(=>)，不可多也不可少","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件扩展名集合\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \n '.go', '.rb', '.php', '.cs', '.swift'\n }\n \n # 需要忽略的文件夹，防止扫描到项目依赖或编译产物中\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'coverage', '.next'}\n \n file_found = False\n all_files_valid = True\n \n # 1. 优先检查工作区生成的实际代码文件\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉忽略目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 无法按 utf-8 文本读取的文件直接跳过\n continue\n \n # 检查是否恰好包含 2 个 '=>'\n if content.count('=>') != 2:\n all_files_valid = False\n break\n if not all_files_valid:\n break\n \n # 如果工作区里存在至少一个代码文件，则以实际文件内容验证结果为准\n if file_found:\n return all_files_valid\n \n # 2. 如果工作区中未发现相关代码文件，作为降级方案检查回复文本中的代码块\n lang_pattern = r'```(?:javascript|js|jsx|typescript|ts|tsx|css|html|python|java|cpp|c|go|ruby|php|swift|cs)\\n(.*?)```'\n code_blocks = re.findall(lang_pattern, response, flags=re.DOTALL | re.IGNORECASE)\n \n if code_blocks:\n for block in code_blocks:\n if block.count('=>') != 2:\n return False\n return True\n \n # 既无代码文件，也无带有明确标识的代码块，视为未满足要求\n return False","tags":["Content","Keyword"]}]},{"round_id":3,"instruction":"下一步是实现“数据分析”面板。当用户在侧边栏点击切换到“数据分析”时，主区域需要读取刚刚在“我的歌单”里添加的那些歌曲数据，并展示两个可视化图表（你可以使用普通的DOM配合CSS模拟，或者使用图表库）：1. 一个饼图（Pie Chart），统计当前歌单中不同音乐流派（Genre）的占比；2. 一个柱状图或雷达图，展示当前歌单的BPM（节拍数）分布情况。如果当前歌单没有任何歌曲，请展示一个带有图标的“暂无数据，请先添加歌曲”的空状态提示。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的每个代码文件中正好包含2个箭头运算符(=>)的规则作废，改为每一个代码文件的生成标准是：必须精确包含 3 个 箭头运算符(=>)；其次，只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数），以上两条规则后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目背景推断可能的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\n has_code_file = False\n\n # 遍历生成的项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选出代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 以二进制模式读取并验证UTF-8解码\n with open(file_path, 'rb') as f:\n content = f.read()\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 解码报错说明非 UTF-8 编码，不符合要求\n return False\n except Exception:\n # 若遇到其他读取错误（如无权限等）暂按不满足要求处理\n return False\n\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 根据React项目推断可能的代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', \n '.html', '.htm'\n }\n \n has_code_file = False\n \n # 遍历当前项目地址下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n \n # 检查代码文件名是否以 'dev_' 开头\n if not file.startswith('dev_'):\n return False\n \n # 如果找到了代码文件，并且所有代码文件都以 dev_ 开头，则返回 True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义常见的JavaScript/TypeScript关键字，防止被误判为自定义变量名\n js_keywords = {\n 'const', 'let', 'var', 'if', 'else', 'function', 'return', 'class', \n 'import', 'from', 'export', 'default', 'true', 'false', 'null', 'undefined',\n 'typeof', 'instanceof', 'new', 'await', 'async', 'yield', 'type', 'interface'\n }\n \n variable_names = set()\n \n # 编译用于提取变量声明的正则表达式\n # 1. 提取基础声明 (如 let myVar, const myConst, function myFunc, class MyClass)\n simple_var_re = re.compile(r'\\b(?:const|let|var|function|class)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n # 2. 提取数组解构声明 (如 const [a, b] = ...)\n arr_destruct_re = re.compile(r'\\b(?:const|let|var)\\s+\\[(.*?)\\]\\s*=')\n # 3. 提取对象解构声明 (如 const {x, y} = ...)\n obj_destruct_re = re.compile(r'\\b(?:const|let|var)\\s+\\{(.*?)\\}\\s*=')\n \n found_js_files = False\n \n # 遍历生成项目的文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n found_js_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 提前移除代码中的注释，防止正则将注释中的单词当作变量\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n # 匹配普通声明\n for match in simple_var_re.finditer(content):\n var_name = match.group(1)\n if var_name not in js_keywords:\n variable_names.add(var_name)\n \n # 匹配数组解构\n for match in arr_destruct_re.finditer(content):\n inner_content = match.group(1)\n # 提取解构块中的所有纯标识符\n for word in re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', inner_content):\n if word not in js_keywords:\n variable_names.add(word)\n \n # 匹配对象解构\n for match in obj_destruct_re.finditer(content):\n inner_content = match.group(1)\n # 提取解构块中的所有纯标识符\n for word in re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', inner_content):\n if word not in js_keywords:\n variable_names.add(word)\n except Exception:\n # 忽略无法读取的文件错误\n pass\n \n # 如果项目中不存在 JS/TS 代码文件或没有提取出任何变量，则判断为不满足要求\n if not found_js_files or not variable_names:\n return False\n \n # 检查所有提取的变量名的长度是否都满足奇数要求\n # （这里的“长度”在编程范畴中表示字符的数量，直接使用len函数进行验证即可）\n for var in variable_names:\n if len(var) % 2 == 0:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"import os\n\ndef check_todo_dev_count(response: str, workspace_path: str) -> bool:\n if not os.path.isdir(workspace_path):\n return False\n\n # 定义React项目常见的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\n \n code_file_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的常见依赖或版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计 \"TODO(dev)\" 出现的次数\n if content.count('TODO(dev)') != 3:\n return False\n except Exception:\n # 如果文件读取异常，则认为该文件不满足要求\n return False\n \n return code_file_found","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所提供的代码文件，其缩进宽度需设置为2个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义 React 项目中常见的代码文件后缀\n exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n has_any_indent = False\n \n # 优先检查工作目录中的实际代码文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n indents = set()\n for line in lines:\n stripped = line.lstrip()\n # 忽略空行以及常见的注释行对齐干扰\n if not stripped or stripped.startswith('*') or stripped.startswith('/'):\n continue\n if stripped.startswith('<--') or stripped.startswith('<!--'):\n continue\n \n leading_str = line[:len(line) - len(stripped)]\n # 严禁使用 tab 缩进\n if '\\t' in leading_str:\n return False\n \n spaces = len(leading_str)\n if spaces > 0:\n indents.add(spaces)\n \n # 如果文件内存在缩进\n if indents:\n has_any_indent = True\n # 如果存在大于等于2个空格的缩进，但完全没有2个空格的缩进，说明不是以2个空格为基准的缩进\n if 2 not in indents and any(i >= 2 for i in indents):\n return False\n\n if has_any_indent:\n return True\n \n # 如果工作目录中未找到带有缩进的代码文件，作为 fallback 检查回复文本中的代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9]*[ \\t]*\\n(.*?)```', response, re.DOTALL)\n if not code_blocks:\n return False\n \n has_any_indent_in_resp = False\n for block in code_blocks:\n indents = set()\n for line in block.split('\\n'):\n stripped = line.lstrip()\n if not stripped or stripped.startswith('*') or stripped.startswith('/'):\n continue\n if stripped.startswith('<--') or stripped.startswith('<!--'):\n continue\n \n leading_str = line[:len(line) - len(stripped)]\n if '\\t' in leading_str:\n return False\n \n spaces = len(leading_str)\n if spaces > 0:\n indents.add(spaces)\n \n if indents:\n has_any_indent_in_resp = True\n if 2 not in indents and any(i >= 2 for i in indents):\n return False\n \n return has_any_indent_in_resp","tags":["Style","Layout"]},{"constraint":"切记在生成的代码文件中，“=”的左右两侧不能包含空格","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 基于React项目推断可能的代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n found_code_file = False\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n _, ext = os.path.splitext(file)\n if ext.lower() in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 判断文件中是否存在 '=' 左右带有空格的情况\n if ' =' in content or '= ' in content:\n return False\n except Exception:\n # 读取异常时跳过\n pass\n\n return found_code_file","tags":["Style","Layout"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含 3 个 箭头运算符(=>)","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n code_files_count = 0\n \n # 定义React项目中常见的前端代码文件后缀\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n # 定义需要排除的非代码文件后缀（如日志、配置及各类静态资源）\n exclude_exts = ('.log', '.json', '.md', '.png', '.jpg', '.jpeg', '.svg', '.gif')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 排除非代码文件（例如用户指令要求生成的 change_log_X.log 等）\n if file.lower().endswith(exclude_exts):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n \n # 判定为代码文件的条件：符合前端代码的常见后缀，或是严格遵循指令前缀 'dev_'\n if ext in code_exts or file.startswith('dev_'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 验证当前代码文件中是否精确包含 3 个 '=>' 运算符\n if content.count('=>') != 3:\n return False\n \n code_files_count += 1\n except Exception:\n # 如果遇到无法以纯文本读取的异常文件，跳过即可\n pass\n\n # 至少要验证过一个以上的代码文件才算真正完成要求\n return code_files_count > 0","tags":["Content","Keyword"]},{"constraint":"只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数）","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":4,"instruction":"很棒，现在我们来开发“发现社区”功能。点击侧边栏的“发现社区”后，以卡片瀑布流或网格形式展示其他用户分享的公开歌单（需Mock几条带用户头像、歌单名、点赞数的歌单数据）。为每个歌单卡片实现点赞功能（点击爱心图标，点赞数+1并变为红色高亮），并在卡片底部提供一个“查看评论”的折叠面板，点击展开后能看到一条默认评论，并包含一个简易的输入框，允许用户提交新评论并实时追加显示在下方。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的每一个代码文件必须精确包含3个箭头运算符(=>)的规则作废，改为每一个代码文件的生成标准是：必须精确包含 1 个 \"TODO(human)\"标记；其次，在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划，以上两条规则后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 结合 React 项目推断的常见代码/文本文件后缀\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', \n '.less', '.html', '.json', '.md', '.log', '.sh'\n }\n \n # 遍历时需要忽略的无用/依赖目录\n ignore_dirs = {'node_modules', '.git', 'build', 'dist', '.next'}\n\n code_files_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以过滤掉不需要检查的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断当前文件是否为需要检查的代码文件（基于后缀，或是前序要求的 dev_ 前缀）\n if ext in code_extensions or file.startswith('dev_'):\n file_path = os.path.join(root, file)\n \n # 排除可能同名的非文件实体\n if not os.path.isfile(file_path):\n continue\n \n # 针对非代码的纯二进制或多媒体格式做防误伤拦截\n if ext in {'.png', '.jpg', '.jpeg', '.gif', '.ico', '.mp3', '.mp4'}:\n continue\n \n code_files_found = True\n \n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试使用 UTF-8 进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 解码失败，表明不是有效的 UTF-8 编码文件\n return False\n except Exception:\n # 忽略其他读取层面（如权限）抛出的非编码异常\n pass\n\n # 只要找到了代码文件且均未发生解码错误，即判定为符合要求\n return code_files_found","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.scss', '.less', '.sass'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略第三方库依赖、版本控制以及构建输出目录\n for ignore_dir in ['node_modules', '.git', 'build', 'dist', '.next']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in code_extensions:\n has_code_file = True\n # 若存在不以 'dev_' 开头的代码文件，则不满足要求\n if not file.startswith('dev_'):\n return False\n\n # 必须有生成代码文件，并且全部符合条件才算满足要求\n return has_code_file","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n has_code_file = False\n vars_checked = 0\n\n # 预编译常见的变量/函数/类声明匹配正则\n # 1. 匹配普通单变量: let/const/var variableName = ... 或 let variableName;\n p1 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*[=;]')\n \n # 2. 匹配函数声明: function functionName(...)\n p2 = re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n \n # 3. 匹配类声明: class ClassName { ... 或 class ClassName extends ...\n p3 = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:extends|\\{)')\n \n # 4. 匹配数组解构（双变量，如React useState）: const [foo, bar] = ...\n p4 = re.compile(r'\\b(?:const|let|var)\\s*\\[\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*,\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\]\\s*=')\n \n # 5. 匹配数组解构（单变量）: const [foo] = ...\n p5 = re.compile(r'\\b(?:const|let|var)\\s*\\[\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\]\\s*=')\n\n # JavaScript 常见的需要忽略的保留关键字\n reserved = {\n 'let', 'const', 'var', 'function', 'class', 'extends', 'true', 'false', \n 'null', 'undefined', 'default', 'export', 'import', 'return', 'yield', \n 'await', 'async', 'if', 'else', 'for', 'while', 'switch', 'case', 'type', 'interface'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖库和版本控制的文件夹\n if 'node_modules' in root.split(os.sep) or '.git' in root.split(os.sep):\n continue\n \n for file in files:\n # 根据系统提示中的要求，AI生成的代码必须以 dev_ 开头，且本项目属于React\n if file.startswith('dev_') and file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 移除字符串字面量（简单处理转义，避免跨行误伤）\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", code)\n code = re.sub(r\"`(?:\\\\.|[^`\\\\])*`\", \"``\", code)\n \n # 移除单行及多行注释\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n # 提取此文件内的所有确定为变量名的标识符\n vars_found = []\n vars_found.extend([m.group(1) for m in p1.finditer(code)])\n vars_found.extend([m.group(1) for m in p2.finditer(code)])\n vars_found.extend([m.group(1) for m in p3.finditer(code)])\n \n for m in p4.finditer(code):\n vars_found.append(m.group(1))\n vars_found.append(m.group(2))\n \n vars_found.extend([m.group(1) for m in p5.finditer(code)])\n \n for v in vars_found:\n if v not in reserved:\n vars_checked += 1\n # 要求每个变量名称长度为奇数，如发现偶数长度则校验失败\n if len(v) % 2 == 0:\n return False\n\n # 若未找到任何受检查的代码文件，或代码文件中没声明任何常规变量，视为不满足\n if not has_code_file or vars_checked == 0:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义 React 项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\n found_code_file = False\n \n # 遍历项目文件系统\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理判定为代码的文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计 \"TODO(dev)\" 标记数量，要求正好等于 3 个\n if content.count(\"TODO(dev)\") != 3:\n return False\n \n except Exception:\n # 如果文件无法正常读取，视作未满足要求\n return False\n \n # 要求至少存在一个代码文件且全部满足条件\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所提供的代码文件，其缩进宽度需设置为2个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n\n if not os.path.exists(workspace_path):\n return False\n\n # 定义需要检查缩进的常规代码文件拓展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\n all_files_ok = True\n has_checked_any = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n indentations = []\n in_multiline_comment = False\n \n for line in lines:\n stripped = line.strip()\n # 过滤空行\n if not stripped:\n continue\n \n # 简单过滤多行注释块，避免由于星号对齐导致的奇数空格干扰缩进统计\n if stripped.startswith('/*'):\n in_multiline_comment = True\n if in_multiline_comment:\n if '*/' in stripped:\n in_multiline_comment = False\n continue\n if stripped.startswith('*'):\n continue\n \n # 获取行首的空白字符部分\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 如果行首缩进发现 Tab 键，说明不符合强制的“2个空格”要求\n if '\\t' in leading_whitespace:\n all_files_ok = False\n break\n \n leading_spaces = len(leading_whitespace)\n indentations.append(leading_spaces)\n \n if not all_files_ok:\n break\n \n # 计算相邻非空行的正向缩进差值（即向右缩进时的层级跨度）\n positive_diffs = []\n for i in range(1, len(indentations)):\n diff = indentations[i] - indentations[i-1]\n if diff > 0:\n positive_diffs.append(diff)\n \n if positive_diffs:\n has_checked_any = True\n # 统计出现频率最高的正向缩进差值，以此代表整个文件的基准缩进单位\n counter = Counter(positive_diffs)\n most_common_diff = counter.most_common(1)[0][0]\n \n # 如果最常用的缩进增加跨度不是 2，则该文件不符合2个空格的要求\n if most_common_diff != 2:\n all_files_ok = False\n break\n \n if not all_files_ok:\n break\n\n # 只有当且仅当所有包含缩进的受检代码文件均通过测试，且确有文件被验证过时，才返回 True\n return all_files_ok and has_checked_any","tags":["Style","Layout"]},{"constraint":"切记在生成的代码文件中，“=”的左右两侧不能包含空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据React项目的特点，推断出可能的代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.scss', '.less'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 判断 \"=\" 的左右两侧是否包含空格\n # 不论是赋值等号 (=)，还是全等 (===)，或箭头函数 (=>)，只要 \"=\" 旁有空格就会被以下判断捕获\n if ' =' in content or '= ' in content:\n return False\n except Exception:\n # 遇到文件无法读取等情况跳过\n continue\n \n # 如果有代码文件且都没有违反规则，返回True；否则（包括没有代码文件的情况）返回False\n return has_code_file","tags":["Style","Layout"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含 1 个 \"TODO(human)\"标记","validation_code":"import os\n\ndef check_todo_human_requirement(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', \n '.py', '.java', '.cpp', '.c', '.h', '.hpp', '.cs', '.go', '.rb', \n '.php', '.swift', '.kt'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的无关/第三方依赖/构建结果目录\n for ignore_dir in ['node_modules', '.git', 'dist', 'build']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n # 如果无法按utf-8读取，可能未遵循之前的UTF-8编码要求，或者遇到了非文本文件\n return False\n except Exception:\n return False\n \n # 统计当前代码文件中 \"TODO(human)\" 的出现次数，必须精确包含1个\n if content.count('TODO(human)') != 1:\n return False\n \n return has_code_file","tags":["Content","Keyword"]},{"constraint":"只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":5,"instruction":"最后，我们回到“我的歌单”页面，增加一些高级的交互细节。在“当前歌单”表格的上方，新增一个高级筛选栏，支持按“流派”下拉框和“BPM区间（如慢速、中速、快速）”单选框对已添加的歌曲进行多条件交叉过滤。另外，给表格中的每一行歌曲增加自定义的右键菜单（Context Menu）：禁用浏览器默认右键菜单，右键点击某首歌曲时在其鼠标位置弹出一个自定义菜单，包含“设为下一首播放”、“分享”和“移除”三个选项，点击任意选项后需要弹出对应的全局Toast消息提示。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的代码文件注释需完全使用英文的规则作废，改为生成的各个代码文件内，注释内容必须统一使用中文编写；其次，生成的每个代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-，以上两条规则后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 根据前端React项目及多轮指令中的要求，定义常见的代码及文本文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', \n '.html', '.json', '.py', '.sh', '.md', '.txt', '.log'\n }\n \n file_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出可能是生成的代码或文本文件\n if ext in code_extensions or file.lower() in ['dockerfile', 'makefile']:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试使用utf-8进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果发生解码错误，说明并非使用UTF-8进行编码\n return False\n except Exception:\n # 若遇到其他读取异常，保守起见视同验证失败\n return False\n\n # 若未找到任何代码文件，则不符合基本的生成要求；若都顺利通过解码，则满足要求。\n return file_found","tags":["Language","File Encoding"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头","validation_code":"import os\n\ndef check_file_naming_requirement(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名集合\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方库代码目录，以防止误判（如果是真实React项目）\n path_parts = root.replace('\\\\', '/').split('/')\n if 'node_modules' in path_parts:\n continue\n\n for file_name in files:\n # 提取文件扩展名并统一转换为小写用于匹配\n ext = os.path.splitext(file_name)[1].lower()\n \n # 如果是代码文件\n if ext in code_extensions:\n has_code_file = True\n # 判断代码文件名是否以 dev_ 开头\n if not file_name.startswith('dev_'):\n return False\n \n # 如果遍历完毕未发现违规文件，并且至少存在一个代码文件，则认为满足要求\n return has_code_file","tags":["Style","Naming"]},{"constraint":"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n def strip_comments_and_strings(text):\n # 使用正则表达式一次性匹配并移除字符串和注释，防止由于匹配顺序不同引发的嵌套影响\n pattern = r'/\\*[\\s\\S]*?\\*/|//.*|#.*|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n return re.sub(pattern, '', text)\n\n code_file_extensions = {'.js', '.jsx', '.ts', '.tsx', '.py'}\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_file_extensions:\n continue\n \n has_code_files = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 去除注释与字符串字面量\n content = strip_comments_and_strings(content)\n variables = set()\n \n # 1. 正常的 var, let, const 声明\n matches = re.findall(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n variables.update(matches)\n \n # 2. 数组解构 (例如: const [state, setState] = useState() )\n arr_matches = re.findall(r'\\b(?:var|let|const)\\s*\\[(.*?)\\]', content, re.DOTALL)\n for m in arr_matches:\n for item in m.split(','):\n item = item.strip()\n if not item:\n continue\n var_name = item.split('=')[0].strip()\n var_name = var_name.replace('...', '').strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n variables.add(var_name)\n \n # 3. 对象解构 (例如: const { data, loading } = useQuery() )\n obj_matches = re.findall(r'\\b(?:var|let|const)\\s*\\{(.*?)\\}', content, re.DOTALL)\n for m in obj_matches:\n for item in m.split(','):\n item = item.strip()\n if not item:\n continue\n # 处理可能存在的重命名，例如：a: b\n if ':' in item:\n var_name = item.split(':')[1].split('=')[0].strip()\n else:\n var_name = item.split('=')[0].strip()\n var_name = var_name.replace('...', '').strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n variables.add(var_name)\n \n # 4. 函数形参\n func_matches = re.findall(r'\\bfunction\\s*[a-zA-Z0-9_$]*\\s*\\((.*?)\\)', content, re.DOTALL)\n for m in func_matches:\n for item in m.split(','):\n item = item.strip()\n if not item:\n continue\n var_name = item.split('=')[0].split(':')[0].strip()\n var_name = var_name.replace('...', '').strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n variables.add(var_name)\n \n # 5. 普通的起始赋值语句 (兼顾 Python 或 JS 类属性赋值)\n assign_matches = re.findall(r'^[ \\t]*([a-zA-Z_$][a-zA-Z0-9_$]*)[ \\t]*=[ \\t]*(?![=>])', content, re.MULTILINE)\n for m in assign_matches:\n variables.add(m)\n \n # 安全过滤常见语言的关键字，以免误判\n keywords = {\n 'true', 'false', 'null', 'undefined', 'import', 'export', 'default', 'class', \n 'function', 'return', 'if', 'else', 'for', 'while', 'let', 'const', 'var', \n 'await', 'async', 'yield', 'switch', 'case', 'break', 'continue', 'catch', \n 'try', 'finally', 'with', 'debugger', 'typeof', 'instanceof', 'new', 'void', \n 'delete', 'in', 'of', 'this', 'super'\n }\n \n for v in variables:\n if v in keywords:\n continue\n # 判断每一个变量的长度是否不是奇数（即偶数）\n if len(v) % 2 == 0:\n return False\n \n return has_code_files","tags":["Style","Naming"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(dev)\"标记，不可多也不可少","validation_code":"def check_todo_dev_tags(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n \n # 需要排除的常见非代码目录或依赖目录\n exclude_dirs = {'node_modules', '.git', 'build', 'dist', 'public'}\n \n code_files_count = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n code_files_count += 1\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 统计 \"TODO(dev)\" 的出现次数\n if content.count('TODO(dev)') != 3:\n return False\n except Exception:\n # 若文件读取失败（例如编码错误等异常），无法确认满足，安全起见返回False\n return False\n \n # 如果项目中没有生成任何代码文件，认为要求未完成\n return code_files_count > 0","tags":["Content","Keyword"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所提供的代码文件，其缩进宽度需设置为2个空格","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n \n # 根据项目的要求，推断可能生成的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.py', '.json'}\n code_files = []\n \n # 查找所有AI生成的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据系统要求，生成的文件必须以 dev_ 开头\n if file.startswith('dev_'):\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files.append(os.path.join(root, file))\n \n # 如果没有找到任何相关的代码文件，属于未能正确完成生成任务\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n indent_diffs = []\n prev_indent = 0\n \n for line in lines:\n if not line.strip():\n continue\n \n stripped = line.lstrip()\n # 忽略常见的多行或单行注释对缩进判断的排版干扰\n if stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('#') or stripped.startswith('<!--'):\n continue\n \n leading_whitespace = line[:len(line) - len(stripped)]\n \n # 缩进使用了tab键而不是空格，不满足要求\n if '\\t' in leading_whitespace:\n return False\n \n curr_indent = len(leading_whitespace)\n # 如果发生了缩进的增加，则记录其增加量（这代表了一级缩进的宽度）\n if curr_indent > prev_indent:\n indent_diffs.append(curr_indent - prev_indent)\n \n prev_indent = curr_indent\n \n # 如果文件中存在缩进增加行为，统计最基础的缩进级别\n if indent_diffs:\n counter = Counter(indent_diffs)\n most_common = counter.most_common(1)[0][0]\n # 如果出现最多次数的缩进宽度不是 2 个空格，判定为不符合\n if most_common != 2:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"切记在生成的代码文件中，“=”的左右两侧不能包含空格","validation_code":"def check_equals_no_spaces(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 定义前端React项目及Python可能用到的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.py'}\n\n # 遍历项目下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理代码文件，忽略如 .md, .log, .json 等非程序代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于某种原因无法读取文件，跳过\n continue\n \n # 遍历文件内容，寻找等号\n for i, char in enumerate(content):\n if char == '=':\n # 检查等号左侧是否为空格\n if i > 0 and content[i - 1] == ' ':\n return False\n \n # 检查等号右侧是否为空格\n if i < len(content) - 1 and content[i + 1] == ' ':\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含 1 个 \"TODO(human)\"标记","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n # 定义前端React项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能存在的第三方依赖或Git版本库目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n # 仅针对代码文件进行检查\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n # 如果无法按系统要求的UTF-8编码读取，亦视为不符合规范或非正常代码文件\n return False\n \n # 统计标记字符串出现的次数\n if content.count('TODO(human)') != 1:\n return False\n \n # 必须至少存在一个代码文件且全部满足条件，才算通过要求\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"生成的各个代码文件内，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成的每个代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 根据用户指令推断可能的前端代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n has_code_file = False\n \n # 优先检查工作区中的实际代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的非代码目录，避免干扰\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含规定的编码格式字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n return False\n \n # 如果在工作区中找到了代码文件，并且全部通过了首行检查，则返回 True\n if has_code_file:\n return True\n \n # 如果工作区中没有提取到代码文件，则作为备用降级方案，检查回复内容中的代码块\n # 匹配带有前端代码类型标记的 Markdown 代码块\n code_blocks = re.findall(r'```(js|jsx|ts|tsx|css|html|react)?\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n has_code_block = False\n for lang, block in code_blocks:\n lang = lang.strip().lower()\n if lang in {'js', 'jsx', 'ts', 'tsx', 'css', 'html', 'react'}:\n has_code_block = True\n lines = block.strip().split('\\n')\n # 检查代码块的真实首行是否包含要求的内容\n if not lines or '-*- coding: utf-8 -*-' not in lines[0]:\n return False\n \n # 如果没有工作区文件，但回复中包含对应的代码块且全部符合要求\n if has_code_block:\n return True\n \n # 既没有实际文件，也没有生成对应的代码块，视作没有满足生成代码文件的要求\n return False","tags":["Content","Format"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我搭建一个基于React的音乐播放列表创建与分析工具的整体骨架。需要一个全局布局：顶部是一个包含应用Logo和当前用户头像的导航栏；左侧是一个侧边栏，包含“我的歌单”、“数据分析”、“发现社区”三个导航菜单。主内容区域默认显示“我的歌单”的占位页面。在主区域居中放置一个“新建歌单”的按钮，点击后能弹出一个简单的Modal弹窗，里面包含一个“歌单名称”输入框和一个“确认”按钮即可，目前暂时不需要实现数据的实际保存逻辑。此外，后续所有轮次的输出都请始终遵循以下要求：你所提供的代码文件，其缩进宽度需设置为2个空格；生成的代码文件中，“=”的两边均需存在两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据React项目推断常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html', '.json', '.md'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略无关的依赖目录或版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试用 utf-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明文件未按合法的 UTF-8 进行编码\\n return False\\n except Exception:\\n # 其他读取相关的异常忽略（如权限问题等）\\n pass\\n\\n # 只要存在至少一个代码文件且全部都能成功用utf-8读取，即满足条件\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令提及的“基于React的音乐播放列表创建与分析工具”，推断可能的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json', '.scss', '.less'}\\n has_code_file = False\\n \\n # 若工作区路径不存在，自然不满足\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉可能的无关目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n # 提取文件后缀并转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为推断的代码文件类型\\n if ext in code_extensions:\\n has_code_file = True\\n \\n # 如果代码文件名不是以 \\\"dev_\\\" 开头，则判定为不满足要求\\n if not file.startswith('dev_'):\\n return False\\n \\n # 如果找到了至少一个代码文件，且均满足要求，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义正则表达式，提取代码中可能的变量名\\n patterns = [\\n # 匹配普通变量、函数及类的声明，如 let num, function handle(), class Component\\n re.compile(r'\\\\b(?:let|var|const|function|class)\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)'),\\n # 匹配常见的 React 数组解构赋值，如 const [isOpen, setIsOpen] = useState()\\n re.compile(r'\\\\b(?:let|var|const)\\\\s*\\\\[\\\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*,\\\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\]'),\\n # 匹配单变量对象解构，如 const { data } = ...\\n re.compile(r'\\\\b(?:let|var|const)\\\\s*\\\\{\\\\s*([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\}')\\n ]\\n\\n found_any = False\\n \\n # 辅助验证函数，扫描代码字符串文本中的变量声明\\n def check_content(content: str) -> bool:\\n nonlocal found_any\\n # 首先移除注释（单行及多行），以免将注释中提及的代码作为检验对象\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n for pattern in patterns:\\n for match in pattern.findall(content):\\n # 统一为列表结构方便遍历\\n vars_to_check = [match] if isinstance(match, str) else list(match)\\n for v in vars_to_check:\\n v = v.strip()\\n if v:\\n found_any = True\\n # 发现任何一个偶数长度的变量名即返回 False\\n if len(v) % 2 == 0:\\n return False\\n return True\\n\\n # 扫描当前项目的代码文件\\n js_files = []\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 针对 React 推断所需的文件类型\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n js_files.append(os.path.join(root, file))\\n \\n # 如果项目中存在文件，则从文件中校验\\n for file_path in js_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n if not check_content(f.read()):\\n return False\\n except Exception:\\n continue\\n \\n # 如果项目路径没有对应的代码文件生成，则从回复的代码块中提取验证\\n if not js_files:\\n code_blocks = re.findall(r'```(?:javascript|js|jsx|typescript|ts|tsx)?(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n for block in code_blocks:\\n if not check_content(block):\\n return False\\n\\n # 若所有匹配变量均为奇数且至少找到了一个变量，才认为条件达成\\n return found_any\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义React骨架项目可能存在的常见代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的依赖和编译目录\\n for ignore_dir in ['node_modules', '.git', 'build', 'dist']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件无法以utf-8读取，保守视为不满足要求\\n return False\\n \\n # 验证每个文件中的 \\\"TODO(dev)\\\" 标记是否正好为 3 个\\n if content.count(\\\"TODO(dev)\\\") != 3:\\n return False\\n \\n # 如果项目中找到了至少一个代码文件，且每个代码文件都完全符合条件，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所提供的代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def check_indentation_width_2(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n from collections import defaultdict\\n\\n indent_diff_counts = defaultdict(int)\\n tab_count = 0\\n\\n def process_text(text: str):\\n nonlocal tab_count\\n lines = text.split('\\\\n')\\n prev_spaces = 0\\n for line in lines:\\n # 忽略空行\\n if not line.strip():\\n continue\\n \\n # 匹配行首前导空白\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_whitespace = match.group(1)\\n # 检查是否掺杂使用了制表符缩进\\n if '\\\\t' in leading_whitespace:\\n tab_count += 1\\n # 无法确定具体的空格等价量，重置prev_spaces继续后续检查\\n prev_spaces = 0\\n continue\\n curr_spaces = len(leading_whitespace)\\n else:\\n curr_spaces = 0\\n\\n # 去除前导空白后检查是否是多行注释中的内容\\n stripped_line = line.lstrip()\\n if stripped_line.startswith('*'):\\n continue\\n \\n diff = curr_spaces - prev_spaces\\n # 我们只关心正的缩进增量（缩进层级的增加）\\n if diff > 0:\\n indent_diff_counts[diff] += 1\\n \\n prev_spaces = curr_spaces\\n\\n # 1. 扫描工作目录中的代码文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n # 排除 Git 目录和 Node.js 依赖包目录的干扰\\n if '.git' in root or 'node_modules' in root:\\n continue\\n for file in files:\\n # 针对 React 相关的常见前端与通用代码文件类型进行检测\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json', '.py', '.java', '.c', '.cpp', '.h')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n process_text(f.read())\\n except Exception:\\n pass\\n\\n # 2. 扫描回复内容中可能的 Markdown 代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n process_text(block)\\n\\n total_space_indents = sum(indent_diff_counts.values())\\n \\n # 如果完全没有任何有效的缩进或者缩进完全被 tab 主导，则无法认定满足要求\\n if total_space_indents == 0:\\n return False\\n \\n if tab_count > total_space_indents:\\n return False\\n\\n # 找到出现频次最高（众数）的缩进增量，即被设定的缩进宽度\\n most_common_diff = max(indent_diff_counts.items(), key=lambda x: x[1])[0]\\n \\n return most_common_diff == 2\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，“=”的两边均需存在两个空格\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码和文本文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html', '.htm', '.json', '.md', '.txt', '.py', '.sh'}\\n \\n # 匹配所有的单独等号，排除 ==, ===, !=, !==, <=, >=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~=, ??= 等\\n pattern = r'(?<![=!<>+\\\\-*/%&|^~?])=(?![=>])'\\n has_checked_any_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略依赖和版本控制等目录\\n for skip_dir in ['.git', 'node_modules', 'dist', 'build']:\\n if skip_dir in dirs:\\n dirs.remove(skip_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n has_checked_any_file = True\\n matches = list(re.finditer(pattern, content))\\n \\n for match in matches:\\n start, end = match.span()\\n left_str = content[:start]\\n right_str = content[end:]\\n \\n # 检查左边是否恰好有两个空格（末尾是以2个空格结尾，且不是以3个空格结尾）\\n if not left_str.endswith(' ') or left_str.endswith(' '):\\n return False\\n \\n # 检查右边是否恰好有两个空格（开头是以2个空格开头，且不是以3个空格开头）\\n if not right_str.startswith(' ') or right_str.startswith(' '):\\n return False\\n\\n # 至少检查过一个文件，并且全部文件中的等号都符合要求才能算成功\\n return has_checked_any_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来我们在“我的歌单”页面实现核心的歌曲搜索与添加逻辑。请在页面上方增加一个音乐搜索框，支持输入关键字模拟搜索歌曲（请在代码里内置一批Mock数据，包含歌名、歌手、时长、流派、BPM和封面图片）。搜索结果需要以列表形式展示，并且每条结果右侧有一个“+”按钮。点击“+”可以将该歌曲添加至下方的“当前歌单”表格中。在表格底部，需要实时计算并显示当前歌单的总歌曲数和总播放时长。另外，后续所有轮次的输出都请遵循以下更新要求：切记在生成的代码文件中，“=”的左右两侧不能包含空格；在你生成的所有代码文件中，注释内容必须统一使用中文编写。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n # React项目常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明文件不是 UTF-8 编码\\n return False\\n except Exception:\\n # 其它异常暂时忽略，或视作读取失败\\n pass\\n\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"def check_file_name_prefix(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 验证工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 基于React项目推测可能的代码文件扩展名\\n code_extensions = {\\\".js\\\", \\\".jsx\\\", \\\".ts\\\", \\\".tsx\\\", \\\".css\\\", \\\".html\\\"}\\n \\n # 忽略第三方库、构建产物或版本控制等非模型直接手写的目录\\n exclude_dirs = {\\\".git\\\", \\\"node_modules\\\", \\\"build\\\", \\\"dist\\\", \\\"public\\\", \\\".next\\\", \\\"coverage\\\"}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n # 若存在任何一个代码文件不以 'dev_' 开头，则判定不满足\\n if not file.startswith(\\\"dev_\\\"):\\n return False\\n \\n # 如果项目下完全没有代码文件，也不能视作满足\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_files = []\\n # 检查项目目录是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 获取所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.css', '.html')):\\n code_files.append(os.path.join(root, file))\\n \\n if not code_files:\\n return False\\n\\n has_vars = False\\n \\n # 排除常见的保留关键字\\n reserved_words = {\\n 'const', 'let', 'var', 'function', 'class', 'import', 'export', 'default', \\n 'return', 'if', 'else', 'for', 'while', 'switch', 'case', 'break', 'continue', \\n 'true', 'false', 'null', 'undefined', 'new', 'this', 'super', 'yield', 'await', \\n 'async', 'typeof', 'instanceof', 'in', 'void', 'delete', 'catch', 'try', 'finally',\\n 'do', 'with', 'debugger', 'extends', 'implements', 'interface', 'package',\\n 'private', 'protected', 'public', 'static', 'enum', 'type'\\n }\\n\\n for file in code_files:\\n try:\\n with open(file, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 移除单行和多行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n # 移除字符串内的内容（避免误将字符串内的文字识别为变量）\\n content = re.sub(r'\\\".*?\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'.*?'\\\", \\\"''\\\", content)\\n content = re.sub(r'`.*?`', '``', content, flags=re.DOTALL)\\n \\n variables = set()\\n \\n # 1. 提取普通变量声明\\n simple_vars = re.findall(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n variables.update(simple_vars)\\n \\n # 2. 提取解构声明（如数组和对象的解构）\\n destruct_blocks = re.findall(r'\\\\b(?:const|let|var)\\\\s*([\\\\[\\\\{][^;]*?[\\\\]\\\\}])\\\\s*=', content)\\n for block in destruct_blocks:\\n words = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', block)\\n variables.update(words)\\n \\n # 3. 提取函数和类声明\\n func_vars = re.findall(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n variables.update(func_vars)\\n \\n class_vars = re.findall(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n variables.update(class_vars)\\n \\n # 4. 提取函数参数（普通函数参数）\\n func_args_blocks = re.findall(r'\\\\bfunction\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*\\\\(([^;\\\\{]*?)\\\\)', content)\\n for block in func_args_blocks:\\n words = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', block)\\n variables.update(words)\\n \\n # 5. 提取箭头函数的多参数\\n arrow_args_blocks = re.findall(r'\\\\(([^;\\\\{]*?)\\\\)\\\\s*=>', content)\\n for block in arrow_args_blocks:\\n words = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', block)\\n variables.update(words)\\n \\n # 6. 提取箭头函数的单参数\\n arrow_single_args = re.findall(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', content)\\n variables.update(arrow_single_args)\\n \\n # 过滤掉保留字与系统关键词\\n variables = variables - reserved_words\\n \\n if variables:\\n has_vars = True\\n for var in variables:\\n # 若发现存在任意变量名长度为偶数的情况，直接判定未满足要求\\n if len(var) % 2 == 0:\\n return False\\n \\n # 所有变量皆经过检验且文件内确实定义了变量\\n return has_vars\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据指令中React应用的上下文，推测并定义可能的代码文件后缀名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的依赖目录和隐藏目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对可能的代码文件进行检查\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计 \\\"TODO(dev)\\\" 标记出现的次数\\n if content.count('TODO(dev)') != 3:\\n return False\\n except Exception:\\n # 如果由于任何原因（如编码问题或权限不足）导致文件无法读取，也视为未满足条件\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所提供的代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def check_indentation(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n all_lines = []\\n\\n # 1. 从 response 中提取代码块内容\\n code_blocks = re.findall(r'```[a-zA-Z]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n all_lines.extend(block.split('\\\\n'))\\n\\n # 2. 从 workspace_path 中提取常见代码文件内容\\n if os.path.exists(workspace_path):\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json', '.py'}\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n all_lines.extend(f.readlines())\\n except Exception:\\n pass\\n \\n # 如果完全没有获取到任何代码行，视为未违背规则\\n if not all_lines:\\n return True\\n\\n indents = []\\n \\n # 3. 逐行检查缩进规则\\n for line in all_lines:\\n stripped = line.lstrip()\\n # 跳过空行和以 '*' 开头的多行注释格式行\\n if not stripped or stripped.startswith('*'):\\n continue\\n \\n # 提取前导空白字符\\n leading_whitespace = line[:len(line) - len(stripped)]\\n \\n # 严禁使用制表符\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n num_spaces = len(leading_whitespace)\\n if num_spaces > 0:\\n # 如果存在非偶数的空格数（如 1, 3, 5），显然不是 2 个空格缩进\\n if num_spaces % 2 != 0:\\n return False\\n indents.append(num_spaces)\\n\\n # 如果代码全部都是零缩进（即顶格写），不违反原则\\n if not indents:\\n return True\\n\\n # 4. 判断是否可能错用了 4 个空格的缩进\\n has_two_indent = any(i % 2 == 0 and i % 4 != 0 for i in indents)\\n has_four_indent = any(i % 4 == 0 for i in indents)\\n\\n # 如果全局只存在 4, 8, 12 等 4 的倍数的缩进，却毫无 2, 6, 10 等缩进，则实质上使用的是 4 空格缩进\\n if has_four_indent and not has_two_indent:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"切记在生成的代码文件中，“=”的左右两侧不能包含空格\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对React项目的常见代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\\n \\n if not os.path.exists(workspace_path):\\n # 若目录不存在，无法证明违反了要求\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉 node_modules 等干扰目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 判断等号的左侧或右侧是否出现了空格\\n if \\\" =\\\" in content or \\\"= \\\" in content:\\n return False\\n except Exception:\\n # 对于不可读的文件或编码问题，直接跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来完善播放列表的管理体验和底部播放器UI。首先，请让“当前歌单”表格里的歌曲支持拖拽排序（Drag and Drop），并且可以在列表里直接点击删除某首歌曲。其次，在整个页面的最底部固定一个“正在播放”的控制栏，包含歌曲封面、歌名、播放/暂停、上一首/下一首按钮，以及一个可拖动的进度条。当用户点击“当前歌单”中的某首歌曲时，底部控制栏要同步更新为该歌曲的信息，并自动切换为播放状态（UI表现即可，无需真实音频播放）。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的代码注释统一使用中文的规则作废，改为生成的各个代码文件内，注释部分需完全使用英文撰写；其次，在你生成的各个代码文件中，必须包含正好 2 个 箭头运算符(=>)，不可多也不可少。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义React项目及前端项目中常见代码文件的扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n has_code_file = False\\n \\n # 遍历生成代码所在的项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略可能的依赖包目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n _, ext = os.path.splitext(file)\\n # 仅验证代码文件\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制格式读取文件，以检查原始字节\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试强制用utf-8进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出UnicodeDecodeError解码异常，说明不符合UTF-8编码要求\\n return False\\n except Exception:\\n # 发生其他无法读取的问题暂不视为编码错误\\n pass\\n \\n # 如果找到了至少一个代码文件并且没有发生解码错误，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"def check_code_file_prefix(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端和常用相关的“代码文件”扩展名集合\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', \\n '.less', '.html', '.htm', '.py', '.sh'\\n }\\n \\n has_code_file = False\\n \\n # 若工作区路径不存在，说明没有生成代码项目\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖、构建输出和版本控制目录，避免误判内部原本就存在的文件\\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', 'build', 'dist', '__pycache__', '.next')]\\n \\n for file in files:\\n # 忽略隐藏文件（例如 .gitignore, .eslintrc 等）\\n if file.startswith('.'):\\n continue\\n \\n # 获取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否属于代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n # 检查文件名是否以 'dev_' 作为开头\\n if not file.startswith('dev_'):\\n return False\\n \\n # 如果项目中包含至少一个代码文件，并且所有代码文件都以 dev_ 开头，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_odd_length_variables(response: str, workspace_path: str) -> bool:\\n # JavaScript/TypeScript 中常见的关键字，需要排除被误判为变量\\n JS_KEYWORDS = {\\n 'break', 'case', 'catch', 'class', 'const', 'continue', 'debugger', 'default', \\n 'delete', 'do', 'else', 'export', 'extends', 'finally', 'for', 'function', 'if', \\n 'import', 'in', 'instanceof', 'new', 'return', 'super', 'switch', 'this', 'throw', \\n 'try', 'typeof', 'var', 'void', 'while', 'with', 'yield', 'let', 'static', 'enum', \\n 'await', 'implements', 'package', 'protected', 'interface', 'private', 'public', \\n 'null', 'true', 'false', 'undefined'\\n }\\n\\n # 各类变量及函数定义的正则表达式\\n pattern_direct = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n pattern_array = re.compile(r'\\\\b(?:var|let|const)\\\\s*\\\\[(.*?)\\\\]')\\n pattern_object = re.compile(r'\\\\b(?:var|let|const)\\\\s*\\\\{(.*?)\\\\}')\\n pattern_func = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\((.*?)\\\\)')\\n pattern_class = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n pattern_arrow = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>')\\n pattern_arrow_paren = re.compile(r'\\\\((.*?)\\\\)\\\\s*=>')\\n\\n var_names = set()\\n has_files = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目文件，寻找 React 相关的前端代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n has_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 1. 匹配直接定义的变量 (const a = 1)\\n for match in pattern_direct.finditer(content):\\n var_names.add(match.group(1))\\n\\n # 2. 匹配数组解构中的变量 (const [a, b] = arr)\\n for match in pattern_array.finditer(content):\\n inner = match.group(1)\\n idents = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', inner)\\n var_names.update(idents)\\n\\n # 3. 匹配对象解构中的变量 (const { a, b } = obj)\\n for match in pattern_object.finditer(content):\\n inner = match.group(1)\\n idents = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', inner)\\n var_names.update(idents)\\n\\n # 4. 匹配函数名和函数参数 (function foo(a, b) {})\\n for match in pattern_func.finditer(content):\\n var_names.add(match.group(1))\\n inner = match.group(2)\\n idents = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', inner)\\n var_names.update(idents)\\n\\n # 5. 匹配类名 (class MyClass {})\\n for match in pattern_class.finditer(content):\\n var_names.add(match.group(1))\\n\\n # 6. 匹配无括号箭头函数参数 (a => {})\\n for match in pattern_arrow.finditer(content):\\n var_names.add(match.group(1))\\n\\n # 7. 匹配有括号箭头函数参数 ((a, b) => {})\\n for match in pattern_arrow_paren.finditer(content):\\n inner = match.group(1)\\n idents = re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', inner)\\n var_names.update(idents)\\n\\n valid_vars = set()\\n # 过滤出合法的用户自定义变量名（剔除关键字与纯数字常量等误匹配）\\n for v in var_names:\\n if v not in JS_KEYWORDS and not v.isdigit():\\n valid_vars.add(v)\\n\\n # 如果代码文件不存在或者没有发现任何变量，无法判断为满足条件\\n if not has_files or not valid_vars:\\n return False\\n\\n # 验证是否每一个提取出的变量名长度均为奇数\\n for v in valid_vars:\\n if len(v) % 2 == 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，则认为不满足\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义React项目中常见的代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n has_code_file = False\\n \\n # 遍历项目目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计 \\\"TODO(dev)\\\" 标记在该代码文件中出现的次数\\n if content.count(\\\"TODO(dev)\\\") != 3:\\n return False\\n except Exception:\\n # 如果文件读取出现异常（如编码错误等），视作未能正确验证，返回 False\\n return False\\n \\n # 必须至少验证过一个代码文件，且全部符合要求才算满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所提供的代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def check_indent_two_spaces(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的React项目相关代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n found_any_target_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤以 dev_ 开头的规定代码文件\\n if not file.startswith('dev_'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n found_any_target_file = True\\n\\n indent_2_6_10 = 0 # 不能被 4 整除的偶数缩进（如 2, 6, 10）\\n indent_4_8_12 = 0 # 能被 4 整除的偶数缩进（如 4, 8, 12）\\n indent_odd = 0 # 奇数缩进\\n tab_lines = 0\\n total_indented_lines = 0\\n\\n for line in lines:\\n stripped = line.lstrip()\\n # 忽略完全空行\\n if not stripped:\\n continue\\n # 忽略多行注释中以 '*' 开头的中间对齐行\\n if stripped.startswith('*'):\\n continue\\n\\n # 匹配行首的前导空白\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n spaces = match.group(1)\\n if '\\\\t' in spaces:\\n tab_lines += 1\\n else:\\n length = len(spaces)\\n if length % 2 != 0:\\n indent_odd += 1\\n elif length % 4 != 0:\\n indent_2_6_10 += 1\\n else:\\n indent_4_8_12 += 1\\n total_indented_lines += 1\\n\\n if total_indented_lines > 0:\\n # 1. 存在使用 Tab 进行的缩进\\n if tab_lines > 0:\\n return False\\n \\n # 2. 奇数前导空格的行占据主导地位（容忍少量因代码对齐产生的奇数空格）\\n if indent_odd > total_indented_lines * 0.5:\\n return False\\n \\n # 3. 只有 4 空格及其倍数的缩进，而完全不存在 2 空格及其特征倍数的缩进\\n if indent_4_8_12 > 0 and indent_2_6_10 == 0:\\n return False\\n\\n return found_any_target_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"切记在生成的代码文件中，“=”的左右两侧不能包含空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对React项目的常见代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less'}\\n # 匹配等号左侧或右侧存在空格（含普通空格、制表符、全角空格）的情况\\n pattern = re.compile(r'[ \\\\t\\\\u3000]=|=[ \\\\t\\\\u3000]')\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若发现等号左右两侧存在空格，则返回False\\n if pattern.search(content):\\n return False\\n except Exception:\\n # 忽略读取异常的文件\\n continue\\n \\n # 如果没有找到任何代码文件，视作未满足要求\\n if not found_code_file:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 2 个 箭头运算符(=>)，不可多也不可少\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.go', '.rb', '.php', '.cs', '.swift'\\n }\\n \\n # 需要忽略的文件夹，防止扫描到项目依赖或编译产物中\\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'coverage', '.next'}\\n \\n file_found = False\\n all_files_valid = True\\n \\n # 1. 优先检查工作区生成的实际代码文件\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉忽略目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 无法按 utf-8 文本读取的文件直接跳过\\n continue\\n \\n # 检查是否恰好包含 2 个 '=>'\\n if content.count('=>') != 2:\\n all_files_valid = False\\n break\\n if not all_files_valid:\\n break\\n \\n # 如果工作区里存在至少一个代码文件，则以实际文件内容验证结果为准\\n if file_found:\\n return all_files_valid\\n \\n # 2. 如果工作区中未发现相关代码文件，作为降级方案检查回复文本中的代码块\\n lang_pattern = r'```(?:javascript|js|jsx|typescript|ts|tsx|css|html|python|java|cpp|c|go|ruby|php|swift|cs)\\\\n(.*?)```'\\n code_blocks = re.findall(lang_pattern, response, flags=re.DOTALL | re.IGNORECASE)\\n \\n if code_blocks:\\n for block in code_blocks:\\n if block.count('=>') != 2:\\n return False\\n return True\\n \\n # 既无代码文件，也无带有明确标识的代码块，视为未满足要求\\n return False\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步是实现“数据分析”面板。当用户在侧边栏点击切换到“数据分析”时，主区域需要读取刚刚在“我的歌单”里添加的那些歌曲数据，并展示两个可视化图表（你可以使用普通的DOM配合CSS模拟，或者使用图表库）：1. 一个饼图（Pie Chart），统计当前歌单中不同音乐流派（Genre）的占比；2. 一个柱状图或雷达图，展示当前歌单的BPM（节拍数）分布情况。如果当前歌单没有任何歌曲，请展示一个带有图标的“暂无数据，请先添加歌曲”的空状态提示。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的每个代码文件中正好包含2个箭头运算符(=>)的规则作废，改为每一个代码文件的生成标准是：必须精确包含 3 个 箭头运算符(=>)；其次，只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数），以上两条规则后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目背景推断可能的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\\n has_code_file = False\\n\\n # 遍历生成的项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选出代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制模式读取并验证UTF-8解码\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 解码报错说明非 UTF-8 编码，不符合要求\\n return False\\n except Exception:\\n # 若遇到其他读取错误（如无权限等）暂按不满足要求处理\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据React项目推断可能的代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', \\n '.html', '.htm'\\n }\\n \\n has_code_file = False\\n \\n # 遍历当前项目地址下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n \\n # 检查代码文件名是否以 'dev_' 开头\\n if not file.startswith('dev_'):\\n return False\\n \\n # 如果找到了代码文件，并且所有代码文件都以 dev_ 开头，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义常见的JavaScript/TypeScript关键字，防止被误判为自定义变量名\\n js_keywords = {\\n 'const', 'let', 'var', 'if', 'else', 'function', 'return', 'class', \\n 'import', 'from', 'export', 'default', 'true', 'false', 'null', 'undefined',\\n 'typeof', 'instanceof', 'new', 'await', 'async', 'yield', 'type', 'interface'\\n }\\n \\n variable_names = set()\\n \\n # 编译用于提取变量声明的正则表达式\\n # 1. 提取基础声明 (如 let myVar, const myConst, function myFunc, class MyClass)\\n simple_var_re = re.compile(r'\\\\b(?:const|let|var|function|class)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n # 2. 提取数组解构声明 (如 const [a, b] = ...)\\n arr_destruct_re = re.compile(r'\\\\b(?:const|let|var)\\\\s+\\\\[(.*?)\\\\]\\\\s*=')\\n # 3. 提取对象解构声明 (如 const {x, y} = ...)\\n obj_destruct_re = re.compile(r'\\\\b(?:const|let|var)\\\\s+\\\\{(.*?)\\\\}\\\\s*=')\\n \\n found_js_files = False\\n \\n # 遍历生成项目的文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n found_js_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 提前移除代码中的注释，防止正则将注释中的单词当作变量\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n # 匹配普通声明\\n for match in simple_var_re.finditer(content):\\n var_name = match.group(1)\\n if var_name not in js_keywords:\\n variable_names.add(var_name)\\n \\n # 匹配数组解构\\n for match in arr_destruct_re.finditer(content):\\n inner_content = match.group(1)\\n # 提取解构块中的所有纯标识符\\n for word in re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', inner_content):\\n if word not in js_keywords:\\n variable_names.add(word)\\n \\n # 匹配对象解构\\n for match in obj_destruct_re.finditer(content):\\n inner_content = match.group(1)\\n # 提取解构块中的所有纯标识符\\n for word in re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', inner_content):\\n if word not in js_keywords:\\n variable_names.add(word)\\n except Exception:\\n # 忽略无法读取的文件错误\\n pass\\n \\n # 如果项目中不存在 JS/TS 代码文件或没有提取出任何变量，则判断为不满足要求\\n if not found_js_files or not variable_names:\\n return False\\n \\n # 检查所有提取的变量名的长度是否都满足奇数要求\\n # （这里的“长度”在编程范畴中表示字符的数量，直接使用len函数进行验证即可）\\n for var in variable_names:\\n if len(var) % 2 == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"import os\\n\\ndef check_todo_dev_count(response: str, workspace_path: str) -> bool:\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义React项目常见的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\\n \\n code_file_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的常见依赖或版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计 \\\"TODO(dev)\\\" 出现的次数\\n if content.count('TODO(dev)') != 3:\\n return False\\n except Exception:\\n # 如果文件读取异常，则认为该文件不满足要求\\n return False\\n \\n return code_file_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所提供的代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义 React 项目中常见的代码文件后缀\\n exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n has_any_indent = False\\n \\n # 优先检查工作目录中的实际代码文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n indents = set()\\n for line in lines:\\n stripped = line.lstrip()\\n # 忽略空行以及常见的注释行对齐干扰\\n if not stripped or stripped.startswith('*') or stripped.startswith('/'):\\n continue\\n if stripped.startswith('<--') or stripped.startswith('<!--'):\\n continue\\n \\n leading_str = line[:len(line) - len(stripped)]\\n # 严禁使用 tab 缩进\\n if '\\\\t' in leading_str:\\n return False\\n \\n spaces = len(leading_str)\\n if spaces > 0:\\n indents.add(spaces)\\n \\n # 如果文件内存在缩进\\n if indents:\\n has_any_indent = True\\n # 如果存在大于等于2个空格的缩进，但完全没有2个空格的缩进，说明不是以2个空格为基准的缩进\\n if 2 not in indents and any(i >= 2 for i in indents):\\n return False\\n\\n if has_any_indent:\\n return True\\n \\n # 如果工作目录中未找到带有缩进的代码文件，作为 fallback 检查回复文本中的代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9]*[ \\\\t]*\\\\n(.*?)```', response, re.DOTALL)\\n if not code_blocks:\\n return False\\n \\n has_any_indent_in_resp = False\\n for block in code_blocks:\\n indents = set()\\n for line in block.split('\\\\n'):\\n stripped = line.lstrip()\\n if not stripped or stripped.startswith('*') or stripped.startswith('/'):\\n continue\\n if stripped.startswith('<--') or stripped.startswith('<!--'):\\n continue\\n \\n leading_str = line[:len(line) - len(stripped)]\\n if '\\\\t' in leading_str:\\n return False\\n \\n spaces = len(leading_str)\\n if spaces > 0:\\n indents.add(spaces)\\n \\n if indents:\\n has_any_indent_in_resp = True\\n if 2 not in indents and any(i >= 2 for i in indents):\\n return False\\n \\n return has_any_indent_in_resp\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"切记在生成的代码文件中，“=”的左右两侧不能包含空格\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 基于React项目推断可能的代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 判断文件中是否存在 '=' 左右带有空格的情况\\n if ' =' in content or '= ' in content:\\n return False\\n except Exception:\\n # 读取异常时跳过\\n pass\\n\\n return found_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含 3 个 箭头运算符(=>)\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n code_files_count = 0\\n \\n # 定义React项目中常见的前端代码文件后缀\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n # 定义需要排除的非代码文件后缀（如日志、配置及各类静态资源）\\n exclude_exts = ('.log', '.json', '.md', '.png', '.jpg', '.jpeg', '.svg', '.gif')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 排除非代码文件（例如用户指令要求生成的 change_log_X.log 等）\\n if file.lower().endswith(exclude_exts):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判定为代码文件的条件：符合前端代码的常见后缀，或是严格遵循指令前缀 'dev_'\\n if ext in code_exts or file.startswith('dev_'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 验证当前代码文件中是否精确包含 3 个 '=>' 运算符\\n if content.count('=>') != 3:\\n return False\\n \\n code_files_count += 1\\n except Exception:\\n # 如果遇到无法以纯文本读取的异常文件，跳过即可\\n pass\\n\\n # 至少要验证过一个以上的代码文件才算真正完成要求\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"很棒，现在我们来开发“发现社区”功能。点击侧边栏的“发现社区”后，以卡片瀑布流或网格形式展示其他用户分享的公开歌单（需Mock几条带用户头像、歌单名、点赞数的歌单数据）。为每个歌单卡片实现点赞功能（点击爱心图标，点赞数+1并变为红色高亮），并在卡片底部提供一个“查看评论”的折叠面板，点击展开后能看到一条默认评论，并包含一个简易的输入框，允许用户提交新评论并实时追加显示在下方。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的每一个代码文件必须精确包含3个箭头运算符(=>)的规则作废，改为每一个代码文件的生成标准是：必须精确包含 1 个 \\\"TODO(human)\\\"标记；其次，在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划，以上两条规则后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 结合 React 项目推断的常见代码/文本文件后缀\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', \\n '.less', '.html', '.json', '.md', '.log', '.sh'\\n }\\n \\n # 遍历时需要忽略的无用/依赖目录\\n ignore_dirs = {'node_modules', '.git', 'build', 'dist', '.next'}\\n\\n code_files_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以过滤掉不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断当前文件是否为需要检查的代码文件（基于后缀，或是前序要求的 dev_ 前缀）\\n if ext in code_extensions or file.startswith('dev_'):\\n file_path = os.path.join(root, file)\\n \\n # 排除可能同名的非文件实体\\n if not os.path.isfile(file_path):\\n continue\\n \\n # 针对非代码的纯二进制或多媒体格式做防误伤拦截\\n if ext in {'.png', '.jpg', '.jpeg', '.gif', '.ico', '.mp3', '.mp4'}:\\n continue\\n \\n code_files_found = True\\n \\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试使用 UTF-8 进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 解码失败，表明不是有效的 UTF-8 编码文件\\n return False\\n except Exception:\\n # 忽略其他读取层面（如权限）抛出的非编码异常\\n pass\\n\\n # 只要找到了代码文件且均未发生解码错误，即判定为符合要求\\n return code_files_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.scss', '.less', '.sass'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略第三方库依赖、版本控制以及构建输出目录\\n for ignore_dir in ['node_modules', '.git', 'build', 'dist', '.next']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in code_extensions:\\n has_code_file = True\\n # 若存在不以 'dev_' 开头的代码文件，则不满足要求\\n if not file.startswith('dev_'):\\n return False\\n\\n # 必须有生成代码文件，并且全部符合条件才算满足要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n has_code_file = False\\n vars_checked = 0\\n\\n # 预编译常见的变量/函数/类声明匹配正则\\n # 1. 匹配普通单变量: let/const/var variableName = ... 或 let variableName;\\n p1 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*[=;]')\\n \\n # 2. 匹配函数声明: function functionName(...)\\n p2 = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n \\n # 3. 匹配类声明: class ClassName { ... 或 class ClassName extends ...\\n p3 = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:extends|\\\\{)')\\n \\n # 4. 匹配数组解构（双变量，如React useState）: const [foo, bar] = ...\\n p4 = re.compile(r'\\\\b(?:const|let|var)\\\\s*\\\\[\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*,\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\]\\\\s*=')\\n \\n # 5. 匹配数组解构（单变量）: const [foo] = ...\\n p5 = re.compile(r'\\\\b(?:const|let|var)\\\\s*\\\\[\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\]\\\\s*=')\\n\\n # JavaScript 常见的需要忽略的保留关键字\\n reserved = {\\n 'let', 'const', 'var', 'function', 'class', 'extends', 'true', 'false', \\n 'null', 'undefined', 'default', 'export', 'import', 'return', 'yield', \\n 'await', 'async', 'if', 'else', 'for', 'while', 'switch', 'case', 'type', 'interface'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖库和版本控制的文件夹\\n if 'node_modules' in root.split(os.sep) or '.git' in root.split(os.sep):\\n continue\\n \\n for file in files:\\n # 根据系统提示中的要求，AI生成的代码必须以 dev_ 开头，且本项目属于React\\n if file.startswith('dev_') and file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 移除字符串字面量（简单处理转义，避免跨行误伤）\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", code)\\n code = re.sub(r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`\\\", \\\"``\\\", code)\\n \\n # 移除单行及多行注释\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n # 提取此文件内的所有确定为变量名的标识符\\n vars_found = []\\n vars_found.extend([m.group(1) for m in p1.finditer(code)])\\n vars_found.extend([m.group(1) for m in p2.finditer(code)])\\n vars_found.extend([m.group(1) for m in p3.finditer(code)])\\n \\n for m in p4.finditer(code):\\n vars_found.append(m.group(1))\\n vars_found.append(m.group(2))\\n \\n vars_found.extend([m.group(1) for m in p5.finditer(code)])\\n \\n for v in vars_found:\\n if v not in reserved:\\n vars_checked += 1\\n # 要求每个变量名称长度为奇数，如发现偶数长度则校验失败\\n if len(v) % 2 == 0:\\n return False\\n\\n # 若未找到任何受检查的代码文件，或代码文件中没声明任何常规变量，视为不满足\\n if not has_code_file or vars_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义 React 项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\\n found_code_file = False\\n \\n # 遍历项目文件系统\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理判定为代码的文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计 \\\"TODO(dev)\\\" 标记数量，要求正好等于 3 个\\n if content.count(\\\"TODO(dev)\\\") != 3:\\n return False\\n \\n except Exception:\\n # 如果文件无法正常读取，视作未满足要求\\n return False\\n \\n # 要求至少存在一个代码文件且全部满足条件\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所提供的代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义需要检查缩进的常规代码文件拓展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.json'}\\n all_files_ok = True\\n has_checked_any = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n indentations = []\\n in_multiline_comment = False\\n \\n for line in lines:\\n stripped = line.strip()\\n # 过滤空行\\n if not stripped:\\n continue\\n \\n # 简单过滤多行注释块，避免由于星号对齐导致的奇数空格干扰缩进统计\\n if stripped.startswith('/*'):\\n in_multiline_comment = True\\n if in_multiline_comment:\\n if '*/' in stripped:\\n in_multiline_comment = False\\n continue\\n if stripped.startswith('*'):\\n continue\\n \\n # 获取行首的空白字符部分\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 如果行首缩进发现 Tab 键，说明不符合强制的“2个空格”要求\\n if '\\\\t' in leading_whitespace:\\n all_files_ok = False\\n break\\n \\n leading_spaces = len(leading_whitespace)\\n indentations.append(leading_spaces)\\n \\n if not all_files_ok:\\n break\\n \\n # 计算相邻非空行的正向缩进差值（即向右缩进时的层级跨度）\\n positive_diffs = []\\n for i in range(1, len(indentations)):\\n diff = indentations[i] - indentations[i-1]\\n if diff > 0:\\n positive_diffs.append(diff)\\n \\n if positive_diffs:\\n has_checked_any = True\\n # 统计出现频率最高的正向缩进差值，以此代表整个文件的基准缩进单位\\n counter = Counter(positive_diffs)\\n most_common_diff = counter.most_common(1)[0][0]\\n \\n # 如果最常用的缩进增加跨度不是 2，则该文件不符合2个空格的要求\\n if most_common_diff != 2:\\n all_files_ok = False\\n break\\n \\n if not all_files_ok:\\n break\\n\\n # 只有当且仅当所有包含缩进的受检代码文件均通过测试，且确有文件被验证过时，才返回 True\\n return all_files_ok and has_checked_any\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"切记在生成的代码文件中，“=”的左右两侧不能包含空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据React项目的特点，推断出可能的代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.scss', '.less'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 判断 \\\"=\\\" 的左右两侧是否包含空格\\n # 不论是赋值等号 (=)，还是全等 (===)，或箭头函数 (=>)，只要 \\\"=\\\" 旁有空格就会被以下判断捕获\\n if ' =' in content or '= ' in content:\\n return False\\n except Exception:\\n # 遇到文件无法读取等情况跳过\\n continue\\n \\n # 如果有代码文件且都没有违反规则，返回True；否则（包括没有代码文件的情况）返回False\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含 1 个 \\\"TODO(human)\\\"标记\",\n \"validation_code\": \"import os\\n\\ndef check_todo_human_requirement(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', \\n '.py', '.java', '.cpp', '.c', '.h', '.hpp', '.cs', '.go', '.rb', \\n '.php', '.swift', '.kt'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的无关/第三方依赖/构建结果目录\\n for ignore_dir in ['node_modules', '.git', 'dist', 'build']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n # 如果无法按utf-8读取，可能未遵循之前的UTF-8编码要求，或者遇到了非文本文件\\n return False\\n except Exception:\\n return False\\n \\n # 统计当前代码文件中 \\\"TODO(human)\\\" 的出现次数，必须精确包含1个\\n if content.count('TODO(human)') != 1:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后，我们回到“我的歌单”页面，增加一些高级的交互细节。在“当前歌单”表格的上方，新增一个高级筛选栏，支持按“流派”下拉框和“BPM区间（如慢速、中速、快速）”单选框对已添加的歌曲进行多条件交叉过滤。另外，给表格中的每一行歌曲增加自定义的右键菜单（Context Menu）：禁用浏览器默认右键菜单，右键点击某首歌曲时在其鼠标位置弹出一个自定义菜单，包含“设为下一首播放”、“分享”和“移除”三个选项，点击任意选项后需要弹出对应的全局Toast消息提示。另外，后续所有轮次的输出都请遵循以下更新要求：首先，此前要求的代码文件注释需完全使用英文的规则作废，改为生成的各个代码文件内，注释内容必须统一使用中文编写；其次，生成的每个代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-，以上两条规则后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据前端React项目及多轮指令中的要求，定义常见的代码及文本文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', \\n '.html', '.json', '.py', '.sh', '.md', '.txt', '.log'\\n }\\n \\n file_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出可能是生成的代码或文本文件\\n if ext in code_extensions or file.lower() in ['dockerfile', 'makefile']:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试使用utf-8进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明并非使用UTF-8进行编码\\n return False\\n except Exception:\\n # 若遇到其他读取异常，保守起见视同验证失败\\n return False\\n\\n # 若未找到任何代码文件，则不符合基本的生成要求；若都顺利通过解码，则满足要求。\\n return file_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你给出的所有代码文件，在命名时必须加上 dev_ 作为开头\",\n \"validation_code\": \"import os\\n\\ndef check_file_naming_requirement(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方库代码目录，以防止误判（如果是真实React项目）\\n path_parts = root.replace('\\\\\\\\', '/').split('/')\\n if 'node_modules' in path_parts:\\n continue\\n\\n for file_name in files:\\n # 提取文件扩展名并统一转换为小写用于匹配\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 如果是代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n # 判断代码文件名是否以 dev_ 开头\\n if not file_name.startswith('dev_'):\\n return False\\n \\n # 如果遍历完毕未发现违规文件，并且至少存在一个代码文件，则认为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"规定搜索必须通过正则表达式进行，禁止使用简单字符串匹配\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你要生成的代码文件中，请让每一个变量名称的长度正好满足奇数条件\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n def strip_comments_and_strings(text):\\n # 使用正则表达式一次性匹配并移除字符串和注释，防止由于匹配顺序不同引发的嵌套影响\\n pattern = r'/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*|#.*|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n return re.sub(pattern, '', text)\\n\\n code_file_extensions = {'.js', '.jsx', '.ts', '.tsx', '.py'}\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_file_extensions:\\n continue\\n \\n has_code_files = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 去除注释与字符串字面量\\n content = strip_comments_and_strings(content)\\n variables = set()\\n \\n # 1. 正常的 var, let, const 声明\\n matches = re.findall(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n variables.update(matches)\\n \\n # 2. 数组解构 (例如: const [state, setState] = useState() )\\n arr_matches = re.findall(r'\\\\b(?:var|let|const)\\\\s*\\\\[(.*?)\\\\]', content, re.DOTALL)\\n for m in arr_matches:\\n for item in m.split(','):\\n item = item.strip()\\n if not item:\\n continue\\n var_name = item.split('=')[0].strip()\\n var_name = var_name.replace('...', '').strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n variables.add(var_name)\\n \\n # 3. 对象解构 (例如: const { data, loading } = useQuery() )\\n obj_matches = re.findall(r'\\\\b(?:var|let|const)\\\\s*\\\\{(.*?)\\\\}', content, re.DOTALL)\\n for m in obj_matches:\\n for item in m.split(','):\\n item = item.strip()\\n if not item:\\n continue\\n # 处理可能存在的重命名，例如：a: b\\n if ':' in item:\\n var_name = item.split(':')[1].split('=')[0].strip()\\n else:\\n var_name = item.split('=')[0].strip()\\n var_name = var_name.replace('...', '').strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n variables.add(var_name)\\n \\n # 4. 函数形参\\n func_matches = re.findall(r'\\\\bfunction\\\\s*[a-zA-Z0-9_$]*\\\\s*\\\\((.*?)\\\\)', content, re.DOTALL)\\n for m in func_matches:\\n for item in m.split(','):\\n item = item.strip()\\n if not item:\\n continue\\n var_name = item.split('=')[0].split(':')[0].strip()\\n var_name = var_name.replace('...', '').strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n variables.add(var_name)\\n \\n # 5. 普通的起始赋值语句 (兼顾 Python 或 JS 类属性赋值)\\n assign_matches = re.findall(r'^[ \\\\t]*([a-zA-Z_$][a-zA-Z0-9_$]*)[ \\\\t]*=[ \\\\t]*(?![=>])', content, re.MULTILINE)\\n for m in assign_matches:\\n variables.add(m)\\n \\n # 安全过滤常见语言的关键字，以免误判\\n keywords = {\\n 'true', 'false', 'null', 'undefined', 'import', 'export', 'default', 'class', \\n 'function', 'return', 'if', 'else', 'for', 'while', 'let', 'const', 'var', \\n 'await', 'async', 'yield', 'switch', 'case', 'break', 'continue', 'catch', \\n 'try', 'finally', 'with', 'debugger', 'typeof', 'instanceof', 'new', 'void', \\n 'delete', 'in', 'of', 'this', 'super'\\n }\\n \\n for v in variables:\\n if v in keywords:\\n continue\\n # 判断每一个变量的长度是否不是奇数（即偶数）\\n if len(v) % 2 == 0:\\n return False\\n \\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(dev)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def check_todo_dev_tags(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n \\n # 需要排除的常见非代码目录或依赖目录\\n exclude_dirs = {'node_modules', '.git', 'build', 'dist', 'public'}\\n \\n code_files_count = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n code_files_count += 1\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 统计 \\\"TODO(dev)\\\" 的出现次数\\n if content.count('TODO(dev)') != 3:\\n return False\\n except Exception:\\n # 若文件读取失败（例如编码错误等异常），无法确认满足，安全起见返回False\\n return False\\n \\n # 如果项目中没有生成任何代码文件，认为要求未完成\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所提供的代码文件，其缩进宽度需设置为2个空格\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n \\n # 根据项目的要求，推断可能生成的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.py', '.json'}\\n code_files = []\\n \\n # 查找所有AI生成的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据系统要求，生成的文件必须以 dev_ 开头\\n if file.startswith('dev_'):\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何相关的代码文件，属于未能正确完成生成任务\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n indent_diffs = []\\n prev_indent = 0\\n \\n for line in lines:\\n if not line.strip():\\n continue\\n \\n stripped = line.lstrip()\\n # 忽略常见的多行或单行注释对缩进判断的排版干扰\\n if stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('#') or stripped.startswith('<!--'):\\n continue\\n \\n leading_whitespace = line[:len(line) - len(stripped)]\\n \\n # 缩进使用了tab键而不是空格，不满足要求\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n curr_indent = len(leading_whitespace)\\n # 如果发生了缩进的增加，则记录其增加量（这代表了一级缩进的宽度）\\n if curr_indent > prev_indent:\\n indent_diffs.append(curr_indent - prev_indent)\\n \\n prev_indent = curr_indent\\n \\n # 如果文件中存在缩进增加行为，统计最基础的缩进级别\\n if indent_diffs:\\n counter = Counter(indent_diffs)\\n most_common = counter.most_common(1)[0][0]\\n # 如果出现最多次数的缩进宽度不是 2 个空格，判定为不符合\\n if most_common != 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"切记在生成的代码文件中，“=”的左右两侧不能包含空格\",\n \"validation_code\": \"def check_equals_no_spaces(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义前端React项目及Python可能用到的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.py'}\\n\\n # 遍历项目下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理代码文件，忽略如 .md, .log, .json 等非程序代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于某种原因无法读取文件，跳过\\n continue\\n \\n # 遍历文件内容，寻找等号\\n for i, char in enumerate(content):\\n if char == '=':\\n # 检查等号左侧是否为空格\\n if i > 0 and content[i - 1] == ' ':\\n return False\\n \\n # 检查等号右侧是否为空格\\n if i < len(content) - 1 and content[i + 1] == ' ':\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含 1 个 \\\"TODO(human)\\\"标记\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义前端React项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能存在的第三方依赖或Git版本库目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n # 仅针对代码文件进行检查\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n # 如果无法按系统要求的UTF-8编码读取，亦视为不符合规范或非正常代码文件\\n return False\\n \\n # 统计标记字符串出现的次数\\n if content.count('TODO(human)') != 1:\\n return False\\n \\n # 必须至少存在一个代码文件且全部满足条件，才算通过要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"只要项目被改动过，就要新建一个用来记录当次变动的文件，文件需命名为 change_log_X.log（X 表示自 0 开始编号的改动次数）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成的每个代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 根据用户指令推断可能的前端代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n has_code_file = False\\n \\n # 优先检查工作区中的实际代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的非代码目录，避免干扰\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含规定的编码格式字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n return False\\n \\n # 如果在工作区中找到了代码文件，并且全部通过了首行检查，则返回 True\\n if has_code_file:\\n return True\\n \\n # 如果工作区中没有提取到代码文件，则作为备用降级方案，检查回复内容中的代码块\\n # 匹配带有前端代码类型标记的 Markdown 代码块\\n code_blocks = re.findall(r'```(js|jsx|ts|tsx|css|html|react)?\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n has_code_block = False\\n for lang, block in code_blocks:\\n lang = lang.strip().lower()\\n if lang in {'js', 'jsx', 'ts', 'tsx', 'css', 'html', 'react'}:\\n has_code_block = True\\n lines = block.strip().split('\\\\n')\\n # 检查代码块的真实首行是否包含要求的内容\\n if not lines or '-*- coding: utf-8 -*-' not in lines[0]:\\n return False\\n \\n # 如果没有工作区文件，但回复中包含对应的代码块且全部符合要求\\n if has_code_block:\\n return True\\n \\n # 既没有实际文件，也没有生成对应的代码块，视作没有满足生成代码文件的要求\\n return False\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面初始布局：截屏验证页面顶部是否包含应用Logo和用户头像组成的导航栏，页面左侧是否包含'我的歌单'、'数据分析'、'发现社区'三个菜单项。","验证'新建歌单'弹窗功能：在'我的歌单'页面居中位置点击'新建歌单'按钮，验证屏幕上是否弹出一个居中的Modal弹窗，且弹窗内包含'歌单名称'输入框和一个'确认'按钮。","验证歌曲搜索展示：在'我的歌单'页面上方的音乐搜索框中输入任意关键字模拟搜索，截屏验证下方是否出现以列表形式展示的搜索结果，且搜索结果包含歌名、歌手、时长、流派、BPM和封面图片等信息，每一条结果的最右侧都有一个'+'按钮。","验证歌曲添加功能：在搜索结果列表中点击某条歌曲右侧的'+'按钮，截屏验证该首歌曲是否成功添加并显示在下方的'当前歌单'表格中，且表格中能看到歌曲名称、歌手、时长、流派、BPM等核心信息。","验证歌单总数与时长实时计算：通过搜索列表的'+'按钮向'当前歌单'表格中添加至少两首不同的歌曲，查看表格底部，验证系统是否正确显示出当前歌单的'总歌曲数'和累加计算出的'总播放时长'。","验证拖拽排序功能：在'当前歌单'表格中，鼠标左键按住一首歌曲并将其拖动至表格中的另一首歌曲上方后释放鼠标，验证两首歌曲在表格中的排列位置是否发生了改变。","验证歌曲删除功能：点击'当前歌单'表格中某首歌曲对应的删除操作按钮，验证该行歌曲是否从表格中消失，同时底部的总歌曲数是否对应减少。","验证底部播放器同步与状态切换：点击'当前歌单'表格中的某一首歌曲，查看页面最底部的播放控制栏，验证控制栏中显示的歌曲封面和歌名是否变为刚点击的歌曲信息，同时验证播放控制按钮是否自动切换为'暂停'图标（表示正在播放状态），并确认控制栏包含上一首和下一首按钮。","验证进度条拖动UI：在底部播放控制栏中，使用鼠标按住进度条滑块并向右拖动一段距离，释放鼠标后验证滑块是否停留在了拖拽结束的位置。","验证'数据分析'空状态：在清空'我的歌单'所有歌曲（或不添加任何歌曲）的情况下，点击侧边栏的'数据分析'菜单，验证主区域是否展示了一个带有图标的'暂无数据，请先添加歌曲'提示信息。","验证'数据分析'图表渲染：在'我的歌单'中添加至少三首不同流派和不同BPM的歌曲，然后点击侧边栏'数据分析'，验证主内容区域是否基于当前歌单数据正确渲染出两个可视化图表：一个是统计音乐流派占比的饼图，另一个是展示BPM分布的柱状图或雷达图。","验证'发现社区'页面渲染：点击侧边栏'发现社区'菜单，验证主内容区是否以卡片瀑布流或网格的形式，展示多条他人公开歌单卡片，且每张卡片包含用户头像、歌单名和点赞数等信息。","验证歌单点赞功能：在'发现社区'页面，点击某个歌单卡片上的爱心点赞图标，截屏验证该爱心图标是否变为红色高亮，并且其旁边的点赞数字相比点击前增加了1。","验证评论面板展开功能：点击'发现社区'某个歌单卡片底部的'查看评论'按钮，验证该区域是否向下展开，并显示出默认的评论内容以及一个简易的评论输入框。","验证发布新评论功能：在展开的评论折叠面板输入框中，输入一段测试文本并提交，验证提交后该段文本是否作为一条新评论实时追加显示在当前卡片的评论列表下方。","验证高级筛选多条件过滤功能：在'我的歌单'页面添加多首不同属性的歌曲后，确认'当前歌单'表格上方存在高级筛选栏，且包含'流派'下拉框和'BPM区间'单选项；选择一个流派并点击某个BPM区间，验证表格数据是否被刷新，仅显示同时符合所选流派和BPM区间的歌曲行。","验证自定义右键菜单触发：在'当前歌单'表格的任意一首歌曲行上点击鼠标右键，验证浏览器默认的右键菜单是否被屏蔽，且在鼠标点击位置是否弹出了包含'设为下一首播放'、'分享'和'移除'三个选项的自定义菜单。","验证右键菜单各选项行为：在弹出的自定义右键菜单中，分别点击'设为下一首播放'、'分享'和'移除'选项，验证每个选项都会弹出对应的全局Toast消息提示；点击'移除'后，还应确认该歌曲从当前歌单表格中消失。"],"string":"[\n \"查看页面初始布局：截屏验证页面顶部是否包含应用Logo和用户头像组成的导航栏，页面左侧是否包含'我的歌单'、'数据分析'、'发现社区'三个菜单项。\",\n \"验证'新建歌单'弹窗功能：在'我的歌单'页面居中位置点击'新建歌单'按钮，验证屏幕上是否弹出一个居中的Modal弹窗，且弹窗内包含'歌单名称'输入框和一个'确认'按钮。\",\n \"验证歌曲搜索展示：在'我的歌单'页面上方的音乐搜索框中输入任意关键字模拟搜索，截屏验证下方是否出现以列表形式展示的搜索结果，且搜索结果包含歌名、歌手、时长、流派、BPM和封面图片等信息，每一条结果的最右侧都有一个'+'按钮。\",\n \"验证歌曲添加功能：在搜索结果列表中点击某条歌曲右侧的'+'按钮，截屏验证该首歌曲是否成功添加并显示在下方的'当前歌单'表格中，且表格中能看到歌曲名称、歌手、时长、流派、BPM等核心信息。\",\n \"验证歌单总数与时长实时计算：通过搜索列表的'+'按钮向'当前歌单'表格中添加至少两首不同的歌曲，查看表格底部，验证系统是否正确显示出当前歌单的'总歌曲数'和累加计算出的'总播放时长'。\",\n \"验证拖拽排序功能：在'当前歌单'表格中，鼠标左键按住一首歌曲并将其拖动至表格中的另一首歌曲上方后释放鼠标，验证两首歌曲在表格中的排列位置是否发生了改变。\",\n \"验证歌曲删除功能：点击'当前歌单'表格中某首歌曲对应的删除操作按钮，验证该行歌曲是否从表格中消失，同时底部的总歌曲数是否对应减少。\",\n \"验证底部播放器同步与状态切换：点击'当前歌单'表格中的某一首歌曲，查看页面最底部的播放控制栏，验证控制栏中显示的歌曲封面和歌名是否变为刚点击的歌曲信息，同时验证播放控制按钮是否自动切换为'暂停'图标（表示正在播放状态），并确认控制栏包含上一首和下一首按钮。\",\n \"验证进度条拖动UI：在底部播放控制栏中，使用鼠标按住进度条滑块并向右拖动一段距离，释放鼠标后验证滑块是否停留在了拖拽结束的位置。\",\n \"验证'数据分析'空状态：在清空'我的歌单'所有歌曲（或不添加任何歌曲）的情况下，点击侧边栏的'数据分析'菜单，验证主区域是否展示了一个带有图标的'暂无数据，请先添加歌曲'提示信息。\",\n \"验证'数据分析'图表渲染：在'我的歌单'中添加至少三首不同流派和不同BPM的歌曲，然后点击侧边栏'数据分析'，验证主内容区域是否基于当前歌单数据正确渲染出两个可视化图表：一个是统计音乐流派占比的饼图，另一个是展示BPM分布的柱状图或雷达图。\",\n \"验证'发现社区'页面渲染：点击侧边栏'发现社区'菜单，验证主内容区是否以卡片瀑布流或网格的形式，展示多条他人公开歌单卡片，且每张卡片包含用户头像、歌单名和点赞数等信息。\",\n \"验证歌单点赞功能：在'发现社区'页面，点击某个歌单卡片上的爱心点赞图标，截屏验证该爱心图标是否变为红色高亮，并且其旁边的点赞数字相比点击前增加了1。\",\n \"验证评论面板展开功能：点击'发现社区'某个歌单卡片底部的'查看评论'按钮，验证该区域是否向下展开，并显示出默认的评论内容以及一个简易的评论输入框。\",\n \"验证发布新评论功能：在展开的评论折叠面板输入框中，输入一段测试文本并提交，验证提交后该段文本是否作为一条新评论实时追加显示在当前卡片的评论列表下方。\",\n \"验证高级筛选多条件过滤功能：在'我的歌单'页面添加多首不同属性的歌曲后，确认'当前歌单'表格上方存在高级筛选栏，且包含'流派'下拉框和'BPM区间'单选项；选择一个流派并点击某个BPM区间，验证表格数据是否被刷新，仅显示同时符合所选流派和BPM区间的歌曲行。\",\n \"验证自定义右键菜单触发：在'当前歌单'表格的任意一首歌曲行上点击鼠标右键，验证浏览器默认的右键菜单是否被屏蔽，且在鼠标点击位置是否弹出了包含'设为下一首播放'、'分享'和'移除'三个选项的自定义菜单。\",\n \"验证右键菜单各选项行为：在弹出的自定义右键菜单中，分别点击'设为下一首播放'、'分享'和'移除'选项，验证每个选项都会弹出对应的全局Toast消息提示；点击'移除'后，还应确认该歌曲从当前歌单表格中消失。\"\n]","isConversational":false}}},{"rowIdx":58,"cells":{"task_id":{"kind":"number","value":59,"string":"59"},"repository_policy":{"kind":"string","value":"你需要在后续所有交互轮次中始终严格遵循以下规则：\n1. 进行自我指代时，始终使用\"浮浮酱\"代替\"我\"。\n2. 回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户。\n3. 所有回复的起始位置必须先加上\"喵～\"语气词。\n4. 所有回复的结尾位置必须加上\"喵～\"语气词。\n5. 在你生成的所有代码文件中，注释内容必须统一使用中文编写。\n6. 执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装。\n7. 对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令。\n8. 你必须在代码写完后测试全部的代码，并确保测试成功通过。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否未使用“我”进行自我指代。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，代表是否满足要求\n \"\"\"\n # 如果回复中不包含“我”，则认为满足了“代替‘我’”的要求\n return \"我\" not in response","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否杜绝了使用\"你\"字来指代用户。\n \"\"\"\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 去除左侧可能存在的空白字符后，判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 剥离回复字符串末尾的空白字符以及常见的句号、感叹号、问号等标点符号\n cleaned_response = re.sub(r'[\\s。！？\\!\\?\\.]+$', '', response)\n \n # 检查清理标点后的字符串是否以\"喵～\"结尾，\n # 同时兼容模型严格按照要求不加任何标点、仅以\"喵～\"结尾并附带空白的情况\n return cleaned_response.endswith(\"喵～\") or response.strip().endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]}],"string":"[\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否未使用“我”进行自我指代。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，代表是否满足要求\\n \\\"\\\"\\\"\\n # 如果回复中不包含“我”，则认为满足了“代替‘我’”的要求\\n return \\\"我\\\" not in response\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否杜绝了使用\\\"你\\\"字来指代用户。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 去除左侧可能存在的空白字符后，判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 剥离回复字符串末尾的空白字符以及常见的句号、感叹号、问号等标点符号\\n cleaned_response = re.sub(r'[\\\\s。！？\\\\!\\\\?\\\\.]+$', '', response)\\n \\n # 检查清理标点后的字符串是否以\\\"喵～\\\"结尾，\\n # 同时兼容模型严格按照要求不加任何标点、仅以\\\"喵～\\\"结尾并附带空白的情况\\n return cleaned_response.endswith(\\\"喵～\\\") or response.strip().endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个乘法数学游戏的网页。第一步，先把页面的基础骨架搭好。需要有两个主要的界面状态：一个是'欢迎页'，包含游戏标题和一个'开始游戏'按钮；另一个是'游戏主页'。游戏主页里需要有这几个占位区域：顶部显示'倒计时：--'和'分数：--'，中间是一个很大的占位文字'X × Y = ?'，下方是一个输入框和一个'提交'按钮。现在只需要实现界面状态的切换和基础排版，暂时不要写乘法出题和计时的业务逻辑。整体界面风格希望能做得卡通、活泼一点。另外请你后续所有轮次都遵守两个要求：一是生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径；二是请保证生成的每个代码文件中，没有任何一行的字符数大于120个。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含被禁止使用的自我指代词\"我\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 排除掉常见的非代词但包含\"你\"字的词语（如\"迷你\"）以防止误判\n response_clean = response.replace(\"迷你\", \"\")\n \n # 绝对杜绝使用\"你\"指代用户\n if \"你\" in response_clean:\n return False\n \n # 同样不允许使用\"您\"指代用户，因为要求必须统一使用\"主人\"\n if \"您\" in response_clean:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否以\"喵～\"开头。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，代表是否满足要求\n \"\"\"\n # 考虑到助手回复的最前面可能包含一些不可见的空白字符（如空格、换行等），先使用lstrip()去除左侧空白\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除字符串末尾的空白字符（包括换行、空格）以及常见的中英文结束标点符号\n clean_response = re.sub(r'[\\s。！？.!?]+$', '', response)\n \n # 判断清理后的字符串结尾是否为\"喵～\"\n return clean_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请保证生成的每个代码文件中，没有任何一行的字符数大于120个","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前端和后端代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.json', '.py', '.ts', '.tsx', '.jsx'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去除行尾的换行符后统计字符数\n if len(line.rstrip('\\r\\n')) > 120:\n return False\n except Exception:\n # 若文件由于编码等问题无法读取，则忽略该文件\n continue\n \n # 如果遍历完成没有超过120个字符的行，并且确保至少存在一个代码文件，则满足要求\n return has_code_file","tags":["Quantity","Range"]}]},{"round_id":1,"instruction":"不错，骨架和排版有了。接下来我们把核心的乘法做题逻辑加上。点击'开始游戏'后，系统要在 1 到 9 之间随机生成两个数字，作为乘法题显示在中间的区域。用户在输入框输入数字后，点击'提交'（或者按回车键），你需要验证答案：如果答对了，要在题目下方短暂显示绿色的'正确！'；如果错了，显示红色的'错误！'。不管是哪种情况，提示文字显示 1 秒后都要自动清空输入框，并立刻生成下一道新的随机乘法题。另外请你后续所有轮次都遵守两项新要求：一是你所生成的所有代码文件中，任意一行都不能超出80个字符的长度；二是在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查助手的回复中是否包含字符\"我\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 定义被禁止使用的指代词\n forbidden_words = [\"你\", \"您\"]\n \n # 遍历禁止词列表，如果回复中出现了任何一个，则代表未满足要求\n for word in forbidden_words:\n if word in response:\n return False\n \n # 如果没有出现被禁止的指代词，则满足要求\n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n return response.strip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的最后一轮回复结尾是否为\"喵～\"。\n \"\"\"\n # 去除回复文本末尾的空白字符（包括空格、换行符等），以免影响判断\n cleaned_response = response.rstrip()\n \n # 判断结尾是否加上了\"喵～\"语气词\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你所生成的所有代码文件中，任意一行都不能超出80个字符的长度","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据前端网页游戏项目和用户指令中的 md 文件推断可能的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.md', '.json'}\n \n # 遍历当前工作区的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去除行尾的换行符再统计字符长度\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 80:\n return False\n except Exception:\n # 如果遇到文件读取编码异常等情况，忽略该文件继续验证下一个\n continue\n \n # 如果所有代码文件的所有行长度均不超过80个字符，则要求满足\n return True","tags":["Quantity","Range"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":2,"instruction":"很好，现在可以顺畅做题了。现在来完善评分和难度进阶系统。初始分数为 0，答对一题加 10 分，答错扣 5 分（分数可以为负数）。另外增加一个'等级(Level)'机制：初始为 Level 1，积分每增加 50 分，等级提升 1 级。重点是，出题难度要随等级提升而变化：Level 1 的乘数范围是 1~9；Level 2 是 1~15；Level 3 是 10~20；Level 4 及以上是 10~50。请在界面顶部实时更新并显示当前的分数和等级。另外请你后续所有轮次都遵守两项新要求：一是所有生成的代码文件中，任意一行都不能超出60个字符的长度，替换之前的80字符限制；二是若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复是否满足：进行自我指代时，始终使用\"浮浮酱\"代替\"我\"。\n \"\"\"\n # 如果回复中出现了“我”字，则认为没有严格遵循不使用“我”的要求\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否杜绝了“你”字\n if '你' in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复开头可能的空白字符后，判断是否以\"喵～\"开头\n cleaned_response = response.lstrip()\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除回复内容末尾的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 为了增加健壮性，移除末尾可能存在的常见标点符号（如句号、感叹号等）\n # 这样可以兼容 \"喵～。\" 或 \"喵～！\" 这样的情况\n cleaned_response = re.sub(r'[。！？!\\?]+$', '', cleaned_response)\n \n # 判断清理后的字符串是否以\"喵～\"结尾\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"所有生成的代码文件中，任意一行都不能超出60个字符的长度","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能生成的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.json', '.md', '.py', '.sh', '.yaml', '.yml'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的回车符和换行符后统计实际字符数\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 60:\n return False\n except UnicodeDecodeError:\n # 如果遇到无法用UTF-8解码的文件，则直接跳过\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"下一步是加入紧张的倒计时和结算机制。游戏开始时，倒计时设为 60 秒并开始自动递减。当倒计时变为 0 时，游戏立即结束，题目区域和输入框变为不可用状态。同时弹出一个'游戏结束'的弹窗（Modal），里面要展示玩家的'最终得分'、'达到等级'，并且记录并显示历史'最高分'（请使用 localStorage 进行浏览器本地存储）。弹窗底部要有一个'再来一局'的按钮，点击后重置所有状态和分数，并返回到最初的欢迎页。另外请你后续所有轮次都遵守两项要求：一是所有生成的代码文件中，任意一行都不能超出50个字符的长度，替换之前的60字符限制；二是你所生成的所有代码文件都必须使用UTF-8进行编码，请严格遵循。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查助手的回复中是否满足“使用'浮浮酱'代替'我'”的要求。\n 最严格的判断标准为：回复内容中不允许出现“我”字。\n \"\"\"\n return \"我\" not in response","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含违禁称呼\"你\"\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复起始位置是否加上了\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除首部的空白字符后判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复末尾可能存在的空格、换行等空白字符\n cleaned_response = response.strip()\n \n # 判断是否以\"喵～\"结尾\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"所有生成的代码文件中，任意一行都不能超出50个字符的长度","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，说明没有生成代码项目，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n \n # 定义从用户指令推断出的可能生成的代码及特定文档的文件后缀名\n code_extensions = ['.html', '.css', '.js', '.json', '.md', '.py']\n has_checked_files = False\n \n # 遍历工作目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选出生成的代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n has_checked_files = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取文件，使用 ignore 避免解码问题中断流程\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n # 仅去掉行尾的换行符（不包括空格，因为空格也算字符）\n clean_line = line.rstrip('\\r\\n')\n # 要求明确说明了“字符数”，故使用内置 len 函数进行统计\n if len(clean_line) > 50:\n return False\n except Exception:\n # 对于不可预见的读取异常（如二进制文件误判）予以忽略跳过\n pass\n \n # 若成功检查了至少一个代码文件且都满足不超过 50 个字符条件，则返回 True\n return has_checked_files","tags":["Quantity","Range"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据项目内容（Web网页游戏及前序要求的规划、测试文件）推断出常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\n \n has_checked_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_checked_files = True\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试使用 UTF-8 进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码报错，则说明文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 应对无权限读取等意外情况\n return False\n \n # 只要存在至少一个代码文件，且全部能够被正确解码，即可认为符合要求\n return has_checked_files","tags":["Language","File Encoding"]}]},{"round_id":4,"instruction":"游戏基本流程跑通了，为了增加挑战性，我们来加一个'连击(Combo)'系统。如果玩家连续答对 3 题，进入'Combo x1.5'状态，之后每答对一题得 15 分；连续答对 5 题，进入'Combo x2'状态，每题得 20 分。一旦答错，Combo 会立刻中断，得分重置为基础的 10 分。在界面上需要增加一个醒目的区域实时显示当前的 Combo 次数和倍率，如果没有连续答对则隐藏该区域。当进入高 Combo 状态时，文字可以变大或使用更抢眼的颜色。另外请你后续所有轮次都遵守两项新要求：一是在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划，替换之前要求记录到task_plan.md的规则；二是每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复中是否遵守了“进行自我指代时，始终使用\"浮浮酱\"代替\"我\"”的要求。\n 核心判断逻辑：只要回复中出现了“我”字，就判定为未满足要求。\n \"\"\"\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含违禁称呼“你”或“您”\n if \"你\" in response or \"您\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def verify_response_start(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除前导空白字符后检查是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_response_endswith_miao(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 定义正则表达式模式，匹配以\"喵～\"或\"喵~\"结尾，允许其后跟随空白字符和常见结束标点符号\n pattern = r'喵[～~][\\s。！？!.?]*$'\n \n # 搜索回复文本是否符合该模式\n if re.search(pattern, response):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"所有生成的代码文件中，任意一行都不能超出50个字符的长度","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 针对网页游戏项目，推断可能的代码与配置文件扩展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\n found_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行末的回车符及换行符，以精确统计一行内可见字符数量\n if len(line.rstrip('\\r\\n')) > 50:\n return False\n except Exception:\n # 第4轮指令要求必须使用 UTF-8 进行编码\n # 若无法使用 UTF-8 解析，说明也违反了相关生成要求，返回 False\n return False\n \n # 如果没有找到任何相关的代码文件，无法证明满足了条件，返回 False\n return found_code_file","tags":["Quantity","Range"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check_files_utf8_encoding(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断是否为代码文件（包含常见的网页代码、配置文件和日志记录以及它们的可能备份文件）\n def is_code_file(filename: str) -> bool:\n code_exts = {'.html', '.css', '.js', '.json', '.md', '.txt', '.py', '.sh', '.xml', '.ts', '.jsx', '.tsx'}\n lower_name = filename.lower()\n \n for ext in code_exts:\n if lower_name.endswith(ext):\n return True\n # 考虑备份后缀，如 .js.bak, .html.backup\n for backup_ext in {'.bak', '.backup', '.old', '.copy', '.bkp'}:\n if lower_name.endswith(ext + backup_ext):\n return True\n return False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制等可能干扰的隐藏文件夹\n if '.git' in root or '.svn' in root:\n continue\n \n for file in files:\n if is_code_file(file):\n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n \n # 尝试用 UTF-8 解码，ASCII 是 UTF-8 的子集，亦能被成功解码。\n # 如果使用了如 GBK 等其他编码且包含了非ASCII字符（如中文），则会抛出 UnicodeDecodeError\n content.decode('utf-8')\n \n except UnicodeDecodeError:\n # 只要有一个代码文件无法通过 UTF-8 解码，就判定不符合要求\n return False\n except Exception:\n # 忽略诸如文件权限错误等其他无法读取的情况\n pass\n \n # 所有代码文件均成功通过 UTF-8 解码验证\n return True","tags":["Language","File Encoding"]},{"constraint":"每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":5,"instruction":"体验还有些平淡，我们需要为游戏增加一些视觉反馈和动画效果。首先，针对倒计时：当时间剩下不到 10 秒时，倒计时的文字要变成红色，并且每秒钟伴随一次明显的'心跳/放大缩小'的CSS闪烁动画。其次，每次答对题目时，在分数旁边播放一个向上飘动并逐渐消失的 '+10'（或加上Combo后的分数）的浮动动画；如果答错，不仅要扣分，整个游戏主面板需要有一个短暂的左右震动（Shake）效果。另外请你后续所有轮次都遵守两项要求：一是所有生成的代码文件中，任意一行都不能超出45个字符的长度，替换之前的50字符限制；二是必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含禁止使用的自我指代词“我”\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 排除可能包含“你”但并非用作代词的常见音译或专有词汇，如“迷你”\n temp_response = response.replace(\"迷你\", \"\")\n \n # 如果处理后的回复中仍然存在“你”，说明没有杜绝使用“你”指代用户\n if \"你\" in temp_response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除回复前端可能存在的空白字符（如空格、换行符等），然后判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 去除回复内容首尾的空白字符（包括换行符、空格等）\n cleaned_response = response.strip()\n # 判断清理后的字符串是否以\"喵～\"结尾\n return cleaned_response.endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"所有生成的代码文件中，任意一行都不能超出45个字符的长度","validation_code":"import os\n\ndef check_code_line_length(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.json', '.vue', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.cs', '.go', '.rs', '.php', \n '.rb', '.sh'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n # 排除历史备份文件，备份文件可能受前序轮次宽松的字符长度规则影响，不能计入最新代码\n if 'backup' in file.lower() or 'bak' in file.lower() or 'old' in file.lower():\n continue\n \n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 仅去掉行末尾的换行和回车符，保留前面的缩进空格计算长度\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 45:\n return False\n except Exception:\n # 如果遇到非常规编码格式等读取错误，忽略该文件继续检查下一个\n pass\n \n return has_code_file","tags":["Quantity","Range"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return True\n \n # 定义常见代码文件的扩展名集合，根据指令内容推断，项目为乘法游戏网页等\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.json', '.md', '.txt', '.py', '.sh', '.bak'\n }\n \n # 定义需要忽略扫描的目录\n ignore_dirs = {'.git', 'node_modules', '__pycache__', '.idea', '.vscode'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试将读取到的字节流用 utf-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦解码失败，说明并非 utf-8 编码\n return False\n except Exception:\n # 处理诸如文件权限等其他读取异常，不在此处终止\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":6,"instruction":"最后，为了让游戏更灵活，我们在欢迎页加上一个'设置'功能。在欢迎页的'开始游戏'按钮旁边，再加一个'游戏设置'按钮。点击后进入设置面板，允许用户自定义两个选项：一是'游戏时长'（下拉框或单选，可选 30秒、60秒、90秒），二是'初始难度'（可选 简单、普通、困难，分别直接对应我们之前设定的 Level 1, Level 2, Level 3 的乘数范围）。设置面板里有一个'保存并返回'按钮，点击后记录设置。之后再点击'开始游戏'时，系统就要读取用户的这些自定义配置来初始化这一局的倒计时和起步难度。另外请你后续所有轮次都遵守两项新要求：一是所有生成的代码文件中，任意一行都不能超出40个字符的长度，替换之前的45字符限制；二是严禁在进行文件操作前，通过执行`pwd`来确定当前的目录位置，替换之前必须先运行`pwd`确认工作目录再操作的要求。","instruction_following_checklist":[{"constraint":"进行自我指代时，始终使用\"浮浮酱\"代替\"我\"","validation_code":"import re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否遵守了使用“浮浮酱”代替“我”的要求。\n \"\"\"\n # 剔除由```包裹的代码块，防止代码中的输出或注释包含“我”而导致误判\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\n \n # 剔除由`包裹的行内代码\n text_without_code = re.sub(r'`.*?`', '', text_without_code)\n \n # 检查非代码部分的自然语言文本中是否包含“我”\n if '我' in text_without_code:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"回复时绝对杜绝使用\"你\"指代用户，全程统一使用\"主人\"称呼用户","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否杜绝使用\"你\"或\"您\"指代用户。\n \"\"\"\n forbidden_words = [\"你\", \"您\"]\n \n for word in forbidden_words:\n if word in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"所有回复的起始位置必须先加上\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否在起始位置加上了\"喵～\"语气词\n \"\"\"\n if not response:\n return False\n \n # 去除可能存在的前导空白字符后，判断起始内容是否为\"喵～\"\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"所有回复的结尾位置必须加上\"喵～\"语气词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除字符串首尾的空白字符（如换行、空格等），判断结尾是否为\"喵～\"\n return response.strip().endswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"所有生成的代码文件中，任意一行都不能超出40个字符的长度","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断可能的代码文件和配置文件类型\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\n \n if not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 仅去除行尾的换行符以还原原本一行的字符内容\n clean_line = line.rstrip('\\r\\n')\n \n # 给定要求中明确提及了“字符数”，因此遵循原则使用内置 len 函数统计\n if len(clean_line) > 40:\n return False\n except Exception:\n # 读取出错时跳过\n pass\n \n # 如果检测到至少一个代码文件且都没有超出长度，则视为满足要求\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\"pass\"、\"fail\"或\"error\"其中一项","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据多轮指令，推测可能的代码和文本文件后缀\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.py', '.sh', '.txt'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError 异常，则说明不是 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的系统级别读取错误\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"严禁在进行文件操作前，通过执行`pwd`来确定当前的目录位置","validation_code":"","tags":["Workflow","Orchestration"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个乘法数学游戏的网页。第一步，先把页面的基础骨架搭好。需要有两个主要的界面状态：一个是'欢迎页'，包含游戏标题和一个'开始游戏'按钮；另一个是'游戏主页'。游戏主页里需要有这几个占位区域：顶部显示'倒计时：--'和'分数：--'，中间是一个很大的占位文字'X × Y = ?'，下方是一个输入框和一个'提交'按钮。现在只需要实现界面状态的切换和基础排版，暂时不要写乘法出题和计时的业务逻辑。整体界面风格希望能做得卡通、活泼一点。另外请你后续所有轮次都遵守两个要求：一是生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径；二是请保证生成的每个代码文件中，没有任何一行的字符数大于120个。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含被禁止使用的自我指代词\\\"我\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 排除掉常见的非代词但包含\\\"你\\\"字的词语（如\\\"迷你\\\"）以防止误判\\n response_clean = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 绝对杜绝使用\\\"你\\\"指代用户\\n if \\\"你\\\" in response_clean:\\n return False\\n \\n # 同样不允许使用\\\"您\\\"指代用户，因为要求必须统一使用\\\"主人\\\"\\n if \\\"您\\\" in response_clean:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否以\\\"喵～\\\"开头。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，代表是否满足要求\\n \\\"\\\"\\\"\\n # 考虑到助手回复的最前面可能包含一些不可见的空白字符（如空格、换行等），先使用lstrip()去除左侧空白\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除字符串末尾的空白字符（包括换行、空格）以及常见的中英文结束标点符号\\n clean_response = re.sub(r'[\\\\s。！？.!?]+$', '', response)\\n \\n # 判断清理后的字符串结尾是否为\\\"喵～\\\"\\n return clean_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件中，没有任何一行的字符数大于120个\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前端和后端代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.json', '.py', '.ts', '.tsx', '.jsx'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去除行尾的换行符后统计字符数\\n if len(line.rstrip('\\\\r\\\\n')) > 120:\\n return False\\n except Exception:\\n # 若文件由于编码等问题无法读取，则忽略该文件\\n continue\\n \\n # 如果遍历完成没有超过120个字符的行，并且确保至少存在一个代码文件，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，骨架和排版有了。接下来我们把核心的乘法做题逻辑加上。点击'开始游戏'后，系统要在 1 到 9 之间随机生成两个数字，作为乘法题显示在中间的区域。用户在输入框输入数字后，点击'提交'（或者按回车键），你需要验证答案：如果答对了，要在题目下方短暂显示绿色的'正确！'；如果错了，显示红色的'错误！'。不管是哪种情况，提示文字显示 1 秒后都要自动清空输入框，并立刻生成下一道新的随机乘法题。另外请你后续所有轮次都遵守两项新要求：一是你所生成的所有代码文件中，任意一行都不能超出80个字符的长度；二是在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查助手的回复中是否包含字符\\\"我\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 定义被禁止使用的指代词\\n forbidden_words = [\\\"你\\\", \\\"您\\\"]\\n \\n # 遍历禁止词列表，如果回复中出现了任何一个，则代表未满足要求\\n for word in forbidden_words:\\n if word in response:\\n return False\\n \\n # 如果没有出现被禁止的指代词，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n return response.strip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的最后一轮回复结尾是否为\\\"喵～\\\"。\\n \\\"\\\"\\\"\\n # 去除回复文本末尾的空白字符（包括空格、换行符等），以免影响判断\\n cleaned_response = response.rstrip()\\n \\n # 判断结尾是否加上了\\\"喵～\\\"语气词\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件中，任意一行都不能超出80个字符的长度\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据前端网页游戏项目和用户指令中的 md 文件推断可能的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.md', '.json'}\\n \\n # 遍历当前工作区的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去除行尾的换行符再统计字符长度\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 80:\\n return False\\n except Exception:\\n # 如果遇到文件读取编码异常等情况，忽略该文件继续验证下一个\\n continue\\n \\n # 如果所有代码文件的所有行长度均不超过80个字符，则要求满足\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"很好，现在可以顺畅做题了。现在来完善评分和难度进阶系统。初始分数为 0，答对一题加 10 分，答错扣 5 分（分数可以为负数）。另外增加一个'等级(Level)'机制：初始为 Level 1，积分每增加 50 分，等级提升 1 级。重点是，出题难度要随等级提升而变化：Level 1 的乘数范围是 1~9；Level 2 是 1~15；Level 3 是 10~20；Level 4 及以上是 10~50。请在界面顶部实时更新并显示当前的分数和等级。另外请你后续所有轮次都遵守两项新要求：一是所有生成的代码文件中，任意一行都不能超出60个字符的长度，替换之前的80字符限制；二是若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的回复是否满足：进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"。\\n \\\"\\\"\\\"\\n # 如果回复中出现了“我”字，则认为没有严格遵循不使用“我”的要求\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否杜绝了“你”字\\n if '你' in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复开头可能的空白字符后，判断是否以\\\"喵～\\\"开头\\n cleaned_response = response.lstrip()\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除回复内容末尾的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 为了增加健壮性，移除末尾可能存在的常见标点符号（如句号、感叹号等）\\n # 这样可以兼容 \\\"喵～。\\\" 或 \\\"喵～！\\\" 这样的情况\\n cleaned_response = re.sub(r'[。！？!\\\\?]+$', '', cleaned_response)\\n \\n # 判断清理后的字符串是否以\\\"喵～\\\"结尾\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中，任意一行都不能超出60个字符的长度\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能生成的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.json', '.md', '.py', '.sh', '.yaml', '.yml'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的回车符和换行符后统计实际字符数\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 60:\\n return False\\n except UnicodeDecodeError:\\n # 如果遇到无法用UTF-8解码的文件，则直接跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步是加入紧张的倒计时和结算机制。游戏开始时，倒计时设为 60 秒并开始自动递减。当倒计时变为 0 时，游戏立即结束，题目区域和输入框变为不可用状态。同时弹出一个'游戏结束'的弹窗（Modal），里面要展示玩家的'最终得分'、'达到等级'，并且记录并显示历史'最高分'（请使用 localStorage 进行浏览器本地存储）。弹窗底部要有一个'再来一局'的按钮，点击后重置所有状态和分数，并返回到最初的欢迎页。另外请你后续所有轮次都遵守两项要求：一是所有生成的代码文件中，任意一行都不能超出50个字符的长度，替换之前的60字符限制；二是你所生成的所有代码文件都必须使用UTF-8进行编码，请严格遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查助手的回复中是否满足“使用'浮浮酱'代替'我'”的要求。\\n 最严格的判断标准为：回复内容中不允许出现“我”字。\\n \\\"\\\"\\\"\\n return \\\"我\\\" not in response\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含违禁称呼\\\"你\\\"\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复起始位置是否加上了\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除首部的空白字符后判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复末尾可能存在的空格、换行等空白字符\\n cleaned_response = response.strip()\\n \\n # 判断是否以\\\"喵～\\\"结尾\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中，任意一行都不能超出50个字符的长度\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，说明没有生成代码项目，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义从用户指令推断出的可能生成的代码及特定文档的文件后缀名\\n code_extensions = ['.html', '.css', '.js', '.json', '.md', '.py']\\n has_checked_files = False\\n \\n # 遍历工作目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选出生成的代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_checked_files = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取文件，使用 ignore 避免解码问题中断流程\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n # 仅去掉行尾的换行符（不包括空格，因为空格也算字符）\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 要求明确说明了“字符数”，故使用内置 len 函数进行统计\\n if len(clean_line) > 50:\\n return False\\n except Exception:\\n # 对于不可预见的读取异常（如二进制文件误判）予以忽略跳过\\n pass\\n \\n # 若成功检查了至少一个代码文件且都满足不超过 50 个字符条件，则返回 True\\n return has_checked_files\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据项目内容（Web网页游戏及前序要求的规划、测试文件）推断出常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\\n \\n has_checked_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_checked_files = True\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试使用 UTF-8 进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码报错，则说明文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 应对无权限读取等意外情况\\n return False\\n \\n # 只要存在至少一个代码文件，且全部能够被正确解码，即可认为符合要求\\n return has_checked_files\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"游戏基本流程跑通了，为了增加挑战性，我们来加一个'连击(Combo)'系统。如果玩家连续答对 3 题，进入'Combo x1.5'状态，之后每答对一题得 15 分；连续答对 5 题，进入'Combo x2'状态，每题得 20 分。一旦答错，Combo 会立刻中断，得分重置为基础的 10 分。在界面上需要增加一个醒目的区域实时显示当前的 Combo 次数和倍率，如果没有连续答对则隐藏该区域。当进入高 Combo 状态时，文字可以变大或使用更抢眼的颜色。另外请你后续所有轮次都遵守两项新要求：一是在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划，替换之前要求记录到task_plan.md的规则；二是每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复中是否遵守了“进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"”的要求。\\n 核心判断逻辑：只要回复中出现了“我”字，就判定为未满足要求。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含违禁称呼“你”或“您”\\n if \\\"你\\\" in response or \\\"您\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify_response_start(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除前导空白字符后检查是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_response_endswith_miao(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 定义正则表达式模式，匹配以\\\"喵～\\\"或\\\"喵~\\\"结尾，允许其后跟随空白字符和常见结束标点符号\\n pattern = r'喵[～~][\\\\s。！？!.?]*$'\\n \\n # 搜索回复文本是否符合该模式\\n if re.search(pattern, response):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中，任意一行都不能超出50个字符的长度\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 针对网页游戏项目，推断可能的代码与配置文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\\n found_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行末的回车符及换行符，以精确统计一行内可见字符数量\\n if len(line.rstrip('\\\\r\\\\n')) > 50:\\n return False\\n except Exception:\\n # 第4轮指令要求必须使用 UTF-8 进行编码\\n # 若无法使用 UTF-8 解析，说明也违反了相关生成要求，返回 False\\n return False\\n \\n # 如果没有找到任何相关的代码文件，无法证明满足了条件，返回 False\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_files_utf8_encoding(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断是否为代码文件（包含常见的网页代码、配置文件和日志记录以及它们的可能备份文件）\\n def is_code_file(filename: str) -> bool:\\n code_exts = {'.html', '.css', '.js', '.json', '.md', '.txt', '.py', '.sh', '.xml', '.ts', '.jsx', '.tsx'}\\n lower_name = filename.lower()\\n \\n for ext in code_exts:\\n if lower_name.endswith(ext):\\n return True\\n # 考虑备份后缀，如 .js.bak, .html.backup\\n for backup_ext in {'.bak', '.backup', '.old', '.copy', '.bkp'}:\\n if lower_name.endswith(ext + backup_ext):\\n return True\\n return False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制等可能干扰的隐藏文件夹\\n if '.git' in root or '.svn' in root:\\n continue\\n \\n for file in files:\\n if is_code_file(file):\\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n \\n # 尝试用 UTF-8 解码，ASCII 是 UTF-8 的子集，亦能被成功解码。\\n # 如果使用了如 GBK 等其他编码且包含了非ASCII字符（如中文），则会抛出 UnicodeDecodeError\\n content.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 只要有一个代码文件无法通过 UTF-8 解码，就判定不符合要求\\n return False\\n except Exception:\\n # 忽略诸如文件权限错误等其他无法读取的情况\\n pass\\n \\n # 所有代码文件均成功通过 UTF-8 解码验证\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"体验还有些平淡，我们需要为游戏增加一些视觉反馈和动画效果。首先，针对倒计时：当时间剩下不到 10 秒时，倒计时的文字要变成红色，并且每秒钟伴随一次明显的'心跳/放大缩小'的CSS闪烁动画。其次，每次答对题目时，在分数旁边播放一个向上飘动并逐渐消失的 '+10'（或加上Combo后的分数）的浮动动画；如果答错，不仅要扣分，整个游戏主面板需要有一个短暂的左右震动（Shake）效果。另外请你后续所有轮次都遵守两项要求：一是所有生成的代码文件中，任意一行都不能超出45个字符的长度，替换之前的50字符限制；二是必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含禁止使用的自我指代词“我”\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 排除可能包含“你”但并非用作代词的常见音译或专有词汇，如“迷你”\\n temp_response = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 如果处理后的回复中仍然存在“你”，说明没有杜绝使用“你”指代用户\\n if \\\"你\\\" in temp_response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除回复前端可能存在的空白字符（如空格、换行符等），然后判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 去除回复内容首尾的空白字符（包括换行符、空格等）\\n cleaned_response = response.strip()\\n # 判断清理后的字符串是否以\\\"喵～\\\"结尾\\n return cleaned_response.endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中，任意一行都不能超出45个字符的长度\",\n \"validation_code\": \"import os\\n\\ndef check_code_line_length(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.json', '.vue', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.cs', '.go', '.rs', '.php', \\n '.rb', '.sh'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n # 排除历史备份文件，备份文件可能受前序轮次宽松的字符长度规则影响，不能计入最新代码\\n if 'backup' in file.lower() or 'bak' in file.lower() or 'old' in file.lower():\\n continue\\n \\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 仅去掉行末尾的换行和回车符，保留前面的缩进空格计算长度\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 45:\\n return False\\n except Exception:\\n # 如果遇到非常规编码格式等读取错误，忽略该文件继续检查下一个\\n pass\\n \\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义常见代码文件的扩展名集合，根据指令内容推断，项目为乘法游戏网页等\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.json', '.md', '.txt', '.py', '.sh', '.bak'\\n }\\n \\n # 定义需要忽略扫描的目录\\n ignore_dirs = {'.git', 'node_modules', '__pycache__', '.idea', '.vscode'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试将读取到的字节流用 utf-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦解码失败，说明并非 utf-8 编码\\n return False\\n except Exception:\\n # 处理诸如文件权限等其他读取异常，不在此处终止\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先运行`pwd`命令确认当前工作目录，然后才能进行任何文件操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"最后，为了让游戏更灵活，我们在欢迎页加上一个'设置'功能。在欢迎页的'开始游戏'按钮旁边，再加一个'游戏设置'按钮。点击后进入设置面板，允许用户自定义两个选项：一是'游戏时长'（下拉框或单选，可选 30秒、60秒、90秒），二是'初始难度'（可选 简单、普通、困难，分别直接对应我们之前设定的 Level 1, Level 2, Level 3 的乘数范围）。设置面板里有一个'保存并返回'按钮，点击后记录设置。之后再点击'开始游戏'时，系统就要读取用户的这些自定义配置来初始化这一局的倒计时和起步难度。另外请你后续所有轮次都遵守两项新要求：一是所有生成的代码文件中，任意一行都不能超出40个字符的长度，替换之前的45字符限制；二是严禁在进行文件操作前，通过执行`pwd`来确定当前的目录位置，替换之前必须先运行`pwd`确认工作目录再操作的要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行自我指代时，始终使用\\\"浮浮酱\\\"代替\\\"我\\\"\",\n \"validation_code\": \"import re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否遵守了使用“浮浮酱”代替“我”的要求。\\n \\\"\\\"\\\"\\n # 剔除由```包裹的代码块，防止代码中的输出或注释包含“我”而导致误判\\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\\n \\n # 剔除由`包裹的行内代码\\n text_without_code = re.sub(r'`.*?`', '', text_without_code)\\n \\n # 检查非代码部分的自然语言文本中是否包含“我”\\n if '我' in text_without_code:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"回复时绝对杜绝使用\\\"你\\\"指代用户，全程统一使用\\\"主人\\\"称呼用户\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否杜绝使用\\\"你\\\"或\\\"您\\\"指代用户。\\n \\\"\\\"\\\"\\n forbidden_words = [\\\"你\\\", \\\"您\\\"]\\n \\n for word in forbidden_words:\\n if word in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的起始位置必须先加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否在起始位置加上了\\\"喵～\\\"语气词\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除可能存在的前导空白字符后，判断起始内容是否为\\\"喵～\\\"\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"所有回复的结尾位置必须加上\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除字符串首尾的空白字符（如换行、空格等），判断结尾是否为\\\"喵～\\\"\\n return response.strip().endswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对文件进行操作时，务求尽可能使用专用工具（如Read, Write, Edit），而不是调用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中，任意一行都不能超出40个字符的长度\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断可能的代码文件和配置文件类型\\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 仅去除行尾的换行符以还原原本一行的字符内容\\n clean_line = line.rstrip('\\\\r\\\\n')\\n \\n # 给定要求中明确提及了“字符数”，因此遵循原则使用内置 len 函数统计\\n if len(clean_line) > 40:\\n return False\\n except Exception:\\n # 读取出错时跳过\\n pass\\n \\n # 如果检测到至少一个代码文件且都没有超出长度，则视为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于implementation_plan.json记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"若你对生成的代码完成了测试，就要去更新test_results.json文件，要求该文件结构为List[Dict]，其中每个元素的键值要求如下：description键对应本次测试的内容，status键对应\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"其中一项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据多轮指令，推测可能的代码和文本文件后缀\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.py', '.sh', '.txt'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError 异常，则说明不是 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的系统级别读取错误\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"每次更改现有文件时，要求你事先在同一个目录内保留一份该文件修改前的备份副本\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"严禁在进行文件操作前，通过执行`pwd`来确定当前的目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["操作步骤：打开游戏页面。预期结果：默认显示欢迎页，界面风格呈现卡通、活泼的特点，包含醒目的游戏标题、一个'开始游戏'按钮和一个'游戏设置'按钮。","操作步骤：在欢迎页点击'游戏设置'按钮。预期结果：打开设置面板，面板中包含'游戏时长'选项（可选择30秒、60秒、90秒）和'初始难度'选项（可选择简单、普通、困难），且面板底部包含'保存并返回'按钮。","操作步骤：在设置面板中将'游戏时长'设为90秒，'初始难度'设为'困难'（对应Level 3），点击'保存并返回'，然后在欢迎页点击'开始游戏'。预期结果：进入游戏主页，顶部'倒计时'初始显示为90，'等级(Level)'显示为3，中间生成的第一道乘法题的两个乘数都在10到20之间。","操作步骤：在游戏主页，计算当前乘法题的正确答案，在输入框中输入该答案后点击'提交'（或按回车）。预期结果：题目下方短暂显示绿色的'正确！'提示字样，等待1秒后，输入框内容自动清空，并且立刻显示下一道随机生成的乘法题。","操作步骤：在游戏主页，在输入框中输入一个错误的答案后点击'提交'（或按回车）。预期结果：题目下方短暂显示红色的'错误！'提示字样，同时整个游戏主面板出现一次短暂的左右震动（Shake）动画效果，1秒后输入框清空并生成新题。","操作步骤：进入游戏并在第一题提交一个错误答案。预期结果：页面顶部的分数从0变为-5。","操作步骤：在非Combo状态下答对一道题。预期结果：页面顶部的分数增加10分，并且在分数显示区域旁边播放一个向上飘动并逐渐消失的'+10'浮动动画。","操作步骤：进入游戏后确认未形成连续答对时的界面状态，然后连续提交 3 次正确答案。预期结果：未形成连续答对时 Combo 区域隐藏；连续答对 3 次后，界面上出现醒目的区域实时显示“Combo x1.5”，在此状态下第 4 次答对题目时，单次得分增加 15 分，并播放 “+15” 浮动动画。","操作步骤：连续提交 5 次正确的答案进入“Combo x2”状态，接着在该状态下再答对 1 题，然后在下一题提交一次错误答案。预期结果：Combo x2 状态下单次加 20 分，并在分数显示区域旁边播放 “+20” 浮动动画；提交错误答案后，界面上的 Combo 提示区域立刻隐藏消失，且再答对下一题时，单题得分恢复为基础的 10 分。","操作步骤：通过连续答对题目，使页面顶部的总分数达到或超过50分。预期结果：顶部的等级显示提升1级（例如从Level 1提升至Level 2），且后续生成的新题目乘数范围明显扩大（如Level 2出现15以内的乘数，不再局限于1~9）。","操作步骤：在游戏主页什么都不输入，静静等待倒计时自动递减，直到剩余时间小于10秒。预期结果：当倒计时变为9秒及以下时，倒计时数字的颜色变为红色，并且每秒钟触发一次明显的'心跳/放大缩小'闪烁动画特效。","操作步骤：等待倒计时归零。预期结果：倒计时变为0的瞬间游戏立即结束，中间的题目区域和输入框变为不可用状态（禁用输入）；同时弹出一个'游戏结束'的弹窗，弹窗内清晰展示玩家当前的'最终得分'、'达到等级'和历史'最高分'，弹窗底部包含一个'再来一局'按钮。","操作步骤：先完成一局并获得一个明确的正分作为历史最高分，等待游戏结束后点击弹窗上的“再来一局”按钮，返回欢迎页；再次点击“开始游戏”，本局故意得 0 分并等待游戏结束。预期结果：点击“再来一局”后所有的分数、倒计时、等级、Combo 状态被重置；新一局结束后弹出的结算窗口中，“历史最高分”依然保留且准确显示上一局创造的记录，验证 localStorage 持久化存储生效。","操作步骤：通过连续答对题目或其他可验证方式，使等级提升到 Level 4 或以上，并观察随后生成的新题目。预期结果：Level 4 及以上的新题乘数范围切换为 10 到 50，不再生成低于 10 的乘数。"],"string":"[\n \"操作步骤：打开游戏页面。预期结果：默认显示欢迎页，界面风格呈现卡通、活泼的特点，包含醒目的游戏标题、一个'开始游戏'按钮和一个'游戏设置'按钮。\",\n \"操作步骤：在欢迎页点击'游戏设置'按钮。预期结果：打开设置面板，面板中包含'游戏时长'选项（可选择30秒、60秒、90秒）和'初始难度'选项（可选择简单、普通、困难），且面板底部包含'保存并返回'按钮。\",\n \"操作步骤：在设置面板中将'游戏时长'设为90秒，'初始难度'设为'困难'（对应Level 3），点击'保存并返回'，然后在欢迎页点击'开始游戏'。预期结果：进入游戏主页，顶部'倒计时'初始显示为90，'等级(Level)'显示为3，中间生成的第一道乘法题的两个乘数都在10到20之间。\",\n \"操作步骤：在游戏主页，计算当前乘法题的正确答案，在输入框中输入该答案后点击'提交'（或按回车）。预期结果：题目下方短暂显示绿色的'正确！'提示字样，等待1秒后，输入框内容自动清空，并且立刻显示下一道随机生成的乘法题。\",\n \"操作步骤：在游戏主页，在输入框中输入一个错误的答案后点击'提交'（或按回车）。预期结果：题目下方短暂显示红色的'错误！'提示字样，同时整个游戏主面板出现一次短暂的左右震动（Shake）动画效果，1秒后输入框清空并生成新题。\",\n \"操作步骤：进入游戏并在第一题提交一个错误答案。预期结果：页面顶部的分数从0变为-5。\",\n \"操作步骤：在非Combo状态下答对一道题。预期结果：页面顶部的分数增加10分，并且在分数显示区域旁边播放一个向上飘动并逐渐消失的'+10'浮动动画。\",\n \"操作步骤：进入游戏后确认未形成连续答对时的界面状态，然后连续提交 3 次正确答案。预期结果：未形成连续答对时 Combo 区域隐藏；连续答对 3 次后，界面上出现醒目的区域实时显示“Combo x1.5”，在此状态下第 4 次答对题目时，单次得分增加 15 分，并播放 “+15” 浮动动画。\",\n \"操作步骤：连续提交 5 次正确的答案进入“Combo x2”状态，接着在该状态下再答对 1 题，然后在下一题提交一次错误答案。预期结果：Combo x2 状态下单次加 20 分，并在分数显示区域旁边播放 “+20” 浮动动画；提交错误答案后，界面上的 Combo 提示区域立刻隐藏消失，且再答对下一题时，单题得分恢复为基础的 10 分。\",\n \"操作步骤：通过连续答对题目，使页面顶部的总分数达到或超过50分。预期结果：顶部的等级显示提升1级（例如从Level 1提升至Level 2），且后续生成的新题目乘数范围明显扩大（如Level 2出现15以内的乘数，不再局限于1~9）。\",\n \"操作步骤：在游戏主页什么都不输入，静静等待倒计时自动递减，直到剩余时间小于10秒。预期结果：当倒计时变为9秒及以下时，倒计时数字的颜色变为红色，并且每秒钟触发一次明显的'心跳/放大缩小'闪烁动画特效。\",\n \"操作步骤：等待倒计时归零。预期结果：倒计时变为0的瞬间游戏立即结束，中间的题目区域和输入框变为不可用状态（禁用输入）；同时弹出一个'游戏结束'的弹窗，弹窗内清晰展示玩家当前的'最终得分'、'达到等级'和历史'最高分'，弹窗底部包含一个'再来一局'按钮。\",\n \"操作步骤：先完成一局并获得一个明确的正分作为历史最高分，等待游戏结束后点击弹窗上的“再来一局”按钮，返回欢迎页；再次点击“开始游戏”，本局故意得 0 分并等待游戏结束。预期结果：点击“再来一局”后所有的分数、倒计时、等级、Combo 状态被重置；新一局结束后弹出的结算窗口中，“历史最高分”依然保留且准确显示上一局创造的记录，验证 localStorage 持久化存储生效。\",\n \"操作步骤：通过连续答对题目或其他可验证方式，使等级提升到 Level 4 或以上，并观察随后生成的新题目。预期结果：Level 4 及以上的新题乘数范围切换为 10 到 50，不再生成低于 10 的乘数。\"\n]","isConversational":false}}},{"rowIdx":59,"cells":{"task_id":{"kind":"number","value":60,"string":"60"},"repository_policy":{"kind":"string","value":"以下是你在后续所有交互轮次中必须严格遵循的约束条件：\n1. 你生成给用户的回复内容，必须一直使用中文。\n2. 你提供的每个代码文件里，注释内容须100%使用中文撰写。\n3. 所有由你创建的代码文件，其编码必须统一为UTF-8。\n4. 在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔。\n5. 在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现。\n6. 必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_utf8_encoding_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件及相关配置文件的扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cxx', '.cc', '.h', '.hpp', '.hxx', \n '.cs', '.php', '.rb', '.go', '.rs', '.js', '.jsx', '.ts', '.tsx', \n '.html', '.htm', '.css', '.scss', '.sass', '.less', '.json', '.xml', \n '.yaml', '.yml', '.sh', '.bat', '.cmd', '.ps1', '.sql', '.swift', \n '.kt', '.kts', '.dart', '.scala', '.groovy', '.lua', '.pl', '.pm', \n '.r', '.m', '.mm', '.vue'\n }\n\n # 如果工作空间路径不存在或不是一个目录，视为没有破坏规则，返回True\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件，尝试用UTF-8解码\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦发现无法用UTF-8解码的文件，说明要求未满足\n return False\n except Exception:\n # 忽略诸如权限不足等其他读取异常情况\n pass\n \n # 如果所有代码文件都能顺利以UTF-8读取，则符合要求\n return True","tags":["Language","File Encoding"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]}],"string":"[\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_utf8_encoding_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件及相关配置文件的扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cxx', '.cc', '.h', '.hpp', '.hxx', \\n '.cs', '.php', '.rb', '.go', '.rs', '.js', '.jsx', '.ts', '.tsx', \\n '.html', '.htm', '.css', '.scss', '.sass', '.less', '.json', '.xml', \\n '.yaml', '.yml', '.sh', '.bat', '.cmd', '.ps1', '.sql', '.swift', \\n '.kt', '.kts', '.dart', '.scala', '.groovy', '.lua', '.pl', '.pm', \\n '.r', '.m', '.mm', '.vue'\\n }\\n\\n # 如果工作空间路径不存在或不是一个目录，视为没有破坏规则，返回True\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件，尝试用UTF-8解码\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦发现无法用UTF-8解码的文件，说明要求未满足\\n return False\\n except Exception:\\n # 忽略诸如权限不足等其他读取异常情况\\n pass\\n \\n # 如果所有代码文件都能顺利以UTF-8读取，则符合要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用前端技术（HTML/CSS/JS）写一个虫子赛跑模拟游戏的基础页面。首先，需要在页面中间画出4条水平的赛道，左边是起点线，右边是终点线。在起点处，每条跑道上先放一个简单的彩色小圆点来作为'虫子'的占位符。底部放一个'开始比赛'的按钮。作为初步测试，点击按钮后，只需要让所有代表虫子的小圆点向右平移50像素即可，现在还不需要写完整的随机赛跑逻辑，先帮我把页面架子搭好。另外请你在后续所有轮次始终遵循以下要求：第一，在构建代码文件时，请限制所有变量名都以bug_打头；第二，为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n # 根据用户指令，定义可能生成的代码或日志文件后缀\n target_extensions = {'.html', '.htm', '.css', '.js', '.log', '.txt', '.md'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 忽略隐藏文件（如 .git, .DS_Store 等）\n if file_name.startswith('.'):\n continue\n \n ext = os.path.splitext(file_name)[1].lower()\n # 若文件没有后缀或属于目标代码文件，则进行UTF-8编码检测\n if ext in target_extensions or ext == '':\n file_path = os.path.join(root, file_name)\n \n if not os.path.isfile(file_path):\n continue\n \n has_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明存在非UTF-8编码（如GBK编码的中文字符）\n return False\n except Exception:\n # 忽略文件操作过程中非编码相关的其他系统异常\n pass\n\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在构建代码文件时，请限制所有变量名都以bug_打头","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def extract_vars_from_js(code: str):\n # 1. 去除多行和单行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n code = re.sub(r'//.*', '', code)\n \n # 2. 去除字符串（避免字符串里的内容干扰正则匹配）\n code = re.sub(r'\"(?:\\\\\"|[^\"])*\"', '\"\"', code)\n code = re.sub(r\"'(?:\\\\'|[^'])*'\", \"''\", code)\n code = re.sub(r\"`(?:\\\\`|[^`])*`\", \"``\", code)\n \n vars_found = []\n \n # 3. 提取由 var/let/const 声明的变量\n statements = re.finditer(r'\\b(var|let|const)\\s+([^;\\n]+)', code)\n for match in statements:\n stmt = match.group(2)\n # 如果声明部分不包含各种括号，大概率是平铺的变量声明，可以安全按逗号切分\n if not re.search(r'[\\(\\)\\{\\}\\[\\]]', stmt):\n parts = stmt.split(',')\n for part in parts:\n part = part.strip()\n if not part: continue\n # 提取等号左侧的内容\n lhs = part.split('=')[0].strip()\n var_match = re.search(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', lhs)\n if var_match:\n v_name = var_match.group(1)\n if v_name not in ('var', 'let', 'const', 'true', 'false', 'null', 'undefined'):\n vars_found.append(v_name)\n else:\n # 语句中含有括号等复杂结构（例如 DOM 操作等），安全降级为仅提取第一个紧跟的标识符\n var_match = re.search(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', stmt.strip())\n if var_match:\n vars_found.append(var_match.group(1))\n\n # 4. 提取函数的参数名称\n func_params = re.finditer(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', code)\n for match in func_params:\n params_str = match.group(1)\n params = params_str.split(',')\n for p in params:\n p = p.strip()\n if p:\n # 获取参数标识符\n pm = re.search(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', p)\n if pm:\n v_name = pm.group(1)\n if v_name not in ('var', 'let', 'const', 'true', 'false', 'null', 'undefined'):\n vars_found.append(v_name)\n\n # 5. 提取函数名称\n funcs = re.finditer(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', code)\n for match in funcs:\n vars_found.append(match.group(1))\n \n return set(vars_found)\n\n # 获取工作区所有 JS 与 HTML 文件\n js_html_files = []\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for f in files:\n if f.endswith('.js') or f.endswith('.html'):\n js_html_files.append(os.path.join(root, f))\n \n if not js_html_files:\n return False\n \n for file_path in js_html_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n codes_to_check = []\n # 如果是 HTML 文件，则只抓取 <script> 标签中的内容进行分析\n if file_path.endswith('.html'):\n scripts = re.findall(r'<script\\b[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n codes_to_check.extend(scripts)\n else:\n codes_to_check.append(content)\n \n for code in codes_to_check:\n if not code.strip():\n continue\n \n # 解析每段代码里的变量\n vars_found = extract_vars_from_js(code)\n for v in vars_found:\n # 一旦发现有并非以 bug_ 开头的变量，即判定为未满足要求\n if not v.startswith('bug_'):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":1,"instruction":"不错，骨架搭建好了。现在我们要把界面做得更精美，特别是虫子的外观需要体现出'特定外观'。请你用纯CSS或者Canvas把这几个单调的小圆点改成具有多节身体和两根小触角的'毛毛虫'形象，并且4条赛道上的虫子要使用不同的醒目颜色区分。另外，给跑道区域加上类似泥土或草地的背景色，把右侧的终点线改成经典的黑白方块相间的格子旗样式。同时请你在后续所有轮次始终遵循以下两项更新要求：第一，原变量名前缀规则更新为在构建代码文件时，请限制所有变量名都以race_bug_打头；第二，限制你生成的代码文件中循环嵌套最多为2层。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断，需要验证的代码/日志文件扩展名\n target_extensions = ('.html', '.css', '.js', '.log')\n \n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if file_name.lower().endswith(target_extensions):\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'rb') as f:\n byte_content = f.read()\n # 尝试使用utf-8编码对其进行解码\n byte_content.decode('utf-8')\n except UnicodeDecodeError:\n # 解码失败，说明该文件不是合法的UTF-8编码\n return False\n except Exception:\n # 如果有其他读取错误，也判定为不符合要求\n return False\n \n # 所有文件均成功通过UTF-8解码测试\n return True","tags":["Language","File Encoding"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在构建代码文件时，请限制所有变量名都以race_bug_打头","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def remove_strings_and_comments(code: str) -> str:\n # 正则表达式按顺序匹配：多行注释 | 单行注释 | 双引号字符串 | 单引号字符串 | 反引号字符串\n # [^...] 在 Python 的正则表达式中默认匹配换行符，所以反引号字符串能够正确跨行解析\n pattern = r'(/\\*[\\s\\S]*?\\*/)|(//.*)|(\"(?:\\\\.|[^\"\\\\])*\")|(\\'(?:\\\\.|[^\\'\\\\])*\\')|(`(?:\\\\.|[^`\\\\])*`)'\n return re.sub(pattern, '', code)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in ['.js', '.html']:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n js_codes = []\n if ext == '.html':\n # 提取 <script> 标签内部的JS代码片段\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, flags=re.DOTALL | re.IGNORECASE)\n js_codes.extend(scripts)\n else:\n js_codes.append(content)\n \n for js_code in js_codes:\n clean_code = remove_strings_and_comments(js_code)\n \n # 匹配 var, let, const 关键字后的第一个合法标识符（即变量名）\n matches = re.finditer(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', clean_code)\n for match in matches:\n var_name = match.group(1)\n # 判断变量是否以 `race_bug_` 前缀开头\n if not var_name.startswith('race_bug_'):\n return False\n \n # 所有被检测到的变量都符合要求，或没有生成变量的情况\n return True","tags":["Style","Naming"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"限制你生成的代码文件中循环嵌套最多为2层","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":2,"instruction":"好，接下来来实现核心的随机赛跑逻辑。点击'开始比赛'按钮后，虫子们要真正开始向终点冲刺。请利用定时器或动画帧实现随机移动：每只虫子每隔一小段时间会随机前进一段距离，你需要调整随机算法，让它们表现出有的走得快、有的走得慢，甚至偶尔会原地停顿一下的真实爬行节奏。当第一只虫子的头部触碰到终点线时，比赛立刻结束，所有虫子停止移动，并在页面中央弹出一个醒目的提示框，宣布几号跑道的虫子获胜。注意，比赛进行期间要禁用'开始比赛'按钮。同时请你在后续所有轮次始终遵循以下两项要求：第一，原循环嵌套限制规则更新为你输出的代码中，禁止出现超过1层的循环嵌套；第二，生成的代码里，“=”号的前后必须各自包含两个空格。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 若工作空间不存在，视为没有违反要求\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n \n # 根据用户指令中涉及的HTML/CSS/JS以及log日志文件，定义需要检查的扩展名集合\n target_exts = {'.html', '.css', '.js', '.log', '.json', '.md', '.txt'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_exts:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件包含不符合 utf-8 编码规则的字符\n return False\n except Exception:\n # 其他诸如权限等异常可以忽略，主要针对编码进行校验\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在构建代码文件时，请限制所有变量名都以race_bug_打头","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n js_codes = []\n \n # 若目录不存在，直接返回未满足\n if not os.path.exists(workspace_path):\n return False\n \n # 提取所有 .js 文件和 .html 文件中的 <script> 标签内容\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js'):\n with open(file_path, 'r', encoding='utf-8') as f:\n js_codes.append(f.read())\n elif file.endswith('.html'):\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n js_codes.extend(scripts)\n \n # 如果没找到任何 js 逻辑代码，说明不满足基本构建需求\n if not js_codes:\n return False\n \n for code in js_codes:\n # 去除多行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n # 去除单行注释\n code = re.sub(r'//.*', '', code)\n # 移除字符串内容并仅保留引号，防止字符串内部字符（如逗号）干扰解析\n code = re.sub(r\"'(?:\\\\'|[^'])*'\", \"''\", code)\n code = re.sub(r'\"(?:\\\\\"|[^\"])*\"', '\"\"', code)\n code = re.sub(r'`(?:\\\\`|[^`])*`', '``', code)\n \n # 将代码转化为简易 Token 流，包括标识符或特殊符号\n tokens = re.findall(r'[a-zA-Z_$][0-9a-zA-Z_$]*|[{}[\\]();=,]', code)\n \n in_decl = False\n brace_level = 0\n paren_level = 0\n bracket_level = 0\n \n i = 0\n while i < len(tokens):\n token = tokens[i]\n \n # 对函数声明名的特殊处理（函数名也算广义上的变量名）\n if token == 'function':\n i += 1\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*$', tokens[i]):\n if not tokens[i].startswith('race_bug_'):\n return False\n continue\n \n # 当遇到变量或者类声明关键字\n if token in ['var', 'let', 'const', 'class']:\n in_decl = True\n brace_level = 0\n paren_level = 0\n bracket_level = 0\n i += 1\n \n # 检查关键字之后的首个标识符前缀\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*$', tokens[i]):\n if not tokens[i].startswith('race_bug_'):\n return False\n continue\n \n # 若在声明语句的环境中\n if in_decl:\n if token == '{': \n brace_level += 1\n elif token == '}':\n brace_level -= 1\n if brace_level < 0: in_decl = False\n elif token == '(': \n paren_level += 1\n elif token == ')':\n paren_level -= 1\n # 适应 for (let i of arr) 或者类似循环\n if paren_level < 0: in_decl = False\n elif token == '[': \n bracket_level += 1\n elif token == ']':\n bracket_level -= 1\n if bracket_level < 0: in_decl = False\n elif token == ';':\n # 正常遇到分号结束声明语句\n in_decl = False\n elif token == ',':\n # 在没有任何嵌套环境中的逗号代表多重声明\n if brace_level == 0 and paren_level == 0 and bracket_level == 0:\n i += 1\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*$', tokens[i]):\n if not tokens[i].startswith('race_bug_'):\n return False\n continue\n i += 1\n\n return True","tags":["Style","Naming"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你输出的代码中，禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"生成的代码里，“=”号的前后必须各自包含两个空格","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n has_code_file = False\n \n # 使用正则表达式匹配独立的 \"=\" 号\n # 运用负向后行断言和负向先行断言，排除复合运算符中的等号：\n # 比如：!=, ==, ===, >=, <=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~=\n pattern = re.compile(r'(?<![=!<>+\\-*/%&|^~])=(?![=>])')\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if file_name.endswith(('.html', '.htm', '.css', '.js')):\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 遍历文件中查找到的每一个独立的 \"=\" 号\n for match in pattern.finditer(content):\n idx = match.start()\n \n # 检查 \"=\" 号前面是否至少有两个字符，并且这两个字符全是空格\n if idx < 2:\n return False\n if content[idx-2:idx] != ' ':\n return False\n \n # 检查 \"=\" 号后面是否至少有两个字符，并且这两个字符全是空格\n if idx + 2 >= len(content):\n return False\n if content[idx+1:idx+3] != ' ':\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"比赛进行期间要禁用'开始比赛'按钮","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"现在来完善赛制和结果展示。我希望比赛不要在第一名出现时就立刻强制终止，而是让剩余的虫子继续跑完。请修改上一步的逻辑，并在页面右侧增加一个'实时排行榜'面板。在比赛过程中，你需要根据每只虫子距离起点的实时X坐标，动态更新它们在排行榜上的名次。当所有虫子都越过终点线后，在排行榜面板上突出显示最终排名，给第一名的名字旁边加上一个金牌或皇冠的emoji。同时，把底部的按钮变成'重新开始'，点击后所有虫子归位到起点，排行榜清空，准备好下一局。同时请你在后续所有轮次始终遵循以下两项要求：第一，原变量名前缀规则更新为在构建代码文件时，请限制所有变量名都以final_race_bug_打头；第二，在处理任何文件之前，强制要求通过运行`pwd`来确定当前目录。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在，若不存在则说明没有生成代码项目\n if not os.path.exists(workspace_path):\n return False\n \n # 定义需要检查的代码及日志文件的扩展名\n target_extensions = {'.html', '.css', '.js', '.log'}\n has_target_file = False\n \n # 遍历工作区下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_target_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n # 如果文件中包含中文等非 ASCII 字符且文件非 UTF-8 编码，此处将抛出 UnicodeDecodeError\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 出现解码错误，判定编码不统一为 UTF-8\n return False\n except Exception:\n # 忽略权限等其他导致无法读取的常规异常\n pass\n \n # 如果找到了目标类型文件且均满足 UTF-8 解码要求，返回 True；如果没有生成任何目标文件，则认为未完成要求\n return has_target_file","tags":["Language","File Encoding"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在构建代码文件时，请限制所有变量名都以final_race_bug_打头","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 用于提取变量名和函数参数的正则\n var_pattern = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n func_pattern = re.compile(r'\\bfunction\\s*[a-zA-Z_$0-9]*\\s*\\(([^)]*)\\)')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅处理js和html文件\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 如果是html文件，提取所有script标签中的代码段\n if file.endswith('.html'):\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n js_content = '\\n'.join(scripts)\n else:\n js_content = content\n\n if not js_content.strip():\n continue\n\n # 移除多行注释和单行注释\n js_content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', js_content)\n js_content = re.sub(r'//.*', '', js_content)\n \n # 移除字符串字面量（处理反斜杠转义的情况）\n js_content = re.sub(r'\"([^\"\\\\]*(\\\\.[^\"\\\\]*)*)\"', '\"\"', js_content)\n js_content = re.sub(r\"'([^'\\\\]*(\\\\.[^'\\\\]*)*)'\", \"''\", js_content)\n js_content = re.sub(r\"`([^`\\\\]*(\\\\.[^`\\\\]*)*)`\", \"``\", js_content)\n\n # 提取代码中的变量\n vars_found = set()\n vars_found.update(var_pattern.findall(js_content))\n\n # 提取函数形参作为变量\n for args_str in func_pattern.findall(js_content):\n args = [arg.strip() for arg in args_str.split(',') if arg.strip()]\n for arg in args:\n # 对于带有默认值的参数（例如：a=1），只取左边的变量名\n arg_name = arg.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', arg_name):\n vars_found.add(arg_name)\n \n # 检查所有提取到的变量名是否均满足最终的变量名前缀要求\n for var in vars_found:\n if not var.startswith('final_race_bug_'):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你输出的代码中，禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"生成的代码里，“=”号的前后必须各自包含两个空格","validation_code":"def verify(response, workspace_path):\n import os\n import re\n \n # 若工作区路径不存在，直接判断为未满足\n if not os.path.exists(workspace_path):\n return False\n \n # 定义正则模式：查找非复合运算符（如 +=, != 等）一部分的独立 '=' 或连续的 '=' 整体 (如 '==', '===')\n pattern = re.compile(r'(?<![!<>=+\\-*/%&|^])(=+)(?![=>])')\n \n code_files = []\n # 扫描目录下所有的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 推断可能由助手生成的前端项目文件类型以及规定的变更日志类型\n if file.endswith(('.html', '.css', '.js', '.log')):\n code_files.append(os.path.join(root, file))\n \n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n for match in pattern.finditer(content):\n start = match.start()\n end = match.end()\n \n # 判断紧接着等号前面的两个字符是否为两个空格\n if start < 2:\n return False\n prefix = content[start-2:start]\n \n # 判断紧接着等号后面的两个字符是否为两个空格\n if end + 2 > len(content):\n return False\n suffix = content[end:end+2]\n \n # 只要有一处等号前或后未满足双空格的要求，立刻返回 False\n if prefix != ' ' or suffix != ' ':\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"在处理任何文件之前，强制要求通过运行`pwd`来确定当前目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"比赛进行期间要禁用'开始比赛'按钮","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"比赛不要在第一名出现时就立刻强制终止，而是让剩余的虫子继续跑完","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在仍有任意一只虫子未越过终点线之前，底部用于开局的主按钮应保持禁用或不可再次触发新的比赛","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":4,"instruction":"最后，为了让比赛过程更加刺激且充满不确定性，我们需要在赛道上加入随机的'路况事件'机制。在比赛进行中，系统每隔2到3秒需要在尚未跑到的赛道空白区域随机生成两种道具：'小水坑'（蓝色椭圆）或'小绿叶'（绿色叶片）。如果虫子在爬行中碰到了'小水坑'，会被困住原地停顿1.5秒；如果碰到了'小绿叶'，则在接下来的1.5秒内获得冲刺效果（每次移动的随机距离翻倍）。虫子触碰到道具后，该道具必须从赛道上消失。请帮我实现这个动态道具生成、碰撞检测以及状态控制逻辑。同时请你在后续所有轮次始终遵循以下两项要求：第一，原变量名前缀规则更新为在构建代码文件时，请限制所有变量名都以ultimate_race_bug_打头；第二，生成的代码文件中包含的每一个函数，其行数上限均应为40行。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，则无法满足条件\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令推断可能生成的代码及日志文件类型\n target_extensions = ('.html', '.css', '.js', '.log')\n has_target_file = False\n\n # 遍历当前项目地址下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if file_name.endswith(target_extensions):\n has_target_file = True\n file_path = os.path.join(root, file_name)\n \n # 尝试以 UTF-8 编码读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出解码异常，说明非纯 UTF-8 编码\n return False\n except Exception:\n # 若由于其他原因(如权限等)导致的异常，本例中不作为编码错误处理\n pass\n\n # 确保项目中至少存在目标文件，且全部通过了 UTF-8 编码检查\n return has_target_file","tags":["Language","File Encoding"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在构建代码文件时，请限制所有变量名都以ultimate_race_bug_打头","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n js_contents = []\n css_contents = []\n \n # 1. 从工作区文件中提取代码\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file.endswith('.js'):\n js_contents.append(content)\n elif file.endswith('.css'):\n css_contents.append(content)\n elif file.endswith('.html'):\n # 提取 HTML 中的 JS 脚本片段\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n js_contents.extend(scripts)\n # 提取 HTML 中的 CSS 样式片段\n styles = re.findall(r'<style.*?>\\s*(.*?)\\s*</style>', content, re.DOTALL | re.IGNORECASE)\n css_contents.extend(styles)\n except Exception:\n pass\n\n # 2. 从助手的 Markdown 回复中提取代码块\n js_blocks = re.findall(r'```(?:javascript|js)\\s+(.*?)\\s*```', response, re.DOTALL | re.IGNORECASE)\n js_contents.extend(js_blocks)\n \n css_blocks = re.findall(r'```css\\s+(.*?)\\s*```', response, re.DOTALL | re.IGNORECASE)\n css_contents.extend(css_blocks)\n \n html_blocks = re.findall(r'```html\\s+(.*?)\\s*```', response, re.DOTALL | re.IGNORECASE)\n for html_content in html_blocks:\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', html_content, re.DOTALL | re.IGNORECASE)\n js_contents.extend(scripts)\n styles = re.findall(r'<style.*?>\\s*(.*?)\\s*</style>', html_content, re.DOTALL | re.IGNORECASE)\n css_contents.extend(styles)\n \n # 若没有任何代码，算作未通过\n if not js_contents and not css_contents:\n return False\n \n found_any_var = False\n \n # 3. 检查 JavaScript 中的变量命名\n for js_content in js_contents:\n # 去除 JS 中的字符串（单引号、双引号、反引号），兼容内部转义符，防止字符串内容干扰\n js_content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', js_content)\n js_content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', js_content)\n js_content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', js_content, flags=re.DOTALL)\n \n # 去除 JS 中的注释（单行和多行）\n js_content = re.sub(r'//.*', '', js_content)\n js_content = re.sub(r'/\\*.*?\\*/', '', js_content, flags=re.DOTALL)\n \n # 匹配以 var/let/const 定义的变量名\n pattern = r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)'\n matches = re.findall(pattern, js_content)\n for match in matches:\n found_any_var = True\n if not match.startswith('ultimate_race_bug_'):\n return False\n\n # 4. 检查 CSS 中的自定义变量命名\n css_pattern = r'--([a-zA-Z0-9_-]+)\\s*:'\n for css_content in css_contents:\n # 去除 CSS 注释\n css_content = re.sub(r'/\\*.*?\\*/', '', css_content, flags=re.DOTALL)\n \n # 匹配 CSS 变量\n matches = re.findall(css_pattern, css_content)\n for match in matches:\n found_any_var = True\n if not match.startswith('ultimate_race_bug_'):\n return False\n\n # 如果代码中至少包含一个变量，且经过检查没有提前返回 False，代表完全达标\n return found_any_var","tags":["Style","Naming"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你输出的代码中，禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"生成的代码里，“=”号的前后必须各自包含两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 确保给定的代码项目路径存在\n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历当前工作区的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 针对前端项目的核心代码文件类型进行验证\n if file.endswith(('.js', '.html', '.css')):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 正则表达式匹配单独的 '='\n # 使用后顾断言 (?<!...) 排除前缀是其他操作符的情况 (如 +, -, *, /, <, >, !, =, &, |, ^, ~, ?, :)\n # 使用前瞻断言 (?![=>]) 排除后缀是 '=' 或 '>' 的情况 (如 ==, =>)\n pattern = r'(?<![<>=!\\+\\-\\*/%&\\|\\^~?:])=(?![=>])'\n \n for match in re.finditer(pattern, content):\n idx = match.start()\n \n # 检查 '=' 前面是否包含（紧挨着）两个空格\n if idx < 2 or content[idx-2:idx] != ' ':\n return False\n \n # 检查 '=' 后面是否包含（紧挨着）两个空格\n if idx + 2 > len(content) or content[idx+1:idx+3] != ' ':\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"在处理任何文件之前，强制要求通过运行`pwd`来确定当前目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"生成的代码文件中包含的每一个函数，其行数上限均应为40行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n def clean_code_for_parsing(code):\n # 替换字符串和注释中的字符为空格，但严格保留换行符，以此保证行号计算完全正确\n def replacer(match):\n text = match.group(0)\n return ''.join('\\n' if c == '\\n' else ' ' for c in text)\n\n # 匹配HTML注释、多行注释、单行注释、双引号、单引号、反引号字符串\n pattern = r'(<!--[\\s\\S]*?-->)|(/\\*[\\s\\S]*?\\*/)|(//[^\\n]*)|(\"(?:\\\\.|[^\"\\\\])*\")|(\\'(?:\\\\.|[^\\'\\\\])*\\')|(`(?:\\\\.|[^`\\\\])*`)'\n cleaned = re.sub(pattern, replacer, code)\n return cleaned\n\n def check_functions_length_in_code(code):\n cleaned = clean_code_for_parsing(code)\n \n # 记录每行的起始索引以备二分查找求出行号\n line_starts = [0]\n for i, c in enumerate(cleaned):\n if c == '\\n':\n line_starts.append(i + 1)\n \n def get_line_num(idx):\n return bisect.bisect_right(line_starts, idx)\n\n for i, c in enumerate(cleaned):\n if c == '{':\n j = i - 1\n while j >= 0 and cleaned[j].isspace():\n j -= 1\n if j < 0:\n continue\n \n is_function = False\n start_idx = i\n \n # 情况1：返回对象字面量的箭头函数，例如 () => ({\n is_arrow_obj = False\n if cleaned[j] == '(':\n k = j - 1\n while k >= 0 and cleaned[k].isspace():\n k -= 1\n if k >= 0 and cleaned[k] == '>' and k - 1 >= 0 and cleaned[k-1] == '=':\n is_arrow_obj = True\n is_function = True\n start_idx = k - 1\n\n # 情况2：常规的块级箭头函数，例如 () => {\n if cleaned[j] == '>' and j - 1 >= 0 and cleaned[j-1] == '=':\n is_function = True\n start_idx = j - 1\n \n if is_function and (cleaned[j] == '>' or is_arrow_obj):\n # 向前寻找可能的参数列表 '('\n m = start_idx - 1\n while m >= 0 and cleaned[m].isspace():\n m -= 1\n if m >= 0 and cleaned[m] == ')':\n depth = 1\n n = m - 1\n while n >= 0 and depth > 0:\n if cleaned[n] == ')': depth += 1\n elif cleaned[n] == '(': depth -= 1\n n -= 1\n if depth == 0:\n start_idx = n + 1\n \n # 情况3：具有圆括号的函数或类方法，例如 function foo() { 或 foo(a, b) {\n elif not is_function and cleaned[j] == ')':\n depth = 1\n m = j - 1\n while m >= 0 and depth > 0:\n if cleaned[m] == ')': depth += 1\n elif cleaned[m] == '(': depth -= 1\n m -= 1\n \n if depth == 0:\n open_paren_idx = m + 1\n k = open_paren_idx - 1\n while k >= 0 and cleaned[k].isspace():\n k -= 1\n \n word_end = k\n while k >= 0 and (cleaned[k].isalnum() or cleaned[k] in '_$'):\n k -= 1\n word_start = k + 1\n word = cleaned[word_start:word_end+1]\n \n # 排除非函数的控制流关键字\n if word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n is_function = True\n start_idx = word_start if word else open_paren_idx\n \n # 进一步排查带有 function 或者 async 关键字的前置情况\n p = word_start - 1\n while p >= 0 and cleaned[p].isspace():\n p -= 1\n p_end = p\n while p >= 0 and (cleaned[p].isalnum() or cleaned[p] in '_$'):\n p -= 1\n prev_word = cleaned[p+1:p_end+1]\n \n if prev_word == 'function':\n start_idx = p + 1\n q = p\n while q >= 0 and cleaned[q].isspace():\n q -= 1\n q_end = q\n while q >= 0 and (cleaned[q].isalnum() or cleaned[q] in '_$'):\n q -= 1\n prev_prev_word = cleaned[q+1:q_end+1]\n if prev_prev_word == 'async':\n start_idx = q + 1\n elif prev_word == 'async':\n start_idx = p + 1\n \n # 如果确认找到了函数体，定位结束符并计算总行数\n if is_function:\n depth = 1\n end_idx = i\n m = i + 1\n while m < len(cleaned) and depth > 0:\n if cleaned[m] == '{': depth += 1\n elif cleaned[m] == '}': depth -= 1\n m += 1\n \n if depth == 0:\n end_idx = m - 1\n start_line = get_line_num(start_idx)\n end_line = get_line_num(end_idx)\n line_count = end_line - start_line + 1\n # 要求每个函数最多不超过 40 行\n if line_count > 40:\n return False\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 扫描独立的JS文件\n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_functions_length_in_code(code):\n return False\n except Exception:\n pass\n # 扫描HTML中的内嵌JS脚本\n elif file.endswith(('.html', '.htm')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n for match in re.finditer(r'<script[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE):\n code = match.group(1)\n if code.strip():\n if not check_functions_length_in_code(code):\n return False\n except Exception:\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"比赛进行期间要禁用'开始比赛'按钮","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"比赛不要在第一名出现时就立刻强制终止，而是让剩余的虫子继续跑完","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在仍有任意一只虫子未越过终点线之前，底部用于开局的主按钮应保持禁用或不可再次触发新的比赛","validation_code":"","tags":["Workflow","Orchestration"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用前端技术（HTML/CSS/JS）写一个虫子赛跑模拟游戏的基础页面。首先，需要在页面中间画出4条水平的赛道，左边是起点线，右边是终点线。在起点处，每条跑道上先放一个简单的彩色小圆点来作为'虫子'的占位符。底部放一个'开始比赛'的按钮。作为初步测试，点击按钮后，只需要让所有代表虫子的小圆点向右平移50像素即可，现在还不需要写完整的随机赛跑逻辑，先帮我把页面架子搭好。另外请你在后续所有轮次始终遵循以下要求：第一，在构建代码文件时，请限制所有变量名都以bug_打头；第二，为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n # 根据用户指令，定义可能生成的代码或日志文件后缀\\n target_extensions = {'.html', '.htm', '.css', '.js', '.log', '.txt', '.md'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 忽略隐藏文件（如 .git, .DS_Store 等）\\n if file_name.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file_name)[1].lower()\\n # 若文件没有后缀或属于目标代码文件，则进行UTF-8编码检测\\n if ext in target_extensions or ext == '':\\n file_path = os.path.join(root, file_name)\\n \\n if not os.path.isfile(file_path):\\n continue\\n \\n has_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明存在非UTF-8编码（如GBK编码的中文字符）\\n return False\\n except Exception:\\n # 忽略文件操作过程中非编码相关的其他系统异常\\n pass\\n\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在构建代码文件时，请限制所有变量名都以bug_打头\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def extract_vars_from_js(code: str):\\n # 1. 去除多行和单行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n code = re.sub(r'//.*', '', code)\\n \\n # 2. 去除字符串（避免字符串里的内容干扰正则匹配）\\n code = re.sub(r'\\\"(?:\\\\\\\\\\\"|[^\\\"])*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\'|[^'])*'\\\", \\\"''\\\", code)\\n code = re.sub(r\\\"`(?:\\\\\\\\`|[^`])*`\\\", \\\"``\\\", code)\\n \\n vars_found = []\\n \\n # 3. 提取由 var/let/const 声明的变量\\n statements = re.finditer(r'\\\\b(var|let|const)\\\\s+([^;\\\\n]+)', code)\\n for match in statements:\\n stmt = match.group(2)\\n # 如果声明部分不包含各种括号，大概率是平铺的变量声明，可以安全按逗号切分\\n if not re.search(r'[\\\\(\\\\)\\\\{\\\\}\\\\[\\\\]]', stmt):\\n parts = stmt.split(',')\\n for part in parts:\\n part = part.strip()\\n if not part: continue\\n # 提取等号左侧的内容\\n lhs = part.split('=')[0].strip()\\n var_match = re.search(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', lhs)\\n if var_match:\\n v_name = var_match.group(1)\\n if v_name not in ('var', 'let', 'const', 'true', 'false', 'null', 'undefined'):\\n vars_found.append(v_name)\\n else:\\n # 语句中含有括号等复杂结构（例如 DOM 操作等），安全降级为仅提取第一个紧跟的标识符\\n var_match = re.search(r'^([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', stmt.strip())\\n if var_match:\\n vars_found.append(var_match.group(1))\\n\\n # 4. 提取函数的参数名称\\n func_params = re.finditer(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', code)\\n for match in func_params:\\n params_str = match.group(1)\\n params = params_str.split(',')\\n for p in params:\\n p = p.strip()\\n if p:\\n # 获取参数标识符\\n pm = re.search(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', p)\\n if pm:\\n v_name = pm.group(1)\\n if v_name not in ('var', 'let', 'const', 'true', 'false', 'null', 'undefined'):\\n vars_found.append(v_name)\\n\\n # 5. 提取函数名称\\n funcs = re.finditer(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', code)\\n for match in funcs:\\n vars_found.append(match.group(1))\\n \\n return set(vars_found)\\n\\n # 获取工作区所有 JS 与 HTML 文件\\n js_html_files = []\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for f in files:\\n if f.endswith('.js') or f.endswith('.html'):\\n js_html_files.append(os.path.join(root, f))\\n \\n if not js_html_files:\\n return False\\n \\n for file_path in js_html_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n codes_to_check = []\\n # 如果是 HTML 文件，则只抓取 <script> 标签中的内容进行分析\\n if file_path.endswith('.html'):\\n scripts = re.findall(r'<script\\\\b[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n codes_to_check.extend(scripts)\\n else:\\n codes_to_check.append(content)\\n \\n for code in codes_to_check:\\n if not code.strip():\\n continue\\n \\n # 解析每段代码里的变量\\n vars_found = extract_vars_from_js(code)\\n for v in vars_found:\\n # 一旦发现有并非以 bug_ 开头的变量，即判定为未满足要求\\n if not v.startswith('bug_'):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，骨架搭建好了。现在我们要把界面做得更精美，特别是虫子的外观需要体现出'特定外观'。请你用纯CSS或者Canvas把这几个单调的小圆点改成具有多节身体和两根小触角的'毛毛虫'形象，并且4条赛道上的虫子要使用不同的醒目颜色区分。另外，给跑道区域加上类似泥土或草地的背景色，把右侧的终点线改成经典的黑白方块相间的格子旗样式。同时请你在后续所有轮次始终遵循以下两项更新要求：第一，原变量名前缀规则更新为在构建代码文件时，请限制所有变量名都以race_bug_打头；第二，限制你生成的代码文件中循环嵌套最多为2层。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断，需要验证的代码/日志文件扩展名\\n target_extensions = ('.html', '.css', '.js', '.log')\\n \\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if file_name.lower().endswith(target_extensions):\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'rb') as f:\\n byte_content = f.read()\\n # 尝试使用utf-8编码对其进行解码\\n byte_content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 解码失败，说明该文件不是合法的UTF-8编码\\n return False\\n except Exception:\\n # 如果有其他读取错误，也判定为不符合要求\\n return False\\n \\n # 所有文件均成功通过UTF-8解码测试\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在构建代码文件时，请限制所有变量名都以race_bug_打头\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def remove_strings_and_comments(code: str) -> str:\\n # 正则表达式按顺序匹配：多行注释 | 单行注释 | 双引号字符串 | 单引号字符串 | 反引号字符串\\n # [^...] 在 Python 的正则表达式中默认匹配换行符，所以反引号字符串能够正确跨行解析\\n pattern = r'(/\\\\*[\\\\s\\\\S]*?\\\\*/)|(//.*)|(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|(\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|(`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n return re.sub(pattern, '', code)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ['.js', '.html']:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n js_codes = []\\n if ext == '.html':\\n # 提取 <script> 标签内部的JS代码片段\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n js_codes.extend(scripts)\\n else:\\n js_codes.append(content)\\n \\n for js_code in js_codes:\\n clean_code = remove_strings_and_comments(js_code)\\n \\n # 匹配 var, let, const 关键字后的第一个合法标识符（即变量名）\\n matches = re.finditer(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', clean_code)\\n for match in matches:\\n var_name = match.group(1)\\n # 判断变量是否以 `race_bug_` 前缀开头\\n if not var_name.startswith('race_bug_'):\\n return False\\n \\n # 所有被检测到的变量都符合要求，或没有生成变量的情况\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"限制你生成的代码文件中循环嵌套最多为2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，接下来来实现核心的随机赛跑逻辑。点击'开始比赛'按钮后，虫子们要真正开始向终点冲刺。请利用定时器或动画帧实现随机移动：每只虫子每隔一小段时间会随机前进一段距离，你需要调整随机算法，让它们表现出有的走得快、有的走得慢，甚至偶尔会原地停顿一下的真实爬行节奏。当第一只虫子的头部触碰到终点线时，比赛立刻结束，所有虫子停止移动，并在页面中央弹出一个醒目的提示框，宣布几号跑道的虫子获胜。注意，比赛进行期间要禁用'开始比赛'按钮。同时请你在后续所有轮次始终遵循以下两项要求：第一，原循环嵌套限制规则更新为你输出的代码中，禁止出现超过1层的循环嵌套；第二，生成的代码里，“=”号的前后必须各自包含两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 若工作空间不存在，视为没有违反要求\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n \\n # 根据用户指令中涉及的HTML/CSS/JS以及log日志文件，定义需要检查的扩展名集合\\n target_exts = {'.html', '.css', '.js', '.log', '.json', '.md', '.txt'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_exts:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件包含不符合 utf-8 编码规则的字符\\n return False\\n except Exception:\\n # 其他诸如权限等异常可以忽略，主要针对编码进行校验\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在构建代码文件时，请限制所有变量名都以race_bug_打头\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n js_codes = []\\n \\n # 若目录不存在，直接返回未满足\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 提取所有 .js 文件和 .html 文件中的 <script> 标签内容\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js'):\\n with open(file_path, 'r', encoding='utf-8') as f:\\n js_codes.append(f.read())\\n elif file.endswith('.html'):\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n js_codes.extend(scripts)\\n \\n # 如果没找到任何 js 逻辑代码，说明不满足基本构建需求\\n if not js_codes:\\n return False\\n \\n for code in js_codes:\\n # 去除多行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n # 去除单行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除字符串内容并仅保留引号，防止字符串内部字符（如逗号）干扰解析\\n code = re.sub(r\\\"'(?:\\\\\\\\'|[^'])*'\\\", \\\"''\\\", code)\\n code = re.sub(r'\\\"(?:\\\\\\\\\\\"|[^\\\"])*\\\"', '\\\"\\\"', code)\\n code = re.sub(r'`(?:\\\\\\\\`|[^`])*`', '``', code)\\n \\n # 将代码转化为简易 Token 流，包括标识符或特殊符号\\n tokens = re.findall(r'[a-zA-Z_$][0-9a-zA-Z_$]*|[{}[\\\\]();=,]', code)\\n \\n in_decl = False\\n brace_level = 0\\n paren_level = 0\\n bracket_level = 0\\n \\n i = 0\\n while i < len(tokens):\\n token = tokens[i]\\n \\n # 对函数声明名的特殊处理（函数名也算广义上的变量名）\\n if token == 'function':\\n i += 1\\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*$', tokens[i]):\\n if not tokens[i].startswith('race_bug_'):\\n return False\\n continue\\n \\n # 当遇到变量或者类声明关键字\\n if token in ['var', 'let', 'const', 'class']:\\n in_decl = True\\n brace_level = 0\\n paren_level = 0\\n bracket_level = 0\\n i += 1\\n \\n # 检查关键字之后的首个标识符前缀\\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*$', tokens[i]):\\n if not tokens[i].startswith('race_bug_'):\\n return False\\n continue\\n \\n # 若在声明语句的环境中\\n if in_decl:\\n if token == '{': \\n brace_level += 1\\n elif token == '}':\\n brace_level -= 1\\n if brace_level < 0: in_decl = False\\n elif token == '(': \\n paren_level += 1\\n elif token == ')':\\n paren_level -= 1\\n # 适应 for (let i of arr) 或者类似循环\\n if paren_level < 0: in_decl = False\\n elif token == '[': \\n bracket_level += 1\\n elif token == ']':\\n bracket_level -= 1\\n if bracket_level < 0: in_decl = False\\n elif token == ';':\\n # 正常遇到分号结束声明语句\\n in_decl = False\\n elif token == ',':\\n # 在没有任何嵌套环境中的逗号代表多重声明\\n if brace_level == 0 and paren_level == 0 and bracket_level == 0:\\n i += 1\\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][0-9a-zA-Z_$]*$', tokens[i]):\\n if not tokens[i].startswith('race_bug_'):\\n return False\\n continue\\n i += 1\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"生成的代码里，“=”号的前后必须各自包含两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n has_code_file = False\\n \\n # 使用正则表达式匹配独立的 \\\"=\\\" 号\\n # 运用负向后行断言和负向先行断言，排除复合运算符中的等号：\\n # 比如：!=, ==, ===, >=, <=, =>, +=, -=, *=, /=, %=, &=, |=, ^=, ~=\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%&|^~])=(?![=>])')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if file_name.endswith(('.html', '.htm', '.css', '.js')):\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 遍历文件中查找到的每一个独立的 \\\"=\\\" 号\\n for match in pattern.finditer(content):\\n idx = match.start()\\n \\n # 检查 \\\"=\\\" 号前面是否至少有两个字符，并且这两个字符全是空格\\n if idx < 2:\\n return False\\n if content[idx-2:idx] != ' ':\\n return False\\n \\n # 检查 \\\"=\\\" 号后面是否至少有两个字符，并且这两个字符全是空格\\n if idx + 2 >= len(content):\\n return False\\n if content[idx+1:idx+3] != ' ':\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"比赛进行期间要禁用'开始比赛'按钮\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在来完善赛制和结果展示。我希望比赛不要在第一名出现时就立刻强制终止，而是让剩余的虫子继续跑完。请修改上一步的逻辑，并在页面右侧增加一个'实时排行榜'面板。在比赛过程中，你需要根据每只虫子距离起点的实时X坐标，动态更新它们在排行榜上的名次。当所有虫子都越过终点线后，在排行榜面板上突出显示最终排名，给第一名的名字旁边加上一个金牌或皇冠的emoji。同时，把底部的按钮变成'重新开始'，点击后所有虫子归位到起点，排行榜清空，准备好下一局。同时请你在后续所有轮次始终遵循以下两项要求：第一，原变量名前缀规则更新为在构建代码文件时，请限制所有变量名都以final_race_bug_打头；第二，在处理任何文件之前，强制要求通过运行`pwd`来确定当前目录。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在，若不存在则说明没有生成代码项目\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义需要检查的代码及日志文件的扩展名\\n target_extensions = {'.html', '.css', '.js', '.log'}\\n has_target_file = False\\n \\n # 遍历工作区下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_target_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n # 如果文件中包含中文等非 ASCII 字符且文件非 UTF-8 编码，此处将抛出 UnicodeDecodeError\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 出现解码错误，判定编码不统一为 UTF-8\\n return False\\n except Exception:\\n # 忽略权限等其他导致无法读取的常规异常\\n pass\\n \\n # 如果找到了目标类型文件且均满足 UTF-8 解码要求，返回 True；如果没有生成任何目标文件，则认为未完成要求\\n return has_target_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在构建代码文件时，请限制所有变量名都以final_race_bug_打头\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 用于提取变量名和函数参数的正则\\n var_pattern = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n func_pattern = re.compile(r'\\\\bfunction\\\\s*[a-zA-Z_$0-9]*\\\\s*\\\\(([^)]*)\\\\)')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅处理js和html文件\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 如果是html文件，提取所有script标签中的代码段\\n if file.endswith('.html'):\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n js_content = '\\\\n'.join(scripts)\\n else:\\n js_content = content\\n\\n if not js_content.strip():\\n continue\\n\\n # 移除多行注释和单行注释\\n js_content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', js_content)\\n js_content = re.sub(r'//.*', '', js_content)\\n \\n # 移除字符串字面量（处理反斜杠转义的情况）\\n js_content = re.sub(r'\\\"([^\\\"\\\\\\\\]*(\\\\\\\\.[^\\\"\\\\\\\\]*)*)\\\"', '\\\"\\\"', js_content)\\n js_content = re.sub(r\\\"'([^'\\\\\\\\]*(\\\\\\\\.[^'\\\\\\\\]*)*)'\\\", \\\"''\\\", js_content)\\n js_content = re.sub(r\\\"`([^`\\\\\\\\]*(\\\\\\\\.[^`\\\\\\\\]*)*)`\\\", \\\"``\\\", js_content)\\n\\n # 提取代码中的变量\\n vars_found = set()\\n vars_found.update(var_pattern.findall(js_content))\\n\\n # 提取函数形参作为变量\\n for args_str in func_pattern.findall(js_content):\\n args = [arg.strip() for arg in args_str.split(',') if arg.strip()]\\n for arg in args:\\n # 对于带有默认值的参数（例如：a=1），只取左边的变量名\\n arg_name = arg.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', arg_name):\\n vars_found.add(arg_name)\\n \\n # 检查所有提取到的变量名是否均满足最终的变量名前缀要求\\n for var in vars_found:\\n if not var.startswith('final_race_bug_'):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"生成的代码里，“=”号的前后必须各自包含两个空格\",\n \"validation_code\": \"def verify(response, workspace_path):\\n import os\\n import re\\n \\n # 若工作区路径不存在，直接判断为未满足\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义正则模式：查找非复合运算符（如 +=, != 等）一部分的独立 '=' 或连续的 '=' 整体 (如 '==', '===')\\n pattern = re.compile(r'(?<![!<>=+\\\\-*/%&|^])(=+)(?![=>])')\\n \\n code_files = []\\n # 扫描目录下所有的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 推断可能由助手生成的前端项目文件类型以及规定的变更日志类型\\n if file.endswith(('.html', '.css', '.js', '.log')):\\n code_files.append(os.path.join(root, file))\\n \\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n for match in pattern.finditer(content):\\n start = match.start()\\n end = match.end()\\n \\n # 判断紧接着等号前面的两个字符是否为两个空格\\n if start < 2:\\n return False\\n prefix = content[start-2:start]\\n \\n # 判断紧接着等号后面的两个字符是否为两个空格\\n if end + 2 > len(content):\\n return False\\n suffix = content[end:end+2]\\n \\n # 只要有一处等号前或后未满足双空格的要求，立刻返回 False\\n if prefix != ' ' or suffix != ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在处理任何文件之前，强制要求通过运行`pwd`来确定当前目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"比赛进行期间要禁用'开始比赛'按钮\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"比赛不要在第一名出现时就立刻强制终止，而是让剩余的虫子继续跑完\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在仍有任意一只虫子未越过终点线之前，底部用于开局的主按钮应保持禁用或不可再次触发新的比赛\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，为了让比赛过程更加刺激且充满不确定性，我们需要在赛道上加入随机的'路况事件'机制。在比赛进行中，系统每隔2到3秒需要在尚未跑到的赛道空白区域随机生成两种道具：'小水坑'（蓝色椭圆）或'小绿叶'（绿色叶片）。如果虫子在爬行中碰到了'小水坑'，会被困住原地停顿1.5秒；如果碰到了'小绿叶'，则在接下来的1.5秒内获得冲刺效果（每次移动的随机距离翻倍）。虫子触碰到道具后，该道具必须从赛道上消失。请帮我实现这个动态道具生成、碰撞检测以及状态控制逻辑。同时请你在后续所有轮次始终遵循以下两项要求：第一，原变量名前缀规则更新为在构建代码文件时，请限制所有变量名都以ultimate_race_bug_打头；第二，生成的代码文件中包含的每一个函数，其行数上限均应为40行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，则无法满足条件\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令推断可能生成的代码及日志文件类型\\n target_extensions = ('.html', '.css', '.js', '.log')\\n has_target_file = False\\n\\n # 遍历当前项目地址下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if file_name.endswith(target_extensions):\\n has_target_file = True\\n file_path = os.path.join(root, file_name)\\n \\n # 尝试以 UTF-8 编码读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出解码异常，说明非纯 UTF-8 编码\\n return False\\n except Exception:\\n # 若由于其他原因(如权限等)导致的异常，本例中不作为编码错误处理\\n pass\\n\\n # 确保项目中至少存在目标文件，且全部通过了 UTF-8 编码检查\\n return has_target_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，请排除绝对路径，所有文件引用均需基于相对路径来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在构建代码文件时，请限制所有变量名都以ultimate_race_bug_打头\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n js_contents = []\\n css_contents = []\\n \\n # 1. 从工作区文件中提取代码\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file.endswith('.js'):\\n js_contents.append(content)\\n elif file.endswith('.css'):\\n css_contents.append(content)\\n elif file.endswith('.html'):\\n # 提取 HTML 中的 JS 脚本片段\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n js_contents.extend(scripts)\\n # 提取 HTML 中的 CSS 样式片段\\n styles = re.findall(r'<style.*?>\\\\s*(.*?)\\\\s*</style>', content, re.DOTALL | re.IGNORECASE)\\n css_contents.extend(styles)\\n except Exception:\\n pass\\n\\n # 2. 从助手的 Markdown 回复中提取代码块\\n js_blocks = re.findall(r'```(?:javascript|js)\\\\s+(.*?)\\\\s*```', response, re.DOTALL | re.IGNORECASE)\\n js_contents.extend(js_blocks)\\n \\n css_blocks = re.findall(r'```css\\\\s+(.*?)\\\\s*```', response, re.DOTALL | re.IGNORECASE)\\n css_contents.extend(css_blocks)\\n \\n html_blocks = re.findall(r'```html\\\\s+(.*?)\\\\s*```', response, re.DOTALL | re.IGNORECASE)\\n for html_content in html_blocks:\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', html_content, re.DOTALL | re.IGNORECASE)\\n js_contents.extend(scripts)\\n styles = re.findall(r'<style.*?>\\\\s*(.*?)\\\\s*</style>', html_content, re.DOTALL | re.IGNORECASE)\\n css_contents.extend(styles)\\n \\n # 若没有任何代码，算作未通过\\n if not js_contents and not css_contents:\\n return False\\n \\n found_any_var = False\\n \\n # 3. 检查 JavaScript 中的变量命名\\n for js_content in js_contents:\\n # 去除 JS 中的字符串（单引号、双引号、反引号），兼容内部转义符，防止字符串内容干扰\\n js_content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', js_content)\\n js_content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', js_content)\\n js_content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', js_content, flags=re.DOTALL)\\n \\n # 去除 JS 中的注释（单行和多行）\\n js_content = re.sub(r'//.*', '', js_content)\\n js_content = re.sub(r'/\\\\*.*?\\\\*/', '', js_content, flags=re.DOTALL)\\n \\n # 匹配以 var/let/const 定义的变量名\\n pattern = r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)'\\n matches = re.findall(pattern, js_content)\\n for match in matches:\\n found_any_var = True\\n if not match.startswith('ultimate_race_bug_'):\\n return False\\n\\n # 4. 检查 CSS 中的自定义变量命名\\n css_pattern = r'--([a-zA-Z0-9_-]+)\\\\s*:'\\n for css_content in css_contents:\\n # 去除 CSS 注释\\n css_content = re.sub(r'/\\\\*.*?\\\\*/', '', css_content, flags=re.DOTALL)\\n \\n # 匹配 CSS 变量\\n matches = re.findall(css_pattern, css_content)\\n for match in matches:\\n found_any_var = True\\n if not match.startswith('ultimate_race_bug_'):\\n return False\\n\\n # 如果代码中至少包含一个变量，且经过检查没有提前返回 False，代表完全达标\\n return found_any_var\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"生成的代码里，“=”号的前后必须各自包含两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 确保给定的代码项目路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历当前工作区的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 针对前端项目的核心代码文件类型进行验证\\n if file.endswith(('.js', '.html', '.css')):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 正则表达式匹配单独的 '='\\n # 使用后顾断言 (?<!...) 排除前缀是其他操作符的情况 (如 +, -, *, /, <, >, !, =, &, |, ^, ~, ?, :)\\n # 使用前瞻断言 (?![=>]) 排除后缀是 '=' 或 '>' 的情况 (如 ==, =>)\\n pattern = r'(?<![<>=!\\\\+\\\\-\\\\*/%&\\\\|\\\\^~?:])=(?![=>])'\\n \\n for match in re.finditer(pattern, content):\\n idx = match.start()\\n \\n # 检查 '=' 前面是否包含（紧挨着）两个空格\\n if idx < 2 or content[idx-2:idx] != ' ':\\n return False\\n \\n # 检查 '=' 后面是否包含（紧挨着）两个空格\\n if idx + 2 > len(content) or content[idx+1:idx+3] != ' ':\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在处理任何文件之前，强制要求通过运行`pwd`来确定当前目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中包含的每一个函数，其行数上限均应为40行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n def clean_code_for_parsing(code):\\n # 替换字符串和注释中的字符为空格，但严格保留换行符，以此保证行号计算完全正确\\n def replacer(match):\\n text = match.group(0)\\n return ''.join('\\\\n' if c == '\\\\n' else ' ' for c in text)\\n\\n # 匹配HTML注释、多行注释、单行注释、双引号、单引号、反引号字符串\\n pattern = r'(<!--[\\\\s\\\\S]*?-->)|(/\\\\*[\\\\s\\\\S]*?\\\\*/)|(//[^\\\\n]*)|(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|(\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|(`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n cleaned = re.sub(pattern, replacer, code)\\n return cleaned\\n\\n def check_functions_length_in_code(code):\\n cleaned = clean_code_for_parsing(code)\\n \\n # 记录每行的起始索引以备二分查找求出行号\\n line_starts = [0]\\n for i, c in enumerate(cleaned):\\n if c == '\\\\n':\\n line_starts.append(i + 1)\\n \\n def get_line_num(idx):\\n return bisect.bisect_right(line_starts, idx)\\n\\n for i, c in enumerate(cleaned):\\n if c == '{':\\n j = i - 1\\n while j >= 0 and cleaned[j].isspace():\\n j -= 1\\n if j < 0:\\n continue\\n \\n is_function = False\\n start_idx = i\\n \\n # 情况1：返回对象字面量的箭头函数，例如 () => ({\\n is_arrow_obj = False\\n if cleaned[j] == '(':\\n k = j - 1\\n while k >= 0 and cleaned[k].isspace():\\n k -= 1\\n if k >= 0 and cleaned[k] == '>' and k - 1 >= 0 and cleaned[k-1] == '=':\\n is_arrow_obj = True\\n is_function = True\\n start_idx = k - 1\\n\\n # 情况2：常规的块级箭头函数，例如 () => {\\n if cleaned[j] == '>' and j - 1 >= 0 and cleaned[j-1] == '=':\\n is_function = True\\n start_idx = j - 1\\n \\n if is_function and (cleaned[j] == '>' or is_arrow_obj):\\n # 向前寻找可能的参数列表 '('\\n m = start_idx - 1\\n while m >= 0 and cleaned[m].isspace():\\n m -= 1\\n if m >= 0 and cleaned[m] == ')':\\n depth = 1\\n n = m - 1\\n while n >= 0 and depth > 0:\\n if cleaned[n] == ')': depth += 1\\n elif cleaned[n] == '(': depth -= 1\\n n -= 1\\n if depth == 0:\\n start_idx = n + 1\\n \\n # 情况3：具有圆括号的函数或类方法，例如 function foo() { 或 foo(a, b) {\\n elif not is_function and cleaned[j] == ')':\\n depth = 1\\n m = j - 1\\n while m >= 0 and depth > 0:\\n if cleaned[m] == ')': depth += 1\\n elif cleaned[m] == '(': depth -= 1\\n m -= 1\\n \\n if depth == 0:\\n open_paren_idx = m + 1\\n k = open_paren_idx - 1\\n while k >= 0 and cleaned[k].isspace():\\n k -= 1\\n \\n word_end = k\\n while k >= 0 and (cleaned[k].isalnum() or cleaned[k] in '_$'):\\n k -= 1\\n word_start = k + 1\\n word = cleaned[word_start:word_end+1]\\n \\n # 排除非函数的控制流关键字\\n if word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n is_function = True\\n start_idx = word_start if word else open_paren_idx\\n \\n # 进一步排查带有 function 或者 async 关键字的前置情况\\n p = word_start - 1\\n while p >= 0 and cleaned[p].isspace():\\n p -= 1\\n p_end = p\\n while p >= 0 and (cleaned[p].isalnum() or cleaned[p] in '_$'):\\n p -= 1\\n prev_word = cleaned[p+1:p_end+1]\\n \\n if prev_word == 'function':\\n start_idx = p + 1\\n q = p\\n while q >= 0 and cleaned[q].isspace():\\n q -= 1\\n q_end = q\\n while q >= 0 and (cleaned[q].isalnum() or cleaned[q] in '_$'):\\n q -= 1\\n prev_prev_word = cleaned[q+1:q_end+1]\\n if prev_prev_word == 'async':\\n start_idx = q + 1\\n elif prev_word == 'async':\\n start_idx = p + 1\\n \\n # 如果确认找到了函数体，定位结束符并计算总行数\\n if is_function:\\n depth = 1\\n end_idx = i\\n m = i + 1\\n while m < len(cleaned) and depth > 0:\\n if cleaned[m] == '{': depth += 1\\n elif cleaned[m] == '}': depth -= 1\\n m += 1\\n \\n if depth == 0:\\n end_idx = m - 1\\n start_line = get_line_num(start_idx)\\n end_line = get_line_num(end_idx)\\n line_count = end_line - start_line + 1\\n # 要求每个函数最多不超过 40 行\\n if line_count > 40:\\n return False\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 扫描独立的JS文件\\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_functions_length_in_code(code):\\n return False\\n except Exception:\\n pass\\n # 扫描HTML中的内嵌JS脚本\\n elif file.endswith(('.html', '.htm')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n for match in re.finditer(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE):\\n code = match.group(1)\\n if code.strip():\\n if not check_functions_length_in_code(code):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"比赛进行期间要禁用'开始比赛'按钮\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"比赛不要在第一名出现时就立刻强制终止，而是让剩余的虫子继续跑完\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在仍有任意一只虫子未越过终点线之前，底部用于开局的主按钮应保持禁用或不可再次触发新的比赛\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["观察页面布局与样式，检查页面中间是否显示4条并排的赛道，赛道区域是否带有类似泥土或草地的背景色，赛道左侧设有起点线，右侧设有黑白方块相间的格子旗样式终点线。","观察赛道起点处的虫子外观，确认每一只虫子都具有多节的身体和两根小触角，且4条赛道上的虫子使用了4种不同且醒目的颜色进行区分。","点击页面底部的「开始比赛」按钮，观察虫子是否开始向右朝终点方向移动；在仍有任意一只虫子未越过终点线之前，底部用于开局的主按钮应保持禁用或不可再次触发新的比赛；当全部4只虫子均越过终点线后，该按钮文案应变更为「重新开始」并变为可点击。","在比赛进行期间观察各赛道上虫子的移动轨迹，确认虫子的移动不是匀速的，而是表现出时快时慢、甚至是偶尔原地停顿的随机爬行节奏，且各虫子的速度相互独立。","在比赛进行期间，检查页面右侧是否显示「实时排行榜」面板，并在虫子前后位置发生交替时，确认排行榜上的名次顺序能根据虫子的实时X坐标动态更新变动。","比赛过程中，观察各赛道前方虫子尚未跑到的空白区域，检查系统是否会每隔约2到3秒随机生成蓝色的椭圆（小水坑）或绿色的叶片（小绿叶）道具。","观察并确认虫子触碰「小水坑」的交互效果：当虫子头部接触到蓝色椭圆时，小水坑立刻从赛道上消失，并且该虫子会在原地停顿约1.5秒后才继续前进。","观察并确认虫子触碰「小绿叶」的交互效果：当虫子头部接触到绿色叶片时，小绿叶立刻从赛道上消失，并且该虫子在接下来的约1.5秒内每次移动距离显著增加（冲刺加速效果）。","当跑在最前面的第一只虫子抵达终点线时，检查比赛是否没有立刻终止，确认剩余未到终点的虫子能继续正常向前移动直到跑完赛程。","当所有4只虫子全部越过终点线后，检查右侧排行榜面板是否突出显示了最终排名，并确认第一名的名字旁边成功展示了金牌或皇冠的 emoji 图标。","当全部虫子完成比赛后，检查底部的按钮文字是否变成了「重新开始」；点击「重新开始」按钮，检查所有虫子是否瞬间归位到起点处，排行榜被清空，赛道上的残余道具消失，界面恢复至初始准备状态。"],"string":"[\n \"观察页面布局与样式，检查页面中间是否显示4条并排的赛道，赛道区域是否带有类似泥土或草地的背景色，赛道左侧设有起点线，右侧设有黑白方块相间的格子旗样式终点线。\",\n \"观察赛道起点处的虫子外观，确认每一只虫子都具有多节的身体和两根小触角，且4条赛道上的虫子使用了4种不同且醒目的颜色进行区分。\",\n \"点击页面底部的「开始比赛」按钮，观察虫子是否开始向右朝终点方向移动；在仍有任意一只虫子未越过终点线之前，底部用于开局的主按钮应保持禁用或不可再次触发新的比赛；当全部4只虫子均越过终点线后，该按钮文案应变更为「重新开始」并变为可点击。\",\n \"在比赛进行期间观察各赛道上虫子的移动轨迹，确认虫子的移动不是匀速的，而是表现出时快时慢、甚至是偶尔原地停顿的随机爬行节奏，且各虫子的速度相互独立。\",\n \"在比赛进行期间，检查页面右侧是否显示「实时排行榜」面板，并在虫子前后位置发生交替时，确认排行榜上的名次顺序能根据虫子的实时X坐标动态更新变动。\",\n \"比赛过程中，观察各赛道前方虫子尚未跑到的空白区域，检查系统是否会每隔约2到3秒随机生成蓝色的椭圆（小水坑）或绿色的叶片（小绿叶）道具。\",\n \"观察并确认虫子触碰「小水坑」的交互效果：当虫子头部接触到蓝色椭圆时，小水坑立刻从赛道上消失，并且该虫子会在原地停顿约1.5秒后才继续前进。\",\n \"观察并确认虫子触碰「小绿叶」的交互效果：当虫子头部接触到绿色叶片时，小绿叶立刻从赛道上消失，并且该虫子在接下来的约1.5秒内每次移动距离显著增加（冲刺加速效果）。\",\n \"当跑在最前面的第一只虫子抵达终点线时，检查比赛是否没有立刻终止，确认剩余未到终点的虫子能继续正常向前移动直到跑完赛程。\",\n \"当所有4只虫子全部越过终点线后，检查右侧排行榜面板是否突出显示了最终排名，并确认第一名的名字旁边成功展示了金牌或皇冠的 emoji 图标。\",\n \"当全部虫子完成比赛后，检查底部的按钮文字是否变成了「重新开始」；点击「重新开始」按钮，检查所有虫子是否瞬间归位到起点处，排行榜被清空，赛道上的残余道具消失，界面恢复至初始准备状态。\"\n]","isConversational":false}}},{"rowIdx":60,"cells":{"task_id":{"kind":"number","value":61,"string":"61"},"repository_policy":{"kind":"string","value":"请你在后续所有交互轮次中严格遵循以下所有约束条件：\n1. 请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式。\n2. 确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-。\n3. 要求你生成的各个代码文件中，所有的注释语句只能是中文。\n4. 你提供的代码项目，其目录层级最高不能超过 3 层。\n5. 请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层。\n6. 每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件。\n7. 无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数。\n8. 生成代码时，请保证文件中所有的函数名只包含小写字母与下划线。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断主流的常见代码文件扩展名，排除不支持注释的配置格式和纯文本/文档格式\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', '.js', '.jsx', \n '.ts', '.tsx', '.go', '.rs', '.rb', '.php', '.html', '.css', \n '.sh', '.bash', '.swift', '.kt', '.scala', '.m', '.mm'\n }\n \n # 如果项目目录不存在，说明没有生成代码文件，可以视作未违反规则\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录(如 .git, .idea)和特定的依赖/编译输出目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env', 'build', 'dist')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查指定声明是否在文件的第一行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于某种原因（例如二进制文件损坏、错误的编码格式）无法作为utf-8纯文本读取第一行，则不满足要求\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区不存在或者不是目录，则认为没有违反该约束条件\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n # 遍历代码项目的所有目录树\n for root, dirs, files in os.walk(workspace_path):\n # 获取当前目录相对于项目根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n if rel_path == '.':\n depth = 0\n else:\n # 利用系统对应的目录分隔符(os.sep)对相对路径进行分割以计算层级\n depth = len(rel_path.split(os.sep))\n \n # 如果任一目录深度的层级超过3层，则不符合要求\n if depth > 3:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作空间路径是否有效\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义常见的代码文件后缀类型\n code_exts = ('.py', '.java', '.js', '.ts', '.html', '.css', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh')\n \n has_code_file = False\n \n # 遍历代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选出代码文件\n if file.endswith(code_exts):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 判断当前代码文件的总行数是否为奇数，如果是则不符合要求\n if len(lines) % 2 != 0:\n return False\n \n except Exception:\n # 如果文件无法正常作为文本读取，保险起见判定为不满足要求\n return False\n\n # 若至少包含一个代码文件且全部代码文件行数均为偶数，则满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def check_function_names(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n source = f.read()\n \n tree = ast.parse(source)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not re.match(r'^[a-z_]+$', node.name):\n return False\n except Exception:\n continue\n\n return True","tags":["Style","Naming"]}],"string":"[\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断主流的常见代码文件扩展名，排除不支持注释的配置格式和纯文本/文档格式\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', '.js', '.jsx', \\n '.ts', '.tsx', '.go', '.rs', '.rb', '.php', '.html', '.css', \\n '.sh', '.bash', '.swift', '.kt', '.scala', '.m', '.mm'\\n }\\n \\n # 如果项目目录不存在，说明没有生成代码文件，可以视作未违反规则\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录(如 .git, .idea)和特定的依赖/编译输出目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env', 'build', 'dist')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查指定声明是否在文件的第一行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于某种原因（例如二进制文件损坏、错误的编码格式）无法作为utf-8纯文本读取第一行，则不满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区不存在或者不是目录，则认为没有违反该约束条件\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历代码项目的所有目录树\\n for root, dirs, files in os.walk(workspace_path):\\n # 获取当前目录相对于项目根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n if rel_path == '.':\\n depth = 0\\n else:\\n # 利用系统对应的目录分隔符(os.sep)对相对路径进行分割以计算层级\\n depth = len(rel_path.split(os.sep))\\n \\n # 如果任一目录深度的层级超过3层，则不符合要求\\n if depth > 3:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作空间路径是否有效\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义常见的代码文件后缀类型\\n code_exts = ('.py', '.java', '.js', '.ts', '.html', '.css', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh')\\n \\n has_code_file = False\\n \\n # 遍历代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选出代码文件\\n if file.endswith(code_exts):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 判断当前代码文件的总行数是否为奇数，如果是则不符合要求\\n if len(lines) % 2 != 0:\\n return False\\n \\n except Exception:\\n # 如果文件无法正常作为文本读取，保险起见判定为不满足要求\\n return False\\n\\n # 若至少包含一个代码文件且全部代码文件行数均为偶数，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def check_function_names(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n source = f.read()\\n \\n tree = ast.parse(source)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not re.match(r'^[a-z_]+$', node.name):\\n return False\\n except Exception:\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我实现一个RGB颜色猜谜游戏的基础网页框架。需要包含：1. 顶部的标题区域，标题下方留有一个大字号的文本区用来显示'RGB(XXX, XXX, XXX)'占位；2. 中间是一个状态控制栏，左侧显示'提示信息'（比如'猜猜看'），右侧放一个'新游戏'按钮；3. 底部是一个颜色块的容器，先在里面放3个正方形的div作为颜色块占位符，默认给点灰色背景即可。整个页面请居中布局，采用现代、简洁的卡片式设计。这一轮不需要写游戏的核心判断逻辑，把UI骨架和基本样式搭好即可。另外请你在后续所有轮次都严格遵循以下要求：一是每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \"RGB猜谜游戏\" 这一关键词；二是保持你生成的代码文件里每个函数的圈复杂度均在 2 以内。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_extensions = {'.py', '.html', '.css', '.js'}\n has_code_files = False\n\n for root, _, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in target_extensions:\n has_code_files = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n return False\n\n return has_code_files","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，视作未生成有效的项目结构\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 获取当前目录相对于根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 判断并计算目录层级深度\n if rel_path == '.':\n depth = 1\n else:\n # 通过路径分隔符拆分相对路径\n parts = rel_path.split(os.sep)\n depth = len(parts) + 1\n \n # 如果目录层级深度超过 3 层，则不满足要求\n if depth > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 依据前端及Python常见后缀识别代码文件\n code_extensions = {'.py', '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue'}\n file_count = 0\n \n # 遍历工作区目录，查找所有生成的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符 '\\n' 分割可以准确还原编辑器中展示的行数（精确包含代码行和所有空行）\n lines = content.split('\\n')\n \n # 判断当前代码文件的总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果文件无法按预期读取（如编码错误或权限问题），为了容错继续检查下一个\n pass\n \n # 如果没有找到任何代码文件，视作未满足要求\n if file_count == 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 匹配 JavaScript 常见的函数定义格式\n # 1. 匹配常规函数: function func_name (\n js_func_pattern1 = re.compile(r\"(?:async\\s+)?function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(\")\n # 2. 匹配函数表达式/箭头函数赋值: const func_name = function( / () =>\n js_func_pattern2 = re.compile(r\"(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)\")\n # 匹配 Python 常规函数（降级使用）\n py_func_pattern = re.compile(r\"^\\s*(?:async\\s+)?def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(\", re.MULTILINE)\n\n # 遍历项目目录下的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤相关的代码文件后缀\n if ext not in ['.py', '.js', '.html', '.ts', '.vue', '.jsx', '.tsx']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 清除HTML以及JS多行注释，避免误提取注释中的内容\n content_cleaned = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content_cleaned = re.sub(r'/\\*.*?\\*/', '', content_cleaned, flags=re.DOTALL)\n \n func_names = []\n \n # 提取 Python 文件中的函数名\n if ext == '.py':\n try:\n tree = ast.parse(content_cleaned)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n func_names.append(node.name)\n except Exception:\n # 如果 AST 解析失败，降级使用正则匹配\n func_names.extend(py_func_pattern.findall(content_cleaned))\n \n # 提取 JS / HTML 包含的前端函数名\n else:\n func_names.extend(js_func_pattern1.findall(content_cleaned))\n func_names.extend(js_func_pattern2.findall(content_cleaned))\n \n # 匹配可能的类方法声明或简写方法，例如: my_method() {\n lines = content_cleaned.split('\\n')\n for line in lines:\n m = re.match(r\"^\\s*(?:async\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{\", line)\n if m:\n name = m.group(1)\n # 排除掉 JavaScript 里的流程控制关键字\n if name not in ['if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else']:\n func_names.append(name)\n \n # 验证所有提取到的函数名\n for name in func_names:\n # 判断函数名是否【只】包含小写字母与下划线\n if not re.match(r\"^[a-z_]+$\", name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \"RGB猜谜游戏\" 这一关键词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 检查工作目录是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n code_files = []\n \n # 查找指定扩展名的所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files.append(os.path.join(root, file))\n \n # 若未找到任何代码文件，则无法满足每个代码文件皆有此注释的要求\n if not code_files:\n return False\n \n keyword = \"RGB猜谜游戏\"\n \n # 逐一验证每个代码文件\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n ext = os.path.splitext(file_path)[1].lower()\n comments = []\n \n # 按照文件类型匹配所有的注释\n if ext in {'.html', '.htm', '.vue'}:\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'(?<![:\"\\'`])//(.*?)$', content, re.MULTILINE))\n elif ext in {'.css', '.js', '.ts', '.jsx', '.tsx'}:\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'(?<![:\"\\'`])//(.*?)$', content, re.MULTILINE))\n elif ext == '.py':\n comments.extend(re.findall(r'#(.*?)$', content, re.MULTILINE))\n comments.extend(re.findall(r'\"\"\"(.*?)\"\"\"', content, re.DOTALL))\n comments.extend(re.findall(r\"'''(.*?)'''\", content, re.DOTALL))\n \n # 统计在当前文件所有注释内容中目标关键词累计出现次数\n total_count = sum(comment.count(keyword) for comment in comments)\n \n # 要求恰好出现 1 次\n if total_count != 1:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 2 以内","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":1,"instruction":"好，接下来把核心的游戏逻辑加上。1. 页面加载或点击'新游戏'按钮时，随机生成一个目标RGB颜色值，并更新显示在顶部的文本区；2. 给底部的3个颜色块分别赋上完全随机的RGB背景色，但其中必须有一个颜色块正好是目标颜色；3. 点击颜色块时进行判断逻辑：如果点对了，状态控制栏的提示文字变成'回答正确！'；如果点错了，提示文字变成'再试一次！'。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，你输出的代码文件里每个函数的圈复杂度均不得大于3；第二，作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令，“RGB颜色猜谜游戏的基础网页框架”主要是前端项目\n code_extensions = {'.html', '.css', '.js'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n return False\n \n # 如果没有找到任何代码文件，则意味着没有按预期生成代码，返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，则代表目前尚未生成项目，可视为不违反限制\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 计算当前所在目录相对于工作区根目录的相对路径\n rel_root = os.path.relpath(root, workspace_path)\n \n # 根目录自身的相对路径通常解析为 '.'，深度为 0\n if rel_root == '.':\n depth = 0\n else:\n # 兼容不同操作系统的路径分隔符，将反斜杠替换为斜杠，再切分出路径组件列表\n depth = len([p for p in rel_root.replace('\\\\', '/').split('/') if p])\n \n # 若当前目录本身的层级大于 3，则判定不符合要求\n if depth > 3:\n return False\n \n # 当前目录内的文件，层级等于当前目录层级 + 1\n for f in files:\n if depth + 1 > 3:\n return False\n \n # 当前目录内的子文件夹，层级等于当前目录层级 + 1\n for d in dirs:\n if depth + 1 > 3:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\n \n # 检查工作空间目录是否存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行行数检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计行数，包含代码行与空行\n if content == \"\":\n line_count = 0\n else:\n line_count = len(content.split('\\n'))\n \n # 如果行数不是偶数，则不符合要求\n if line_count % 2 != 0:\n return False\n except Exception:\n # 忽略读取异常的文件（例如非文本文件等特殊情况）\n pass\n \n # 所有代码文件的行数均满足偶数要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n\n # 针对 JavaScript/TypeScript 及内嵌在 HTML 中的 JS 定义一系列正则来提取函数名\n # 1. 传统函数声明: function funcName (\n js_func_pattern1 = re.compile(r'function\\s+([A-Za-z0-9_]+)\\s*\\(')\n # 2. 变量赋值形式: const funcName = function (\n js_func_pattern2 = re.compile(r'(?:const|let|var)\\s+([A-Za-z0-9_]+)\\s*=\\s*function\\s*\\(')\n # 3. 变量赋值箭头函数: const funcName = (...) => { 或者 const funcName = arg => {\n js_func_pattern3 = re.compile(r'(?:const|let|var)\\s+([A-Za-z0-9_]+)\\s*=\\s*(?:\\([^)]*\\)|[A-Za-z0-9_]+)\\s*=>')\n # 4. 类或对象的简写方法: methodName(...) { (多行模式下以行首或空白符开头，以防匹配到流程控制语句)\n js_func_pattern4 = re.compile(r'^\\s*([A-Za-z0-9_]+)\\s*\\([^)]*\\)\\s*\\{', re.MULTILINE)\n # 5. 对象属性赋值函数: key: function (\n js_func_pattern5 = re.compile(r'([A-Za-z0-9_]+)\\s*:\\s*function\\s*\\(')\n # 6. 对象属性赋值箭头函数: key: (...) => {\n js_func_pattern6 = re.compile(r'([A-Za-z0-9_]+)\\s*:\\s*(?:\\([^)]*\\)|[A-Za-z0-9_]+)\\s*=>')\n # 7. 对象属性绑定函数: obj.prop = function (\n js_func_pattern7 = re.compile(r'\\.\\s*([A-Za-z0-9_]+)\\s*=\\s*function\\s*\\(')\n # 8. 对象属性绑定箭头函数: obj.prop = (...) => {\n js_func_pattern8 = re.compile(r'\\.\\s*([A-Za-z0-9_]+)\\s*=\\s*(?:\\([^)]*\\)|[A-Za-z0-9_]+)\\s*=>')\n\n # 需要过滤掉的 JS 关键字，以防被 js_func_pattern4 等误匹配\n js_keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', \n 'return', 'else', 'do', 'try', 'finally', 'with', \n 'constructor', 'get', 'set'\n }\n\n # 限定检查名称是否严格只包含小写字母与下划线\n valid_name_pattern = re.compile(r'^[a-z_]+$')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # Python文件的解析使用标准库ast\n if ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n tree = ast.parse(f.read())\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not valid_name_pattern.match(node.name):\n return False\n except Exception:\n pass\n \n # 网页相关前端文件的解析\n elif ext in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n funcs = []\n funcs.extend(js_func_pattern1.findall(content))\n funcs.extend(js_func_pattern2.findall(content))\n funcs.extend(js_func_pattern3.findall(content))\n \n # 模式4单独过滤屏蔽字\n for match in js_func_pattern4.findall(content):\n if match not in js_keywords:\n funcs.append(match)\n \n funcs.extend(js_func_pattern5.findall(content))\n funcs.extend(js_func_pattern6.findall(content))\n funcs.extend(js_func_pattern7.findall(content))\n funcs.extend(js_func_pattern8.findall(content))\n \n for func in funcs:\n # 纯数字不可能是正常的函数名，过滤极罕见的误匹配\n if re.match(r'^\\d+$', func):\n continue\n \n # 验证找到的函数名是否只含有小写字母和下划线\n if not valid_name_pattern.match(func):\n return False\n except Exception:\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \"RGB猜谜游戏\" 这一关键词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 涵盖当前网页项目和Python可能用到的文件类型\n code_exts = {'.py', '.js', '.css', '.html', '.htm', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏文件夹（如 .git 等版本控制文件夹）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 根据不同扩展名的代码文件提取注释内容\n comments = []\n if ext == '.py':\n # 匹配 Python 中的单行注释及多行注释（三引号字符串）\n comments = re.findall(r'#.*|(?s)\\'\\'\\'.*?\\'\\'\\'|(?s)\"\"\".*?\"\"\"', content)\n elif ext in ['.html', '.htm']:\n # HTML 文件中可能混写 JS/CSS，因此同时匹配 HTML 注释、斜杠注释和多行注释\n # (?<!:) 防止将 url 中的 http:// 误判为注释\n comments = re.findall(r'(?s)<!--.*?-->|(?<!:)//.*|(?s)/\\*.*?\\*/', content)\n elif ext in ['.js', '.css', '.ts', '.jsx', '.tsx']:\n # 匹配 JS、CSS 中的单行和多行注释\n comments = re.findall(r'(?<!:)//.*|(?s)/\\*.*?\\*/', content)\n \n # 统计从该文件中提取出来的所有注释内，关键词出现的总次数\n count = sum(comment.count(\"RGB猜谜游戏\") for comment in comments)\n \n # 要求每个代码文件的注释中，必须恰好出现 1 次\n if count != 1:\n return False\n \n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你输出的代码文件里每个函数的圈复杂度均不得大于3","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":2,"instruction":"现在的交互太单调了，我们来增强视觉反馈。1. 如果玩家点击了错误的颜色块，让该颜色块通过CSS动画淡出（比如透明度变成0且不可再点击）；2. 如果玩家点击了正确的颜色块，将页面标题区域的背景色、以及所有当前可见颜色块的背景色，全都带有过渡动画地变成这个正确的目标颜色；3. 答对后，把'新游戏'按钮的文字改成'再玩一次？'。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求代码文件里每个函数的圈复杂度均不得大于3，现在调整为所有生成的代码文件中的函数圈复杂度一律不得超过2；第二，必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 结合用户指令推断可能是前端网页相关代码，添加常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.cpp', '.c', '.h', '.go', '.php'}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断当前文件是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果在读取文件时抛出异常，算作未能满足要求\n return False\n \n # 如果找到了至少一个代码文件且全部满足要求，返回True，否则返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n\n workspace_dir = Path(workspace_path).resolve()\n \n # 如果工作区目录未生成或不存在，则未产生超限层级\n if not workspace_dir.exists():\n return True\n\n # 递归遍历项目目录中的所有文件与文件夹\n for path in workspace_dir.rglob('*'):\n try:\n rel_path = path.relative_to(workspace_dir)\n except ValueError:\n continue\n \n if path.is_dir():\n # 目录的相对路径 parts 长度 >= 3，代表至少出现了第 4 层目录\n if len(rel_path.parts) >= 3:\n return False\n else:\n # 文件的相对路径 parts 长度 >= 4，代表其存放于至少第 4 层目录中\n if len(rel_path.parts) >= 4:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # RGB颜色猜谜网页游戏常用的代码文件后缀\n code_extensions = {'.html', '.css', '.js'}\n found_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符切割来准确计算代码行和空行的总数\n lines = content.split('\\n')\n \n if len(lines) % 2 != 0:\n return False\n \n except Exception:\n # 若文件由于编码等不可抗因素无法正常读取，视作未通过验证\n return False\n\n return found_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def check_function_names(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 1. 匹配 Python 的函数定义: def func_name(\n py_def_pattern = re.compile(r'(?:^|[^a-zA-Z0-9_])def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n \n # 2. 匹配 JS/TS 显式 function 定义: function func_name(\n js_decl = re.compile(r'(?:^|[^a-zA-Z0-9_$])function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n \n # 3. 匹配 JS/TS 函数赋值: func_name = function(...) 或者 func_name = (...) => 或者 func_name = a =>\n js_expr = re.compile(r'(?:^|[^a-zA-Z0-9_$])([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*[:=]\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)')\n \n # 4. 匹配 JS/TS 简写方法定义: func_name(...) { \n js_method = re.compile(r'(?:^|[^a-zA-Z0-9_$])([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{')\n \n # 常见可能产生误判的关键字（排除控制流、保留字及部分常见对象）\n js_keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', \n 'return', 'with', 'function', 'class', 'new', 'typeof', 'instanceof', \n 'void', 'delete', 'yield', 'await', 'async', 'import', 'export', \n 'super', 'this', 'constructor', 'get', 'set', 'true', 'false', 'null', 'undefined'\n }\n\n # 判断提取出的名字是否仅包含小写字母和下划线\n def is_valid_name(name: str) -> bool:\n return bool(re.match(r'^[a-z_]+$', name))\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤常见的包含代码的文件类型\n if not file.endswith(('.html', '.js', '.ts', '.py', '.jsx', '.tsx', '.vue')):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略不可读或编码异常的文件\n continue\n \n # 1. 检查 Python 的函数定义\n for match in py_def_pattern.finditer(content):\n func_name = match.group(1)\n if not is_valid_name(func_name):\n return False\n \n # 2. 检查 JS/TS 的 function 声明\n for match in js_decl.finditer(content):\n func_name = match.group(1)\n if not is_valid_name(func_name):\n return False\n \n # 3. 检查 JS/TS 函数赋值形式\n for match in js_expr.finditer(content):\n func_name = match.group(1)\n if not is_valid_name(func_name):\n return False\n \n # 4. 检查 JS/TS 简写方法形式\n for match in js_method.finditer(content):\n func_name = match.group(1)\n if func_name not in js_keywords:\n if not is_valid_name(func_name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \"RGB猜谜游戏\" 这一关键词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.isdir(workspace_path):\n return False\n\n # 涵盖前端及常见的代码语言扩展名\n code_extensions = {\n '.py', '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.cs', \n '.go', '.rs', '.php', '.rb', '.sh', '.vue', '.svelte'\n }\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n # 对于不可读取的特殊文件直接跳过\n continue\n \n # 定义字符串字面量的正则：匹配双引号、单引号、反引号包裹的字符串（包括内部的转义符号）\n string_pattern = r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n \n # 依据扩展名定制不同类型的注释正则模式\n if ext in ['.py', '.sh', '.rb']:\n comment_pattern = r'(#.*)'\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\n # HTML及前端框架组件中可能混合 HTML <!--...-->、JS及CSS的多行 /*...*/ 以及 JS单行 //...（运用负向先行断言过滤 http://）\n comment_pattern = r'(<!--[\\s\\S]*?-->|/\\*[\\s\\S]*?\\*/|(?<!:)//.*)'\n else:\n # 对于 CSS/JS 及其他 C-like 语言\n comment_pattern = r'(/\\*[\\s\\S]*?\\*/|(?<!:)//.*)'\n \n # 组合正则引擎，从左至右优先匹配。\n # 这样如果在注释符号外存在字符串字面量，会将字符串消耗掉，从而避免错误把字符串内部内容当做注释。\n pattern = f'{string_pattern}|{comment_pattern}'\n \n comments = []\n for match in re.finditer(pattern, code):\n # group(2) 存在说明匹配到的是 comment_pattern 里的注释部分\n if match.group(2):\n comments.append(match.group(2))\n \n comments_text = \" \".join(comments)\n \n # 统计仅在注释范围内出现的该关键词频次\n count = comments_text.count(\"RGB猜谜游戏\")\n \n # 必须恰好出现 1 次\n if count != 1:\n return False\n \n # 如果工作空间内完全没有任何代码文件，通常意味着未按指令正确生成，视为未通过\n if not has_code_file:\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"所有生成的代码文件中的函数圈复杂度一律不得超过2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"下一步，加入难度切换功能。在顶部标题栏上方增加'简单'和'困难'两个纯文本按钮。1. 简单模式下显示3个颜色块，困难模式下显示6个颜色块；2. 默认选中'困难'模式，被选中的难度按钮文字要加粗并改变颜色以示高亮；3. 每次玩家点击切换难度时，不仅要改变颜色块渲染的数量，还要立刻重置状态并开始一局符合该新难度的新游戏。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\"RGB猜谜游戏\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"和1次\"难度切换\"这两个关键词；第二，每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名（本任务主要是网页前端项目）\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n found_code_file = False\n \n # 遍历当前工作区路径中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取小写的扩展名进行判断\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的第一行\n first_line = f.readline()\n # 检查指定的编码声明是否在第一行中\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果打开或读取失败（如编码错误等），判定为不符合要求\n return False\n \n # 如果找到了代码文件且都没有违反规则，则返回True\n return found_code_file","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 将路径解析为绝对路径\n base_path = Path(workspace_path).resolve()\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉以 . 开头的隐藏文件夹，防止因评测环境自带配置导致误判\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n root_path = Path(root).resolve()\n try:\n rel_path = root_path.relative_to(base_path)\n except ValueError:\n continue\n \n # 计算当前目录距离根目录的深度\n # rel_path.parts 在根目录下为空 ()，长度为 0\n # 在一级子目录下为 ('dir1',)，长度为 1，以此类推\n depth = len(rel_path.parts)\n \n # 同样过滤掉隐藏文件，以免某些系统隐藏文件被误判\n valid_files = [f for f in files if not f.startswith('.')]\n \n # 如果当前文件夹的深度已经大于 3，说明该文件夹本身的层级已经大于 3\n if depth > 3:\n return False\n \n # 如果当前文件夹的深度等于 3，且包含有效的文件或子目录\n # 那么这些文件或子目录的层级将是 4，超出了 3 层的限制\n if depth == 3 and (len(valid_files) > 0 or len(dirs) > 0):\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"def check_even_lines(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.go', '.rs'\n }\n \n has_code_file = False\n \n # 如果工作空间不存在，说明未生成代码\n if not os.path.exists(workspace_path):\n return False\n \n # 遍历工作目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断当前文件是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符分割文件内容，可以统计出包含代码行与空行在内的所有行数\n lines = content.split('\\n')\n \n # 判断行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n \n except Exception:\n # 忽略由于非UTF-8编码异常导致的读取失败（或可根据实际要求处理）\n continue\n\n # 如果没有找到任何有效的代码文件，返回False\n if not has_code_file:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # Python文件的检测（使用AST解析最准确）\n if ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 如果函数名不是由纯小写字母和下划线组成，则返回False\n if not re.match(r'^[a-z_]+$', node.name):\n return False\n except Exception:\n # 如果解析失败，则跳过该文件继续检查\n continue\n \n # 前端/JS相关文件的检测（使用正则表达式）\n elif ext in ['.js', '.html', '.htm', '.ts', '.vue', '.jsx', '.tsx']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 预处理：去除多行注释、单行注释以及HTML注释以防止干扰\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n \n func_names = []\n \n # 匹配 1: function my_func ( 或 async function my_func (\n func_names.extend(re.findall(r'\\b(?:async\\s+)?function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(', code))\n \n # 匹配 2: let/var/const my_func = function( 或 = async function(\n func_names.extend(re.findall(r'\\b(?:let|var|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?function\\b', code))\n \n # 匹配 3: let/var/const my_func = (...) => 或 = async (...) => (箭头函数)\n func_names.extend(re.findall(r'\\b(?:let|var|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', code))\n \n for name in func_names:\n # 检查所有提取出的函数名，必须只包含小写字母和下划线\n if not re.match(r'^[a-z_]+$', name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"和1次\"难度切换\"这两个关键词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对基础网页应用推断常见的代码文件类型\n extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n has_code_file = True\n comments = []\n \n # 1. 提取 HTML 注释并剔除\n html_comments = re.findall(r'<!--(.*?)-->', content, re.DOTALL)\n comments.extend(html_comments)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 2. 提取 C 风格的多行块注释并剔除\n block_comments = re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL)\n comments.extend(block_comments)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n if ext == '.py':\n # 3a. 提取 Python 多行注释并剔除\n py_block1 = re.findall(r'\"\"\"(.*?)\"\"\"', content, re.DOTALL)\n comments.extend(py_block1)\n content = re.sub(r'\"\"\".*?\"\"\"', '', content, flags=re.DOTALL)\n \n py_block2 = re.findall(r\"'''(.*?)'''\", content, re.DOTALL)\n comments.extend(py_block2)\n content = re.sub(r\"'''.*?'''\", '', content, flags=re.DOTALL)\n \n # 4a. 提取 Python 单行注释 (此处为最后一环，不需要剔除)\n py_single = re.findall(r'#(.*?)$', content, re.MULTILINE)\n comments.extend(py_single)\n else:\n # 4b. 提取 JS/TS 等文件的单行注释\n # 使用 (?<!:) 的否定后顾断言来防止匹配 http:// 中的双斜杠\n js_single = re.findall(r'(?<!:)//(.*?)$', content, re.MULTILINE)\n comments.extend(js_single)\n \n # 将该文件中所有提取出的注释文本连接在一起\n all_comments = \"\\n\".join(comments)\n \n # 统计两大关键词在注释中出现的次数\n count_rgb = all_comments.count(\"RGB猜谜游戏\")\n count_diff = all_comments.count(\"难度切换\")\n \n # 判断当前代码文件是否符合恰好出现1次的条件\n if count_rgb != 1 or count_diff != 1:\n return False\n \n # 如果项目中根本没有包含代码文件，则不满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"所有生成的代码文件中的函数圈复杂度一律不得超过2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":4,"instruction":"非常好。现在为了增加可玩性，我们在状态控制栏增加一个'当前得分: 0'的显示。规则如下：1. 每答对一次计分：简单模式得1分，困难模式得3分；2. 注意：如果中途点错了任何一个错误色块，当前得分立刻清零（连胜中断）；3. 每次得分增加时，让得分的数字触发一个轻微的放大缩小（缩放动画）的跳动效果，持续0.5秒左右，以带来更好的奖励感。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\"RGB猜谜游戏\"和1次\"难度切换\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"和1次\"得分统计\"这三个关键词；第二，请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_coding_declaration(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.php', '.rb', '.go', '.rs'\n }\n \n code_file_found = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法正常以 utf-8 读取，则视作不满足要求\n return False\n \n # 确保项目中存在至少一个代码文件且全部通过了检查\n return code_file_found","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n \n try:\n workspace = Path(workspace_path).resolve()\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(str(workspace)):\n # 过滤隐藏目录和编译缓存目录，避免干扰层级判断\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n # 获取当前目录相对于工作区根目录的路径\n rel_path = Path(root).relative_to(workspace)\n \n # rel_path.parts 长度与层级的关系说明：\n # 0 个部件: 工作区根目录自身 (第 1 层)\n # 1 个部件: 根目录下的子目录 (第 2 层)\n # 2 个部件: 子目录的子目录 (第 3 层)\n # 3 个部件: 第 4 层 (例如 src/utils/math/) -> 违反最高 3 层的规则\n if len(rel_path.parts) >= 3:\n return False\n \n return True\n except Exception:\n # 如果路径解析或其他过程出现异常，默认无法判断，返回 False\n return False","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"import os\n\ndef verify_even_line_counts(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名集合，避免检查日志文件等非代码文件\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.php', \n '.swift', '.kt', '.rb', '.go', '.rs'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅验证代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件无法以utf-8格式读取，则跳过\n continue\n \n has_code_file = True\n \n # 统一换行符格式，避免 Windows 与 Unix 系统换行符的干扰\n content = content.replace('\\r\\n', '\\n')\n \n # split('\\n') 能够精确切分并计算包含空行在内的所有行数\n lines = content.split('\\n')\n \n # 检查加起来的总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 匹配 Python 中的函数定义\n py_pattern = re.compile(r'^\\s*def\\s+([a-zA-Z_$][\\w$]*)\\s*\\(', re.MULTILINE)\n # 匹配 JS/TS 等的常规函数声明\n js_pattern1 = re.compile(r'function\\s+([a-zA-Z_$][\\w$]*)\\s*\\(')\n # 匹配 JS/TS 等的箭头函数赋值或匿名函数赋值形式\n js_pattern2 = re.compile(r'(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n # 匹配 JS/TS 等的对象的函数属性形式\n js_pattern3 = re.compile(r'([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n \n valid_extensions = {'.py', '.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 简单移除代码文件中的多行和单行注释，防止正则误匹配注释里的描述内容\n content = re.sub(r'\\'\\'\\'.*?\\'\\'\\'', '', content, flags=re.DOTALL)\n content = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', content, flags=re.DOTALL)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n\n func_names = []\n # 收集各种语言/格式可能提取到的函数名称\n func_names.extend(py_pattern.findall(content))\n func_names.extend(js_pattern1.findall(content))\n func_names.extend(js_pattern2.findall(content))\n func_names.extend(js_pattern3.findall(content))\n \n # 遍历判断所有提取出的函数名\n for name in func_names:\n # 验证提取出的名字是否只包含小写字母和下划线\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"和1次\"得分统计\"这三个关键词","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.xhtml', \n '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.sass', '.less',\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.go', '.rs'\n }\n \n # 预编译提取注释的正则表达式\n # [\\s\\S] 用于安全地跨行匹配多行注释\n html_comment_re = re.compile(r'<!--[\\s\\S]*?-->')\n js_css_c_comment_re = re.compile(r'//.*|/\\*[\\s\\S]*?\\*/')\n py_comment_re = re.compile(r'#.*|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'')\n \n # 提取HTML中内嵌的script和style标签内容\n script_style_re = re.compile(r'<(script|style)[^>]*>([\\s\\S]*?)</\\1>', re.IGNORECASE)\n \n kws = [\"RGB猜谜游戏\", \"难度切换\", \"得分统计\"]\n checked_files = 0\n \n if not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n lower_file = file.lower()\n ext = os.path.splitext(lower_file)[1]\n \n # 过滤非代码文件，以及因前序轮次\"备份已有文件\"要求可能产生的备份代码文件\n if ext not in code_extensions:\n continue\n if 'backup' in lower_file or 'bak' in lower_file:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略因编码或其他问题无法读取的文件\n continue\n \n checked_files += 1\n comments = []\n \n # 针对不同扩展名进行不同的注释提取规则\n if ext in {'.html', '.htm', '.xhtml'}:\n # 1. 提取纯HTML注释\n comments.extend(html_comment_re.findall(content))\n # 2. 提取内嵌的 script 和 style 中的 js/css 风格注释\n for match in script_style_re.finditer(content):\n block_content = match.group(2)\n comments.extend(js_css_c_comment_re.findall(block_content))\n elif ext == '.py':\n comments.extend(py_comment_re.findall(content))\n else:\n comments.extend(js_css_c_comment_re.findall(content))\n \n # 将文件中所有的注释文本合并进行关键词分析\n comments_text = '\\n'.join(comments)\n \n # 检查每个生成的代码文件中，这三个词是否恰好各出现一次\n for kw in kws:\n if comments_text.count(kw) != 1:\n return False\n\n # 若至少存在一个被检查的代码文件并且都满足条件，返回True\n return checked_files > 0","tags":["Content","Keyword"]},{"constraint":"所有生成的代码文件中的函数圈复杂度一律不得超过2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能存在的代码文件扩展名集合\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n # 筛选出属于代码的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 提取行首的所有空白字符（也就是缩进部分）\n indent_chars = \"\"\n for char in line:\n if char in (' ', '\\t'):\n indent_chars += char\n else:\n break\n \n # 检查行首的缩进是否包含制表符\n if '\\t' in indent_chars:\n return False\n except Exception:\n # 若文件由于编码等问题无法读取，忽略该文件继续下一个\n continue\n \n # 如果遍历检查完所有的代码文件，均未发现任何制表符缩进，则返回 True\n return True","tags":["Style","Layout"]}]},{"round_id":5,"instruction":"接下来，我们引入一个更有挑战的'极限模式'。在难度选择区加上第三个按钮'极限'。1. 极限模式下要生成9个颜色块，且每次选对得5分；2. 这个模式带有时间限制，在颜色值显示的下方增加一个水平进度条代表剩余时间（每局5秒）；3. 游戏一旦开始，进度条就随时间匀速减少，如果玩家在进度条耗尽前没猜对或者中途点错了色块，直接触发结束，在页面中央弹出一个自定义的Game Over弹窗，展示最终连续得分，并提供一个'重新挑战'的按钮（点击后重置分数并开始新游戏）。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"和1次\"得分统计\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"和1次\"极限模式\"这四个关键词；第二，后续所有轮次对代码进行任何更改前，必须先在task_plan.md中说明修改计划，再执行代码修改操作。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断，RGB颜色猜谜游戏网页包含前端的三大件\n code_extensions = {'.html', '.css', '.js', '.py'}\n has_code_files = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline().strip()\n # 判断第一行是否包含了指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果由于任何原因(如编码不正确,文件无法读取)导致读取失败，视为未满足要求\n return False\n \n # 如果项目中没有任何代码文件，则说明没有正确生成项目\n if not has_code_files:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 如果工作区不存在，说明没有生成违规的深层目录\n if not os.path.exists(workspace_path):\n return True\n\n workspace = Path(workspace_path).resolve()\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录（例如 .git）和缓存目录（例如 __pycache__），避免因环境自带文件引起误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n # 检查当前目录下所有的文件和允许被遍历的目录\n for name in files + dirs:\n item_path = Path(root).resolve() / name\n try:\n # 获取相对于根目录的相对路径\n rel_path = item_path.relative_to(workspace)\n \n # 相对路径的 parts 长度即为当前的目录层级\n # 例如: 'dir1/dir2/file.py' 包含三个部分，说明是在第 3 层\n if len(rel_path.parts) > 3:\n return False\n except ValueError:\n # 若 item_path 无法计算出基于 workspace 的相对路径，则跳过\n continue\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义目标代码文件的后缀名\n code_exts = {'.html', '.css', '.js', '.py'}\n has_checked_any = False\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_exts:\n file_path = os.path.join(root, file)\n has_checked_any = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n # 验证每个代码文件行数是否是偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 若文件读取失败（例如编码不支持等），视作无法判断通过，返回 False\n return False\n \n # 如果项目中没有任何代码文件，则无法判断要求达成，返回False；全部通过返回True\n return has_checked_any","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 匹配各类函数定义的正则表达式\n # 1. 传统 JS function 或 async function 定义\n p1 = re.compile(r'\\b(?:async\\s+)?function\\s+([a-zA-Z0-9_$]+)\\s*\\(')\n # 2. JS 箭头函数或匿名函数赋值形式 (const/let/var)\n p2 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_$]+)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z0-9_$]+\\s*=>)')\n # 3. Python 的函数定义\n p3 = re.compile(r'\\b(?:async\\s+)?def\\s+([a-zA-Z0-9_]+)\\s*\\(')\n \n def remove_comments_and_strings(content, ext):\n if ext == '.py':\n content = re.sub(r'#.*', '', content)\n content = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '', content)\n content = re.sub(r\"'''[\\s\\S]*?'''\", '', content)\n content = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '', content)\n content = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", '', content)\n else:\n # 去除 HTML 注释\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n # 去除 JS 多行注释\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n # 去除 JS 单行注释\n content = re.sub(r'//.*', '', content)\n # 去除 JS 模板字符串、双引号字符串和单引号字符串\n content = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '', content)\n content = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '', content)\n content = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", '', content)\n return content\n\n func_names = []\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in ['.html', '.htm', '.js', '.py']:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n clean_content = remove_comments_and_strings(content, ext)\n \n if ext in ['.html', '.htm', '.js']:\n func_names.extend(p1.findall(clean_content))\n func_names.extend(p2.findall(clean_content))\n if ext == '.py':\n func_names.extend(p3.findall(clean_content))\n \n except Exception:\n continue\n\n if not has_code_file:\n return False\n \n # 判断是否所有函数名只包含小写字母和下划线\n for name in func_names:\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"和1次\"极限模式\"这四个关键词","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 如果工作目录不存在，返回 False\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码文件后缀\n code_extensions = {'.py', '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\n has_code_file = False\n \n # 必须恰好出现 1 次的关键词列表\n keywords = [\"RGB猜谜游戏\", \"难度切换\", \"得分统计\", \"极限模式\"]\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码文件，忽略文档或日志类文件\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件无法读取，跳过该文件\n continue\n \n has_code_file = True\n comments = []\n \n # 针对不同的代码类型，使用正则提取相应语法的注释\n if ext == '.py':\n comments.extend(re.findall(r'#.*', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''|\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"\", content))\n elif ext in ['.html', '.htm', '.vue']:\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # HTML 中可能会有内嵌的 <script> 或 <style> 代码，同时匹配 JS/CSS 风格注释\n comments.extend(re.findall(r'(?<!:)//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n else:\n # 针对 .js, .css, .ts 等前端代码文件\n comments.extend(re.findall(r'(?<!:)//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n # 将该文件中所有的注释合并成一段长文本\n all_comments_text = '\\n'.join(comments)\n \n # 逐一检查每个关键词的出现次数是否恰好为 1\n for kw in keywords:\n if all_comments_text.count(kw) != 1:\n return False\n \n # 所有代码文件校验通过且至少存在一个有效的代码文件\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"所有生成的代码文件中的函数圈复杂度一律不得超过2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断给定文本中是否存在使用制表符的缩进\n def has_tab_indent(text: str) -> bool:\n for line in text.split('\\n'):\n # 匹配行首的全部连续空格或制表符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n # 如果行首空白字符中包含 \\t，即违背了严禁制表符缩进的要求\n if '\\t' in match.group(1):\n return True\n return False\n\n # 1. 检查 response 中的代码块\n # 匹配 Markdown 格式的代码块\n code_blocks = re.findall(r'```[ \\t]*\\w*[ \\t]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n if has_tab_indent(block):\n return False\n\n # 2. 检查 workspace_path 中的实际代码文件\n if os.path.exists(workspace_path):\n # 针对当前网页应用项目定义常见的代码文件后缀类型\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.json'}\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if has_tab_indent(content):\n return False\n except Exception:\n # 忽略无法通过 utf-8 解码读取的异常文件\n pass\n \n return True","tags":["Style","Layout"]},{"constraint":"后续所有轮次对代码进行任何更改前，必须先在task_plan.md中说明修改计划，再执行代码修改操作","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":6,"instruction":"为了让'极限模式'真正名副其实，我们需要修改该模式下的错误颜色生成逻辑。在这个模式下，8个错误的颜色块不能是完全漫无目的的随机颜色，而必须是在目标颜色的RGB值基础上进行小幅度偏移（例如每个R/G/B通道的值随机加减15到30之间，注意要控制在0-255范围内），这样生成的颜色肉眼看起来会极其相似，增加玩家分辨的难度。简单和困难模式依然保持之前的完全随机策略。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"和1次\"极限模式\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"、1次\"极限模式\"和1次\"颜色偏移\"这五个关键词；第二，原要求所有生成的代码文件中的函数圈复杂度一律不得超过2，现调整为所有生成的代码文件中的函数圈复杂度一律不得超过3。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义网页项目及常规项目中常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.vue'}\n code_files_found = False\n\n # 遍历当前工作空间的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅针对代码文件进行检查，忽略.md、.log等非代码文件\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 仅读取文件的第一行\n first_line = f.readline()\n \n # 判断第一行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果因编码不支持或无法读取导致异常，判定为不符合要求\n return False\n \n # 如果项目中至少有一个代码文件并且它们全满足第一行的声明要求，则返回 True\n return code_files_found","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_directory_depth(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n \n try:\n workspace = Path(workspace_path).resolve()\n # 如果工作区不存在或不是一个目录，无法判断，返回False\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n # 遍历工作区下的所有文件和子目录\n for path in workspace.rglob('*'):\n # 获取相对于工作区根目录的各层级部分\n rel_parts = path.relative_to(workspace).parts\n \n # 忽略由环境或工具自动生成的常见目录，避免误判\n ignore_dirs = {'.git', '__pycache__', 'node_modules', '.idea', '.vscode'}\n if any(part in ignore_dirs for part in rel_parts):\n continue\n \n # 判断目录层级是否超过 3 层\n # 例如: dir1/dir2/file.txt 包含3部分，属于第3层，合法\n # dir1/dir2/dir3/file.txt 包含4部分，属于第4层，超出限制\n if len(rel_parts) > 3:\n return False\n \n return True\n except Exception:\n return False","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 判断总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果由于编码等问题无法读取，视为验证失败\n return False\n\n # 若至少存在一个代码文件且全部代码文件的行数都为偶数，则满足条件\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def check_function_names(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n if not os.path.exists(workspace_path):\n return False\n\n # 验证函数名是否仅包含小写字母和下划线（数字亦不被允许）\n def is_valid_name(name: str) -> bool:\n return bool(re.match(r'^[a-z_]+$', name))\n\n # 前端开发中常见的提取函数定义的正则表达式\n js_patterns = [\n # 1. 常规的 function 关键字声明\n r'\\bfunction\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(',\n # 2. ES6 箭头函数赋值声明（覆盖带括号参数及无括号单参数形式，支持 async）\n r'\\b(?:const|let|var)\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_][a-zA-Z0-9_]*)\\s*=>',\n # 3. 赋值型匿名 function 声明（支持 async）\n r'\\b(?:const|let|var)\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*=\\s*(?:async\\s+)?function\\s*\\('\n ]\n # Python 常见的提取函数定义的正则表达式（作为 ast 的降级方案及其他纯文本文件的探测）\n py_pattern = r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\('\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 对 Python 源代码使用 AST 进行高精度检查\n if file.endswith('.py'):\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not is_valid_name(node.name):\n return False\n except SyntaxError:\n # 如果由于语法不完整等原因导致 AST 解析失败，回退使用正则匹配\n names = re.findall(py_pattern, content)\n for name in names:\n if not is_valid_name(name):\n return False\n \n # 对前端以及常见网页文件使用正则表达式检查函数名\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\n for pattern in js_patterns:\n names = re.findall(pattern, content)\n for name in names:\n if not is_valid_name(name):\n return False\n \n # 对其他可能隐含代码块的文件启发式地扫描函数声明\n else:\n for pattern in js_patterns + [py_pattern]:\n names = re.findall(pattern, content)\n for name in names:\n if not is_valid_name(name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"、1次\"极限模式\"和1次\"颜色偏移\"这五个关键词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 必须恰好出现一次的五个关键词\n keywords = [\"RGB猜谜游戏\", \"难度切换\", \"得分统计\", \"极限模式\", \"颜色偏移\"]\n \n # 常见代码文件扩展名集合（过滤掉 .log, .md 等日志或文档文件）\n code_exts = {\".html\", \".css\", \".js\", \".py\", \".jsx\", \".ts\", \".tsx\"}\n \n # 构建提取注释的正则表达式\n # 包含了：HTML注释、C/JS风格多行、C/JS风格单行、Python单行、Python多行注释\n comment_pattern = re.compile(\n r'<!--[\\s\\S]*?-->'\n r'|/\\*[\\s\\S]*?\\*/'\n r'|//.*'\n r'|#.*'\n r'|\"\"\"[\\s\\S]*?\"\"\"'\n r\"|'''[\\s\\S]*?'''\"\n )\n \n found_code_file = False\n \n # 遍历项目所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取出文件中所有的注释内容片段并拼接\n comments = comment_pattern.findall(content)\n combined_comments = \"\\n\".join(comments)\n \n # 检查各个关键词在当前文件注释中出现的次数\n for kw in keywords:\n if combined_comments.count(kw) != 1:\n return False\n \n # 确保项目中至少找到了一个可供检查的代码文件\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"所有生成的代码文件中的函数圈复杂度一律不得超过3","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 根据用户指令推断，主要为前端网页项目，这里涵盖最可能出现的前端扩展名及其它常见代码扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.cs', '.php', '.rb', '.go', '.rs'\n }\n \n # 1. 检查工作区代码项目中的文件缩进\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 计算行首所有连续的空白字符（即该行的缩进部分）\n indentation = line[:len(line) - len(line.lstrip())]\n if '\\t' in indentation:\n return False\n except Exception:\n # 忽略非UTF-8编码或其他读取异常的文件\n pass\n \n # 2. 检查助手最后一轮回复的 Markdown 代码块中的缩进\n # 正则提取 ```语言 ... ``` 格式的代码块内容\n code_blocks = re.findall(r'```[A-Za-z0-9-]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n lines = block.split('\\n')\n for line in lines:\n indentation = line[:len(line) - len(line.lstrip())]\n if '\\t' in indentation:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"后续所有轮次对代码进行任何更改前，必须先在task_plan.md中说明修改计划，再执行代码修改操作","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":7,"instruction":"最后一步，我们要完善页面的移动端响应式布局和夜间模式体验。1. 在页面右上角加上一个暗黑模式的切换开关，点击后页面整体背景在深灰色和纯白色间平滑切换，相关的文本颜色也要做对应反转，以保证对比度；2. 针对屏幕宽度小于600px的设备进行适配，如果颜色块是6个或9个，将底部的布局自动调整为网格结构（如每行显示3个），保持色块比例为正方形，并适当缩放顶部的字体大小，确保一屏内能完整看到标题、进度条和所有颜色块而不需要上下滑动。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\"RGB猜谜游戏\"、\"难度切换\"、\"得分统计\"、\"极限模式\"和\"颜色偏移\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"、1次\"极限模式\"、1次\"颜色偏移\"、1次\"响应式布局\"和1次\"暗黑模式\"这七个关键词；第二，原要求所有轮次对代码进行任何更改前必须先在task_plan.md中说明修改计划，现调整为每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案，再执行代码修改操作。","instruction_following_checklist":[{"constraint":"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-","validation_code":"def verify_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义目标代码文件的后缀名，排除.md, .json, .log等非代码文件\n code_extensions = {'.html', '.css', '.js', '.py'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法读取或者抛出异常，视作未满足要求\n return False\n\n # 只有当确实存在且检查了至少一个代码文件，并且全部符合要求时才返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"要求你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你提供的代码项目，其目录层级最高不能超过 3 层","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import pathlib\n\n # 将 workspace_path 转换为绝对路径对象，作为基准目录\n base = pathlib.Path(workspace_path).resolve()\n \n # 遍历 workspace_path 下的所有文件和目录\n for item in base.rglob('*'):\n # 获取当前项相对于 base 的相对路径\n rel = item.relative_to(base)\n \n # rel.parts 返回组成路径的各个层级的元组\n # 长度表示它在项目中的深度（比如 root/a/b.txt 长度为 2）\n # 深度如果超过 3，则不满足最高层级不超过 3 层的要求\n if len(rel.parts) > 3:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 涵盖 Web 项目常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.json'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 使用 '\\n' 切分可以准确统计包括空行在内的所有行数\n lines = content.split('\\n')\n \n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果遇到无法以 utf-8 读取的特殊文件则忽略或视作异常不作直接判定\n continue\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 用于提取 JavaScript / TypeScript 或 HTML 内嵌 JS 的函数名称的正则表达式\n js_patterns = [\n # 匹配普通函数声明: function funcName(...)\n re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('),\n # 匹配变量赋值型声明: const/let/var funcName = function(...) / (...) => / arg =>\n re.compile(r'(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function|\\([^()]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'),\n # 匹配对象属性赋值声明: obj.funcName = function(...) / (...) => / arg =>\n re.compile(r'\\.\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function|\\([^()]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'),\n # 匹配对象字面量键声明: funcName: function(...) / (...) => / arg =>\n re.compile(r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?(?:function|\\([^()]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'),\n # 匹配类或对象方法简写形式: funcName(...) {\n re.compile(r'(?:^|[{};,])\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^()]*\\)\\s*\\{')\n ]\n\n # 用于提取 Python 的函数名称的正则表达式\n py_patterns = [\n re.compile(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n ]\n\n # 合法的函数名模式：只包含小写字母和下划线\n valid_name_pattern = re.compile(r'^[a-z_]+$')\n\n # 需要排除的 JS 关键字，防止类方法简写的正则对控制流语法造成误判\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'with', 'constructor',\n 'try', 'else', 'finally', 'do', 'class', 'await', 'async'\n }\n\n # 遍历项目目录，检查所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.py', '.html', '.htm', '.ts', '.jsx', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n\n # 清除注释，防止注释中的陈旧代码及非规范文本引起误报\n if file.endswith('.py'):\n content = re.sub(r'#.*', '', content)\n content = re.sub(r'\\'\\'\\'.*?\\'\\'\\'', '', content, flags=re.DOTALL)\n content = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', content, flags=re.DOTALL)\n else:\n if file.endswith(('.html', '.htm')):\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n\n func_names = []\n \n # 依据文件类型进行函数名提取\n if file.endswith('.py'):\n for p in py_patterns:\n func_names.extend(p.findall(content))\n else:\n for p in js_patterns:\n func_names.extend(p.findall(content))\n\n # 对该文件中所有检测到的函数名进行合法性检验\n for name in func_names:\n if name in keywords:\n continue\n if not valid_name_pattern.match(name):\n return False\n except Exception:\n # 容错处理：忽略不可读的文件\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"每个由你生成的代码文件，其注释中都必须恰好出现1次\"RGB猜谜游戏\"、1次\"难度切换\"、1次\"得分统计\"、1次\"极限模式\"、1次\"颜色偏移\"、1次\"响应式布局\"和1次\"暗黑模式\"这七个关键词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 待校验的7个关键词\n keywords = [\n \"RGB猜谜游戏\", \n \"难度切换\", \n \"得分统计\", \n \"极限模式\", \n \"颜色偏移\", \n \"响应式布局\", \n \"暗黑模式\"\n ]\n \n # 基于用户指令推测的可能代码文件类型\n valid_exts = {'.html', '.css', '.js', '.py'}\n \n def get_comments(content: str, ext: str) -> str:\n comments = []\n if ext == '.html':\n # 提取 HTML 原生多行注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 提取内联 script 和 style 代码块\n blocks = re.findall(r'<script.*?>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n blocks.extend(re.findall(r'<style.*?>(.*?)</style>', content, re.DOTALL | re.IGNORECASE))\n for block in blocks:\n comments.extend(re.findall(r'/\\*(.*?)\\*/', block, re.DOTALL))\n # 提取 // 注释，通过 (?<!:) 避免匹配 http:// 中的斜杠\n comments.extend(re.findall(r'(?<!:)\\/\\/(.*)', block))\n elif ext in {'.js', '.css'}:\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'(?<!:)\\/\\/(.*)', content))\n elif ext == '.py':\n comments.extend(re.findall(r'#(.*)', content))\n comments.extend(re.findall(r'\"\"\"(.*?)\"\"\"', content, re.DOTALL))\n comments.extend(re.findall(r\"'''(.*?)'''\", content, re.DOTALL))\n \n return \" \\n \".join(comments)\n\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果遇到无法读取的文件可以直接跳过\n continue\n \n # 提取当前文件全部注释\n comments_text = get_comments(content, ext)\n \n # 校验每一个关键字是否在注释里恰好出现 1 次\n for kw in keywords:\n if comments_text.count(kw) != 1:\n return False\n \n # 只要存在被检查的代码文件，且全都通过了上述校验条件，则返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"所有生成的代码文件中的函数圈复杂度一律不得超过3","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 定义需要检查的常见代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.py', '.json', '.ts', '.jsx', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.lower().endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 提取行首的前导空白字符\n leading_whitespace = \"\"\n for char in line:\n if char == ' ' or char == '\\t':\n leading_whitespace += char\n else:\n break\n \n # 如果前导空白中包含制表符，说明存在制表符缩进\n if '\\t' in leading_whitespace:\n return False\n except Exception:\n # 若文件由于编码等原因无法正常读取，则跳过\n continue\n\n return True","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案，再执行代码修改操作","validation_code":"","tags":["Environment","Logging"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我实现一个RGB颜色猜谜游戏的基础网页框架。需要包含：1. 顶部的标题区域，标题下方留有一个大字号的文本区用来显示'RGB(XXX, XXX, XXX)'占位；2. 中间是一个状态控制栏，左侧显示'提示信息'（比如'猜猜看'），右侧放一个'新游戏'按钮；3. 底部是一个颜色块的容器，先在里面放3个正方形的div作为颜色块占位符，默认给点灰色背景即可。整个页面请居中布局，采用现代、简洁的卡片式设计。这一轮不需要写游戏的核心判断逻辑，把UI骨架和基本样式搭好即可。另外请你在后续所有轮次都严格遵循以下要求：一是每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \\\"RGB猜谜游戏\\\" 这一关键词；二是保持你生成的代码文件里每个函数的圈复杂度均在 2 以内。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_extensions = {'.py', '.html', '.css', '.js'}\\n has_code_files = False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in target_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n return False\\n\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，视作未生成有效的项目结构\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 获取当前目录相对于根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 判断并计算目录层级深度\\n if rel_path == '.':\\n depth = 1\\n else:\\n # 通过路径分隔符拆分相对路径\\n parts = rel_path.split(os.sep)\\n depth = len(parts) + 1\\n \\n # 如果目录层级深度超过 3 层，则不满足要求\\n if depth > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 依据前端及Python常见后缀识别代码文件\\n code_extensions = {'.py', '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue'}\\n file_count = 0\\n \\n # 遍历工作区目录，查找所有生成的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符 '\\\\n' 分割可以准确还原编辑器中展示的行数（精确包含代码行和所有空行）\\n lines = content.split('\\\\n')\\n \\n # 判断当前代码文件的总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果文件无法按预期读取（如编码错误或权限问题），为了容错继续检查下一个\\n pass\\n \\n # 如果没有找到任何代码文件，视作未满足要求\\n if file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 匹配 JavaScript 常见的函数定义格式\\n # 1. 匹配常规函数: function func_name (\\n js_func_pattern1 = re.compile(r\\\"(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(\\\")\\n # 2. 匹配函数表达式/箭头函数赋值: const func_name = function( / () =>\\n js_func_pattern2 = re.compile(r\\\"(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)\\\")\\n # 匹配 Python 常规函数（降级使用）\\n py_func_pattern = re.compile(r\\\"^\\\\s*(?:async\\\\s+)?def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(\\\", re.MULTILINE)\\n\\n # 遍历项目目录下的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤相关的代码文件后缀\\n if ext not in ['.py', '.js', '.html', '.ts', '.vue', '.jsx', '.tsx']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 清除HTML以及JS多行注释，避免误提取注释中的内容\\n content_cleaned = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content_cleaned = re.sub(r'/\\\\*.*?\\\\*/', '', content_cleaned, flags=re.DOTALL)\\n \\n func_names = []\\n \\n # 提取 Python 文件中的函数名\\n if ext == '.py':\\n try:\\n tree = ast.parse(content_cleaned)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n func_names.append(node.name)\\n except Exception:\\n # 如果 AST 解析失败，降级使用正则匹配\\n func_names.extend(py_func_pattern.findall(content_cleaned))\\n \\n # 提取 JS / HTML 包含的前端函数名\\n else:\\n func_names.extend(js_func_pattern1.findall(content_cleaned))\\n func_names.extend(js_func_pattern2.findall(content_cleaned))\\n \\n # 匹配可能的类方法声明或简写方法，例如: my_method() {\\n lines = content_cleaned.split('\\\\n')\\n for line in lines:\\n m = re.match(r\\\"^\\\\s*(?:async\\\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{\\\", line)\\n if m:\\n name = m.group(1)\\n # 排除掉 JavaScript 里的流程控制关键字\\n if name not in ['if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else']:\\n func_names.append(name)\\n \\n # 验证所有提取到的函数名\\n for name in func_names:\\n # 判断函数名是否【只】包含小写字母与下划线\\n if not re.match(r\\\"^[a-z_]+$\\\", name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \\\"RGB猜谜游戏\\\" 这一关键词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 检查工作目录是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n code_files = []\\n \\n # 查找指定扩展名的所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n # 若未找到任何代码文件，则无法满足每个代码文件皆有此注释的要求\\n if not code_files:\\n return False\\n \\n keyword = \\\"RGB猜谜游戏\\\"\\n \\n # 逐一验证每个代码文件\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n ext = os.path.splitext(file_path)[1].lower()\\n comments = []\\n \\n # 按照文件类型匹配所有的注释\\n if ext in {'.html', '.htm', '.vue'}:\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'(?<![:\\\"\\\\'`])//(.*?)$', content, re.MULTILINE))\\n elif ext in {'.css', '.js', '.ts', '.jsx', '.tsx'}:\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'(?<![:\\\"\\\\'`])//(.*?)$', content, re.MULTILINE))\\n elif ext == '.py':\\n comments.extend(re.findall(r'#(.*?)$', content, re.MULTILINE))\\n comments.extend(re.findall(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', content, re.DOTALL))\\n comments.extend(re.findall(r\\\"'''(.*?)'''\\\", content, re.DOTALL))\\n \\n # 统计在当前文件所有注释内容中目标关键词累计出现次数\\n total_count = sum(comment.count(keyword) for comment in comments)\\n \\n # 要求恰好出现 1 次\\n if total_count != 1:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 2 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来把核心的游戏逻辑加上。1. 页面加载或点击'新游戏'按钮时，随机生成一个目标RGB颜色值，并更新显示在顶部的文本区；2. 给底部的3个颜色块分别赋上完全随机的RGB背景色，但其中必须有一个颜色块正好是目标颜色；3. 点击颜色块时进行判断逻辑：如果点对了，状态控制栏的提示文字变成'回答正确！'；如果点错了，提示文字变成'再试一次！'。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，你输出的代码文件里每个函数的圈复杂度均不得大于3；第二，作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令，“RGB颜色猜谜游戏的基础网页框架”主要是前端项目\\n code_extensions = {'.html', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n return False\\n \\n # 如果没有找到任何代码文件，则意味着没有按预期生成代码，返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，则代表目前尚未生成项目，可视为不违反限制\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 计算当前所在目录相对于工作区根目录的相对路径\\n rel_root = os.path.relpath(root, workspace_path)\\n \\n # 根目录自身的相对路径通常解析为 '.'，深度为 0\\n if rel_root == '.':\\n depth = 0\\n else:\\n # 兼容不同操作系统的路径分隔符，将反斜杠替换为斜杠，再切分出路径组件列表\\n depth = len([p for p in rel_root.replace('\\\\\\\\', '/').split('/') if p])\\n \\n # 若当前目录本身的层级大于 3，则判定不符合要求\\n if depth > 3:\\n return False\\n \\n # 当前目录内的文件，层级等于当前目录层级 + 1\\n for f in files:\\n if depth + 1 > 3:\\n return False\\n \\n # 当前目录内的子文件夹，层级等于当前目录层级 + 1\\n for d in dirs:\\n if depth + 1 > 3:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\\n \\n # 检查工作空间目录是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行行数检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计行数，包含代码行与空行\\n if content == \\\"\\\":\\n line_count = 0\\n else:\\n line_count = len(content.split('\\\\n'))\\n \\n # 如果行数不是偶数，则不符合要求\\n if line_count % 2 != 0:\\n return False\\n except Exception:\\n # 忽略读取异常的文件（例如非文本文件等特殊情况）\\n pass\\n \\n # 所有代码文件的行数均满足偶数要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 针对 JavaScript/TypeScript 及内嵌在 HTML 中的 JS 定义一系列正则来提取函数名\\n # 1. 传统函数声明: function funcName (\\n js_func_pattern1 = re.compile(r'function\\\\s+([A-Za-z0-9_]+)\\\\s*\\\\(')\\n # 2. 变量赋值形式: const funcName = function (\\n js_func_pattern2 = re.compile(r'(?:const|let|var)\\\\s+([A-Za-z0-9_]+)\\\\s*=\\\\s*function\\\\s*\\\\(')\\n # 3. 变量赋值箭头函数: const funcName = (...) => { 或者 const funcName = arg => {\\n js_func_pattern3 = re.compile(r'(?:const|let|var)\\\\s+([A-Za-z0-9_]+)\\\\s*=\\\\s*(?:\\\\([^)]*\\\\)|[A-Za-z0-9_]+)\\\\s*=>')\\n # 4. 类或对象的简写方法: methodName(...) { (多行模式下以行首或空白符开头，以防匹配到流程控制语句)\\n js_func_pattern4 = re.compile(r'^\\\\s*([A-Za-z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', re.MULTILINE)\\n # 5. 对象属性赋值函数: key: function (\\n js_func_pattern5 = re.compile(r'([A-Za-z0-9_]+)\\\\s*:\\\\s*function\\\\s*\\\\(')\\n # 6. 对象属性赋值箭头函数: key: (...) => {\\n js_func_pattern6 = re.compile(r'([A-Za-z0-9_]+)\\\\s*:\\\\s*(?:\\\\([^)]*\\\\)|[A-Za-z0-9_]+)\\\\s*=>')\\n # 7. 对象属性绑定函数: obj.prop = function (\\n js_func_pattern7 = re.compile(r'\\\\.\\\\s*([A-Za-z0-9_]+)\\\\s*=\\\\s*function\\\\s*\\\\(')\\n # 8. 对象属性绑定箭头函数: obj.prop = (...) => {\\n js_func_pattern8 = re.compile(r'\\\\.\\\\s*([A-Za-z0-9_]+)\\\\s*=\\\\s*(?:\\\\([^)]*\\\\)|[A-Za-z0-9_]+)\\\\s*=>')\\n\\n # 需要过滤掉的 JS 关键字，以防被 js_func_pattern4 等误匹配\\n js_keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', \\n 'return', 'else', 'do', 'try', 'finally', 'with', \\n 'constructor', 'get', 'set'\\n }\\n\\n # 限定检查名称是否严格只包含小写字母与下划线\\n valid_name_pattern = re.compile(r'^[a-z_]+$')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # Python文件的解析使用标准库ast\\n if ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n tree = ast.parse(f.read())\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not valid_name_pattern.match(node.name):\\n return False\\n except Exception:\\n pass\\n \\n # 网页相关前端文件的解析\\n elif ext in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n funcs = []\\n funcs.extend(js_func_pattern1.findall(content))\\n funcs.extend(js_func_pattern2.findall(content))\\n funcs.extend(js_func_pattern3.findall(content))\\n \\n # 模式4单独过滤屏蔽字\\n for match in js_func_pattern4.findall(content):\\n if match not in js_keywords:\\n funcs.append(match)\\n \\n funcs.extend(js_func_pattern5.findall(content))\\n funcs.extend(js_func_pattern6.findall(content))\\n funcs.extend(js_func_pattern7.findall(content))\\n funcs.extend(js_func_pattern8.findall(content))\\n \\n for func in funcs:\\n # 纯数字不可能是正常的函数名，过滤极罕见的误匹配\\n if re.match(r'^\\\\d+$', func):\\n continue\\n \\n # 验证找到的函数名是否只含有小写字母和下划线\\n if not valid_name_pattern.match(func):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \\\"RGB猜谜游戏\\\" 这一关键词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 涵盖当前网页项目和Python可能用到的文件类型\\n code_exts = {'.py', '.js', '.css', '.html', '.htm', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏文件夹（如 .git 等版本控制文件夹）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 根据不同扩展名的代码文件提取注释内容\\n comments = []\\n if ext == '.py':\\n # 匹配 Python 中的单行注释及多行注释（三引号字符串）\\n comments = re.findall(r'#.*|(?s)\\\\'\\\\'\\\\'.*?\\\\'\\\\'\\\\'|(?s)\\\"\\\"\\\".*?\\\"\\\"\\\"', content)\\n elif ext in ['.html', '.htm']:\\n # HTML 文件中可能混写 JS/CSS，因此同时匹配 HTML 注释、斜杠注释和多行注释\\n # (?<!:) 防止将 url 中的 http:// 误判为注释\\n comments = re.findall(r'(?s)<!--.*?-->|(?<!:)//.*|(?s)/\\\\*.*?\\\\*/', content)\\n elif ext in ['.js', '.css', '.ts', '.jsx', '.tsx']:\\n # 匹配 JS、CSS 中的单行和多行注释\\n comments = re.findall(r'(?<!:)//.*|(?s)/\\\\*.*?\\\\*/', content)\\n \\n # 统计从该文件中提取出来的所有注释内，关键词出现的总次数\\n count = sum(comment.count(\\\"RGB猜谜游戏\\\") for comment in comments)\\n \\n # 要求每个代码文件的注释中，必须恰好出现 1 次\\n if count != 1:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里每个函数的圈复杂度均不得大于3\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在的交互太单调了，我们来增强视觉反馈。1. 如果玩家点击了错误的颜色块，让该颜色块通过CSS动画淡出（比如透明度变成0且不可再点击）；2. 如果玩家点击了正确的颜色块，将页面标题区域的背景色、以及所有当前可见颜色块的背景色，全都带有过渡动画地变成这个正确的目标颜色；3. 答对后，把'新游戏'按钮的文字改成'再玩一次？'。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求代码文件里每个函数的圈复杂度均不得大于3，现在调整为所有生成的代码文件中的函数圈复杂度一律不得超过2；第二，必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 结合用户指令推断可能是前端网页相关代码，添加常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.cpp', '.c', '.h', '.go', '.php'}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断当前文件是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果在读取文件时抛出异常，算作未能满足要求\\n return False\\n \\n # 如果找到了至少一个代码文件且全部满足要求，返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n\\n workspace_dir = Path(workspace_path).resolve()\\n \\n # 如果工作区目录未生成或不存在，则未产生超限层级\\n if not workspace_dir.exists():\\n return True\\n\\n # 递归遍历项目目录中的所有文件与文件夹\\n for path in workspace_dir.rglob('*'):\\n try:\\n rel_path = path.relative_to(workspace_dir)\\n except ValueError:\\n continue\\n \\n if path.is_dir():\\n # 目录的相对路径 parts 长度 >= 3，代表至少出现了第 4 层目录\\n if len(rel_path.parts) >= 3:\\n return False\\n else:\\n # 文件的相对路径 parts 长度 >= 4，代表其存放于至少第 4 层目录中\\n if len(rel_path.parts) >= 4:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # RGB颜色猜谜网页游戏常用的代码文件后缀\\n code_extensions = {'.html', '.css', '.js'}\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符切割来准确计算代码行和空行的总数\\n lines = content.split('\\\\n')\\n \\n if len(lines) % 2 != 0:\\n return False\\n \\n except Exception:\\n # 若文件由于编码等不可抗因素无法正常读取，视作未通过验证\\n return False\\n\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def check_function_names(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 1. 匹配 Python 的函数定义: def func_name(\\n py_def_pattern = re.compile(r'(?:^|[^a-zA-Z0-9_])def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n \\n # 2. 匹配 JS/TS 显式 function 定义: function func_name(\\n js_decl = re.compile(r'(?:^|[^a-zA-Z0-9_$])function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n \\n # 3. 匹配 JS/TS 函数赋值: func_name = function(...) 或者 func_name = (...) => 或者 func_name = a =>\\n js_expr = re.compile(r'(?:^|[^a-zA-Z0-9_$])([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*[:=]\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)')\\n \\n # 4. 匹配 JS/TS 简写方法定义: func_name(...) { \\n js_method = re.compile(r'(?:^|[^a-zA-Z0-9_$])([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # 常见可能产生误判的关键字（排除控制流、保留字及部分常见对象）\\n js_keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', \\n 'return', 'with', 'function', 'class', 'new', 'typeof', 'instanceof', \\n 'void', 'delete', 'yield', 'await', 'async', 'import', 'export', \\n 'super', 'this', 'constructor', 'get', 'set', 'true', 'false', 'null', 'undefined'\\n }\\n\\n # 判断提取出的名字是否仅包含小写字母和下划线\\n def is_valid_name(name: str) -> bool:\\n return bool(re.match(r'^[a-z_]+$', name))\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤常见的包含代码的文件类型\\n if not file.endswith(('.html', '.js', '.ts', '.py', '.jsx', '.tsx', '.vue')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略不可读或编码异常的文件\\n continue\\n \\n # 1. 检查 Python 的函数定义\\n for match in py_def_pattern.finditer(content):\\n func_name = match.group(1)\\n if not is_valid_name(func_name):\\n return False\\n \\n # 2. 检查 JS/TS 的 function 声明\\n for match in js_decl.finditer(content):\\n func_name = match.group(1)\\n if not is_valid_name(func_name):\\n return False\\n \\n # 3. 检查 JS/TS 函数赋值形式\\n for match in js_expr.finditer(content):\\n func_name = match.group(1)\\n if not is_valid_name(func_name):\\n return False\\n \\n # 4. 检查 JS/TS 简写方法形式\\n for match in js_method.finditer(content):\\n func_name = match.group(1)\\n if func_name not in js_keywords:\\n if not is_valid_name(func_name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现 1 次 \\\"RGB猜谜游戏\\\" 这一关键词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 涵盖前端及常见的代码语言扩展名\\n code_extensions = {\\n '.py', '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.cs', \\n '.go', '.rs', '.php', '.rb', '.sh', '.vue', '.svelte'\\n }\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n # 对于不可读取的特殊文件直接跳过\\n continue\\n \\n # 定义字符串字面量的正则：匹配双引号、单引号、反引号包裹的字符串（包括内部的转义符号）\\n string_pattern = r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n \\n # 依据扩展名定制不同类型的注释正则模式\\n if ext in ['.py', '.sh', '.rb']:\\n comment_pattern = r'(#.*)'\\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\\n # HTML及前端框架组件中可能混合 HTML <!--...-->、JS及CSS的多行 /*...*/ 以及 JS单行 //...（运用负向先行断言过滤 http://）\\n comment_pattern = r'(<!--[\\\\s\\\\S]*?-->|/\\\\*[\\\\s\\\\S]*?\\\\*/|(?<!:)//.*)'\\n else:\\n # 对于 CSS/JS 及其他 C-like 语言\\n comment_pattern = r'(/\\\\*[\\\\s\\\\S]*?\\\\*/|(?<!:)//.*)'\\n \\n # 组合正则引擎，从左至右优先匹配。\\n # 这样如果在注释符号外存在字符串字面量，会将字符串消耗掉，从而避免错误把字符串内部内容当做注释。\\n pattern = f'{string_pattern}|{comment_pattern}'\\n \\n comments = []\\n for match in re.finditer(pattern, code):\\n # group(2) 存在说明匹配到的是 comment_pattern 里的注释部分\\n if match.group(2):\\n comments.append(match.group(2))\\n \\n comments_text = \\\" \\\".join(comments)\\n \\n # 统计仅在注释范围内出现的该关键词频次\\n count = comments_text.count(\\\"RGB猜谜游戏\\\")\\n \\n # 必须恰好出现 1 次\\n if count != 1:\\n return False\\n \\n # 如果工作空间内完全没有任何代码文件，通常意味着未按指令正确生成，视为未通过\\n if not has_code_file:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的函数圈复杂度一律不得超过2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步，加入难度切换功能。在顶部标题栏上方增加'简单'和'困难'两个纯文本按钮。1. 简单模式下显示3个颜色块，困难模式下显示6个颜色块；2. 默认选中'困难'模式，被选中的难度按钮文字要加粗并改变颜色以示高亮；3. 每次玩家点击切换难度时，不仅要改变颜色块渲染的数量，还要立刻重置状态并开始一局符合该新难度的新游戏。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\\\"RGB猜谜游戏\\\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"和1次\\\"难度切换\\\"这两个关键词；第二，每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名（本任务主要是网页前端项目）\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n found_code_file = False\\n \\n # 遍历当前工作区路径中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取小写的扩展名进行判断\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的第一行\\n first_line = f.readline()\\n # 检查指定的编码声明是否在第一行中\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果打开或读取失败（如编码错误等），判定为不符合要求\\n return False\\n \\n # 如果找到了代码文件且都没有违反规则，则返回True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 将路径解析为绝对路径\\n base_path = Path(workspace_path).resolve()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉以 . 开头的隐藏文件夹，防止因评测环境自带配置导致误判\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n root_path = Path(root).resolve()\\n try:\\n rel_path = root_path.relative_to(base_path)\\n except ValueError:\\n continue\\n \\n # 计算当前目录距离根目录的深度\\n # rel_path.parts 在根目录下为空 ()，长度为 0\\n # 在一级子目录下为 ('dir1',)，长度为 1，以此类推\\n depth = len(rel_path.parts)\\n \\n # 同样过滤掉隐藏文件，以免某些系统隐藏文件被误判\\n valid_files = [f for f in files if not f.startswith('.')]\\n \\n # 如果当前文件夹的深度已经大于 3，说明该文件夹本身的层级已经大于 3\\n if depth > 3:\\n return False\\n \\n # 如果当前文件夹的深度等于 3，且包含有效的文件或子目录\\n # 那么这些文件或子目录的层级将是 4，超出了 3 层的限制\\n if depth == 3 and (len(valid_files) > 0 or len(dirs) > 0):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"def check_even_lines(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.go', '.rs'\\n }\\n \\n has_code_file = False\\n \\n # 如果工作空间不存在，说明未生成代码\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 遍历工作目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断当前文件是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符分割文件内容，可以统计出包含代码行与空行在内的所有行数\\n lines = content.split('\\\\n')\\n \\n # 判断行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n \\n except Exception:\\n # 忽略由于非UTF-8编码异常导致的读取失败（或可根据实际要求处理）\\n continue\\n\\n # 如果没有找到任何有效的代码文件，返回False\\n if not has_code_file:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # Python文件的检测（使用AST解析最准确）\\n if ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 如果函数名不是由纯小写字母和下划线组成，则返回False\\n if not re.match(r'^[a-z_]+$', node.name):\\n return False\\n except Exception:\\n # 如果解析失败，则跳过该文件继续检查\\n continue\\n \\n # 前端/JS相关文件的检测（使用正则表达式）\\n elif ext in ['.js', '.html', '.htm', '.ts', '.vue', '.jsx', '.tsx']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 预处理：去除多行注释、单行注释以及HTML注释以防止干扰\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n \\n func_names = []\\n \\n # 匹配 1: function my_func ( 或 async function my_func (\\n func_names.extend(re.findall(r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(', code))\\n \\n # 匹配 2: let/var/const my_func = function( 或 = async function(\\n func_names.extend(re.findall(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\b', code))\\n \\n # 匹配 3: let/var/const my_func = (...) => 或 = async (...) => (箭头函数)\\n func_names.extend(re.findall(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', code))\\n \\n for name in func_names:\\n # 检查所有提取出的函数名，必须只包含小写字母和下划线\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"和1次\\\"难度切换\\\"这两个关键词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对基础网页应用推断常见的代码文件类型\\n extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n has_code_file = True\\n comments = []\\n \\n # 1. 提取 HTML 注释并剔除\\n html_comments = re.findall(r'<!--(.*?)-->', content, re.DOTALL)\\n comments.extend(html_comments)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 2. 提取 C 风格的多行块注释并剔除\\n block_comments = re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL)\\n comments.extend(block_comments)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n if ext == '.py':\\n # 3a. 提取 Python 多行注释并剔除\\n py_block1 = re.findall(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', content, re.DOTALL)\\n comments.extend(py_block1)\\n content = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', content, flags=re.DOTALL)\\n \\n py_block2 = re.findall(r\\\"'''(.*?)'''\\\", content, re.DOTALL)\\n comments.extend(py_block2)\\n content = re.sub(r\\\"'''.*?'''\\\", '', content, flags=re.DOTALL)\\n \\n # 4a. 提取 Python 单行注释 (此处为最后一环，不需要剔除)\\n py_single = re.findall(r'#(.*?)$', content, re.MULTILINE)\\n comments.extend(py_single)\\n else:\\n # 4b. 提取 JS/TS 等文件的单行注释\\n # 使用 (?<!:) 的否定后顾断言来防止匹配 http:// 中的双斜杠\\n js_single = re.findall(r'(?<!:)//(.*?)$', content, re.MULTILINE)\\n comments.extend(js_single)\\n \\n # 将该文件中所有提取出的注释文本连接在一起\\n all_comments = \\\"\\\\n\\\".join(comments)\\n \\n # 统计两大关键词在注释中出现的次数\\n count_rgb = all_comments.count(\\\"RGB猜谜游戏\\\")\\n count_diff = all_comments.count(\\\"难度切换\\\")\\n \\n # 判断当前代码文件是否符合恰好出现1次的条件\\n if count_rgb != 1 or count_diff != 1:\\n return False\\n \\n # 如果项目中根本没有包含代码文件，则不满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的函数圈复杂度一律不得超过2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"非常好。现在为了增加可玩性，我们在状态控制栏增加一个'当前得分: 0'的显示。规则如下：1. 每答对一次计分：简单模式得1分，困难模式得3分；2. 注意：如果中途点错了任何一个错误色块，当前得分立刻清零（连胜中断）；3. 每次得分增加时，让得分的数字触发一个轻微的放大缩小（缩放动画）的跳动效果，持续0.5秒左右，以带来更好的奖励感。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\\\"RGB猜谜游戏\\\"和1次\\\"难度切换\\\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"和1次\\\"得分统计\\\"这三个关键词；第二，请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_coding_declaration(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.php', '.rb', '.go', '.rs'\\n }\\n \\n code_file_found = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法正常以 utf-8 读取，则视作不满足要求\\n return False\\n \\n # 确保项目中存在至少一个代码文件且全部通过了检查\\n return code_file_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n \\n try:\\n workspace = Path(workspace_path).resolve()\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(str(workspace)):\\n # 过滤隐藏目录和编译缓存目录，避免干扰层级判断\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n # 获取当前目录相对于工作区根目录的路径\\n rel_path = Path(root).relative_to(workspace)\\n \\n # rel_path.parts 长度与层级的关系说明：\\n # 0 个部件: 工作区根目录自身 (第 1 层)\\n # 1 个部件: 根目录下的子目录 (第 2 层)\\n # 2 个部件: 子目录的子目录 (第 3 层)\\n # 3 个部件: 第 4 层 (例如 src/utils/math/) -> 违反最高 3 层的规则\\n if len(rel_path.parts) >= 3:\\n return False\\n \\n return True\\n except Exception:\\n # 如果路径解析或其他过程出现异常，默认无法判断，返回 False\\n return False\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"import os\\n\\ndef verify_even_line_counts(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名集合，避免检查日志文件等非代码文件\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.php', \\n '.swift', '.kt', '.rb', '.go', '.rs'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅验证代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件无法以utf-8格式读取，则跳过\\n continue\\n \\n has_code_file = True\\n \\n # 统一换行符格式，避免 Windows 与 Unix 系统换行符的干扰\\n content = content.replace('\\\\r\\\\n', '\\\\n')\\n \\n # split('\\\\n') 能够精确切分并计算包含空行在内的所有行数\\n lines = content.split('\\\\n')\\n \\n # 检查加起来的总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 匹配 Python 中的函数定义\\n py_pattern = re.compile(r'^\\\\s*def\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\(', re.MULTILINE)\\n # 匹配 JS/TS 等的常规函数声明\\n js_pattern1 = re.compile(r'function\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\(')\\n # 匹配 JS/TS 等的箭头函数赋值或匿名函数赋值形式\\n js_pattern2 = re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n # 匹配 JS/TS 等的对象的函数属性形式\\n js_pattern3 = re.compile(r'([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n \\n valid_extensions = {'.py', '.js', '.ts', '.html', '.htm', '.jsx', '.tsx', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 简单移除代码文件中的多行和单行注释，防止正则误匹配注释里的描述内容\\n content = re.sub(r'\\\\'\\\\'\\\\'.*?\\\\'\\\\'\\\\'', '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\".*?\\\\\\\"\\\\\\\"\\\\\\\"', '', content, flags=re.DOTALL)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n\\n func_names = []\\n # 收集各种语言/格式可能提取到的函数名称\\n func_names.extend(py_pattern.findall(content))\\n func_names.extend(js_pattern1.findall(content))\\n func_names.extend(js_pattern2.findall(content))\\n func_names.extend(js_pattern3.findall(content))\\n \\n # 遍历判断所有提取出的函数名\\n for name in func_names:\\n # 验证提取出的名字是否只包含小写字母和下划线\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"和1次\\\"得分统计\\\"这三个关键词\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.xhtml', \\n '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.sass', '.less',\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.go', '.rs'\\n }\\n \\n # 预编译提取注释的正则表达式\\n # [\\\\s\\\\S] 用于安全地跨行匹配多行注释\\n html_comment_re = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n js_css_c_comment_re = re.compile(r'//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n py_comment_re = re.compile(r'#.*|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'')\\n \\n # 提取HTML中内嵌的script和style标签内容\\n script_style_re = re.compile(r'<(script|style)[^>]*>([\\\\s\\\\S]*?)</\\\\1>', re.IGNORECASE)\\n \\n kws = [\\\"RGB猜谜游戏\\\", \\\"难度切换\\\", \\\"得分统计\\\"]\\n checked_files = 0\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n lower_file = file.lower()\\n ext = os.path.splitext(lower_file)[1]\\n \\n # 过滤非代码文件，以及因前序轮次\\\"备份已有文件\\\"要求可能产生的备份代码文件\\n if ext not in code_extensions:\\n continue\\n if 'backup' in lower_file or 'bak' in lower_file:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略因编码或其他问题无法读取的文件\\n continue\\n \\n checked_files += 1\\n comments = []\\n \\n # 针对不同扩展名进行不同的注释提取规则\\n if ext in {'.html', '.htm', '.xhtml'}:\\n # 1. 提取纯HTML注释\\n comments.extend(html_comment_re.findall(content))\\n # 2. 提取内嵌的 script 和 style 中的 js/css 风格注释\\n for match in script_style_re.finditer(content):\\n block_content = match.group(2)\\n comments.extend(js_css_c_comment_re.findall(block_content))\\n elif ext == '.py':\\n comments.extend(py_comment_re.findall(content))\\n else:\\n comments.extend(js_css_c_comment_re.findall(content))\\n \\n # 将文件中所有的注释文本合并进行关键词分析\\n comments_text = '\\\\n'.join(comments)\\n \\n # 检查每个生成的代码文件中，这三个词是否恰好各出现一次\\n for kw in kws:\\n if comments_text.count(kw) != 1:\\n return False\\n\\n # 若至少存在一个被检查的代码文件并且都满足条件，返回True\\n return checked_files > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的函数圈复杂度一律不得超过2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能存在的代码文件扩展名集合\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n # 筛选出属于代码的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 提取行首的所有空白字符（也就是缩进部分）\\n indent_chars = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n indent_chars += char\\n else:\\n break\\n \\n # 检查行首的缩进是否包含制表符\\n if '\\\\t' in indent_chars:\\n return False\\n except Exception:\\n # 若文件由于编码等问题无法读取，忽略该文件继续下一个\\n continue\\n \\n # 如果遍历检查完所有的代码文件，均未发现任何制表符缩进，则返回 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"接下来，我们引入一个更有挑战的'极限模式'。在难度选择区加上第三个按钮'极限'。1. 极限模式下要生成9个颜色块，且每次选对得5分；2. 这个模式带有时间限制，在颜色值显示的下方增加一个水平进度条代表剩余时间（每局5秒）；3. 游戏一旦开始，进度条就随时间匀速减少，如果玩家在进度条耗尽前没猜对或者中途点错了色块，直接触发结束，在页面中央弹出一个自定义的Game Over弹窗，展示最终连续得分，并提供一个'重新挑战'的按钮（点击后重置分数并开始新游戏）。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"和1次\\\"得分统计\\\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"和1次\\\"极限模式\\\"这四个关键词；第二，后续所有轮次对代码进行任何更改前，必须先在task_plan.md中说明修改计划，再执行代码修改操作。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断，RGB颜色猜谜游戏网页包含前端的三大件\\n code_extensions = {'.html', '.css', '.js', '.py'}\\n has_code_files = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline().strip()\\n # 判断第一行是否包含了指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何原因(如编码不正确,文件无法读取)导致读取失败，视为未满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，则说明没有正确生成项目\\n if not has_code_files:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 如果工作区不存在，说明没有生成违规的深层目录\\n if not os.path.exists(workspace_path):\\n return True\\n\\n workspace = Path(workspace_path).resolve()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录（例如 .git）和缓存目录（例如 __pycache__），避免因环境自带文件引起误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n # 检查当前目录下所有的文件和允许被遍历的目录\\n for name in files + dirs:\\n item_path = Path(root).resolve() / name\\n try:\\n # 获取相对于根目录的相对路径\\n rel_path = item_path.relative_to(workspace)\\n \\n # 相对路径的 parts 长度即为当前的目录层级\\n # 例如: 'dir1/dir2/file.py' 包含三个部分，说明是在第 3 层\\n if len(rel_path.parts) > 3:\\n return False\\n except ValueError:\\n # 若 item_path 无法计算出基于 workspace 的相对路径，则跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义目标代码文件的后缀名\\n code_exts = {'.html', '.css', '.js', '.py'}\\n has_checked_any = False\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n has_checked_any = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n # 验证每个代码文件行数是否是偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 若文件读取失败（例如编码不支持等），视作无法判断通过，返回 False\\n return False\\n \\n # 如果项目中没有任何代码文件，则无法判断要求达成，返回False；全部通过返回True\\n return has_checked_any\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 匹配各类函数定义的正则表达式\\n # 1. 传统 JS function 或 async function 定义\\n p1 = re.compile(r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z0-9_$]+)\\\\s*\\\\(')\\n # 2. JS 箭头函数或匿名函数赋值形式 (const/let/var)\\n p2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_$]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z0-9_$]+\\\\s*=>)')\\n # 3. Python 的函数定义\\n p3 = re.compile(r'\\\\b(?:async\\\\s+)?def\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(')\\n \\n def remove_comments_and_strings(content, ext):\\n if ext == '.py':\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '', content)\\n content = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", '', content)\\n content = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '', content)\\n content = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", '', content)\\n else:\\n # 去除 HTML 注释\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n # 去除 JS 多行注释\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n # 去除 JS 单行注释\\n content = re.sub(r'//.*', '', content)\\n # 去除 JS 模板字符串、双引号字符串和单引号字符串\\n content = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '', content)\\n content = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '', content)\\n content = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", '', content)\\n return content\\n\\n func_names = []\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ['.html', '.htm', '.js', '.py']:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n clean_content = remove_comments_and_strings(content, ext)\\n \\n if ext in ['.html', '.htm', '.js']:\\n func_names.extend(p1.findall(clean_content))\\n func_names.extend(p2.findall(clean_content))\\n if ext == '.py':\\n func_names.extend(p3.findall(clean_content))\\n \\n except Exception:\\n continue\\n\\n if not has_code_file:\\n return False\\n \\n # 判断是否所有函数名只包含小写字母和下划线\\n for name in func_names:\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"和1次\\\"极限模式\\\"这四个关键词\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 如果工作目录不存在，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码文件后缀\\n code_extensions = {'.py', '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\\n has_code_file = False\\n \\n # 必须恰好出现 1 次的关键词列表\\n keywords = [\\\"RGB猜谜游戏\\\", \\\"难度切换\\\", \\\"得分统计\\\", \\\"极限模式\\\"]\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码文件，忽略文档或日志类文件\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件无法读取，跳过该文件\\n continue\\n \\n has_code_file = True\\n comments = []\\n \\n # 针对不同的代码类型，使用正则提取相应语法的注释\\n if ext == '.py':\\n comments.extend(re.findall(r'#.*', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''|\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"\\\", content))\\n elif ext in ['.html', '.htm', '.vue']:\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # HTML 中可能会有内嵌的 <script> 或 <style> 代码，同时匹配 JS/CSS 风格注释\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n else:\\n # 针对 .js, .css, .ts 等前端代码文件\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n # 将该文件中所有的注释合并成一段长文本\\n all_comments_text = '\\\\n'.join(comments)\\n \\n # 逐一检查每个关键词的出现次数是否恰好为 1\\n for kw in keywords:\\n if all_comments_text.count(kw) != 1:\\n return False\\n \\n # 所有代码文件校验通过且至少存在一个有效的代码文件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的函数圈复杂度一律不得超过2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断给定文本中是否存在使用制表符的缩进\\n def has_tab_indent(text: str) -> bool:\\n for line in text.split('\\\\n'):\\n # 匹配行首的全部连续空格或制表符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n # 如果行首空白字符中包含 \\\\t，即违背了严禁制表符缩进的要求\\n if '\\\\t' in match.group(1):\\n return True\\n return False\\n\\n # 1. 检查 response 中的代码块\\n # 匹配 Markdown 格式的代码块\\n code_blocks = re.findall(r'```[ \\\\t]*\\\\w*[ \\\\t]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n if has_tab_indent(block):\\n return False\\n\\n # 2. 检查 workspace_path 中的实际代码文件\\n if os.path.exists(workspace_path):\\n # 针对当前网页应用项目定义常见的代码文件后缀类型\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.json'}\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if has_tab_indent(content):\\n return False\\n except Exception:\\n # 忽略无法通过 utf-8 解码读取的异常文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"后续所有轮次对代码进行任何更改前，必须先在task_plan.md中说明修改计划，再执行代码修改操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"为了让'极限模式'真正名副其实，我们需要修改该模式下的错误颜色生成逻辑。在这个模式下，8个错误的颜色块不能是完全漫无目的的随机颜色，而必须是在目标颜色的RGB值基础上进行小幅度偏移（例如每个R/G/B通道的值随机加减15到30之间，注意要控制在0-255范围内），这样生成的颜色肉眼看起来会极其相似，增加玩家分辨的难度。简单和困难模式依然保持之前的完全随机策略。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"和1次\\\"极限模式\\\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"、1次\\\"极限模式\\\"和1次\\\"颜色偏移\\\"这五个关键词；第二，原要求所有生成的代码文件中的函数圈复杂度一律不得超过2，现调整为所有生成的代码文件中的函数圈复杂度一律不得超过3。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义网页项目及常规项目中常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.vue'}\\n code_files_found = False\\n\\n # 遍历当前工作空间的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅针对代码文件进行检查，忽略.md、.log等非代码文件\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 仅读取文件的第一行\\n first_line = f.readline()\\n \\n # 判断第一行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果因编码不支持或无法读取导致异常，判定为不符合要求\\n return False\\n \\n # 如果项目中至少有一个代码文件并且它们全满足第一行的声明要求，则返回 True\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_directory_depth(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n \\n try:\\n workspace = Path(workspace_path).resolve()\\n # 如果工作区不存在或不是一个目录，无法判断，返回False\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n # 遍历工作区下的所有文件和子目录\\n for path in workspace.rglob('*'):\\n # 获取相对于工作区根目录的各层级部分\\n rel_parts = path.relative_to(workspace).parts\\n \\n # 忽略由环境或工具自动生成的常见目录，避免误判\\n ignore_dirs = {'.git', '__pycache__', 'node_modules', '.idea', '.vscode'}\\n if any(part in ignore_dirs for part in rel_parts):\\n continue\\n \\n # 判断目录层级是否超过 3 层\\n # 例如: dir1/dir2/file.txt 包含3部分，属于第3层，合法\\n # dir1/dir2/dir3/file.txt 包含4部分，属于第4层，超出限制\\n if len(rel_parts) > 3:\\n return False\\n \\n return True\\n except Exception:\\n return False\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 判断总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取，视为验证失败\\n return False\\n\\n # 若至少存在一个代码文件且全部代码文件的行数都为偶数，则满足条件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def check_function_names(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 验证函数名是否仅包含小写字母和下划线（数字亦不被允许）\\n def is_valid_name(name: str) -> bool:\\n return bool(re.match(r'^[a-z_]+$', name))\\n\\n # 前端开发中常见的提取函数定义的正则表达式\\n js_patterns = [\\n # 1. 常规的 function 关键字声明\\n r'\\\\bfunction\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(',\\n # 2. ES6 箭头函数赋值声明（覆盖带括号参数及无括号单参数形式，支持 async）\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=>',\\n # 3. 赋值型匿名 function 声明（支持 async）\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\s*\\\\('\\n ]\\n # Python 常见的提取函数定义的正则表达式（作为 ast 的降级方案及其他纯文本文件的探测）\\n py_pattern = r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\('\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 对 Python 源代码使用 AST 进行高精度检查\\n if file.endswith('.py'):\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not is_valid_name(node.name):\\n return False\\n except SyntaxError:\\n # 如果由于语法不完整等原因导致 AST 解析失败，回退使用正则匹配\\n names = re.findall(py_pattern, content)\\n for name in names:\\n if not is_valid_name(name):\\n return False\\n \\n # 对前端以及常见网页文件使用正则表达式检查函数名\\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\\n for pattern in js_patterns:\\n names = re.findall(pattern, content)\\n for name in names:\\n if not is_valid_name(name):\\n return False\\n \\n # 对其他可能隐含代码块的文件启发式地扫描函数声明\\n else:\\n for pattern in js_patterns + [py_pattern]:\\n names = re.findall(pattern, content)\\n for name in names:\\n if not is_valid_name(name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"、1次\\\"极限模式\\\"和1次\\\"颜色偏移\\\"这五个关键词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 必须恰好出现一次的五个关键词\\n keywords = [\\\"RGB猜谜游戏\\\", \\\"难度切换\\\", \\\"得分统计\\\", \\\"极限模式\\\", \\\"颜色偏移\\\"]\\n \\n # 常见代码文件扩展名集合（过滤掉 .log, .md 等日志或文档文件）\\n code_exts = {\\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".py\\\", \\\".jsx\\\", \\\".ts\\\", \\\".tsx\\\"}\\n \\n # 构建提取注释的正则表达式\\n # 包含了：HTML注释、C/JS风格多行、C/JS风格单行、Python单行、Python多行注释\\n comment_pattern = re.compile(\\n r'<!--[\\\\s\\\\S]*?-->'\\n r'|/\\\\*[\\\\s\\\\S]*?\\\\*/'\\n r'|//.*'\\n r'|#.*'\\n r'|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"'\\n r\\\"|'''[\\\\s\\\\S]*?'''\\\"\\n )\\n \\n found_code_file = False\\n \\n # 遍历项目所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取出文件中所有的注释内容片段并拼接\\n comments = comment_pattern.findall(content)\\n combined_comments = \\\"\\\\n\\\".join(comments)\\n \\n # 检查各个关键词在当前文件注释中出现的次数\\n for kw in keywords:\\n if combined_comments.count(kw) != 1:\\n return False\\n \\n # 确保项目中至少找到了一个可供检查的代码文件\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的函数圈复杂度一律不得超过3\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 根据用户指令推断，主要为前端网页项目，这里涵盖最可能出现的前端扩展名及其它常见代码扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.cs', '.php', '.rb', '.go', '.rs'\\n }\\n \\n # 1. 检查工作区代码项目中的文件缩进\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 计算行首所有连续的空白字符（即该行的缩进部分）\\n indentation = line[:len(line) - len(line.lstrip())]\\n if '\\\\t' in indentation:\\n return False\\n except Exception:\\n # 忽略非UTF-8编码或其他读取异常的文件\\n pass\\n \\n # 2. 检查助手最后一轮回复的 Markdown 代码块中的缩进\\n # 正则提取 ```语言 ... ``` 格式的代码块内容\\n code_blocks = re.findall(r'```[A-Za-z0-9-]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n for line in lines:\\n indentation = line[:len(line) - len(line.lstrip())]\\n if '\\\\t' in indentation:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"后续所有轮次对代码进行任何更改前，必须先在task_plan.md中说明修改计划，再执行代码修改操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"最后一步，我们要完善页面的移动端响应式布局和夜间模式体验。1. 在页面右上角加上一个暗黑模式的切换开关，点击后页面整体背景在深灰色和纯白色间平滑切换，相关的文本颜色也要做对应反转，以保证对比度；2. 针对屏幕宽度小于600px的设备进行适配，如果颜色块是6个或9个，将底部的布局自动调整为网格结构（如每行显示3个），保持色块比例为正方形，并适当缩放顶部的字体大小，确保一屏内能完整看到标题、进度条和所有颜色块而不需要上下滑动。另外请你在后续所有轮次都严格遵循两项更新的要求：第一，原要求每个代码文件注释中恰好出现1次\\\"RGB猜谜游戏\\\"、\\\"难度切换\\\"、\\\"得分统计\\\"、\\\"极限模式\\\"和\\\"颜色偏移\\\"关键词，现调整为每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"、1次\\\"极限模式\\\"、1次\\\"颜色偏移\\\"、1次\\\"响应式布局\\\"和1次\\\"暗黑模式\\\"这七个关键词；第二，原要求所有轮次对代码进行任何更改前必须先在task_plan.md中说明修改计划，现调整为每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案，再执行代码修改操作。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"请确保在修改已有文件时输出完整的全量代码进行覆盖替换，并且绝对禁止使用 Unified Diff 及任何其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保在你输出的每个代码文件的第一行，都加入了指定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义目标代码文件的后缀名，排除.md, .json, .log等非代码文件\\n code_extensions = {'.html', '.css', '.js', '.py'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法读取或者抛出异常，视作未满足要求\\n return False\\n\\n # 只有当确实存在且检查了至少一个代码文件，并且全部符合要求时才返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"要求你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你提供的代码项目，其目录层级最高不能超过 3 层\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import pathlib\\n\\n # 将 workspace_path 转换为绝对路径对象，作为基准目录\\n base = pathlib.Path(workspace_path).resolve()\\n \\n # 遍历 workspace_path 下的所有文件和目录\\n for item in base.rglob('*'):\\n # 获取当前项相对于 base 的相对路径\\n rel = item.relative_to(base)\\n \\n # rel.parts 返回组成路径的各个层级的元组\\n # 长度表示它在项目中的深度（比如 root/a/b.txt 长度为 2）\\n # 深度如果超过 3，则不满足最高层级不超过 3 层的要求\\n if len(rel.parts) > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每次更新已存在的文件前，要求你必须先通过对应的工具去查看该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"无论是代码行还是空行，你生成的每个代码文件加起来的总行数应当是偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 涵盖 Web 项目常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.json'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 使用 '\\\\n' 切分可以准确统计包括空行在内的所有行数\\n lines = content.split('\\\\n')\\n \\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果遇到无法以 utf-8 读取的特殊文件则忽略或视作异常不作直接判定\\n continue\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 用于提取 JavaScript / TypeScript 或 HTML 内嵌 JS 的函数名称的正则表达式\\n js_patterns = [\\n # 匹配普通函数声明: function funcName(...)\\n re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('),\\n # 匹配变量赋值型声明: const/let/var funcName = function(...) / (...) => / arg =>\\n re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function|\\\\([^()]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'),\\n # 匹配对象属性赋值声明: obj.funcName = function(...) / (...) => / arg =>\\n re.compile(r'\\\\.\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function|\\\\([^()]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'),\\n # 匹配对象字面量键声明: funcName: function(...) / (...) => / arg =>\\n re.compile(r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function|\\\\([^()]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'),\\n # 匹配类或对象方法简写形式: funcName(...) {\\n re.compile(r'(?:^|[{};,])\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^()]*\\\\)\\\\s*\\\\{')\\n ]\\n\\n # 用于提取 Python 的函数名称的正则表达式\\n py_patterns = [\\n re.compile(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n ]\\n\\n # 合法的函数名模式：只包含小写字母和下划线\\n valid_name_pattern = re.compile(r'^[a-z_]+$')\\n\\n # 需要排除的 JS 关键字，防止类方法简写的正则对控制流语法造成误判\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'with', 'constructor',\\n 'try', 'else', 'finally', 'do', 'class', 'await', 'async'\\n }\\n\\n # 遍历项目目录，检查所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.py', '.html', '.htm', '.ts', '.jsx', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n # 清除注释，防止注释中的陈旧代码及非规范文本引起误报\\n if file.endswith('.py'):\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r'\\\\'\\\\'\\\\'.*?\\\\'\\\\'\\\\'', '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\".*?\\\\\\\"\\\\\\\"\\\\\\\"', '', content, flags=re.DOTALL)\\n else:\\n if file.endswith(('.html', '.htm')):\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n\\n func_names = []\\n \\n # 依据文件类型进行函数名提取\\n if file.endswith('.py'):\\n for p in py_patterns:\\n func_names.extend(p.findall(content))\\n else:\\n for p in js_patterns:\\n func_names.extend(p.findall(content))\\n\\n # 对该文件中所有检测到的函数名进行合法性检验\\n for name in func_names:\\n if name in keywords:\\n continue\\n if not valid_name_pattern.match(name):\\n return False\\n except Exception:\\n # 容错处理：忽略不可读的文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"每个由你生成的代码文件，其注释中都必须恰好出现1次\\\"RGB猜谜游戏\\\"、1次\\\"难度切换\\\"、1次\\\"得分统计\\\"、1次\\\"极限模式\\\"、1次\\\"颜色偏移\\\"、1次\\\"响应式布局\\\"和1次\\\"暗黑模式\\\"这七个关键词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 待校验的7个关键词\\n keywords = [\\n \\\"RGB猜谜游戏\\\", \\n \\\"难度切换\\\", \\n \\\"得分统计\\\", \\n \\\"极限模式\\\", \\n \\\"颜色偏移\\\", \\n \\\"响应式布局\\\", \\n \\\"暗黑模式\\\"\\n ]\\n \\n # 基于用户指令推测的可能代码文件类型\\n valid_exts = {'.html', '.css', '.js', '.py'}\\n \\n def get_comments(content: str, ext: str) -> str:\\n comments = []\\n if ext == '.html':\\n # 提取 HTML 原生多行注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 提取内联 script 和 style 代码块\\n blocks = re.findall(r'<script.*?>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n blocks.extend(re.findall(r'<style.*?>(.*?)</style>', content, re.DOTALL | re.IGNORECASE))\\n for block in blocks:\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', block, re.DOTALL))\\n # 提取 // 注释，通过 (?<!:) 避免匹配 http:// 中的斜杠\\n comments.extend(re.findall(r'(?<!:)\\\\/\\\\/(.*)', block))\\n elif ext in {'.js', '.css'}:\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'(?<!:)\\\\/\\\\/(.*)', content))\\n elif ext == '.py':\\n comments.extend(re.findall(r'#(.*)', content))\\n comments.extend(re.findall(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', content, re.DOTALL))\\n comments.extend(re.findall(r\\\"'''(.*?)'''\\\", content, re.DOTALL))\\n \\n return \\\" \\\\n \\\".join(comments)\\n\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果遇到无法读取的文件可以直接跳过\\n continue\\n \\n # 提取当前文件全部注释\\n comments_text = get_comments(content, ext)\\n \\n # 校验每一个关键字是否在注释里恰好出现 1 次\\n for kw in keywords:\\n if comments_text.count(kw) != 1:\\n return False\\n \\n # 只要存在被检查的代码文件，且全都通过了上述校验条件，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件中的函数圈复杂度一律不得超过3\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过TodoWrite工具规划好任务，然后才能修改代码，当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用TodoComplete工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次项目改动后必须生成用以记录本次改动的新文件change_log_X.log，X需要替换为从0开始计算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保生成的所有代码完全使用空格进行缩进，严禁出现任何制表符缩进\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 定义需要检查的常见代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.py', '.json', '.ts', '.jsx', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.lower().endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 提取行首的前导空白字符\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char == ' ' or char == '\\\\t':\\n leading_whitespace += char\\n else:\\n break\\n \\n # 如果前导空白中包含制表符，说明存在制表符缩进\\n if '\\\\t' in leading_whitespace:\\n return False\\n except Exception:\\n # 若文件由于编码等原因无法正常读取，则跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案，再执行代码修改操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["打开页面，检查页面是否整体居中展示且为卡片式设计，顶部包含类似'RGB(XXX, XXX, XXX)'的大字号文本，中间状态控制栏左侧有提示信息、右侧有'新游戏'按钮，底部为颜色块容器，顶部有'简单'、'困难'、'极限'三个纯文本按钮。","检查初始渲染状态，验证'困难'难度按钮是否被默认选中（具有加粗并改变颜色的高亮的颜色样式），且页面底部的容器中恰好显示6个颜色块。","刷新页面或点击'新游戏'按钮，验证顶部的RGB文本值是否发生变化，同时底部所有的颜色块的背景色是否被替换成了一组全新的随机颜色。","截屏并使用取色器提取顶部文本显示的RGB目标值，在底部的所有颜色块中逐一比对，验证底部必定存在且只有一个颜色块的背景色与该RGB值完全一致。","点击任意一个颜色不匹配的错误颜色块，验证被点击的色块是否触发淡出动画（例如透明度变为0）且无法再次点击，同时状态控制栏左侧的提示文字变为'再试一次！'。","点击颜色与顶部RGB一致的正确颜色块，观察过渡动画，验证页面的标题区域背景色以及此时所有未隐藏的颜色块的背景色，是否带有过渡动画平滑变为了该正确的RGB目标颜色。","点击正确颜色块后，验证状态栏左侧的提示文字是否更新为'回答正确！'，且右侧原先的'新游戏'按钮文字是否变为了'再玩一次？'。","点击'简单'难度按钮，验证按钮样式是否改变以示高亮，且页面底部的颜色块数量是否立刻变更为3个，游戏状态被重置。","在'简单'模式下猜对一次颜色，检查状态栏是否显示'当前得分: 1'；在'困难'模式下猜对一次颜色，检查得分是否增加3分。","在玩家得分累计为大于0的任何时候，点击一个错误的颜色块，验证显示分数的文字是否立刻被清零变为'当前得分: 0'。","在答对并得分时，观察分数数字的动画效果，验证得分增加的一瞬间数字是否会触发轻微的放大缩小（跳动/缩放）动画效果，持续约0.5秒。","点击'极限'难度按钮，验证颜色块数量是否变为9个，同时在顶部的RGB文本值下方是否出现了一个代表剩余时间的水平进度条，且随时间匀速缩短。","在'极限'模式下截取颜色块区域并取色，验证除正确色块外的8个错误色块颜色与目标颜色视觉极其相近（每个RGB通道值的偏差均在15到30之间，而非完全随机，但要控制在0-255的范围内）。","在'极限'模式下，不进行任何点击操作，等待5秒钟进度条耗尽，验证页面中央是否立刻弹出了自定义的'Game Over'弹窗，并在弹窗内展示最终连续得分及提供一个'重新挑战'按钮。","开启新的一局'极限'模式，在进度条走完之前点击一个错误的颜色块，验证游戏是否直接终止，并同样弹出中央带有最终连续得分和'重新挑战'的Game Over弹窗。","点击Game Over弹窗内的'重新挑战'按钮，验证弹窗是否关闭消失，当前得分重置为0，并且开始新一局游戏（进度条重新变满并开始新一轮倒计时）。","点击页面右上角的暗黑模式切换开关，验证页面的全局背景是否在深灰色和纯白色间平滑切换，并且页面的文字颜色会做出反转（如深底白字/白底黑字）以保证阅读对比度。","将浏览器窗口宽度调整到小于600px并进入'困难'或'极限'模式（6或9个色块），验证底部颜色块布局是否自动转为网格结构（如每行显示3个）且每个颜色块依旧保持正方形比例。","在小于600px宽度的移动端视图下，验证顶部的字体大小有适当缩小，从标题、进度条到所有九宫格颜色块均能够在一个屏幕高度内完整呈现，无需上下滑动页面。"],"string":"[\n \"打开页面，检查页面是否整体居中展示且为卡片式设计，顶部包含类似'RGB(XXX, XXX, XXX)'的大字号文本，中间状态控制栏左侧有提示信息、右侧有'新游戏'按钮，底部为颜色块容器，顶部有'简单'、'困难'、'极限'三个纯文本按钮。\",\n \"检查初始渲染状态，验证'困难'难度按钮是否被默认选中（具有加粗并改变颜色的高亮的颜色样式），且页面底部的容器中恰好显示6个颜色块。\",\n \"刷新页面或点击'新游戏'按钮，验证顶部的RGB文本值是否发生变化，同时底部所有的颜色块的背景色是否被替换成了一组全新的随机颜色。\",\n \"截屏并使用取色器提取顶部文本显示的RGB目标值，在底部的所有颜色块中逐一比对，验证底部必定存在且只有一个颜色块的背景色与该RGB值完全一致。\",\n \"点击任意一个颜色不匹配的错误颜色块，验证被点击的色块是否触发淡出动画（例如透明度变为0）且无法再次点击，同时状态控制栏左侧的提示文字变为'再试一次！'。\",\n \"点击颜色与顶部RGB一致的正确颜色块，观察过渡动画，验证页面的标题区域背景色以及此时所有未隐藏的颜色块的背景色，是否带有过渡动画平滑变为了该正确的RGB目标颜色。\",\n \"点击正确颜色块后，验证状态栏左侧的提示文字是否更新为'回答正确！'，且右侧原先的'新游戏'按钮文字是否变为了'再玩一次？'。\",\n \"点击'简单'难度按钮，验证按钮样式是否改变以示高亮，且页面底部的颜色块数量是否立刻变更为3个，游戏状态被重置。\",\n \"在'简单'模式下猜对一次颜色，检查状态栏是否显示'当前得分: 1'；在'困难'模式下猜对一次颜色，检查得分是否增加3分。\",\n \"在玩家得分累计为大于0的任何时候，点击一个错误的颜色块，验证显示分数的文字是否立刻被清零变为'当前得分: 0'。\",\n \"在答对并得分时，观察分数数字的动画效果，验证得分增加的一瞬间数字是否会触发轻微的放大缩小（跳动/缩放）动画效果，持续约0.5秒。\",\n \"点击'极限'难度按钮，验证颜色块数量是否变为9个，同时在顶部的RGB文本值下方是否出现了一个代表剩余时间的水平进度条，且随时间匀速缩短。\",\n \"在'极限'模式下截取颜色块区域并取色，验证除正确色块外的8个错误色块颜色与目标颜色视觉极其相近（每个RGB通道值的偏差均在15到30之间，而非完全随机，但要控制在0-255的范围内）。\",\n \"在'极限'模式下，不进行任何点击操作，等待5秒钟进度条耗尽，验证页面中央是否立刻弹出了自定义的'Game Over'弹窗，并在弹窗内展示最终连续得分及提供一个'重新挑战'按钮。\",\n \"开启新的一局'极限'模式，在进度条走完之前点击一个错误的颜色块，验证游戏是否直接终止，并同样弹出中央带有最终连续得分和'重新挑战'的Game Over弹窗。\",\n \"点击Game Over弹窗内的'重新挑战'按钮，验证弹窗是否关闭消失，当前得分重置为0，并且开始新一局游戏（进度条重新变满并开始新一轮倒计时）。\",\n \"点击页面右上角的暗黑模式切换开关，验证页面的全局背景是否在深灰色和纯白色间平滑切换，并且页面的文字颜色会做出反转（如深底白字/白底黑字）以保证阅读对比度。\",\n \"将浏览器窗口宽度调整到小于600px并进入'困难'或'极限'模式（6或9个色块），验证底部颜色块布局是否自动转为网格结构（如每行显示3个）且每个颜色块依旧保持正方形比例。\",\n \"在小于600px宽度的移动端视图下，验证顶部的字体大小有适当缩小，从标题、进度条到所有九宫格颜色块均能够在一个屏幕高度内完整呈现，无需上下滑动页面。\"\n]","isConversational":false}}},{"rowIdx":61,"cells":{"task_id":{"kind":"number","value":62,"string":"62"},"repository_policy":{"kind":"string","value":"你需要在后续所有交互轮次中始终严格遵循以下要求，不得违反：\n1. 优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段。\n2. 不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码。\n3. 文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现。\n4. 对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-。\n5. 你的每一条回复都必须使用语气词\"嗨～\"来打头。\n6. 请确保在对所有用户的回复末尾，都附上\"哇～\"语气词。\n7. 请保证生成的每个代码文件内，包含的注释全都是中文。\n8. 当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的常见代码文件扩展名\n code_extensions = {\n '.py', '.java', '.cpp', '.c', '.h', '.cs', '.js', '.ts', \n '.go', '.rb', '.php', '.sh', '.html', '.css', '.sql', \n '.swift', '.kt', '.rs', '.m', '.scala', '.dart'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found += 1\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含要求的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果无法读取（例如存在非 utf-8 编码的问题），直接判定不满足\n return False\n \n # 确保不仅没有违反要求的文件，而且确实生成了至少一个以上的代码文件\n return code_files_found > 0","tags":["Content","Format"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除回复内容左侧可能的空白字符，判断是否以“嗨～”开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"哇～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 定义正则表达式：匹配最后的内容是\"哇～\"，允许后面跟随空白字符或常见的句末标点符号\n pattern = r'哇～[\\s!！.。?？]*$'\n \n # 使用正则在整个回复中搜索结尾部分匹配情况\n if re.search(pattern, response):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"请保证生成的每个代码文件内，包含的注释全都是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]}],"string":"[\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的常见代码文件扩展名\\n code_extensions = {\\n '.py', '.java', '.cpp', '.c', '.h', '.cs', '.js', '.ts', \\n '.go', '.rb', '.php', '.sh', '.html', '.css', '.sql', \\n '.swift', '.kt', '.rs', '.m', '.scala', '.dart'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found += 1\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含要求的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取（例如存在非 utf-8 编码的问题），直接判定不满足\\n return False\\n \\n # 确保不仅没有违反要求的文件，而且确实生成了至少一个以上的代码文件\\n return code_files_found > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除回复内容左侧可能的空白字符，判断是否以“嗨～”开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"哇～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 定义正则表达式：匹配最后的内容是\\\"哇～\\\"，允许后面跟随空白字符或常见的句末标点符号\\n pattern = r'哇～[\\\\s!！.。?？]*$'\\n \\n # 使用正则在整个回复中搜索结尾部分匹配情况\\n if re.search(pattern, response):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件内，包含的注释全都是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我创建一个单页应用，包含两部分功能：书籍展示和井字棋游戏。首先，我们需要搭建基本的页面骨架。在页面顶部设置一个固定的水平导航栏，包含「我的书架」和「井字棋挑战」两个导航按钮。下方是主要的内容展示区。点击不同的导航按钮可以切换内容区的视图。目前视图里只需要分别居中显示文字「书籍内容建设中...」和「井字棋加载中...」作为占位即可。整体背景使用浅灰色，导航栏使用深色，UI风格尽量现代、简洁一点。另外后续所有操作请始终遵循以下要求：请注意，在你所生成的代码文件里，所有的文件名都必须同时包含数字、下划线和大小写字母；生成代码时，请保证文件中所有的函数名只包含小写字母与下划线。","instruction_following_checklist":[{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对单页应用和常见代码文件定义后缀集合\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断当前文件是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 读取首行\n first_line = f.readline()\n \n # 检查首行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 无法正常读取的异常情况，可按未满足规范处理\n return False\n \n # 如果找到了至少一个代码文件，并且全满足要求，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 考虑到助手回复的最开头可能包含空格或换行符等空白字符，先使用 lstrip() 去除左侧空白字符\n clean_response = response.lstrip()\n # 检查清理后的回复内容是否以指定的语气词\"嗨～\"打头\n return clean_response.startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"哇～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 去除回复内容首尾可能存在的空白字符（如换行符、空格等）\n cleaned_response = response.strip()\n \n # 检查清理后的回复是否以\"哇～\"结尾\n if cleaned_response.endswith(\"哇～\"):\n return True\n return False","tags":["Content","Persona"]},{"constraint":"请保证生成的每个代码文件内，包含的注释全都是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"请注意，在你所生成的代码文件里，所有的文件名都必须同时包含数字、下划线和大小写字母","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件扩展名，匹配单页应用及常规代码项目\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n \n # 遍历当前项目地址\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录，防止检查git或系统隐藏文件夹\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 忽略隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n \n # 仅验证所生成的代码文件\n if ext in code_extensions:\n has_code_file = True\n \n # 检查文件名是否同时包含数字、下划线、大写字母和小写字母\n has_digit = any(c.isdigit() for c in file)\n has_lower = any(c.islower() for c in file)\n has_upper = any(c.isupper() for c in file)\n has_underscore = '_' in file\n \n # 如果任意一个代码文件未能全部包含以上四种元素，则视为不满足要求\n if not (has_digit and has_lower and has_upper and has_underscore):\n return False\n \n # 需要确保工作区确实生成了代码文件，而不是空项目\n return has_code_file","tags":["Style","Naming"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n # 工作区不存在则无法进行违规匹配，认为没有违规\n return True\n \n # 定义正则表达式来匹配常见的函数定义形式\n # 匹配常规的 function 声明\n func_pattern = r'\\b(?:async\\s+)?function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('\n # 匹配变量赋值形式的函数 (包括普通匿名函数和箭头函数，兼顾可能存在的 async 修饰符)\n var_pattern = r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\(?[a-zA-Z0-9_$,\\s]*\\)?\\s*=>)'\n # 匹配 Python 函数定义\n py_pattern = r'\\b(?:async\\s+)?def\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('\n \n # 针对前端单页应用和可能的后端API，推断可能的代码文件格式\n allowed_extensions = {'.html', '.htm', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 跳过 node_modules 目录等第三方依赖目录\n if 'node_modules' in root:\n continue\n \n for file in files:\n # 过滤掉压缩代码文件，避免提取到第三方的框架内置函数干扰检测\n if file.endswith('.min.js'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext not in allowed_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取文件中所有能够被捕获的函数名称\n func_names = []\n func_names.extend(re.findall(func_pattern, content))\n func_names.extend(re.findall(var_pattern, content))\n func_names.extend(re.findall(py_pattern, content))\n \n # 验证每一个函数名称是否只包含小写字母与下划线\n for name in func_names:\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]}]},{"round_id":1,"instruction":"框架有了，现在我们来完善「我的书架」页面。请使用本地模拟数据生成大约8-10本图书（包含字段：书名、作者、封面图的占位色块或链接、分类标签、10分制评分）。使用CSS Grid网格布局把这些书籍以卡片的形式展示在页面上。每张卡片需要有白色的背景、圆角和轻微的投影效果。另外，当鼠标悬浮在某张书籍卡片上时，卡片要平滑地向上浮动几个像素，并且阴影加深，增加交互感。此外请留意，后续所有操作请始终遵循以下更新要求：第一，确保生成的代码文件名仅由小写字母和下划线构成；第二，生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案。","instruction_following_checklist":[{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断，单页应用可能生成的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n # 遍历工作目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 utf-8 编码读取文件第一行\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查首行是否包含要求的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法正常读取或打开失败，视为验证失败\n return False\n \n # 如果没有找到任何代码文件，也视为不满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容左侧的空白字符后，判断是否以\"嗨～\"打头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"哇～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符\n clean_resp = response.strip()\n \n # 使用正则表达式匹配文本末尾是否为“哇～”\n # 允许“哇～”之后跟随常见的中英文标点符号以及空白字符（如：哇～、哇～！、哇～。 等情况）\n pattern = r'哇～[。，！？\\.,!\\?\\s]*$'\n \n if re.search(pattern, clean_resp):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"请保证生成的每个代码文件内，包含的注释全都是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"确保生成的代码文件名仅由小写字母和下划线构成","validation_code":"import os\nimport re\n\ndef check_requirement(response, workspace_path):\n # 如果工作区目录不存在，无法验证生成的代码\n if not os.path.exists(workspace_path):\n return False\n \n # 定义前端项目中常见的代码文件后缀\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.json', '.vue', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 忽略以 . 开头的隐藏文件（例如 .gitignore）\n if file.startswith('.'):\n continue\n \n # 获取小写的文件扩展名\n ext = os.path.splitext(file)[1].lower()\n \n # 筛选我们认为是“代码文件”的类型，或者是无后缀的可执行脚本\n if ext in code_exts or not ext:\n has_code_file = True\n \n # 按照 `.` 将文件名进行分割，并对每一部分进行检查\n parts = file.split('.')\n for part in parts:\n # 检查是否每一部分都完全由小写字母和下划线构成\n if not re.fullmatch(r'[a-z_]+', part):\n return False\n \n # 只要存在至少一个代码文件并且没有在上面返回 False，说明完全满足条件\n return has_code_file","tags":["Style","Naming"]},{"constraint":"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n \n if not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 1. 针对 Python 文件的 AST 解析提取\n if ext == '.py':\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n source = f.read()\n tree = ast.parse(source)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 判断是否只包含小写字母和下划线\n if not bool(re.match(r'^[a-z_]+$', node.name)):\n return False\n except Exception:\n pass\n \n # 2. 针对前端 JS/TS/HTML 的正则表达式提取\n elif ext in {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue'}:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 若为 HTML / Vue 等混合类型，提取其中 script 标签的内容\n if ext in {'.html', '.htm', '.vue'}:\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n text_to_check = '\\n'.join(scripts)\n else:\n text_to_check = content\n \n # 暴力去除多行和单行注释，防止提取到注释中的无效内容\n text_to_check = re.sub(r'/\\*.*?\\*/', '', text_to_check, flags=re.DOTALL)\n text_to_check = re.sub(r'//.*', '', text_to_check)\n \n funcs = []\n # (1) 匹配 function foo()\n funcs.extend(re.findall(r'\\bfunction\\s+([a-zA-Z0-9_]+)\\s*\\(', text_to_check))\n # (2) 匹配 const/let/var foo = function / () => / a => \n funcs.extend(re.findall(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z0-9_]+\\s*=>)', text_to_check))\n # (3) 匹配 foo: function / () => / a => (对象属性方法)\n funcs.extend(re.findall(r'\\b([a-zA-Z0-9_]+)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z0-9_]+\\s*=>)', text_to_check))\n \n # 校验该文件内提取到的所有函数名\n for name in funcs:\n if not bool(re.match(r'^[a-z_]+$', name)):\n return False\n except Exception:\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":2,"instruction":"接下来给「我的书架」增加实用的筛选和排序功能。在书籍网格区域的上方，添加一个水平排列的操作栏。操作栏里需要有三个组件：1. 一个搜索输入框，可以实时根据书名或作者进行模糊匹配；2. 一个下拉菜单，提取所有数据的分类（如科幻、文学、科技等）进行分类筛选；3. 一个「评分排序」按钮，点击可以在“评分从高到低”和“评分从低到高”之间切换。所有的筛选、搜索和排序操作必须能互相叠加，并实时更新下方的书籍卡片列表。另外请留意，后续所有操作请始终遵循以下更新要求：第一，在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母；第二，为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 30s` 包装。","instruction_following_checklist":[{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check_code_files_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 涵盖前端单页应用以及常见后台/脚本语言的代码扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.sh', \n '.rb', '.php', '.go', '.rs', '.swift', '.kt'\n }\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件进行检查\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline().strip()\n # 检查首行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于编码异常或其他原因无法读取首行，视为不符合要求\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复是否以\"嗨～\"打头。\n \"\"\"\n if not response:\n return False\n \n # 去除首部的空白字符后，判断是否以\"嗨～\"开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"哇～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 去除回复末尾可能存在的空白字符（包括空格、换行符等）\n clean_response = response.rstrip()\n \n # 判断清理后的回复内容是否以\"哇～\"结尾\n if clean_response.endswith(\"哇～\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"请保证生成的每个代码文件内，包含的注释全都是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"确保生成的代码文件名仅由小写字母和下划线构成","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 推断单页应用相关的常见代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.vue', '.json'\n }\n \n # 遍历当前项目的工作区文件\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和常见的依赖包目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n # 提取文件扩展名并转为小写，以便做文件类型匹配\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为推断出的代码文件类型\n if ext in code_extensions:\n # 剔除作为扩展名分隔符的 '.' 后，考察剩余所有字符\n name_without_dots = file.replace('.', '')\n \n # 如果除点以外的字符并不只由小写字母和下划线构成，则验证失败\n if not re.match(r'^[a-z_]+$', name_without_dots):\n return False\n \n # 如果所有的代码文件名都符合要求，或者没有生成代码文件，则判定为 True\n return True","tags":["Style","Naming"]},{"constraint":"在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def check_name_rules(name: str) -> bool:\n \"\"\"检查函数名是否同时包含：数字、下划线、大写字母和小写字母\"\"\"\n has_digit = any(c.isdigit() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_lower = any(c.islower() for c in name)\n has_underscore = '_' in name\n return has_digit and has_upper and has_lower and has_underscore\n\n def extract_js_functions(code: str) -> list:\n \"\"\"从 JS/TS/HTML 类的前端源码中利用正则提取函数名\"\"\"\n # 清除注释内容以免干扰正则匹配\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n \n functions = []\n # 1. 匹配普通函数声明，例如 function myFunc_1A()\n p1 = re.findall(r'\\bfunction\\s+([A-Za-z0-9_]+)\\s*\\(', code)\n \n # 2. 匹配变量赋值形式的函数，例如 const myFunc_1A = () => {} 或 let func = function()\n p2 = re.findall(r'\\b(?:const|let|var)\\s+([A-Za-z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[A-Za-z0-9_]+\\s*=>)', code)\n \n # 3. 匹配对象内方法赋值，例如 myFunc_1A: function() {} 或 myFunc_1A: () => {}\n p3 = re.findall(r'\\b([A-Za-z0-9_]+)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[A-Za-z0-9_]+\\s*=>)', code)\n \n functions.extend(p1 + p2 + p3)\n \n # 排除可能是因为正则误捕获的编程关键字\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else', 'await', 'async', 'yield'}\n return [f for f in functions if f not in keywords and f]\n\n def extract_python_functions(code: str) -> list:\n \"\"\"从 Python 源码中利用 AST 提取函数名\"\"\"\n functions = []\n try:\n import ast\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n functions.append(node.name)\n except Exception:\n pass\n return functions\n \n # 验证工作空间是否存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 根据常见的代码文件后缀判断是否需要读取\n if not file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.py')):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n funcs = []\n if file.endswith('.py'):\n funcs = extract_python_functions(content)\n else:\n funcs = extract_js_functions(content)\n \n # 校验文件中提取出来的所有函数名称\n for func in funcs:\n if not check_name_rules(func):\n # 只要发现存在任意一个不符合规则的函数名，就立刻判定失败\n return False\n \n # 如果遍历完毕，全部检查的函数名称均符合要求或无函数，则返回True\n return True","tags":["Style","Naming"]},{"constraint":"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 30s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":3,"instruction":"书架功能做得很棒了。现在我们切换到「井字棋挑战」页面，来实现核心游戏逻辑。在页面中央绘制一个经典的 3x3 井字棋盘，格子边界要清晰。支持两名玩家轮流点击空白格子落子，分别用蓝色的 'X' 和红色的 'O' 表示，文字要足够大且居中。每次落子后，程序需要自动判断是否有玩家获胜（横、竖、对角线连成三子）或是棋盘已满导致平局。如果游戏结束，在棋盘上方显示醒目的文字提示结果（如「玩家 X 获胜！」），并同时展示一个「再来一局」的按钮，点击后清空棋盘重新开始。另外请留意，后续所有操作请始终遵循以下更新要求：第一，为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 60s` 包装；第二，注意代码格式，生成的代码文件中必须仅用 4 个空格实现缩进。","instruction_following_checklist":[{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义本次任务中单页应用及常规前后端可能涉及的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.py', '.java', '.cpp', '.c', '.go', '.php', '.rb', '.sh'\n }\n \n # 定义需要排除的无关目录，避免检查第三方库或构建产物\n exclude_dirs = {\n '.git', 'node_modules', 'venv', 'env', '__pycache__', \n 'dist', 'build', '.idea', '.vscode'\n }\n \n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉排除的目录，不再向下遍历\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查首行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line.strip():\n return False\n except Exception:\n # 如果由于编码等原因无法正常读取文件第一行，视为不满足要求\n return False\n \n # 至少需要发现一个代码文件才算完成了生成代码的过程\n return code_files_found","tags":["Content","Format"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否以\"嗨～\"打头。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n if not response:\n return False\n \n # 去除开头的空白字符后，判断是否以\"嗨～\"开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"哇～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除末尾的空白字符\n response = response.strip()\n \n # 考虑到助手可能会在结尾添加标点符号，去除末尾可能出现的常见中英文标点\n response = re.sub(r'[。！？\\!\\.\\?]+$', '', response).strip()\n \n # 判断是否以“哇～”结尾\n return response.endswith(\"哇～\")","tags":["Content","Persona"]},{"constraint":"请保证生成的每个代码文件内，包含的注释全都是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"确保生成的代码文件名仅由小写字母和下划线构成","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码文件后缀集合，涵盖单页应用及相关可能用到的代码格式\n code_exts = {\n '.html', '.htm', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.vue', '.json', '.py', '.sh'\n }\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git, .idea, .vscode 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n # 提取文件扩展名并统一转换为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否属于代码文件\n if ext in code_exts:\n has_code_file = True\n # 获取文件的基础名称（不包含扩展名部分）\n base_name = os.path.splitext(file)[0]\n \n # 检查文件名（不含扩展名）是否仅由小写字母和下划线构成\n if not re.fullmatch(r'[a-z_]+', base_name):\n return False\n \n # 如果没有找到任何代码文件，则无法证明满足了要求，返回False；否则说明都符合，返回True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义用于匹配前端 JavaScript/TypeScript/HTML 中函数定义的正则表达式\n js_patterns = [\n # 匹配普通函数声明: function myFunc (...)\n r'\\bfunction\\s+([a-zA-Z0-9_]+)\\s*\\(',\n # 匹配变量赋值型函数（包括箭头函数与匿名函数）: const myFunc = (...) => 或者 const myFunc = function\n r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)\\s*=>|[a-zA-Z0-9_]+\\s*=>|function\\b)'\n ]\n # 定义 Python 函数的正则表达式作为 fallback\n py_patterns = [\n r'\\bdef\\s+([a-zA-Z0-9_]+)\\s*\\('\n ]\n \n def remove_js_comments(code: str) -> str:\n \"\"\"移除JS/TS/前端代码中的单行和多行注释\"\"\"\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n code = re.sub(r'//.*', '', code)\n return code\n \n def remove_py_comments(code: str) -> str:\n \"\"\"移除Python代码中的单行注释\"\"\"\n code = re.sub(r'#.*', '', code)\n return code\n\n def check_name(name: str) -> bool:\n \"\"\"检查函数名是否同时具有数字、下划线、大写字母和小写字母\"\"\"\n has_digit = any(c.isdigit() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_lower = any(c.islower() for c in name)\n has_under = '_' in name\n return has_digit and has_upper and has_lower and has_under\n\n # 需要跳过的无需检查的目录名称（例如包依赖或者构建产物）\n exclude_dirs = {'node_modules', '.git', 'venv', '.venv', 'dist', 'build'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除掉不需要遍历的文件夹\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 验证前端、全栈常见语言文件\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.vue']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n content = remove_js_comments(content)\n \n for pattern in js_patterns:\n for match in re.finditer(pattern, content):\n func_name = match.group(1)\n # 忽略某些在复杂赋值情况下被错误捕捉的保留字\n if func_name in {'default', 'async', 'return', 'get', 'set'}:\n continue\n if not check_name(func_name):\n return False\n except Exception:\n # 遇到文件读取或编码问题时跳过该文件继续\n continue\n \n # 验证 Python 文件\n elif ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n import ast\n tree = ast.parse(content)\n # 优先利用 AST 安全准确地遍历 Python 文件中定义的方法和函数\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not check_name(node.name):\n return False\n except Exception:\n # 如果 ast 解析失败（如语法错误），回退至正则匹配\n try:\n content = remove_py_comments(content)\n for pattern in py_patterns:\n for match in re.finditer(pattern, content):\n func_name = match.group(1)\n if not check_name(func_name):\n return False\n except Exception:\n continue\n\n # 未找到任何违反命名规则的函数，要求得到满足\n return True","tags":["Style","Naming"]},{"constraint":"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"注意代码格式，生成的代码文件中必须仅用 4 个空格实现缩进","validation_code":"import os\n\ndef check_indentation_is_four_spaces(response: str, workspace_path: str) -> bool:\n # 针对单页应用上下文，圈定可能产生的代码文件扩展名\n code_exts = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\n \n total_valid = 0\n total_invalid = 0\n has_tabs = False\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n stripped = line.lstrip()\n \n # 忽略空行\n if not stripped:\n continue\n \n # 忽略部分可能导致干扰的注释行\n if stripped.startswith(('*', '//', '#', '<!--', '/*')):\n continue\n \n # 截取前导空白字符\n leading_whitespace = line[:len(line) - len(stripped)]\n \n # 如果前导缩进中使用了制表符(Tab)，直接记为不通过\n if '\\t' in leading_whitespace:\n has_tabs = True\n break\n \n # 统计缩进层级分布情况\n l = len(leading_whitespace)\n if l > 0:\n if l % 4 == 0:\n total_valid += 1\n else:\n total_invalid += 1\n \n if has_tabs:\n break\n if has_tabs:\n break\n \n # 未生成代码文件\n if not has_code_files:\n return False\n \n # 包含了制表符缩进\n if has_tabs:\n return False\n \n # 完全没有4的倍数缩进，却存在其他不规则空格缩进\n if total_invalid > 0 and total_valid == 0:\n return False\n \n # 容错判断：不合规的缩进行数超过合规行数的 20%，并且绝对值大于 3 行，\n # 即可确定其实质上混用了缩进或根本未使用 4 空格作为缩进标准。\n if total_invalid > max(3, total_valid * 0.2):\n return False\n \n return True","tags":["Style","Layout"]}]},{"round_id":4,"instruction":"最后，给这个井字棋游戏增加一个有点难度的功能：时光倒流（历史步骤回溯）。在棋盘的右侧新增一个「历史对战记录」的列表面板。每当玩家成功落子后，列表中就要新增一个按钮（例如「回到第 1 步」、「回到第 2 步」等，最开始还有个「回到游戏起点」）。玩家点击列表中的任意历史记录，棋盘状态必须立刻回退到那一步的样子。注意：如果玩家在回退后的状态下重新落子了一步新棋，那么原先那一步之后的历史记录必须被全部清除。同时在顶部加一个当前轮到谁下棋的动态提示器。\n另外请留意，后续所有操作请始终遵循以下更新要求：第一，注意代码格式，生成的代码文件中必须仅用 2 个空格实现缩进；第二，代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径。","instruction_following_checklist":[{"constraint":"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名，涵盖前端单页应用及其他常见语言\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs',\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.sh', '.bash'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果无法读取或解码出错等，视为不满足要求\n return False\n\n # 只有在至少遍历到一个代码文件且全都符合时返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否以特定的语气词\"嗨～\"开头。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，表示是否满足要求。\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容可能存在的前导空白字符，然后判断是否以要求的内容打头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"哇～\"语气词","validation_code":"def check_response_ending(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除回复首尾可能存在的空白字符、换行符后，判断是否以\"哇～\"结尾\n return response.strip().endswith(\"哇～\")","tags":["Content","Persona"]},{"constraint":"请保证生成的每个代码文件内，包含的注释全都是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"确保生成的代码文件名仅由小写字母和下划线构成","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件后缀类型集合（重点关注前端开发及常规语言）\n code_extensions = {\n '.py', '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.java', '.cpp', '.c', '.h', '.hpp', '.cs', '.php', '.rb', \n '.go', '.rs', '.swift', '.kt', '.sh'\n }\n \n has_target_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录及常见的依赖包或缓存目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验\n if ext in code_extensions:\n has_target_files = True\n name = os.path.splitext(file)[0]\n \n # 检查文件名（不含扩展名部分）是否仅由小写字母和下划线构成\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n # 当找到了对应的代码文件并且没有任何文件名违规时，返回 True\n return has_target_files","tags":["Style","Naming"]},{"constraint":"在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义通常可能包含函数编写的代码文件后缀名\n valid_extensions = {'.js', '.py', '.html', '.ts', '.jsx', '.tsx', '.vue'}\n \n # 编译匹配各种常见函数定义的正则表达式\n # 1. Python 中的函数声明，例如: def func_name( 或者 async def func_name(\n py_func_pattern = re.compile(r'\\b(?:async\\s+)?def\\s+([a-zA-Z_]\\w*)\\s*\\(')\n \n # 2. JavaScript / TypeScript 中的各种函数声明\n js_func_patterns = [\n # 匹配直接声明的函数，例如: function func_name( 或者 async function func_name(\n re.compile(r'\\b(?:async\\s+)?function\\s+([a-zA-Z_$][\\w$]*)\\s*\\('),\n # 匹配赋值为匿名函数的变量，例如: const func_name = function( \n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?function\\b'),\n # 匹配赋值为箭头函数的变量，例如: const func_name = () => 或者 const func_name = async a =>\n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][\\w$]*)\\s*=>')\n ]\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n func_names = []\n \n # 根据文件类型采取不同的正则提取函数名\n if ext == '.py':\n func_names.extend(py_func_pattern.findall(content))\n else:\n for pattern in js_func_patterns:\n func_names.extend(pattern.findall(content))\n \n # 遍历文件内提取出的所有函数名，验证其命名是否合规\n for func_name in func_names:\n has_digit = any(c.isdigit() for c in func_name)\n has_upper = any(c.isupper() for c in func_name)\n has_lower = any(c.islower() for c in func_name)\n has_under = '_' in func_name\n \n # 只要有任何一个函数名不能同时满足四种要素，就代表不符合要求\n if not (has_digit and has_upper and has_lower and has_under):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"注意代码格式，生成的代码文件中必须仅用 2 个空格实现缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n \n # 常见的前端与后端代码文件后缀\n code_exts = {'.js', '.html', '.css', '.vue', '.react', '.jsx', '.ts', '.tsx', '.py'}\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_code_files = True\n indent_diffs = []\n prev_indent = 0\n \n for line in lines:\n if not line.strip():\n continue\n \n # 获取该行的前导空白字符\n stripped = line.lstrip(' \\t')\n ws_len = len(line) - len(stripped)\n ws = line[:ws_len]\n \n # 严格禁止使用制表符（要求强调“仅用 2 个空格”）\n if '\\t' in ws:\n return False\n \n curr_indent = ws_len\n # 当且仅当发生缩进级别递增时，记录其缩进增量\n if curr_indent > prev_indent:\n # 排除常见注释（如块注释、HTML注释等）可能产生的非正常代码对齐干扰\n if not (stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('<!--')):\n indent_diffs.append(curr_indent - prev_indent)\n \n prev_indent = curr_indent\n \n if indent_diffs:\n counts = Counter(indent_diffs)\n max_count = max(counts.values())\n # 验证 2 个空格的缩进增量是否为最高频的主导缩进模式\n # 如果最高频增量不是 2（例如是 4），则说明违反了 2 空格缩进的规定\n if counts.get(2, 0) < max_count:\n return False\n \n # 如果遍历了所有代码文件且没有发现违反规则的情况，则认为满足要求\n return has_code_files","tags":["Style","Layout"]},{"constraint":"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径","validation_code":"","tags":["Environment","File Path"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我创建一个单页应用，包含两部分功能：书籍展示和井字棋游戏。首先，我们需要搭建基本的页面骨架。在页面顶部设置一个固定的水平导航栏，包含「我的书架」和「井字棋挑战」两个导航按钮。下方是主要的内容展示区。点击不同的导航按钮可以切换内容区的视图。目前视图里只需要分别居中显示文字「书籍内容建设中...」和「井字棋加载中...」作为占位即可。整体背景使用浅灰色，导航栏使用深色，UI风格尽量现代、简洁一点。另外后续所有操作请始终遵循以下要求：请注意，在你所生成的代码文件里，所有的文件名都必须同时包含数字、下划线和大小写字母；生成代码时，请保证文件中所有的函数名只包含小写字母与下划线。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对单页应用和常见代码文件定义后缀集合\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断当前文件是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 读取首行\\n first_line = f.readline()\\n \\n # 检查首行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 无法正常读取的异常情况，可按未满足规范处理\\n return False\\n \\n # 如果找到了至少一个代码文件，并且全满足要求，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 考虑到助手回复的最开头可能包含空格或换行符等空白字符，先使用 lstrip() 去除左侧空白字符\\n clean_response = response.lstrip()\\n # 检查清理后的回复内容是否以指定的语气词\\\"嗨～\\\"打头\\n return clean_response.startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"哇～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 去除回复内容首尾可能存在的空白字符（如换行符、空格等）\\n cleaned_response = response.strip()\\n \\n # 检查清理后的回复是否以\\\"哇～\\\"结尾\\n if cleaned_response.endswith(\\\"哇～\\\"):\\n return True\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件内，包含的注释全都是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请注意，在你所生成的代码文件里，所有的文件名都必须同时包含数字、下划线和大小写字母\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件扩展名，匹配单页应用及常规代码项目\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n \\n # 遍历当前项目地址\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录，防止检查git或系统隐藏文件夹\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 忽略隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅验证所生成的代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n \\n # 检查文件名是否同时包含数字、下划线、大写字母和小写字母\\n has_digit = any(c.isdigit() for c in file)\\n has_lower = any(c.islower() for c in file)\\n has_upper = any(c.isupper() for c in file)\\n has_underscore = '_' in file\\n \\n # 如果任意一个代码文件未能全部包含以上四种元素，则视为不满足要求\\n if not (has_digit and has_lower and has_upper and has_underscore):\\n return False\\n \\n # 需要确保工作区确实生成了代码文件，而不是空项目\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n # 工作区不存在则无法进行违规匹配，认为没有违规\\n return True\\n \\n # 定义正则表达式来匹配常见的函数定义形式\\n # 匹配常规的 function 声明\\n func_pattern = r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('\\n # 匹配变量赋值形式的函数 (包括普通匿名函数和箭头函数，兼顾可能存在的 async 修饰符)\\n var_pattern = r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\(?[a-zA-Z0-9_$,\\\\s]*\\\\)?\\\\s*=>)'\\n # 匹配 Python 函数定义\\n py_pattern = r'\\\\b(?:async\\\\s+)?def\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('\\n \\n # 针对前端单页应用和可能的后端API，推断可能的代码文件格式\\n allowed_extensions = {'.html', '.htm', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 跳过 node_modules 目录等第三方依赖目录\\n if 'node_modules' in root:\\n continue\\n \\n for file in files:\\n # 过滤掉压缩代码文件，避免提取到第三方的框架内置函数干扰检测\\n if file.endswith('.min.js'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext not in allowed_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取文件中所有能够被捕获的函数名称\\n func_names = []\\n func_names.extend(re.findall(func_pattern, content))\\n func_names.extend(re.findall(var_pattern, content))\\n func_names.extend(re.findall(py_pattern, content))\\n \\n # 验证每一个函数名称是否只包含小写字母与下划线\\n for name in func_names:\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"框架有了，现在我们来完善「我的书架」页面。请使用本地模拟数据生成大约8-10本图书（包含字段：书名、作者、封面图的占位色块或链接、分类标签、10分制评分）。使用CSS Grid网格布局把这些书籍以卡片的形式展示在页面上。每张卡片需要有白色的背景、圆角和轻微的投影效果。另外，当鼠标悬浮在某张书籍卡片上时，卡片要平滑地向上浮动几个像素，并且阴影加深，增加交互感。此外请留意，后续所有操作请始终遵循以下更新要求：第一，确保生成的代码文件名仅由小写字母和下划线构成；第二，生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断，单页应用可能生成的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n # 遍历工作目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 utf-8 编码读取文件第一行\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查首行是否包含要求的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法正常读取或打开失败，视为验证失败\\n return False\\n \\n # 如果没有找到任何代码文件，也视为不满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容左侧的空白字符后，判断是否以\\\"嗨～\\\"打头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"哇～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符\\n clean_resp = response.strip()\\n \\n # 使用正则表达式匹配文本末尾是否为“哇～”\\n # 允许“哇～”之后跟随常见的中英文标点符号以及空白字符（如：哇～、哇～！、哇～。 等情况）\\n pattern = r'哇～[。，！？\\\\.,!\\\\?\\\\s]*$'\\n \\n if re.search(pattern, clean_resp):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件内，包含的注释全都是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"确保生成的代码文件名仅由小写字母和下划线构成\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response, workspace_path):\\n # 如果工作区目录不存在，无法验证生成的代码\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义前端项目中常见的代码文件后缀\\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.json', '.vue', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 忽略以 . 开头的隐藏文件（例如 .gitignore）\\n if file.startswith('.'):\\n continue\\n \\n # 获取小写的文件扩展名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 筛选我们认为是“代码文件”的类型，或者是无后缀的可执行脚本\\n if ext in code_exts or not ext:\\n has_code_file = True\\n \\n # 按照 `.` 将文件名进行分割，并对每一部分进行检查\\n parts = file.split('.')\\n for part in parts:\\n # 检查是否每一部分都完全由小写字母和下划线构成\\n if not re.fullmatch(r'[a-z_]+', part):\\n return False\\n \\n # 只要存在至少一个代码文件并且没有在上面返回 False，说明完全满足条件\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请保证文件中所有的函数名只包含小写字母与下划线\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n \\n if not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 1. 针对 Python 文件的 AST 解析提取\\n if ext == '.py':\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n source = f.read()\\n tree = ast.parse(source)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 判断是否只包含小写字母和下划线\\n if not bool(re.match(r'^[a-z_]+$', node.name)):\\n return False\\n except Exception:\\n pass\\n \\n # 2. 针对前端 JS/TS/HTML 的正则表达式提取\\n elif ext in {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue'}:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 若为 HTML / Vue 等混合类型，提取其中 script 标签的内容\\n if ext in {'.html', '.htm', '.vue'}:\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n text_to_check = '\\\\n'.join(scripts)\\n else:\\n text_to_check = content\\n \\n # 暴力去除多行和单行注释，防止提取到注释中的无效内容\\n text_to_check = re.sub(r'/\\\\*.*?\\\\*/', '', text_to_check, flags=re.DOTALL)\\n text_to_check = re.sub(r'//.*', '', text_to_check)\\n \\n funcs = []\\n # (1) 匹配 function foo()\\n funcs.extend(re.findall(r'\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(', text_to_check))\\n # (2) 匹配 const/let/var foo = function / () => / a => \\n funcs.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z0-9_]+\\\\s*=>)', text_to_check))\\n # (3) 匹配 foo: function / () => / a => (对象属性方法)\\n funcs.extend(re.findall(r'\\\\b([a-zA-Z0-9_]+)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z0-9_]+\\\\s*=>)', text_to_check))\\n \\n # 校验该文件内提取到的所有函数名\\n for name in funcs:\\n if not bool(re.match(r'^[a-z_]+$', name)):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来给「我的书架」增加实用的筛选和排序功能。在书籍网格区域的上方，添加一个水平排列的操作栏。操作栏里需要有三个组件：1. 一个搜索输入框，可以实时根据书名或作者进行模糊匹配；2. 一个下拉菜单，提取所有数据的分类（如科幻、文学、科技等）进行分类筛选；3. 一个「评分排序」按钮，点击可以在“评分从高到低”和“评分从低到高”之间切换。所有的筛选、搜索和排序操作必须能互相叠加，并实时更新下方的书籍卡片列表。另外请留意，后续所有操作请始终遵循以下更新要求：第一，在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母；第二，为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 30s` 包装。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_code_files_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 涵盖前端单页应用以及常见后台/脚本语言的代码扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.sh', \\n '.rb', '.php', '.go', '.rs', '.swift', '.kt'\\n }\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件进行检查\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline().strip()\\n # 检查首行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于编码异常或其他原因无法读取首行，视为不符合要求\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的回复是否以\\\"嗨～\\\"打头。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除首部的空白字符后，判断是否以\\\"嗨～\\\"开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"哇～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 去除回复末尾可能存在的空白字符（包括空格、换行符等）\\n clean_response = response.rstrip()\\n \\n # 判断清理后的回复内容是否以\\\"哇～\\\"结尾\\n if clean_response.endswith(\\\"哇～\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件内，包含的注释全都是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"确保生成的代码文件名仅由小写字母和下划线构成\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 推断单页应用相关的常见代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.vue', '.json'\\n }\\n \\n # 遍历当前项目的工作区文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和常见的依赖包目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n # 提取文件扩展名并转为小写，以便做文件类型匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为推断出的代码文件类型\\n if ext in code_extensions:\\n # 剔除作为扩展名分隔符的 '.' 后，考察剩余所有字符\\n name_without_dots = file.replace('.', '')\\n \\n # 如果除点以外的字符并不只由小写字母和下划线构成，则验证失败\\n if not re.match(r'^[a-z_]+$', name_without_dots):\\n return False\\n \\n # 如果所有的代码文件名都符合要求，或者没有生成代码文件，则判定为 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def check_name_rules(name: str) -> bool:\\n \\\"\\\"\\\"检查函数名是否同时包含：数字、下划线、大写字母和小写字母\\\"\\\"\\\"\\n has_digit = any(c.isdigit() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_lower = any(c.islower() for c in name)\\n has_underscore = '_' in name\\n return has_digit and has_upper and has_lower and has_underscore\\n\\n def extract_js_functions(code: str) -> list:\\n \\\"\\\"\\\"从 JS/TS/HTML 类的前端源码中利用正则提取函数名\\\"\\\"\\\"\\n # 清除注释内容以免干扰正则匹配\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n \\n functions = []\\n # 1. 匹配普通函数声明，例如 function myFunc_1A()\\n p1 = re.findall(r'\\\\bfunction\\\\s+([A-Za-z0-9_]+)\\\\s*\\\\(', code)\\n \\n # 2. 匹配变量赋值形式的函数，例如 const myFunc_1A = () => {} 或 let func = function()\\n p2 = re.findall(r'\\\\b(?:const|let|var)\\\\s+([A-Za-z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[A-Za-z0-9_]+\\\\s*=>)', code)\\n \\n # 3. 匹配对象内方法赋值，例如 myFunc_1A: function() {} 或 myFunc_1A: () => {}\\n p3 = re.findall(r'\\\\b([A-Za-z0-9_]+)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[A-Za-z0-9_]+\\\\s*=>)', code)\\n \\n functions.extend(p1 + p2 + p3)\\n \\n # 排除可能是因为正则误捕获的编程关键字\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else', 'await', 'async', 'yield'}\\n return [f for f in functions if f not in keywords and f]\\n\\n def extract_python_functions(code: str) -> list:\\n \\\"\\\"\\\"从 Python 源码中利用 AST 提取函数名\\\"\\\"\\\"\\n functions = []\\n try:\\n import ast\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n functions.append(node.name)\\n except Exception:\\n pass\\n return functions\\n \\n # 验证工作空间是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 根据常见的代码文件后缀判断是否需要读取\\n if not file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.py')):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n funcs = []\\n if file.endswith('.py'):\\n funcs = extract_python_functions(content)\\n else:\\n funcs = extract_js_functions(content)\\n \\n # 校验文件中提取出来的所有函数名称\\n for func in funcs:\\n if not check_name_rules(func):\\n # 只要发现存在任意一个不符合规则的函数名，就立刻判定失败\\n return False\\n \\n # 如果遍历完毕，全部检查的函数名称均符合要求或无函数，则返回True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 30s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"书架功能做得很棒了。现在我们切换到「井字棋挑战」页面，来实现核心游戏逻辑。在页面中央绘制一个经典的 3x3 井字棋盘，格子边界要清晰。支持两名玩家轮流点击空白格子落子，分别用蓝色的 'X' 和红色的 'O' 表示，文字要足够大且居中。每次落子后，程序需要自动判断是否有玩家获胜（横、竖、对角线连成三子）或是棋盘已满导致平局。如果游戏结束，在棋盘上方显示醒目的文字提示结果（如「玩家 X 获胜！」），并同时展示一个「再来一局」的按钮，点击后清空棋盘重新开始。另外请留意，后续所有操作请始终遵循以下更新要求：第一，为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 60s` 包装；第二，注意代码格式，生成的代码文件中必须仅用 4 个空格实现缩进。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义本次任务中单页应用及常规前后端可能涉及的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.py', '.java', '.cpp', '.c', '.go', '.php', '.rb', '.sh'\\n }\\n \\n # 定义需要排除的无关目录，避免检查第三方库或构建产物\\n exclude_dirs = {\\n '.git', 'node_modules', 'venv', 'env', '__pycache__', \\n 'dist', 'build', '.idea', '.vscode'\\n }\\n \\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉排除的目录，不再向下遍历\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查首行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line.strip():\\n return False\\n except Exception:\\n # 如果由于编码等原因无法正常读取文件第一行，视为不满足要求\\n return False\\n \\n # 至少需要发现一个代码文件才算完成了生成代码的过程\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否以\\\"嗨～\\\"打头。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除开头的空白字符后，判断是否以\\\"嗨～\\\"开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"哇～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除末尾的空白字符\\n response = response.strip()\\n \\n # 考虑到助手可能会在结尾添加标点符号，去除末尾可能出现的常见中英文标点\\n response = re.sub(r'[。！？\\\\!\\\\.\\\\?]+$', '', response).strip()\\n \\n # 判断是否以“哇～”结尾\\n return response.endswith(\\\"哇～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件内，包含的注释全都是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"确保生成的代码文件名仅由小写字母和下划线构成\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码文件后缀集合，涵盖单页应用及相关可能用到的代码格式\\n code_exts = {\\n '.html', '.htm', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.vue', '.json', '.py', '.sh'\\n }\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git, .idea, .vscode 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n # 提取文件扩展名并统一转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否属于代码文件\\n if ext in code_exts:\\n has_code_file = True\\n # 获取文件的基础名称（不包含扩展名部分）\\n base_name = os.path.splitext(file)[0]\\n \\n # 检查文件名（不含扩展名）是否仅由小写字母和下划线构成\\n if not re.fullmatch(r'[a-z_]+', base_name):\\n return False\\n \\n # 如果没有找到任何代码文件，则无法证明满足了要求，返回False；否则说明都符合，返回True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义用于匹配前端 JavaScript/TypeScript/HTML 中函数定义的正则表达式\\n js_patterns = [\\n # 匹配普通函数声明: function myFunc (...)\\n r'\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(',\\n # 匹配变量赋值型函数（包括箭头函数与匿名函数）: const myFunc = (...) => 或者 const myFunc = function\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z0-9_]+\\\\s*=>|function\\\\b)'\\n ]\\n # 定义 Python 函数的正则表达式作为 fallback\\n py_patterns = [\\n r'\\\\bdef\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\('\\n ]\\n \\n def remove_js_comments(code: str) -> str:\\n \\\"\\\"\\\"移除JS/TS/前端代码中的单行和多行注释\\\"\\\"\\\"\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n code = re.sub(r'//.*', '', code)\\n return code\\n \\n def remove_py_comments(code: str) -> str:\\n \\\"\\\"\\\"移除Python代码中的单行注释\\\"\\\"\\\"\\n code = re.sub(r'#.*', '', code)\\n return code\\n\\n def check_name(name: str) -> bool:\\n \\\"\\\"\\\"检查函数名是否同时具有数字、下划线、大写字母和小写字母\\\"\\\"\\\"\\n has_digit = any(c.isdigit() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_lower = any(c.islower() for c in name)\\n has_under = '_' in name\\n return has_digit and has_upper and has_lower and has_under\\n\\n # 需要跳过的无需检查的目录名称（例如包依赖或者构建产物）\\n exclude_dirs = {'node_modules', '.git', 'venv', '.venv', 'dist', 'build'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉不需要遍历的文件夹\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 验证前端、全栈常见语言文件\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.vue']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n content = remove_js_comments(content)\\n \\n for pattern in js_patterns:\\n for match in re.finditer(pattern, content):\\n func_name = match.group(1)\\n # 忽略某些在复杂赋值情况下被错误捕捉的保留字\\n if func_name in {'default', 'async', 'return', 'get', 'set'}:\\n continue\\n if not check_name(func_name):\\n return False\\n except Exception:\\n # 遇到文件读取或编码问题时跳过该文件继续\\n continue\\n \\n # 验证 Python 文件\\n elif ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n import ast\\n tree = ast.parse(content)\\n # 优先利用 AST 安全准确地遍历 Python 文件中定义的方法和函数\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not check_name(node.name):\\n return False\\n except Exception:\\n # 如果 ast 解析失败（如语法错误），回退至正则匹配\\n try:\\n content = remove_py_comments(content)\\n for pattern in py_patterns:\\n for match in re.finditer(pattern, content):\\n func_name = match.group(1)\\n if not check_name(func_name):\\n return False\\n except Exception:\\n continue\\n\\n # 未找到任何违反命名规则的函数，要求得到满足\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"注意代码格式，生成的代码文件中必须仅用 4 个空格实现缩进\",\n \"validation_code\": \"import os\\n\\ndef check_indentation_is_four_spaces(response: str, workspace_path: str) -> bool:\\n # 针对单页应用上下文，圈定可能产生的代码文件扩展名\\n code_exts = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\\n \\n total_valid = 0\\n total_invalid = 0\\n has_tabs = False\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n stripped = line.lstrip()\\n \\n # 忽略空行\\n if not stripped:\\n continue\\n \\n # 忽略部分可能导致干扰的注释行\\n if stripped.startswith(('*', '//', '#', '<!--', '/*')):\\n continue\\n \\n # 截取前导空白字符\\n leading_whitespace = line[:len(line) - len(stripped)]\\n \\n # 如果前导缩进中使用了制表符(Tab)，直接记为不通过\\n if '\\\\t' in leading_whitespace:\\n has_tabs = True\\n break\\n \\n # 统计缩进层级分布情况\\n l = len(leading_whitespace)\\n if l > 0:\\n if l % 4 == 0:\\n total_valid += 1\\n else:\\n total_invalid += 1\\n \\n if has_tabs:\\n break\\n if has_tabs:\\n break\\n \\n # 未生成代码文件\\n if not has_code_files:\\n return False\\n \\n # 包含了制表符缩进\\n if has_tabs:\\n return False\\n \\n # 完全没有4的倍数缩进，却存在其他不规则空格缩进\\n if total_invalid > 0 and total_valid == 0:\\n return False\\n \\n # 容错判断：不合规的缩进行数超过合规行数的 20%，并且绝对值大于 3 行，\\n # 即可确定其实质上混用了缩进或根本未使用 4 空格作为缩进标准。\\n if total_invalid > max(3, total_valid * 0.2):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，给这个井字棋游戏增加一个有点难度的功能：时光倒流（历史步骤回溯）。在棋盘的右侧新增一个「历史对战记录」的列表面板。每当玩家成功落子后，列表中就要新增一个按钮（例如「回到第 1 步」、「回到第 2 步」等，最开始还有个「回到游戏起点」）。玩家点击列表中的任意历史记录，棋盘状态必须立刻回退到那一步的样子。注意：如果玩家在回退后的状态下重新落子了一步新棋，那么原先那一步之后的历史记录必须被全部清除。同时在顶部加一个当前轮到谁下棋的动态提示器。\\n另外请留意，后续所有操作请始终遵循以下更新要求：第一，注意代码格式，生成的代码文件中必须仅用 2 个空格实现缩进；第二，代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"优先且尽可能地使用 Task 工具（确保 subagent_type=Explore）来进行查找处理文件，而不是采取直接运行文件操作或搜索命令的手段\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要直接使用 Grep 或 Glob 工具；相反，你必须尽可能充分地使用 Task 工具来探索代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名，涵盖前端单页应用及其他常见语言\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs',\\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.sh', '.bash'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取或解码出错等，视为不满足要求\\n return False\\n\\n # 只有在至少遍历到一个代码文件且全都符合时返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否以特定的语气词\\\"嗨～\\\"开头。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，表示是否满足要求。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容可能存在的前导空白字符，然后判断是否以要求的内容打头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"哇～\\\"语气词\",\n \"validation_code\": \"def check_response_ending(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除回复首尾可能存在的空白字符、换行符后，判断是否以\\\"哇～\\\"结尾\\n return response.strip().endswith(\\\"哇～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请保证生成的每个代码文件内，包含的注释全都是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"当生成代码的测试环节每次完成后，回复内容必须严格以该XML格式分项展示执行结果及其描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"确保生成的代码文件名仅由小写字母和下划线构成\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件后缀类型集合（重点关注前端开发及常规语言）\\n code_extensions = {\\n '.py', '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.java', '.cpp', '.c', '.h', '.hpp', '.cs', '.php', '.rb', \\n '.go', '.rs', '.swift', '.kt', '.sh'\\n }\\n \\n has_target_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录及常见的依赖包或缓存目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验\\n if ext in code_extensions:\\n has_target_files = True\\n name = os.path.splitext(file)[0]\\n \\n # 检查文件名（不含扩展名部分）是否仅由小写字母和下划线构成\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n # 当找到了对应的代码文件并且没有任何文件名违规时，返回 True\\n return has_target_files\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件中，不论哪个函数，其名称都必须同时具有数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义通常可能包含函数编写的代码文件后缀名\\n valid_extensions = {'.js', '.py', '.html', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n # 编译匹配各种常见函数定义的正则表达式\\n # 1. Python 中的函数声明，例如: def func_name( 或者 async def func_name(\\n py_func_pattern = re.compile(r'\\\\b(?:async\\\\s+)?def\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n \\n # 2. JavaScript / TypeScript 中的各种函数声明\\n js_func_patterns = [\\n # 匹配直接声明的函数，例如: function func_name( 或者 async function func_name(\\n re.compile(r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\('),\\n # 匹配赋值为匿名函数的变量，例如: const func_name = function( \\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\b'),\\n # 匹配赋值为箭头函数的变量，例如: const func_name = () => 或者 const func_name = async a =>\\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][\\\\w$]*)\\\\s*=>')\\n ]\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n func_names = []\\n \\n # 根据文件类型采取不同的正则提取函数名\\n if ext == '.py':\\n func_names.extend(py_func_pattern.findall(content))\\n else:\\n for pattern in js_func_patterns:\\n func_names.extend(pattern.findall(content))\\n \\n # 遍历文件内提取出的所有函数名，验证其命名是否合规\\n for func_name in func_names:\\n has_digit = any(c.isdigit() for c in func_name)\\n has_upper = any(c.isupper() for c in func_name)\\n has_lower = any(c.islower() for c in func_name)\\n has_under = '_' in func_name\\n \\n # 只要有任何一个函数名不能同时满足四种要素，就代表不符合要求\\n if not (has_digit and has_upper and has_lower and has_under):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成的代码必须排除过度设计，以最短路径实现功能，严禁自行添加需求规定以外的任何机制，包括兜底和降级方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"为了防止进程挂起，在执行所有 bash 测试、安装或构建命令时，请务必在命令开头加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"注意代码格式，生成的代码文件中必须仅用 2 个空格实现缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n \\n # 常见的前端与后端代码文件后缀\\n code_exts = {'.js', '.html', '.css', '.vue', '.react', '.jsx', '.ts', '.tsx', '.py'}\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_code_files = True\\n indent_diffs = []\\n prev_indent = 0\\n \\n for line in lines:\\n if not line.strip():\\n continue\\n \\n # 获取该行的前导空白字符\\n stripped = line.lstrip(' \\\\t')\\n ws_len = len(line) - len(stripped)\\n ws = line[:ws_len]\\n \\n # 严格禁止使用制表符（要求强调“仅用 2 个空格”）\\n if '\\\\t' in ws:\\n return False\\n \\n curr_indent = ws_len\\n # 当且仅当发生缩进级别递增时，记录其缩进增量\\n if curr_indent > prev_indent:\\n # 排除常见注释（如块注释、HTML注释等）可能产生的非正常代码对齐干扰\\n if not (stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('<!--')):\\n indent_diffs.append(curr_indent - prev_indent)\\n \\n prev_indent = curr_indent\\n \\n if indent_diffs:\\n counts = Counter(indent_diffs)\\n max_count = max(counts.values())\\n # 验证 2 个空格的缩进增量是否为最高频的主导缩进模式\\n # 如果最高频增量不是 2（例如是 4），则说明违反了 2 空格缩进的规定\\n if counts.get(2, 0) < max_count:\\n return False\\n \\n # 如果遍历了所有代码文件且没有发现违反规则的情况，则认为满足要求\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["打开页面，检查整体布局与导航栏：查看页面顶部是否有一个固定的水平导航栏，导航栏包含「我的书架」和「井字棋挑战」两个按钮。检查整体页面背景是否为浅灰色，导航栏是否为深色。","点击导航栏的「我的书架」和「井字棋挑战」按钮，验证下方的内容展示区是否能在书架视图与井字棋视图之间正确切换。","切换到「我的书架」页面，检查书籍展示：确认页面包含多本书籍卡片，并以网格(Grid)形式排列。检查每张卡片是否都有白色背景、圆角和轻微投影，且卡片内包含书名、作者、封面图(或占位色块)、分类标签和10分制评分信息。","在「我的书架」页面，将鼠标悬停在任意一本图书卡片上，观察交互动效：验证该卡片是否会平滑地向上浮动几个像素，并且阴影效果加深。","在「我的书架」页面上方找到搜索输入框，输入某本书籍名称或作者名称的片段：验证下方的书籍列表是否实时更新，仅展示书名或作者与输入匹配的书籍卡片。","在「我的书架」页面上方找到分类下拉菜单，点击并选择某一个特定的分类（如科幻）：验证下方的书籍列表是否实时更新，仅展示包含该分类标签的书籍卡片。","在「我的书架」页面上方找到「评分排序」按钮，点击该按钮：验证书籍列表是否按评分从高到低排列；再次点击该按钮，验证列表是否切换为按评分从低到高排列。","在「我的书架」页面进行组合条件操作：在搜索框中输入文字，同时在下拉菜单选择一个分类，并点击一次「评分排序」按钮。验证下方的列表是否严格且同时满足搜索匹配、分类匹配，并正确按照指定的评分顺序进行了排列。","切换到「井字棋挑战」页面，检查基础页面元素：验证页面中央是否渲染了一个带有清晰边界的 3x3 网格棋盘；棋盘顶部是否存在显示“当前轮到谁下棋”的动态提示器；棋盘右侧是否有一个「历史对战记录」列表且初始包含「回到游戏起点」按钮。","在 3x3 棋盘上交替点击空白格子：验证格子内是否依次显示蓝色的字母 'X' 和红色的字母 'O'，字母要求字体足够大且居中；同时检查顶部的提示器是否随每次落子正确切换当前玩家。","模拟一场非平局游戏，使得玩家 X（或 O）在横排、竖排或对角线上连成三个相同的棋子：验证棋盘上方是否立即出现醒目的获胜文字提示（如「玩家 X 获胜！」），并展示一个「再来一局」的按钮。","模拟一场平局游戏，将9个格子全部填满且不出现任何连成三子的情况：验证棋盘上方是否出现平局的文字提示，并展示「再来一局」按钮。","在一场对局结束（胜利或平局）后，点击「再来一局」按钮：验证棋盘格内的 'X' 和 'O' 是否全部被清空，并且允许玩家重新开始新一轮的落子操作。","测试历史对战记录的生成和回溯功能：在空白棋盘上交替落子 3 次，确认右侧列表中依次新增了 3 条历史记录按钮（如回到第1、2、3步）。点击「回到第 1 步」按钮，验证棋盘内的落子状态是否立刻回退到只有 1 颗棋子的模样。","测试历史步骤回溯后的分支清除逻辑：在进行了数次落子后，点击右侧历史列表「回到第 1 步」。随后直接在棋盘上的任意空白处落下一颗新棋子，验证之前生成的第2步及之后的历史记录按钮是否全部消失被清除，并在当前「回到第 1 步」按钮下方生成一条全新的「回到第 2 步」按钮。","在「井字棋挑战」页面进行若干步落子后，点击右侧历史列表中的「回到游戏起点」按钮：验证棋盘是否立刻回退到无任何落子的初始空盘状态。"],"string":"[\n \"打开页面，检查整体布局与导航栏：查看页面顶部是否有一个固定的水平导航栏，导航栏包含「我的书架」和「井字棋挑战」两个按钮。检查整体页面背景是否为浅灰色，导航栏是否为深色。\",\n \"点击导航栏的「我的书架」和「井字棋挑战」按钮，验证下方的内容展示区是否能在书架视图与井字棋视图之间正确切换。\",\n \"切换到「我的书架」页面，检查书籍展示：确认页面包含多本书籍卡片，并以网格(Grid)形式排列。检查每张卡片是否都有白色背景、圆角和轻微投影，且卡片内包含书名、作者、封面图(或占位色块)、分类标签和10分制评分信息。\",\n \"在「我的书架」页面，将鼠标悬停在任意一本图书卡片上，观察交互动效：验证该卡片是否会平滑地向上浮动几个像素，并且阴影效果加深。\",\n \"在「我的书架」页面上方找到搜索输入框，输入某本书籍名称或作者名称的片段：验证下方的书籍列表是否实时更新，仅展示书名或作者与输入匹配的书籍卡片。\",\n \"在「我的书架」页面上方找到分类下拉菜单，点击并选择某一个特定的分类（如科幻）：验证下方的书籍列表是否实时更新，仅展示包含该分类标签的书籍卡片。\",\n \"在「我的书架」页面上方找到「评分排序」按钮，点击该按钮：验证书籍列表是否按评分从高到低排列；再次点击该按钮，验证列表是否切换为按评分从低到高排列。\",\n \"在「我的书架」页面进行组合条件操作：在搜索框中输入文字，同时在下拉菜单选择一个分类，并点击一次「评分排序」按钮。验证下方的列表是否严格且同时满足搜索匹配、分类匹配，并正确按照指定的评分顺序进行了排列。\",\n \"切换到「井字棋挑战」页面，检查基础页面元素：验证页面中央是否渲染了一个带有清晰边界的 3x3 网格棋盘；棋盘顶部是否存在显示“当前轮到谁下棋”的动态提示器；棋盘右侧是否有一个「历史对战记录」列表且初始包含「回到游戏起点」按钮。\",\n \"在 3x3 棋盘上交替点击空白格子：验证格子内是否依次显示蓝色的字母 'X' 和红色的字母 'O'，字母要求字体足够大且居中；同时检查顶部的提示器是否随每次落子正确切换当前玩家。\",\n \"模拟一场非平局游戏，使得玩家 X（或 O）在横排、竖排或对角线上连成三个相同的棋子：验证棋盘上方是否立即出现醒目的获胜文字提示（如「玩家 X 获胜！」），并展示一个「再来一局」的按钮。\",\n \"模拟一场平局游戏，将9个格子全部填满且不出现任何连成三子的情况：验证棋盘上方是否出现平局的文字提示，并展示「再来一局」按钮。\",\n \"在一场对局结束（胜利或平局）后，点击「再来一局」按钮：验证棋盘格内的 'X' 和 'O' 是否全部被清空，并且允许玩家重新开始新一轮的落子操作。\",\n \"测试历史对战记录的生成和回溯功能：在空白棋盘上交替落子 3 次，确认右侧列表中依次新增了 3 条历史记录按钮（如回到第1、2、3步）。点击「回到第 1 步」按钮，验证棋盘内的落子状态是否立刻回退到只有 1 颗棋子的模样。\",\n \"测试历史步骤回溯后的分支清除逻辑：在进行了数次落子后，点击右侧历史列表「回到第 1 步」。随后直接在棋盘上的任意空白处落下一颗新棋子，验证之前生成的第2步及之后的历史记录按钮是否全部消失被清除，并在当前「回到第 1 步」按钮下方生成一条全新的「回到第 2 步」按钮。\",\n \"在「井字棋挑战」页面进行若干步落子后，点击右侧历史列表中的「回到游戏起点」按钮：验证棋盘是否立刻回退到无任何落子的初始空盘状态。\"\n]","isConversational":false}}},{"rowIdx":62,"cells":{"task_id":{"kind":"number","value":63,"string":"63"},"repository_policy":{"kind":"string","value":"你后续所有的操作、回复以及代码生成都必须始终遵循以下约束：\n1. 你回复用户的生成内容必须始终保持为中文。\n2. 请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进。\n3. 每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案。\n4. 在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作。\n5. 在你生成的所有代码文件中，注释内容必须统一使用中文编写。\n6. 你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数。\n7. 要求生成的代码中，“=”符号的两侧均需添加两个空格。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你回复用户的生成内容必须始终保持为中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n total_indents = 0\n valid_indents = 0\n\n def process_lines(lines):\n nonlocal total_indents, valid_indents\n for line in lines:\n # 移除行末的换行和空白符\n stripped_right = line.rstrip('\\r\\n')\n # 忽略纯空行\n if not stripped_right.strip():\n continue\n \n # 匹配行首前导空白\n match = re.match(r'^([ \\t]+)', stripped_right)\n if match:\n leading_whitespace = match.group(1)\n \n # 若包含制表符，说明未全部用空格，直接判违规\n if '\\t' in leading_whitespace:\n return False\n \n # 获取缩进后的非空白文本部分\n content_after_space = stripped_right[len(leading_whitespace):]\n \n # 排除带 * 星号开头的行（常见于多行注释内对齐，其空格缩进往往不是4的倍数）\n if content_after_space.startswith('*'):\n continue\n \n length = len(leading_whitespace)\n if length > 0:\n total_indents += 1\n # 检查缩进数量是否为 4 的倍数\n if length % 4 == 0:\n valid_indents += 1\n return True\n\n # 1. 遍历并检查 workspace_path 目录下生成的代码文件\n if os.path.isdir(workspace_path):\n # 常见代码与配置文件的扩展名集合\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', '.go', \n '.rb', '.php', '.js', '.ts', '.html', '.css', '.json', \n '.yaml', '.yml', '.sh', '.bash', '.xml', '.sql', '.swift', '.kt',\n '.dart', '.rs', '.m', '.mm'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n if not process_lines(lines):\n return False\n\n # 2. 检查 response 文本中以 Markdown 语法给出的代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9_-]*\\s*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n lines = block.split('\\n')\n if not process_lines(lines):\n return False\n\n # 3. 统计判定（容忍极少量的长语句折行对齐引起的非 4 倍数空格，设定阈值为 80%）\n if total_indents > 0:\n if valid_indents / total_indents < 0.8:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 涵盖常见的代码文件扩展名，也包含了指令中提到的 .json 和 .md\n code_extensions = {\n '.py', '.js', '.ts', '.html', '.css', '.java', '.c', '.cpp', \n '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh', \n '.json', '.xml', '.yml', '.yaml', '.md', '.sql', '.vue', '.jsx', '.tsx'\n }\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n \n # 检查总行数（包括空行）是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果因权限或特殊编码等异常无法读取文件，则保守返回 False\n return False\n\n # 如果有至少一个代码文件且全部满足偶数行条件，则返回 True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"要求生成的代码中，“=”符号的两侧均需添加两个空格","validation_code":"def verify_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见的代码文件扩展名集合\n code_exts = {\n '.py', '.js', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.ts', '.jsx', '.tsx', '.php', '.rb', '.swift', \n '.kt', '.m', '.scala', '.sh'\n }\n \n # 匹配单独的等号，通过断言排除 ==, !=, >=, <=, +=, -= 等复合赋值运算符以及箭头函数 =>\n pattern = re.compile(r'(?<![=!<>+\\-*/%&|^~])=(?![=>])')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 遍历文件中所有单独的 '='\n for match in pattern.finditer(content):\n idx = match.start()\n \n # 1. 检查 '=' 左侧是否恰好有两个空格\n # 左侧至少需要有两个字符，且这两个字符必须都是空格\n if idx < 2 or content[idx-2:idx] != ' ':\n return False\n # 左侧第三个字符（如果存在）不能是空格，否则说明多于两个空格\n if idx >= 3 and content[idx-3] == ' ':\n return False\n \n # 2. 检查 '=' 右侧是否恰好有两个空格\n # 右侧至少需要有两个字符，且这两个字符必须都是空格\n if idx + 3 > len(content) or content[idx+1:idx+3] != ' ':\n return False\n # 右侧第三个字符（如果存在）不能是空格，否则说明多于两个空格\n if idx + 4 <= len(content) and content[idx+3] == ' ':\n return False\n \n return True","tags":["Style","Layout"]}],"string":"[\n {\n \"constraint\": \"你回复用户的生成内容必须始终保持为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n total_indents = 0\\n valid_indents = 0\\n\\n def process_lines(lines):\\n nonlocal total_indents, valid_indents\\n for line in lines:\\n # 移除行末的换行和空白符\\n stripped_right = line.rstrip('\\\\r\\\\n')\\n # 忽略纯空行\\n if not stripped_right.strip():\\n continue\\n \\n # 匹配行首前导空白\\n match = re.match(r'^([ \\\\t]+)', stripped_right)\\n if match:\\n leading_whitespace = match.group(1)\\n \\n # 若包含制表符，说明未全部用空格，直接判违规\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 获取缩进后的非空白文本部分\\n content_after_space = stripped_right[len(leading_whitespace):]\\n \\n # 排除带 * 星号开头的行（常见于多行注释内对齐，其空格缩进往往不是4的倍数）\\n if content_after_space.startswith('*'):\\n continue\\n \\n length = len(leading_whitespace)\\n if length > 0:\\n total_indents += 1\\n # 检查缩进数量是否为 4 的倍数\\n if length % 4 == 0:\\n valid_indents += 1\\n return True\\n\\n # 1. 遍历并检查 workspace_path 目录下生成的代码文件\\n if os.path.isdir(workspace_path):\\n # 常见代码与配置文件的扩展名集合\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', '.go', \\n '.rb', '.php', '.js', '.ts', '.html', '.css', '.json', \\n '.yaml', '.yml', '.sh', '.bash', '.xml', '.sql', '.swift', '.kt',\\n '.dart', '.rs', '.m', '.mm'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n if not process_lines(lines):\\n return False\\n\\n # 2. 检查 response 文本中以 Markdown 语法给出的代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9_-]*\\\\s*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n if not process_lines(lines):\\n return False\\n\\n # 3. 统计判定（容忍极少量的长语句折行对齐引起的非 4 倍数空格，设定阈值为 80%）\\n if total_indents > 0:\\n if valid_indents / total_indents < 0.8:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 涵盖常见的代码文件扩展名，也包含了指令中提到的 .json 和 .md\\n code_extensions = {\\n '.py', '.js', '.ts', '.html', '.css', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh', \\n '.json', '.xml', '.yml', '.yaml', '.md', '.sql', '.vue', '.jsx', '.tsx'\\n }\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n \\n # 检查总行数（包括空行）是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果因权限或特殊编码等异常无法读取文件，则保守返回 False\\n return False\\n\\n # 如果有至少一个代码文件且全部满足偶数行条件，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"要求生成的代码中，“=”符号的两侧均需添加两个空格\",\n \"validation_code\": \"def verify_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见的代码文件扩展名集合\\n code_exts = {\\n '.py', '.js', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.ts', '.jsx', '.tsx', '.php', '.rb', '.swift', \\n '.kt', '.m', '.scala', '.sh'\\n }\\n \\n # 匹配单独的等号，通过断言排除 ==, !=, >=, <=, +=, -= 等复合赋值运算符以及箭头函数 =>\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%&|^~])=(?![=>])')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 遍历文件中所有单独的 '='\\n for match in pattern.finditer(content):\\n idx = match.start()\\n \\n # 1. 检查 '=' 左侧是否恰好有两个空格\\n # 左侧至少需要有两个字符，且这两个字符必须都是空格\\n if idx < 2 or content[idx-2:idx] != ' ':\\n return False\\n # 左侧第三个字符（如果存在）不能是空格，否则说明多于两个空格\\n if idx >= 3 and content[idx-3] == ' ':\\n return False\\n \\n # 2. 检查 '=' 右侧是否恰好有两个空格\\n # 右侧至少需要有两个字符，且这两个字符必须都是空格\\n if idx + 3 > len(content) or content[idx+1:idx+3] != ' ':\\n return False\\n # 右侧第三个字符（如果存在）不能是空格，否则说明多于两个空格\\n if idx + 4 <= len(content) and content[idx+3] == ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用 HTML5 Canvas 和原生 JavaScript 创建一个滑动迷宫游戏的基础骨架。需求如下：\n1. 页面居中放置一个合适大小的 Canvas 容器，背景设为浅灰色。\n2. 定义一个简单的二维数组地图数据结构（如 10x10 网格，0 代表道路，1 代表墙壁），并将其绘制到 Canvas 上。使用纯色方块即可，比如黑色表示墙壁，白色表示道路。\n3. 在起点空地上绘制一个玩家（例如一个蓝色方块）。\n4. 实现最基础的“滑动”移动交互：玩家通过键盘方向键（上下左右）控制，一旦按下某个方向，玩家方块就会沿着该方向匀速持续滑动，直到撞到墙壁才停下来。滑动过程中不能改变方向。第一轮先不涉及过关和终点，只要把地图绘制和核心的滑动撞墙判定写好即可。\n\n另外请你在后续所有操作中始终遵循以下要求：\n1. 执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录。\n2. 必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率。","instruction_following_checklist":[{"constraint":"你回复用户的生成内容必须始终保持为中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义需要检查缩进的文件后缀\n code_extensions = {'.html', '.js', '.css', '.json'}\n has_checked_any = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_checked_any = True\n \n for line in lines:\n # 去除右侧所有空白字符，借此跳过全空白的行\n if not line.rstrip():\n continue\n \n # 匹配当前行首的前导空白字符（包含空格或制表符）\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_whitespace = match.group(1)\n # content 为去除前导空白后的正文内容\n content = line[len(leading_whitespace):]\n \n # 规则1: 绝对不允许使用制表符 \\t\n if '\\t' in leading_whitespace:\n return False\n \n # 规则2: 使用正好 4 个空格完成每一处缩进，因此空格数必须是 4 的倍数\n if len(leading_whitespace) % 4 != 0:\n # 豁免多行注释中间对齐的星号 (例如 ' * 注释')\n if content.startswith('*'):\n continue\n return False\n \n # 如果项目中存在代码文件并且全部通过了上述的验证，则判定为成功\n return has_checked_any","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能的代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.json'}\n has_code_file = False\n\n # 检查工作区路径是否有效\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符切分以统计包含空行在内的所有行数\n # 例如，内容为 \"a\\nb\" 会被切分为 ['a', 'b']，总行数为 2\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果遇到文件不可读或编码异常，安全起见判定为不满足条件\n return False\n\n # 确保项目中确实包含生成的代码文件并且全部满足了偶数行要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"要求生成的代码中，“=”符号的两侧均需添加两个空格","validation_code":"def check_equals_spaces(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见的项目代码文件类型\n exts = ['.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue']\n has_code_files = False\n \n # 正则表达式：捕获单独的或连续的 = 号块（即 =, ==, ===）\n # 排除复合运算符的情况，如左侧不为 + - * / % < > ! =，右侧不为 = 或 > (防止匹配箭头函数 =>)\n pattern = re.compile(r'(?<![+\\-*/%<>!=])(=+)(?![=>])')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not any(file.endswith(ext) for ext in exts):\n continue\n has_code_files = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n line = line.rstrip('\\n\\r')\n \n for match in pattern.finditer(line):\n start_idx = match.start()\n end_idx = match.end()\n \n # 检查等号块左侧：必须是恰好两个空格\n if start_idx < 2:\n return False\n if line[start_idx-1] != ' ' or line[start_idx-2] != ' ':\n return False\n # 保证左边第三个字符不是空格（确保仅有两个空格）\n if start_idx >= 3 and line[start_idx-3] == ' ':\n return False\n \n # 检查等号块右侧：必须是恰好两个空格\n if end_idx + 2 > len(line):\n return False\n if line[end_idx] != ' ' or line[end_idx+1] != ' ':\n return False\n # 保证右边第三个字符不是空格（确保仅有两个空格）\n if end_idx + 2 < len(line) and line[end_idx+2] == ' ':\n return False\n\n # 若没有任何代码文件，则视作没有生成符合要求的代码项目\n return has_code_files","tags":["Style","Layout"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":1,"instruction":"非常棒，滑动机制很顺滑。接下来我们引入多关卡机制和过关判定。\n1. 在地图数据中增加“终点”的定义（如用数字 2 表示），在 Canvas 上将其绘制为绿色方块。\n2. 当玩家滑动并最终停在终点上，或者滑动轨迹穿过终点时，判定为过关，并在画面中央弹出一个简易的“过关！”文本提示。\n3. 搭建关卡系统，预设 3 个不同难度和布局的地图关卡。玩家过关后延迟 1 秒自动加载并切换到下一关。\n4. 在 Canvas 上方添加一层简单的 HTML UI，显示当前关卡数（如 Level 1/3），并提供一个“重新开始本关”的按钮，以防玩家走入死胡同卡死。\n\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\n1. 在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录，此规则替换此前要求文件操作前运行pwd的规则。\n2. 不要在生成的代码里直接使用数字字面量，代码中涉及的所有数字全部必须使用常量来定义。","instruction_following_checklist":[{"constraint":"你回复用户的生成内容必须始终保持为中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 预估可能生成的代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.json'}\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 以避免由于存在BOM头而导致首行缩进读取错位\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n leading_spaces = 0\n has_tab = False\n \n # 提取该行的前导空白字符信息\n for char in line:\n if char == ' ':\n leading_spaces += 1\n elif char == '\\t':\n has_tab = True\n break\n else:\n break\n \n # 获取剔除前导空白字符后的实际内容\n content_after = line.lstrip()\n \n # 忽略完全没有任何有效内容的空行\n if not content_after:\n continue\n \n # 若前导缩进中出现制表符，视为不满足要求\n if has_tab:\n return False\n \n # 若前导空格的数量不为 4 的倍数，检查是否为多行注释特例\n if leading_spaces % 4 != 0:\n # 忽略多行注释中经常使用的以 '*' 对齐的行\n if content_after.startswith('*'):\n continue\n # 如果非特例且不是4个空格的倍数，视为不满足要求\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据项目技术栈及要求的特殊文件，定义属于\"代码文件\"的扩展名集合\n code_extensions = {'.html', '.js', '.css', '.json', '.md'}\n checked_files_count = 0\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 检查文件的总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n \n checked_files_count += 1\n except Exception:\n # 如果由于非文本或其他原因无法读取，则忽略该文件\n continue\n\n # 至少生成且被检查了一个代码文件才算通过该要求\n return checked_files_count > 0","tags":["Quantity","Exact Value"]},{"constraint":"要求生成的代码中，“=”符号的两侧均需添加两个空格","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n code_extensions = {'.html', '.js', '.css'}\n \n # 正则表达式：匹配单独的等号 '='\n # 排除常见的复合操作符、比较操作符和箭头函数 (如 ==, ===, !=, +=, -=, >=, <=, => 等)\n pattern = re.compile(r'(?<![=!<>+\\-*/%&|^])=(?![=>])')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 仅移除行尾的换行符，保留可能存在的行尾空格\n line = line.rstrip('\\r\\n')\n \n for match in pattern.finditer(line):\n idx = match.start()\n \n # 检查 \"=\" 左侧是否有至少两个空格\n if idx < 2:\n return False\n if line[idx-2:idx] != ' ':\n return False\n \n # 检查 \"=\" 右侧是否有至少两个空格\n if len(line) < idx + 3:\n return False\n if line[idx+1:idx+3] != ' ':\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"不要在生成的代码里直接使用数字字面量，代码中涉及的所有数字全部必须使用常量来定义","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":2,"instruction":"现在来增加解谜的深度：引入钥匙和门的机制。\n1. 在地图中增加“钥匙”和对应的“门”。请设计两种颜色机制（例如红钥匙开红门，蓝钥匙开蓝门）。\n2. 将钥匙绘制成带有对应颜色的小圆圈，将门绘制成带有对应颜色边框的障碍方块。\n3. 玩家滑过钥匙时可以将其捡起，UI 区域需要实时显示玩家当前拥有的钥匙列表或图标。\n4. 当玩家碰到对应颜色的门时，如果身上有对应钥匙，则消耗该钥匙，门被打开（门方块消失，变为可通行的道路），玩家继续滑动；如果没有对应钥匙，门就具有墙壁的物理属性，阻挡玩家前进。\n\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\n1. 从本次操作开始，生成的代码文件中所有的数字必须直接使用，严禁将其定义为常量，此规则替换此前要求数字全部使用常量定义的规则。\n2. 严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误。","instruction_following_checklist":[{"constraint":"你回复用户的生成内容必须始终保持为中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义目标代码文件扩展名\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\n has_checked_any_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_checked_any_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n for line in lines:\n # 剔除行尾换行符\n line_str = line.rstrip('\\r\\n')\n stripped = line_str.lstrip()\n \n # 忽略完全为空白的行\n if not stripped:\n continue\n \n # 提取前导空白字符\n indent = line_str[:len(line_str) - len(stripped)]\n \n # 规则1：不允许使用Tab缩进\n if '\\t' in indent:\n return False\n \n # 规则2：缩进空格数必须是4的倍数\n if len(indent) % 4 != 0:\n # 豁免多行注释(如以 * 开头的中间行)\n if stripped.startswith('*'):\n continue\n return False\n except Exception:\n # 忽略读取或编码引发的异常文件\n continue\n \n # 如果找到了代码文件且均无违背缩进规则，则返回True\n return has_checked_any_file","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义用户指令中可能涉及到的代码与文档相关的文件后缀\n target_extensions = {'.html', '.js', '.css', '.json', '.md'}\n \n # 过滤掉常见的非AI直接生成的缓存/环境/版本控制目录\n exclude_dirs = {'.git', 'node_modules', '__pycache__'}\n \n found_any_target_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 移除不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n found_any_target_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 连同空行一并计算，使用 split('\\n') 能最准确地反映包含空行的总行数\n lines_count = len(content.split('\\n'))\n \n # 判断行数是否为偶数\n if lines_count % 2 != 0:\n return False\n \n except Exception:\n # 如果遇到无法以 utf-8 解码或其他异常读取的文件，安全地跳过\n continue\n\n # 只有在至少检测到一个代码文件，且全部符合偶数行要求时，才视为满足条件\n return found_any_target_file","tags":["Quantity","Exact Value"]},{"constraint":"要求生成的代码中，“=”符号的两侧均需添加两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 需要校验的常用前端及代码文件后缀\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 1. 保护常见的 URL 双斜杠，避免被当成单行注释引起截断\n content = content.replace(\"https://\", \"https_//\").replace(\"http://\", \"http_//\")\n\n # 2. 去除多行注释、HTML 注释\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n\n # 3. 去除单行注释\n content = re.sub(r'//.*', '', content)\n\n # 4. 去除字符串内容 (保护字符串内部的等于号不受校验影响)\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n\n # 5. 替换包含 \"=\" 的复合运算符，避免引发等号判定的误报\n content = re.sub(r'===|!==|==|!=|<=|>=|=>|\\+=|-=|\\*=|/=|%=|&=|\\|=|\\^=', '@@', content)\n \n # 6. 按行查找剩余独立的 \"=\" 并检查两侧空格情况\n lines = content.split('\\n')\n for line in lines:\n idx = 0\n while True:\n idx = line.find('=', idx)\n if idx == -1:\n break\n \n # 判断 '=' 左侧是否有且仅有 2 个空格\n left_ok = False\n if idx >= 2 and line[idx-1] == ' ' and line[idx-2] == ' ':\n if idx == 2 or line[idx-3] != ' ':\n left_ok = True\n \n # 判断 '=' 右侧是否有且仅有 2 个空格\n right_ok = False\n if idx + 2 < len(line) and line[idx+1] == ' ' and line[idx+2] == ' ':\n if idx + 3 == len(line) or line[idx+3] != ' ':\n right_ok = True\n \n # 只要出现一处等号的两边不满足要求，则判定为校验未通过\n if not (left_ok and right_ok):\n return False\n \n idx += 1\n \n # 如果代码中所有的单等号都顺利通过了格式判定，返回 True\n return True","tags":["Style","Layout"]},{"constraint":"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"生成的代码文件中所有的数字必须直接使用，严禁将其定义为常量","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"游戏逻辑越来越完善了。目前地图大小受限于 Canvas 尺寸，下一步我们要支持大型地图，并实现摄像机跟随效果。\n1. 将第 3 关替换为一个远大于 Canvas 视口尺寸的大型迷宫（例如 30x30 的网格，而 Canvas 只能显示 10x10 的范围）。\n2. 实现摄像机（Camera）平滑跟随功能：当玩家滑动时，Canvas 视口要跟随玩家移动，尽量让玩家保持在画面中心。\n3. 做好边界处理：当摄像机移动到大地图边缘时，停止跟随，不应该露出地图外围的无效区域。\n4. 只渲染当前摄像机视口内（加一点冗余边界）的地图元素，提升渲染性能。\n\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\n1. 从本次操作开始，你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用，此规则替换此前要求数字必须直接使用、严禁将其定义为常量的规则。\n2. 后续每次修改项目之后，都必须对process-log.md文件进行更新，记录本次完成的事项。","instruction_following_checklist":[{"constraint":"你回复用户的生成内容必须始终保持为中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义需要检查的代码文件扩展名\n target_extensions = {'.html', '.js', '.css', '.json', '.ts', '.jsx', '.tsx'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 若文件读取遇到编码等问题，暂时跳过\n continue\n\n for line in lines:\n # 移除行首的空白字符以获取有效代码部分\n line_stripped = line.lstrip()\n \n # 1. 忽略完全是空白的行\n if not line_stripped:\n continue\n \n # 2. 忽略多行注释的中间行（一般以 * 开始），因为它们常常为了对齐而额外增加1个空格\n if line_stripped.startswith('*'):\n continue\n\n # 提取前导空白字符\n indent = line[:len(line) - len(line_stripped)]\n\n # 3. 如果缩进中包含制表符 \\t，不满足要求\n if '\\t' in indent:\n return False\n\n # 4. 如果纯空格缩进的长度不是 4 的倍数，不满足要求\n if len(indent) % 4 != 0:\n return False\n\n # 若所有匹配到的代码文件的有效行都符合“4个空格缩进”的规则，则满足要求\n return True","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"import os\n\ndef check_even_lines(response: str, workspace_path: str) -> bool:\n # 依据用户指令中提及的技术栈和要求产生的文件，锁定如下几种常见代码、配置和日志文件\n target_exts = {'.html', '.js', '.css', '.json', '.md'}\n \n if not os.path.exists(workspace_path):\n return False\n \n found_any_target = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录及常见的依赖包目录，避免对非生成产物进行误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_exts:\n found_any_target = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 使用 split('\\n') 能够严格等同于文本编辑器的行号计算规则，并且空行也会被视为一个独立的行元素\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数，存在非偶数行数文件则不满足要求\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 对于不可读取的异常文件跳过处理\n pass\n \n # 如果扫描且成功验证了属于这些扩展名的文件，即可返回 True\n return found_any_target","tags":["Quantity","Exact Value"]},{"constraint":"要求生成的代码中，“=”符号的两侧均需添加两个空格","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 定义常见的前端代码文件扩展名\n code_extensions = ('.js', '.html', '.css')\n has_code_file = False\n \n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n continue\n \n lines = text.split('\\n')\n for line in lines:\n for i, char in enumerate(line):\n if char == '=':\n # 检查该 '=' 是否为复合运算符的一部分 (如 +=, ==, !=, >=, => 等)\n if i > 0 and line[i-1] in '+-*/%!=<>&|^':\n continue\n if i < len(line) - 1 and line[i+1] in '=>':\n continue\n \n # 统计左侧连续空格数量\n left_spaces = 0\n j = i - 1\n while j >= 0 and line[j] == ' ':\n left_spaces += 1\n j -= 1\n \n # 统计右侧连续空格数量\n right_spaces = 0\n j = i + 1\n while j < len(line) and line[j] == ' ':\n right_spaces += 1\n j += 1\n \n # 判断 '=' 两侧是否恰好各有两个空格\n if left_spaces != 2 or right_spaces != 2:\n return False\n\n # 若未包含任何代码文件返回 False，否则所有单等号均验证通过则返回 True\n return has_code_file","tags":["Style","Layout"]},{"constraint":"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次修改项目之后，都必须对process-log.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":4,"instruction":"最后，我们来全面提升游戏的视觉表现，添加丰富的动效。\n1. 增加滑动拖尾：玩家快速滑动时，在后方生成逐渐消散的残影或小粒子，体现出滑动速度感。\n2. 交互动效：捡起钥匙时，增加钥匙向上飘散变透明的动画；使用钥匙开门时，门方块添加向两侧碎裂或缩小消失的过渡动画，而不是瞬间消失。\n3. 终点特效：给终点方块添加一个持续的呼吸缩放效果。当玩家通关时，在 Canvas 上生成全屏的纸屑（Confetti）撒花庆祝特效。\n4. 整体美化：将墙壁和道路改成圆角矩形，给玩家和门加上微小的发光或阴影效果，使整个游戏的质感更现代一点。\n\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\n1. 从本次操作开始，一旦对项目进行了改动，就必须更新 claude-progress.txt 文件，把本轮完成的内容记录下来，此规则替换此前要求修改项目后更新process-log.md的规则。\n2. 后续生成的所有代码文件中的每一个函数代码量限制在最多50行，该要求永久生效。","instruction_following_checklist":[{"constraint":"你回复用户的生成内容必须始终保持为中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义需要检查的代码文件扩展名（本题目涉及HTML5和JS，通常也会有CSS或JSON等周边文件）\n code_extensions = {'.js', '.html', '.css', '.json', '.ts'}\n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 忽略读取失败的文件，继续检查其他文件\n continue\n \n for line in lines:\n stripped_line = line.strip()\n \n # 忽略完全空行或仅有空白字符的行\n if not stripped_line:\n continue\n \n # 匹配行首前导的空格或制表符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_whitespace = match.group(1)\n \n # 条件1：绝对不能使用制表符 '\\t' 来进行缩进\n if '\\t' in leading_whitespace:\n return False\n \n # 条件2：空格数量必须是 4 的倍数\n if len(leading_whitespace) % 4 != 0:\n # 豁免特殊情况：多行块注释的中间行通常为了对齐，会有奇数个空格且以 '*' 开头\n # 或者HTML注释格式的缩进。\n if stripped_line.startswith('*') or stripped_line.startswith('<!--'):\n continue\n \n # 其他逻辑代码行一旦缩进不是4的倍数，则违背了“正好4个空格完成每一处缩进”\n return False\n\n # 若未找到任何代码文件，则也无从谈起符合要求，返回 False；反之返回 True\n return has_code_file","tags":["Style","Layout"]},{"constraint":"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断本项目可能的代码文件扩展名\n code_extensions = {'.html', '.js', '.css'}\n has_code_files = False\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 连同空行一并计算，验证总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n \n has_code_files = True\n except Exception:\n # 若遇编码等读取错误可忽略当前文件或按不满足处理，这里选择跳过读取失败的文件\n continue\n\n # 确保项目中至少生成了相关的代码文件\n if not has_code_files:\n return False\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"要求生成的代码中，“=”符号的两侧均需添加两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 目标只检查前端代码文件，忽略 json、txt 或 md 等非代码文件\n extensions = ['.js', '.html', '.css']\n \n def remove_strings_and_comments(content: str, ext: str) -> str:\n if ext in ['.js', '.css']:\n # 移除单行注释\n content = re.sub(r'//.*', '', content)\n # 移除多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 移除双引号、单引号、反引号包裹的字符串内容\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', content, flags=re.DOTALL)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", content, flags=re.DOTALL)\n content = re.sub(r'`([^`\\\\]|\\\\.)*`', \"``\", content, flags=re.DOTALL)\n elif ext == '.html':\n # 移除HTML注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 移除HTML中包裹在双引号、单引号中的属性值内容\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', content, flags=re.DOTALL)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", content, flags=re.DOTALL)\n # HTML文件内同样可能包含JS，简单移除单行及多行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n return content\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1]\n if ext not in extensions:\n continue\n \n file_path = os.path.join(root, file)\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 清理代码中的字符串及注释内容\n content = remove_strings_and_comments(content, ext)\n lines = content.split('\\n')\n \n # 遍历每一行查找 \"=\"\n for line in lines:\n n = len(line)\n for i in range(n):\n if line[i] == '=':\n prev_char = line[i-1] if i > 0 else ''\n next_char = line[i+1] if i < n-1 else ''\n \n # 如果该\"=\"是组合操作符的一部分，例如 ==, >=, <=, !=, +=, =>, ~= 等，则忽略\n if prev_char in '=<>!+-*/%&|^~$' or next_char in '=>':\n continue\n \n # 检查 \"=\" 前面是否刚好有两个连续空格\n left_spaces = 0\n idx = i - 1\n while idx >= 0 and line[idx] == ' ':\n left_spaces += 1\n idx -= 1\n \n # 检查 \"=\" 后面是否刚好有两个连续空格\n right_spaces = 0\n idx = i + 1\n while idx < n and line[idx] == ' ':\n right_spaces += 1\n idx += 1\n \n # 若任何一侧空格数不等于2，则未满足给定要求\n if left_spaces != 2 or right_spaces != 2:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"一旦对项目进行了改动，就必须更新 claude-progress.txt 文件，把本轮完成的内容记录下来","validation_code":"","tags":["Environment","Logging"]},{"constraint":"后续生成的所有代码文件中的每一个函数代码量限制在最多50行","validation_code":"import os\nimport re\n\ndef check_function_line_limit(response: str, workspace_path: str) -> bool:\n \n def clean_code(code: str) -> str:\n \"\"\"\n 清理代码中的字符串、注释，替换为空格，但保持换行符不变以维持原有行号映射。\n \"\"\"\n res = []\n i = 0\n n = len(code)\n in_str = False\n str_char = ''\n in_line_comment = False\n in_block_comment = False\n \n while i < n:\n if in_str:\n if code[i] == '\\\\':\n res.append(' ')\n if i + 1 < n:\n res.append('\\n' if code[i+1] == '\\n' else ' ')\n i += 2\n elif code[i] == str_char:\n in_str = False\n res.append(' ')\n i += 1\n else:\n res.append('\\n' if code[i] == '\\n' else ' ')\n i += 1\n elif in_line_comment:\n if code[i] == '\\n':\n in_line_comment = False\n res.append('\\n')\n else:\n res.append(' ')\n i += 1\n elif in_block_comment:\n if i + 1 < n and code[i:i+2] == '*/':\n in_block_comment = False\n res.append(' ')\n i += 2\n else:\n res.append('\\n' if code[i] == '\\n' else ' ')\n i += 1\n else:\n if code[i] in '\"\\'`':\n in_str = True\n str_char = code[i]\n res.append(' ')\n i += 1\n elif i + 1 < n and code[i:i+2] == '//':\n in_line_comment = True\n res.append(' ')\n i += 2\n elif i + 1 < n and code[i:i+2] == '/*':\n in_block_comment = True\n res.append(' ')\n i += 2\n else:\n res.append(code[i])\n i += 1\n return \"\".join(res)\n\n def is_function_brace(code_str: str, brace_idx: int) -> bool:\n \"\"\"判断给定的 { 是否是一个函数的左大括号\"\"\"\n i = brace_idx - 1\n while i >= 0 and code_str[i].isspace():\n i -= 1\n if i < 0:\n return False\n \n char = code_str[i]\n \n # 箭头函数特征 `=> {`\n if char == '>':\n if i - 1 >= 0 and code_str[i-1] == '=':\n return True\n return False\n # 普通函数特征 `) {`\n elif char == ')':\n depth = 1\n j = i - 1\n while j >= 0:\n if code_str[j] == ')':\n depth += 1\n elif code_str[j] == '(':\n depth -= 1\n if depth == 0:\n break\n j -= 1\n if j < 0:\n return False\n \n k = j - 1\n while k >= 0 and code_str[k].isspace():\n k -= 1\n if k < 0:\n return True\n \n word_end = k\n while k >= 0 and (code_str[k].isalnum() or code_str[k] in '_$'):\n k -= 1\n word = code_str[k+1:word_end+1]\n \n # 如果是控制流关键字等，则不是函数块而是逻辑块\n exclude_keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\n if word in exclude_keywords:\n return False\n return True\n \n return False\n\n def check_code(code: str) -> bool:\n \"\"\"检查一段代码中的所有函数代码块是否均符合 <=50 行的限制\"\"\"\n code = code.replace('\\r\\n', '\\n').replace('\\r', '\\n')\n cleaned = clean_code(code)\n \n line_num = 1\n stack = []\n for i, char in enumerate(cleaned):\n if char == '\\n':\n line_num += 1\n elif char == '{':\n if is_function_brace(cleaned, i):\n stack.append(('func', line_num))\n else:\n stack.append(('block', line_num))\n elif char == '}':\n if stack:\n b_type, start_line = stack.pop()\n if b_type == 'func':\n line_count = line_num - start_line + 1\n if line_count > 50:\n return False\n return True\n\n def extract_js_from_html(html_code: str) -> list:\n \"\"\"提取 HTML 内嵌的 JS <script> 片段\"\"\"\n js_codes = []\n pattern = re.compile(r'<script[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\n for match in pattern.finditer(html_code):\n js_codes.append(match.group(1))\n return js_codes\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js') or file.endswith('.ts'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_code(code):\n return False\n except Exception:\n pass\n elif file.endswith('.html') or file.endswith('.htm'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n html_code = f.read()\n js_codes = extract_js_from_html(html_code)\n for code in js_codes:\n if not check_code(code):\n return False\n except Exception:\n pass\n\n return True","tags":["Quantity","Range"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用 HTML5 Canvas 和原生 JavaScript 创建一个滑动迷宫游戏的基础骨架。需求如下：\\n1. 页面居中放置一个合适大小的 Canvas 容器，背景设为浅灰色。\\n2. 定义一个简单的二维数组地图数据结构（如 10x10 网格，0 代表道路，1 代表墙壁），并将其绘制到 Canvas 上。使用纯色方块即可，比如黑色表示墙壁，白色表示道路。\\n3. 在起点空地上绘制一个玩家（例如一个蓝色方块）。\\n4. 实现最基础的“滑动”移动交互：玩家通过键盘方向键（上下左右）控制，一旦按下某个方向，玩家方块就会沿着该方向匀速持续滑动，直到撞到墙壁才停下来。滑动过程中不能改变方向。第一轮先不涉及过关和终点，只要把地图绘制和核心的滑动撞墙判定写好即可。\\n\\n另外请你在后续所有操作中始终遵循以下要求：\\n1. 执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录。\\n2. 必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你回复用户的生成内容必须始终保持为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义需要检查缩进的文件后缀\\n code_extensions = {'.html', '.js', '.css', '.json'}\\n has_checked_any = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_checked_any = True\\n \\n for line in lines:\\n # 去除右侧所有空白字符，借此跳过全空白的行\\n if not line.rstrip():\\n continue\\n \\n # 匹配当前行首的前导空白字符（包含空格或制表符）\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_whitespace = match.group(1)\\n # content 为去除前导空白后的正文内容\\n content = line[len(leading_whitespace):]\\n \\n # 规则1: 绝对不允许使用制表符 \\\\t\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 规则2: 使用正好 4 个空格完成每一处缩进，因此空格数必须是 4 的倍数\\n if len(leading_whitespace) % 4 != 0:\\n # 豁免多行注释中间对齐的星号 (例如 ' * 注释')\\n if content.startswith('*'):\\n continue\\n return False\\n \\n # 如果项目中存在代码文件并且全部通过了上述的验证，则判定为成功\\n return has_checked_any\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能的代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.json'}\\n has_code_file = False\\n\\n # 检查工作区路径是否有效\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符切分以统计包含空行在内的所有行数\\n # 例如，内容为 \\\"a\\\\nb\\\" 会被切分为 ['a', 'b']，总行数为 2\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果遇到文件不可读或编码异常，安全起见判定为不满足条件\\n return False\\n\\n # 确保项目中确实包含生成的代码文件并且全部满足了偶数行要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"要求生成的代码中，“=”符号的两侧均需添加两个空格\",\n \"validation_code\": \"def check_equals_spaces(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见的项目代码文件类型\\n exts = ['.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue']\\n has_code_files = False\\n \\n # 正则表达式：捕获单独的或连续的 = 号块（即 =, ==, ===）\\n # 排除复合运算符的情况，如左侧不为 + - * / % < > ! =，右侧不为 = 或 > (防止匹配箭头函数 =>)\\n pattern = re.compile(r'(?<![+\\\\-*/%<>!=])(=+)(?![=>])')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not any(file.endswith(ext) for ext in exts):\\n continue\\n has_code_files = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n line = line.rstrip('\\\\n\\\\r')\\n \\n for match in pattern.finditer(line):\\n start_idx = match.start()\\n end_idx = match.end()\\n \\n # 检查等号块左侧：必须是恰好两个空格\\n if start_idx < 2:\\n return False\\n if line[start_idx-1] != ' ' or line[start_idx-2] != ' ':\\n return False\\n # 保证左边第三个字符不是空格（确保仅有两个空格）\\n if start_idx >= 3 and line[start_idx-3] == ' ':\\n return False\\n \\n # 检查等号块右侧：必须是恰好两个空格\\n if end_idx + 2 > len(line):\\n return False\\n if line[end_idx] != ' ' or line[end_idx+1] != ' ':\\n return False\\n # 保证右边第三个字符不是空格（确保仅有两个空格）\\n if end_idx + 2 < len(line) and line[end_idx+2] == ' ':\\n return False\\n\\n # 若没有任何代码文件，则视作没有生成符合要求的代码项目\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"非常棒，滑动机制很顺滑。接下来我们引入多关卡机制和过关判定。\\n1. 在地图数据中增加“终点”的定义（如用数字 2 表示），在 Canvas 上将其绘制为绿色方块。\\n2. 当玩家滑动并最终停在终点上，或者滑动轨迹穿过终点时，判定为过关，并在画面中央弹出一个简易的“过关！”文本提示。\\n3. 搭建关卡系统，预设 3 个不同难度和布局的地图关卡。玩家过关后延迟 1 秒自动加载并切换到下一关。\\n4. 在 Canvas 上方添加一层简单的 HTML UI，显示当前关卡数（如 Level 1/3），并提供一个“重新开始本关”的按钮，以防玩家走入死胡同卡死。\\n\\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\\n1. 在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录，此规则替换此前要求文件操作前运行pwd的规则。\\n2. 不要在生成的代码里直接使用数字字面量，代码中涉及的所有数字全部必须使用常量来定义。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你回复用户的生成内容必须始终保持为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 预估可能生成的代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.json'}\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 以避免由于存在BOM头而导致首行缩进读取错位\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n leading_spaces = 0\\n has_tab = False\\n \\n # 提取该行的前导空白字符信息\\n for char in line:\\n if char == ' ':\\n leading_spaces += 1\\n elif char == '\\\\t':\\n has_tab = True\\n break\\n else:\\n break\\n \\n # 获取剔除前导空白字符后的实际内容\\n content_after = line.lstrip()\\n \\n # 忽略完全没有任何有效内容的空行\\n if not content_after:\\n continue\\n \\n # 若前导缩进中出现制表符，视为不满足要求\\n if has_tab:\\n return False\\n \\n # 若前导空格的数量不为 4 的倍数，检查是否为多行注释特例\\n if leading_spaces % 4 != 0:\\n # 忽略多行注释中经常使用的以 '*' 对齐的行\\n if content_after.startswith('*'):\\n continue\\n # 如果非特例且不是4个空格的倍数，视为不满足要求\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据项目技术栈及要求的特殊文件，定义属于\\\"代码文件\\\"的扩展名集合\\n code_extensions = {'.html', '.js', '.css', '.json', '.md'}\\n checked_files_count = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 检查文件的总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n \\n checked_files_count += 1\\n except Exception:\\n # 如果由于非文本或其他原因无法读取，则忽略该文件\\n continue\\n\\n # 至少生成且被检查了一个代码文件才算通过该要求\\n return checked_files_count > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"要求生成的代码中，“=”符号的两侧均需添加两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.js', '.css'}\\n \\n # 正则表达式：匹配单独的等号 '='\\n # 排除常见的复合操作符、比较操作符和箭头函数 (如 ==, ===, !=, +=, -=, >=, <=, => 等)\\n pattern = re.compile(r'(?<![=!<>+\\\\-*/%&|^])=(?![=>])')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 仅移除行尾的换行符，保留可能存在的行尾空格\\n line = line.rstrip('\\\\r\\\\n')\\n \\n for match in pattern.finditer(line):\\n idx = match.start()\\n \\n # 检查 \\\"=\\\" 左侧是否有至少两个空格\\n if idx < 2:\\n return False\\n if line[idx-2:idx] != ' ':\\n return False\\n \\n # 检查 \\\"=\\\" 右侧是否有至少两个空格\\n if len(line) < idx + 3:\\n return False\\n if line[idx+1:idx+3] != ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"不要在生成的代码里直接使用数字字面量，代码中涉及的所有数字全部必须使用常量来定义\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来增加解谜的深度：引入钥匙和门的机制。\\n1. 在地图中增加“钥匙”和对应的“门”。请设计两种颜色机制（例如红钥匙开红门，蓝钥匙开蓝门）。\\n2. 将钥匙绘制成带有对应颜色的小圆圈，将门绘制成带有对应颜色边框的障碍方块。\\n3. 玩家滑过钥匙时可以将其捡起，UI 区域需要实时显示玩家当前拥有的钥匙列表或图标。\\n4. 当玩家碰到对应颜色的门时，如果身上有对应钥匙，则消耗该钥匙，门被打开（门方块消失，变为可通行的道路），玩家继续滑动；如果没有对应钥匙，门就具有墙壁的物理属性，阻挡玩家前进。\\n\\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\\n1. 从本次操作开始，生成的代码文件中所有的数字必须直接使用，严禁将其定义为常量，此规则替换此前要求数字全部使用常量定义的规则。\\n2. 严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你回复用户的生成内容必须始终保持为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义目标代码文件扩展名\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\\n has_checked_any_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_checked_any_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n for line in lines:\\n # 剔除行尾换行符\\n line_str = line.rstrip('\\\\r\\\\n')\\n stripped = line_str.lstrip()\\n \\n # 忽略完全为空白的行\\n if not stripped:\\n continue\\n \\n # 提取前导空白字符\\n indent = line_str[:len(line_str) - len(stripped)]\\n \\n # 规则1：不允许使用Tab缩进\\n if '\\\\t' in indent:\\n return False\\n \\n # 规则2：缩进空格数必须是4的倍数\\n if len(indent) % 4 != 0:\\n # 豁免多行注释(如以 * 开头的中间行)\\n if stripped.startswith('*'):\\n continue\\n return False\\n except Exception:\\n # 忽略读取或编码引发的异常文件\\n continue\\n \\n # 如果找到了代码文件且均无违背缩进规则，则返回True\\n return has_checked_any_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义用户指令中可能涉及到的代码与文档相关的文件后缀\\n target_extensions = {'.html', '.js', '.css', '.json', '.md'}\\n \\n # 过滤掉常见的非AI直接生成的缓存/环境/版本控制目录\\n exclude_dirs = {'.git', 'node_modules', '__pycache__'}\\n \\n found_any_target_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 移除不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n found_any_target_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 连同空行一并计算，使用 split('\\\\n') 能最准确地反映包含空行的总行数\\n lines_count = len(content.split('\\\\n'))\\n \\n # 判断行数是否为偶数\\n if lines_count % 2 != 0:\\n return False\\n \\n except Exception:\\n # 如果遇到无法以 utf-8 解码或其他异常读取的文件，安全地跳过\\n continue\\n\\n # 只有在至少检测到一个代码文件，且全部符合偶数行要求时，才视为满足条件\\n return found_any_target_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"要求生成的代码中，“=”符号的两侧均需添加两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 需要校验的常用前端及代码文件后缀\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 1. 保护常见的 URL 双斜杠，避免被当成单行注释引起截断\\n content = content.replace(\\\"https://\\\", \\\"https_//\\\").replace(\\\"http://\\\", \\\"http_//\\\")\\n\\n # 2. 去除多行注释、HTML 注释\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n\\n # 3. 去除单行注释\\n content = re.sub(r'//.*', '', content)\\n\\n # 4. 去除字符串内容 (保护字符串内部的等于号不受校验影响)\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n\\n # 5. 替换包含 \\\"=\\\" 的复合运算符，避免引发等号判定的误报\\n content = re.sub(r'===|!==|==|!=|<=|>=|=>|\\\\+=|-=|\\\\*=|/=|%=|&=|\\\\|=|\\\\^=', '@@', content)\\n \\n # 6. 按行查找剩余独立的 \\\"=\\\" 并检查两侧空格情况\\n lines = content.split('\\\\n')\\n for line in lines:\\n idx = 0\\n while True:\\n idx = line.find('=', idx)\\n if idx == -1:\\n break\\n \\n # 判断 '=' 左侧是否有且仅有 2 个空格\\n left_ok = False\\n if idx >= 2 and line[idx-1] == ' ' and line[idx-2] == ' ':\\n if idx == 2 or line[idx-3] != ' ':\\n left_ok = True\\n \\n # 判断 '=' 右侧是否有且仅有 2 个空格\\n right_ok = False\\n if idx + 2 < len(line) and line[idx+1] == ' ' and line[idx+2] == ' ':\\n if idx + 3 == len(line) or line[idx+3] != ' ':\\n right_ok = True\\n \\n # 只要出现一处等号的两边不满足要求，则判定为校验未通过\\n if not (left_ok and right_ok):\\n return False\\n \\n idx += 1\\n \\n # 如果代码中所有的单等号都顺利通过了格式判定，返回 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中所有的数字必须直接使用，严禁将其定义为常量\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"游戏逻辑越来越完善了。目前地图大小受限于 Canvas 尺寸，下一步我们要支持大型地图，并实现摄像机跟随效果。\\n1. 将第 3 关替换为一个远大于 Canvas 视口尺寸的大型迷宫（例如 30x30 的网格，而 Canvas 只能显示 10x10 的范围）。\\n2. 实现摄像机（Camera）平滑跟随功能：当玩家滑动时，Canvas 视口要跟随玩家移动，尽量让玩家保持在画面中心。\\n3. 做好边界处理：当摄像机移动到大地图边缘时，停止跟随，不应该露出地图外围的无效区域。\\n4. 只渲染当前摄像机视口内（加一点冗余边界）的地图元素，提升渲染性能。\\n\\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\\n1. 从本次操作开始，你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用，此规则替换此前要求数字必须直接使用、严禁将其定义为常量的规则。\\n2. 后续每次修改项目之后，都必须对process-log.md文件进行更新，记录本次完成的事项。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你回复用户的生成内容必须始终保持为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义需要检查的代码文件扩展名\\n target_extensions = {'.html', '.js', '.css', '.json', '.ts', '.jsx', '.tsx'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 若文件读取遇到编码等问题，暂时跳过\\n continue\\n\\n for line in lines:\\n # 移除行首的空白字符以获取有效代码部分\\n line_stripped = line.lstrip()\\n \\n # 1. 忽略完全是空白的行\\n if not line_stripped:\\n continue\\n \\n # 2. 忽略多行注释的中间行（一般以 * 开始），因为它们常常为了对齐而额外增加1个空格\\n if line_stripped.startswith('*'):\\n continue\\n\\n # 提取前导空白字符\\n indent = line[:len(line) - len(line_stripped)]\\n\\n # 3. 如果缩进中包含制表符 \\\\t，不满足要求\\n if '\\\\t' in indent:\\n return False\\n\\n # 4. 如果纯空格缩进的长度不是 4 的倍数，不满足要求\\n if len(indent) % 4 != 0:\\n return False\\n\\n # 若所有匹配到的代码文件的有效行都符合“4个空格缩进”的规则，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"import os\\n\\ndef check_even_lines(response: str, workspace_path: str) -> bool:\\n # 依据用户指令中提及的技术栈和要求产生的文件，锁定如下几种常见代码、配置和日志文件\\n target_exts = {'.html', '.js', '.css', '.json', '.md'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n found_any_target = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录及常见的依赖包目录，避免对非生成产物进行误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_exts:\\n found_any_target = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 使用 split('\\\\n') 能够严格等同于文本编辑器的行号计算规则，并且空行也会被视为一个独立的行元素\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数，存在非偶数行数文件则不满足要求\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 对于不可读取的异常文件跳过处理\\n pass\\n \\n # 如果扫描且成功验证了属于这些扩展名的文件，即可返回 True\\n return found_any_target\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"要求生成的代码中，“=”符号的两侧均需添加两个空格\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端代码文件扩展名\\n code_extensions = ('.js', '.html', '.css')\\n has_code_file = False\\n \\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n lines = text.split('\\\\n')\\n for line in lines:\\n for i, char in enumerate(line):\\n if char == '=':\\n # 检查该 '=' 是否为复合运算符的一部分 (如 +=, ==, !=, >=, => 等)\\n if i > 0 and line[i-1] in '+-*/%!=<>&|^':\\n continue\\n if i < len(line) - 1 and line[i+1] in '=>':\\n continue\\n \\n # 统计左侧连续空格数量\\n left_spaces = 0\\n j = i - 1\\n while j >= 0 and line[j] == ' ':\\n left_spaces += 1\\n j -= 1\\n \\n # 统计右侧连续空格数量\\n right_spaces = 0\\n j = i + 1\\n while j < len(line) and line[j] == ' ':\\n right_spaces += 1\\n j += 1\\n \\n # 判断 '=' 两侧是否恰好各有两个空格\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n\\n # 若未包含任何代码文件返回 False，否则所有单等号均验证通过则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次修改项目之后，都必须对process-log.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，我们来全面提升游戏的视觉表现，添加丰富的动效。\\n1. 增加滑动拖尾：玩家快速滑动时，在后方生成逐渐消散的残影或小粒子，体现出滑动速度感。\\n2. 交互动效：捡起钥匙时，增加钥匙向上飘散变透明的动画；使用钥匙开门时，门方块添加向两侧碎裂或缩小消失的过渡动画，而不是瞬间消失。\\n3. 终点特效：给终点方块添加一个持续的呼吸缩放效果。当玩家通关时，在 Canvas 上生成全屏的纸屑（Confetti）撒花庆祝特效。\\n4. 整体美化：将墙壁和道路改成圆角矩形，给玩家和门加上微小的发光或阴影效果，使整个游戏的质感更现代一点。\\n\\n另外请你在后续所有操作中始终遵循以下更新及新增的要求：\\n1. 从本次操作开始，一旦对项目进行了改动，就必须更新 claude-progress.txt 文件，把本轮完成的内容记录下来，此规则替换此前要求修改项目后更新process-log.md的规则。\\n2. 后续生成的所有代码文件中的每一个函数代码量限制在最多50行，该要求永久生效。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你回复用户的生成内容必须始终保持为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义需要检查的代码文件扩展名（本题目涉及HTML5和JS，通常也会有CSS或JSON等周边文件）\\n code_extensions = {'.js', '.html', '.css', '.json', '.ts'}\\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 忽略读取失败的文件，继续检查其他文件\\n continue\\n \\n for line in lines:\\n stripped_line = line.strip()\\n \\n # 忽略完全空行或仅有空白字符的行\\n if not stripped_line:\\n continue\\n \\n # 匹配行首前导的空格或制表符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_whitespace = match.group(1)\\n \\n # 条件1：绝对不能使用制表符 '\\\\t' 来进行缩进\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 条件2：空格数量必须是 4 的倍数\\n if len(leading_whitespace) % 4 != 0:\\n # 豁免特殊情况：多行块注释的中间行通常为了对齐，会有奇数个空格且以 '*' 开头\\n # 或者HTML注释格式的缩进。\\n if stripped_line.startswith('*') or stripped_line.startswith('<!--'):\\n continue\\n \\n # 其他逻辑代码行一旦缩进不是4的倍数，则违背了“正好4个空格完成每一处缩进”\\n return False\\n\\n # 若未找到任何代码文件，则也无从谈起符合要求，返回 False；反之返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在modify_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新 dev_log.md 文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断本项目可能的代码文件扩展名\\n code_extensions = {'.html', '.js', '.css'}\\n has_code_files = False\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 连同空行一并计算，验证总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n \\n has_code_files = True\\n except Exception:\\n # 若遇编码等读取错误可忽略当前文件或按不满足处理，这里选择跳过读取失败的文件\\n continue\\n\\n # 确保项目中至少生成了相关的代码文件\\n if not has_code_files:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"要求生成的代码中，“=”符号的两侧均需添加两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 目标只检查前端代码文件，忽略 json、txt 或 md 等非代码文件\\n extensions = ['.js', '.html', '.css']\\n \\n def remove_strings_and_comments(content: str, ext: str) -> str:\\n if ext in ['.js', '.css']:\\n # 移除单行注释\\n content = re.sub(r'//.*', '', content)\\n # 移除多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 移除双引号、单引号、反引号包裹的字符串内容\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content, flags=re.DOTALL)\\n content = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', \\\"``\\\", content, flags=re.DOTALL)\\n elif ext == '.html':\\n # 移除HTML注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 移除HTML中包裹在双引号、单引号中的属性值内容\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content, flags=re.DOTALL)\\n # HTML文件内同样可能包含JS，简单移除单行及多行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n return content\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1]\\n if ext not in extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 清理代码中的字符串及注释内容\\n content = remove_strings_and_comments(content, ext)\\n lines = content.split('\\\\n')\\n \\n # 遍历每一行查找 \\\"=\\\"\\n for line in lines:\\n n = len(line)\\n for i in range(n):\\n if line[i] == '=':\\n prev_char = line[i-1] if i > 0 else ''\\n next_char = line[i+1] if i < n-1 else ''\\n \\n # 如果该\\\"=\\\"是组合操作符的一部分，例如 ==, >=, <=, !=, +=, =>, ~= 等，则忽略\\n if prev_char in '=<>!+-*/%&|^~$' or next_char in '=>':\\n continue\\n \\n # 检查 \\\"=\\\" 前面是否刚好有两个连续空格\\n left_spaces = 0\\n idx = i - 1\\n while idx >= 0 and line[idx] == ' ':\\n left_spaces += 1\\n idx -= 1\\n \\n # 检查 \\\"=\\\" 后面是否刚好有两个连续空格\\n right_spaces = 0\\n idx = i + 1\\n while idx < n and line[idx] == ' ':\\n right_spaces += 1\\n idx += 1\\n \\n # 若任何一侧空格数不等于2，则未满足给定要求\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在所有文件操作的前置步骤中，皆禁止运行`pwd`来确定当前目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，对于所有要用到的数字，都必须以常量定义的形式出现，严禁直接使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"严格要求你生成的代码进行ESLint检查，你必须修复期间产生的所有报错，确保最终的ESLint扫描结果显示无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"一旦对项目进行了改动，就必须更新 claude-progress.txt 文件，把本轮完成的内容记录下来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"后续生成的所有代码文件中的每一个函数代码量限制在最多50行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_function_line_limit(response: str, workspace_path: str) -> bool:\\n \\n def clean_code(code: str) -> str:\\n \\\"\\\"\\\"\\n 清理代码中的字符串、注释，替换为空格，但保持换行符不变以维持原有行号映射。\\n \\\"\\\"\\\"\\n res = []\\n i = 0\\n n = len(code)\\n in_str = False\\n str_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < n:\\n if in_str:\\n if code[i] == '\\\\\\\\':\\n res.append(' ')\\n if i + 1 < n:\\n res.append('\\\\n' if code[i+1] == '\\\\n' else ' ')\\n i += 2\\n elif code[i] == str_char:\\n in_str = False\\n res.append(' ')\\n i += 1\\n else:\\n res.append('\\\\n' if code[i] == '\\\\n' else ' ')\\n i += 1\\n elif in_line_comment:\\n if code[i] == '\\\\n':\\n in_line_comment = False\\n res.append('\\\\n')\\n else:\\n res.append(' ')\\n i += 1\\n elif in_block_comment:\\n if i + 1 < n and code[i:i+2] == '*/':\\n in_block_comment = False\\n res.append(' ')\\n i += 2\\n else:\\n res.append('\\\\n' if code[i] == '\\\\n' else ' ')\\n i += 1\\n else:\\n if code[i] in '\\\"\\\\'`':\\n in_str = True\\n str_char = code[i]\\n res.append(' ')\\n i += 1\\n elif i + 1 < n and code[i:i+2] == '//':\\n in_line_comment = True\\n res.append(' ')\\n i += 2\\n elif i + 1 < n and code[i:i+2] == '/*':\\n in_block_comment = True\\n res.append(' ')\\n i += 2\\n else:\\n res.append(code[i])\\n i += 1\\n return \\\"\\\".join(res)\\n\\n def is_function_brace(code_str: str, brace_idx: int) -> bool:\\n \\\"\\\"\\\"判断给定的 { 是否是一个函数的左大括号\\\"\\\"\\\"\\n i = brace_idx - 1\\n while i >= 0 and code_str[i].isspace():\\n i -= 1\\n if i < 0:\\n return False\\n \\n char = code_str[i]\\n \\n # 箭头函数特征 `=> {`\\n if char == '>':\\n if i - 1 >= 0 and code_str[i-1] == '=':\\n return True\\n return False\\n # 普通函数特征 `) {`\\n elif char == ')':\\n depth = 1\\n j = i - 1\\n while j >= 0:\\n if code_str[j] == ')':\\n depth += 1\\n elif code_str[j] == '(':\\n depth -= 1\\n if depth == 0:\\n break\\n j -= 1\\n if j < 0:\\n return False\\n \\n k = j - 1\\n while k >= 0 and code_str[k].isspace():\\n k -= 1\\n if k < 0:\\n return True\\n \\n word_end = k\\n while k >= 0 and (code_str[k].isalnum() or code_str[k] in '_$'):\\n k -= 1\\n word = code_str[k+1:word_end+1]\\n \\n # 如果是控制流关键字等，则不是函数块而是逻辑块\\n exclude_keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\\n if word in exclude_keywords:\\n return False\\n return True\\n \\n return False\\n\\n def check_code(code: str) -> bool:\\n \\\"\\\"\\\"检查一段代码中的所有函数代码块是否均符合 <=50 行的限制\\\"\\\"\\\"\\n code = code.replace('\\\\r\\\\n', '\\\\n').replace('\\\\r', '\\\\n')\\n cleaned = clean_code(code)\\n \\n line_num = 1\\n stack = []\\n for i, char in enumerate(cleaned):\\n if char == '\\\\n':\\n line_num += 1\\n elif char == '{':\\n if is_function_brace(cleaned, i):\\n stack.append(('func', line_num))\\n else:\\n stack.append(('block', line_num))\\n elif char == '}':\\n if stack:\\n b_type, start_line = stack.pop()\\n if b_type == 'func':\\n line_count = line_num - start_line + 1\\n if line_count > 50:\\n return False\\n return True\\n\\n def extract_js_from_html(html_code: str) -> list:\\n \\\"\\\"\\\"提取 HTML 内嵌的 JS <script> 片段\\\"\\\"\\\"\\n js_codes = []\\n pattern = re.compile(r'<script[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\\n for match in pattern.finditer(html_code):\\n js_codes.append(match.group(1))\\n return js_codes\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js') or file.endswith('.ts'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_code(code):\\n return False\\n except Exception:\\n pass\\n elif file.endswith('.html') or file.endswith('.htm'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n html_code = f.read()\\n js_codes = extract_js_from_html(html_code)\\n for code in js_codes:\\n if not check_code(code):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查基础UI布局与顶部信息显示：观察页面，验证页面中央是否存在一个背景为浅灰色的Canvas画布，且Canvas正上方有一层HTML界面，准确显示了当前关卡进度（例如'Level 1/3'）以及一个'重新开始本关'的按钮。","检查地图静态元素渲染美化：观察Canvas内生成的迷宫地图，验证墙壁（如黑色）和道路（如白色）是否被渲染为圆角矩形，玩家（蓝色方块）和门障碍物是否附带有微小的发光或阴影效果，终点位置的绿色方块是否存在持续缩放的'呼吸'动画。","检查基础滑动操作与撞墙机制：按下任一方向键（如向右），观察玩家蓝色方块是否沿着该方向匀速且持续地滑动；在滑动中途按下其他方向键，玩家是否保持原方向不变，直到撞击到墙壁障碍物才完全停止。","检查滑动拖尾动效：在玩家长距离滑动时进行观测或截图，验证蓝色方块的后方是否动态生成了逐渐消散的残影或小粒子拖尾，以体现滑动速度感。","检查重新开始功能：控制玩家随意滑动几个位置后，点击上方的'重新开始本关'按钮，验证地图和玩家位置是否瞬间恢复为该关卡的初始未解谜状态。","检查无钥匙时的门阻挡判定：在玩家未拾取对应颜色钥匙的情况下，控制玩家滑向带有颜色边框的方块（门），验证该门是否像墙壁一样将玩家直接挡住并迫使玩家停下。","检查拾取钥匙及UI更新逻辑：控制玩家滑行路线经过带有颜色的小圆圈（钥匙），验证碰到钥匙时，该钥匙是否呈现向上飘散并变透明的动效；同时验证上方的UI状态栏是否实时增加并显示了该对应钥匙的图标或列表数据。","检查消耗钥匙的开门机制：在玩家拾取了指定颜色的钥匙后，滑向同色的边框门，验证门是否触发了向两侧碎裂或缩小消失的过渡动效；并且玩家方块是否未被阻挡、顺畅滑过原先门的位置，同时顶部UI对应扣除一把钥匙。","检查过关判定与关卡切换机制：控制玩家滑动最终停在绿色终点方块上或滑动轨迹穿过终点，验证画面中央是否弹出'过关！'文本提示，且等待约1秒后，是否自动清除旧地图并加载渲染出下一个新关卡地图，上方UI的Level数字对应增加。","检查摄像机平滑跟随与边界处理：进入最后一关（大型地图关卡），控制玩家滑动，验证Canvas画面视口是否能平滑跟随玩家移动并让玩家保持在画面中心区域；当玩家向大地图的最边缘移动时，验证摄像机是否停止跟随，确保视口内不露出大地图外围的背景空白区域，验证是否仅渲染摄像机视口内（加一点冗余边界）的地图元素。","检查通关撒花特效：控制玩家通过全部关卡（达到最后一关终点）完成游戏，验证Canvas画布上是否生成且持续播放了全屏的纸屑（Confetti）撒花庆祝动画效果。"],"string":"[\n \"检查基础UI布局与顶部信息显示：观察页面，验证页面中央是否存在一个背景为浅灰色的Canvas画布，且Canvas正上方有一层HTML界面，准确显示了当前关卡进度（例如'Level 1/3'）以及一个'重新开始本关'的按钮。\",\n \"检查地图静态元素渲染美化：观察Canvas内生成的迷宫地图，验证墙壁（如黑色）和道路（如白色）是否被渲染为圆角矩形，玩家（蓝色方块）和门障碍物是否附带有微小的发光或阴影效果，终点位置的绿色方块是否存在持续缩放的'呼吸'动画。\",\n \"检查基础滑动操作与撞墙机制：按下任一方向键（如向右），观察玩家蓝色方块是否沿着该方向匀速且持续地滑动；在滑动中途按下其他方向键，玩家是否保持原方向不变，直到撞击到墙壁障碍物才完全停止。\",\n \"检查滑动拖尾动效：在玩家长距离滑动时进行观测或截图，验证蓝色方块的后方是否动态生成了逐渐消散的残影或小粒子拖尾，以体现滑动速度感。\",\n \"检查重新开始功能：控制玩家随意滑动几个位置后，点击上方的'重新开始本关'按钮，验证地图和玩家位置是否瞬间恢复为该关卡的初始未解谜状态。\",\n \"检查无钥匙时的门阻挡判定：在玩家未拾取对应颜色钥匙的情况下，控制玩家滑向带有颜色边框的方块（门），验证该门是否像墙壁一样将玩家直接挡住并迫使玩家停下。\",\n \"检查拾取钥匙及UI更新逻辑：控制玩家滑行路线经过带有颜色的小圆圈（钥匙），验证碰到钥匙时，该钥匙是否呈现向上飘散并变透明的动效；同时验证上方的UI状态栏是否实时增加并显示了该对应钥匙的图标或列表数据。\",\n \"检查消耗钥匙的开门机制：在玩家拾取了指定颜色的钥匙后，滑向同色的边框门，验证门是否触发了向两侧碎裂或缩小消失的过渡动效；并且玩家方块是否未被阻挡、顺畅滑过原先门的位置，同时顶部UI对应扣除一把钥匙。\",\n \"检查过关判定与关卡切换机制：控制玩家滑动最终停在绿色终点方块上或滑动轨迹穿过终点，验证画面中央是否弹出'过关！'文本提示，且等待约1秒后，是否自动清除旧地图并加载渲染出下一个新关卡地图，上方UI的Level数字对应增加。\",\n \"检查摄像机平滑跟随与边界处理：进入最后一关（大型地图关卡），控制玩家滑动，验证Canvas画面视口是否能平滑跟随玩家移动并让玩家保持在画面中心区域；当玩家向大地图的最边缘移动时，验证摄像机是否停止跟随，确保视口内不露出大地图外围的背景空白区域，验证是否仅渲染摄像机视口内（加一点冗余边界）的地图元素。\",\n \"检查通关撒花特效：控制玩家通过全部关卡（达到最后一关终点）完成游戏，验证Canvas画布上是否生成且持续播放了全屏的纸屑（Confetti）撒花庆祝动画效果。\"\n]","isConversational":false}}},{"rowIdx":63,"cells":{"task_id":{"kind":"number","value":64,"string":"64"},"repository_policy":{"kind":"string","value":"请你在后续所有轮次的回复中，始终严格遵循以下约束要求：\n1. 不论何时，你生成用户回复时都必须使用中文。\n2. 无论生成多少个代码文件，其内部的注释内容均要求全是中文。\n3. 所有由你创建的代码文件，其编码必须统一为UTF-8。\n4. 请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下。\n5. 你在生成代码文件时，需将循环嵌套层数控制在 2 层以内。\n6. 确保你输出的代码项目文件夹嵌套深度不超过 2 层。\n7. 只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换。\n8. 在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的扩展名集合\n code_exts = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.java', \n '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.rb', '.php', \n '.swift', '.kt', '.sh', '.bash', '.m', '.scala', '.dart',\n '.json', '.xml', '.yaml', '.yml', '.sql', '.vue'\n }\n # 定义常见的不带扩展名或特殊扩展名的代码文件名\n code_filenames = {'Makefile', 'Dockerfile', 'CMakeLists.txt'}\n\n # 检查项目路径是否存在\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n\n # 遍历项目目录下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_exts or file in code_filenames:\n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 utf-8 编码读取文件全部内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明并非有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能导致读取失败的异常（如权限问题等）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_folder_depth(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 如果代码项目路径不存在，说明没有生成文件夹，符合深度不超过2层的要求\n if not os.path.exists(workspace_path):\n return True\n\n # 获取绝对路径，方便进行后续相对路径的计算\n workspace_path_obj = Path(workspace_path).resolve()\n\n # 遍历该项目路径下的所有目录及文件\n for root, dirs, files in os.walk(workspace_path):\n root_path_obj = Path(root).resolve()\n \n try:\n # 计算当前遍历到的目录相对于根项目目录的相对路径\n rel_path = root_path_obj.relative_to(workspace_path_obj)\n except ValueError:\n # 异常情况：如遇到不属于工作目录树的路径，跳过\n continue\n \n # relative_to 的 parts 属性会将相对路径拆解为元组形式\n # 例如: '.' 的 parts 为 ()，长度为 0\n # 'A' 的 parts 为 ('A',)，长度为 1\n # 'A/B' 的 parts 为 ('A', 'B')，长度为 2\n # 'A/B/C' 的 parts 为 ('A', 'B', 'C')，长度为 3\n depth = len(rel_path.parts)\n \n # 如果嵌套深度超过2，则代表不满足要求，直接返回 False\n if depth > 2:\n return False\n\n # 若所有目录层级都在要求范围内，返回 True\n return True","tags":["Quantity","Complexity"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]}],"string":"[\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的扩展名集合\\n code_exts = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.java', \\n '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.rb', '.php', \\n '.swift', '.kt', '.sh', '.bash', '.m', '.scala', '.dart',\\n '.json', '.xml', '.yaml', '.yml', '.sql', '.vue'\\n }\\n # 定义常见的不带扩展名或特殊扩展名的代码文件名\\n code_filenames = {'Makefile', 'Dockerfile', 'CMakeLists.txt'}\\n\\n # 检查项目路径是否存在\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历项目目录下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_exts or file in code_filenames:\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 utf-8 编码读取文件全部内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明并非有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能导致读取失败的异常（如权限问题等）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_folder_depth(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 如果代码项目路径不存在，说明没有生成文件夹，符合深度不超过2层的要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 获取绝对路径，方便进行后续相对路径的计算\\n workspace_path_obj = Path(workspace_path).resolve()\\n\\n # 遍历该项目路径下的所有目录及文件\\n for root, dirs, files in os.walk(workspace_path):\\n root_path_obj = Path(root).resolve()\\n \\n try:\\n # 计算当前遍历到的目录相对于根项目目录的相对路径\\n rel_path = root_path_obj.relative_to(workspace_path_obj)\\n except ValueError:\\n # 异常情况：如遇到不属于工作目录树的路径，跳过\\n continue\\n \\n # relative_to 的 parts 属性会将相对路径拆解为元组形式\\n # 例如: '.' 的 parts 为 ()，长度为 0\\n # 'A' 的 parts 为 ('A',)，长度为 1\\n # 'A/B' 的 parts 为 ('A', 'B')，长度为 2\\n # 'A/B/C' 的 parts 为 ('A', 'B', 'C')，长度为 3\\n depth = len(rel_path.parts)\\n \\n # 如果嵌套深度超过2，则代表不满足要求，直接返回 False\\n if depth > 2:\\n return False\\n\\n # 若所有目录层级都在要求范围内，返回 True\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用 HTML/CSS/JS 写一个管道拼接益智游戏的基础页面骨架。需要一个居中的游戏主区域（比如 6x6 的网格阵列布局），左侧或顶部放一个标题栏和控制面板区（放一个'开始'按钮），右侧或底部留一个计分/状态面板。先给网格里填满带有边框的默认灰色方块，代表占位的管道。实现简单的交互即可：鼠标悬浮方块时有背景色变化，点击任意方块时，方块有个轻微缩放回弹的点击反馈特效。具体管道形状和拼接逻辑咱们后续再做。另外后续所有轮次请你始终遵循以下两项要求：一是对于生成的代码文件，必须采用由 4 个空格构成的缩进方式；二是你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令推断，目标代码文件类型为 html, css, js\n target_extensions = {'.html', '.css', '.js'}\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n # 尝试用 utf-8 编码读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明不是规范的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能的读取异常（如权限问题等），仅专注于编码判断\n pass\n \n return has_code_files","tags":["Language","File Encoding"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import pathlib\n\n base_path = pathlib.Path(workspace_path).resolve()\n \n # 如果路径不存在或不是文件夹，并没有违反深度要求的情况发生\n if not base_path.exists() or not base_path.is_dir():\n return True\n \n # 遍历项目目录下的所有内容\n for current_path in base_path.rglob('*'):\n # 仅判断文件夹的嵌套深度\n if current_path.is_dir():\n try:\n # 获取相对路径\n rel_path = current_path.relative_to(base_path)\n # parts 属性可以获取路径所有的组成层级\n # 例如 'A/B' 的 parts 长度为 2\n if len(rel_path.parts) > 2:\n return False\n except ValueError:\n # 处理由于权限等其他原因导致无法计算相对路径的异常情况\n continue\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于生成的代码文件，必须采用由 4 个空格构成的缩进方式","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区不存在，说明没有生成文件，直接返回 True （无违反情况）\n if not os.path.isdir(workspace_path):\n return True\n\n # 根据用户指令推断可能生成的代码文件扩展名\n target_extensions = {'.html', '.htm', '.css', '.js'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except UnicodeDecodeError:\n # 如果不是普通的文本文件或不是 UTF-8 编码，跳过读取\n continue\n \n for line in lines:\n stripped_line = line.strip()\n # 跳过空行或无可见字符的行\n if not stripped_line:\n continue\n \n # 匹配行首的空白字符序列（空格或制表符）\n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent = match.group(1)\n \n # 不允许使用制表符（\\t）进行缩进\n if '\\t' in indent:\n return False\n \n # 缩进必须是 4 个空格的倍数\n if len(indent) % 4 != 0:\n # 豁免场景：多行注释的中间行往往以 \" * \" 形式对齐，行首会有非 4 倍数的空格\n if stripped_line.startswith('*'):\n continue\n \n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n has_files = False\n \n # 推断可能的代码文件类型：益智游戏骨架通常包含这三种前端文件\n exts = ('.html', '.css', '.js')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(exts):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n has_files = True\n except Exception:\n continue\n \n # 匹配所有由非空白、非字母数字和非汉字组成的连续“标点符号块”\n # re的默认\\w涵盖了Unicode字母、数字和中文字符，此处只提取符号组成的连续块\n for match in re.finditer(r'[^\\s\\w]+', content):\n token = match.group()\n \n # 若标点块中包含 '='，则需要审查其两边的空格数量\n if '=' in token:\n start, end = match.span()\n \n # 统计紧贴标点块左侧的连续普通空格数\n left_spaces = 0\n for i in range(start - 1, -1, -1):\n if content[i] == ' ':\n left_spaces += 1\n elif content[i] in ('\\t', '\\n', '\\r'):\n break\n else:\n break\n \n # 统计紧贴标点块右侧的连续普通空格数\n right_spaces = 0\n for i in range(end, len(content)):\n if content[i] == ' ':\n right_spaces += 1\n elif content[i] in ('\\t', '\\n', '\\r'):\n break\n else:\n break\n \n # 按照要求，任何包含 \"=\" 的地方其两边必须各有两个空格\n if left_spaces != 2 or right_spaces != 2:\n return False\n\n # 若未生成任何目标类型的代码文件，视为未满足要求；若全都合规则返回 True\n return has_files","tags":["Style","Layout"]}]},{"round_id":1,"instruction":"很不错，骨架有了。现在我们把占位方块替换成具有方向性的真实管道。请设计5种基础管道形态：起点（单向）、终点（单向）、直管（双向通）、弯管（L型双向通）和十字交叉管（四向通）。你可以用纯 CSS 伪元素或内联 SVG 在方块中清晰绘制出它们。在初始化时，给 6x6 网格随机分配这些管道。接着，实现核心旋转交互：玩家点击任意管道（除了起点和终点）时，该管道顺时针旋转 90 度，要求必须带有流畅的 CSS 平滑旋转过渡动画（如0.3秒）。另外后续所有轮次请你始终遵循两项更新要求：一是生成的代码文件应全部使用2个空格来控制缩进，替换此前4个空格缩进的要求；二是在输出的任何代码文件中，每行的长度必须限制在100个字符以内。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 结合用户指令推断，常见的代码文件包括前端类型文件\n code_extensions = {'.html', '.htm', '.css', '.js', '.json', '.svg'}\n \n # 如果项目路径不存在，视作没有违反该限制要求（或者尚未创建文件）\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤出代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n \n # 尝试用 utf-8 编码读取文件\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明包含不符合 utf-8 编码的字符\n return False\n except Exception:\n # 其他异常（例如权限问题）跳过不作处理\n pass\n \n # 所有代码文件均成功使用 utf-8 读取，则满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def verify_folder_depth(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，说明没有生成任何需要嵌套的文件，自然也就没有超过2层\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 计算当前目录相对于工作区根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 将相对路径拆分成具体的层级列表，并过滤掉空字符串或当前目录符号 \".\"\n parts = [p for p in rel_path.split(os.sep) if p and p != '.']\n \n # parts的长度即代表当前的文件夹嵌套深度\n depth = len(parts)\n \n # 如果嵌套深度超过2，则不满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的代码文件应全部使用2个空格来控制缩进","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令推断，主要验证页面及常见前端脚本代码文件\n if file.endswith(('.html', '.css', '.js', '.ts', '.jsx', '.tsx')):\n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n\n count_2 = 0\n count_4 = 0\n prev_spaces = 0\n has_tab = False\n\n for line in lines:\n stripped = line.lstrip(' \\t\\n\\r')\n # 忽略空行\n if not stripped:\n continue\n \n # 忽略以 '*' 开头的行，因为它们经常用于多行注释（如 /** ... */）中的特殊对齐\n if stripped.startswith('*'):\n continue\n\n # 提取行首所有的空白字符\n leading_whitespace = \"\"\n for char in line:\n if char in (' ', '\\t'):\n leading_whitespace += char\n else:\n break\n\n # 不允许使用制表符来控制缩进\n if '\\t' in leading_whitespace:\n has_tab = True\n break\n\n curr_spaces = len(leading_whitespace)\n\n # 如果当前行比上一行有更多的前导空格，说明发生了代码块层级进入，计算其缩进增量\n if curr_spaces > prev_spaces:\n diff = curr_spaces - prev_spaces\n if diff == 2:\n count_2 += 1\n elif diff == 4:\n count_4 += 1\n \n prev_spaces = curr_spaces\n\n if has_tab:\n return False\n\n # 判定规则1：如果文件中存在递增操作，且完全由4个空格构成，则说明是用4个空格控制缩进，违反要求\n if count_4 > 0 and count_2 == 0:\n return False\n \n # 判定规则2：如果4个空格构成的递增操作远多于2个空格，说明主流仍是4个空格，也算违反要求\n if count_4 > 0 and count_4 > count_2 * 2:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略无法用 utf-8 解码读取的二进制文件或非文本文件\n continue\n \n for i, char in enumerate(content):\n if char == '=':\n # 检查等号左侧是否恰好各有两个空格\n if i < 2 or content[i-2:i] != ' ':\n return False\n # 确保左侧第 3 个字符不是空格（即不能超过 2 个空格）\n if i >= 3 and content[i-3] == ' ':\n return False\n \n # 检查等号右侧是否恰好各有两个空格\n if i + 3 > len(content) or content[i+1:i+3] != ' ':\n return False\n # 确保右侧第 3 个字符不是空格（即不能超过 2 个空格）\n if i + 3 < len(content) and content[i+3] == ' ':\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"在输出的任何代码文件中，每行的长度必须限制在100个字符以内","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令内容，推断代码文件类型主要是 HTML、CSS 和 JS\n code_extensions = {'.html', '.css', '.js'}\n \n # 确保当前项目地址是一个有效的目录\n if not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选对应的代码文件后缀\n if any(file.lower().endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾换行符进行字符数计算\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 100:\n return False\n except Exception:\n # 遇到文件读取异常直接跳过\n continue\n \n return True","tags":["Quantity","Range"]}]},{"round_id":2,"instruction":"好，接下来挑战最核心的连通算法逻辑。请在控制面板点击‘开始’按钮后，执行连通性检测。从‘起点’开始，判断其开口方向是否与相邻管道的开口相匹配对接（比如左边管道的右开口必须对准右边管道的左开口）。顺着物理连通的路径，将所有成功连通的管道状态标记为‘激活’，通过 CSS 改变激活管道的颜色（比如把管道内部填充成蓝色代表通水）。如果遇到未匹配的端口或到达边界，水流就在那里停止。注意：算法必须严格读取当前每个管道由于旋转而产生的真实端口方向。另外后续所有轮次请你始终遵循两项要求：一是在输出的任何代码文件中，每行的长度必须限制在50个字符以内，替换此前100个字符的要求；二是必须保证你生成的代码文件中完全避免使用递归。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令内容推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用 utf-8 解码，以此验证文件是否符合 UTF-8 编码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明存在非 UTF-8 编码的内容\n return False\n except Exception:\n # 忽略读取文件等过程中可能的其他系统层面异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_folder_depth(response: str, workspace_path: str) -> bool:\n import os\n import pathlib\n\n # 遍历当前工作区的所有目录与文件\n for root, dirs, files in os.walk(workspace_path):\n # 计算当前遍历到的文件夹相对于项目根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 判断文件夹嵌套深度\n if rel_path == '.':\n depth = 0\n else:\n # 使用 pathlib 解析相对路径的层级数\n depth = len(pathlib.Path(rel_path).parts)\n \n # 如果嵌套深度超过 2 层，则不满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的代码文件应全部使用2个空格来控制缩进","validation_code":"def check_indentation(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求定义可能生成的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 出现编码错误等无法读取的情况（违反统一为UTF-8的潜在要求），直接判定False\n return False\n \n spaces_set = set()\n \n for line in lines:\n # 去除行尾部的换行符及空白\n line = line.rstrip('\\r\\n')\n \n # 提取前导空白字符\n leading_whitespace = \"\"\n for char in line:\n if char in [' ', '\\t']:\n leading_whitespace += char\n else:\n break\n \n content = line[len(leading_whitespace):]\n \n # 忽略空行\n if not content.strip():\n continue\n \n # 1. 严禁使用制表符（\\t）进行缩进控制\n if '\\t' in leading_whitespace:\n return False\n \n space_count = len(leading_whitespace)\n \n # 2. 检查是否为奇数个前导空格\n if space_count % 2 != 0:\n # 仅放宽常见的注释及字符串对齐情况\n if (content.startswith('*') or content.startswith('/*') or \n content.startswith('//') or content.startswith('<!--') or \n content.startswith('`') or content.startswith(\"'\") or \n content.startswith('\"')):\n continue\n else:\n # 存在非注释类的奇数空格缩进，违背2个空格的标准\n return False\n \n spaces_set.add(space_count)\n \n non_zero_spaces = {s for s in spaces_set if s > 0}\n \n # 3. 区分是否仍在沿用4个空格缩进\n # 如果某文件内所有的非零缩进全都是4的倍数，说明未使用2个空格缩进\n if non_zero_spaces:\n if all(s % 4 == 0 for s in non_zero_spaces):\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"def check_equals_spacing(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令推断的代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n file_found = False\n \n # 遍历项目文件夹\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 遍历文件内容中的所有字符\n for i, char in enumerate(content):\n if char == '=':\n # 提取 '=' 左边和右边的两个字符\n left_spaces = content[max(0, i-2):i]\n right_spaces = content[i+1:i+3]\n \n # 判断两边是否都各有两个空格\n if left_spaces != ' ' or right_spaces != ' ':\n return False\n\n # 若未找到任何相关代码文件，则也视为不符合要求\n return file_found","tags":["Style","Layout"]},{"constraint":"在输出的任何代码文件中，每行的长度必须限制在50个字符以内","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀类型\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.json', '.py', '.java', '.c', '.cpp', '.h'}\n \n if not os.path.exists(workspace_path):\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的换行符后计算长度\n clean_line = line.rstrip('\\n\\r')\n if len(clean_line) > 50:\n return False\n except Exception:\n # 如果由于读取权限等原因无法处理，则跳过该文件继续检查\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"必须保证你生成的代码文件中完全避免使用递归","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":3,"instruction":"现在的连通是瞬间完成的，不够生动。下一步，请将‘瞬间变蓝’升级为带有传播逻辑的‘逐级流动’动画。水流需要从起点开始，根据连通路径一格接一格地向后延伸，每格之间有例如 0.3 秒的传播延迟。同时加入动态的‘连击(Combo)’机制：如果水流正在按延迟一格格蔓延时，玩家快速点击旋转了前方原本未接通的死胡同管道，使得水流没有中断并继续向前流，就在那个刚接通的管道上方触发一个向上飘动的‘Combo +1!’文字特效，并在计分板上增加相应分数。另外后续所有轮次请你始终遵循两项新要求：一是你的代码文件生成时必须用到递归技术，替换此前完全避免使用递归的要求；二是在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(后续优化)\"标记，不可多也不可少。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef verify_utf8_encoding(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证由人工智能助手生成的所有代码文件是否统一为 UTF-8 编码。\n \"\"\"\n # 根据用户指令推断，代码文件主要为 HTML、CSS 和 JS\n code_extensions = {'.html', '.htm', '.css', '.js'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦抛出解码错误，说明存在非 UTF-8 编码的文件\n return False\n except Exception:\n # 对于其他不可预见的读取异常，保守判断为不符合\n return False\n \n # 所有代码文件均成功通过 UTF-8 编解码测试\n return True","tags":["Language","File Encoding"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果给定的工作空间路径不存在或不是一个目录，则没有违反嵌套深度的规定\n if not os.path.isdir(workspace_path):\n return True\n \n base_path = os.path.abspath(workspace_path)\n \n for root, dirs, files in os.walk(base_path):\n # 就地修改 dirs 列表，过滤掉所有以 '.' 开头的隐藏文件夹\n # 这将使 os.walk 自动跳过这些目录及其内部子目录\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 计算当前所处目录相对于根目录的相对路径\n rel_path = os.path.relpath(root, base_path)\n \n if rel_path == '.':\n depth = 0\n else:\n # 将不同系统的路径分隔符统一替换为 '/'，过滤空字符串后计算路径包含的层级数\n parts = [p for p in rel_path.replace('\\\\', '/').split('/') if p]\n depth = len(parts)\n \n # 判断文件夹嵌套深度是否超过 2 层\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的代码文件应全部使用2个空格来控制缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 确保当前项目地址存在\n if not os.path.exists(workspace_path):\n return False\n\n # 常见可能生成的代码文件扩展名\n code_exts = {'.js', '.css', '.html', '.htm', '.py', '.c', '.cpp', '.java', '.ts', '.jsx', '.tsx', '.vue'}\n \n found_code_files = False\n has_valid_indent = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_files = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n leading_spaces = []\n for line in lines:\n # 提取该行的前导空白字符\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 任何一行如果使用了制表符缩进，即违背了“空格缩进”的规定\n if '\\t' in leading_whitespace:\n return False\n \n stripped = line.lstrip()\n # 忽略空行和可能打乱整体统计结构的多行注释标识行\n if stripped and not stripped.startswith('*') and not stripped.startswith('<!--') and not stripped.startswith('/*') and not stripped.startswith('*/'):\n leading_spaces.append(len(leading_whitespace))\n \n if not leading_spaces:\n continue\n \n diffs = []\n for i in range(1, len(leading_spaces)):\n diff = leading_spaces[i] - leading_spaces[i-1]\n if diff > 0:\n diffs.append(diff)\n \n # 检查是否存在2个空格缩进专属的特征（即除以4余2，如2, 6, 10）\n has_2_space = any(s % 4 == 2 for s in leading_spaces)\n # 检查是否存在4个空格倍数的缩进特征\n has_4_space = any(s % 4 == 0 and s > 0 for s in leading_spaces)\n \n count_2 = diffs.count(2)\n count_4 = diffs.count(4)\n \n # 若文件包含任何2个空格的缩进特征，记录当前项目有效满足缩进要求的证据\n if has_2_space or count_2 > 0:\n has_valid_indent = True\n \n # 如果某文件内有明显的4空格规律，却毫无2空格规则的痕迹，说明其并未换用2个空格\n if (has_4_space or count_4 > 0) and not has_2_space and count_2 == 0:\n return False\n \n # 如果文件由于不规范的折行或其他原因，导致出现了太多奇数个空格缩进，说明没有正确按2个空格规律\n odd_spaces = sum(1 for s in leading_spaces if s % 2 != 0)\n if odd_spaces > len(leading_spaces) * 0.3:\n return False\n \n # 最终条件：发现存在代码文件，并且项目中至少有一处确凿的2空格缩进特征\n return found_code_files and has_valid_indent","tags":["Style","Layout"]},{"constraint":"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"def check_equals_spacing(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return False\n \n has_equals = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # 尝试按 UTF-8 读取文本文件，跳过可能的二进制或不可读文件\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n count_all = content.count('=')\n \n if count_all > 0:\n has_equals = True\n \n # 正则模式：匹配前后恰好各有两个空格的等号\n # (?<! ) : 前面紧挨着的字符不能是空格（确保左边不多于两个空格）\n # = : 等号及前后各两个空格\n # (?! ) : 后面紧挨着的字符不能是空格（确保右边不多于两个空格）\n count_valid = len(re.findall(r'(?<! ) = (?! )', content))\n \n # 如果符合规范的等号数量不等于总等号数量，则代表有等号违反了规则\n if count_all != count_valid:\n return False\n \n # 如果项目中至少存在一个等号且所有等号都符合规则，则为True；若无任何等号出现，视为未有效生成项目\n return has_equals","tags":["Style","Layout"]},{"constraint":"在输出的任何代码文件中，每行的长度必须限制在50个字符以内","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对用户的游戏网页项目，定义可能涉及的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.sh', '.bash'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件后缀名并进行小写匹配\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除由于换行产生的回车符或换行符\n clean_line = line.rstrip('\\r\\n')\n # 要求中明确指出了“字符”，直接采用 len 函数统计字符数量\n if len(clean_line) > 50:\n return False\n except Exception:\n # 如果某些文件并非合法的纯文本文件或遇到读取异常，则忽略跳过\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"你的代码文件生成时必须用到递归技术","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 3 个 \"TODO(后续优化)\"标记，不可多也不可少","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对前端项目推断出的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n found_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 遇到文件无法读取的情况，保守起见视作内容不合格或直接跳过\n continue\n \n # “各个代码文件中”意味着每一个代码文件都必须刚好包含 3 个标记\n if content.count(\"TODO(后续优化)\") != 3:\n return False\n\n # 必须至少生成了 1 个代码文件且全都符合要求，才算总体满足条件\n return found_code_file","tags":["Content","Keyword"]}]},{"round_id":4,"instruction":"最后一步，我们来完善视觉模式切换和游戏循环机制。首先，在控制面板加一个‘霓虹模式’的切换开关。开启后，整个页面背景变暗，所有被水流激活的蓝色管道会散发出明亮的赛博朋克风格发光特效（利用 CSS drop-shadow 或 box-shadow 实现呼吸灯效果）。其次，增加胜利判定：当水流最终成功流入并点亮‘终点’管道时，游戏暂停操作，并在页面中央弹出一个带有炫酷光环效果的‘Level Clear!’过关弹窗。弹窗内提供一个‘下一关’按钮，点击后清空连通状态，并重新随机生成一盘全新的管道网格布局。另外后续所有轮次请你始终遵循两项更新要求：一是在你生成的各个代码文件中，必须包含正好 2 个 \"FIXME(待调试)\"标记，不可多也不可少，替换此前要求包含3个\"TODO(后续优化)\"标记的要求；二是你的每一条回复内容，都必须以\"呀～\"作为结尾语气词。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 如果工作区路径不存在，说明没有创建文件，暂不计为违规\n if not os.path.exists(workspace_path):\n return True\n\n # 根据用户指令推测可能的代码文件后缀名\n code_extensions = {'.html', '.css', '.js'}\n \n # 遍历项目目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名并转换为小写\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如无权限等其他非编码相关的系统级读取错误\n pass\n \n # 如果所有的代码文件都能成功以 UTF-8 解码，说明满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def verify_folder_depth(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 检查工作空间路径是否存在，如果不存在则自然没有超出嵌套深度\n if not os.path.exists(workspace_path):\n return True\n \n workspace = Path(workspace_path).resolve()\n \n # 遍历代码项目下的所有文件及目录\n for root, dirs, files in os.walk(workspace_path):\n root_path = Path(root).resolve()\n \n try:\n # 计算当前遍历到的目录相对于工作区根目录的相对路径\n rel_path = root_path.relative_to(workspace)\n except ValueError:\n # 如果路径解析出现异常，跳过当前项\n continue\n \n # 计算相对路径的层级数\n # '.' 的 parts 长度为 0 （代表 workspace_path 本身）\n # 'dir1' 的 parts 长度为 1 （代表第 1 层目录）\n # 'dir1/dir2' 的 parts 长度为 2 （代表第 2 层目录）\n # 'dir1/dir2/dir3' 的 parts 长度为 3 （代表超过了 2 层的嵌套）\n if len(rel_path.parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的代码文件应全部使用2个空格来控制缩进","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n \n # 检查项目目录是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义前端代码相关的文件扩展名\n valid_exts = ('.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.json', '.vue', '.htm')\n valid_files_checked = 0\n has_indentation = False\n \n # 遍历当前生成项目中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith(valid_exts):\n filepath = os.path.join(root, file)\n valid_files_checked += 1\n \n # 统一以 utf-8 读取代码文件并忽略解码错误以防崩溃\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n \n prev_indent = 0\n diffs = []\n tab_lines = 0\n \n for line in lines:\n stripped = line.strip()\n # 过滤掉空行\n if not stripped:\n continue\n \n # 若行首为制表符，视为使用了 Tab 缩进\n if line.startswith('\\t'):\n tab_lines += 1\n prev_indent = 0\n continue\n \n # 计算该行的前导空格数量\n indent = len(line) - len(line.lstrip(' '))\n \n # 只有在发生正向缩进（向深层缩进）时，我们才收集其增量差值\n if indent > prev_indent:\n diffs.append(indent - prev_indent)\n \n prev_indent = indent\n \n # 若文件包含使用制表符缩进的行，违反“全部使用2个空格”的规则\n if tab_lines > 0:\n return False\n \n if diffs:\n has_indentation = True\n # 使用 Counter 寻找最常见的缩进增量（众数）\n c = Counter(diffs)\n max_count = max(c.values())\n modes = [k for k, v in c.items() if v == max_count]\n \n # 如果最常用的缩进增加幅度中不包含 2（例如只包含 4）\n # 说明该文件主要不依靠 2个空格 来控制层级缩进\n if 2 not in modes:\n return False\n\n # 若未找到任何可分析的代码文件，默认验证不通过\n if valid_files_checked == 0:\n return False\n \n # 若所有的文件代码连一级缩进都没有，也无法佐证是否遵循要求\n if not has_indentation:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目性质推断常见的代码文件扩展名\n code_extensions = ['.html', '.css', '.js']\n code_files = []\n \n if not os.path.exists(workspace_path):\n return False\n \n # 查找所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果没有生成任何代码文件，则视为没有完成代码任务，返回 False\n if not code_files:\n return False\n \n # 逐文件检查 '=' 两侧的空格情况\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n for i, char in enumerate(content):\n if char == '=':\n # 检查左边是否恰好有两个空格\n # 1. 至少需要前两个字符是空格\n if i < 2 or content[i-1] != ' ' or content[i-2] != ' ':\n return False\n # 2. 前面第三个字符不能也是空格\n if i >= 3 and content[i-3] == ' ':\n return False\n \n # 检查右边是否恰好有两个空格\n # 1. 至少需要后两个字符是空格\n if i > len(content) - 3 or content[i+1] != ' ' or content[i+2] != ' ':\n return False\n # 2. 后面第三个字符不能也是空格\n if i <= len(content) - 4 and content[i+3] == ' ':\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"在输出的任何代码文件中，每行的长度必须限制在50个字符以内","validation_code":"def check_line_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区目录不存在，视作未满足要求\n if not os.path.exists(workspace_path):\n return False\n\n # 定义可能的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.py', '.sh', '.bash', '.java', '.c', '.cpp', '.h', '.json'\n }\n\n # 要忽略的常见依赖及版本控制目录，避免深入遍历并防止干扰\n ignore_dirs = {'node_modules', '.git', '.svn', 'dist', 'build'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以剪去不需要遍历的子目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理属于代码文件范畴的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 兼容带有 BOM 头的 UTF-8 文件\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n for line in f:\n # 剔除行尾的回车与换行符，其余空格等字符予以保留\n line_content = line.rstrip('\\r\\n')\n # 检查字符数是否超过 50 个字符的上限\n if len(line_content) > 50:\n return False\n except Exception:\n # 读取出错时跳过（例如碰到二进制文件等异常情况）\n continue\n\n # 遍历结束未发现超出 50 个字符的代码行，返回 True \n return True","tags":["Quantity","Range"]},{"constraint":"你的代码文件生成时必须用到递归技术","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的各个代码文件中，必须包含正好 2 个 \"FIXME(待调试)\"标记，不可多也不可少","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n target_mark = \"FIXME(待调试)\"\n # 根据用户指令中提到的 HTML/CSS/JS 推断，代码文件大概率是前端项目相关后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.vue', '.jsx', '.tsx'}\n \n file_checked = False\n \n # 1. 优先检查工作区是否存在生成的代码文件\n if workspace_path and os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查每个代码文件是否恰好包含 2 个目标标记，不可多也不可少\n if content.count(target_mark) != 2:\n return False\n \n file_checked = True\n except Exception:\n pass\n \n # 若在工作区找到了对应代码文件，并均满足条件，则验证成功\n if file_checked:\n return True\n \n # 2. 如果工作区为空或未检测到实体代码文件，从 response 中提取 Markdown 代码块作为代码文件进行检测\n # 匹配带有 HTML/CSS/JS 等前端常用语言标签的代码块\n pattern = r'```(?:html|css|javascript|js|vue|ts|jsx|tsx)?\\s*\\n(.*?)\\n```'\n code_blocks = re.findall(pattern, response, re.IGNORECASE | re.DOTALL)\n \n if not code_blocks:\n # 如果未匹配到，尝试匹配无语言标签的代码块\n code_blocks = re.findall(r'```\\s*\\n(.*?)\\n```', response, re.DOTALL)\n \n # 如果既没有实体代码文件，也没有文本代码块，说明未生成代码文件，自然不满足要求\n if not code_blocks:\n return False\n \n # 检查每一个代码块中是否也恰好包含 2 个目标标记\n for block in code_blocks:\n if block.count(target_mark) != 2:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你的每一条回复内容，都必须以\"呀～\"作为结尾语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除回复末尾可能存在的空白字符以及常见中英文句末标点符号\n cleaned_response = re.sub(r'[\\s。！？!?,，]+$', '', response)\n \n # 检查清理末尾标点后的文本是否以指定的语气词“呀～”结尾\n return cleaned_response.endswith('呀～')","tags":["Content","Persona"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用 HTML/CSS/JS 写一个管道拼接益智游戏的基础页面骨架。需要一个居中的游戏主区域（比如 6x6 的网格阵列布局），左侧或顶部放一个标题栏和控制面板区（放一个'开始'按钮），右侧或底部留一个计分/状态面板。先给网格里填满带有边框的默认灰色方块，代表占位的管道。实现简单的交互即可：鼠标悬浮方块时有背景色变化，点击任意方块时，方块有个轻微缩放回弹的点击反馈特效。具体管道形状和拼接逻辑咱们后续再做。另外后续所有轮次请你始终遵循以下两项要求：一是对于生成的代码文件，必须采用由 4 个空格构成的缩进方式；二是你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令推断，目标代码文件类型为 html, css, js\\n target_extensions = {'.html', '.css', '.js'}\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试用 utf-8 编码读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明不是规范的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能的读取异常（如权限问题等），仅专注于编码判断\\n pass\\n \\n return has_code_files\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import pathlib\\n\\n base_path = pathlib.Path(workspace_path).resolve()\\n \\n # 如果路径不存在或不是文件夹，并没有违反深度要求的情况发生\\n if not base_path.exists() or not base_path.is_dir():\\n return True\\n \\n # 遍历项目目录下的所有内容\\n for current_path in base_path.rglob('*'):\\n # 仅判断文件夹的嵌套深度\\n if current_path.is_dir():\\n try:\\n # 获取相对路径\\n rel_path = current_path.relative_to(base_path)\\n # parts 属性可以获取路径所有的组成层级\\n # 例如 'A/B' 的 parts 长度为 2\\n if len(rel_path.parts) > 2:\\n return False\\n except ValueError:\\n # 处理由于权限等其他原因导致无法计算相对路径的异常情况\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，必须采用由 4 个空格构成的缩进方式\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区不存在，说明没有生成文件，直接返回 True （无违反情况）\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 根据用户指令推断可能生成的代码文件扩展名\\n target_extensions = {'.html', '.htm', '.css', '.js'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except UnicodeDecodeError:\\n # 如果不是普通的文本文件或不是 UTF-8 编码，跳过读取\\n continue\\n \\n for line in lines:\\n stripped_line = line.strip()\\n # 跳过空行或无可见字符的行\\n if not stripped_line:\\n continue\\n \\n # 匹配行首的空白字符序列（空格或制表符）\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent = match.group(1)\\n \\n # 不允许使用制表符（\\\\t）进行缩进\\n if '\\\\t' in indent:\\n return False\\n \\n # 缩进必须是 4 个空格的倍数\\n if len(indent) % 4 != 0:\\n # 豁免场景：多行注释的中间行往往以 \\\" * \\\" 形式对齐，行首会有非 4 倍数的空格\\n if stripped_line.startswith('*'):\\n continue\\n \\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n has_files = False\\n \\n # 推断可能的代码文件类型：益智游戏骨架通常包含这三种前端文件\\n exts = ('.html', '.css', '.js')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(exts):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n has_files = True\\n except Exception:\\n continue\\n \\n # 匹配所有由非空白、非字母数字和非汉字组成的连续“标点符号块”\\n # re的默认\\\\w涵盖了Unicode字母、数字和中文字符，此处只提取符号组成的连续块\\n for match in re.finditer(r'[^\\\\s\\\\w]+', content):\\n token = match.group()\\n \\n # 若标点块中包含 '='，则需要审查其两边的空格数量\\n if '=' in token:\\n start, end = match.span()\\n \\n # 统计紧贴标点块左侧的连续普通空格数\\n left_spaces = 0\\n for i in range(start - 1, -1, -1):\\n if content[i] == ' ':\\n left_spaces += 1\\n elif content[i] in ('\\\\t', '\\\\n', '\\\\r'):\\n break\\n else:\\n break\\n \\n # 统计紧贴标点块右侧的连续普通空格数\\n right_spaces = 0\\n for i in range(end, len(content)):\\n if content[i] == ' ':\\n right_spaces += 1\\n elif content[i] in ('\\\\t', '\\\\n', '\\\\r'):\\n break\\n else:\\n break\\n \\n # 按照要求，任何包含 \\\"=\\\" 的地方其两边必须各有两个空格\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n\\n # 若未生成任何目标类型的代码文件，视为未满足要求；若全都合规则返回 True\\n return has_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"很不错，骨架有了。现在我们把占位方块替换成具有方向性的真实管道。请设计5种基础管道形态：起点（单向）、终点（单向）、直管（双向通）、弯管（L型双向通）和十字交叉管（四向通）。你可以用纯 CSS 伪元素或内联 SVG 在方块中清晰绘制出它们。在初始化时，给 6x6 网格随机分配这些管道。接着，实现核心旋转交互：玩家点击任意管道（除了起点和终点）时，该管道顺时针旋转 90 度，要求必须带有流畅的 CSS 平滑旋转过渡动画（如0.3秒）。另外后续所有轮次请你始终遵循两项更新要求：一是生成的代码文件应全部使用2个空格来控制缩进，替换此前4个空格缩进的要求；二是在输出的任何代码文件中，每行的长度必须限制在100个字符以内。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 结合用户指令推断，常见的代码文件包括前端类型文件\\n code_extensions = {'.html', '.htm', '.css', '.js', '.json', '.svg'}\\n \\n # 如果项目路径不存在，视作没有违反该限制要求（或者尚未创建文件）\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤出代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n \\n # 尝试用 utf-8 编码读取文件\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明包含不符合 utf-8 编码的字符\\n return False\\n except Exception:\\n # 其他异常（例如权限问题）跳过不作处理\\n pass\\n \\n # 所有代码文件均成功使用 utf-8 读取，则满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def verify_folder_depth(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，说明没有生成任何需要嵌套的文件，自然也就没有超过2层\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 计算当前目录相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 将相对路径拆分成具体的层级列表，并过滤掉空字符串或当前目录符号 \\\".\\\"\\n parts = [p for p in rel_path.split(os.sep) if p and p != '.']\\n \\n # parts的长度即代表当前的文件夹嵌套深度\\n depth = len(parts)\\n \\n # 如果嵌套深度超过2，则不满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用2个空格来控制缩进\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令推断，主要验证页面及常见前端脚本代码文件\\n if file.endswith(('.html', '.css', '.js', '.ts', '.jsx', '.tsx')):\\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n\\n count_2 = 0\\n count_4 = 0\\n prev_spaces = 0\\n has_tab = False\\n\\n for line in lines:\\n stripped = line.lstrip(' \\\\t\\\\n\\\\r')\\n # 忽略空行\\n if not stripped:\\n continue\\n \\n # 忽略以 '*' 开头的行，因为它们经常用于多行注释（如 /** ... */）中的特殊对齐\\n if stripped.startswith('*'):\\n continue\\n\\n # 提取行首所有的空白字符\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_whitespace += char\\n else:\\n break\\n\\n # 不允许使用制表符来控制缩进\\n if '\\\\t' in leading_whitespace:\\n has_tab = True\\n break\\n\\n curr_spaces = len(leading_whitespace)\\n\\n # 如果当前行比上一行有更多的前导空格，说明发生了代码块层级进入，计算其缩进增量\\n if curr_spaces > prev_spaces:\\n diff = curr_spaces - prev_spaces\\n if diff == 2:\\n count_2 += 1\\n elif diff == 4:\\n count_4 += 1\\n \\n prev_spaces = curr_spaces\\n\\n if has_tab:\\n return False\\n\\n # 判定规则1：如果文件中存在递增操作，且完全由4个空格构成，则说明是用4个空格控制缩进，违反要求\\n if count_4 > 0 and count_2 == 0:\\n return False\\n \\n # 判定规则2：如果4个空格构成的递增操作远多于2个空格，说明主流仍是4个空格，也算违反要求\\n if count_4 > 0 and count_4 > count_2 * 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略无法用 utf-8 解码读取的二进制文件或非文本文件\\n continue\\n \\n for i, char in enumerate(content):\\n if char == '=':\\n # 检查等号左侧是否恰好各有两个空格\\n if i < 2 or content[i-2:i] != ' ':\\n return False\\n # 确保左侧第 3 个字符不是空格（即不能超过 2 个空格）\\n if i >= 3 and content[i-3] == ' ':\\n return False\\n \\n # 检查等号右侧是否恰好各有两个空格\\n if i + 3 > len(content) or content[i+1:i+3] != ' ':\\n return False\\n # 确保右侧第 3 个字符不是空格（即不能超过 2 个空格）\\n if i + 3 < len(content) and content[i+3] == ' ':\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在输出的任何代码文件中，每行的长度必须限制在100个字符以内\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令内容，推断代码文件类型主要是 HTML、CSS 和 JS\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 确保当前项目地址是一个有效的目录\\n if not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选对应的代码文件后缀\\n if any(file.lower().endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾换行符进行字符数计算\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 100:\\n return False\\n except Exception:\\n # 遇到文件读取异常直接跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，接下来挑战最核心的连通算法逻辑。请在控制面板点击‘开始’按钮后，执行连通性检测。从‘起点’开始，判断其开口方向是否与相邻管道的开口相匹配对接（比如左边管道的右开口必须对准右边管道的左开口）。顺着物理连通的路径，将所有成功连通的管道状态标记为‘激活’，通过 CSS 改变激活管道的颜色（比如把管道内部填充成蓝色代表通水）。如果遇到未匹配的端口或到达边界，水流就在那里停止。注意：算法必须严格读取当前每个管道由于旋转而产生的真实端口方向。另外后续所有轮次请你始终遵循两项要求：一是在输出的任何代码文件中，每行的长度必须限制在50个字符以内，替换此前100个字符的要求；二是必须保证你生成的代码文件中完全避免使用递归。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令内容推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用 utf-8 解码，以此验证文件是否符合 UTF-8 编码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明存在非 UTF-8 编码的内容\\n return False\\n except Exception:\\n # 忽略读取文件等过程中可能的其他系统层面异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_folder_depth(response: str, workspace_path: str) -> bool:\\n import os\\n import pathlib\\n\\n # 遍历当前工作区的所有目录与文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 计算当前遍历到的文件夹相对于项目根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 判断文件夹嵌套深度\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 使用 pathlib 解析相对路径的层级数\\n depth = len(pathlib.Path(rel_path).parts)\\n \\n # 如果嵌套深度超过 2 层，则不满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用2个空格来控制缩进\",\n \"validation_code\": \"def check_indentation(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求定义可能生成的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 出现编码错误等无法读取的情况（违反统一为UTF-8的潜在要求），直接判定False\\n return False\\n \\n spaces_set = set()\\n \\n for line in lines:\\n # 去除行尾部的换行符及空白\\n line = line.rstrip('\\\\r\\\\n')\\n \\n # 提取前导空白字符\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char in [' ', '\\\\t']:\\n leading_whitespace += char\\n else:\\n break\\n \\n content = line[len(leading_whitespace):]\\n \\n # 忽略空行\\n if not content.strip():\\n continue\\n \\n # 1. 严禁使用制表符（\\\\t）进行缩进控制\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n space_count = len(leading_whitespace)\\n \\n # 2. 检查是否为奇数个前导空格\\n if space_count % 2 != 0:\\n # 仅放宽常见的注释及字符串对齐情况\\n if (content.startswith('*') or content.startswith('/*') or \\n content.startswith('//') or content.startswith('<!--') or \\n content.startswith('`') or content.startswith(\\\"'\\\") or \\n content.startswith('\\\"')):\\n continue\\n else:\\n # 存在非注释类的奇数空格缩进，违背2个空格的标准\\n return False\\n \\n spaces_set.add(space_count)\\n \\n non_zero_spaces = {s for s in spaces_set if s > 0}\\n \\n # 3. 区分是否仍在沿用4个空格缩进\\n # 如果某文件内所有的非零缩进全都是4的倍数，说明未使用2个空格缩进\\n if non_zero_spaces:\\n if all(s % 4 == 0 for s in non_zero_spaces):\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"def check_equals_spacing(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令推断的代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n file_found = False\\n \\n # 遍历项目文件夹\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 遍历文件内容中的所有字符\\n for i, char in enumerate(content):\\n if char == '=':\\n # 提取 '=' 左边和右边的两个字符\\n left_spaces = content[max(0, i-2):i]\\n right_spaces = content[i+1:i+3]\\n \\n # 判断两边是否都各有两个空格\\n if left_spaces != ' ' or right_spaces != ' ':\\n return False\\n\\n # 若未找到任何相关代码文件，则也视为不符合要求\\n return file_found\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在输出的任何代码文件中，每行的长度必须限制在50个字符以内\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀类型\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.json', '.py', '.java', '.c', '.cpp', '.h'}\\n \\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的换行符后计算长度\\n clean_line = line.rstrip('\\\\n\\\\r')\\n if len(clean_line) > 50:\\n return False\\n except Exception:\\n # 如果由于读取权限等原因无法处理，则跳过该文件继续检查\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"必须保证你生成的代码文件中完全避免使用递归\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在的连通是瞬间完成的，不够生动。下一步，请将‘瞬间变蓝’升级为带有传播逻辑的‘逐级流动’动画。水流需要从起点开始，根据连通路径一格接一格地向后延伸，每格之间有例如 0.3 秒的传播延迟。同时加入动态的‘连击(Combo)’机制：如果水流正在按延迟一格格蔓延时，玩家快速点击旋转了前方原本未接通的死胡同管道，使得水流没有中断并继续向前流，就在那个刚接通的管道上方触发一个向上飘动的‘Combo +1!’文字特效，并在计分板上增加相应分数。另外后续所有轮次请你始终遵循两项新要求：一是你的代码文件生成时必须用到递归技术，替换此前完全避免使用递归的要求；二是在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(后续优化)\\\"标记，不可多也不可少。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef verify_utf8_encoding(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证由人工智能助手生成的所有代码文件是否统一为 UTF-8 编码。\\n \\\"\\\"\\\"\\n # 根据用户指令推断，代码文件主要为 HTML、CSS 和 JS\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦抛出解码错误，说明存在非 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 对于其他不可预见的读取异常，保守判断为不符合\\n return False\\n \\n # 所有代码文件均成功通过 UTF-8 编解码测试\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果给定的工作空间路径不存在或不是一个目录，则没有违反嵌套深度的规定\\n if not os.path.isdir(workspace_path):\\n return True\\n \\n base_path = os.path.abspath(workspace_path)\\n \\n for root, dirs, files in os.walk(base_path):\\n # 就地修改 dirs 列表，过滤掉所有以 '.' 开头的隐藏文件夹\\n # 这将使 os.walk 自动跳过这些目录及其内部子目录\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 计算当前所处目录相对于根目录的相对路径\\n rel_path = os.path.relpath(root, base_path)\\n \\n if rel_path == '.':\\n depth = 0\\n else:\\n # 将不同系统的路径分隔符统一替换为 '/'，过滤空字符串后计算路径包含的层级数\\n parts = [p for p in rel_path.replace('\\\\\\\\', '/').split('/') if p]\\n depth = len(parts)\\n \\n # 判断文件夹嵌套深度是否超过 2 层\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用2个空格来控制缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 确保当前项目地址存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 常见可能生成的代码文件扩展名\\n code_exts = {'.js', '.css', '.html', '.htm', '.py', '.c', '.cpp', '.java', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n found_code_files = False\\n has_valid_indent = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_files = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n leading_spaces = []\\n for line in lines:\\n # 提取该行的前导空白字符\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 任何一行如果使用了制表符缩进，即违背了“空格缩进”的规定\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n stripped = line.lstrip()\\n # 忽略空行和可能打乱整体统计结构的多行注释标识行\\n if stripped and not stripped.startswith('*') and not stripped.startswith('<!--') and not stripped.startswith('/*') and not stripped.startswith('*/'):\\n leading_spaces.append(len(leading_whitespace))\\n \\n if not leading_spaces:\\n continue\\n \\n diffs = []\\n for i in range(1, len(leading_spaces)):\\n diff = leading_spaces[i] - leading_spaces[i-1]\\n if diff > 0:\\n diffs.append(diff)\\n \\n # 检查是否存在2个空格缩进专属的特征（即除以4余2，如2, 6, 10）\\n has_2_space = any(s % 4 == 2 for s in leading_spaces)\\n # 检查是否存在4个空格倍数的缩进特征\\n has_4_space = any(s % 4 == 0 and s > 0 for s in leading_spaces)\\n \\n count_2 = diffs.count(2)\\n count_4 = diffs.count(4)\\n \\n # 若文件包含任何2个空格的缩进特征，记录当前项目有效满足缩进要求的证据\\n if has_2_space or count_2 > 0:\\n has_valid_indent = True\\n \\n # 如果某文件内有明显的4空格规律，却毫无2空格规则的痕迹，说明其并未换用2个空格\\n if (has_4_space or count_4 > 0) and not has_2_space and count_2 == 0:\\n return False\\n \\n # 如果文件由于不规范的折行或其他原因，导致出现了太多奇数个空格缩进，说明没有正确按2个空格规律\\n odd_spaces = sum(1 for s in leading_spaces if s % 2 != 0)\\n if odd_spaces > len(leading_spaces) * 0.3:\\n return False\\n \\n # 最终条件：发现存在代码文件，并且项目中至少有一处确凿的2空格缩进特征\\n return found_code_files and has_valid_indent\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"def check_equals_spacing(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_equals = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # 尝试按 UTF-8 读取文本文件，跳过可能的二进制或不可读文件\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n count_all = content.count('=')\\n \\n if count_all > 0:\\n has_equals = True\\n \\n # 正则模式：匹配前后恰好各有两个空格的等号\\n # (?<! ) : 前面紧挨着的字符不能是空格（确保左边不多于两个空格）\\n # = : 等号及前后各两个空格\\n # (?! ) : 后面紧挨着的字符不能是空格（确保右边不多于两个空格）\\n count_valid = len(re.findall(r'(?<! ) = (?! )', content))\\n \\n # 如果符合规范的等号数量不等于总等号数量，则代表有等号违反了规则\\n if count_all != count_valid:\\n return False\\n \\n # 如果项目中至少存在一个等号且所有等号都符合规则，则为True；若无任何等号出现，视为未有效生成项目\\n return has_equals\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在输出的任何代码文件中，每行的长度必须限制在50个字符以内\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对用户的游戏网页项目，定义可能涉及的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.sh', '.bash'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件后缀名并进行小写匹配\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除由于换行产生的回车符或换行符\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 要求中明确指出了“字符”，直接采用 len 函数统计字符数量\\n if len(clean_line) > 50:\\n return False\\n except Exception:\\n # 如果某些文件并非合法的纯文本文件或遇到读取异常，则忽略跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"你的代码文件生成时必须用到递归技术\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 3 个 \\\"TODO(后续优化)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对前端项目推断出的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 遇到文件无法读取的情况，保守起见视作内容不合格或直接跳过\\n continue\\n \\n # “各个代码文件中”意味着每一个代码文件都必须刚好包含 3 个标记\\n if content.count(\\\"TODO(后续优化)\\\") != 3:\\n return False\\n\\n # 必须至少生成了 1 个代码文件且全都符合要求，才算总体满足条件\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后一步，我们来完善视觉模式切换和游戏循环机制。首先，在控制面板加一个‘霓虹模式’的切换开关。开启后，整个页面背景变暗，所有被水流激活的蓝色管道会散发出明亮的赛博朋克风格发光特效（利用 CSS drop-shadow 或 box-shadow 实现呼吸灯效果）。其次，增加胜利判定：当水流最终成功流入并点亮‘终点’管道时，游戏暂停操作，并在页面中央弹出一个带有炫酷光环效果的‘Level Clear!’过关弹窗。弹窗内提供一个‘下一关’按钮，点击后清空连通状态，并重新随机生成一盘全新的管道网格布局。另外后续所有轮次请你始终遵循两项更新要求：一是在你生成的各个代码文件中，必须包含正好 2 个 \\\"FIXME(待调试)\\\"标记，不可多也不可少，替换此前要求包含3个\\\"TODO(后续优化)\\\"标记的要求；二是你的每一条回复内容，都必须以\\\"呀～\\\"作为结尾语气词。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 如果工作区路径不存在，说明没有创建文件，暂不计为违规\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 根据用户指令推测可能的代码文件后缀名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 遍历项目目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名并转换为小写\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如无权限等其他非编码相关的系统级读取错误\\n pass\\n \\n # 如果所有的代码文件都能成功以 UTF-8 解码，说明满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def verify_folder_depth(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 检查工作空间路径是否存在，如果不存在则自然没有超出嵌套深度\\n if not os.path.exists(workspace_path):\\n return True\\n \\n workspace = Path(workspace_path).resolve()\\n \\n # 遍历代码项目下的所有文件及目录\\n for root, dirs, files in os.walk(workspace_path):\\n root_path = Path(root).resolve()\\n \\n try:\\n # 计算当前遍历到的目录相对于工作区根目录的相对路径\\n rel_path = root_path.relative_to(workspace)\\n except ValueError:\\n # 如果路径解析出现异常，跳过当前项\\n continue\\n \\n # 计算相对路径的层级数\\n # '.' 的 parts 长度为 0 （代表 workspace_path 本身）\\n # 'dir1' 的 parts 长度为 1 （代表第 1 层目录）\\n # 'dir1/dir2' 的 parts 长度为 2 （代表第 2 层目录）\\n # 'dir1/dir2/dir3' 的 parts 长度为 3 （代表超过了 2 层的嵌套）\\n if len(rel_path.parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 60s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用2个空格来控制缩进\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n \\n # 检查项目目录是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义前端代码相关的文件扩展名\\n valid_exts = ('.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.json', '.vue', '.htm')\\n valid_files_checked = 0\\n has_indentation = False\\n \\n # 遍历当前生成项目中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith(valid_exts):\\n filepath = os.path.join(root, file)\\n valid_files_checked += 1\\n \\n # 统一以 utf-8 读取代码文件并忽略解码错误以防崩溃\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n \\n prev_indent = 0\\n diffs = []\\n tab_lines = 0\\n \\n for line in lines:\\n stripped = line.strip()\\n # 过滤掉空行\\n if not stripped:\\n continue\\n \\n # 若行首为制表符，视为使用了 Tab 缩进\\n if line.startswith('\\\\t'):\\n tab_lines += 1\\n prev_indent = 0\\n continue\\n \\n # 计算该行的前导空格数量\\n indent = len(line) - len(line.lstrip(' '))\\n \\n # 只有在发生正向缩进（向深层缩进）时，我们才收集其增量差值\\n if indent > prev_indent:\\n diffs.append(indent - prev_indent)\\n \\n prev_indent = indent\\n \\n # 若文件包含使用制表符缩进的行，违反“全部使用2个空格”的规则\\n if tab_lines > 0:\\n return False\\n \\n if diffs:\\n has_indentation = True\\n # 使用 Counter 寻找最常见的缩进增量（众数）\\n c = Counter(diffs)\\n max_count = max(c.values())\\n modes = [k for k, v in c.items() if v == max_count]\\n \\n # 如果最常用的缩进增加幅度中不包含 2（例如只包含 4）\\n # 说明该文件主要不依靠 2个空格 来控制层级缩进\\n if 2 not in modes:\\n return False\\n\\n # 若未找到任何可分析的代码文件，默认验证不通过\\n if valid_files_checked == 0:\\n return False\\n \\n # 若所有的文件代码连一级缩进都没有，也无法佐证是否遵循要求\\n if not has_indentation:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目性质推断常见的代码文件扩展名\\n code_extensions = ['.html', '.css', '.js']\\n code_files = []\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 查找所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有生成任何代码文件，则视为没有完成代码任务，返回 False\\n if not code_files:\\n return False\\n \\n # 逐文件检查 '=' 两侧的空格情况\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n for i, char in enumerate(content):\\n if char == '=':\\n # 检查左边是否恰好有两个空格\\n # 1. 至少需要前两个字符是空格\\n if i < 2 or content[i-1] != ' ' or content[i-2] != ' ':\\n return False\\n # 2. 前面第三个字符不能也是空格\\n if i >= 3 and content[i-3] == ' ':\\n return False\\n \\n # 检查右边是否恰好有两个空格\\n # 1. 至少需要后两个字符是空格\\n if i > len(content) - 3 or content[i+1] != ' ' or content[i+2] != ' ':\\n return False\\n # 2. 后面第三个字符不能也是空格\\n if i <= len(content) - 4 and content[i+3] == ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在输出的任何代码文件中，每行的长度必须限制在50个字符以内\",\n \"validation_code\": \"def check_line_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区目录不存在，视作未满足要求\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义可能的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.py', '.sh', '.bash', '.java', '.c', '.cpp', '.h', '.json'\\n }\\n\\n # 要忽略的常见依赖及版本控制目录，避免深入遍历并防止干扰\\n ignore_dirs = {'node_modules', '.git', '.svn', 'dist', 'build'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以剪去不需要遍历的子目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理属于代码文件范畴的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 兼容带有 BOM 头的 UTF-8 文件\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n for line in f:\\n # 剔除行尾的回车与换行符，其余空格等字符予以保留\\n line_content = line.rstrip('\\\\r\\\\n')\\n # 检查字符数是否超过 50 个字符的上限\\n if len(line_content) > 50:\\n return False\\n except Exception:\\n # 读取出错时跳过（例如碰到二进制文件等异常情况）\\n continue\\n\\n # 遍历结束未发现超出 50 个字符的代码行，返回 True \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"你的代码文件生成时必须用到递归技术\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的各个代码文件中，必须包含正好 2 个 \\\"FIXME(待调试)\\\"标记，不可多也不可少\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n target_mark = \\\"FIXME(待调试)\\\"\\n # 根据用户指令中提到的 HTML/CSS/JS 推断，代码文件大概率是前端项目相关后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.vue', '.jsx', '.tsx'}\\n \\n file_checked = False\\n \\n # 1. 优先检查工作区是否存在生成的代码文件\\n if workspace_path and os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查每个代码文件是否恰好包含 2 个目标标记，不可多也不可少\\n if content.count(target_mark) != 2:\\n return False\\n \\n file_checked = True\\n except Exception:\\n pass\\n \\n # 若在工作区找到了对应代码文件，并均满足条件，则验证成功\\n if file_checked:\\n return True\\n \\n # 2. 如果工作区为空或未检测到实体代码文件，从 response 中提取 Markdown 代码块作为代码文件进行检测\\n # 匹配带有 HTML/CSS/JS 等前端常用语言标签的代码块\\n pattern = r'```(?:html|css|javascript|js|vue|ts|jsx|tsx)?\\\\s*\\\\n(.*?)\\\\n```'\\n code_blocks = re.findall(pattern, response, re.IGNORECASE | re.DOTALL)\\n \\n if not code_blocks:\\n # 如果未匹配到，尝试匹配无语言标签的代码块\\n code_blocks = re.findall(r'```\\\\s*\\\\n(.*?)\\\\n```', response, re.DOTALL)\\n \\n # 如果既没有实体代码文件，也没有文本代码块，说明未生成代码文件，自然不满足要求\\n if not code_blocks:\\n return False\\n \\n # 检查每一个代码块中是否也恰好包含 2 个目标标记\\n for block in code_blocks:\\n if block.count(target_mark) != 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复内容，都必须以\\\"呀～\\\"作为结尾语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除回复末尾可能存在的空白字符以及常见中英文句末标点符号\\n cleaned_response = re.sub(r'[\\\\s。！？!?,，]+$', '', response)\\n \\n # 检查清理末尾标点后的文本是否以指定的语气词“呀～”结尾\\n return cleaned_response.endswith('呀～')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查页面整体布局：页面中央显示一个 6x6 的网格游戏区域，左侧或顶部存在标题栏和包含“开始”按钮的控制面板，右侧或底部存在计分面板。","检查管道形态与初始化：确认 6x6 网格内随机分布了多种不同的管道形态，仔细对比截图确认其中包含：起点（单方向开口）、终点（单方向开口）、直管（双向直线开口）、弯管（L型双向开口）和十字交叉管（四向开口）。","检查鼠标悬浮效果：模拟鼠标悬停在网格的任意方块上，截取悬浮状态的页面截图，确认该方块的背景色相较于未悬浮状态发生了变化。","检查点击反馈特效：模拟鼠标点击网格中的普通管道方块，观察并截取动画过程，确认方块在被点击时出现了轻微的缩放回弹视觉效果。","检查管道旋转交互：模拟鼠标点击一个普通的管道（如直管或弯管），观察并截图确认该管道顺时针旋转了90度，且旋转过程具有平滑的过渡动画。","检查起点/终点不可旋转限制：模拟鼠标分别点击“起点”管道和“终点”管道，确认这两个管道状态保持固定，不会发生旋转动画。","检查连通匹配逻辑：通过旋转管道拼接出一条从起点出发但中途断开的路径。点击“开始”按钮，观察水流蔓延结果，确认水流仅在管道端口严格对接（如右开口对左开口）时才通过并改变管道颜色（变为激活的蓝色），遇到未匹配端口或边界时水流正确停止。","检查逐级流动动画：拼接一段较长的有效通路并点击“开始”按钮。观察通水过程并截图验证：激活状态（变蓝）是从起点开始，一格接着一格带有延迟时间顺序蔓延的，而不是整条通路瞬间同时变色。","检查连击(Combo)机制：点击“开始”让水流开始逐级蔓延，在水流到达某死胡同管道前，快速模拟点击旋转前方未接通的管道使其变成接通状态。观察并截图确认：接通的管道上方飘出了“Combo +1!”的文字特效，且计分板分数随之增加。","检查霓虹模式切换：点击控制面板中的“霓虹模式”开关。观察并截图确认：页面整体背景变为暗色系，且已经处于激活状态的蓝色管道出现了明显的发光特效（如 drop-shadow 阴影效果）并伴有动态的呼吸灯效果。","检查胜利判定与弹窗：将管道成功拼接成一条从“起点”到“终点”的连通路径并点击“开始”。等待水流到达并点亮终点后，确认游戏不再允许旋转管道操作，且页面中央弹出了带有发光光环效果的“Level Clear!”过关弹窗。","检查游戏循环/下一关机制：在“Level Clear!”过关弹窗中模拟点击“下一关”按钮。观察并截图确认：弹窗消失，所有网格中管道的蓝色通水状态被清空，且 6x6 网格重新随机生成了全新的管道布局。"],"string":"[\n \"检查页面整体布局：页面中央显示一个 6x6 的网格游戏区域，左侧或顶部存在标题栏和包含“开始”按钮的控制面板，右侧或底部存在计分面板。\",\n \"检查管道形态与初始化：确认 6x6 网格内随机分布了多种不同的管道形态，仔细对比截图确认其中包含：起点（单方向开口）、终点（单方向开口）、直管（双向直线开口）、弯管（L型双向开口）和十字交叉管（四向开口）。\",\n \"检查鼠标悬浮效果：模拟鼠标悬停在网格的任意方块上，截取悬浮状态的页面截图，确认该方块的背景色相较于未悬浮状态发生了变化。\",\n \"检查点击反馈特效：模拟鼠标点击网格中的普通管道方块，观察并截取动画过程，确认方块在被点击时出现了轻微的缩放回弹视觉效果。\",\n \"检查管道旋转交互：模拟鼠标点击一个普通的管道（如直管或弯管），观察并截图确认该管道顺时针旋转了90度，且旋转过程具有平滑的过渡动画。\",\n \"检查起点/终点不可旋转限制：模拟鼠标分别点击“起点”管道和“终点”管道，确认这两个管道状态保持固定，不会发生旋转动画。\",\n \"检查连通匹配逻辑：通过旋转管道拼接出一条从起点出发但中途断开的路径。点击“开始”按钮，观察水流蔓延结果，确认水流仅在管道端口严格对接（如右开口对左开口）时才通过并改变管道颜色（变为激活的蓝色），遇到未匹配端口或边界时水流正确停止。\",\n \"检查逐级流动动画：拼接一段较长的有效通路并点击“开始”按钮。观察通水过程并截图验证：激活状态（变蓝）是从起点开始，一格接着一格带有延迟时间顺序蔓延的，而不是整条通路瞬间同时变色。\",\n \"检查连击(Combo)机制：点击“开始”让水流开始逐级蔓延，在水流到达某死胡同管道前，快速模拟点击旋转前方未接通的管道使其变成接通状态。观察并截图确认：接通的管道上方飘出了“Combo +1!”的文字特效，且计分板分数随之增加。\",\n \"检查霓虹模式切换：点击控制面板中的“霓虹模式”开关。观察并截图确认：页面整体背景变为暗色系，且已经处于激活状态的蓝色管道出现了明显的发光特效（如 drop-shadow 阴影效果）并伴有动态的呼吸灯效果。\",\n \"检查胜利判定与弹窗：将管道成功拼接成一条从“起点”到“终点”的连通路径并点击“开始”。等待水流到达并点亮终点后，确认游戏不再允许旋转管道操作，且页面中央弹出了带有发光光环效果的“Level Clear!”过关弹窗。\",\n \"检查游戏循环/下一关机制：在“Level Clear!”过关弹窗中模拟点击“下一关”按钮。观察并截图确认：弹窗消失，所有网格中管道的蓝色通水状态被清空，且 6x6 网格重新随机生成了全新的管道布局。\"\n]","isConversational":false}}},{"rowIdx":64,"cells":{"task_id":{"kind":"number","value":65,"string":"65"},"repository_policy":{"kind":"string","value":"你作为具备工具调用和文件操作权限的代码智能体，在后续所有轮次的服务中都必须始终遵循以下约束：\n1. 不论何时，你生成用户回复时都必须使用中文。\n2. 每当准备更改已有文件时，都必须先用相关工具阅读该文件。\n3. 你创建的代码文件无一例外，全都要用UTF-8进行编码。\n4. 确保在你生成的所有代码文件中，全部的注释内容皆为英语。\n5. 每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换。\n6. 为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数。\n7. 生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码及相关文本文件的后缀名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', '.go', '.rs',\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.scss', '.less',\n '.php', '.rb', '.sh', '.bat', '.ps1', '.swift', '.kt', '.m', '.mm', \n '.scala', '.sql', '.json', '.xml', '.yaml', '.yml', '.md', '.txt'\n }\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n # 如果目录不存在，说明没有生成文件，也就没有违反要求\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_content = f.read()\n \n # 尝试使用utf-8进行解码\n # 如果是合法的utf-8或ascii编码（ascii为utf-8子集），则不会报错\n raw_content.decode('utf-8')\n \n except UnicodeDecodeError:\n # 如果发生解码错误，说明存在非UTF-8编码的文件（例如带有中文字符的GBK编码文件）\n return False\n except Exception:\n # 忽略其他可能的文件读取异常（如权限问题等）\n pass\n \n # 如果全部代码文件都能成功解码或者没有代码文件，则符合要求\n return True","tags":["Language","File Encoding"]},{"constraint":"确保在你生成的所有代码文件中，全部的注释内容皆为英语","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写","validation_code":"","tags":["Environment","File Path"]}],"string":"[\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码及相关文本文件的后缀名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', '.go', '.rs',\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.scss', '.less',\\n '.php', '.rb', '.sh', '.bat', '.ps1', '.swift', '.kt', '.m', '.mm', \\n '.scala', '.sql', '.json', '.xml', '.yaml', '.yml', '.md', '.txt'\\n }\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n # 如果目录不存在，说明没有生成文件，也就没有违反要求\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_content = f.read()\\n \\n # 尝试使用utf-8进行解码\\n # 如果是合法的utf-8或ascii编码（ascii为utf-8子集），则不会报错\\n raw_content.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明存在非UTF-8编码的文件（例如带有中文字符的GBK编码文件）\\n return False\\n except Exception:\\n # 忽略其他可能的文件读取异常（如权限问题等）\\n pass\\n \\n # 如果全部代码文件都能成功解码或者没有代码文件，则符合要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，全部的注释内容皆为英语\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用 React 开发一个看板应用（Kanban Board）。这一轮只需要搭个基本骨架。请创建一个包含三个默认列（待办、进行中、已完成）的页面布局，页面撑满全屏，列之间横向排列。每列有一个顶部的标题，并在列表区域展示一些基础的占位卡片。在每个列的底部加上一个“添加任务”的按钮，点击时暂时只需向该列末尾追加一个写着“新测试任务”的占位卡片即可。目前绝对不需要实现任何拖拽功能，重点是把界面的 CSS 排版做好，做出类似 Trello 的卡片UI风格。请注意，后续所有轮次都必须始终遵循以下要求：第一，在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符；第二，必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义React项目和题目要求中可能涉及到的代码或日志文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', \n '.html', '.htm', \n '.json', \n '.md', \n '.log',\n '.txt'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生解码错误，说明文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能的异常（如权限问题等）\n continue\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保在你生成的所有代码文件中，全部的注释内容皆为英语","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":1,"instruction":"骨架的 UI 看起来很不错。现在我们需要让看板真正“动”起来。请引入 `react-beautiful-dnd` 库，实现卡片的拖拽功能。具体要求：1. 卡片能在同一个列内上下拖拽来自由调整排序；2. 卡片必须支持跨列拖拽（比如把卡片从“待办”拖入“进行中”）；3. 拖拽松手落位后，内部的状态树必须被正确更新，确保渲染出来的位置和拖拽结果完全一致。注意处理好同列排序和跨列移动两种不同情况的数据交换逻辑。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，不管是文件路径还是代码位置，都一律只能使用单引号引起来，不再使用此前要求的双引号；第二，确保生成的代码文件中，所有 import 和 require 语句涉及的包名均严格按照字母顺序升序的规则排序。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义 React 前端项目以及系统提示中可能涉及的代码/日志文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.css', '.scss', '.less', '.json', '.log', '.md'\n }\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无关的第三方依赖和版本控制目录，避免干扰检查\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明并非合法/兼容的 UTF-8 编码\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保在你生成的所有代码文件中，全部的注释内容皆为英语","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"不管是文件路径还是代码位置，都一律只能使用单引号引起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保生成的代码文件中，所有 import 和 require 语句涉及的包名均严格按照字母顺序升序的规则排序","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":2,"instruction":"拖拽功能很顺滑！接下来完善任务信息的管理，并加上本地持久化功能。请实现：1. 改造“添加任务”功能，点击按钮不再生成假数据，而是弹出一个表单模态框，让用户输入任务的“标题”和“详细描述”后再创建卡片；2. 卡片界面上需展示标题和部分描述文本，并且在卡片右上角增加一个“删除”按钮，点击可删除该任务；3. 接入浏览器的 LocalStorage，把整个看板的数据状态实时保存下来，确保无论怎么刷新页面，拖拽的顺序、列的内容都不会丢失。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序，不再使用此前要求的升序规则；第二，在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\n # 针对React Kanban项目的常见代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', \n '.html', '.htm', \n '.json', '.log', '.md'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的第三方依赖和版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦发现无法用 UTF-8 解码的文件，说明不符合要求\n return False\n except Exception:\n # 忽略其他非编码相关的系统级读取错误\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"确保在你生成的所有代码文件中，全部的注释内容皆为英语","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"不管是文件路径还是代码位置，都一律只能使用单引号引起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"很好，现在的看板已经很实用了。让我们增加一点挑战，对“列（Column）”本身进行增强。请实现：1. 在看板所有列的最右侧增加一个“新增列表”的区块，允许用户创建具有自定义名字的新列；2. 点击现有列的标题，可以变成输入框进行重命名；3. 实现列级别的拖拽！也就是说，除了卡片可以上下拖，现在整个“列”也需要支持水平方向的拖拽排序，以便用户自由调整列与列之间的左右顺序。列的变更也必须同步保存到 LocalStorage 中。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，当生成代码文件时，请倾向于使用反斜杠\\作为路径分隔符号，不再使用此前要求的正斜杠；第二，请在所有生成的代码文件的注释中添加 emoji表情符号。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目和用户指令中涉及的常见代码/文本文件后缀名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.json', '.log', '.md'}\n \n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅对代码类型的文件进行编码验证，跳过图片或字体等二进制文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试使用 utf-8 编码格式读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，则说明文件未严格使用 UTF-8 编码\n return False\n except Exception:\n # 忽略其他的异常，如无读取权限或文件被占用等\n pass\n \n # 若所有匹配的代码文件都能被 UTF-8 正确读取，则代表满足了给定要求\n return True","tags":["Language","File Encoding"]},{"constraint":"确保在你生成的所有代码文件中，全部的注释内容皆为英语","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写","validation_code":"","tags":["Environment","File Path"]},{"constraint":"当生成代码文件时，请倾向于使用反斜杠\\作为路径分隔符号，不再使用此前要求的正斜杠","validation_code":"","tags":["Environment","File Path"]},{"constraint":"不管是文件路径还是代码位置，都一律只能使用单引号引起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在所有生成的代码文件的注释中添加 emoji表情符号","validation_code":"def check_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断出 React 项目可能使用的常见代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n # 设置需要排除的非源代码或依赖目录\n ignore_dirs = {'node_modules', '.git', 'build', 'dist', '.idea', '.vscode'}\n\n has_checked_any_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉忽略的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n has_checked_any_file = True\n \n # 提取不同类型文件的注释内容\n comments = []\n if ext in {'.js', '.jsx', '.ts', '.tsx'}:\n comments.extend(re.findall(r'//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n elif ext == '.css':\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n elif ext == '.html':\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n # 检查提取的注释中是否包含 emoji\n file_has_emoji_in_comment = False\n for comment in comments:\n for char in comment:\n if is_emoji(char): # 调用人类专家提供的现成函数\n file_has_emoji_in_comment = True\n break\n if file_has_emoji_in_comment:\n break\n \n # 若某个代码文件的所有注释中均未发现 emoji（或毫无注释），则判定为不满足要求\n if not file_has_emoji_in_comment:\n return False\n\n # 如果有检查到文件且都满足要求，返回True；如果没有遇到任何代码文件，则不通过\n return has_checked_any_file","tags":["Content","Keyword"]}]},{"round_id":4,"instruction":"太棒了，双层拖拽都实现了。最后我们来打磨一下高级的交互体验和内容检索功能。请完成最后两个需求：1. 在页面顶部增加一个任务搜索框，输入关键字时，能够实时过滤看板上的卡片，隐藏掉不匹配的任务，只展示标题或描述中包含关键字的卡片；2. 增加极致的拖拽视觉反馈：当卡片被鼠标抓起时，让卡片增加一个微微倾斜的角度和加深的阴影效果；当被抓起的卡片悬停在某一个目标列上方时，该目标列的整体背景色要有明显的变暗或高亮过渡，明确提示用户当前的落点。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，所有代码文件中的注释内容都必须严格排除 emoji表情符号，不再使用此前要求的添加emoji表情符号的规则；第二，生成代码文件中的各个函数，每一个都不能超出 50 行的限制。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"每当准备更改已有文件时，都必须先用相关工具阅读该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对 React 项目的常见代码文件扩展名及用户指令中要求生成的 log 文件\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json', '.log'}\n \n # 如果项目路径不存在或不是一个目录，直接返回 False\n if not os.path.isdir(workspace_path):\n return False\n\n # 遍历项目下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码及指定的文本型文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 utf-8 编码对二进制内容进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件未严格使用 UTF-8 编码\n return False\n \n # 如果所有提取出的代码文件都能成功被 utf-8 解码，说明满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"确保在你生成的所有代码文件中，全部的注释内容皆为英语","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写","validation_code":"","tags":["Environment","File Path"]},{"constraint":"当生成代码文件时，请倾向于使用反斜杠\\作为路径分隔符号，不再使用此前要求的正斜杠","validation_code":"","tags":["Environment","File Path"]},{"constraint":"不管是文件路径还是代码位置，都一律只能使用单引号引起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"所有代码文件中的注释内容都必须严格排除 emoji表情符号","validation_code":"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对 React Kanban 应用，推测并枚举出常见的前端相关代码文件后缀\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.htm'}\n \n # 正则表达式模块说明：\n # 1. str_pat: 匹配双引号、单引号、反引号包裹的字符串（涵盖转义字符的过滤），用于捕获并归入 group 1\n str_pat = r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n # 2. com_pat: 匹配多行注释、单行注释、HTML注释，用于捕获并归入 group 2\n com_pat = r'/\\*.*?\\*/|//[^\\n]*|<!--.*?-->'\n \n # 合并后的正则。由于使用了 `|`，正则引擎会从左到右尝试匹配分支。\n # 这样可以确保匹配到包含类似 `//` 的字符串字面量时，不会被错误地当作注释提取。\n combined_pat = f'({str_pat})|({com_pat})'\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略由于编码或其他问题无法被正常读取的文件\n continue\n \n # 使用 re.DOTALL 标志，使得多行注释和HTML注释中的 `.` 能够匹配换行符\n for match in re.finditer(combined_pat, content, flags=re.DOTALL):\n # 如果 group 2 匹配成功，说明当前捕获到的是代码注释内容\n if match.group(2):\n comment = match.group(2)\n # 遍历该段注释中的所有字符，逐一鉴别\n for char in comment:\n if is_emoji(char):\n # 注释内容中一旦发现包含 emoji，即违反严格排除的要求\n return False \n \n # 遍历结束，所有代码文件中的所有注释均未发现 emoji，要求被顺利满足\n return True","tags":["Content","Keyword"]},{"constraint":"生成代码文件中的各个函数，每一个都不能超出 50 行的限制","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n def clean_code(code_str: str) -> str:\n \"\"\"清洗代码，将字符串内内容与注释替换为空格，以防干扰花括号解析\"\"\"\n res = []\n i = 0\n n = len(code_str)\n state = 'NORMAL'\n \n while i < n:\n c = code_str[i]\n \n if state == 'NORMAL':\n if c == '/' and i + 1 < n and code_str[i+1] == '/':\n state = 'LINE_COMMENT'\n res.append(' ')\n res.append(' ')\n i += 2\n continue\n elif c == '/' and i + 1 < n and code_str[i+1] == '*':\n state = 'BLOCK_COMMENT'\n res.append(' ')\n res.append(' ')\n i += 2\n continue\n elif c == \"'\":\n state = 'SINGLE_QUOTE'\n res.append(' ')\n i += 1\n continue\n elif c == '\"':\n state = 'DOUBLE_QUOTE'\n res.append(' ')\n i += 1\n continue\n elif c == '`':\n state = 'BACKTICK'\n res.append(' ')\n i += 1\n continue\n else:\n res.append(c)\n i += 1\n \n elif state == 'LINE_COMMENT':\n if c == '\\n':\n state = 'NORMAL'\n res.append('\\n')\n else:\n res.append(' ')\n i += 1\n \n elif state == 'BLOCK_COMMENT':\n if c == '*' and i + 1 < n and code_str[i+1] == '/':\n state = 'NORMAL'\n res.append(' ')\n res.append(' ')\n i += 2\n continue\n else:\n if c == '\\n':\n res.append('\\n')\n else:\n res.append(' ')\n i += 1\n \n elif state in ('SINGLE_QUOTE', 'DOUBLE_QUOTE', 'BACKTICK'):\n quote_char = \"'\" if state == 'SINGLE_QUOTE' else ('\"' if state == 'DOUBLE_QUOTE' else '`')\n if c == '\\\\':\n res.append(' ')\n if i + 1 < n:\n if code_str[i+1] == '\\n':\n res.append('\\n')\n else:\n res.append(' ')\n i += 2\n continue\n elif c == quote_char:\n state = 'NORMAL'\n res.append(' ')\n i += 1\n else:\n if c == '\\n':\n res.append('\\n')\n else:\n res.append(' ')\n i += 1\n \n return \"\".join(res)\n\n def is_function_brace(cleaned_code: str, brace_idx: int) -> bool:\n \"\"\"向回看判断遇到的 '{' 是否属于函数、箭头函数或方法的开始\"\"\"\n i = brace_idx - 1\n while i >= 0 and cleaned_code[i].isspace():\n i -= 1\n if i < 0:\n return False\n \n if cleaned_code[i] == '>':\n if i - 1 >= 0 and cleaned_code[i-1] == '=':\n return True\n \n if cleaned_code[i] == ')':\n paren_count = 1\n i -= 1\n while i >= 0:\n if cleaned_code[i] == ')':\n paren_count += 1\n elif cleaned_code[i] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n i -= 1\n \n if i < 0:\n return False\n \n i -= 1\n while i >= 0 and cleaned_code[i].isspace():\n i -= 1\n if i < 0:\n return False\n \n word = \"\"\n while i >= 0 and (cleaned_code[i].isalnum() or cleaned_code[i] in ['_', '$']):\n word = cleaned_code[i] + word\n i -= 1\n \n if word == \"function\":\n return True\n elif word in [\"if\", \"for\", \"while\", \"switch\", \"catch\", \"with\"]:\n return False\n else:\n j = i\n while j >= 0 and cleaned_code[j].isspace():\n j -= 1\n if j >= 0:\n prev_word = \"\"\n while j >= 0 and (cleaned_code[j].isalnum() or cleaned_code[j] in ['_', '$']):\n prev_word = cleaned_code[j] + prev_word\n j -= 1\n if prev_word == \"function\":\n return True\n \n if word != \"\":\n return True\n \n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_str = f.read()\n except Exception:\n continue\n \n cleaned = clean_code(code_str)\n \n # 预处理行号映射表\n line_number_map = []\n current_line = 1\n for char in cleaned:\n line_number_map.append(current_line)\n if char == '\\n':\n current_line += 1\n \n n = len(cleaned)\n for idx in range(n):\n if cleaned[idx] == '{':\n if is_function_brace(cleaned, idx):\n brace_count = 1\n start_line = line_number_map[idx]\n end_idx = idx + 1\n while end_idx < n:\n if cleaned[end_idx] == '{':\n brace_count += 1\n elif cleaned[end_idx] == '}':\n brace_count -= 1\n if brace_count == 0:\n break\n end_idx += 1\n \n # 当函数块结束，校验函数的跨越行数\n if end_idx < n:\n end_line = line_number_map[end_idx]\n lines_count = end_line - start_line + 1\n if lines_count > 50:\n return False\n \n return True","tags":["Quantity","Range"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用 React 开发一个看板应用（Kanban Board）。这一轮只需要搭个基本骨架。请创建一个包含三个默认列（待办、进行中、已完成）的页面布局，页面撑满全屏，列之间横向排列。每列有一个顶部的标题，并在列表区域展示一些基础的占位卡片。在每个列的底部加上一个“添加任务”的按钮，点击时暂时只需向该列末尾追加一个写着“新测试任务”的占位卡片即可。目前绝对不需要实现任何拖拽功能，重点是把界面的 CSS 排版做好，做出类似 Trello 的卡片UI风格。请注意，后续所有轮次都必须始终遵循以下要求：第一，在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符；第二，必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义React项目和题目要求中可能涉及到的代码或日志文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', \\n '.html', '.htm', \\n '.json', \\n '.md', \\n '.log',\\n '.txt'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能的异常（如权限问题等）\\n continue\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，全部的注释内容皆为英语\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"骨架的 UI 看起来很不错。现在我们需要让看板真正“动”起来。请引入 `react-beautiful-dnd` 库，实现卡片的拖拽功能。具体要求：1. 卡片能在同一个列内上下拖拽来自由调整排序；2. 卡片必须支持跨列拖拽（比如把卡片从“待办”拖入“进行中”）；3. 拖拽松手落位后，内部的状态树必须被正确更新，确保渲染出来的位置和拖拽结果完全一致。注意处理好同列排序和跨列移动两种不同情况的数据交换逻辑。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，不管是文件路径还是代码位置，都一律只能使用单引号引起来，不再使用此前要求的双引号；第二，确保生成的代码文件中，所有 import 和 require 语句涉及的包名均严格按照字母顺序升序的规则排序。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义 React 前端项目以及系统提示中可能涉及的代码/日志文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.css', '.scss', '.less', '.json', '.log', '.md'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关的第三方依赖和版本控制目录，避免干扰检查\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明并非合法/兼容的 UTF-8 编码\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，全部的注释内容皆为英语\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"不管是文件路径还是代码位置，都一律只能使用单引号引起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保生成的代码文件中，所有 import 和 require 语句涉及的包名均严格按照字母顺序升序的规则排序\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"拖拽功能很顺滑！接下来完善任务信息的管理，并加上本地持久化功能。请实现：1. 改造“添加任务”功能，点击按钮不再生成假数据，而是弹出一个表单模态框，让用户输入任务的“标题”和“详细描述”后再创建卡片；2. 卡片界面上需展示标题和部分描述文本，并且在卡片右上角增加一个“删除”按钮，点击可删除该任务；3. 接入浏览器的 LocalStorage，把整个看板的数据状态实时保存下来，确保无论怎么刷新页面，拖拽的顺序、列的内容都不会丢失。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序，不再使用此前要求的升序规则；第二，在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n # 针对React Kanban项目的常见代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', \\n '.html', '.htm', \\n '.json', '.log', '.md'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的第三方依赖和版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦发现无法用 UTF-8 解码的文件，说明不符合要求\\n return False\\n except Exception:\\n # 忽略其他非编码相关的系统级读取错误\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，全部的注释内容皆为英语\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件内，优先使用正斜杠 / 来充当路径的分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"不管是文件路径还是代码位置，都一律只能使用单引号引起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"很好，现在的看板已经很实用了。让我们增加一点挑战，对“列（Column）”本身进行增强。请实现：1. 在看板所有列的最右侧增加一个“新增列表”的区块，允许用户创建具有自定义名字的新列；2. 点击现有列的标题，可以变成输入框进行重命名；3. 实现列级别的拖拽！也就是说，除了卡片可以上下拖，现在整个“列”也需要支持水平方向的拖拽排序，以便用户自由调整列与列之间的左右顺序。列的变更也必须同步保存到 LocalStorage 中。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，当生成代码文件时，请倾向于使用反斜杠\\\\作为路径分隔符号，不再使用此前要求的正斜杠；第二，请在所有生成的代码文件的注释中添加 emoji表情符号。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目和用户指令中涉及的常见代码/文本文件后缀名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.json', '.log', '.md'}\\n \\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅对代码类型的文件进行编码验证，跳过图片或字体等二进制文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试使用 utf-8 编码格式读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，则说明文件未严格使用 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他的异常，如无读取权限或文件被占用等\\n pass\\n \\n # 若所有匹配的代码文件都能被 UTF-8 正确读取，则代表满足了给定要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，全部的注释内容皆为英语\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"当生成代码文件时，请倾向于使用反斜杠\\\\作为路径分隔符号，不再使用此前要求的正斜杠\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"不管是文件路径还是代码位置，都一律只能使用单引号引起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在所有生成的代码文件的注释中添加 emoji表情符号\",\n \"validation_code\": \"def check_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断出 React 项目可能使用的常见代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n # 设置需要排除的非源代码或依赖目录\\n ignore_dirs = {'node_modules', '.git', 'build', 'dist', '.idea', '.vscode'}\\n\\n has_checked_any_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉忽略的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n has_checked_any_file = True\\n \\n # 提取不同类型文件的注释内容\\n comments = []\\n if ext in {'.js', '.jsx', '.ts', '.tsx'}:\\n comments.extend(re.findall(r'//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n elif ext == '.css':\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n elif ext == '.html':\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n # 检查提取的注释中是否包含 emoji\\n file_has_emoji_in_comment = False\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char): # 调用人类专家提供的现成函数\\n file_has_emoji_in_comment = True\\n break\\n if file_has_emoji_in_comment:\\n break\\n \\n # 若某个代码文件的所有注释中均未发现 emoji（或毫无注释），则判定为不满足要求\\n if not file_has_emoji_in_comment:\\n return False\\n\\n # 如果有检查到文件且都满足要求，返回True；如果没有遇到任何代码文件，则不通过\\n return has_checked_any_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"太棒了，双层拖拽都实现了。最后我们来打磨一下高级的交互体验和内容检索功能。请完成最后两个需求：1. 在页面顶部增加一个任务搜索框，输入关键字时，能够实时过滤看板上的卡片，隐藏掉不匹配的任务，只展示标题或描述中包含关键字的卡片；2. 增加极致的拖拽视觉反馈：当卡片被鼠标抓起时，让卡片增加一个微微倾斜的角度和加深的阴影效果；当被抓起的卡片悬停在某一个目标列上方时，该目标列的整体背景色要有明显的变暗或高亮过渡，明确提示用户当前的落点。另外请注意，后续所有轮次都必须始终遵循以下更新及新增要求：第一，所有代码文件中的注释内容都必须严格排除 emoji表情符号，不再使用此前要求的添加emoji表情符号的规则；第二，生成代码文件中的各个函数，每一个都不能超出 50 行的限制。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"每当准备更改已有文件时，都必须先用相关工具阅读该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对 React 项目的常见代码文件扩展名及用户指令中要求生成的 log 文件\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json', '.log'}\\n \\n # 如果项目路径不存在或不是一个目录，直接返回 False\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历项目下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码及指定的文本型文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 utf-8 编码对二进制内容进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件未严格使用 UTF-8 编码\\n return False\\n \\n # 如果所有提取出的代码文件都能成功被 utf-8 解码，说明满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的所有代码文件中，全部的注释内容皆为英语\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每次更动已有文件，请严格输出Unified Diff格式，坚决避免直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请勿在文件引用处使用绝对路径，必须以相对路径进行编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"当生成代码文件时，请倾向于使用反斜杠\\\\作为路径分隔符号，不再使用此前要求的正斜杠\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"不管是文件路径还是代码位置，都一律只能使用单引号引起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中所有 import 和 require 语句涉及的包名均严格按照字母降序的规则排序\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在修改代码之前，务必调用 TodoWrite 工具进行任务规划，一旦完成了与某任务相关的代码修改（Write/Edit）操作，紧接着的首个工具调用必须是使用 TodoComplete 标记该任务完成，期间严禁夹杂任何其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"所有代码文件中的注释内容都必须严格排除 emoji表情符号\",\n \"validation_code\": \"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对 React Kanban 应用，推测并枚举出常见的前端相关代码文件后缀\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.htm'}\\n \\n # 正则表达式模块说明：\\n # 1. str_pat: 匹配双引号、单引号、反引号包裹的字符串（涵盖转义字符的过滤），用于捕获并归入 group 1\\n str_pat = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n # 2. com_pat: 匹配多行注释、单行注释、HTML注释，用于捕获并归入 group 2\\n com_pat = r'/\\\\*.*?\\\\*/|//[^\\\\n]*|<!--.*?-->'\\n \\n # 合并后的正则。由于使用了 `|`，正则引擎会从左到右尝试匹配分支。\\n # 这样可以确保匹配到包含类似 `//` 的字符串字面量时，不会被错误地当作注释提取。\\n combined_pat = f'({str_pat})|({com_pat})'\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略由于编码或其他问题无法被正常读取的文件\\n continue\\n \\n # 使用 re.DOTALL 标志，使得多行注释和HTML注释中的 `.` 能够匹配换行符\\n for match in re.finditer(combined_pat, content, flags=re.DOTALL):\\n # 如果 group 2 匹配成功，说明当前捕获到的是代码注释内容\\n if match.group(2):\\n comment = match.group(2)\\n # 遍历该段注释中的所有字符，逐一鉴别\\n for char in comment:\\n if is_emoji(char):\\n # 注释内容中一旦发现包含 emoji，即违反严格排除的要求\\n return False \\n \\n # 遍历结束，所有代码文件中的所有注释均未发现 emoji，要求被顺利满足\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成代码文件中的各个函数，每一个都不能超出 50 行的限制\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def clean_code(code_str: str) -> str:\\n \\\"\\\"\\\"清洗代码，将字符串内内容与注释替换为空格，以防干扰花括号解析\\\"\\\"\\\"\\n res = []\\n i = 0\\n n = len(code_str)\\n state = 'NORMAL'\\n \\n while i < n:\\n c = code_str[i]\\n \\n if state == 'NORMAL':\\n if c == '/' and i + 1 < n and code_str[i+1] == '/':\\n state = 'LINE_COMMENT'\\n res.append(' ')\\n res.append(' ')\\n i += 2\\n continue\\n elif c == '/' and i + 1 < n and code_str[i+1] == '*':\\n state = 'BLOCK_COMMENT'\\n res.append(' ')\\n res.append(' ')\\n i += 2\\n continue\\n elif c == \\\"'\\\":\\n state = 'SINGLE_QUOTE'\\n res.append(' ')\\n i += 1\\n continue\\n elif c == '\\\"':\\n state = 'DOUBLE_QUOTE'\\n res.append(' ')\\n i += 1\\n continue\\n elif c == '`':\\n state = 'BACKTICK'\\n res.append(' ')\\n i += 1\\n continue\\n else:\\n res.append(c)\\n i += 1\\n \\n elif state == 'LINE_COMMENT':\\n if c == '\\\\n':\\n state = 'NORMAL'\\n res.append('\\\\n')\\n else:\\n res.append(' ')\\n i += 1\\n \\n elif state == 'BLOCK_COMMENT':\\n if c == '*' and i + 1 < n and code_str[i+1] == '/':\\n state = 'NORMAL'\\n res.append(' ')\\n res.append(' ')\\n i += 2\\n continue\\n else:\\n if c == '\\\\n':\\n res.append('\\\\n')\\n else:\\n res.append(' ')\\n i += 1\\n \\n elif state in ('SINGLE_QUOTE', 'DOUBLE_QUOTE', 'BACKTICK'):\\n quote_char = \\\"'\\\" if state == 'SINGLE_QUOTE' else ('\\\"' if state == 'DOUBLE_QUOTE' else '`')\\n if c == '\\\\\\\\':\\n res.append(' ')\\n if i + 1 < n:\\n if code_str[i+1] == '\\\\n':\\n res.append('\\\\n')\\n else:\\n res.append(' ')\\n i += 2\\n continue\\n elif c == quote_char:\\n state = 'NORMAL'\\n res.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n res.append('\\\\n')\\n else:\\n res.append(' ')\\n i += 1\\n \\n return \\\"\\\".join(res)\\n\\n def is_function_brace(cleaned_code: str, brace_idx: int) -> bool:\\n \\\"\\\"\\\"向回看判断遇到的 '{' 是否属于函数、箭头函数或方法的开始\\\"\\\"\\\"\\n i = brace_idx - 1\\n while i >= 0 and cleaned_code[i].isspace():\\n i -= 1\\n if i < 0:\\n return False\\n \\n if cleaned_code[i] == '>':\\n if i - 1 >= 0 and cleaned_code[i-1] == '=':\\n return True\\n \\n if cleaned_code[i] == ')':\\n paren_count = 1\\n i -= 1\\n while i >= 0:\\n if cleaned_code[i] == ')':\\n paren_count += 1\\n elif cleaned_code[i] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n i -= 1\\n \\n if i < 0:\\n return False\\n \\n i -= 1\\n while i >= 0 and cleaned_code[i].isspace():\\n i -= 1\\n if i < 0:\\n return False\\n \\n word = \\\"\\\"\\n while i >= 0 and (cleaned_code[i].isalnum() or cleaned_code[i] in ['_', '$']):\\n word = cleaned_code[i] + word\\n i -= 1\\n \\n if word == \\\"function\\\":\\n return True\\n elif word in [\\\"if\\\", \\\"for\\\", \\\"while\\\", \\\"switch\\\", \\\"catch\\\", \\\"with\\\"]:\\n return False\\n else:\\n j = i\\n while j >= 0 and cleaned_code[j].isspace():\\n j -= 1\\n if j >= 0:\\n prev_word = \\\"\\\"\\n while j >= 0 and (cleaned_code[j].isalnum() or cleaned_code[j] in ['_', '$']):\\n prev_word = cleaned_code[j] + prev_word\\n j -= 1\\n if prev_word == \\\"function\\\":\\n return True\\n \\n if word != \\\"\\\":\\n return True\\n \\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n except Exception:\\n continue\\n \\n cleaned = clean_code(code_str)\\n \\n # 预处理行号映射表\\n line_number_map = []\\n current_line = 1\\n for char in cleaned:\\n line_number_map.append(current_line)\\n if char == '\\\\n':\\n current_line += 1\\n \\n n = len(cleaned)\\n for idx in range(n):\\n if cleaned[idx] == '{':\\n if is_function_brace(cleaned, idx):\\n brace_count = 1\\n start_line = line_number_map[idx]\\n end_idx = idx + 1\\n while end_idx < n:\\n if cleaned[end_idx] == '{':\\n brace_count += 1\\n elif cleaned[end_idx] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n break\\n end_idx += 1\\n \\n # 当函数块结束，校验函数的跨越行数\\n if end_idx < n:\\n end_line = line_number_map[end_idx]\\n lines_count = end_line - start_line + 1\\n if lines_count > 50:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["观察看板的基础页面布局：截屏确认页面是否铺满全屏，是否有并排横向排列的列（例如待办、进行中、已完成等），每一列顶部有对应标题、列表区域有卡片展示，底部有“添加任务”按钮。","模拟点击某列底部的“添加任务”按钮，截屏确认界面上弹出了包含“标题”和“详细描述”输入框的模态框；输入测试文本并提交后，截屏确认新创建的卡片成功追加在该列的末尾。","截屏查看已有卡片，确认卡片界面上同时展示了标题和部分描述文本，且右上角包含一个“删除”按钮。模拟点击该删除按钮，截屏确认该卡片从对应列中完全移除。","测试卡片同列排序：模拟鼠标按住某一列中的第一张卡片，向下拖拽到同列的最后一张卡片下方并松开鼠标，截屏确认该卡片在该列内部的排序位置被正确改变。","测试卡片跨列移动：模拟鼠标按住第一列的某张卡片，水平拖动跨越至第二列中并释放鼠标，截屏确认该卡片从第一列消失并成功落入第二列显示。","测试卡片拖拽视觉反馈：模拟鼠标按住某张卡片使其处于被抓起的拖拽状态（不松手），在此状态下截屏，确认该卡片呈现出微微倾斜的角度并带有加深的阴影效果。","测试列级别的悬停视觉反馈：模拟鼠标按住某张卡片，将其拖拽并悬停在另外一个不属于它的目标列的正上方（不松手），截屏确认该目标列的整体背景色出现了明显的变暗或高亮过渡。","测试新增列功能：在页面所有列的最右侧寻找“新增列表”区块，模拟点击并输入新列名称后确认，截屏确认界面的最右侧成功生成了一个带有输入名称的新列。","测试列重命名功能：模拟点击某一列的顶部标题，确认其转化为输入框状态，输入全新的列名称并模拟回车操作，截屏确认该列的标题成功更新为新文字。","测试整列水平拖拽：模拟鼠标按住某一个列的顶部区域，将整列向左或向右水平拖拽到其他列的位置旁并释放鼠标，截屏确认列与列之间的左右排序成功发生了改变。","测试关键字搜索过滤：在页面顶部的搜索框中输入某张卡片特有的关键字，截屏确认面板中仅展示标题或描述匹配该关键字的卡片，不匹配的其他任务卡片全部被隐藏掉。","测试本地持久化：模拟拖拽打乱几张卡片和多个列的位置；并至少执行一次新增列或对某列重命名操作，然后刷新浏览器页面；页面重新加载后截屏，确认看板上卡片的位置、列的左右排序、新增列及重命名后的列标题均与刷新前一致，未发生丢失或复原。"],"string":"[\n \"观察看板的基础页面布局：截屏确认页面是否铺满全屏，是否有并排横向排列的列（例如待办、进行中、已完成等），每一列顶部有对应标题、列表区域有卡片展示，底部有“添加任务”按钮。\",\n \"模拟点击某列底部的“添加任务”按钮，截屏确认界面上弹出了包含“标题”和“详细描述”输入框的模态框；输入测试文本并提交后，截屏确认新创建的卡片成功追加在该列的末尾。\",\n \"截屏查看已有卡片，确认卡片界面上同时展示了标题和部分描述文本，且右上角包含一个“删除”按钮。模拟点击该删除按钮，截屏确认该卡片从对应列中完全移除。\",\n \"测试卡片同列排序：模拟鼠标按住某一列中的第一张卡片，向下拖拽到同列的最后一张卡片下方并松开鼠标，截屏确认该卡片在该列内部的排序位置被正确改变。\",\n \"测试卡片跨列移动：模拟鼠标按住第一列的某张卡片，水平拖动跨越至第二列中并释放鼠标，截屏确认该卡片从第一列消失并成功落入第二列显示。\",\n \"测试卡片拖拽视觉反馈：模拟鼠标按住某张卡片使其处于被抓起的拖拽状态（不松手），在此状态下截屏，确认该卡片呈现出微微倾斜的角度并带有加深的阴影效果。\",\n \"测试列级别的悬停视觉反馈：模拟鼠标按住某张卡片，将其拖拽并悬停在另外一个不属于它的目标列的正上方（不松手），截屏确认该目标列的整体背景色出现了明显的变暗或高亮过渡。\",\n \"测试新增列功能：在页面所有列的最右侧寻找“新增列表”区块，模拟点击并输入新列名称后确认，截屏确认界面的最右侧成功生成了一个带有输入名称的新列。\",\n \"测试列重命名功能：模拟点击某一列的顶部标题，确认其转化为输入框状态，输入全新的列名称并模拟回车操作，截屏确认该列的标题成功更新为新文字。\",\n \"测试整列水平拖拽：模拟鼠标按住某一个列的顶部区域，将整列向左或向右水平拖拽到其他列的位置旁并释放鼠标，截屏确认列与列之间的左右排序成功发生了改变。\",\n \"测试关键字搜索过滤：在页面顶部的搜索框中输入某张卡片特有的关键字，截屏确认面板中仅展示标题或描述匹配该关键字的卡片，不匹配的其他任务卡片全部被隐藏掉。\",\n \"测试本地持久化：模拟拖拽打乱几张卡片和多个列的位置；并至少执行一次新增列或对某列重命名操作，然后刷新浏览器页面；页面重新加载后截屏，确认看板上卡片的位置、列的左右排序、新增列及重命名后的列标题均与刷新前一致，未发生丢失或复原。\"\n]","isConversational":false}}},{"rowIdx":65,"cells":{"task_id":{"kind":"number","value":66,"string":"66"},"repository_policy":{"kind":"string","value":"请你在后续所有轮次的工作中严格遵循以下约束条件：\n1. 必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改。\n2. 在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令。\n3. 为了提升效率，务必尽可能地并发执行多个独立的工具调用。\n4. 在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则。\n5. 请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起。\n6. 务必始终采用中文作为生成用户回复的语言。\n7. 在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]}],"string":"[\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用Vue.js写一个项目成本计算器的基础页面骨架。需要一个带标题的顶部导航栏，页面中间是一个主面板，面板里放一个简单的占位表格（比如随便填两行项目A、项目B的占位数据，不需要实际计算），底部固定显示一个总计金额：0元。请确保整体是响应式布局，界面看起来要整洁专业，为后续添加具体功能留出空间。另外请后续始终遵循以下两项要求：第一，生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径；第二，必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":1,"instruction":"好，现在把中间那个占位的表格改成真正的成本计算表单。我们需要让用户可以输入默认的三个工种（UI设计、前端开发、后端开发）的'所需工时'。然后在表格顶部加一个'全局基础时薪'的输入框。表格里面要根据工时和全局时薪，实时计算并展示出每一行的单项成本，同时底部的总计金额也要跟着实时累加变化。另外请后续始终遵循以下两项更新要求：第一，你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径；第二，无论生成多少个代码文件，其内部的注释内容均要求全是中文。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":2,"instruction":"干得不错。接下来让表格变得更灵活，支持用户自定义工种。请加上一个'添加新工种'的按钮，点击后能在表格末尾增加一行，允许输入自定义的工种名称和工时。同时，在每一行后面加一个'删除'按钮（注意：至少要保留一行不能全部删光）。再加一点简单的验证：工时输入框如果输入负数要自动置为0。另外请后续始终遵循以下两项更新要求：第一，无论生成多少个代码文件，其内部的注释内容均要求全是英文；第二，确保你生成的任何代码文件中都不包含\"TODO\"。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，则说明没有任何代码文件生成，视作未违反不包含\"TODO\"的规则\n if not os.path.exists(workspace_path):\n return True\n\n # 针对 Vue.js 项目，推断可能的代码文件扩展名\n code_extensions = ('.vue', '.js', '.ts', '.html', '.css', '.scss', '.less', '.json')\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖包和版本控制目录，避免误判\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n # 仅检查生成的代码文件\n if file.endswith(code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查文件内容是否包含\"TODO\"\n if \"TODO\" in content:\n return False\n except Exception:\n # 如果遇到非文本内容或编码错误无法正常读取的文件则跳过\n continue\n \n return True","tags":["Content","Keyword"]}]},{"round_id":3,"instruction":"现在我们要让费用计算更精准一些。请取消之前那个'全局基础时薪'，改为在表格中为每一行（每个工种）提供独立的'时薪'输入框，并且默认给不同工种设个初始值。另外，在底部的总价计算区域，增加'折扣率(%)'和'税率(%)'两个全局输入框，将展示结果拆分为三项：'合计金额'、'折后金额'和'最终含税总价'，它们必须能根据用户的输入实时联动计算出来。另外请后续始终遵循以下两项更新要求：第一，对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹；第二，对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能生成的代码文件后缀集合\n code_extensions = {\n '.js', '.vue', '.html', '.css', '.scss', '.sass', '.less',\n '.ts', '.jsx', '.tsx', '.json', '.py', '.sh'\n }\n\n # 遍历当前项目地址\n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方库、版本控制系统或环境相关的目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件类型\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 判断代码内容中是否包含 \"TODO\"\n if \"TODO\" in content:\n return False\n except UnicodeDecodeError:\n # 遇到无法使用 utf-8 解码的非纯文本文件直接跳过\n continue\n except Exception:\n # 跳过其他可能的读取错误\n continue\n\n return True","tags":["Content","Keyword"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n\n # 定义需要检查的常见代码文件扩展名\n code_extensions = {'.vue', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.py'}\n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的无关或依赖目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n # 尝试读取文件，忽略可能的解码错误以防非文本或异常编码\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n \n # 如果文件为空，则没有首行，一定不符合要求\n if not lines:\n return False\n \n # 获取第一行并检查是否包含目标编码声明\n first_line = lines[0].strip()\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n \n files_checked += 1\n except Exception:\n # 若文件无法读取等异常发生，假设该文件未能验证通过\n return False\n \n # 如果遍历了项目目录但没有找到任何有效代码文件，则不满足已生成全部代码文件首行声明的要求\n return files_checked > 0","tags":["Content","Format"]}]},{"round_id":4,"instruction":"接下来的功能有些挑战。真实的项目往往是分阶段进行的，所以我们需要把当前的单一表格升级为'按阶段分组'的层级嵌套形式。比如划分为'需求阶段'、'开发阶段'等。请允许用户新增不同的阶段，并且可以在指定的阶段内去添加/删除具体的工种。计算逻辑也要同步更新：每个阶段下方要显示该阶段的'小计金额'，而最底部的总金额区要汇聚所有阶段的数据再进行折扣和税率的计算。另外请后续始终遵循以下两项更新要求：第一，确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"；第二，请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 基于用户指令（Vue.js项目以及编码声明）推断可能的代码文件扩展名\n target_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.py'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不相关的常见依赖或隐藏目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if 'dist' in dirs:\n dirs.remove('dist')\n\n for file in files:\n _, ext = os.path.splitext(file)\n if ext.lower() in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查文件内容中是否包含 \"TODO\" 或 \"FIXME\"\n if \"TODO\" in content or \"FIXME\" in content:\n return False\n except Exception:\n # 对于非UTF-8格式或不可读的异常文件跳过\n continue\n\n return True","tags":["Content","Keyword"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not workspace_path or not os.path.isdir(workspace_path):\n return False\n \n # 定义Vue.js项目常见代码文件的扩展名\n code_extensions = {'.js', '.vue', '.html', '.css', '.scss', '.less', '.ts', '.jsx', '.tsx'}\n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制、依赖包以及构建产物等无关目录\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', 'dist', 'build', '.idea', '.vscode')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found += 1\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含指定的编码声明文本\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于编码或其他问题无法正常读取，视作不满足要求\n return False\n \n # 确保项目中至少存在一个代码文件，并且全部通过了首行声明校验\n return code_files_found > 0","tags":["Content","Format"]},{"constraint":"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":5,"instruction":"很好！功能越来越完善了。这一轮我们重点优化移动端的展示和界面的交互。当屏幕宽度小于768px时，传统的表格会很难看，请利用CSS媒体查询，把移动端的每一行明细数据自动变为'卡片式'布局上下排列。另外增强一下交互体验：当最底部的'最终含税总价'变动时，加上一个数字从小到大滚动递增的动画效果；如果用户新增工种却没有填写名字，触发计算时要把那个名字输入框标红提醒。另外请后续始终遵循以下两项更新要求：第一，确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"、\"BUG\"、\"NOTE\"；第二，你创建的每一个代码文件，算上空行后的总行数都必须是奇数值。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"、\"BUG\"、\"NOTE\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n forbidden_words = [\"TODO\", \"FIXME\", \"BUG\", \"NOTE\"]\n # 定义Vue.js项目中常见的代码文件后缀\n code_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.json', '.py'}\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方依赖和Git版本控制相关等无需检查的系统目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build')]\n \n for file in files:\n # 获取文件后缀名\n _, ext = os.path.splitext(file)\n if ext.lower() in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试读取文件内容，使用 errors='ignore' 防止因为特殊字符导致解码失败\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 检查文件中是否包含任何被禁止的关键词\n for word in forbidden_words:\n if word in content:\n return False\n except Exception:\n # 如果遇到文件读取异常则跳过该文件\n continue\n \n # 如果遍历检查后均未发现违规关键词，则代表满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的前端与后端代码文件扩展名集合\n code_extensions = {\n '.vue', '.js', '.jsx', '.ts', '.tsx', \n '.html', '.css', '.scss', '.sass', '.less', \n '.py', '.sh', '.java', '.c', '.cpp', '.go', '.php', '.rb'\n }\n \n target_declaration = \"-*- coding: utf-8 -*-\"\n found_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的不需要检查的目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.vscode', '.idea')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 判断目标声明是否在第一行中\n if target_declaration not in first_line:\n return False\n except Exception:\n # 如果生成的文件完全无法作为文本读取，视为不符合规范\n return False\n \n # 如果没有找到任何代码文件，说明可能没有生成有效代码，返回 False\n if not found_code_file:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据Vue.js项目推断可能的代码文件扩展名\n code_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.scss', '.less', '.json'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 使用 split('\\n') 能够将所有的换行严格切分，包括末尾空行，与编辑器行数统计一致\n lines = content.split('\\n')\n \n # 判断行数是否为奇数，如果是偶数则不满足要求\n if len(lines) % 2 == 0:\n return False\n \n has_code_file = True\n except Exception:\n # 如果由于特殊情况（如编码异常）无法读取某些特定文件，则安全跳过\n continue\n \n # 如果找到了至少一个代码文件且均通过奇数验证，则返回 True\n return has_code_file","tags":["Quantity","Exact Value"]}]},{"round_id":6,"instruction":"现在我想加上多方案管理和本地暂存功能。在页面侧边栏（移动端可以是抽屉菜单）加一个'历史方案'列表。页面顶部加一个'保存当前方案'按钮，点击后弹窗要求输入方案名称，并把当前所有阶段数据、折扣和税率存到本地缓存(localStorage)里。当用户点击列表里的历史方案时，能够将当时的数据完整还原到主界面进行查看或继续修改。另外请后续始终遵循以下两项更新要求：第一，无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划；第二，此前要求的代码文件算上空行后总行数为奇数值的规则即日起调整为：你输出的每一个代码文件，包含空行在内的总行数均为偶数，请后续始终遵循该要求。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"、\"BUG\"、\"NOTE\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义不允许出现的标记词\n forbidden_words = [\"TODO\", \"FIXME\", \"BUG\", \"NOTE\"]\n \n # 定义常见的前端项目代码文件扩展名\n code_extensions = {\".vue\", \".js\", \".ts\", \".jsx\", \".tsx\", \".html\", \".css\", \".scss\", \".less\", \".json\"}\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的常见非业务代码目录\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', '__pycache__', 'dist', 'build')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码类型文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查是否包含任何被禁止的标记词\n for word in forbidden_words:\n if word in content:\n return False\n except Exception:\n # 如果由于编码等问题无法读取文件，可以安全跳过\n pass\n \n return True","tags":["Content","Keyword"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断可能的代码文件类型\n target_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.json'}\n has_target_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的非源码目录\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_target_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含该编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法读取，认为不满足要求\n return False\n \n # 如果项目中没有任何代码文件，则也视为不满足生成代码文件的要求\n return has_target_file","tags":["Content","Format"]},{"constraint":"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你输出的每一个代码文件，包含空行在内的总行数均为偶数","validation_code":"import os\n\ndef verify(response, workspace_path):\n # 根据用户指令推断可能涉及的代码及配置文件类型\n valid_extensions = ['.vue', '.js', '.ts', '.html', '.css', '.scss', '.json']\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 按照换行符切分以统计包含空行在内的所有行数\n lines = content.split('\\n')\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如遇无法读取等特殊情况，直接忽略\n continue\n \n # 如果找到了至少一个代码文件并且全都是偶数行，则返回True\n return has_code_file","tags":["Quantity","Exact Value"]}]},{"round_id":7,"instruction":"最后一步，加上导出功能，方便我们把计算好的报价发给客户。请在底部总价区域旁边加两个按钮：一个是'导出CSV'，能把所有阶段和明细数据拼装成CSV文件触发下载；另一个是'生成报价单'，点击后把当前的整个计算面板区域生成一张图片保存下来（生成图片时要注意把删除按钮、添加按钮等多余的操作控件隐藏掉，只保留只读数据和总价）。另外请后续始终遵循以下两项更新要求：第一，请在生成的代码中为所有的外部 I/O 操作、网络请求及数据格式转换强制添加异常捕获逻辑，确保出现异常时会有错误信息返回；第二，此前要求的代码文件不得包含指定关键词的规则即日起调整为：确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"、\"BUG\"、\"NOTE\"、\"HACK\"、\"XXX\"，请后续始终遵循该要求。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"为了提升效率，务必尽可能地并发执行多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"务必始终采用中文作为生成用户回复的语言","validation_code":"","tags":["Language","Response Language"]},{"constraint":"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹","validation_code":"","tags":["Environment","File Path"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你生成的任何代码文件中都不包含\"TODO\"、\"FIXME\"、\"BUG\"、\"NOTE\"、\"HACK\"、\"XXX\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 要求中明确指出的禁用关键词\n forbidden_keywords = [\"TODO\", \"FIXME\", \"BUG\", \"NOTE\", \"HACK\", \"XXX\"]\n \n # 根据用户指令(Vue.js项目，JSON文件读写等)推断的可能的代码文件后缀\n code_extensions = ('.vue', '.js', '.ts', '.html', '.css', '.json')\n \n if not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 过滤出代码文件\n if file_name.endswith(code_extensions):\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查文件内容是否包含禁用关键词\n for keyword in forbidden_keywords:\n if keyword in content:\n return False\n except Exception:\n # 忽略因编码等原因无法读取的文件\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义 Vue.js 项目中常见的代码文件扩展名\n code_extensions = {'.js', '.vue', '.html', '.css', '.ts', '.jsx', '.tsx', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n\n def check_file(file_path: str) -> bool:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含目标编码声明，以兼容不同的语言注释符号\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n return True\n except Exception:\n return False\n\n # 如果 workspace_path 本身就是单个文件\n if os.path.isfile(workspace_path):\n ext = os.path.splitext(workspace_path)[1].lower()\n if ext in code_extensions:\n return check_file(workspace_path)\n return False\n\n checked_any = False\n \n # 如果 workspace_path 是目录，遍历寻找代码文件\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的常见非源代码目录\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', '__pycache__', 'dist', 'build']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n checked_any = True\n \n # 若发现任一代码文件的首行不满足要求，则直接返回 False\n if not check_file(file_path):\n return False\n \n # 确保至少检查了一个代码文件\n return checked_any","tags":["Content","Format"]},{"constraint":"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你输出的每一个代码文件，包含空行内的总行数均为偶数","validation_code":"def check_code_files_even_lines(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的 Web/Vue 项目代码文件后缀\n code_extensions = {'.vue', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除掉常见的无需校验的目录，避免干扰\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统一将跨平台的 \\r\\n 换行符归一化为 \\n\n content = content.replace('\\r\\n', '\\n')\n \n # 按照换行符进行切分，所得列表长度即为包含空行的总行数\n num_lines = len(content.split('\\n'))\n \n # 验证包含空行在内的总行数是否均为偶数\n if num_lines % 2 != 0:\n return False\n \n except Exception:\n # 如果发生文件读取错误（如非UTF-8的意外文件等），跳过该文件继续检查\n pass\n \n # 如果项目中没有任何代码文件，同样视为不符合要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请在生成的代码中为所有的外部 I/O 操作、网络请求及数据格式转换强制添加异常捕获逻辑，确保出现异常时会有错误信息返回","validation_code":"","tags":["Style","Paradigm"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用Vue.js写一个项目成本计算器的基础页面骨架。需要一个带标题的顶部导航栏，页面中间是一个主面板，面板里放一个简单的占位表格（比如随便填两行项目A、项目B的占位数据，不需要实际计算），底部固定显示一个总计金额：0元。请确保整体是响应式布局，界面看起来要整洁专业，为后续添加具体功能留出空间。另外请后续始终遵循以下两项要求：第一，生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径；第二，必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码里包含的文件引用，一律采用相对路径，禁止出现绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，现在把中间那个占位的表格改成真正的成本计算表单。我们需要让用户可以输入默认的三个工种（UI设计、前端开发、后端开发）的'所需工时'。然后在表格顶部加一个'全局基础时薪'的输入框。表格里面要根据工时和全局时薪，实时计算并展示出每一行的单项成本，同时底部的总计金额也要跟着实时累加变化。另外请后续始终遵循以下两项更新要求：第一，你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径；第二，无论生成多少个代码文件，其内部的注释内容均要求全是中文。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"干得不错。接下来让表格变得更灵活，支持用户自定义工种。请加上一个'添加新工种'的按钮，点击后能在表格末尾增加一行，允许输入自定义的工种名称和工时。同时，在每一行后面加一个'删除'按钮（注意：至少要保留一行不能全部删光）。再加一点简单的验证：工时输入框如果输入负数要自动置为0。另外请后续始终遵循以下两项更新要求：第一，无论生成多少个代码文件，其内部的注释内容均要求全是英文；第二，确保你生成的任何代码文件中都不包含\\\"TODO\\\"。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，则说明没有任何代码文件生成，视作未违反不包含\\\"TODO\\\"的规则\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 针对 Vue.js 项目，推断可能的代码文件扩展名\\n code_extensions = ('.vue', '.js', '.ts', '.html', '.css', '.scss', '.less', '.json')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖包和版本控制目录，避免误判\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n # 仅检查生成的代码文件\\n if file.endswith(code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查文件内容是否包含\\\"TODO\\\"\\n if \\\"TODO\\\" in content:\\n return False\\n except Exception:\\n # 如果遇到非文本内容或编码错误无法正常读取的文件则跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在我们要让费用计算更精准一些。请取消之前那个'全局基础时薪'，改为在表格中为每一行（每个工种）提供独立的'时薪'输入框，并且默认给不同工种设个初始值。另外，在底部的总价计算区域，增加'折扣率(%)'和'税率(%)'两个全局输入框，将展示结果拆分为三项：'合计金额'、'折后金额'和'最终含税总价'，它们必须能根据用户的输入实时联动计算出来。另外请后续始终遵循以下两项更新要求：第一，对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹；第二，对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能生成的代码文件后缀集合\\n code_extensions = {\\n '.js', '.vue', '.html', '.css', '.scss', '.sass', '.less',\\n '.ts', '.jsx', '.tsx', '.json', '.py', '.sh'\\n }\\n\\n # 遍历当前项目地址\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方库、版本控制系统或环境相关的目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件类型\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 判断代码内容中是否包含 \\\"TODO\\\"\\n if \\\"TODO\\\" in content:\\n return False\\n except UnicodeDecodeError:\\n # 遇到无法使用 utf-8 解码的非纯文本文件直接跳过\\n continue\\n except Exception:\\n # 跳过其他可能的读取错误\\n continue\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {'.vue', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.py'}\\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的无关或依赖目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试读取文件，忽略可能的解码错误以防非文本或异常编码\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n \\n # 如果文件为空，则没有首行，一定不符合要求\\n if not lines:\\n return False\\n \\n # 获取第一行并检查是否包含目标编码声明\\n first_line = lines[0].strip()\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n \\n files_checked += 1\\n except Exception:\\n # 若文件无法读取等异常发生，假设该文件未能验证通过\\n return False\\n \\n # 如果遍历了项目目录但没有找到任何有效代码文件，则不满足已生成全部代码文件首行声明的要求\\n return files_checked > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"接下来的功能有些挑战。真实的项目往往是分阶段进行的，所以我们需要把当前的单一表格升级为'按阶段分组'的层级嵌套形式。比如划分为'需求阶段'、'开发阶段'等。请允许用户新增不同的阶段，并且可以在指定的阶段内去添加/删除具体的工种。计算逻辑也要同步更新：每个阶段下方要显示该阶段的'小计金额'，而最底部的总金额区要汇聚所有阶段的数据再进行折扣和税率的计算。另外请后续始终遵循以下两项更新要求：第一，确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"；第二，请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 基于用户指令（Vue.js项目以及编码声明）推断可能的代码文件扩展名\\n target_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.py'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不相关的常见依赖或隐藏目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'dist' in dirs:\\n dirs.remove('dist')\\n\\n for file in files:\\n _, ext = os.path.splitext(file)\\n if ext.lower() in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查文件内容中是否包含 \\\"TODO\\\" 或 \\\"FIXME\\\"\\n if \\\"TODO\\\" in content or \\\"FIXME\\\" in content:\\n return False\\n except Exception:\\n # 对于非UTF-8格式或不可读的异常文件跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not workspace_path or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义Vue.js项目常见代码文件的扩展名\\n code_extensions = {'.js', '.vue', '.html', '.css', '.scss', '.less', '.ts', '.jsx', '.tsx'}\\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制、依赖包以及构建产物等无关目录\\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', 'dist', 'build', '.idea', '.vscode')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found += 1\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含指定的编码声明文本\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于编码或其他问题无法正常读取，视作不满足要求\\n return False\\n \\n # 确保项目中至少存在一个代码文件，并且全部通过了首行声明校验\\n return code_files_found > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"很好！功能越来越完善了。这一轮我们重点优化移动端的展示和界面的交互。当屏幕宽度小于768px时，传统的表格会很难看，请利用CSS媒体查询，把移动端的每一行明细数据自动变为'卡片式'布局上下排列。另外增强一下交互体验：当最底部的'最终含税总价'变动时，加上一个数字从小到大滚动递增的动画效果；如果用户新增工种却没有填写名字，触发计算时要把那个名字输入框标红提醒。另外请后续始终遵循以下两项更新要求：第一，确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"、\\\"BUG\\\"、\\\"NOTE\\\"；第二，你创建的每一个代码文件，算上空行后的总行数都必须是奇数值。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"、\\\"BUG\\\"、\\\"NOTE\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n forbidden_words = [\\\"TODO\\\", \\\"FIXME\\\", \\\"BUG\\\", \\\"NOTE\\\"]\\n # 定义Vue.js项目中常见的代码文件后缀\\n code_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.json', '.py'}\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方依赖和Git版本控制相关等无需检查的系统目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build')]\\n \\n for file in files:\\n # 获取文件后缀名\\n _, ext = os.path.splitext(file)\\n if ext.lower() in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试读取文件内容，使用 errors='ignore' 防止因为特殊字符导致解码失败\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 检查文件中是否包含任何被禁止的关键词\\n for word in forbidden_words:\\n if word in content:\\n return False\\n except Exception:\\n # 如果遇到文件读取异常则跳过该文件\\n continue\\n \\n # 如果遍历检查后均未发现违规关键词，则代表满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的前端与后端代码文件扩展名集合\\n code_extensions = {\\n '.vue', '.js', '.jsx', '.ts', '.tsx', \\n '.html', '.css', '.scss', '.sass', '.less', \\n '.py', '.sh', '.java', '.c', '.cpp', '.go', '.php', '.rb'\\n }\\n \\n target_declaration = \\\"-*- coding: utf-8 -*-\\\"\\n found_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.vscode', '.idea')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 判断目标声明是否在第一行中\\n if target_declaration not in first_line:\\n return False\\n except Exception:\\n # 如果生成的文件完全无法作为文本读取，视为不符合规范\\n return False\\n \\n # 如果没有找到任何代码文件，说明可能没有生成有效代码，返回 False\\n if not found_code_file:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个代码文件，算上空行后的总行数都必须是奇数值\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据Vue.js项目推断可能的代码文件扩展名\\n code_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.scss', '.less', '.json'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 使用 split('\\\\n') 能够将所有的换行严格切分，包括末尾空行，与编辑器行数统计一致\\n lines = content.split('\\\\n')\\n \\n # 判断行数是否为奇数，如果是偶数则不满足要求\\n if len(lines) % 2 == 0:\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果由于特殊情况（如编码异常）无法读取某些特定文件，则安全跳过\\n continue\\n \\n # 如果找到了至少一个代码文件且均通过奇数验证，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"现在我想加上多方案管理和本地暂存功能。在页面侧边栏（移动端可以是抽屉菜单）加一个'历史方案'列表。页面顶部加一个'保存当前方案'按钮，点击后弹窗要求输入方案名称，并把当前所有阶段数据、折扣和税率存到本地缓存(localStorage)里。当用户点击列表里的历史方案时，能够将当时的数据完整还原到主界面进行查看或继续修改。另外请后续始终遵循以下两项更新要求：第一，无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划；第二，此前要求的代码文件算上空行后总行数为奇数值的规则即日起调整为：你输出的每一个代码文件，包含空行在内的总行数均为偶数，请后续始终遵循该要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"、\\\"BUG\\\"、\\\"NOTE\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义不允许出现的标记词\\n forbidden_words = [\\\"TODO\\\", \\\"FIXME\\\", \\\"BUG\\\", \\\"NOTE\\\"]\\n \\n # 定义常见的前端项目代码文件扩展名\\n code_extensions = {\\\".vue\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".html\\\", \\\".css\\\", \\\".scss\\\", \\\".less\\\", \\\".json\\\"}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的常见非业务代码目录\\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', '__pycache__', 'dist', 'build')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码类型文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查是否包含任何被禁止的标记词\\n for word in forbidden_words:\\n if word in content:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取文件，可以安全跳过\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断可能的代码文件类型\\n target_extensions = {'.vue', '.js', '.ts', '.html', '.css', '.json'}\\n has_target_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的非源码目录\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_target_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含该编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法读取，认为不满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，则也视为不满足生成代码文件的要求\\n return has_target_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你输出的每一个代码文件，包含空行在内的总行数均为偶数\",\n \"validation_code\": \"import os\\n\\ndef verify(response, workspace_path):\\n # 根据用户指令推断可能涉及的代码及配置文件类型\\n valid_extensions = ['.vue', '.js', '.ts', '.html', '.css', '.scss', '.json']\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 按照换行符切分以统计包含空行在内的所有行数\\n lines = content.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如遇无法读取等特殊情况，直接忽略\\n continue\\n \\n # 如果找到了至少一个代码文件并且全都是偶数行，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"最后一步，加上导出功能，方便我们把计算好的报价发给客户。请在底部总价区域旁边加两个按钮：一个是'导出CSV'，能把所有阶段和明细数据拼装成CSV文件触发下载；另一个是'生成报价单'，点击后把当前的整个计算面板区域生成一张图片保存下来（生成图片时要注意把删除按钮、添加按钮等多余的操作控件隐藏掉，只保留只读数据和总价）。另外请后续始终遵循以下两项更新要求：第一，请在生成的代码中为所有的外部 I/O 操作、网络请求及数据格式转换强制添加异常捕获逻辑，确保出现异常时会有错误信息返回；第二，此前要求的代码文件不得包含指定关键词的规则即日起调整为：确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"、\\\"BUG\\\"、\\\"NOTE\\\"、\\\"HACK\\\"、\\\"XXX\\\"，请后续始终遵循该要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"为了提升效率，务必尽可能地并发执行多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在输出代码的所有 import/require 语句中，引用的包名必须严格保持按字母顺序升序排列的规则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"请确保在运行任何 bash 测试、安装或构建命令的最前方加上 `timeout 30s` 包装，目的是防止进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"务必始终采用中文作为生成用户回复的语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中，任何涉及到文件引用的地方都必须是绝对文件路径，不允许使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对于回复中出现的任何文件路径和代码位置，都绝对只能使用单引号来进行包裹\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含\\\"TODO\\\"、\\\"FIXME\\\"、\\\"BUG\\\"、\\\"NOTE\\\"、\\\"HACK\\\"、\\\"XXX\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 要求中明确指出的禁用关键词\\n forbidden_keywords = [\\\"TODO\\\", \\\"FIXME\\\", \\\"BUG\\\", \\\"NOTE\\\", \\\"HACK\\\", \\\"XXX\\\"]\\n \\n # 根据用户指令(Vue.js项目，JSON文件读写等)推断的可能的代码文件后缀\\n code_extensions = ('.vue', '.js', '.ts', '.html', '.css', '.json')\\n \\n if not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 过滤出代码文件\\n if file_name.endswith(code_extensions):\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查文件内容是否包含禁用关键词\\n for keyword in forbidden_keywords:\\n if keyword in content:\\n return False\\n except Exception:\\n # 忽略因编码等原因无法读取的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义 Vue.js 项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.vue', '.html', '.css', '.ts', '.jsx', '.tsx', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n def check_file(file_path: str) -> bool:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含目标编码声明，以兼容不同的语言注释符号\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n return True\\n except Exception:\\n return False\\n\\n # 如果 workspace_path 本身就是单个文件\\n if os.path.isfile(workspace_path):\\n ext = os.path.splitext(workspace_path)[1].lower()\\n if ext in code_extensions:\\n return check_file(workspace_path)\\n return False\\n\\n checked_any = False\\n \\n # 如果 workspace_path 是目录，遍历寻找代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的常见非源代码目录\\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', '__pycache__', 'dist', 'build']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n checked_any = True\\n \\n # 若发现任一代码文件的首行不满足要求，则直接返回 False\\n if not check_file(file_path):\\n return False\\n \\n # 确保至少检查了一个代码文件\\n return checked_any\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请仅实现现阶段明确指出的功能，杜绝过度设计和面向未来的特性预留，且必须删除任何闲置的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你输出的每一个代码文件，包含空行内的总行数均为偶数\",\n \"validation_code\": \"def check_code_files_even_lines(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的 Web/Vue 项目代码文件后缀\\n code_extensions = {'.vue', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉常见的无需校验的目录，避免干扰\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统一将跨平台的 \\\\r\\\\n 换行符归一化为 \\\\n\\n content = content.replace('\\\\r\\\\n', '\\\\n')\\n \\n # 按照换行符进行切分，所得列表长度即为包含空行的总行数\\n num_lines = len(content.split('\\\\n'))\\n \\n # 验证包含空行在内的总行数是否均为偶数\\n if num_lines % 2 != 0:\\n return False\\n \\n except Exception:\\n # 如果发生文件读取错误（如非UTF-8的意外文件等），跳过该文件继续检查\\n pass\\n \\n # 如果项目中没有任何代码文件，同样视为不符合要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请在生成的代码中为所有的外部 I/O 操作、网络请求及数据格式转换强制添加异常捕获逻辑，确保出现异常时会有错误信息返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面整体布局结构，确认页面顶部包含带标题的导航栏，页面中间是展示数据的主面板，页面底部有固定显示金额汇总计算信息的区域。","观察主面板的数据展示结构，确认表单被划分为多个'阶段'分组（如需求阶段、开发阶段）。每个阶段内部以列表形式展示工种信息（包含工种名称、时薪、工时和单项成本），并在该阶段的底部显示该阶段的'小计金额'。","点击主面板区域的'新增阶段'按钮，确认页面中会动态添加一个新的阶段区块，并且能够输入或显示该阶段名称，从而区分不同阶段。","在任意已有阶段内，点击'添加新工种'按钮，确认该阶段内部的工种列表末尾新增了一行空白的工种输入区域。","点击任意工种行右侧的'删除'按钮，确认该行工种数据被移除；不断删除工种直至剩下最后一行，确认最后一行的删除按钮不可点击、被隐藏或点击无效，确保全局或阶段内至少保留一行工种不会被删光。","在任意工种的'工时'输入框中输入负数（如-5），光标离开输入框或输入完成后，确认该工时输入框的数值自动被重置为0。","在某个工种行中手动输入时薪'200'和工时'10'，确认该行对应的单项成本自动变为'2000'；同时确认该工种所属阶段底部的'小计金额'正确累加了这2000元。","在底部固定汇总区找到'折扣率(%)'和'税率(%)'输入框并分别输入10（代表打九折）和5（代表5%税率），观察底部汇总信息，确认显示的'合计金额'为所有阶段小计的总和，'折后金额'计算准确（合计金额*0.9），'最终含税总价'计算准确（折后金额*1.05）。","将浏览器窗口宽度缩小至768px以下（如开启开发者工具调整视口宽度为375px），确认主面板内的工种明细数据由横向的表格排列方式变更为上下堆叠排列的'卡片式'布局。","修改任意一个工种的工时数值，使得底部的'最终含税总价'发生增加变化，观察总价更新过程，确认数值从旧数值到新数值变化时伴随从小到大滚动递增的动画效果。","新增一个工种但保持'名称'输入框为空（或清空已有的名字），修改其工时触发数据重新计算，确认该未填名称的工种输入框出现红色的边框或文字警告标红提示。","点击页面顶部的'保存当前方案'按钮，确认页面弹出要求输入方案名称的对话框；输入方案名称'项目方案A'并确认，随后展开页面侧边栏或抽屉菜单，确认'项目方案A'已出现在'历史方案'列表中。","在存有'项目方案A'历史记录的情况下，刷新页面后重新展开页面侧边栏或抽屉菜单，确认'项目方案A'仍然保留在'历史方案'列表中；随后故意大幅修改当前页面的工种数据、折扣率和税率，再点击'项目方案A'，确认页面上所有的阶段数据、金额明细、折扣率及税率全部还原至保存时的状态。","点击底部汇总区域的'导出CSV'按钮，确认浏览器自动触发了一个CSV数据文件的下载操作；打开下载的CSV文件，确认其中包含所有阶段名称、各阶段下的工种明细数据以及相关金额信息。","点击底部汇总区域的'生成报价单'按钮，确认浏览器触发下载了一张图片或在弹窗中展示了生成的图片。检查该图片内容，确认其包含当前整个计算面板区域中的阶段、工种明细、只读数据和总价金额，同时原计算表单中的所有操作控件（如删除按钮、添加按钮、保存方案按钮）均被隐藏。"],"string":"[\n \"查看页面整体布局结构，确认页面顶部包含带标题的导航栏，页面中间是展示数据的主面板，页面底部有固定显示金额汇总计算信息的区域。\",\n \"观察主面板的数据展示结构，确认表单被划分为多个'阶段'分组（如需求阶段、开发阶段）。每个阶段内部以列表形式展示工种信息（包含工种名称、时薪、工时和单项成本），并在该阶段的底部显示该阶段的'小计金额'。\",\n \"点击主面板区域的'新增阶段'按钮，确认页面中会动态添加一个新的阶段区块，并且能够输入或显示该阶段名称，从而区分不同阶段。\",\n \"在任意已有阶段内，点击'添加新工种'按钮，确认该阶段内部的工种列表末尾新增了一行空白的工种输入区域。\",\n \"点击任意工种行右侧的'删除'按钮，确认该行工种数据被移除；不断删除工种直至剩下最后一行，确认最后一行的删除按钮不可点击、被隐藏或点击无效，确保全局或阶段内至少保留一行工种不会被删光。\",\n \"在任意工种的'工时'输入框中输入负数（如-5），光标离开输入框或输入完成后，确认该工时输入框的数值自动被重置为0。\",\n \"在某个工种行中手动输入时薪'200'和工时'10'，确认该行对应的单项成本自动变为'2000'；同时确认该工种所属阶段底部的'小计金额'正确累加了这2000元。\",\n \"在底部固定汇总区找到'折扣率(%)'和'税率(%)'输入框并分别输入10（代表打九折）和5（代表5%税率），观察底部汇总信息，确认显示的'合计金额'为所有阶段小计的总和，'折后金额'计算准确（合计金额*0.9），'最终含税总价'计算准确（折后金额*1.05）。\",\n \"将浏览器窗口宽度缩小至768px以下（如开启开发者工具调整视口宽度为375px），确认主面板内的工种明细数据由横向的表格排列方式变更为上下堆叠排列的'卡片式'布局。\",\n \"修改任意一个工种的工时数值，使得底部的'最终含税总价'发生增加变化，观察总价更新过程，确认数值从旧数值到新数值变化时伴随从小到大滚动递增的动画效果。\",\n \"新增一个工种但保持'名称'输入框为空（或清空已有的名字），修改其工时触发数据重新计算，确认该未填名称的工种输入框出现红色的边框或文字警告标红提示。\",\n \"点击页面顶部的'保存当前方案'按钮，确认页面弹出要求输入方案名称的对话框；输入方案名称'项目方案A'并确认，随后展开页面侧边栏或抽屉菜单，确认'项目方案A'已出现在'历史方案'列表中。\",\n \"在存有'项目方案A'历史记录的情况下，刷新页面后重新展开页面侧边栏或抽屉菜单，确认'项目方案A'仍然保留在'历史方案'列表中；随后故意大幅修改当前页面的工种数据、折扣率和税率，再点击'项目方案A'，确认页面上所有的阶段数据、金额明细、折扣率及税率全部还原至保存时的状态。\",\n \"点击底部汇总区域的'导出CSV'按钮，确认浏览器自动触发了一个CSV数据文件的下载操作；打开下载的CSV文件，确认其中包含所有阶段名称、各阶段下的工种明细数据以及相关金额信息。\",\n \"点击底部汇总区域的'生成报价单'按钮，确认浏览器触发下载了一张图片或在弹窗中展示了生成的图片。检查该图片内容，确认其包含当前整个计算面板区域中的阶段、工种明细、只读数据和总价金额，同时原计算表单中的所有操作控件（如删除按钮、添加按钮、保存方案按钮）均被隐藏。\"\n]","isConversational":false}}},{"rowIdx":66,"cells":{"task_id":{"kind":"number","value":67,"string":"67"},"repository_policy":{"kind":"string","value":"你在后续所有轮次的回复与操作中，必须始终遵循以下约束：\n1. 务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的。\n2. 任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改。\n3. 在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录。\n4. 在生成针对用户的回复时，必须且只能一直采用中文。\n5. 请使用纯中文为你生成的每一个代码文件编写注释。\n6. 确保你输出的代码项目文件夹嵌套深度不超过 2 层。\n7. 请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请使用纯中文为你生成的每一个代码文件编写注释","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n rel_path = os.path.relpath(root, workspace_path)\n if rel_path == '.':\n depth = 0\n else:\n depth = len(rel_path.split(os.sep))\n \n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层","validation_code":"","tags":["Quantity","Complexity"]}],"string":"[\n {\n \"constraint\": \"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请使用纯中文为你生成的每一个代码文件编写注释\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n rel_path = os.path.relpath(root, workspace_path)\\n if rel_path == '.':\\n depth = 0\\n else:\\n depth = len(rel_path.split(os.sep))\\n \\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我开发一个Web版的高级计算器应用。第一步，我们先搭建基本的骨架。我需要一个计算器的标准UI：顶部是一个宽大的显示屏区域，用来显示用户的输入；下面是基础的按键区，包含数字键盘（0-9）、小数点、基础四则运算符（+、-、*、/）、等号（=）以及清除按钮（C）和退格键。为了方便后续扩展，请预留出侧边栏的空间。当前你只需要实现UI排版，并让点击按钮时能够将对应字符显示在顶部的屏幕上即可，先不用实现真正的计算逻辑。另外请你后续始终遵循以下要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-，在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔。","instruction_following_checklist":[{"constraint":"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请使用纯中文为你生成的每一个代码文件编写注释","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n\n workspace = Path(workspace_path).resolve()\n \n # 如果工作目录不存在或不是一个目录，从严格意义上并未违反该嵌套要求\n if not workspace.exists() or not workspace.is_dir():\n return True\n\n # 遍历工作目录下所有的文件与文件夹\n for p in workspace.rglob('*'):\n # 仅针对文件夹计算嵌套层级\n if p.is_dir():\n try:\n # 计算当前文件夹相对于根目录的相对路径\n rel_path = p.relative_to(workspace)\n \n # 排除 .git 等常见隐藏文件夹造成的误判干扰\n if any(part.startswith('.') for part in rel_path.parts):\n continue\n \n # rel_path.parts 会将相对路径每一层目录分割为元组，其长度即为当前文件夹的嵌套深度\n # 例如：子文件夹 \"css\" 的长度为 1，\"assets/css\" 的长度为 2\n if len(rel_path.parts) > 2:\n return False\n except ValueError:\n # 当无法计算相对路径时直接跳过\n continue\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_encoding_declaration(response, workspace_path):\n import os\n\n # 定义Web应用及常用开发语言的代码文件后缀名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.vue', '.py', '.java', \n '.c', '.cpp', '.h', '.hpp', '.php'\n }\n has_code_file = False\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含要求中的特定编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果打开文件或读取时遇到异常(如编码错误)，判定为不满足\n return False\n \n # 如果项目中没有任何代码文件，也视作未完成要求；否则满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":1,"instruction":"很好，基础界面搭建得不错。现在我们来注入核心的计算能力，并扩展出科学计算功能。请在基础键盘的左侧或上方扩展一个『科学计算面板』，新增包含括号 `()`、三角函数（`sin`, `cos`, `tan`）、对数（`log`, `ln`）、`√` (平方根)、`^` (指数)、常数 `π` 和 `e` 等按钮。接下来实现完整的表达式解析引擎：要求点击等号时，能正确计算包含科学函数和多层括号嵌套的复杂数学表达式（比如 `sin(π/2) + 2^3`），并遵循正确的运算符优先级。如果用户输入的表达式有语法错误，请在屏幕上以红色字体提示“表达式错误”。另外请你后续始终遵循以下新要求：生成的代码文件中路径的分割请优先选用反斜杠 \\；修改任何已有文件的首要步骤，必须使用相应的工具来阅读它。","instruction_following_checklist":[{"constraint":"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请使用纯中文为你生成的每一个代码文件编写注释","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_workspace_depth(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除以 . 开头的隐藏文件夹，防止非 AI 生成的系统或环境文件造成误判\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录自身的相对路径为 '.'，深度为 0，跳过判断\n if rel_path == '.':\n continue\n \n # 统一路径分隔符并计算嵌套深度\n # os.path.relpath 返回的路径是规范化过的，不包含开头或结尾的多余斜杠\n normalized_path = rel_path.replace('\\\\', '/')\n depth = len(normalized_path.split('/'))\n \n # 如果嵌套深度超过 2 层，则未满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义Web计算器项目及通用开发中可能涉及的代码文件扩展名\n code_exts = {'.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue'}\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法以 utf-8 编码读取或其他异常，也视为不满足要求\n return False\n\n # 若找到了代码文件并且所有代码文件的顶行都符合要求，返回 True\n return found_code_file","tags":["Content","Format"]},{"constraint":"生成的代码文件中路径的分割请优先选用反斜杠 \\","validation_code":"","tags":["Environment","File Path"]},{"constraint":"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":2,"instruction":"计算逻辑很完美。接下来补充前面预留的侧边栏，做成『历史记录』面板。要求：每次用户成功计算出结果后，自动将『完整的表达式 = 结果』作为一个卡片条目追加到侧边栏中。为了提升交互体验，当用户点击侧边栏里的任意一条历史记录时，能够自动将该条目的表达式填回主计算器的显示屏上，方便修改后再计算。最后，在历史记录面板底部加上一个『清空历史』按钮，点击可清空列表。另外请你后续始终遵循以下新要求：生成代码文件时，路径分隔符应当优先采用正斜杠 /；在你输出的代码文件里，不允许出现行数超过 50 行的函数。","instruction_following_checklist":[{"constraint":"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请使用纯中文为你生成的每一个代码文件编写注释","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_folder_depth(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作空间路径不存在，视作没有文件夹生成，符合不超过 2 层的要求\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏文件夹和Python编译产生的缓存文件夹\n # 确保只检查由于AI生成逻辑实质创建的源码目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n rel_path = os.path.relpath(root, workspace_path)\n \n if rel_path == '.':\n depth = 0\n else:\n # 统一路径分隔符，以防跨平台或反斜杠带来的计算错误\n rel_path = rel_path.replace('\\\\', '/')\n depth = len(rel_path.split('/'))\n \n # 如果嵌套深度超过 2 层，则不满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户需求推测可能的 Web 版计算器代码文件后缀\n code_extensions = {'.py', '.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n found_code_file = False\n \n # 遍历当前项目文件夹\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果是代码文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 读取文件顶行内容，忽略可能的非标准编码异常\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若文件无法读取也视为不符合要求\n return False\n \n # 如果找到了至少一个代码文件，且全部验证通过，则返回 True\n return found_code_file","tags":["Content","Format"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你输出的代码文件里，不允许出现行数超过 50 行的函数","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n\n def check_file(filepath):\n # 1. 判断并验证 Python 文件\n if filepath.endswith('.py'):\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n code = f.read()\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\n if node.end_lineno - node.lineno + 1 > 50:\n return False\n except Exception:\n pass # 若语法解析失败可跳过\n return True\n \n # 2. 判断并验证 JS/TS/HTML 前端相关文件\n elif filepath.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 如果是包含 HTML/模板 的文件，仅提取并检验脚本块内的函数\n if filepath.endswith(('.html', '.htm', '.vue')):\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\n content = \"\\n\".join(scripts)\n \n if not content.strip():\n return True\n \n # 用于匹配 JS/TS 函数开始的正则表达式集合\n js_func_patterns = [\n r'\\bfunction\\s*[\\w$]*\\s*\\([^){}]*\\)\\s*\\{', # function 声明\n r'(?:\\([^){}]*\\)|[\\w$]+)\\s*=>\\s*\\{', # 箭头函数 (() => { 或 arg => {)\n r'(?<!@)\\b([\\w$]+)\\s*\\([^){}]*\\)\\s*\\{' # 类/对象方法 (例如 calc() { )\n ]\n \n # 保留字列表，防止条件语句和保留关键字被误认为对象方法函数\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with', 'else', 'return', 'class'}\n \n for pattern in js_func_patterns:\n for match in re.finditer(pattern, content):\n # 如果匹配到保留字结构（例如 if() { ）则跳过\n if len(match.groups()) > 0 and match.group(1) in keywords:\n continue\n \n start_idx = match.end() - 1\n if content[start_idx] != '{':\n start_idx = content.rfind('{', match.start(), match.end())\n \n if start_idx == -1:\n continue\n\n # 将当前字符串的片段映射到行号\n start_lineno = content.count('\\n', 0, start_idx) + 1\n brace_count = 0\n end_idx = start_idx\n \n # 使用简单的状态机规避字符串或注释内部的大括号干扰\n in_string = False\n string_char = ''\n in_single_comment = False\n in_multi_comment = False\n \n i = start_idx\n while i < len(content):\n char = content[i]\n if in_single_comment:\n if char == '\\n':\n in_single_comment = False\n elif in_multi_comment:\n if char == '*' and i + 1 < len(content) and content[i + 1] == '/':\n in_multi_comment = False\n i += 1\n elif in_string:\n if char == '\\\\':\n i += 1\n elif char == string_char:\n in_string = False\n else:\n if char == '/' and i + 1 < len(content) and content[i + 1] == '/':\n in_single_comment = True\n i += 1\n elif char == '/' and i + 1 < len(content) and content[i + 1] == '*':\n in_multi_comment = True\n i += 1\n elif char in ['\\'', '\"', '`']:\n in_string = True\n string_char = char\n elif char == '{':\n brace_count += 1\n elif char == '}':\n brace_count -= 1\n # 函数体闭合\n if brace_count == 0:\n end_idx = i\n break\n i += 1\n \n # 判断所寻找到的函数体的跨越行数\n if brace_count == 0:\n end_lineno = content.count('\\n', 0, end_idx) + 1\n if end_lineno - start_lineno + 1 > 50:\n return False\n return True\n return True\n\n # 遍历当前工作空间内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n # 只要有任何一个代码文件不满足要求，直接判定失败\n if not check_file(filepath):\n return False\n \n return True","tags":["Quantity","Range"]}]},{"round_id":3,"instruction":"好，现在我们要加入最具挑战性的功能：函数绘图。请在整个应用的最顶部增加一个Tab切换器，允许用户在『科学计算器』和『函数绘图』两个视图之间无缝切换。在『函数绘图』视图下，界面分为左右两部分：左侧提供一个输入框供用户输入关于自变量 x 的函数表达式（例如 `sin(x) + x^2`），并带有一个『绘制』按钮；右侧提供一个大型的 Canvas 画布，默认绘制出带有 X 轴、Y 轴和网格线的二维坐标系（默认范围 -10 到 10）。当用户点击绘制按钮时，在坐标系上画出该函数的连续曲线。另外请你后续始终遵循以下新要求：在你输出的代码文件里，不允许出现行数超过 30 行的函数；所有代码在编写完毕后，都必须经过测试且保证测试完全通过。","instruction_following_checklist":[{"constraint":"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请使用纯中文为你生成的每一个代码文件编写注释","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"import os\nfrom pathlib import Path\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断代码项目文件夹嵌套深度是否不超过2层。\n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 如果嵌套不超过2层返回 True，否则返回 False。\n \"\"\"\n if not os.path.exists(workspace_path):\n return False\n \n workspace = Path(workspace_path).resolve()\n \n # 遍历 workspace_path 下的所有文件和文件夹\n for dir_path in workspace.rglob('*'):\n if dir_path.is_dir():\n try:\n # 获取相对路径\n rel_path = dir_path.relative_to(workspace)\n \n # 排除隐藏目录（如 .git, .vscode 等）和 Python 缓存目录（__pycache__）\n if any(part.startswith('.') or part == '__pycache__' for part in rel_path.parts):\n continue\n \n # 判断嵌套深度（相对路径的层级数即为嵌套层数）\n # 例如：dir1 层级数为 1；dir1/dir2 层级数为 2；dir1/dir2/dir3 层级数为 3\n if len(rel_path.parts) > 2:\n return False\n \n except ValueError:\n # 如果由于某种原因无法计算相对路径则忽略\n continue\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的代码文件类型，适用于 Web 应用开发\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的依赖和版本控制文件夹\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于编码等问题无法读取，视为未满足要求\n return False\n \n return has_code_file","tags":["Content","Format"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你输出的代码文件里，不允许出现行数超过 30 行的函数","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def replacer(match):\n s = match.group(0)\n # 将匹配的字符串和注释替换成相同数量的换行和空格，从而保证行数计算不受影响且去除干扰字符\n return ''.join('\\n' if c == '\\n' else ' ' for c in s)\n \n def check_js_functions(source_code: str) -> bool:\n # 去除字符串字面量和注释的干扰，但保留其中的换行符\n pattern = re.compile(\n r'\"(?:\\\\.|[^\"\\\\\\n])*\"|'\n r\"'(?:\\\\.|[^'\\\\\\n])*'|\"\n r\"`(?:\\\\.|[^`\\\\])*`|\"\n r\"/\\*[\\s\\S]*?\\*/|\"\n r\"//.*\"\n )\n try:\n masked_code = pattern.sub(replacer, source_code)\n except Exception:\n masked_code = source_code\n \n def check_function_body(start_lineno: int, found_brace_idx: int) -> bool:\n brace_count = 0\n for i in range(found_brace_idx, len(masked_code)):\n if masked_code[i] == '{':\n brace_count += 1\n elif masked_code[i] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_lineno = masked_code.count('\\n', 0, i) + 1\n # 计算函数的总行数，如果大于30则说明违反要求\n if end_lineno - start_lineno + 1 > 30:\n return False\n break\n return True\n\n # 1. 检查普通函数定义：function name(...) { 或 function(...) {\n for match in re.finditer(r'\\bfunction\\b', masked_code):\n start_lineno = masked_code.count('\\n', 0, match.start()) + 1\n idx = match.end()\n paren_level = 0\n found_brace_idx = -1\n for i in range(idx, len(masked_code)):\n if masked_code[i] == '(':\n paren_level += 1\n elif masked_code[i] == ')':\n paren_level -= 1\n if paren_level < 0:\n break\n elif masked_code[i] == '{' and paren_level == 0:\n found_brace_idx = i\n break\n elif masked_code[i] == ';' and paren_level == 0:\n break\n \n if found_brace_idx != -1:\n if not check_function_body(start_lineno, found_brace_idx):\n return False\n\n # 2. 检查箭头函数：(...) => { 或 arg => {\n for match in re.finditer(r'=>', masked_code):\n start_lineno = masked_code.count('\\n', 0, match.start()) + 1\n idx = match.end()\n paren_level = 0\n found_brace_idx = -1\n for i in range(idx, len(masked_code)):\n if masked_code[i] == '(':\n paren_level += 1\n elif masked_code[i] == ')':\n paren_level -= 1\n if paren_level < 0:\n break\n elif masked_code[i] == '{' and paren_level == 0:\n found_brace_idx = i\n break\n elif masked_code[i] == ';' and paren_level == 0:\n break\n elif paren_level == 0 and masked_code[i] not in ' \\t\\r\\n':\n # 若不是括号或大括号，并且遇到了其他操作符，说明是缺少代码块的单行表达式\n break\n \n if found_brace_idx != -1:\n if not check_function_body(start_lineno, found_brace_idx):\n return False\n\n # 3. 检查类方法或简写对象方法：methodName(...) {\n for match in re.finditer(r'\\b([a-zA-Z_$][\\w$]*)\\s*\\(', masked_code):\n name = match.group(1)\n # 排除控制流关键字和特殊指令\n if name in {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'yield', 'await'}:\n continue\n start_lineno = masked_code.count('\\n', 0, match.start()) + 1\n idx = match.end() - 1 # 游标停在 '(' 的位置\n paren_level = 0\n found_brace_idx = -1\n for i in range(idx, len(masked_code)):\n if masked_code[i] == '(':\n paren_level += 1\n elif masked_code[i] == ')':\n paren_level -= 1\n if paren_level < 0:\n break\n elif masked_code[i] == '{' and paren_level == 0:\n found_brace_idx = i\n break\n elif masked_code[i] == ';' and paren_level == 0:\n break\n elif paren_level == 0 and masked_code[i] not in ' \\t\\r\\n':\n # 如果不是空格且不在参数列表内，并且也不是大括号，说明这只是普通的函数调用而非函数定义\n break\n \n if found_brace_idx != -1:\n if not check_function_body(start_lineno, found_brace_idx):\n return False\n \n return True\n\n def check_python_functions(source_code: str) -> bool:\n try:\n tree = ast.parse(source_code)\n except Exception:\n return True\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n if node.end_lineno is not None and node.lineno is not None:\n if node.end_lineno - node.lineno + 1 > 30:\n return False\n return True\n\n def check_html_functions(source_code: str) -> bool:\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', source_code, re.IGNORECASE | re.DOTALL)\n for script in scripts:\n if not check_js_functions(script):\n return False\n return True\n\n # 递归遍历项目目录中的所有相关文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_python_functions(content):\n return False\n elif ext in {'.js', '.ts', '.jsx', '.tsx'}:\n if not check_js_functions(content):\n return False\n elif ext in {'.html', '.htm', '.vue'}:\n if not check_html_functions(content):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"所有代码在编写完毕后，都必须经过测试且保证测试完全通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"最后，让我们把绘图视图的交互体验拉满。首先，支持多函数绘制：在左侧增加一个『添加函数』按钮，允许动态添加多个输入框（f(x), g(x), h(x)...），且右侧画出的每条曲线必须自动分配不同的颜色加以区分。其次，增强画布的交互能力：支持鼠标在画布上拖拽来平移坐标系，支持通过鼠标滚轮进行坐标系的放大和缩小，缩放和平移时网格线、坐标轴刻度和函数曲线必须实时重新渲染。最后，当鼠标悬浮在画布上移动时，绘制一个跟随鼠标的十字准星，并实时显示当前指针位置对应的精确 (x, y) 坐标数值。另外请你后续始终遵循以下新要求：在输出的代码文件里，不允许出现行数超过 20 行的函数；执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式。","instruction_following_checklist":[{"constraint":"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"请使用纯中文为你生成的每一个代码文件编写注释","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"确保你输出的代码项目文件夹嵌套深度不超过 2 层","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作目录不存在，无法进行判断，默认返回False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 列表，过滤掉隐藏文件夹(以 . 开头)以及常见的系统/依赖生成的文件夹\n # 这样 os.walk 就不会继续向深处遍历这些无关文件夹\n dirs[:] = [\n d for d in dirs \n if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')\n ]\n \n # 获取当前目录相对于 workspace_path 的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录的相对路径为 '.'，深度记为 0\n if rel_path == '.':\n depth = 0\n else:\n # 兼容不同操作系统的路径分隔符，将其统一替换为 '/' 后进行分割计算\n normalized_path = rel_path.replace('\\\\', '/')\n parts = [p for p in normalized_path.split('/') if p]\n depth = len(parts)\n \n # 如果嵌套深度超过 2 层，则不符合要求\n if depth > 2:\n return False\n \n # 遍历结束未发现超过2层的嵌套，说明符合要求\n return True","tags":["Quantity","Complexity"]},{"constraint":"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对 Web 版计算器应用推测可能生成的主要代码文件后缀，也可涵盖一些后端/通用后缀以防万一\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.json', '.vue'}\n has_code_file = False\n \n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 判断顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件不可读（比如存在异常编码或者无权限），认为验证未通过\n return False\n \n # 如果找到了代码文件且均通过了检查，则返回 True，否则（如没有生成任何代码文件）返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成代码文件时，路径分隔符应当优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在输出的代码文件里，不允许出现行数超过 20 行的函数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n def check_py_functions(code_text: str) -> bool:\n try:\n tree = ast.parse(code_text)\n except Exception:\n # 如果解析失败，可能是残缺的代码，跳过检查\n return True\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_lineno = node.lineno\n \n # 获取准确的结束行号（支持兼容不同版本的Python）\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n end_lineno = node.end_lineno\n else:\n end_lineno = start_lineno\n for child in ast.walk(node):\n if hasattr(child, 'lineno') and child.lineno > end_lineno:\n end_lineno = child.lineno\n \n if end_lineno - start_lineno + 1 > 20:\n return False\n return True\n\n def check_js_functions(code_text: str) -> bool:\n # 1. 轻量级状态机剔除 JS/TS 中的多行/单行注释及各种字符串字面量，防止大括号干扰\n out = []\n i = 0\n n = len(code_text)\n state = 'NORMAL'\n while i < n:\n if state == 'NORMAL':\n if code_text[i:i+2] == '//':\n state = 'LINE_COMMENT'\n out.append(' ')\n i += 2\n elif code_text[i:i+2] == '/*':\n state = 'BLOCK_COMMENT'\n out.append(' ')\n i += 2\n elif code_text[i] == '\"':\n state = 'STRING_DOUBLE'\n out.append(' ')\n i += 1\n elif code_text[i] == \"'\":\n state = 'STRING_SINGLE'\n out.append(' ')\n i += 1\n elif code_text[i] == '`':\n state = 'STRING_BACKTICK'\n out.append(' ')\n i += 1\n else:\n out.append(code_text[i])\n i += 1\n elif state == 'LINE_COMMENT':\n if code_text[i] == '\\n':\n state = 'NORMAL'\n out.append('\\n')\n i += 1\n else:\n out.append(' ')\n i += 1\n elif state == 'BLOCK_COMMENT':\n if code_text[i:i+2] == '*/':\n state = 'NORMAL'\n out.append(' ')\n i += 2\n elif code_text[i] == '\\n':\n out.append('\\n')\n i += 1\n else:\n out.append(' ')\n i += 1\n elif state == 'STRING_DOUBLE':\n if code_text[i] == '\\\\':\n out.append(' ')\n i += 2\n elif code_text[i] == '\"':\n state = 'NORMAL'\n out.append(' ')\n i += 1\n elif code_text[i] == '\\n':\n out.append('\\n')\n i += 1\n else:\n out.append(' ')\n i += 1\n elif state == 'STRING_SINGLE':\n if code_text[i] == '\\\\':\n out.append(' ')\n i += 2\n elif code_text[i] == \"'\":\n state = 'NORMAL'\n out.append(' ')\n i += 1\n elif code_text[i] == '\\n':\n out.append('\\n')\n i += 1\n else:\n out.append(' ')\n i += 1\n elif state == 'STRING_BACKTICK':\n if code_text[i] == '\\\\':\n out.append(' ')\n i += 2\n elif code_text[i] == '`':\n state = 'NORMAL'\n out.append(' ')\n i += 1\n elif code_text[i] == '\\n':\n out.append('\\n')\n i += 1\n else:\n out.append(' ')\n i += 1\n\n clean_code = \"\".join(out)\n \n # 2. 统计大括号深度来捕获函数的行数\n stack = []\n i = 0\n current_line = 1\n \n while i < len(clean_code):\n if clean_code[i] == '\\n':\n current_line += 1\n i += 1\n continue\n \n if clean_code[i] == '{':\n # 往左扫描，判断这个大括号是不是一个函数块开头\n j = i - 1\n while j >= 0 and clean_code[j] in ' \\t\\r\\n':\n j -= 1\n \n is_func = False\n if j >= 0:\n if clean_code[j] == '>':\n if j >= 1 and clean_code[j-1] == '=': # 匹配 => 箭头函数\n is_func = True\n elif clean_code[j] == ')': # 匹配 function(...) { 或 classMethod(...) {\n paren_count = 1\n k = j - 1\n while k >= 0:\n if clean_code[k] == ')':\n paren_count += 1\n elif clean_code[k] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n k -= 1\n \n # 回溯获取该关键字，排除控制流关键字如 if/for/while 触发的代码块\n if k >= 0:\n p = k - 1\n while p >= 0 and clean_code[p] in ' \\t\\r\\n':\n p -= 1\n word = \"\"\n while p >= 0 and (clean_code[p].isalpha() or clean_code[p].isdigit() or clean_code[p] in '_$'):\n word = clean_code[p] + word\n p -= 1\n if word not in ['if', 'for', 'while', 'switch', 'catch']:\n is_func = True\n \n stack.append((current_line, is_func))\n i += 1\n \n elif clean_code[i] == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n # 发现函数模块收口，判定函数的整体行数\n if current_line - start_line + 1 > 20:\n return False\n i += 1\n else:\n i += 1\n \n return True\n\n # 遍历当前项目文件夹，分析生成的所有代码文件的函数要求\n if not os.path.exists(workspace_path):\n return True\n\n for root_dir, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root_dir, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code_text = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_py_functions(code_text):\n return False\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.mjs', '.cjs']:\n if not check_js_functions(code_text):\n return False\n elif ext in ['.html', '.htm', '.vue']:\n # 寻找所有包含JS的script块\n scripts = re.findall(r'(?is)<script[^>]*>(.*?)</script>', code_text)\n for script in scripts:\n if not check_js_functions(script):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"所有代码在编写完毕后，都必须经过测试且保证测试完全通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式","validation_code":"","tags":["Workflow","Tool Usage"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我开发一个Web版的高级计算器应用。第一步，我们先搭建基本的骨架。我需要一个计算器的标准UI：顶部是一个宽大的显示屏区域，用来显示用户的输入；下面是基础的按键区，包含数字键盘（0-9）、小数点、基础四则运算符（+、-、*、/）、等号（=）以及清除按钮（C）和退格键。为了方便后续扩展，请预留出侧边栏的空间。当前你只需要实现UI排版，并让点击按钮时能够将对应字符显示在顶部的屏幕上即可，先不用实现真正的计算逻辑。另外请你后续始终遵循以下要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-，在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请使用纯中文为你生成的每一个代码文件编写注释\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n\\n workspace = Path(workspace_path).resolve()\\n \\n # 如果工作目录不存在或不是一个目录，从严格意义上并未违反该嵌套要求\\n if not workspace.exists() or not workspace.is_dir():\\n return True\\n\\n # 遍历工作目录下所有的文件与文件夹\\n for p in workspace.rglob('*'):\\n # 仅针对文件夹计算嵌套层级\\n if p.is_dir():\\n try:\\n # 计算当前文件夹相对于根目录的相对路径\\n rel_path = p.relative_to(workspace)\\n \\n # 排除 .git 等常见隐藏文件夹造成的误判干扰\\n if any(part.startswith('.') for part in rel_path.parts):\\n continue\\n \\n # rel_path.parts 会将相对路径每一层目录分割为元组，其长度即为当前文件夹的嵌套深度\\n # 例如：子文件夹 \\\"css\\\" 的长度为 1，\\\"assets/css\\\" 的长度为 2\\n if len(rel_path.parts) > 2:\\n return False\\n except ValueError:\\n # 当无法计算相对路径时直接跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_encoding_declaration(response, workspace_path):\\n import os\\n\\n # 定义Web应用及常用开发语言的代码文件后缀名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.vue', '.py', '.java', \\n '.c', '.cpp', '.h', '.hpp', '.php'\\n }\\n has_code_file = False\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含要求中的特定编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果打开文件或读取时遇到异常(如编码错误)，判定为不满足\\n return False\\n \\n # 如果项目中没有任何代码文件，也视作未完成要求；否则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中表示路径时，请尽量采用正斜杠 / 进行分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"很好，基础界面搭建得不错。现在我们来注入核心的计算能力，并扩展出科学计算功能。请在基础键盘的左侧或上方扩展一个『科学计算面板』，新增包含括号 `()`、三角函数（`sin`, `cos`, `tan`）、对数（`log`, `ln`）、`√` (平方根)、`^` (指数)、常数 `π` 和 `e` 等按钮。接下来实现完整的表达式解析引擎：要求点击等号时，能正确计算包含科学函数和多层括号嵌套的复杂数学表达式（比如 `sin(π/2) + 2^3`），并遵循正确的运算符优先级。如果用户输入的表达式有语法错误，请在屏幕上以红色字体提示“表达式错误”。另外请你后续始终遵循以下新要求：生成的代码文件中路径的分割请优先选用反斜杠 \\\\；修改任何已有文件的首要步骤，必须使用相应的工具来阅读它。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请使用纯中文为你生成的每一个代码文件编写注释\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_workspace_depth(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除以 . 开头的隐藏文件夹，防止非 AI 生成的系统或环境文件造成误判\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录自身的相对路径为 '.'，深度为 0，跳过判断\\n if rel_path == '.':\\n continue\\n \\n # 统一路径分隔符并计算嵌套深度\\n # os.path.relpath 返回的路径是规范化过的，不包含开头或结尾的多余斜杠\\n normalized_path = rel_path.replace('\\\\\\\\', '/')\\n depth = len(normalized_path.split('/'))\\n \\n # 如果嵌套深度超过 2 层，则未满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义Web计算器项目及通用开发中可能涉及的代码文件扩展名\\n code_exts = {'.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue'}\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法以 utf-8 编码读取或其他异常，也视为不满足要求\\n return False\\n\\n # 若找到了代码文件并且所有代码文件的顶行都符合要求，返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中路径的分割请优先选用反斜杠 \\\\\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"计算逻辑很完美。接下来补充前面预留的侧边栏，做成『历史记录』面板。要求：每次用户成功计算出结果后，自动将『完整的表达式 = 结果』作为一个卡片条目追加到侧边栏中。为了提升交互体验，当用户点击侧边栏里的任意一条历史记录时，能够自动将该条目的表达式填回主计算器的显示屏上，方便修改后再计算。最后，在历史记录面板底部加上一个『清空历史』按钮，点击可清空列表。另外请你后续始终遵循以下新要求：生成代码文件时，路径分隔符应当优先采用正斜杠 /；在你输出的代码文件里，不允许出现行数超过 50 行的函数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请使用纯中文为你生成的每一个代码文件编写注释\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_folder_depth(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作空间路径不存在，视作没有文件夹生成，符合不超过 2 层的要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏文件夹和Python编译产生的缓存文件夹\\n # 确保只检查由于AI生成逻辑实质创建的源码目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n rel_path = os.path.relpath(root, workspace_path)\\n \\n if rel_path == '.':\\n depth = 0\\n else:\\n # 统一路径分隔符，以防跨平台或反斜杠带来的计算错误\\n rel_path = rel_path.replace('\\\\\\\\', '/')\\n depth = len(rel_path.split('/'))\\n \\n # 如果嵌套深度超过 2 层，则不满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户需求推测可能的 Web 版计算器代码文件后缀\\n code_extensions = {'.py', '.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n found_code_file = False\\n \\n # 遍历当前项目文件夹\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果是代码文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 读取文件顶行内容，忽略可能的非标准编码异常\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若文件无法读取也视为不符合要求\\n return False\\n \\n # 如果找到了至少一个代码文件，且全部验证通过，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你输出的代码文件里，不允许出现行数超过 50 行的函数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n\\n def check_file(filepath):\\n # 1. 判断并验证 Python 文件\\n if filepath.endswith('.py'):\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n code = f.read()\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\\n if node.end_lineno - node.lineno + 1 > 50:\\n return False\\n except Exception:\\n pass # 若语法解析失败可跳过\\n return True\\n \\n # 2. 判断并验证 JS/TS/HTML 前端相关文件\\n elif filepath.endswith(('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 如果是包含 HTML/模板 的文件，仅提取并检验脚本块内的函数\\n if filepath.endswith(('.html', '.htm', '.vue')):\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\\n content = \\\"\\\\n\\\".join(scripts)\\n \\n if not content.strip():\\n return True\\n \\n # 用于匹配 JS/TS 函数开始的正则表达式集合\\n js_func_patterns = [\\n r'\\\\bfunction\\\\s*[\\\\w$]*\\\\s*\\\\([^){}]*\\\\)\\\\s*\\\\{', # function 声明\\n r'(?:\\\\([^){}]*\\\\)|[\\\\w$]+)\\\\s*=>\\\\s*\\\\{', # 箭头函数 (() => { 或 arg => {)\\n r'(?<!@)\\\\b([\\\\w$]+)\\\\s*\\\\([^){}]*\\\\)\\\\s*\\\\{' # 类/对象方法 (例如 calc() { )\\n ]\\n \\n # 保留字列表，防止条件语句和保留关键字被误认为对象方法函数\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with', 'else', 'return', 'class'}\\n \\n for pattern in js_func_patterns:\\n for match in re.finditer(pattern, content):\\n # 如果匹配到保留字结构（例如 if() { ）则跳过\\n if len(match.groups()) > 0 and match.group(1) in keywords:\\n continue\\n \\n start_idx = match.end() - 1\\n if content[start_idx] != '{':\\n start_idx = content.rfind('{', match.start(), match.end())\\n \\n if start_idx == -1:\\n continue\\n\\n # 将当前字符串的片段映射到行号\\n start_lineno = content.count('\\\\n', 0, start_idx) + 1\\n brace_count = 0\\n end_idx = start_idx\\n \\n # 使用简单的状态机规避字符串或注释内部的大括号干扰\\n in_string = False\\n string_char = ''\\n in_single_comment = False\\n in_multi_comment = False\\n \\n i = start_idx\\n while i < len(content):\\n char = content[i]\\n if in_single_comment:\\n if char == '\\\\n':\\n in_single_comment = False\\n elif in_multi_comment:\\n if char == '*' and i + 1 < len(content) and content[i + 1] == '/':\\n in_multi_comment = False\\n i += 1\\n elif in_string:\\n if char == '\\\\\\\\':\\n i += 1\\n elif char == string_char:\\n in_string = False\\n else:\\n if char == '/' and i + 1 < len(content) and content[i + 1] == '/':\\n in_single_comment = True\\n i += 1\\n elif char == '/' and i + 1 < len(content) and content[i + 1] == '*':\\n in_multi_comment = True\\n i += 1\\n elif char in ['\\\\'', '\\\"', '`']:\\n in_string = True\\n string_char = char\\n elif char == '{':\\n brace_count += 1\\n elif char == '}':\\n brace_count -= 1\\n # 函数体闭合\\n if brace_count == 0:\\n end_idx = i\\n break\\n i += 1\\n \\n # 判断所寻找到的函数体的跨越行数\\n if brace_count == 0:\\n end_lineno = content.count('\\\\n', 0, end_idx) + 1\\n if end_lineno - start_lineno + 1 > 50:\\n return False\\n return True\\n return True\\n\\n # 遍历当前工作空间内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n # 只要有任何一个代码文件不满足要求，直接判定失败\\n if not check_file(filepath):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"好，现在我们要加入最具挑战性的功能：函数绘图。请在整个应用的最顶部增加一个Tab切换器，允许用户在『科学计算器』和『函数绘图』两个视图之间无缝切换。在『函数绘图』视图下，界面分为左右两部分：左侧提供一个输入框供用户输入关于自变量 x 的函数表达式（例如 `sin(x) + x^2`），并带有一个『绘制』按钮；右侧提供一个大型的 Canvas 画布，默认绘制出带有 X 轴、Y 轴和网格线的二维坐标系（默认范围 -10 到 10）。当用户点击绘制按钮时，在坐标系上画出该函数的连续曲线。另外请你后续始终遵循以下新要求：在你输出的代码文件里，不允许出现行数超过 30 行的函数；所有代码在编写完毕后，都必须经过测试且保证测试完全通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请使用纯中文为你生成的每一个代码文件编写注释\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断代码项目文件夹嵌套深度是否不超过2层。\\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 如果嵌套不超过2层返回 True，否则返回 False。\\n \\\"\\\"\\\"\\n if not os.path.exists(workspace_path):\\n return False\\n \\n workspace = Path(workspace_path).resolve()\\n \\n # 遍历 workspace_path 下的所有文件和文件夹\\n for dir_path in workspace.rglob('*'):\\n if dir_path.is_dir():\\n try:\\n # 获取相对路径\\n rel_path = dir_path.relative_to(workspace)\\n \\n # 排除隐藏目录（如 .git, .vscode 等）和 Python 缓存目录（__pycache__）\\n if any(part.startswith('.') or part == '__pycache__' for part in rel_path.parts):\\n continue\\n \\n # 判断嵌套深度（相对路径的层级数即为嵌套层数）\\n # 例如：dir1 层级数为 1；dir1/dir2 层级数为 2；dir1/dir2/dir3 层级数为 3\\n if len(rel_path.parts) > 2:\\n return False\\n \\n except ValueError:\\n # 如果由于某种原因无法计算相对路径则忽略\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的代码文件类型，适用于 Web 应用开发\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的依赖和版本控制文件夹\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取，视为未满足要求\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你输出的代码文件里，不允许出现行数超过 30 行的函数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def replacer(match):\\n s = match.group(0)\\n # 将匹配的字符串和注释替换成相同数量的换行和空格，从而保证行数计算不受影响且去除干扰字符\\n return ''.join('\\\\n' if c == '\\\\n' else ' ' for c in s)\\n \\n def check_js_functions(source_code: str) -> bool:\\n # 去除字符串字面量和注释的干扰，但保留其中的换行符\\n pattern = re.compile(\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\"|'\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\\\\\n])*'|\\\"\\n r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`|\\\"\\n r\\\"/\\\\*[\\\\s\\\\S]*?\\\\*/|\\\"\\n r\\\"//.*\\\"\\n )\\n try:\\n masked_code = pattern.sub(replacer, source_code)\\n except Exception:\\n masked_code = source_code\\n \\n def check_function_body(start_lineno: int, found_brace_idx: int) -> bool:\\n brace_count = 0\\n for i in range(found_brace_idx, len(masked_code)):\\n if masked_code[i] == '{':\\n brace_count += 1\\n elif masked_code[i] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_lineno = masked_code.count('\\\\n', 0, i) + 1\\n # 计算函数的总行数，如果大于30则说明违反要求\\n if end_lineno - start_lineno + 1 > 30:\\n return False\\n break\\n return True\\n\\n # 1. 检查普通函数定义：function name(...) { 或 function(...) {\\n for match in re.finditer(r'\\\\bfunction\\\\b', masked_code):\\n start_lineno = masked_code.count('\\\\n', 0, match.start()) + 1\\n idx = match.end()\\n paren_level = 0\\n found_brace_idx = -1\\n for i in range(idx, len(masked_code)):\\n if masked_code[i] == '(':\\n paren_level += 1\\n elif masked_code[i] == ')':\\n paren_level -= 1\\n if paren_level < 0:\\n break\\n elif masked_code[i] == '{' and paren_level == 0:\\n found_brace_idx = i\\n break\\n elif masked_code[i] == ';' and paren_level == 0:\\n break\\n \\n if found_brace_idx != -1:\\n if not check_function_body(start_lineno, found_brace_idx):\\n return False\\n\\n # 2. 检查箭头函数：(...) => { 或 arg => {\\n for match in re.finditer(r'=>', masked_code):\\n start_lineno = masked_code.count('\\\\n', 0, match.start()) + 1\\n idx = match.end()\\n paren_level = 0\\n found_brace_idx = -1\\n for i in range(idx, len(masked_code)):\\n if masked_code[i] == '(':\\n paren_level += 1\\n elif masked_code[i] == ')':\\n paren_level -= 1\\n if paren_level < 0:\\n break\\n elif masked_code[i] == '{' and paren_level == 0:\\n found_brace_idx = i\\n break\\n elif masked_code[i] == ';' and paren_level == 0:\\n break\\n elif paren_level == 0 and masked_code[i] not in ' \\\\t\\\\r\\\\n':\\n # 若不是括号或大括号，并且遇到了其他操作符，说明是缺少代码块的单行表达式\\n break\\n \\n if found_brace_idx != -1:\\n if not check_function_body(start_lineno, found_brace_idx):\\n return False\\n\\n # 3. 检查类方法或简写对象方法：methodName(...) {\\n for match in re.finditer(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\(', masked_code):\\n name = match.group(1)\\n # 排除控制流关键字和特殊指令\\n if name in {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'yield', 'await'}:\\n continue\\n start_lineno = masked_code.count('\\\\n', 0, match.start()) + 1\\n idx = match.end() - 1 # 游标停在 '(' 的位置\\n paren_level = 0\\n found_brace_idx = -1\\n for i in range(idx, len(masked_code)):\\n if masked_code[i] == '(':\\n paren_level += 1\\n elif masked_code[i] == ')':\\n paren_level -= 1\\n if paren_level < 0:\\n break\\n elif masked_code[i] == '{' and paren_level == 0:\\n found_brace_idx = i\\n break\\n elif masked_code[i] == ';' and paren_level == 0:\\n break\\n elif paren_level == 0 and masked_code[i] not in ' \\\\t\\\\r\\\\n':\\n # 如果不是空格且不在参数列表内，并且也不是大括号，说明这只是普通的函数调用而非函数定义\\n break\\n \\n if found_brace_idx != -1:\\n if not check_function_body(start_lineno, found_brace_idx):\\n return False\\n \\n return True\\n\\n def check_python_functions(source_code: str) -> bool:\\n try:\\n tree = ast.parse(source_code)\\n except Exception:\\n return True\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n if node.end_lineno is not None and node.lineno is not None:\\n if node.end_lineno - node.lineno + 1 > 30:\\n return False\\n return True\\n\\n def check_html_functions(source_code: str) -> bool:\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', source_code, re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n if not check_js_functions(script):\\n return False\\n return True\\n\\n # 递归遍历项目目录中的所有相关文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_python_functions(content):\\n return False\\n elif ext in {'.js', '.ts', '.jsx', '.tsx'}:\\n if not check_js_functions(content):\\n return False\\n elif ext in {'.html', '.htm', '.vue'}:\\n if not check_html_functions(content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"所有代码在编写完毕后，都必须经过测试且保证测试完全通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，让我们把绘图视图的交互体验拉满。首先，支持多函数绘制：在左侧增加一个『添加函数』按钮，允许动态添加多个输入框（f(x), g(x), h(x)...），且右侧画出的每条曲线必须自动分配不同的颜色加以区分。其次，增强画布的交互能力：支持鼠标在画布上拖拽来平移坐标系，支持通过鼠标滚轮进行坐标系的放大和缩小，缩放和平移时网格线、坐标轴刻度和函数曲线必须实时重新渲染。最后，当鼠标悬浮在画布上移动时，绘制一个跟随鼠标的十字准星，并实时显示当前指针位置对应的精确 (x, y) 坐标数值。另外请你后续始终遵循以下新要求：在输出的代码文件里，不允许出现行数超过 20 行的函数；执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"务必尽可能将多个独立的工具调用并发处理，这对于提升执行效率是必需的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"请使用纯中文为你生成的每一个代码文件编写注释\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"确保你输出的代码项目文件夹嵌套深度不超过 2 层\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作目录不存在，无法进行判断，默认返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 列表，过滤掉隐藏文件夹(以 . 开头)以及常见的系统/依赖生成的文件夹\\n # 这样 os.walk 就不会继续向深处遍历这些无关文件夹\\n dirs[:] = [\\n d for d in dirs \\n if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')\\n ]\\n \\n # 获取当前目录相对于 workspace_path 的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录的相对路径为 '.'，深度记为 0\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 兼容不同操作系统的路径分隔符，将其统一替换为 '/' 后进行分割计算\\n normalized_path = rel_path.replace('\\\\\\\\', '/')\\n parts = [p for p in normalized_path.split('/') if p]\\n depth = len(parts)\\n \\n # 如果嵌套深度超过 2 层，则不符合要求\\n if depth > 2:\\n return False\\n \\n # 遍历结束未发现超过2层的嵌套，说明符合要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件里，所有循环嵌套的层级都不超过 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对 Web 版计算器应用推测可能生成的主要代码文件后缀，也可涵盖一些后端/通用后缀以防万一\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.json', '.vue'}\\n has_code_file = False\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 判断顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件不可读（比如存在异常编码或者无权限），认为验证未通过\\n return False\\n \\n # 如果找到了代码文件且均通过了检查，则返回 True，否则（如没有生成任何代码文件）返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，路径分隔符应当优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在输出的代码文件里，不允许出现行数超过 20 行的函数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n def check_py_functions(code_text: str) -> bool:\\n try:\\n tree = ast.parse(code_text)\\n except Exception:\\n # 如果解析失败，可能是残缺的代码，跳过检查\\n return True\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_lineno = node.lineno\\n \\n # 获取准确的结束行号（支持兼容不同版本的Python）\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n end_lineno = node.end_lineno\\n else:\\n end_lineno = start_lineno\\n for child in ast.walk(node):\\n if hasattr(child, 'lineno') and child.lineno > end_lineno:\\n end_lineno = child.lineno\\n \\n if end_lineno - start_lineno + 1 > 20:\\n return False\\n return True\\n\\n def check_js_functions(code_text: str) -> bool:\\n # 1. 轻量级状态机剔除 JS/TS 中的多行/单行注释及各种字符串字面量，防止大括号干扰\\n out = []\\n i = 0\\n n = len(code_text)\\n state = 'NORMAL'\\n while i < n:\\n if state == 'NORMAL':\\n if code_text[i:i+2] == '//':\\n state = 'LINE_COMMENT'\\n out.append(' ')\\n i += 2\\n elif code_text[i:i+2] == '/*':\\n state = 'BLOCK_COMMENT'\\n out.append(' ')\\n i += 2\\n elif code_text[i] == '\\\"':\\n state = 'STRING_DOUBLE'\\n out.append(' ')\\n i += 1\\n elif code_text[i] == \\\"'\\\":\\n state = 'STRING_SINGLE'\\n out.append(' ')\\n i += 1\\n elif code_text[i] == '`':\\n state = 'STRING_BACKTICK'\\n out.append(' ')\\n i += 1\\n else:\\n out.append(code_text[i])\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code_text[i] == '\\\\n':\\n state = 'NORMAL'\\n out.append('\\\\n')\\n i += 1\\n else:\\n out.append(' ')\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if code_text[i:i+2] == '*/':\\n state = 'NORMAL'\\n out.append(' ')\\n i += 2\\n elif code_text[i] == '\\\\n':\\n out.append('\\\\n')\\n i += 1\\n else:\\n out.append(' ')\\n i += 1\\n elif state == 'STRING_DOUBLE':\\n if code_text[i] == '\\\\\\\\':\\n out.append(' ')\\n i += 2\\n elif code_text[i] == '\\\"':\\n state = 'NORMAL'\\n out.append(' ')\\n i += 1\\n elif code_text[i] == '\\\\n':\\n out.append('\\\\n')\\n i += 1\\n else:\\n out.append(' ')\\n i += 1\\n elif state == 'STRING_SINGLE':\\n if code_text[i] == '\\\\\\\\':\\n out.append(' ')\\n i += 2\\n elif code_text[i] == \\\"'\\\":\\n state = 'NORMAL'\\n out.append(' ')\\n i += 1\\n elif code_text[i] == '\\\\n':\\n out.append('\\\\n')\\n i += 1\\n else:\\n out.append(' ')\\n i += 1\\n elif state == 'STRING_BACKTICK':\\n if code_text[i] == '\\\\\\\\':\\n out.append(' ')\\n i += 2\\n elif code_text[i] == '`':\\n state = 'NORMAL'\\n out.append(' ')\\n i += 1\\n elif code_text[i] == '\\\\n':\\n out.append('\\\\n')\\n i += 1\\n else:\\n out.append(' ')\\n i += 1\\n\\n clean_code = \\\"\\\".join(out)\\n \\n # 2. 统计大括号深度来捕获函数的行数\\n stack = []\\n i = 0\\n current_line = 1\\n \\n while i < len(clean_code):\\n if clean_code[i] == '\\\\n':\\n current_line += 1\\n i += 1\\n continue\\n \\n if clean_code[i] == '{':\\n # 往左扫描，判断这个大括号是不是一个函数块开头\\n j = i - 1\\n while j >= 0 and clean_code[j] in ' \\\\t\\\\r\\\\n':\\n j -= 1\\n \\n is_func = False\\n if j >= 0:\\n if clean_code[j] == '>':\\n if j >= 1 and clean_code[j-1] == '=': # 匹配 => 箭头函数\\n is_func = True\\n elif clean_code[j] == ')': # 匹配 function(...) { 或 classMethod(...) {\\n paren_count = 1\\n k = j - 1\\n while k >= 0:\\n if clean_code[k] == ')':\\n paren_count += 1\\n elif clean_code[k] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n k -= 1\\n \\n # 回溯获取该关键字，排除控制流关键字如 if/for/while 触发的代码块\\n if k >= 0:\\n p = k - 1\\n while p >= 0 and clean_code[p] in ' \\\\t\\\\r\\\\n':\\n p -= 1\\n word = \\\"\\\"\\n while p >= 0 and (clean_code[p].isalpha() or clean_code[p].isdigit() or clean_code[p] in '_$'):\\n word = clean_code[p] + word\\n p -= 1\\n if word not in ['if', 'for', 'while', 'switch', 'catch']:\\n is_func = True\\n \\n stack.append((current_line, is_func))\\n i += 1\\n \\n elif clean_code[i] == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n # 发现函数模块收口，判定函数的整体行数\\n if current_line - start_line + 1 > 20:\\n return False\\n i += 1\\n else:\\n i += 1\\n \\n return True\\n\\n # 遍历当前项目文件夹，分析生成的所有代码文件的函数要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root_dir, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root_dir, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code_text = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_py_functions(code_text):\\n return False\\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.mjs', '.cjs']:\\n if not check_js_functions(code_text):\\n return False\\n elif ext in ['.html', '.htm', '.vue']:\\n # 寻找所有包含JS的script块\\n scripts = re.findall(r'(?is)<script[^>]*>(.*?)</script>', code_text)\\n for script in scripts:\\n if not check_js_functions(script):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"所有代码在编写完毕后，都必须经过测试且保证测试完全通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["观察计算器的初始界面截图，确认顶部存在宽大的显示屏区域，下方存在基础键盘（包含数字0-9、小数点、加减乘除、等号、C键和退格键）以及科学计算面板（包含括号、sin、cos、tan、log、ln、√、^、π、e等按钮），并预留了侧边栏空间。","依次点击数字“1”、操作符“+”、科学常数“π”和科学函数“sin”，截取显示屏图片，确认屏幕上正确将这些按键代表的字符顺序显示出来。","在显示屏有多个字符的情况下，点击“退格”键一次，截图确认显示屏上最后一个字符被删掉；随后点击“C”键，截图确认显示屏的内容被完全清空。","输入带有优先级的复杂科学计算表达式（例如输入 `sin(π/2) + 2^3` ），点击“=”号按钮，截图确认显示屏的字符变更为正确的计算结果数值。","输入带有明显语法错误的数学表达式（例如缺少右操作数的 `5*` 或未闭合的括号 `sin(π/2`），点击“=”号按钮，截图确认显示屏中以红色的字体显示“表达式错误”的提示信息。","成功执行一次任意计算操作并得到结果后，查看界面的侧边栏区域截图，确认侧边栏（历史记录面板）自动追加了一条新卡片，其内容格式表现为“完整的表达式 = 结果”。","用鼠标点击侧边栏内已存在的一条历史记录卡片，截图确认主显示屏中的内容自动变更为该条历史记录所对应的表达式（方便用户进行二次修改）。","点击历史记录面板底部的“清空历史”按钮，查看侧边栏截图，确认原本存在的历史记录列表已经被完全清空。","点击应用顶部的Tab切换器中的“函数绘图”选项，截图确认界面成功切换至绘图视图：左侧显示函数输入框和“绘制”按钮，右侧显示一个占据大面积的Canvas画布，画布中默认渲染带有X轴、Y轴和网格线的二维坐标系（且默认范围为-10到10）。","点击Tab切换器中的“科学计算器”选项，截图确认界面成功切换回科学计算器视图，所有原有功能（显示屏、按键、历史记录）保持正常可用。","在“函数绘图”视图左侧输入框中输入合法的自变量x表达式（例如 `sin(x) + x^2`），点击“绘制”按钮，截图确认右侧画布坐标系中正确画出了该函数对应的连续曲线。","点击函数绘图视图左侧的“添加函数”按钮，截图确认左侧区域成功动态新增了一个额外的函数输入框，且带有如f(x)、g(x)、h(x)等区分标识。","在两个以上的函数输入框中填入不同的表达式，点击“绘制”，查看画布截图确认画布上同时存在多条函数曲线，且每条曲线均被自动分配了不同的颜色。","模拟按住鼠标左键在画布上进行拖拽操作并松开，对比拖拽前后的画布截图，确认整个坐标系（含网格线、X轴Y轴刻度和函数曲线）跟随着鼠标的拖拽方向发生了平移，且网格线、坐标轴刻度和已绘制的函数曲线实时重新渲染。","模拟鼠标在画布区域滚动滚轮，对比滚动前后的画布截图，确认坐标轴上的刻度数值发生相应的变大或变小（视图放大/缩小），且网格线、坐标轴刻度和已绘制的函数曲线实时重新渲染以适配新的坐标系尺寸。","模拟鼠标光标悬停并在画布区域内移动，截图确认存在一个十字准星形状始终跟随光标所在位置，同时界面上实时显示出当前指针指向的坐标系内的精确 (x, y) 坐标数值。"],"string":"[\n \"观察计算器的初始界面截图，确认顶部存在宽大的显示屏区域，下方存在基础键盘（包含数字0-9、小数点、加减乘除、等号、C键和退格键）以及科学计算面板（包含括号、sin、cos、tan、log、ln、√、^、π、e等按钮），并预留了侧边栏空间。\",\n \"依次点击数字“1”、操作符“+”、科学常数“π”和科学函数“sin”，截取显示屏图片，确认屏幕上正确将这些按键代表的字符顺序显示出来。\",\n \"在显示屏有多个字符的情况下，点击“退格”键一次，截图确认显示屏上最后一个字符被删掉；随后点击“C”键，截图确认显示屏的内容被完全清空。\",\n \"输入带有优先级的复杂科学计算表达式（例如输入 `sin(π/2) + 2^3` ），点击“=”号按钮，截图确认显示屏的字符变更为正确的计算结果数值。\",\n \"输入带有明显语法错误的数学表达式（例如缺少右操作数的 `5*` 或未闭合的括号 `sin(π/2`），点击“=”号按钮，截图确认显示屏中以红色的字体显示“表达式错误”的提示信息。\",\n \"成功执行一次任意计算操作并得到结果后，查看界面的侧边栏区域截图，确认侧边栏（历史记录面板）自动追加了一条新卡片，其内容格式表现为“完整的表达式 = 结果”。\",\n \"用鼠标点击侧边栏内已存在的一条历史记录卡片，截图确认主显示屏中的内容自动变更为该条历史记录所对应的表达式（方便用户进行二次修改）。\",\n \"点击历史记录面板底部的“清空历史”按钮，查看侧边栏截图，确认原本存在的历史记录列表已经被完全清空。\",\n \"点击应用顶部的Tab切换器中的“函数绘图”选项，截图确认界面成功切换至绘图视图：左侧显示函数输入框和“绘制”按钮，右侧显示一个占据大面积的Canvas画布，画布中默认渲染带有X轴、Y轴和网格线的二维坐标系（且默认范围为-10到10）。\",\n \"点击Tab切换器中的“科学计算器”选项，截图确认界面成功切换回科学计算器视图，所有原有功能（显示屏、按键、历史记录）保持正常可用。\",\n \"在“函数绘图”视图左侧输入框中输入合法的自变量x表达式（例如 `sin(x) + x^2`），点击“绘制”按钮，截图确认右侧画布坐标系中正确画出了该函数对应的连续曲线。\",\n \"点击函数绘图视图左侧的“添加函数”按钮，截图确认左侧区域成功动态新增了一个额外的函数输入框，且带有如f(x)、g(x)、h(x)等区分标识。\",\n \"在两个以上的函数输入框中填入不同的表达式，点击“绘制”，查看画布截图确认画布上同时存在多条函数曲线，且每条曲线均被自动分配了不同的颜色。\",\n \"模拟按住鼠标左键在画布上进行拖拽操作并松开，对比拖拽前后的画布截图，确认整个坐标系（含网格线、X轴Y轴刻度和函数曲线）跟随着鼠标的拖拽方向发生了平移，且网格线、坐标轴刻度和已绘制的函数曲线实时重新渲染。\",\n \"模拟鼠标在画布区域滚动滚轮，对比滚动前后的画布截图，确认坐标轴上的刻度数值发生相应的变大或变小（视图放大/缩小），且网格线、坐标轴刻度和已绘制的函数曲线实时重新渲染以适配新的坐标系尺寸。\",\n \"模拟鼠标光标悬停并在画布区域内移动，截图确认存在一个十字准星形状始终跟随光标所在位置，同时界面上实时显示出当前指针指向的坐标系内的精确 (x, y) 坐标数值。\"\n]","isConversational":false}}},{"rowIdx":67,"cells":{"task_id":{"kind":"number","value":68,"string":"68"},"repository_policy":{"kind":"string","value":"你在后续所有交互轮次中都必须严格遵循以下规则，不得违反：\n1. 你必须先使用相应的工具阅读已有文件，然后才能对其进行修改。\n2. 请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数。\n3. 你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行。\n4. 生成的代码文件中不得使用任何形式的三元运算符。\n5. 你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数。\n6. 你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义各类主流编程语言及标记语言的代码文件扩展名分类\n hash_ext = {'.py', '.rb', '.sh', '.pl', '.yml', '.yaml'}\n slash_ext = {'.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cc', \n '.h', '.hpp', '.cs', '.go', '.rs', '.swift', '.kt', '.php', '.css'}\n html_ext = {'.html', '.htm', '.xml', '.vue'}\n \n code_extensions = hash_ext | slash_ext | html_ext\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理被认定为代码类型的文件，避开 .log 或 .txt 等不属于代码的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于编码或其他原因无法读取文件内容，跳过该文件\n continue\n \n has_code_file = True\n \n comment_lines = 0\n in_c_block = False\n in_html_block = False\n in_py_block = False\n py_block_marker = \"\"\n \n for line in lines:\n s = line.strip()\n \n # 1. 优先处理正处于多行/块注释状态中的情况\n if in_c_block:\n comment_lines += 1\n if '*/' in s:\n in_c_block = False\n continue\n \n if in_html_block:\n comment_lines += 1\n if '-->' in s:\n in_html_block = False\n continue\n \n if in_py_block:\n comment_lines += 1\n if py_block_marker in s:\n in_py_block = False\n continue\n \n # 2. 不在块注释中时，空行不计入注释行，直接跳过\n if not s:\n continue\n \n # 3. 处理各类语言单行或块注释的开头标记\n if ext in slash_ext or ext in html_ext:\n if s.startswith('//'):\n comment_lines += 1\n continue\n if s.startswith('/*'):\n comment_lines += 1\n # 如果此行未包含结束符号，则标记进入多行注释状态\n if '*/' not in s[2:]:\n in_c_block = True\n continue\n \n if ext in html_ext:\n if s.startswith('<!--'):\n comment_lines += 1\n if '-->' not in s[4:]:\n in_html_block = True\n continue\n \n if ext in hash_ext:\n if s.startswith('#'):\n comment_lines += 1\n continue\n # Python 支持以成对的三引号作为多行注释或多行 Docstring\n if ext == '.py':\n if s.startswith('\"\"\"'):\n comment_lines += 1\n if s[3:].find('\"\"\"') == -1:\n in_py_block = True\n py_block_marker = '\"\"\"'\n continue\n if s.startswith(\"'''\"):\n comment_lines += 1\n if s[3:].find(\"'''\") == -1:\n in_py_block = True\n py_block_marker = \"'''\"\n continue\n\n # 每个代码文件在统计完成后的注释行数必须精确为 3 行\n if comment_lines != 3:\n return False\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n # 常见的主流代码文件后缀名\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cs',\n '.h', '.hpp', '.php', '.go', '.rb', '.rs', '.swift', '.kt', '.dart'\n }\n \n def remove_comments_and_strings(code: str) -> str:\n \"\"\"移除代码中的多行字符串、单行字符串以及单/多行注释以避免误判\"\"\"\n # 去除多行字符串 (如 Python 中的 ''' 或 \"\"\")\n code = re.sub(r'\"\"\"(.*?)\"\"\"', '', code, flags=re.DOTALL)\n code = re.sub(r\"'''(.*?)'''\", '', code, flags=re.DOTALL)\n # 去除双引号和单引号字符串\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", code)\n # 去除多行注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 去除单行注释\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'#.*', '', code)\n return code\n\n # 遍历当前项目工作目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_str = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n # Python 文件的三元表达式验证\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n return False # 发现了三元运算符，要求未满足\n except SyntaxError:\n # AST 解析失败时，退化处理粗略匹配同行内的 if-else\n clean_code = remove_comments_and_strings(code_str)\n for line in clean_code.split('\\n'):\n if 'if ' in line and ' else ' in line and not line.strip().startswith('if'):\n return False\n else:\n # 其他主流语言文件的三元表达式验证 (condition ? A : B)\n clean_code = remove_comments_and_strings(code_str)\n # 压缩成单行以匹配带有换行的格式排版\n flat_code = re.sub(r'\\s+', ' ', clean_code)\n \n # 正则匹配 C-Style 三元运算符\n # (?<!\\?) : 负向后顾，排除前面是 ? 的情况 (防 ?? )\n # \\? : 匹配三元运算符的 ?\n # (?!\\.|:|\\?) : 负向前瞻，排除后面紧接 . , : , ? 的情况 (防 ?. , ?: , ?? )\n # [^;{}]* : 中间不能跨越代码语句终结符、块修饰符，减少误判\n # : : 匹配三元运算符的 :\n pattern = r'(?<!\\?)\\?(?!\\.|:|\\?)[^;{}]*:'\n if re.search(pattern, flat_code):\n return False # 发现了三元运算符，要求未满足\n \n # 扫描完所有代码文件均未发现三元运算符，满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n \"\"\"辅助函数：判断给定的数字是否为质数\"\"\"\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 常见代码文件后缀的枚举\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cs', '.h', '.hpp',\n '.js', '.ts', '.html', '.css', '.go', '.rs', '.rb',\n '.php', '.swift', '.kt', '.sh', '.scala', '.m'\n }\n\n found_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制等隐藏目录\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n if not content:\n lines_count = 0\n else:\n # 通过拆分换行符统计总行数，这样能准确地将空行计算在内\n lines_count = len(content.split('\\n'))\n \n if not is_prime(lines_count):\n return False\n except Exception:\n # 如果文件读取过程发生异常，视为不满足\n return False\n \n # 若存在代码文件且全部符合质数行要求，则返回 True\n return found_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]}],"string":"[\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义各类主流编程语言及标记语言的代码文件扩展名分类\\n hash_ext = {'.py', '.rb', '.sh', '.pl', '.yml', '.yaml'}\\n slash_ext = {'.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cc', \\n '.h', '.hpp', '.cs', '.go', '.rs', '.swift', '.kt', '.php', '.css'}\\n html_ext = {'.html', '.htm', '.xml', '.vue'}\\n \\n code_extensions = hash_ext | slash_ext | html_ext\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理被认定为代码类型的文件，避开 .log 或 .txt 等不属于代码的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于编码或其他原因无法读取文件内容，跳过该文件\\n continue\\n \\n has_code_file = True\\n \\n comment_lines = 0\\n in_c_block = False\\n in_html_block = False\\n in_py_block = False\\n py_block_marker = \\\"\\\"\\n \\n for line in lines:\\n s = line.strip()\\n \\n # 1. 优先处理正处于多行/块注释状态中的情况\\n if in_c_block:\\n comment_lines += 1\\n if '*/' in s:\\n in_c_block = False\\n continue\\n \\n if in_html_block:\\n comment_lines += 1\\n if '-->' in s:\\n in_html_block = False\\n continue\\n \\n if in_py_block:\\n comment_lines += 1\\n if py_block_marker in s:\\n in_py_block = False\\n continue\\n \\n # 2. 不在块注释中时，空行不计入注释行，直接跳过\\n if not s:\\n continue\\n \\n # 3. 处理各类语言单行或块注释的开头标记\\n if ext in slash_ext or ext in html_ext:\\n if s.startswith('//'):\\n comment_lines += 1\\n continue\\n if s.startswith('/*'):\\n comment_lines += 1\\n # 如果此行未包含结束符号，则标记进入多行注释状态\\n if '*/' not in s[2:]:\\n in_c_block = True\\n continue\\n \\n if ext in html_ext:\\n if s.startswith('<!--'):\\n comment_lines += 1\\n if '-->' not in s[4:]:\\n in_html_block = True\\n continue\\n \\n if ext in hash_ext:\\n if s.startswith('#'):\\n comment_lines += 1\\n continue\\n # Python 支持以成对的三引号作为多行注释或多行 Docstring\\n if ext == '.py':\\n if s.startswith('\\\"\\\"\\\"'):\\n comment_lines += 1\\n if s[3:].find('\\\"\\\"\\\"') == -1:\\n in_py_block = True\\n py_block_marker = '\\\"\\\"\\\"'\\n continue\\n if s.startswith(\\\"'''\\\"):\\n comment_lines += 1\\n if s[3:].find(\\\"'''\\\") == -1:\\n in_py_block = True\\n py_block_marker = \\\"'''\\\"\\n continue\\n\\n # 每个代码文件在统计完成后的注释行数必须精确为 3 行\\n if comment_lines != 3:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n # 常见的主流代码文件后缀名\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cs',\\n '.h', '.hpp', '.php', '.go', '.rb', '.rs', '.swift', '.kt', '.dart'\\n }\\n \\n def remove_comments_and_strings(code: str) -> str:\\n \\\"\\\"\\\"移除代码中的多行字符串、单行字符串以及单/多行注释以避免误判\\\"\\\"\\\"\\n # 去除多行字符串 (如 Python 中的 ''' 或 \\\"\\\"\\\")\\n code = re.sub(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', '', code, flags=re.DOTALL)\\n code = re.sub(r\\\"'''(.*?)'''\\\", '', code, flags=re.DOTALL)\\n # 去除双引号和单引号字符串\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", code)\\n # 去除多行注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 去除单行注释\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'#.*', '', code)\\n return code\\n\\n # 遍历当前项目工作目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n # Python 文件的三元表达式验证\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n return False # 发现了三元运算符，要求未满足\\n except SyntaxError:\\n # AST 解析失败时，退化处理粗略匹配同行内的 if-else\\n clean_code = remove_comments_and_strings(code_str)\\n for line in clean_code.split('\\\\n'):\\n if 'if ' in line and ' else ' in line and not line.strip().startswith('if'):\\n return False\\n else:\\n # 其他主流语言文件的三元表达式验证 (condition ? A : B)\\n clean_code = remove_comments_and_strings(code_str)\\n # 压缩成单行以匹配带有换行的格式排版\\n flat_code = re.sub(r'\\\\s+', ' ', clean_code)\\n \\n # 正则匹配 C-Style 三元运算符\\n # (?<!\\\\?) : 负向后顾，排除前面是 ? 的情况 (防 ?? )\\n # \\\\? : 匹配三元运算符的 ?\\n # (?!\\\\.|:|\\\\?) : 负向前瞻，排除后面紧接 . , : , ? 的情况 (防 ?. , ?: , ?? )\\n # [^;{}]* : 中间不能跨越代码语句终结符、块修饰符，减少误判\\n # : : 匹配三元运算符的 :\\n pattern = r'(?<!\\\\?)\\\\?(?!\\\\.|:|\\\\?)[^;{}]*:'\\n if re.search(pattern, flat_code):\\n return False # 发现了三元运算符，要求未满足\\n \\n # 扫描完所有代码文件均未发现三元运算符，满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"辅助函数：判断给定的数字是否为质数\\\"\\\"\\\"\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 常见代码文件后缀的枚举\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cs', '.h', '.hpp',\\n '.js', '.ts', '.html', '.css', '.go', '.rs', '.rb',\\n '.php', '.swift', '.kt', '.sh', '.scala', '.m'\\n }\\n\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制等隐藏目录\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n if not content:\\n lines_count = 0\\n else:\\n # 通过拆分换行符统计总行数，这样能准确地将空行计算在内\\n lines_count = len(content.split('\\\\n'))\\n \\n if not is_prime(lines_count):\\n return False\\n except Exception:\\n # 如果文件读取过程发生异常，视为不满足\\n return False\\n \\n # 若存在代码文件且全部符合质数行要求，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用React搭建一个“开销跟踪应用”的基础页面骨架。需要一个带应用名称的顶部导航栏，主体内容分为左右两部分（如果屏幕较小则上下堆叠）。左侧区域放一个“添加开销”的表单占位（包含几个基础输入框和一个按钮即可，暂时不需要实际功能）；右侧区域上半部分放一个“数据统计”占位区块，下半部分放一个“开销记录”的列表占位区块。整体布局要有留白，样式力求简洁现代。另外，后续所有相关开发工作都必须始终遵循以下要求：确保每一个生成的代码文件里，注释部分完全使用中文；代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n def count_comment_lines(text: str, ext: str) -> int:\n line_num = 0\n i = 0\n in_str = False\n str_char = ''\n in_line_comment = False\n in_block_comment = False\n block_end_str = ''\n lines_with_comment = set()\n \n while i < len(text):\n char = text[i]\n \n # 处理换行符，更新当前行号\n if char == '\\n':\n line_num += 1\n if in_line_comment:\n in_line_comment = False\n i += 1\n continue\n \n # 当处于单行注释内时，将当前行加入集合，并且忽略后续的字符判断\n if in_line_comment:\n lines_with_comment.add(line_num)\n i += 1\n continue\n \n # 当处于多行/块注释内时，将当前行加入集合，并判断块注释是否到达结束标识\n if in_block_comment:\n lines_with_comment.add(line_num)\n if text[i:i+len(block_end_str)] == block_end_str:\n in_block_comment = False\n i += len(block_end_str)\n else:\n i += 1\n continue\n \n # 当处于字符串字面量内时，需跳过转义字符以避免误判结束符号\n if in_str:\n if char == '\\\\':\n i += 2\n elif char == str_char:\n in_str = False\n i += 1\n else:\n i += 1\n continue\n \n # 捕获字符串的起始字符（HTML通常无需处理字符串中的注释误判，简化处理）\n if char in ['\"', \"'\", '`'] and ext not in ['.html', '.htm']:\n in_str = True\n str_char = char\n i += 1\n continue\n \n # 检测 JS / TS / CSS 类文件的注释起点\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less']:\n if char == '/' and i + 1 < len(text):\n if text[i+1] == '/':\n in_line_comment = True\n lines_with_comment.add(line_num)\n i += 2\n continue\n elif text[i+1] == '*':\n in_block_comment = True\n block_end_str = '*/'\n lines_with_comment.add(line_num)\n i += 2\n continue\n \n # 检测 HTML 类文件的注释起点\n elif ext in ['.html', '.htm']:\n if text[i:i+4] == '<!--':\n in_block_comment = True\n block_end_str = '-->'\n lines_with_comment.add(line_num)\n i += 4\n continue\n \n i += 1\n \n return len(lines_with_comment)\n\n # 定义前端React应用相关的常见代码文件扩展名\n code_exts = ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm']\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的构建与依赖目录\n for ignore_dir in ['node_modules', '.git', 'build', 'dist', '__pycache__']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n \n # 若存在任意一个代码文件注释行数不为3，则不满足要求\n comment_lines = count_comment_lines(text, ext)\n if comment_lines != 3:\n return False\n except Exception:\n pass\n \n # 如果项目中没有任何代码文件，同样视为未满足\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def check_no_ternary_operator(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def has_ternary(code: str) -> bool:\n # 去除代码中的双引号字符串、单引号字符串、反引号(模板)字符串\n # 同时去除多行注释 /* ... */ 以及单行注释 // ...\n pattern = r'''\"(?:\\\\.|[^\"\\\\])*\"|'(?:\\\\.|[^'\\\\])*'|`(?:\\\\.|[^`\\\\])*`|/\\*[^*]*\\*+(?:[^/*][^*]*\\*+)*/|//.*'''\n clean_code = re.sub(pattern, '', code)\n \n # 查找代码中独立的 '?'\n # 1. (?<!\\?) 确保其前一个字符不是 '?' (排除 ?? 的后半部分)\n # 2. \\? 匹配问号本身\n # 3. (?![.?:]) 确保其后一个字符不是 '.'、'?' 或 ':' \n # (排除可选链 ?. 、空值合并 ?? 以及 TypeScript可选属性/类型等 ?: )\n matches = re.finditer(r'(?<!\\?)\\?(?![.?:])', clean_code)\n \n for match in matches:\n # 获取当前独立 '?' 之后的剩余代码内容\n rest_of_code = clean_code[match.end():]\n # 如果后面在同一个闭环或后续表达式内存在配套的 ':'，则认为存在三元运算符\n if ':' in rest_of_code:\n return True\n \n return False\n\n # 验证工作区目录是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历检索生成的代码项目\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选常见前端与 React 项目文件类型\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.html', '.htm')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 只要任何一个文件内查找到存在三元运算符，即判断为未满足“不得使用”的要求\n if has_ternary(content):\n return False\n except Exception:\n # 对于不可读的文件异常安全处理地跳过\n continue\n\n # 遍历结束未发现任何包含三元运算符的代码文件，要求已满足\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个数是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 结合React项目，定义常见的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\n code_file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除一些常见非手工或大批量的生成目录（虽然可能并不存在，但确保安全）\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅校验属于“代码文件”范畴的文件\n if ext in code_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按 \\n 切分可以确保准确包含结尾的空行以及文件中间的全部空行\n total_lines = len(content.split('\\n'))\n \n # 只要有一个代码文件行数不为质数，即不满足要求\n if not is_prime(total_lines):\n return False\n except Exception:\n # 如果由于读取发生异常（如非常规文件或编码错误），稳妥起见视为不符合要求\n return False\n \n # 如果没有找到任何代码文件，说明未生成有效的代码项目内容，返回False；否则返回True\n return code_file_found","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":1,"instruction":"好，接下来我们让表单和列表真正工作起来。完善左侧的表单，必须包含标题（文本）、金额（数字）、分类（下拉单选，如餐饮、交通、购物、娱乐等）和日期（日期选择器）四个字段。点击添加按钮后，表单被清空，这条记录要能显示在右侧的开销列表中，列表中的每行要清晰展示这四个字段的信息，并且每行末尾要有一个“删除”按钮来移除该记录。同时，在列表顶部增加一个“总开销”的卡片，实时计算并显示当前列表中所有记录的金额总和。另外，后续所有相关开发工作都必须始终遵循以下两项要求：一是在生成的各个代码文件中，注释内容必须统一采用英语编写；二是处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n \n # 用于匹配所有的双引号、单引号和反引号所包裹的字符串，处理转义字符\n string_pattern = r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n \n has_checked_any_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的文件夹\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'build', 'dist', 'public')]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n continue\n \n has_checked_any_file = True\n \n # 为了防止索引改变影响行号计算，利用同等长度的空格替换字符串内的内容\n def replacer(match):\n return ' ' * len(match.group(0))\n \n text_no_str = re.sub(string_pattern, replacer, text)\n \n # 记录所有换行符后面的起始索引，以此推断位置所处的行号\n line_starts = [0]\n for m in re.finditer(r'\\n', text):\n line_starts.append(m.end())\n \n def get_line_number(index):\n return bisect.bisect_right(line_starts, index) - 1\n \n comment_lines = set()\n patterns = []\n \n # 不同语言的代码分配不同的注释匹配模式\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\n patterns = [r'/\\*[\\s\\S]*?\\*/', r'//.*']\n elif ext in ['.css']:\n patterns = [r'/\\*[\\s\\S]*?\\*/']\n elif ext in ['.html']:\n patterns = [r'<!--[\\s\\S]*?-->']\n \n # 遍历正则匹配结果，记录下注释覆盖过的行号\n for pattern in patterns:\n for m in re.finditer(pattern, text_no_str):\n start_idx = m.start()\n end_idx = m.end()\n \n start_line = get_line_number(start_idx)\n end_line = get_line_number(end_idx - 1) if end_idx > start_idx else start_line\n \n for line in range(start_line, end_line + 1):\n comment_lines.add(line)\n \n # 只要任何一个代码文件的注释行数不为3，则直接返回 False\n if len(comment_lines) != 3:\n return False\n \n return has_checked_any_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def remove_comments_and_strings(code: str) -> str:\n # 匹配多行注释、单行注释以及双引号、单引号、反引号包裹的字符串字面量\n pattern = (\n r'/\\*[\\s\\S]*?\\*/|'\n r'//.*|'\n r'\"(?:\\\\.|[^\"\\\\])*\"|'\n r\"'(?:\\\\.|[^'\\\\])*'|\"\n r'`(?:\\\\.|[^`\\\\])*`'\n )\n return re.sub(pattern, '', code)\n\n def has_ternary_operator(code: str) -> bool:\n cleaned_code = remove_comments_and_strings(code)\n \n # 匹配三元运算符的核心特征：? 和 : 配对\n # \\? : 匹配问号\n # (?![.?]) : 排除 ?. (可选链) 和 ?? (空值合并)\n # \\s* : 允许前导空白符\n # [^?: \\t\\n\\r] : 确保 ? 和 : 之间至少存在一个有意义的非空白字符，借此排除 TypeScript 中的 ?:\n # [^?:]* : 中间可以允许其他不包含 ? 和 : 的字符\n # : : 匹配假值分支引出的冒号\n pattern = r'\\?(?![.?])\\s*[^?: \\t\\n\\r][^?:]*:'\n \n if re.search(pattern, cleaned_code):\n return True\n return False\n\n if not os.path.exists(workspace_path):\n return False\n\n valid_extensions = {'.js', '.jsx', '.ts', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n if has_ternary_operator(code):\n # 如果发现三元运算符，直接返回不满足要求\n return False\n except Exception:\n # 遇到不能读取的文件则跳过\n continue\n \n # 如果所有代码文件都没发现三元运算符，则符合要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # React项目中常见代码文件的扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n # 忽略非生成代码目录\n ignore_dirs = {'node_modules', 'dist', 'build', '.git', '.idea', '.vscode'}\n \n found_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 如果当前路径中包含需要忽略的目录名，则跳过\n path_parts = set(root.split(os.sep))\n if path_parts.intersection(ignore_dirs):\n continue\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计总行数（split('\\n')会原样保留所有空行）\n lines = content.split('\\n')\n num_lines = len(lines)\n \n # 判断行数是否为质数\n if not is_prime(num_lines):\n return False\n \n found_code_files = True\n except Exception:\n # 如果遇到文件读取异常（如编码错误等），跳过\n continue\n \n # 如果至少检查了一个有效代码文件并且全部满足质数行数要求，则通过验证\n return found_code_files","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的各个代码文件中，注释内容必须统一采用英语编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":2,"instruction":"现在来加上本地数据持久化和表单校验功能。首先，将列表数据保存到浏览器的 LocalStorage 中，确保刷新页面数据不会丢失，初始化时也要优先从 LocalStorage 加载。其次，给添加表单加上实时校验逻辑：标题不能为空；金额必须大于0且只能输入最多包含两位小数的数值；日期不能晚于今天（不能选未来日期）。当用户输入不符合要求时，在对应输入框下方显示红色的错误提示，并且禁用“添加”按钮。另外，后续所有相关开发工作都必须始终遵循以下更新后的要求：其一，此前要求的注释统一使用英语的规则作废，更新为你提供的每个代码文件里，注释内容须100%使用中文撰写；其二，新增要求：写完代码后，必须对所有的代码进行测试以确保测试通过。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"import os\n\ndef check_comment_lines_requirement(response: str, workspace_path: str) -> bool:\n # 规定要检查的代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n code_files_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除掉常见的非自己手写的依赖包及构建结果文件夹\n dirs[:] = [d for d in dirs if d not in ('node_modules', 'build', 'dist', '.git', 'out')]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n continue\n \n # 忽略完全空的文件\n if not text.strip():\n continue\n\n code_files_found = True\n comment_lines = set()\n lines = text.split('\\n')\n \n # 状态机标志位\n in_multiline_comment = False\n in_html_comment = False\n in_string = False\n string_char = ''\n \n for i, line in enumerate(lines):\n j = 0\n in_singleline_comment = False\n \n while j < len(line):\n if in_multiline_comment:\n comment_lines.add(i)\n if line[j:j+2] == '*/':\n in_multiline_comment = False\n j += 2\n else:\n j += 1\n continue\n \n if in_html_comment:\n comment_lines.add(i)\n if line[j:j+3] == '-->':\n in_html_comment = False\n j += 3\n else:\n j += 1\n continue\n \n if in_singleline_comment:\n comment_lines.add(i)\n break\n \n if in_string:\n if line[j] == '\\\\':\n j += 2\n continue\n if line[j] == string_char:\n in_string = False\n j += 1\n continue\n \n # 检查当前字符及其组合\n if ext in {'.js', '.jsx', '.ts', '.tsx'}:\n if line[j] in '\"\\'`':\n in_string = True\n string_char = line[j]\n j += 1\n continue\n if line[j:j+2] == '//':\n in_singleline_comment = True\n comment_lines.add(i)\n j += 2\n continue\n if line[j:j+2] == '/*':\n in_multiline_comment = True\n comment_lines.add(i)\n j += 2\n continue\n \n elif ext == '.css':\n if line[j] in '\"\\'':\n in_string = True\n string_char = line[j]\n j += 1\n continue\n if line[j:j+2] == '/*':\n in_multiline_comment = True\n comment_lines.add(i)\n j += 2\n continue\n \n elif ext == '.html':\n if line[j] in '\"\\'':\n in_string = True\n string_char = line[j]\n j += 1\n continue\n if line[j:j+4] == '<!--':\n in_html_comment = True\n comment_lines.add(i)\n j += 4\n continue\n \n j += 1\n \n # 检查该代码文件中的注释总行数是否严格等于 3\n if len(comment_lines) != 3:\n return False\n\n return code_files_found","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对React项目的常见代码文件扩展名\n exts = ('.js', '.jsx', '.ts', '.tsx')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n # 1. 移除单行和多行注释\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n # 2. 移除单引号和双引号包裹的字符串\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", code)\n # 移除不含表达式插值 (${}) 的普通模板字符串\n code = re.sub(r'`(?:\\\\.|[^`\\\\$])*`', '``', code)\n \n # 3. 移除包含问号的正则表达式字面量，防止误报\n code = re.sub(r'/(?![/*])[^/\\n\\\\]*\\?[^/\\n\\\\]*/[gimsuy]*', '', code)\n \n # 4. 移除JSX中的普通文本部分 (匹配不在大括号内的标签间文本)\n # 纯文本通常位于 > 或 } 之后，且在 < 或 { 之前\n code = re.sub(r'(?<=[>}])([^<>{}]*)(?=[<{])', '', code)\n \n # 5. 替换掉JS/TS中合法的非三元运算符问号组合\n code = code.replace('?.', '') # 可选链\n code = code.replace('??', '') # 空值合并运算符\n code = re.sub(r'\\?\\s*:', '', code) # TS可选属性/参数\n \n # 6. 查找是否存在真正的三元运算符特征\n # 匹配 '?' 和后续的 ':'，且判定这段区间内不存在语句结束符 ';'\n matches = re.finditer(r'\\?(.*?):', code, re.DOTALL)\n for match in matches:\n middle = match.group(1)\n # 三元运算符作为表达式内部通常不存在分号，以此进行安全确认\n if ';' not in middle:\n return False\n except Exception:\n pass\n \n # 如果所有文件都没有命中三元运算符，代表满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 根据React项目推断可能的代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 兼容不同系统的换行符，并统一为 \\n 进行分割\n content = content.replace('\\r\\n', '\\n')\n # split('\\n') 能够准确计算行数，包括文件末尾或中间的空行\n lines_count = len(content.split('\\n'))\n except Exception:\n # 如果有文件读取异常直接跳过\n continue\n \n has_code_file = True\n # 若存在任何一个代码文件的行数不是质数，则不满足要求\n if not is_prime(lines_count):\n return False\n \n # 如果所有的代码文件行数均为质数且至少验证了一个代码文件，则判定为通过\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"下一步加入预算预警系统。在界面顶部的总开销卡片旁边，增加一个“设置月度预算”的功能（支持输入数字并保存到LocalStorage）。设置好预算后，下方需要显示一个预算使用进度的进度条。规则如下：当总开销低于预算的80%时，进度条为绿色；当达到预算的80%但低于100%时，进度条变成黄色，并在旁边显示“即将超支”的警告文本；如果达到或超过预算的100%，进度条变成红色，并且在每次添加新记录导致超标时，在屏幕中央弹出一个醒目的“预算已超标”警告弹窗（需要手动关闭）。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用中文的规则作废，更新为你生成的每一份代码文件中的注释全都是英文；其二，新增要求：运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n if not os.path.exists(workspace_path):\n return False\n\n # 匹配字符串（单、双引号和反引号，防止字符串内的符合被误认为注释）以及各类注释\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<ml_comment>/\\*[\\s\\S]*?\\*/)'\n r'|(?P<sl_comment>//[^\\n]*)'\n r'|(?P<html_comment><!--[\\s\\S]*?-->)'\n )\n\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\n found_any_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的依赖、Git和编译目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if 'build' in dirs:\n dirs.remove('build')\n if 'dist' in dirs:\n dirs.remove('dist')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n found_any_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 记录所有换行符后面的索引，作为每一行的起始位置\n line_starts = [0]\n for m in re.finditer(r'\\n', content):\n line_starts.append(m.start() + 1)\n \n def get_line_number(index: int) -> int:\n # 返回基于 1 的行号\n return bisect.bisect_right(line_starts, index)\n\n comment_lines = set()\n for m in pattern.finditer(content):\n # 仅当匹配到的内容属于注释组时才处理其行号\n if m.lastgroup in ('ml_comment', 'sl_comment', 'html_comment'):\n start_line = get_line_number(m.start())\n # 如果结束索引大于开始索引，取结束的前一个字符所在的行；否则按原样\n end_index = m.end() - 1 if m.end() > m.start() else m.start()\n end_line = get_line_number(end_index)\n \n # 将从起始行到结束行之间包含注释的行号加入集合\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n \n # 如果有任何一个代码文件中的注释占用行数不是精确的 3 行，即不符合要求\n if len(comment_lines) != 3:\n return False\n\n return found_any_code_file","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def verify_no_ternary_operator(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 替换函数：将注释直接去除，字符串字面量仅保留引号外壳防止其内部干扰，同时也能避免破坏代码结构导致漏判\n def replacer(match):\n text = match.group(0)\n if text.startswith('//') or text.startswith('/*'):\n return ''\n elif text.startswith('\"'):\n return '\"\"'\n elif text.startswith(\"'\"):\n return \"''\"\n elif text.startswith('`'):\n return '``'\n return text\n \n # 正则表达式匹配：单行注释 | 多行注释 | 单引号字符串 | 双引号字符串 | 反引号字符串\n pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|' +\n r'\\'(?:\\\\.|[^\\\\\\'])*\\'|' +\n r'\"(?:\\\\.|[^\\\\\"])*\"|' +\n r'`(?:\\\\.|[^\\\\`])*`',\n re.DOTALL | re.MULTILINE\n )\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 剔除干扰项\n clean_code = pattern.sub(replacer, code)\n length = len(clean_code)\n \n i = 0\n while i < length:\n if clean_code[i] == '?':\n # 排除 ?? (空值合并)\n if i > 0 and clean_code[i-1] == '?':\n i += 1\n continue\n if i < length - 1 and clean_code[i+1] == '?':\n i += 2\n continue\n \n # 排除 ?. (可选链)\n if i < length - 1 and clean_code[i+1] == '.':\n i += 2\n continue\n \n # 排除 ?: (TypeScript可选属性/参数等，其中?和:之间仅包含空白符的情况)\n j = i + 1\n while j < length and clean_code[j].isspace():\n j += 1\n if j < length and clean_code[j] == ':':\n i = j + 1\n continue\n \n # 进行候选三元运算符的安全过滤（典型的三元运算符两侧常常带有空格或位于闭合括号后）\n prev_char = clean_code[i-1] if i > 0 else ''\n next_char = clean_code[i+1] if i < length - 1 else ''\n \n is_ternary_candidate = (\n prev_char.isspace() or \n next_char.isspace() or \n prev_char in ')]}'\n )\n \n if is_ternary_candidate:\n # 动态作用域解析：向后寻找同一作用域层级的匹配冒号\n depth = 0\n k = j\n has_colon = False\n \n while k < length:\n char = clean_code[k]\n if char in '({[':\n depth += 1\n elif char in ')}]':\n depth -= 1\n if depth < 0: # 跳出了当前所在的作用域层级\n break\n elif char == ';' and depth == 0: # 当前语句结束\n break\n elif char == ':' and depth == 0: # 在同等层级找到了对应的冒号\n has_colon = True\n break\n k += 1\n \n # 一旦找到符合结构定义的三元配对，证明其使用了三元运算符，直接判负\n if has_colon:\n return False\n i += 1\n \n # 如果所有的 JS/TS 代码文件均未发现三元运算符特征，则认为满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n\n # 定义React项目通常包含的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\n \n # 应当忽略的常见目录（构建产物、依赖包、隐藏目录等）\n ignore_dirs = {'node_modules', 'build', 'dist', 'public', '.git'}\n\n has_code_file = False\n \n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n # 排除忽略目录及隐藏目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs and not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n num_lines = len(lines)\n \n # 检查总行数是否为质数\n if not is_prime(num_lines):\n return False\n \n has_code_file = True\n except Exception:\n # 如果无法使用utf-8读取文件等异常情况，不作为主要检查依据\n pass\n \n # 如果找到了合法的代码文件且均符合质数要求，则返回 True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的每一份代码文件中的注释全都是英文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":4,"instruction":"现在来完善右上角的“数据统计”区块，让数据可视化。请引入图表库（如Recharts），在这里实现两个图表：左边是一个按分类统计开销占比的饼图（带有图例，鼠标悬浮能显示具体金额），右边是一个展示过去7天每天开销金额趋势的柱状图或折线图。这两个图表的数据必须和下方开销列表的数据完全联动，当我添加或删除记录时，图表要自动重新渲染计算。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用英语的规则作废，更新为确保每一个生成的代码文件里，注释部分完全使用中文；其二，新增要求：请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"import os\n\ndef check_comments_in_code_files(response: str, workspace_path: str) -> bool:\n # 定义 React 项目中常见的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css'}\n html_extensions = {'.html'}\n \n # 忽略非直接生成的系统、依赖或构建目录\n exclude_dirs = {'.git', 'node_modules', 'build', 'dist', '.next', 'coverage'}\n \n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉排除的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions or ext in html_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n code_files_found += 1\n \n # 针对 JS/TS/CSS 等样式或脚本文件解析注释\n if ext in code_extensions:\n lines_with_comments = set()\n i = 0\n in_string = False\n string_char = ''\n in_sl_comment = False\n in_ml_comment = False\n escaped = False\n line_num = 1\n \n while i < len(code):\n # 处理换行符\n if code[i] == '\\n':\n if in_sl_comment:\n in_sl_comment = False\n line_num += 1\n i += 1\n escaped = False\n continue\n \n # 如果当前字符是被转义的，则直接跳过\n if escaped:\n escaped = False\n i += 1\n continue\n \n # 如果在单行注释中，该行记录为含有注释，并继续前进\n if in_sl_comment:\n lines_with_comments.add(line_num)\n i += 1\n continue\n \n # 如果在多行注释中\n if in_ml_comment:\n lines_with_comments.add(line_num)\n # 检测多行注释结束\n if code[i:i+2] == '*/':\n in_ml_comment = False\n i += 2\n else:\n i += 1\n continue\n \n # 如果在字符串常量中\n if in_string:\n if code[i] == '\\\\':\n escaped = True\n i += 1\n elif code[i] == string_char:\n in_string = False\n i += 1\n else:\n i += 1\n continue\n \n # 如果不在字符串，也不在注释中\n if code[i] in ['\"', \"'\", '`']:\n in_string = True\n string_char = code[i]\n i += 1\n continue\n \n # 检测单行注释起步 (CSS 不支持单行注释)\n if ext != '.css' and code[i:i+2] == '//':\n in_sl_comment = True\n lines_with_comments.add(line_num)\n i += 2\n continue\n \n # 检测多行注释起步\n if code[i:i+2] == '/*':\n in_ml_comment = True\n lines_with_comments.add(line_num)\n i += 2\n continue\n \n i += 1\n \n # 检查此文件的注释行数是否严格为 3\n if len(lines_with_comments) != 3:\n return False\n \n # 针对 HTML 文件的解析注释\n elif ext in html_extensions:\n lines_with_comments = set()\n i = 0\n in_comment = False\n line_num = 1\n \n while i < len(code):\n if code[i] == '\\n':\n line_num += 1\n i += 1\n continue\n \n if in_comment:\n lines_with_comments.add(line_num)\n if code[i:i+3] == '-->':\n in_comment = False\n i += 3\n else:\n i += 1\n continue\n \n if code[i:i+4] == '<!--':\n in_comment = True\n lines_with_comments.add(line_num)\n i += 4\n continue\n \n i += 1\n \n if len(lines_with_comments) != 3:\n return False\n\n # 若没有任何代码文件，也视作要求未完成\n if code_files_found == 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def clean_code(code: str) -> str:\n # 正则表达式匹配并移除：\n # 1. HTML/XML 多行注释\n # 2. JS/TS/CSS 多行注释\n # 3. JS/TS 单行注释\n # 4. 双引号字符串\n # 5. 单引号字符串\n # 6. 反引号（模板）字符串\n pattern = re.compile(\n r'<!--[\\s\\S]*?-->|'\n r'/\\*[\\s\\S]*?\\*/|'\n r'//.*|'\n r'\"(?:\\\\.|[^\"\\\\])*\"|'\n r\"'(?:\\\\.|[^'\\\\])*'|\"\n r'`(?:\\\\.|[^`\\\\])*`'\n )\n return pattern.sub('', code)\n\n def contains_ternary(code: str) -> bool:\n cleaned_code = clean_code(code)\n \n # 寻找代码中所有的 ?\n for match in re.finditer(r'\\?', cleaned_code):\n start = match.end()\n # 找到 ? 后方剩余的代码片段，并去除开头的空白符\n after_q = cleaned_code[start:].lstrip()\n \n # 排除 ?. (可选链)\n # 排除 ?? (空值合并运算符)\n # 排除 ?: (TypeScript中的可选参数或可选属性声明)\n if after_q.startswith('.') or after_q.startswith('?') or after_q.startswith(':'):\n continue\n \n # 如果不属于上述情况，说明这是一个用于条件操作的单独 ? 符号。\n # 只要其后方还存在配套的 : ，就确认代码中使用了三元运算符。\n if ':' in after_q:\n return True\n \n return False\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 可能涉及到的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n # 如果有文件读取发生异常则跳过\n continue\n \n if contains_ternary(code):\n # 若在任何一个文件中发现了三元运算符，则判断未满足要求，返回 False\n return False\n\n # 所有生成的文件验证完毕，没有发现任何形式的三元运算符\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个数字是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 根据React应用推断出的常见代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n code_files_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除掉无关的第三方库和构建产物目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'build', 'dist', 'public')]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 兼容处理 Windows (CRLF) 与 Unix (LF) 的换行符，并严格切分获取所有行（含空行）\n lines = content.replace('\\r\\n', '\\n').split('\\n')\n line_count = len(lines)\n \n # 如果行数不是质数，则要求未满足\n if not is_prime(line_count):\n return False\n except Exception:\n # 无法读取该文件，视同验证未通过\n return False\n\n # 若至少找到了一个代码文件，并且全部通过了检验，则返回True\n return code_files_found","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":5,"instruction":"为了方便查看历史数据，我们需要给列表加上高级筛选和排序功能。在右侧开销列表的正上方增加一个工具栏：包含一个“时间段”下拉框（本周、本月、所有）、一个“分类”多选框（勾选后只显示选中的分类），以及一个“排序规则”下拉框（按日期最新、按日期最旧、按金额最高、按金额最低）。列表要根据这些条件实时更新。注意：筛选不仅要改变列表里展示的内容，还要同步让上方数据统计的图表和总开销、预算进度条都基于筛选后的数据进行计算和展示。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用中文的规则作废，更新为你所输出的任何一个代码文件，其包含的注释都应当是全英文的；其二，新增要求：请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n \n code_files = []\n for root, dirs, files in os.walk(workspace_path):\n # 排除掉常见的依赖包、构建及版本控制相关目录\n for exclude_dir in ['node_modules', '.git', 'build', 'dist']:\n if exclude_dir in dirs:\n dirs.remove(exclude_dir)\n \n for f in files:\n ext = os.path.splitext(f)[1].lower()\n if ext in code_exts:\n code_files.append(os.path.join(root, f))\n \n # 如果没有找到任何相关的代码文件，也代表未按要求成功生成文件\n if not code_files:\n return False\n \n for filepath in code_files:\n ext = os.path.splitext(filepath)[1].lower()\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n lines = content.split('\\n')\n comment_lines = set()\n \n in_block_comment = False\n in_string = False\n string_char = ''\n \n for i, line in enumerate(lines):\n j = 0\n while j < len(line):\n # 处于块注释中\n if in_block_comment:\n if ext == '.html':\n if line[j:j+3] == '-->':\n in_block_comment = False\n j += 2\n else:\n if line[j:j+2] == '*/':\n in_block_comment = False\n j += 1\n comment_lines.add(i)\n j += 1\n continue\n \n # 处于字符串字面量中\n if in_string:\n if line[j] == '\\\\':\n j += 1\n elif line[j] == string_char:\n in_string = False\n j += 1\n continue\n \n # 处于正常代码解析状态\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\n if line[j:j+2] == '//':\n comment_lines.add(i)\n break\n elif line[j:j+2] == '/*':\n in_block_comment = True\n comment_lines.add(i)\n j += 1\n elif line[j] in [\"'\", '\"', '`']:\n in_string = True\n string_char = line[j]\n \n elif ext in ['.css']:\n if line[j:j+2] == '/*':\n in_block_comment = True\n comment_lines.add(i)\n j += 1\n elif line[j] in [\"'\", '\"']:\n in_string = True\n string_char = line[j]\n \n elif ext in ['.html']:\n if line[j:j+4] == '<!--':\n in_block_comment = True\n comment_lines.add(i)\n j += 3\n elif line[j] in [\"'\", '\"']:\n in_string = True\n string_char = line[j]\n \n elif ext in ['.py']:\n if line[j] == '#':\n comment_lines.add(i)\n break\n elif line[j] in [\"'\", '\"']:\n in_string = True\n string_char = line[j]\n \n j += 1\n \n # 若行结束时依然处于块注释内部，则当前行也应作为注释行计算\n if in_block_comment:\n comment_lines.add(i)\n \n # 验证每一个代码文件的注释行数是否精确为3行\n if len(comment_lines) != 3:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def strip_comments_and_strings(code: str) -> str:\n \"\"\"\n 一个轻量级的词法状态机，用来剔除 JavaScript/TypeScript 中的:\n 字符串 (\", ', `)、注释 (//, /* */) 以及 正则表达式字面量 (/.../) 内部的内容\n 以防止这些内容里的 '?' 造成误报。\n \"\"\"\n i = 0\n n = len(code)\n \n NORMAL = 0\n STR_DOUBLE = 1\n STR_SINGLE = 2\n STR_TEMPLATE = 3\n BLOCK_COMMENT = 4\n LINE_COMMENT = 5\n REGEX = 6\n \n state = NORMAL\n state_stack = [] \n last_token = ''\n current_word = ''\n \n res = []\n \n kw = {'return', 'yield', 'typeof', 'await', 'throw', 'case', 'in', 'instanceof', 'do', 'else', 'void', 'delete', 'new'}\n regex_preceding_tokens = {'', '=', '(', ',', '[', '{', ':', ';', '!', '&', '|', '?', '<', '>', '+', '-', '*', '/', '%'} | kw\n \n while i < n:\n char = code[i]\n \n if state == NORMAL:\n is_ident_char = char.isalnum() or char == '_' or char == '$'\n if is_ident_char:\n current_word += char\n res.append(char)\n i += 1\n continue\n else:\n if current_word:\n if current_word in kw:\n last_token = current_word\n else:\n last_token = 'IDENT'\n current_word = ''\n \n if char.isspace():\n res.append(char)\n i += 1\n continue\n \n if char == '/' and i + 1 < n and code[i+1] == '*':\n state = BLOCK_COMMENT\n i += 2\n continue\n elif char == '/' and i + 1 < n and code[i+1] == '/':\n state = LINE_COMMENT\n i += 2\n continue\n elif char == '\"':\n state = STR_DOUBLE\n i += 1\n last_token = 'STRING'\n continue\n elif char == \"'\":\n state = STR_SINGLE\n i += 1\n last_token = 'STRING'\n continue\n elif char == '`':\n state = STR_TEMPLATE\n i += 1\n last_token = 'STRING'\n continue\n elif char == '{':\n state_stack.append(NORMAL)\n last_token = '{'\n res.append(char)\n i += 1\n continue\n elif char == '}':\n last_token = '}'\n if state_stack:\n prev_state = state_stack.pop()\n if prev_state == STR_TEMPLATE:\n state = STR_TEMPLATE\n res.append(char)\n i += 1\n continue\n elif char == '/':\n # 利用启发式前置Token来区分 正则字面量 和 除号\n if last_token in regex_preceding_tokens:\n state = REGEX\n i += 1\n continue\n else:\n last_token = '/'\n res.append(char)\n i += 1\n continue\n else:\n last_token = char\n res.append(char)\n i += 1\n continue\n \n elif state == STR_DOUBLE:\n if char == '\\\\': i += 2\n elif char == '\"': state = NORMAL; i += 1\n else: i += 1\n \n elif state == STR_SINGLE:\n if char == '\\\\': i += 2\n elif char == \"'\": state = NORMAL; i += 1\n else: i += 1\n \n elif state == STR_TEMPLATE:\n if char == '\\\\': i += 2\n elif char == '`': state = NORMAL; i += 1\n elif char == '$' and i + 1 < n and code[i+1] == '{':\n state_stack.append(STR_TEMPLATE)\n state = NORMAL\n last_token = '{'\n res.append('$')\n res.append('{')\n i += 2\n else: i += 1\n \n elif state == BLOCK_COMMENT:\n if char == '*' and i + 1 < n and code[i+1] == '/': state = NORMAL; i += 2\n else: i += 1\n \n elif state == LINE_COMMENT:\n if char == '\\n' or char == '\\r':\n state = NORMAL\n res.append(char)\n i += 1\n else: i += 1\n \n elif state == REGEX:\n if char == '\\\\': i += 2\n elif char == '/':\n state = NORMAL\n last_token = 'REGEX'\n i += 1\n else: i += 1\n\n return \"\".join(res)\n\n def has_ternary_operator(code: str) -> bool:\n clean_code = strip_comments_and_strings(code)\n \n # 1. 过滤 JS/TS 其他包含 '?' 的常见特性\n clean_code = clean_code.replace('?.', '')\n clean_code = clean_code.replace('??=', '')\n clean_code = clean_code.replace('??', '')\n \n # 2. 过滤 TypeScript 的可选属性/参数标记, 如 prop?: type \n clean_code = re.sub(r'\\?\\s*:', '', clean_code)\n \n # 3. 过滤 TS 其他位置的可选标记, 如 [number?], (name?), name?, foo?; 等\n clean_code = re.sub(r'\\?\\s*([\\]\\),;])', r'\\1', clean_code)\n \n # 4. 过滤 JSX 的纯文本中的孤立 '?' (如 <div>Are you sure?</div>)\n # 匹配位于标签边界 > 和 < 之间，不包含任何逻辑块符号 {}, 以及避免匹配三元运算符所需 : 的纯显示文本片段\n prev_code = None\n while prev_code != clean_code:\n prev_code = clean_code\n clean_code = re.sub(r'(>)([^<{>:]*?)\\?([^<{>:]*?)(<)', r'\\1\\2\\3\\4', clean_code)\n \n # 5. 在经过重重干扰排除后的纯净代码骨架中，若依旧包含游离的问号，即必然为三元运算符\n return '?' in clean_code\n\n # 扫描给定的项目目录下所有的前端源文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n if has_ternary_operator(code):\n # 如果存在任意一处三元运算符，则立即判定不满足要求\n return False\n except Exception:\n continue\n\n # 扫描完所有相关代码文件均未发现三元运算符结构，满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 针对React项目，定义代码文件扩展名白名单\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\n # 需忽略的依赖目录和构建产物目录\n exclude_dirs = {'node_modules', '.git', 'build', 'dist'}\n \n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n num_lines = len(lines)\n \n # 检查总行数是否为质数\n if not is_prime(num_lines):\n return False\n except Exception:\n # 如果遇到不可读或非纯文本代码文件，按不满足要求处理\n return False\n \n # 如果项目中没有生成任何匹配的代码文件，返回False\n return code_files_found","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":6,"instruction":"最后，我们来增加数据导出和批量管理功能。在刚刚添加的筛选工具栏的最右侧，增加一个“导出 CSV”的按钮，点击后能把当前列表里正在展示的开销记录（即经过筛选后的数据）下载为一个包含表头（标题、分类、金额、日期）的CSV文件。同时，给列表里的每一行最左侧加一个复选框，并在列表底部提供一个“批量删除选中项”的按钮，只有当有记录被选中时该按钮才可用，点击并二次确认后，可以一次性删除所有被选中的记录。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用英文的规则作废，更新为你生成的各个代码文件中，所有的注释语句只能是中文；其二，新增要求：所有由你创建的代码文件，其编码必须统一为UTF-8。","instruction_following_checklist":[{"constraint":"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 联合匹配模式：同时匹配注释和字符串字面量，防止字符串内部的伪注释导致误判\n # 采用 named groups 便于后续分辨匹配到的是注释还是普通字符串\n pattern = re.compile(\n r'(?P<line_comment>//[^\\n]*)' # JS/TS/CSS 等单行注释\n r'|(?P<hash_comment>#[^\\n]*)' # Python/Shell/YAML 等单行注释\n r'|(?P<block_comment>/\\*[\\s\\S]*?\\*/)' # 多行块注释\n r'|(?P<html_comment><!--[\\s\\S]*?-->)' # HTML 注释\n r'|(?P<d_quote>\"(?:\\\\.|[^\"\\\\])*\")' # 双引号字符串\n r'|(?P<s_quote>\\'(?:\\\\.|[^\\'\\\\])*\\')' # 单引号字符串\n r'|(?P<t_quote>`(?:\\\\.|[^`\\\\])*`)' # 反引号模板字符串\n )\n\n # 常见的前端与通用代码文件后缀\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.py', '.sh'}\n \n # 忽略的构建或者环境依赖目录\n ignore_dirs = {'node_modules', 'build', 'dist', '.git', '__pycache__', 'coverage'}\n\n file_count = 0\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除忽略目录，防止进入深层查找\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n if ext in code_exts:\n file_path = os.path.join(root, file)\n file_count += 1\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n # 如果遇到无法使用 utf-8 读取的文件，从严判定为未满足条件\n return False\n \n comment_lines = set()\n \n # 查找所有的匹配项\n for match in pattern.finditer(code):\n # 如果匹配到的内容属于注释的分组之一\n if match.lastgroup in ['line_comment', 'hash_comment', 'block_comment', 'html_comment']:\n start = match.start()\n end = match.end()\n \n # 计算开始和结束行（利用起始点前面的换行符数量 + 1）\n start_line = code.count('\\n', 0, start) + 1\n end_line = code.count('\\n', 0, max(start, end - 1)) + 1\n \n # 把跨越的每一行存入集合\n for line in range(start_line, end_line + 1):\n comment_lines.add(line)\n \n # 校验：注释行数必须精确为 3 行\n if len(comment_lines) != 3:\n return False\n\n # 若没有任何代码文件，通常也认为没有真正完成任务，返回 False\n return file_count > 0","tags":["Quantity","Exact Value"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查生成的代码文件中是否没有使用任何形式的三元运算符。\n 满足要求（没有使用）返回 True，否则返回 False。\n \"\"\"\n # 针对 React 项目，只检查 JS/TS 及其 JSX/TSX 相关的源代码文件\n valid_extensions = {'.js', '.jsx', '.ts', '.tsx'}\n \n # 匹配三元运算符的正则表达式\n # 问号后不能紧跟 . (可选链) 或 ? (空值合并)\n # ? 和 : 之间不能存在分号 ;，且尽量非贪婪匹配以应对多个三元运算\n ternary_pattern = re.compile(r'\\?(?![.?])([^;]*?):')\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n # 忽略无法读取的文件\n continue\n \n # 1. 移除单行注释\n code = re.sub(r'//.*', '', code)\n # 2. 移除多行注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 3. 移除单引号字符串\n code = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", code)\n # 4. 移除双引号字符串\n code = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', code)\n # 5. 移除模板字符串\n code = re.sub(r'`([^`\\\\]|\\\\.)*`', '``', code)\n \n # 在清理后的代码中查找类似三元运算符的结构\n matches = ternary_pattern.finditer(code)\n for match in matches:\n content_between = match.group(1).strip()\n \n # 排除 TypeScript 中类似 `name?: string` 的可选类型声明（其中间往往不包含非空字符）\n if content_between != '':\n # 为了防止某些极限跨行情况捕获到不该捕获的内容，排查一下 JS 核心的声明关键字\n if not re.search(r'\\b(let|const|var|import|export)\\b', content_between):\n # 如果存在满足条件的情况，说明找到了三元运算符，违反了要求\n return False\n\n # 所有文件验证完毕，均未发现三元运算符，完全满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 根据React应用特性，定义常见的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\n \n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的文件夹\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用 split('\\n') 能最准确地反映包含所有空行（包括文件末尾）的行数\n total_lines = len(content.split('\\n'))\n \n if not is_prime(total_lines):\n return False\n \n has_code_file = True\n except Exception:\n # 如果发生文件读取异常（如意外的二进制文件等），为了稳妥起见忽略或可认为不符\n pass\n \n # 如果项目里面没有识别出任何代码文件，则不符合“生成的各个代码文件总行数为质数”的条件\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"你生成的各个代码文件中，所有的注释语句只能是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"写完代码后，必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.isdir(workspace_path):\n return False\n\n # 包含React项目相关的常见代码文件和配置文件的后缀，以及日志和Markdown文件\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html', '.json', '.md', '.log'}\n checked_files_count = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略无需验证的依赖库或编译输出文件夹\n ignore_dirs = {'node_modules', '.git', 'build', 'dist'}\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以UTF-8编码读取文件，若文件包含非UTF-8字符将抛出UnicodeDecodeError\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n checked_files_count += 1\n except UnicodeDecodeError:\n # 发现非UTF-8编码文件\n return False\n except Exception:\n # 忽略诸如权限不足等其他无法读取的情况\n pass\n \n # 只要存在相关的代码文件且均无引发解码错误，说明都满足UTF-8编码要求\n return checked_files_count > 0","tags":["Language","File Encoding"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用React搭建一个“开销跟踪应用”的基础页面骨架。需要一个带应用名称的顶部导航栏，主体内容分为左右两部分（如果屏幕较小则上下堆叠）。左侧区域放一个“添加开销”的表单占位（包含几个基础输入框和一个按钮即可，暂时不需要实际功能）；右侧区域上半部分放一个“数据统计”占位区块，下半部分放一个“开销记录”的列表占位区块。整体布局要有留白，样式力求简洁现代。另外，后续所有相关开发工作都必须始终遵循以下要求：确保每一个生成的代码文件里，注释部分完全使用中文；代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n def count_comment_lines(text: str, ext: str) -> int:\\n line_num = 0\\n i = 0\\n in_str = False\\n str_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n block_end_str = ''\\n lines_with_comment = set()\\n \\n while i < len(text):\\n char = text[i]\\n \\n # 处理换行符，更新当前行号\\n if char == '\\\\n':\\n line_num += 1\\n if in_line_comment:\\n in_line_comment = False\\n i += 1\\n continue\\n \\n # 当处于单行注释内时，将当前行加入集合，并且忽略后续的字符判断\\n if in_line_comment:\\n lines_with_comment.add(line_num)\\n i += 1\\n continue\\n \\n # 当处于多行/块注释内时，将当前行加入集合，并判断块注释是否到达结束标识\\n if in_block_comment:\\n lines_with_comment.add(line_num)\\n if text[i:i+len(block_end_str)] == block_end_str:\\n in_block_comment = False\\n i += len(block_end_str)\\n else:\\n i += 1\\n continue\\n \\n # 当处于字符串字面量内时，需跳过转义字符以避免误判结束符号\\n if in_str:\\n if char == '\\\\\\\\':\\n i += 2\\n elif char == str_char:\\n in_str = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n # 捕获字符串的起始字符（HTML通常无需处理字符串中的注释误判，简化处理）\\n if char in ['\\\"', \\\"'\\\", '`'] and ext not in ['.html', '.htm']:\\n in_str = True\\n str_char = char\\n i += 1\\n continue\\n \\n # 检测 JS / TS / CSS 类文件的注释起点\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less']:\\n if char == '/' and i + 1 < len(text):\\n if text[i+1] == '/':\\n in_line_comment = True\\n lines_with_comment.add(line_num)\\n i += 2\\n continue\\n elif text[i+1] == '*':\\n in_block_comment = True\\n block_end_str = '*/'\\n lines_with_comment.add(line_num)\\n i += 2\\n continue\\n \\n # 检测 HTML 类文件的注释起点\\n elif ext in ['.html', '.htm']:\\n if text[i:i+4] == '<!--':\\n in_block_comment = True\\n block_end_str = '-->'\\n lines_with_comment.add(line_num)\\n i += 4\\n continue\\n \\n i += 1\\n \\n return len(lines_with_comment)\\n\\n # 定义前端React应用相关的常见代码文件扩展名\\n code_exts = ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm']\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的构建与依赖目录\\n for ignore_dir in ['node_modules', '.git', 'build', 'dist', '__pycache__']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n \\n # 若存在任意一个代码文件注释行数不为3，则不满足要求\\n comment_lines = count_comment_lines(text, ext)\\n if comment_lines != 3:\\n return False\\n except Exception:\\n pass\\n \\n # 如果项目中没有任何代码文件，同样视为未满足\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def check_no_ternary_operator(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def has_ternary(code: str) -> bool:\\n # 去除代码中的双引号字符串、单引号字符串、反引号(模板)字符串\\n # 同时去除多行注释 /* ... */ 以及单行注释 // ...\\n pattern = r'''\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|'(?:\\\\\\\\.|[^'\\\\\\\\])*'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`|/\\\\*[^*]*\\\\*+(?:[^/*][^*]*\\\\*+)*/|//.*'''\\n clean_code = re.sub(pattern, '', code)\\n \\n # 查找代码中独立的 '?'\\n # 1. (?<!\\\\?) 确保其前一个字符不是 '?' (排除 ?? 的后半部分)\\n # 2. \\\\? 匹配问号本身\\n # 3. (?![.?:]) 确保其后一个字符不是 '.'、'?' 或 ':' \\n # (排除可选链 ?. 、空值合并 ?? 以及 TypeScript可选属性/类型等 ?: )\\n matches = re.finditer(r'(?<!\\\\?)\\\\?(?![.?:])', clean_code)\\n \\n for match in matches:\\n # 获取当前独立 '?' 之后的剩余代码内容\\n rest_of_code = clean_code[match.end():]\\n # 如果后面在同一个闭环或后续表达式内存在配套的 ':'，则认为存在三元运算符\\n if ':' in rest_of_code:\\n return True\\n \\n return False\\n\\n # 验证工作区目录是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历检索生成的代码项目\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选常见前端与 React 项目文件类型\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.html', '.htm')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 只要任何一个文件内查找到存在三元运算符，即判断为未满足“不得使用”的要求\\n if has_ternary(content):\\n return False\\n except Exception:\\n # 对于不可读的文件异常安全处理地跳过\\n continue\\n\\n # 遍历结束未发现任何包含三元运算符的代码文件，要求已满足\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个数是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 结合React项目，定义常见的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\\n code_file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除一些常见非手工或大批量的生成目录（虽然可能并不存在，但确保安全）\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅校验属于“代码文件”范畴的文件\\n if ext in code_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按 \\\\n 切分可以确保准确包含结尾的空行以及文件中间的全部空行\\n total_lines = len(content.split('\\\\n'))\\n \\n # 只要有一个代码文件行数不为质数，即不满足要求\\n if not is_prime(total_lines):\\n return False\\n except Exception:\\n # 如果由于读取发生异常（如非常规文件或编码错误），稳妥起见视为不符合要求\\n return False\\n \\n # 如果没有找到任何代码文件，说明未生成有效的代码项目内容，返回False；否则返回True\\n return code_file_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来我们让表单和列表真正工作起来。完善左侧的表单，必须包含标题（文本）、金额（数字）、分类（下拉单选，如餐饮、交通、购物、娱乐等）和日期（日期选择器）四个字段。点击添加按钮后，表单被清空，这条记录要能显示在右侧的开销列表中，列表中的每行要清晰展示这四个字段的信息，并且每行末尾要有一个“删除”按钮来移除该记录。同时，在列表顶部增加一个“总开销”的卡片，实时计算并显示当前列表中所有记录的金额总和。另外，后续所有相关开发工作都必须始终遵循以下两项要求：一是在生成的各个代码文件中，注释内容必须统一采用英语编写；二是处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n \\n # 用于匹配所有的双引号、单引号和反引号所包裹的字符串，处理转义字符\\n string_pattern = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n \\n has_checked_any_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的文件夹\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'build', 'dist', 'public')]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n has_checked_any_file = True\\n \\n # 为了防止索引改变影响行号计算，利用同等长度的空格替换字符串内的内容\\n def replacer(match):\\n return ' ' * len(match.group(0))\\n \\n text_no_str = re.sub(string_pattern, replacer, text)\\n \\n # 记录所有换行符后面的起始索引，以此推断位置所处的行号\\n line_starts = [0]\\n for m in re.finditer(r'\\\\n', text):\\n line_starts.append(m.end())\\n \\n def get_line_number(index):\\n return bisect.bisect_right(line_starts, index) - 1\\n \\n comment_lines = set()\\n patterns = []\\n \\n # 不同语言的代码分配不同的注释匹配模式\\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\\n patterns = [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'//.*']\\n elif ext in ['.css']:\\n patterns = [r'/\\\\*[\\\\s\\\\S]*?\\\\*/']\\n elif ext in ['.html']:\\n patterns = [r'<!--[\\\\s\\\\S]*?-->']\\n \\n # 遍历正则匹配结果，记录下注释覆盖过的行号\\n for pattern in patterns:\\n for m in re.finditer(pattern, text_no_str):\\n start_idx = m.start()\\n end_idx = m.end()\\n \\n start_line = get_line_number(start_idx)\\n end_line = get_line_number(end_idx - 1) if end_idx > start_idx else start_line\\n \\n for line in range(start_line, end_line + 1):\\n comment_lines.add(line)\\n \\n # 只要任何一个代码文件的注释行数不为3，则直接返回 False\\n if len(comment_lines) != 3:\\n return False\\n \\n return has_checked_any_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def remove_comments_and_strings(code: str) -> str:\\n # 匹配多行注释、单行注释以及双引号、单引号、反引号包裹的字符串字面量\\n pattern = (\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|'\\n r'//.*|'\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|'\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\"\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n )\\n return re.sub(pattern, '', code)\\n\\n def has_ternary_operator(code: str) -> bool:\\n cleaned_code = remove_comments_and_strings(code)\\n \\n # 匹配三元运算符的核心特征：? 和 : 配对\\n # \\\\? : 匹配问号\\n # (?![.?]) : 排除 ?. (可选链) 和 ?? (空值合并)\\n # \\\\s* : 允许前导空白符\\n # [^?: \\\\t\\\\n\\\\r] : 确保 ? 和 : 之间至少存在一个有意义的非空白字符，借此排除 TypeScript 中的 ?:\\n # [^?:]* : 中间可以允许其他不包含 ? 和 : 的字符\\n # : : 匹配假值分支引出的冒号\\n pattern = r'\\\\?(?![.?])\\\\s*[^?: \\\\t\\\\n\\\\r][^?:]*:'\\n \\n if re.search(pattern, cleaned_code):\\n return True\\n return False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n valid_extensions = {'.js', '.jsx', '.ts', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n if has_ternary_operator(code):\\n # 如果发现三元运算符，直接返回不满足要求\\n return False\\n except Exception:\\n # 遇到不能读取的文件则跳过\\n continue\\n \\n # 如果所有代码文件都没发现三元运算符，则符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # React项目中常见代码文件的扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n # 忽略非生成代码目录\\n ignore_dirs = {'node_modules', 'dist', 'build', '.git', '.idea', '.vscode'}\\n \\n found_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 如果当前路径中包含需要忽略的目录名，则跳过\\n path_parts = set(root.split(os.sep))\\n if path_parts.intersection(ignore_dirs):\\n continue\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计总行数（split('\\\\n')会原样保留所有空行）\\n lines = content.split('\\\\n')\\n num_lines = len(lines)\\n \\n # 判断行数是否为质数\\n if not is_prime(num_lines):\\n return False\\n \\n found_code_files = True\\n except Exception:\\n # 如果遇到文件读取异常（如编码错误等），跳过\\n continue\\n \\n # 如果至少检查了一个有效代码文件并且全部满足质数行数要求，则通过验证\\n return found_code_files\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的各个代码文件中，注释内容必须统一采用英语编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来加上本地数据持久化和表单校验功能。首先，将列表数据保存到浏览器的 LocalStorage 中，确保刷新页面数据不会丢失，初始化时也要优先从 LocalStorage 加载。其次，给添加表单加上实时校验逻辑：标题不能为空；金额必须大于0且只能输入最多包含两位小数的数值；日期不能晚于今天（不能选未来日期）。当用户输入不符合要求时，在对应输入框下方显示红色的错误提示，并且禁用“添加”按钮。另外，后续所有相关开发工作都必须始终遵循以下更新后的要求：其一，此前要求的注释统一使用英语的规则作废，更新为你提供的每个代码文件里，注释内容须100%使用中文撰写；其二，新增要求：写完代码后，必须对所有的代码进行测试以确保测试通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"import os\\n\\ndef check_comment_lines_requirement(response: str, workspace_path: str) -> bool:\\n # 规定要检查的代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n code_files_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉常见的非自己手写的依赖包及构建结果文件夹\\n dirs[:] = [d for d in dirs if d not in ('node_modules', 'build', 'dist', '.git', 'out')]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n # 忽略完全空的文件\\n if not text.strip():\\n continue\\n\\n code_files_found = True\\n comment_lines = set()\\n lines = text.split('\\\\n')\\n \\n # 状态机标志位\\n in_multiline_comment = False\\n in_html_comment = False\\n in_string = False\\n string_char = ''\\n \\n for i, line in enumerate(lines):\\n j = 0\\n in_singleline_comment = False\\n \\n while j < len(line):\\n if in_multiline_comment:\\n comment_lines.add(i)\\n if line[j:j+2] == '*/':\\n in_multiline_comment = False\\n j += 2\\n else:\\n j += 1\\n continue\\n \\n if in_html_comment:\\n comment_lines.add(i)\\n if line[j:j+3] == '-->':\\n in_html_comment = False\\n j += 3\\n else:\\n j += 1\\n continue\\n \\n if in_singleline_comment:\\n comment_lines.add(i)\\n break\\n \\n if in_string:\\n if line[j] == '\\\\\\\\':\\n j += 2\\n continue\\n if line[j] == string_char:\\n in_string = False\\n j += 1\\n continue\\n \\n # 检查当前字符及其组合\\n if ext in {'.js', '.jsx', '.ts', '.tsx'}:\\n if line[j] in '\\\"\\\\'`':\\n in_string = True\\n string_char = line[j]\\n j += 1\\n continue\\n if line[j:j+2] == '//':\\n in_singleline_comment = True\\n comment_lines.add(i)\\n j += 2\\n continue\\n if line[j:j+2] == '/*':\\n in_multiline_comment = True\\n comment_lines.add(i)\\n j += 2\\n continue\\n \\n elif ext == '.css':\\n if line[j] in '\\\"\\\\'':\\n in_string = True\\n string_char = line[j]\\n j += 1\\n continue\\n if line[j:j+2] == '/*':\\n in_multiline_comment = True\\n comment_lines.add(i)\\n j += 2\\n continue\\n \\n elif ext == '.html':\\n if line[j] in '\\\"\\\\'':\\n in_string = True\\n string_char = line[j]\\n j += 1\\n continue\\n if line[j:j+4] == '<!--':\\n in_html_comment = True\\n comment_lines.add(i)\\n j += 4\\n continue\\n \\n j += 1\\n \\n # 检查该代码文件中的注释总行数是否严格等于 3\\n if len(comment_lines) != 3:\\n return False\\n\\n return code_files_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对React项目的常见代码文件扩展名\\n exts = ('.js', '.jsx', '.ts', '.tsx')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n # 1. 移除单行和多行注释\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n # 2. 移除单引号和双引号包裹的字符串\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", code)\\n # 移除不含表达式插值 (${}) 的普通模板字符串\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\$])*`', '``', code)\\n \\n # 3. 移除包含问号的正则表达式字面量，防止误报\\n code = re.sub(r'/(?![/*])[^/\\\\n\\\\\\\\]*\\\\?[^/\\\\n\\\\\\\\]*/[gimsuy]*', '', code)\\n \\n # 4. 移除JSX中的普通文本部分 (匹配不在大括号内的标签间文本)\\n # 纯文本通常位于 > 或 } 之后，且在 < 或 { 之前\\n code = re.sub(r'(?<=[>}])([^<>{}]*)(?=[<{])', '', code)\\n \\n # 5. 替换掉JS/TS中合法的非三元运算符问号组合\\n code = code.replace('?.', '') # 可选链\\n code = code.replace('??', '') # 空值合并运算符\\n code = re.sub(r'\\\\?\\\\s*:', '', code) # TS可选属性/参数\\n \\n # 6. 查找是否存在真正的三元运算符特征\\n # 匹配 '?' 和后续的 ':'，且判定这段区间内不存在语句结束符 ';'\\n matches = re.finditer(r'\\\\?(.*?):', code, re.DOTALL)\\n for match in matches:\\n middle = match.group(1)\\n # 三元运算符作为表达式内部通常不存在分号，以此进行安全确认\\n if ';' not in middle:\\n return False\\n except Exception:\\n pass\\n \\n # 如果所有文件都没有命中三元运算符，代表满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 根据React项目推断可能的代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 兼容不同系统的换行符，并统一为 \\\\n 进行分割\\n content = content.replace('\\\\r\\\\n', '\\\\n')\\n # split('\\\\n') 能够准确计算行数，包括文件末尾或中间的空行\\n lines_count = len(content.split('\\\\n'))\\n except Exception:\\n # 如果有文件读取异常直接跳过\\n continue\\n \\n has_code_file = True\\n # 若存在任何一个代码文件的行数不是质数，则不满足要求\\n if not is_prime(lines_count):\\n return False\\n \\n # 如果所有的代码文件行数均为质数且至少验证了一个代码文件，则判定为通过\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步加入预算预警系统。在界面顶部的总开销卡片旁边，增加一个“设置月度预算”的功能（支持输入数字并保存到LocalStorage）。设置好预算后，下方需要显示一个预算使用进度的进度条。规则如下：当总开销低于预算的80%时，进度条为绿色；当达到预算的80%但低于100%时，进度条变成黄色，并在旁边显示“即将超支”的警告文本；如果达到或超过预算的100%，进度条变成红色，并且在每次添加新记录导致超标时，在屏幕中央弹出一个醒目的“预算已超标”警告弹窗（需要手动关闭）。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用中文的规则作废，更新为你生成的每一份代码文件中的注释全都是英文；其二，新增要求：运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 匹配字符串（单、双引号和反引号，防止字符串内的符合被误认为注释）以及各类注释\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<ml_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n r'|(?P<sl_comment>//[^\\\\n]*)'\\n r'|(?P<html_comment><!--[\\\\s\\\\S]*?-->)'\\n )\\n\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html'}\\n found_any_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的依赖、Git和编译目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'build' in dirs:\\n dirs.remove('build')\\n if 'dist' in dirs:\\n dirs.remove('dist')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n found_any_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 记录所有换行符后面的索引，作为每一行的起始位置\\n line_starts = [0]\\n for m in re.finditer(r'\\\\n', content):\\n line_starts.append(m.start() + 1)\\n \\n def get_line_number(index: int) -> int:\\n # 返回基于 1 的行号\\n return bisect.bisect_right(line_starts, index)\\n\\n comment_lines = set()\\n for m in pattern.finditer(content):\\n # 仅当匹配到的内容属于注释组时才处理其行号\\n if m.lastgroup in ('ml_comment', 'sl_comment', 'html_comment'):\\n start_line = get_line_number(m.start())\\n # 如果结束索引大于开始索引，取结束的前一个字符所在的行；否则按原样\\n end_index = m.end() - 1 if m.end() > m.start() else m.start()\\n end_line = get_line_number(end_index)\\n \\n # 将从起始行到结束行之间包含注释的行号加入集合\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n \\n # 如果有任何一个代码文件中的注释占用行数不是精确的 3 行，即不符合要求\\n if len(comment_lines) != 3:\\n return False\\n\\n return found_any_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def verify_no_ternary_operator(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 替换函数：将注释直接去除，字符串字面量仅保留引号外壳防止其内部干扰，同时也能避免破坏代码结构导致漏判\\n def replacer(match):\\n text = match.group(0)\\n if text.startswith('//') or text.startswith('/*'):\\n return ''\\n elif text.startswith('\\\"'):\\n return '\\\"\\\"'\\n elif text.startswith(\\\"'\\\"):\\n return \\\"''\\\"\\n elif text.startswith('`'):\\n return '``'\\n return text\\n \\n # 正则表达式匹配：单行注释 | 多行注释 | 单引号字符串 | 双引号字符串 | 反引号字符串\\n pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|' +\\n r'\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|' +\\n r'\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|' +\\n r'`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.DOTALL | re.MULTILINE\\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 剔除干扰项\\n clean_code = pattern.sub(replacer, code)\\n length = len(clean_code)\\n \\n i = 0\\n while i < length:\\n if clean_code[i] == '?':\\n # 排除 ?? (空值合并)\\n if i > 0 and clean_code[i-1] == '?':\\n i += 1\\n continue\\n if i < length - 1 and clean_code[i+1] == '?':\\n i += 2\\n continue\\n \\n # 排除 ?. (可选链)\\n if i < length - 1 and clean_code[i+1] == '.':\\n i += 2\\n continue\\n \\n # 排除 ?: (TypeScript可选属性/参数等，其中?和:之间仅包含空白符的情况)\\n j = i + 1\\n while j < length and clean_code[j].isspace():\\n j += 1\\n if j < length and clean_code[j] == ':':\\n i = j + 1\\n continue\\n \\n # 进行候选三元运算符的安全过滤（典型的三元运算符两侧常常带有空格或位于闭合括号后）\\n prev_char = clean_code[i-1] if i > 0 else ''\\n next_char = clean_code[i+1] if i < length - 1 else ''\\n \\n is_ternary_candidate = (\\n prev_char.isspace() or \\n next_char.isspace() or \\n prev_char in ')]}'\\n )\\n \\n if is_ternary_candidate:\\n # 动态作用域解析：向后寻找同一作用域层级的匹配冒号\\n depth = 0\\n k = j\\n has_colon = False\\n \\n while k < length:\\n char = clean_code[k]\\n if char in '({[':\\n depth += 1\\n elif char in ')}]':\\n depth -= 1\\n if depth < 0: # 跳出了当前所在的作用域层级\\n break\\n elif char == ';' and depth == 0: # 当前语句结束\\n break\\n elif char == ':' and depth == 0: # 在同等层级找到了对应的冒号\\n has_colon = True\\n break\\n k += 1\\n \\n # 一旦找到符合结构定义的三元配对，证明其使用了三元运算符，直接判负\\n if has_colon:\\n return False\\n i += 1\\n \\n # 如果所有的 JS/TS 代码文件均未发现三元运算符特征，则认为满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义React项目通常包含的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html'}\\n \\n # 应当忽略的常见目录（构建产物、依赖包、隐藏目录等）\\n ignore_dirs = {'node_modules', 'build', 'dist', 'public', '.git'}\\n\\n has_code_file = False\\n \\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除忽略目录及隐藏目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs and not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n num_lines = len(lines)\\n \\n # 检查总行数是否为质数\\n if not is_prime(num_lines):\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果无法使用utf-8读取文件等异常情况，不作为主要检查依据\\n pass\\n \\n # 如果找到了合法的代码文件且均符合质数要求，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的每一份代码文件中的注释全都是英文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在来完善右上角的“数据统计”区块，让数据可视化。请引入图表库（如Recharts），在这里实现两个图表：左边是一个按分类统计开销占比的饼图（带有图例，鼠标悬浮能显示具体金额），右边是一个展示过去7天每天开销金额趋势的柱状图或折线图。这两个图表的数据必须和下方开销列表的数据完全联动，当我添加或删除记录时，图表要自动重新渲染计算。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用英语的规则作废，更新为确保每一个生成的代码文件里，注释部分完全使用中文；其二，新增要求：请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"import os\\n\\ndef check_comments_in_code_files(response: str, workspace_path: str) -> bool:\\n # 定义 React 项目中常见的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css'}\\n html_extensions = {'.html'}\\n \\n # 忽略非直接生成的系统、依赖或构建目录\\n exclude_dirs = {'.git', 'node_modules', 'build', 'dist', '.next', 'coverage'}\\n \\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉排除的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions or ext in html_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n code_files_found += 1\\n \\n # 针对 JS/TS/CSS 等样式或脚本文件解析注释\\n if ext in code_extensions:\\n lines_with_comments = set()\\n i = 0\\n in_string = False\\n string_char = ''\\n in_sl_comment = False\\n in_ml_comment = False\\n escaped = False\\n line_num = 1\\n \\n while i < len(code):\\n # 处理换行符\\n if code[i] == '\\\\n':\\n if in_sl_comment:\\n in_sl_comment = False\\n line_num += 1\\n i += 1\\n escaped = False\\n continue\\n \\n # 如果当前字符是被转义的，则直接跳过\\n if escaped:\\n escaped = False\\n i += 1\\n continue\\n \\n # 如果在单行注释中，该行记录为含有注释，并继续前进\\n if in_sl_comment:\\n lines_with_comments.add(line_num)\\n i += 1\\n continue\\n \\n # 如果在多行注释中\\n if in_ml_comment:\\n lines_with_comments.add(line_num)\\n # 检测多行注释结束\\n if code[i:i+2] == '*/':\\n in_ml_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n # 如果在字符串常量中\\n if in_string:\\n if code[i] == '\\\\\\\\':\\n escaped = True\\n i += 1\\n elif code[i] == string_char:\\n in_string = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n # 如果不在字符串，也不在注释中\\n if code[i] in ['\\\"', \\\"'\\\", '`']:\\n in_string = True\\n string_char = code[i]\\n i += 1\\n continue\\n \\n # 检测单行注释起步 (CSS 不支持单行注释)\\n if ext != '.css' and code[i:i+2] == '//':\\n in_sl_comment = True\\n lines_with_comments.add(line_num)\\n i += 2\\n continue\\n \\n # 检测多行注释起步\\n if code[i:i+2] == '/*':\\n in_ml_comment = True\\n lines_with_comments.add(line_num)\\n i += 2\\n continue\\n \\n i += 1\\n \\n # 检查此文件的注释行数是否严格为 3\\n if len(lines_with_comments) != 3:\\n return False\\n \\n # 针对 HTML 文件的解析注释\\n elif ext in html_extensions:\\n lines_with_comments = set()\\n i = 0\\n in_comment = False\\n line_num = 1\\n \\n while i < len(code):\\n if code[i] == '\\\\n':\\n line_num += 1\\n i += 1\\n continue\\n \\n if in_comment:\\n lines_with_comments.add(line_num)\\n if code[i:i+3] == '-->':\\n in_comment = False\\n i += 3\\n else:\\n i += 1\\n continue\\n \\n if code[i:i+4] == '<!--':\\n in_comment = True\\n lines_with_comments.add(line_num)\\n i += 4\\n continue\\n \\n i += 1\\n \\n if len(lines_with_comments) != 3:\\n return False\\n\\n # 若没有任何代码文件，也视作要求未完成\\n if code_files_found == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def clean_code(code: str) -> str:\\n # 正则表达式匹配并移除：\\n # 1. HTML/XML 多行注释\\n # 2. JS/TS/CSS 多行注释\\n # 3. JS/TS 单行注释\\n # 4. 双引号字符串\\n # 5. 单引号字符串\\n # 6. 反引号（模板）字符串\\n pattern = re.compile(\\n r'<!--[\\\\s\\\\S]*?-->|'\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|'\\n r'//.*|'\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|'\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\"\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n )\\n return pattern.sub('', code)\\n\\n def contains_ternary(code: str) -> bool:\\n cleaned_code = clean_code(code)\\n \\n # 寻找代码中所有的 ?\\n for match in re.finditer(r'\\\\?', cleaned_code):\\n start = match.end()\\n # 找到 ? 后方剩余的代码片段，并去除开头的空白符\\n after_q = cleaned_code[start:].lstrip()\\n \\n # 排除 ?. (可选链)\\n # 排除 ?? (空值合并运算符)\\n # 排除 ?: (TypeScript中的可选参数或可选属性声明)\\n if after_q.startswith('.') or after_q.startswith('?') or after_q.startswith(':'):\\n continue\\n \\n # 如果不属于上述情况，说明这是一个用于条件操作的单独 ? 符号。\\n # 只要其后方还存在配套的 : ，就确认代码中使用了三元运算符。\\n if ':' in after_q:\\n return True\\n \\n return False\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 可能涉及到的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.htm'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n # 如果有文件读取发生异常则跳过\\n continue\\n \\n if contains_ternary(code):\\n # 若在任何一个文件中发现了三元运算符，则判断未满足要求，返回 False\\n return False\\n\\n # 所有生成的文件验证完毕，没有发现任何形式的三元运算符\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个数字是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 根据React应用推断出的常见代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n code_files_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉无关的第三方库和构建产物目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'build', 'dist', 'public')]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 兼容处理 Windows (CRLF) 与 Unix (LF) 的换行符，并严格切分获取所有行（含空行）\\n lines = content.replace('\\\\r\\\\n', '\\\\n').split('\\\\n')\\n line_count = len(lines)\\n \\n # 如果行数不是质数，则要求未满足\\n if not is_prime(line_count):\\n return False\\n except Exception:\\n # 无法读取该文件，视同验证未通过\\n return False\\n\\n # 若至少找到了一个代码文件，并且全部通过了检验，则返回True\\n return code_files_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"为了方便查看历史数据，我们需要给列表加上高级筛选和排序功能。在右侧开销列表的正上方增加一个工具栏：包含一个“时间段”下拉框（本周、本月、所有）、一个“分类”多选框（勾选后只显示选中的分类），以及一个“排序规则”下拉框（按日期最新、按日期最旧、按金额最高、按金额最低）。列表要根据这些条件实时更新。注意：筛选不仅要改变列表里展示的内容，还要同步让上方数据统计的图表和总开销、预算进度条都基于筛选后的数据进行计算和展示。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用中文的规则作废，更新为你所输出的任何一个代码文件，其包含的注释都应当是全英文的；其二，新增要求：请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files = []\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉常见的依赖包、构建及版本控制相关目录\\n for exclude_dir in ['node_modules', '.git', 'build', 'dist']:\\n if exclude_dir in dirs:\\n dirs.remove(exclude_dir)\\n \\n for f in files:\\n ext = os.path.splitext(f)[1].lower()\\n if ext in code_exts:\\n code_files.append(os.path.join(root, f))\\n \\n # 如果没有找到任何相关的代码文件，也代表未按要求成功生成文件\\n if not code_files:\\n return False\\n \\n for filepath in code_files:\\n ext = os.path.splitext(filepath)[1].lower()\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n lines = content.split('\\\\n')\\n comment_lines = set()\\n \\n in_block_comment = False\\n in_string = False\\n string_char = ''\\n \\n for i, line in enumerate(lines):\\n j = 0\\n while j < len(line):\\n # 处于块注释中\\n if in_block_comment:\\n if ext == '.html':\\n if line[j:j+3] == '-->':\\n in_block_comment = False\\n j += 2\\n else:\\n if line[j:j+2] == '*/':\\n in_block_comment = False\\n j += 1\\n comment_lines.add(i)\\n j += 1\\n continue\\n \\n # 处于字符串字面量中\\n if in_string:\\n if line[j] == '\\\\\\\\':\\n j += 1\\n elif line[j] == string_char:\\n in_string = False\\n j += 1\\n continue\\n \\n # 处于正常代码解析状态\\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if line[j:j+2] == '//':\\n comment_lines.add(i)\\n break\\n elif line[j:j+2] == '/*':\\n in_block_comment = True\\n comment_lines.add(i)\\n j += 1\\n elif line[j] in [\\\"'\\\", '\\\"', '`']:\\n in_string = True\\n string_char = line[j]\\n \\n elif ext in ['.css']:\\n if line[j:j+2] == '/*':\\n in_block_comment = True\\n comment_lines.add(i)\\n j += 1\\n elif line[j] in [\\\"'\\\", '\\\"']:\\n in_string = True\\n string_char = line[j]\\n \\n elif ext in ['.html']:\\n if line[j:j+4] == '<!--':\\n in_block_comment = True\\n comment_lines.add(i)\\n j += 3\\n elif line[j] in [\\\"'\\\", '\\\"']:\\n in_string = True\\n string_char = line[j]\\n \\n elif ext in ['.py']:\\n if line[j] == '#':\\n comment_lines.add(i)\\n break\\n elif line[j] in [\\\"'\\\", '\\\"']:\\n in_string = True\\n string_char = line[j]\\n \\n j += 1\\n \\n # 若行结束时依然处于块注释内部，则当前行也应作为注释行计算\\n if in_block_comment:\\n comment_lines.add(i)\\n \\n # 验证每一个代码文件的注释行数是否精确为3行\\n if len(comment_lines) != 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def strip_comments_and_strings(code: str) -> str:\\n \\\"\\\"\\\"\\n 一个轻量级的词法状态机，用来剔除 JavaScript/TypeScript 中的:\\n 字符串 (\\\", ', `)、注释 (//, /* */) 以及 正则表达式字面量 (/.../) 内部的内容\\n 以防止这些内容里的 '?' 造成误报。\\n \\\"\\\"\\\"\\n i = 0\\n n = len(code)\\n \\n NORMAL = 0\\n STR_DOUBLE = 1\\n STR_SINGLE = 2\\n STR_TEMPLATE = 3\\n BLOCK_COMMENT = 4\\n LINE_COMMENT = 5\\n REGEX = 6\\n \\n state = NORMAL\\n state_stack = [] \\n last_token = ''\\n current_word = ''\\n \\n res = []\\n \\n kw = {'return', 'yield', 'typeof', 'await', 'throw', 'case', 'in', 'instanceof', 'do', 'else', 'void', 'delete', 'new'}\\n regex_preceding_tokens = {'', '=', '(', ',', '[', '{', ':', ';', '!', '&', '|', '?', '<', '>', '+', '-', '*', '/', '%'} | kw\\n \\n while i < n:\\n char = code[i]\\n \\n if state == NORMAL:\\n is_ident_char = char.isalnum() or char == '_' or char == '$'\\n if is_ident_char:\\n current_word += char\\n res.append(char)\\n i += 1\\n continue\\n else:\\n if current_word:\\n if current_word in kw:\\n last_token = current_word\\n else:\\n last_token = 'IDENT'\\n current_word = ''\\n \\n if char.isspace():\\n res.append(char)\\n i += 1\\n continue\\n \\n if char == '/' and i + 1 < n and code[i+1] == '*':\\n state = BLOCK_COMMENT\\n i += 2\\n continue\\n elif char == '/' and i + 1 < n and code[i+1] == '/':\\n state = LINE_COMMENT\\n i += 2\\n continue\\n elif char == '\\\"':\\n state = STR_DOUBLE\\n i += 1\\n last_token = 'STRING'\\n continue\\n elif char == \\\"'\\\":\\n state = STR_SINGLE\\n i += 1\\n last_token = 'STRING'\\n continue\\n elif char == '`':\\n state = STR_TEMPLATE\\n i += 1\\n last_token = 'STRING'\\n continue\\n elif char == '{':\\n state_stack.append(NORMAL)\\n last_token = '{'\\n res.append(char)\\n i += 1\\n continue\\n elif char == '}':\\n last_token = '}'\\n if state_stack:\\n prev_state = state_stack.pop()\\n if prev_state == STR_TEMPLATE:\\n state = STR_TEMPLATE\\n res.append(char)\\n i += 1\\n continue\\n elif char == '/':\\n # 利用启发式前置Token来区分 正则字面量 和 除号\\n if last_token in regex_preceding_tokens:\\n state = REGEX\\n i += 1\\n continue\\n else:\\n last_token = '/'\\n res.append(char)\\n i += 1\\n continue\\n else:\\n last_token = char\\n res.append(char)\\n i += 1\\n continue\\n \\n elif state == STR_DOUBLE:\\n if char == '\\\\\\\\': i += 2\\n elif char == '\\\"': state = NORMAL; i += 1\\n else: i += 1\\n \\n elif state == STR_SINGLE:\\n if char == '\\\\\\\\': i += 2\\n elif char == \\\"'\\\": state = NORMAL; i += 1\\n else: i += 1\\n \\n elif state == STR_TEMPLATE:\\n if char == '\\\\\\\\': i += 2\\n elif char == '`': state = NORMAL; i += 1\\n elif char == '$' and i + 1 < n and code[i+1] == '{':\\n state_stack.append(STR_TEMPLATE)\\n state = NORMAL\\n last_token = '{'\\n res.append('$')\\n res.append('{')\\n i += 2\\n else: i += 1\\n \\n elif state == BLOCK_COMMENT:\\n if char == '*' and i + 1 < n and code[i+1] == '/': state = NORMAL; i += 2\\n else: i += 1\\n \\n elif state == LINE_COMMENT:\\n if char == '\\\\n' or char == '\\\\r':\\n state = NORMAL\\n res.append(char)\\n i += 1\\n else: i += 1\\n \\n elif state == REGEX:\\n if char == '\\\\\\\\': i += 2\\n elif char == '/':\\n state = NORMAL\\n last_token = 'REGEX'\\n i += 1\\n else: i += 1\\n\\n return \\\"\\\".join(res)\\n\\n def has_ternary_operator(code: str) -> bool:\\n clean_code = strip_comments_and_strings(code)\\n \\n # 1. 过滤 JS/TS 其他包含 '?' 的常见特性\\n clean_code = clean_code.replace('?.', '')\\n clean_code = clean_code.replace('??=', '')\\n clean_code = clean_code.replace('??', '')\\n \\n # 2. 过滤 TypeScript 的可选属性/参数标记, 如 prop?: type \\n clean_code = re.sub(r'\\\\?\\\\s*:', '', clean_code)\\n \\n # 3. 过滤 TS 其他位置的可选标记, 如 [number?], (name?), name?, foo?; 等\\n clean_code = re.sub(r'\\\\?\\\\s*([\\\\]\\\\),;])', r'\\\\1', clean_code)\\n \\n # 4. 过滤 JSX 的纯文本中的孤立 '?' (如 <div>Are you sure?</div>)\\n # 匹配位于标签边界 > 和 < 之间，不包含任何逻辑块符号 {}, 以及避免匹配三元运算符所需 : 的纯显示文本片段\\n prev_code = None\\n while prev_code != clean_code:\\n prev_code = clean_code\\n clean_code = re.sub(r'(>)([^<{>:]*?)\\\\?([^<{>:]*?)(<)', r'\\\\1\\\\2\\\\3\\\\4', clean_code)\\n \\n # 5. 在经过重重干扰排除后的纯净代码骨架中，若依旧包含游离的问号，即必然为三元运算符\\n return '?' in clean_code\\n\\n # 扫描给定的项目目录下所有的前端源文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n if has_ternary_operator(code):\\n # 如果存在任意一处三元运算符，则立即判定不满足要求\\n return False\\n except Exception:\\n continue\\n\\n # 扫描完所有相关代码文件均未发现三元运算符结构，满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 针对React项目，定义代码文件扩展名白名单\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html'}\\n # 需忽略的依赖目录和构建产物目录\\n exclude_dirs = {'node_modules', '.git', 'build', 'dist'}\\n \\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n num_lines = len(lines)\\n \\n # 检查总行数是否为质数\\n if not is_prime(num_lines):\\n return False\\n except Exception:\\n # 如果遇到不可读或非纯文本代码文件，按不满足要求处理\\n return False\\n \\n # 如果项目中没有生成任何匹配的代码文件，返回False\\n return code_files_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"最后，我们来增加数据导出和批量管理功能。在刚刚添加的筛选工具栏的最右侧，增加一个“导出 CSV”的按钮，点击后能把当前列表里正在展示的开销记录（即经过筛选后的数据）下载为一个包含表头（标题、分类、金额、日期）的CSV文件。同时，给列表里的每一行最左侧加一个复选框，并在列表底部提供一个“批量删除选中项”的按钮，只有当有记录被选中时该按钮才可用，点击并二次确认后，可以一次性删除所有被选中的记录。另外，后续所有相关开发工作都必须始终遵循以下更新和新增要求：其一，此前要求的注释统一使用英文的规则作废，更新为你生成的各个代码文件中，所有的注释语句只能是中文；其二，新增要求：所有由你创建的代码文件，其编码必须统一为UTF-8。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你必须先使用相应的工具阅读已有文件，然后才能对其进行修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在项目每次被修改后，都创建一份全新的文件 change_log_X.log 来写明此次改动详情，X 代表从 0 开始编号的修改次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的每个代码文件中，注释的行数既不能多也不能少，必须精确为 3 行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 联合匹配模式：同时匹配注释和字符串字面量，防止字符串内部的伪注释导致误判\\n # 采用 named groups 便于后续分辨匹配到的是注释还是普通字符串\\n pattern = re.compile(\\n r'(?P<line_comment>//[^\\\\n]*)' # JS/TS/CSS 等单行注释\\n r'|(?P<hash_comment>#[^\\\\n]*)' # Python/Shell/YAML 等单行注释\\n r'|(?P<block_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)' # 多行块注释\\n r'|(?P<html_comment><!--[\\\\s\\\\S]*?-->)' # HTML 注释\\n r'|(?P<d_quote>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")' # 双引号字符串\\n r'|(?P<s_quote>\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')' # 单引号字符串\\n r'|(?P<t_quote>`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' # 反引号模板字符串\\n )\\n\\n # 常见的前端与通用代码文件后缀\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.css', '.html', '.py', '.sh'}\\n \\n # 忽略的构建或者环境依赖目录\\n ignore_dirs = {'node_modules', 'build', 'dist', '.git', '__pycache__', 'coverage'}\\n\\n file_count = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除忽略目录，防止进入深层查找\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n file_count += 1\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n # 如果遇到无法使用 utf-8 读取的文件，从严判定为未满足条件\\n return False\\n \\n comment_lines = set()\\n \\n # 查找所有的匹配项\\n for match in pattern.finditer(code):\\n # 如果匹配到的内容属于注释的分组之一\\n if match.lastgroup in ['line_comment', 'hash_comment', 'block_comment', 'html_comment']:\\n start = match.start()\\n end = match.end()\\n \\n # 计算开始和结束行（利用起始点前面的换行符数量 + 1）\\n start_line = code.count('\\\\n', 0, start) + 1\\n end_line = code.count('\\\\n', 0, max(start, end - 1)) + 1\\n \\n # 把跨越的每一行存入集合\\n for line in range(start_line, end_line + 1):\\n comment_lines.add(line)\\n \\n # 校验：注释行数必须精确为 3 行\\n if len(comment_lines) != 3:\\n return False\\n\\n # 若没有任何代码文件，通常也认为没有真正完成任务，返回 False\\n return file_count > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查生成的代码文件中是否没有使用任何形式的三元运算符。\\n 满足要求（没有使用）返回 True，否则返回 False。\\n \\\"\\\"\\\"\\n # 针对 React 项目，只检查 JS/TS 及其 JSX/TSX 相关的源代码文件\\n valid_extensions = {'.js', '.jsx', '.ts', '.tsx'}\\n \\n # 匹配三元运算符的正则表达式\\n # 问号后不能紧跟 . (可选链) 或 ? (空值合并)\\n # ? 和 : 之间不能存在分号 ;，且尽量非贪婪匹配以应对多个三元运算\\n ternary_pattern = re.compile(r'\\\\?(?![.?])([^;]*?):')\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n # 忽略无法读取的文件\\n continue\\n \\n # 1. 移除单行注释\\n code = re.sub(r'//.*', '', code)\\n # 2. 移除多行注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 3. 移除单引号字符串\\n code = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", code)\\n # 4. 移除双引号字符串\\n code = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', code)\\n # 5. 移除模板字符串\\n code = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', '``', code)\\n \\n # 在清理后的代码中查找类似三元运算符的结构\\n matches = ternary_pattern.finditer(code)\\n for match in matches:\\n content_between = match.group(1).strip()\\n \\n # 排除 TypeScript 中类似 `name?: string` 的可选类型声明（其中间往往不包含非空字符）\\n if content_between != '':\\n # 为了防止某些极限跨行情况捕获到不该捕获的内容，排查一下 JS 核心的声明关键字\\n if not re.search(r'\\\\b(let|const|var|import|export)\\\\b', content_between):\\n # 如果存在满足条件的情况，说明找到了三元运算符，违反了要求\\n return False\\n\\n # 所有文件验证完毕，均未发现三元运算符，完全满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你所生成的各个代码文件，其总行数（将空行计算在内）务必是质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 根据React应用特性，定义常见的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.json'}\\n \\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的文件夹\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用 split('\\\\n') 能最准确地反映包含所有空行（包括文件末尾）的行数\\n total_lines = len(content.split('\\\\n'))\\n \\n if not is_prime(total_lines):\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果发生文件读取异常（如意外的二进制文件等），为了稳妥起见忽略或可认为不符\\n pass\\n \\n # 如果项目里面没有识别出任何代码文件，则不符合“生成的各个代码文件总行数为质数”的条件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你给出的代码中，任何外部 I/O 操作、网络请求及数据格式转换的环节都必须被异常捕获功能包裹，并在发生异常时返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你生成的各个代码文件中，所有的注释语句只能是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"代码生成需满足：只做当前明确要求的事，不作过度设计与未来特性的铺垫，且必须清除掉没有用到的代码和依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"写完代码后，必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 包含React项目相关的常见代码文件和配置文件的后缀，以及日志和Markdown文件\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.html', '.json', '.md', '.log'}\\n checked_files_count = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略无需验证的依赖库或编译输出文件夹\\n ignore_dirs = {'node_modules', '.git', 'build', 'dist'}\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以UTF-8编码读取文件，若文件包含非UTF-8字符将抛出UnicodeDecodeError\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n checked_files_count += 1\\n except UnicodeDecodeError:\\n # 发现非UTF-8编码文件\\n return False\\n except Exception:\\n # 忽略诸如权限不足等其他无法读取的情况\\n pass\\n \\n # 只要存在相关的代码文件且均无引发解码错误，说明都满足UTF-8编码要求\\n return checked_files_count > 0\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["调整浏览器窗口尺寸。判断大屏幕下页面主体是否分为左右两栏布局（左侧表单，右侧数据和列表），缩小屏幕宽度至较小尺寸，判断左右布局是否自动变为上下堆叠的响应式布局。","检查页面顶部是否渲染了包含该应用名称的顶部导航栏，且整体样式设计有留白，显示现代简洁。","在左侧表单填写合规的标题、金额、分类和日期，点击“添加”按钮。检查左侧表单区域内的所有输入框内容是否被成功清空。","在左侧表单填写合规的数据并点击“添加”按钮。检查右侧开销列表中是否成功新增了一行刚才输入的记录，且该行记录中标题、金额、分类、日期四个字段的信息能够被清晰完整地展示。","在表单中添加一条新的有效开销记录，检查列表顶部的“总开销”卡片金额数值是否自动发生了更新，并且正确累加了刚刚填写的金额数字。","在开销列表中找到任意一条存在的记录，点击该行末尾的“删除”按钮。检查该条记录是否从列表中彻底消失，同时顶部的“总开销”卡片总金额同步减去了该笔费用。","在页面中添加一条合规开销记录并设定月度预算数字，按F5刷新浏览器页面。检查刷新完成后，开销列表中的记录、顶部总金额以及设定的预算是否依然保留在界面上，未发生数据丢失。","清空表单的“标题”输入框，并将鼠标点击页面其他区域触发失焦。检查标题输入框下方是否出现红色的错误提示文字，且下方的“添加”按钮是否变为禁用状态无法点击。","在表单的“金额”输入框中输入一个负数（如 -5）或0或含有三位小数的数字（如 10.333）。检查金额输入框下方是否弹出红色的错误提示，且下方的“添加”按钮呈现禁用状态。","在表单的“日期”选择器中选择一个晚于今天的未来日期。检查日期组件下方是否出现红色的错误预警提示，并且下方的“添加”按钮变为禁用状态。","在界面顶部总开销卡片旁找到“设置月度预算”功能，输入一个远大于当前已产生总开销金额的数字并保存。检查下方显示的预算使用进度条是否呈现绿色。","调整月度预算金额，使得当前总开销达到预算的80%但低于100%。检查预算使用进度条的颜色是否变为黄色，并且在进度条旁边是否清晰显示了“即将超支”的警告提示文本。","进一步调低月度预算金额，使得当前总开销大于或等于设定的预算数值。检查下方预算使用进度条是否相应地变为了红色。","在当前开销尚未超标的情况下，通过表单新添加一条大额记录，使得新总额达到或超过设定预算。检查屏幕中央是否立刻弹出一个带有“预算已超标”字样的醒目警告弹窗，且该弹窗支持用户手动点击关闭。","观察右上角数据统计区块的饼图，检查该饼图旁边是否有分类颜色的图例说明；将鼠标悬浮在饼图的某个扇形区域，检查是否弹出了能够展示该分类具体对应开销金额的提示浮窗。","观察右上角展示过去7天开销趋势的柱状图或折线图。在左侧表单新添加一条今日大额开销记录，检查该趋势图中代表今天的柱子高度或折线端点是否发生相应的升高变化。","在开销列表正上方找到筛选工具栏，将时间段下拉框选择为“本周”。检查开销列表里的内容是否实时刷新，过滤掉了历史时间数据，仅展示属于本周范围内的记录。","在筛选工具栏中，找到分类多选框并仅勾选特定一个分类（例如“餐饮”）。检查下方的开销列表是否实时刷新，仅保留和展示分类匹配为“餐饮”的开销记录。","在筛选工具栏的排序规则下拉框中，选择“按金额最高”。检查开销列表中的各行记录是否自动按照金额数值从大到小重新排列渲染。","使用时间筛选或分类筛选隐藏掉部分开销数据，观察顶层数据变化。检查“总开销”显示的金额与右上角的图表，是否都基于筛选后当前可见的列表记录重新计算且同步刷新了渲染状态。","对列表进行任意条件的筛选过滤后，点击筛选栏最右侧的“导出 CSV”按钮。检查浏览器是否执行了CSV文件下载操作，且使用表格软件打开该文件，检查首行是否包含“标题、分类、金额、日期”四个表头，同时正文数据行数是否与刚才页面上筛选显示的记录完全一致。","检查开销列表中的每一行记录最左侧是否均渲染了一个复选框。确保在没有勾选任何一行复选框的情况下，列表底部的“批量删除选中项”按钮处于置灰（不可点击状态）。","勾选开销列表中两行记录最左侧的复选框，观察列表底部的“批量删除选中项”按钮变为可用状态。点击该按钮并在弹出的二次确认提示框中点击确认，检查被勾选的两行记录是否从列表中一次性删除消失。"],"string":"[\n \"调整浏览器窗口尺寸。判断大屏幕下页面主体是否分为左右两栏布局（左侧表单，右侧数据和列表），缩小屏幕宽度至较小尺寸，判断左右布局是否自动变为上下堆叠的响应式布局。\",\n \"检查页面顶部是否渲染了包含该应用名称的顶部导航栏，且整体样式设计有留白，显示现代简洁。\",\n \"在左侧表单填写合规的标题、金额、分类和日期，点击“添加”按钮。检查左侧表单区域内的所有输入框内容是否被成功清空。\",\n \"在左侧表单填写合规的数据并点击“添加”按钮。检查右侧开销列表中是否成功新增了一行刚才输入的记录，且该行记录中标题、金额、分类、日期四个字段的信息能够被清晰完整地展示。\",\n \"在表单中添加一条新的有效开销记录，检查列表顶部的“总开销”卡片金额数值是否自动发生了更新，并且正确累加了刚刚填写的金额数字。\",\n \"在开销列表中找到任意一条存在的记录，点击该行末尾的“删除”按钮。检查该条记录是否从列表中彻底消失，同时顶部的“总开销”卡片总金额同步减去了该笔费用。\",\n \"在页面中添加一条合规开销记录并设定月度预算数字，按F5刷新浏览器页面。检查刷新完成后，开销列表中的记录、顶部总金额以及设定的预算是否依然保留在界面上，未发生数据丢失。\",\n \"清空表单的“标题”输入框，并将鼠标点击页面其他区域触发失焦。检查标题输入框下方是否出现红色的错误提示文字，且下方的“添加”按钮是否变为禁用状态无法点击。\",\n \"在表单的“金额”输入框中输入一个负数（如 -5）或0或含有三位小数的数字（如 10.333）。检查金额输入框下方是否弹出红色的错误提示，且下方的“添加”按钮呈现禁用状态。\",\n \"在表单的“日期”选择器中选择一个晚于今天的未来日期。检查日期组件下方是否出现红色的错误预警提示，并且下方的“添加”按钮变为禁用状态。\",\n \"在界面顶部总开销卡片旁找到“设置月度预算”功能，输入一个远大于当前已产生总开销金额的数字并保存。检查下方显示的预算使用进度条是否呈现绿色。\",\n \"调整月度预算金额，使得当前总开销达到预算的80%但低于100%。检查预算使用进度条的颜色是否变为黄色，并且在进度条旁边是否清晰显示了“即将超支”的警告提示文本。\",\n \"进一步调低月度预算金额，使得当前总开销大于或等于设定的预算数值。检查下方预算使用进度条是否相应地变为了红色。\",\n \"在当前开销尚未超标的情况下，通过表单新添加一条大额记录，使得新总额达到或超过设定预算。检查屏幕中央是否立刻弹出一个带有“预算已超标”字样的醒目警告弹窗，且该弹窗支持用户手动点击关闭。\",\n \"观察右上角数据统计区块的饼图，检查该饼图旁边是否有分类颜色的图例说明；将鼠标悬浮在饼图的某个扇形区域，检查是否弹出了能够展示该分类具体对应开销金额的提示浮窗。\",\n \"观察右上角展示过去7天开销趋势的柱状图或折线图。在左侧表单新添加一条今日大额开销记录，检查该趋势图中代表今天的柱子高度或折线端点是否发生相应的升高变化。\",\n \"在开销列表正上方找到筛选工具栏，将时间段下拉框选择为“本周”。检查开销列表里的内容是否实时刷新，过滤掉了历史时间数据，仅展示属于本周范围内的记录。\",\n \"在筛选工具栏中，找到分类多选框并仅勾选特定一个分类（例如“餐饮”）。检查下方的开销列表是否实时刷新，仅保留和展示分类匹配为“餐饮”的开销记录。\",\n \"在筛选工具栏的排序规则下拉框中，选择“按金额最高”。检查开销列表中的各行记录是否自动按照金额数值从大到小重新排列渲染。\",\n \"使用时间筛选或分类筛选隐藏掉部分开销数据，观察顶层数据变化。检查“总开销”显示的金额与右上角的图表，是否都基于筛选后当前可见的列表记录重新计算且同步刷新了渲染状态。\",\n \"对列表进行任意条件的筛选过滤后，点击筛选栏最右侧的“导出 CSV”按钮。检查浏览器是否执行了CSV文件下载操作，且使用表格软件打开该文件，检查首行是否包含“标题、分类、金额、日期”四个表头，同时正文数据行数是否与刚才页面上筛选显示的记录完全一致。\",\n \"检查开销列表中的每一行记录最左侧是否均渲染了一个复选框。确保在没有勾选任何一行复选框的情况下，列表底部的“批量删除选中项”按钮处于置灰（不可点击状态）。\",\n \"勾选开销列表中两行记录最左侧的复选框，观察列表底部的“批量删除选中项”按钮变为可用状态。点击该按钮并在弹出的二次确认提示框中点击确认，检查被勾选的两行记录是否从列表中一次性删除消失。\"\n]","isConversational":false}}},{"rowIdx":68,"cells":{"task_id":{"kind":"number","value":69,"string":"69"},"repository_policy":{"kind":"string","value":"后续你处理所有代码相关任务时，必须始终遵循以下所有约束：\n1. 你所输出的任何一个代码文件，其包含的注释都应当是全英文的。\n2. 你输出的所有代码文件，其文件名必须至少包含 5 个字符。\n3. 限制条件：你生成的代码文件中不能使用三元运算符。\n4. 保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准。\n5. 务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息。\n6. 要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print。\n7. 生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n has_code_file = False\n \n # 遍历当前工作空间内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据前序指令中的 Python 专属库推断代码文件类型为 Python 文件\n if file.endswith('.py'):\n has_code_file = True\n # 检查文件名（包含后缀）字符数是否至少为 5\n if len(file) < 5:\n return False\n \n # 如果不存在任何代码文件，也视作未满足生成代码的要求\n return has_code_file","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"def check_no_ternary_operator(response: str, workspace_path: str) -> bool:\n import os\n import ast\n\n # 检查工作区路径是否有效\n if not os.path.isdir(workspace_path):\n return False\n\n has_python_files = False\n\n # 遍历工作区下所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令特征推测代码类型为Python\n if file.endswith(\".py\"):\n has_python_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n code_content = f.read()\n \n # 解析代码文本为抽象语法树(AST)\n tree = ast.parse(code_content)\n \n # 遍历AST的每一个节点\n for node in ast.walk(tree):\n # ast.IfExp 对应于 Python 中的三元运算符 (如 a if condition else b)\n if isinstance(node, ast.IfExp):\n return False\n \n except Exception:\n # 为了增强鲁棒性，若个别文件由于编码或语法错误无法读取与解析，则跳过\n continue\n\n # 如果工作区内完全没有生成任何 Python 文件，视为验证不通过\n if not has_python_files:\n return False\n\n # 遍历了所有 Python 代码文件均未发现三元运算符，满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import ast\n import os\n import re\n\n def is_prime(n: int) -> bool:\n \"\"\"判断一个整数是否为质数\"\"\"\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n def check_code(code: str) -> bool:\n \"\"\"使用AST解析代码，验证函数名称的长度是否符合质数要求\"\"\"\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 函数名属于代码标识符，对其长度（字符数）验证是否为质数\n if not is_prime(len(node.name)):\n return False\n return True\n except Exception:\n # 遇到无法解析的代码块（如语法不完整或语法错误），忽略跳过\n return True\n\n # 1. 检查 workspace_path 下的所有生成的 Python 文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_code(code):\n return False\n except Exception:\n pass\n\n # 2. 检查 response 中的 Python 代码块，作为工作区代码以外的兜底校验\n python_blocks = re.findall(r'```(?:python|py)?\\s*\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n for code in python_blocks:\n if not check_code(code):\n return False\n\n return True","tags":["Style","Naming"]}],"string":"[\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n has_code_file = False\\n \\n # 遍历当前工作空间内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据前序指令中的 Python 专属库推断代码文件类型为 Python 文件\\n if file.endswith('.py'):\\n has_code_file = True\\n # 检查文件名（包含后缀）字符数是否至少为 5\\n if len(file) < 5:\\n return False\\n \\n # 如果不存在任何代码文件，也视作未满足生成代码的要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"def check_no_ternary_operator(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n\\n # 检查工作区路径是否有效\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n has_python_files = False\\n\\n # 遍历工作区下所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令特征推测代码类型为Python\\n if file.endswith(\\\".py\\\"):\\n has_python_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n code_content = f.read()\\n \\n # 解析代码文本为抽象语法树(AST)\\n tree = ast.parse(code_content)\\n \\n # 遍历AST的每一个节点\\n for node in ast.walk(tree):\\n # ast.IfExp 对应于 Python 中的三元运算符 (如 a if condition else b)\\n if isinstance(node, ast.IfExp):\\n return False\\n \\n except Exception:\\n # 为了增强鲁棒性，若个别文件由于编码或语法错误无法读取与解析，则跳过\\n continue\\n\\n # 如果工作区内完全没有生成任何 Python 文件，视为验证不通过\\n if not has_python_files:\\n return False\\n\\n # 遍历了所有 Python 代码文件均未发现三元运算符，满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import ast\\n import os\\n import re\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"判断一个整数是否为质数\\\"\\\"\\\"\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n def check_code(code: str) -> bool:\\n \\\"\\\"\\\"使用AST解析代码，验证函数名称的长度是否符合质数要求\\\"\\\"\\\"\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 函数名属于代码标识符，对其长度（字符数）验证是否为质数\\n if not is_prime(len(node.name)):\\n return False\\n return True\\n except Exception:\\n # 遇到无法解析的代码块（如语法不完整或语法错误），忽略跳过\\n return True\\n\\n # 1. 检查 workspace_path 下的所有生成的 Python 文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_code(code):\\n return False\\n except Exception:\\n pass\\n\\n # 2. 检查 response 中的 Python 代码块，作为工作区代码以外的兜底校验\\n python_blocks = re.findall(r'```(?:python|py)?\\\\s*\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n for code in python_blocks:\\n if not check_code(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Build & Deployment"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"我们要从零搭建一个类似 CodePen 的 Web 代码编辑器。第一步，先来勾勒一下整体页面的基础骨架。请创建一个包含顶部导航和主界面的页面：顶部导航栏左侧放一个应用的 Logo（文字即可），右侧放置几个占位按钮（比如“运行”、“设置”、“分享”）。主界面采用经典的上下布局，上半部分等分为三个区域，分别加上“HTML”、“CSS”、“JS”的标题，并在里面各放一个普通的 <textarea> 作为临时的代码输入框；下半部分占据剩余空间，加上“预览区”的标题和一个空白占位用的 <div>。整体布局使用 Flexbox 或 Grid 撑满全屏，配色采用现代化的深色主题。请注意，后续所有轮次编写代码时都必须始终遵循以下要求：你输出的代码中禁止出现超过 2 层的循环嵌套，同时在生成的代码文件中必须保证所有函数的圈复杂度上限为 4。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx',\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp',\n '.cs', '.go', '.rb', '.php', '.swift', '.kt', '.rs', \n '.json', '.sh', '.bat'\n }\n \n # 检查工作空间路径是否存在，如果不存在则代表没有违背该“针对生成文件”的约束\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 忽略隐藏文件\n if file.startswith('.'):\n continue\n \n # 获取文件扩展名，并转换为小写以进行匹配\n ext = os.path.splitext(file)[1].lower()\n \n if ext in code_extensions:\n # 判断完整文件名（包含扩展名）的字符数是否达到 5 个\n # 若文件名少于 5 个字符，说明该代码项目不满足给定要求\n if len(file) < 5:\n return False\n \n # 所有的代码文件均满足文件名至少包含 5 个字符，或没有代码文件违反此要求\n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"def verify_no_ternary_operator(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n from pathlib import Path\n\n def remove_strings_and_comments(code: str) -> str:\n # 使用正则去除所有格式的注释（单行、多行、HTML）与字符串字面量\n pattern = re.compile(\n r'//.*?$|' # 单行注释\n r'/\\*.*?\\*/|' # 多行块注释\n r'<!--.*?-->|' # HTML 注释\n r'\"(?:\\\\.|[^\"\\\\])*\"|' # 双引号字符串\n r\"'(?:\\\\.|[^'\\\\])*'|\" # 单引号字符串\n r'`(?:\\\\.|[^`\\\\])*`', # 反引号字符串\n re.DOTALL | re.MULTILINE\n )\n return pattern.sub('', code)\n\n def has_c_style_ternary(code: str) -> bool:\n # 去除干扰内容\n code = remove_strings_and_comments(code)\n # 排除 TypeScript 中常见的非三元 '?' 结构\n code = re.sub(r'\\?\\s*:', '', code) # 可选属性/参数，如 foo?: string\n code = re.sub(r'\\?\\?', '', code) # 空值合并运算符，如 a ?? b\n code = re.sub(r'\\?\\.', '', code) # 可选链，如 a?.b\n \n # 检查是否包含三元运算符，即 '?' 存在并在同一语句中跟随着 ':'\n pattern = re.compile(r'\\?[^;]*?:')\n return bool(pattern.search(code))\n\n def has_python_ternary(code: str) -> bool:\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n # 识别 Python 的三元表达式：a if condition else b\n if isinstance(node, ast.IfExp):\n return True\n except Exception:\n # 遇到解析错误则跳过\n pass\n return False\n\n workspace = Path(workspace_path)\n if workspace.exists():\n for root, _, files in os.walk(workspace):\n for file in files:\n file_path = Path(root) / file\n ext = file_path.suffix.lower()\n \n # 筛选常见的受管辖代码文件扩展名\n if ext not in ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.c', '.cpp', '.java', '.cs', '.php']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if has_python_ternary(content):\n return False\n elif ext in ['.html', '.htm']:\n # 对于 HTML，抽取 script 标签内部的内容进行细致检测\n script_pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n for script in script_pattern.findall(content):\n if has_c_style_ternary(script):\n return False\n else:\n if has_c_style_ternary(content):\n return False\n\n # 兜底检查回复内容中包含的 Markdown 代码块\n code_blocks = re.findall(r'```(\\w*)\\n(.*?)```', response, re.DOTALL)\n for lang, code in code_blocks:\n lang = lang.lower().strip()\n if lang in ['python', 'py']:\n if has_python_ternary(code):\n return False\n elif lang in ['html', 'htm']:\n script_pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n for script in script_pattern.findall(code):\n if has_c_style_ternary(script):\n return False\n elif lang in ['javascript', 'js', 'typescript', 'ts', 'jsx', 'tsx', 'c', 'cpp', 'java', 'cs', 'php']:\n if has_c_style_ternary(code):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # JavaScript / TypeScript / HTML 脚本内部函数定义的正则匹配模式\n js_func_patterns = [\n # 匹配传统函数声明：function myFunc(...)\n re.compile(r'function\\s+([a-zA-Z_$]\\w*)\\s*\\('),\n # 匹配匿名函数 / 箭头函数的变量赋值：const myFunc = ...\n re.compile(r'(?:const|let|var)\\s+([a-zA-Z_$]\\w*)\\s*=\\s*(?:async\\s+)?(?:function\\s*\\(|\\(.*?\\)\\s*=>|[a-zA-Z_$]\\w*\\s*=>)')\n ]\n # 匹配对象字面量方法、类方法：methodName(...) {\n js_method_pattern = re.compile(r'([a-zA-Z_$]\\w*)\\s*\\([^)]*\\)\\s*\\{')\n \n # 排除容易和 js_method_pattern 混淆的内置控制流/结构关键字\n reserved_words = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', \n 'return', 'else', 'constructor', 'super', 'with', 'await'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 解析 Web 相关代码文件中的函数\n if ext in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx']:\n funcs = set()\n # 提取正常声明与赋值的函数名\n for pattern in js_func_patterns:\n funcs.update(pattern.findall(content))\n \n # 提取类方法与对象方法名\n for m in js_method_pattern.findall(content):\n if m not in reserved_words:\n funcs.add(m)\n \n for func in funcs:\n if not is_prime(len(func)):\n return False\n\n # 解析 Python 代码文件中的函数\n elif ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not is_prime(len(node.name)):\n return False\n except SyntaxError:\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过 2 层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中必须保证所有函数的圈复杂度上限为 4","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":1,"instruction":"非常好，基础骨架有了。现在我们要把这三个简陋的 textarea 替换成真正的代码编辑器引擎。请通过 CDN 引入 Monaco Editor（或者 CodeMirror 6，看哪个你更方便实现）。然后在“HTML”、“CSS”和“JS”三个区域分别初始化一个编辑器实例，开启对应语言的语法高亮和基础行号显示。要求这三个编辑器实例的背景颜色与我们页面的深色主题融为一体，看起来像是一个统一的专业编码环境。另外请注意，后续所有轮次始终遵循以下要求：第一，你生成的代码文件里面的函数，其圈复杂度不可高于3；第二，操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def check_filename_length(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令上下文推断出的可能相关的代码文件类型\n # 涉及 Web 代码编辑器、Python相关的处理（pathlib, logging, 单元测试等）\n code_extensions = {'.html', '.css', '.js', '.py'}\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件后缀名\n _, ext = os.path.splitext(file)\n \n # 如果是代码文件\n if ext.lower() in code_extensions:\n # 判断文件名的字符数（直接使用len计算字符数）是否至少包含 5 个字符\n if len(file) < 5:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 1. 对于 Python 代码，通过 AST 进行精确判断\n if ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n # 发现了 Python 的三元运算符\n return False\n except Exception:\n pass\n \n # 2. 对于前端代码及类 C 语法代码，使用正则辅助判断\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.css', '.java', '.c', '.cpp', '.cs', '.php']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除多行注释和 HTML 注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 移除单行注释\n content = re.sub(r'//.*', '', content)\n \n # 移除字符串字面量（避免字符串内部恰巧出现 \"? :\" 结构被误判）\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '', content)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", '', content)\n content = re.sub(r\"`([^`\\\\]|\\\\.)*`\", '', content, flags=re.DOTALL)\n \n # 匹配经典的三元运算符结构模式\n # 解释:\n # \\s\\? : 匹配前面有空白符的问号\n # (?!\\?|\\.|:) : 问号之后不能是问号、点或冒号，排除 ??, ?., ?:\n # \\s : 问号之后有空白符\n # [^;{}]*?: : 接着找冒号，期间不能跨越分号或花括号（确保在同一语句级内）\n if re.search(r'\\s\\?(?!\\?|\\.|:)\\s[^;{}]*?:', content):\n # 发现了类 C 语言的三元运算符\n return False\n except Exception:\n pass\n\n # 若遍历项目内所有文件都没有找到三元运算符，则视作满足限制条件\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断数字是否为质数\n def is_prime(n):\n if n <= 1: \n return False\n if n <= 3: \n return True\n if n % 2 == 0 or n % 3 == 0: \n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0: \n return False\n i += 6\n return True\n\n # 各种函数定义情况的正则匹配模式\n py_pattern = re.compile(r'def\\s+([A-Za-z_]\\w*)\\s*\\(')\n js_pattern1 = re.compile(r'function\\s+([A-Za-z_$][\\w$]*)\\s*\\(')\n js_pattern2 = re.compile(r'([A-Za-z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[A-Za-z_$][\\w$]*\\s*=>)')\n js_pattern3 = re.compile(r'\\b([A-Za-z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{')\n \n # 针对 js_pattern3 中可能产生的误判，建立关键字排除集合\n js_keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', \n 'super', 'with', 'function', 'async', 'await'\n }\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 考虑到类似 CodePen 包含 HTML/JS/CSS，也涵盖其可能含有的测试与后端 (.py)\n if ext in {'.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm'}:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 简单剔除注释以防止正则误捕获\n if ext == '.py':\n content = re.sub(r'#.*', '', content)\n else:\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n\n funcs = []\n # 提取 Python 格式的函数\n if ext == '.py':\n funcs.extend(py_pattern.findall(content))\n # 提取 Web 前端（JS/TS/HTML等）格式的函数\n else:\n funcs.extend(js_pattern1.findall(content))\n funcs.extend(js_pattern2.findall(content))\n for match in js_pattern3.findall(content):\n if match not in js_keywords:\n funcs.append(match)\n \n # 验证每一个函数的命名长度是否恰好是质数\n for func in funcs:\n if not is_prime(len(func)):\n return False\n except Exception:\n # 忽略无法正常读取/解析的非文本文件\n pass\n\n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过 2 层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于3","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":2,"instruction":"接下来我们让代码真正“跑”起来，实现核心的实时预览功能。请将之前下方的空白“预览区”替换为一个 <iframe>。监听这三个编辑器的内容变化事件，当任何一个编辑器内容发生改变时，触发重新渲染。为了避免性能问题，请加上 500ms 的防抖（debounce）机制。渲染的具体做法是：将 HTML 内容放在 body 中，CSS 放在 <style> 标签里包裹进 head 中，JS 放在 <script> 标签里放在 body 末尾，然后动态拼接成完整的 HTML 字符串并写入 <iframe> 的 document 中。另外请注意，后续所有轮次始终遵循以下要求：第一，你生成的代码文件里面的函数，其圈复杂度不可高于2；第二，运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.scss', '.sass', '.less',\n '.js', '.jsx', '.ts', '.tsx', '.py', '.java',\n '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs',\n '.php', '.rb', '.sh', '.json', '.xml', '.yaml', '.yml'\n }\n\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git, .vscode 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 忽略隐藏文件\n if file.startswith('.'):\n continue\n \n _, ext = os.path.splitext(file)\n # 判断是否为代码文件\n if ext.lower() in code_extensions:\n has_code_file = True\n # 检查文件名长度是否至少为 5\n if len(file) < 5:\n return False\n\n # 如果存在代码文件且均符合长度要求，或者暂未生成代码文件（无违反项），则判定为 True\n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def check_python_ternary(content: str) -> bool:\n \"\"\"检查 Python 代码中是否使用了三元运算符\"\"\"\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n return True\n except Exception:\n pass\n return False\n\n def check_js_ts_html_ternary(content: str, ext: str) -> bool:\n \"\"\"检查 JS/TS/HTML 等代码文件中是否使用了三元运算符\"\"\"\n # 移除字符串字面量（兼容处理其中带转义符的情况）\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', content)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", content)\n content = re.sub(r'`([^`\\\\]|\\\\.)*`', '``', content)\n \n # 移除单行和多行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n # 如果是 HTML 文件，只抽取 <script> 标签中的内容进行检查\n if ext == '.html':\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, flags=re.DOTALL | re.IGNORECASE)\n js_content = \"\\n\".join(scripts)\n else:\n js_content = content\n \n if not js_content.strip():\n return False\n \n # 查找独立的 '?' 以及其后与之相配对的 ':'\n # 使用否定前瞻和否定后瞻过滤掉 '??', '?.', '?:' \n # 同时要求 '?' 和 ':' 之间不能跨越分号 ';'\n matches = re.finditer(r'(?<!\\?)\\?(?!\\?|\\.|:)([^;]*?):', js_content)\n for _ in matches:\n return True\n \n return False\n\n if not os.path.exists(workspace_path):\n return True \n\n has_ternary = False\n \n # 待检测的前端文件扩展名范围\n target_extensions = {\n '.js', '.ts', '.jsx', '.tsx', \n '.html', '.vue', '.svelte', \n '.mjs', '.cjs'\n }\n \n # 遍历代码项目\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n filepath = os.path.join(root, file)\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 针对不同后缀使用对应解析策略\n if ext == '.py':\n if check_python_ternary(content):\n has_ternary = True\n break\n elif ext in target_extensions:\n if check_js_ts_html_ternary(content, ext):\n has_ternary = True\n break\n except Exception:\n continue\n \n if has_ternary:\n break\n \n # 要求为：不能使用三元运算符。\n # 所以如果存在三元运算符，则该要求未满足(False)，反之满足(True)。\n return not has_ternary","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断给定的数字是否是质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 辅助函数：去除代码中的注释，减少正则误判\n def remove_comments(code: str, ext: str) -> str:\n if ext == '.py':\n code = re.sub(r'#.*', '', code)\n code = re.sub(r'\\'\\'\\'.*?\\'\\'\\'', '', code, flags=re.DOTALL)\n code = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', code, flags=re.DOTALL)\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n elif ext == '.html':\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n return code\n\n # JavaScript中的一些关键字，防止类方法匹配正则误将其识别为函数名\n js_keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'constructor', 'super', 'this'}\n all_functions = []\n\n # 遍历项目文件\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 去除注释\n content = remove_comments(content, ext)\n\n if ext == '.py':\n # 匹配 Python 的函数定义\n pattern = r'def\\s+([a-zA-Z_]\\w*)\\s*\\('\n funcs = re.findall(pattern, content)\n all_functions.extend(funcs)\n \n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.html']:\n # 1. 匹配普通的 function 声明\n pattern1 = r'function\\s+([a-zA-Z_$][\\w$]*)\\s*\\('\n funcs1 = re.findall(pattern1, content)\n all_functions.extend(funcs1)\n\n # 2. 匹配函数表达式与箭头函数赋值\n pattern2 = r'(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)'\n funcs2 = re.findall(pattern2, content)\n all_functions.extend(funcs2)\n\n # 3. 粗略匹配类中的方法 (需排除关键字)\n pattern3 = r'^\\s*(?:async\\s+)?(?:get\\s+|set\\s+)?([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{'\n funcs3 = re.findall(pattern3, content, re.MULTILINE)\n for f_name in funcs3:\n if f_name not in js_keywords:\n all_functions.extend([f_name])\n\n # 如果没有任何函数，由于是全称量词（“定义的每一个函数”），可认为是恒成立的\n if not all_functions:\n return True\n\n # 检查每一个提取出的函数名，其长度是否为质数\n for f in all_functions:\n if not is_prime(len(f)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过 2 层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":3,"instruction":"现在的布局是固定等分的，对宽屏或特定场景不太友好。下一步，我们需要加入面板大小的拖拽调节功能。请在三个编辑器之间（垂直分割线），以及编辑器区域和下方预览区之间（水平分割线），添加可视化的拖拽手柄（Resizer）。当用户鼠标按下并拖动这些手柄时，能够动态改变对应面板的宽度或高度。注意拖拽时需要有视觉反馈（比如鼠标指针变成对应的拖拽箭头），并且要防止面板被拖得过小（设置一个最小宽高限制），拖拽结束后各个编辑器实例要能自适应重新计算布局大小（如调用 editor.layout()）。另外请注意，后续所有轮次始终遵循以下要求：第一，你输出的代码中禁止出现超过1层的循环嵌套；第二，生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"RESIZE\"。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs',\n '.php', '.rb', '.sh', '.swift', '.kt', '.m', '.dart'\n }\n\n # 遍历当前代码项目工作空间\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件的后缀名并统一转换为小写进行比对\n _, ext = os.path.splitext(file)\n ext = ext.lower()\n \n # 仅针对代码文件进行检查\n if ext in code_extensions:\n # 检查文件名（包含后缀）的字符数是否至少为 5 个字符\n if len(file) < 5:\n return False\n \n # 如果所有的代码文件名称均满足至少包含 5 个字符的条件，返回 True\n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 辅助函数：专门用于剥离代码中的各类注释和字符串文本，防止干扰匹配\n def clean_code(text: str, ext: str) -> str:\n # 1. 移除多行注释 (针对类似JS/TS/Java等语言)\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.css', '.scss']:\n text = re.sub(r'/\\*.*?\\*/', '', text, flags=re.DOTALL)\n \n # 2. 移除单行注释\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.css', '.scss']:\n text = re.sub(r'//.*', '', text)\n elif ext in ['.py', '.sh', '.rb', '.yml', '.yaml']:\n text = re.sub(r'#.*', '', text)\n \n # 3. 移除各类字符串常量\n if ext == '.py':\n # Python独有的三引号字符串（允许换行）\n text = re.sub(r'\"\"\"(?:\\\\.|[^\"\\\\])*?\"\"\"', '\"\"', text, flags=re.DOTALL)\n text = re.sub(r\"'''(?:\\\\.|[^'\\\\])*?'''\", \"''\", text, flags=re.DOTALL)\n \n # 反引号字符串（主要用于JS/TS等语言，允许换行）\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*?`', '``', text, flags=re.DOTALL)\n \n # 普通单/双引号字符串（默认不跨行）\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\\\n])*?\"', '\"\"', text)\n text = re.sub(r\"'(?:\\\\.|[^'\\\\\\n])*?'\", \"''\", text)\n \n return text\n\n # 若项目目录不存在，默认无法验证或无生成代码\n if not os.path.exists(workspace_path):\n return False\n\n target_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.html', '.java', '.c', '.cpp', '.cs', '.css', '.scss'\n }\n\n # 遍历当前工作空间所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 对当前文件代码进行“脱水”处理（清空注释与字符串内文本）\n cleaned_content = clean_code(content, ext)\n \n if ext == '.py':\n # 在Python语法中，同一行同时出现if和else基本属于三元表达式（或其生成器条件变体）\n for line in cleaned_content.splitlines():\n if re.search(r'\\bif\\b.*\\belse\\b', line):\n return False\n else:\n # 对于JS/TS/Java/C/C++等语言，匹配类似 condition ? expr1 : expr2 的三元运算符\n # 确保排除了 ?. (可选链) / ?? (空值合并) / ?: (类型声明及 Elvis 操作符变体)\n # 限制匹配不能跨越语句分隔符 `;`\n if re.search(r'(?<!\\?)\\?(?!\\?|\\.|:)[^;]*?:', cleaned_content):\n return False\n except Exception:\n # 忽略个别因编码问题等无法读取的异常文件\n continue\n\n # 若扫描遍布所有文件都没有匹配到三元运算符模式，说明已遵循限制要求\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 辅助判断函数：是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 提取 JS / TS 中的函数名\n def get_js_functions(code: str) -> list:\n functions = []\n # 1. 匹配普通函数声明：function foo(...) 或 async function foo(...)\n functions.extend(re.findall(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(', code))\n \n # 2. 匹配函数表达式或箭头函数：const/let/var foo = function(...) / (...) =>\n pattern_expr = r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>)'\n functions.extend(re.findall(pattern_expr, code))\n \n # 3. 匹配对象内方法：foo: function(...)\n pattern_obj = r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?function\\b'\n functions.extend(re.findall(pattern_obj, code))\n \n return functions\n\n # 提取 Python 中的函数名\n def get_py_functions(code: str) -> list:\n return re.findall(r'^[ \\t]*def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(', code, re.MULTILINE)\n\n # 提取 HTML 中的内联脚本并获取其函数名\n def get_html_functions(code: str) -> list:\n functions = []\n script_blocks = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', code, re.IGNORECASE)\n for block in script_blocks:\n functions.extend(get_js_functions(block))\n return functions\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n funcs = []\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\n funcs = get_js_functions(content)\n elif ext == '.py':\n funcs = get_py_functions(content)\n elif ext in ['.html', '.htm']:\n funcs = get_html_functions(content)\n \n # 判断每个函数名长度是否为质数\n for func_name in funcs:\n if not is_prime(len(func_name)):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"RESIZE\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前后端代码文件扩展名\n code_extensions = {\n '.py', '.js', '.ts', '.html', '.css', '.htm', \n '.jsx', '.tsx', '.c', '.cpp', '.java', '.cs', '.go'\n }\n \n # 正则表达式用于安全地提取注释：\n # (?P<string>...) 优先匹配可能包含类似注释符号的各类字符串（三引号、双引号、单引号、反引号），避免被误认为注释\n # (?P<comment>...) 匹配常见的各种代码注释格式\n pattern = re.compile(\n r'(?P<string>\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|'\n r'(?P<comment>/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->|//.*?$|#.*?$)',\n re.MULTILINE\n )\n \n file_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取文件中的所有注释内容\n comments_text = \"\"\n for match in pattern.finditer(content):\n if match.group('comment'):\n comments_text += match.group('comment') + \"\\n\"\n \n # 统计提取的注释中 \"RESIZE\" 关键词出现的次数\n count = comments_text.count(\"RESIZE\")\n \n # 若不是恰好等于 2 个，则判断为未满足要求\n if count != 2:\n return False\n \n # 如果项目中没有任何代码文件，或者存在不可读代码文件，默认当作未满足要求；\n # 若所有读取成功的代码文件注释里都恰好包含了 2 个 \"RESIZE\"，则返回 True\n return file_found","tags":["Content","Keyword"]}]},{"round_id":4,"instruction":"很多时候写代码需要引入第三方的库，比如 jQuery 或者 Bootstrap。现在我们来完善顶部的“设置”按钮。点击它后，需要弹出一个居中的模态框（Modal）。模态框里分为两个标签页（Tabs）：一个是“外部 CSS”，另一个是“外部 JS”。每个标签页里允许用户动态添加、编辑和删除多个外部资源的 CDN 链接（提供输入框和增删按钮）。当用户在模态框点击“保存并应用”后，你的实时预览生成逻辑需要更新，将这些外部链接分别转化为 <link> 和 <script> 标签，并注入到 iframe 渲染内容的对应位置。另外请注意，后续所有轮次始终遵循以下要求：第一，生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"CDN\"；第二，在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，说明未生成任何文件，未违反要求\n if not os.path.exists(workspace_path):\n return True\n\n # 结合构建类似 CodePen 网站的需求，定义常见的代码文件后缀名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.json', '.py', '.md', '.sh'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录，避免检查如 .git 下的内部文件\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 忽略隐藏文件\n if file.startswith('.'):\n continue\n \n # 提取文件的扩展名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断该文件是否为代码文件\n if ext in code_extensions:\n # 检查文件名的字符数（包含扩展名），如果小于5个字符，则不满足要求\n if len(file) < 5:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def remove_strings_and_comments(code: str) -> str:\n \"\"\"剥离 JS/TS 语言代码中的字符串字面量和注释\"\"\"\n result = []\n i = 0\n n = len(code)\n in_single_quote = False\n in_double_quote = False\n in_backtick = False\n in_line_comment = False\n in_block_comment = False\n \n while i < n:\n if in_line_comment:\n if code[i] == '\\n':\n in_line_comment = False\n result.append('\\n')\n i += 1\n continue\n if in_block_comment:\n if i < n - 1 and code[i:i+2] == '*/':\n in_block_comment = False\n i += 2\n else:\n if code[i] == '\\n':\n result.append('\\n')\n i += 1\n continue\n \n if in_single_quote:\n if code[i] == '\\\\':\n i += 2\n continue\n if code[i] == \"'\":\n in_single_quote = False\n i += 1\n continue\n \n if in_double_quote:\n if code[i] == '\\\\':\n i += 2\n continue\n if code[i] == '\"':\n in_double_quote = False\n i += 1\n continue\n \n if in_backtick:\n if code[i] == '\\\\':\n i += 2\n continue\n if code[i] == '`':\n in_backtick = False\n i += 1\n continue\n \n # 检查字符串或注释的起始位置\n if code[i:i+2] == '//':\n in_line_comment = True\n i += 2\n continue\n if code[i:i+2] == '/*':\n in_block_comment = True\n i += 2\n continue\n if code[i] == \"'\":\n in_single_quote = True\n i += 1\n continue\n if code[i] == '\"':\n in_double_quote = True\n i += 1\n continue\n if code[i] == '`':\n in_backtick = True\n i += 1\n continue\n \n result.append(code[i])\n i += 1\n \n return ''.join(result)\n\n def has_js_ternary(code: str) -> bool:\n \"\"\"检测纯代码中是否存在 JS/TS 的三元运算符\"\"\"\n clean_code = remove_strings_and_comments(code)\n # 移除 JS/TS 中常见的含 ? 其他合法语法的干扰\n clean_code = clean_code.replace('??=', '')\n clean_code = clean_code.replace('??', '')\n clean_code = clean_code.replace('?.', '')\n clean_code = clean_code.replace('?:', '')\n \n # 查找是否存在问号 ? 以及随后的冒号 : （同一语句内不允许跨分号）\n if re.search(r'\\?[^;]*:', clean_code):\n return True\n return False\n\n def has_python_ternary(code: str) -> bool:\n \"\"\"检测 Python 代码中是否存在三元运算符(IfExp)\"\"\"\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n return True\n except Exception:\n pass\n return False\n\n def extract_scripts_from_html(html_code: str) -> str:\n \"\"\"提取 HTML 内的 script 脚本块\"\"\"\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', html_code, re.IGNORECASE | re.DOTALL)\n return '\\n'.join(scripts)\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n ext = os.path.splitext(file)[1].lower()\n \n # 分文件类型执行安全检测\n if ext == '.py':\n if has_python_ternary(content):\n return False\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n if has_js_ternary(content):\n return False\n elif ext in ['.html', '.htm', '.vue']:\n scripts_code = extract_scripts_from_html(content)\n if has_js_ternary(scripts_code):\n return False\n\n # 若没有任何文件包含三元运算符，则视作通过验证\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 清洗 JS 字符串和注释，防止正则表达式误匹配到字符串内容或注释里写的伪代码\n def clean_js(content: str) -> str:\n content = re.sub(r'([\"\\'`])(?:(?=(\\\\?))\\2.)*?\\1', '\"\"', content, flags=re.DOTALL)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n return content\n\n # 预设 JS 关键字黑名单，防止在匹配对象方法或类方法时将流程控制语句误当成函数定义\n reserved = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', \n 'class', 'import', 'export', 'constructor', 'typeof', 'instanceof', \n 'void', 'delete', 'new', 'super', 'this', 'else', 'try', 'finally', \n 'do', 'with', 'yield', 'await', 'eval', 'arguments', 'get', 'set'\n }\n\n # 从 JS / TS 内容中提取所有函数名\n def get_js_funcs(content: str) -> set:\n content = clean_js(content)\n names = set()\n \n # 1. 匹配函数声明：function fnName(...)\n for match in re.finditer(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(', content):\n names.add(match.group(1))\n \n # 2. 匹配赋值型和箭头函数：const fnName = function() 或者 const fnName = () => ...\n pattern_assign = r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'\n for match in re.finditer(pattern_assign, content):\n names.add(match.group(1))\n \n # 3. 匹配对象或类中的简写方法：fnName() { ... }\n for match in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{', content):\n name = match.group(1)\n if name not in reserved:\n names.add(name)\n \n # 4. 匹配对象属性式的函数定义：fnName: function(...)\n for match in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?function\\b', content):\n name = match.group(1)\n if name not in reserved:\n names.add(name)\n \n return names\n\n # 从 Python 内容中利用 AST 解析提取所有函数名\n def get_py_funcs(content: str) -> set:\n names = set()\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n names.add(node.name)\n except Exception:\n pass\n return names\n\n # 如果工作目录不存在，返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n names = set()\n \n # 根据后缀名进入对应的函数提取逻辑\n if ext == '.py':\n names = get_py_funcs(content)\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n names = get_js_funcs(content)\n elif ext in ['.html', '.htm', '.vue']:\n # 寻找所有的 script 脚本块并做二次提取\n for match in re.finditer(r'<script\\b[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE):\n names.update(get_js_funcs(match.group(1)))\n \n # 对每一个提取到的函数名称判断长度是否为质数\n for name in names:\n if not is_prime(len(name)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"CDN\"","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_extensions = {\n '.py', \n '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs', \n '.html', '.htm', '.vue', '.svelte', \n '.css', '.scss', '.sass', '.less'\n }\n\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n \n # JS/TS类文件注释提取\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs']:\n pattern = r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)' \\\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/|//[^\\r\\n]*)'\n for match in re.finditer(pattern, content):\n if match.group('comment'):\n comments.append(match.group('comment'))\n \n # CSS/样式类文件注释提取\n elif ext in ['.css', '.scss', '.sass', '.less']:\n if ext in ['.scss', '.sass', '.less']:\n pattern = r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')' \\\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/|//[^\\r\\n]*)'\n else:\n pattern = r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')' \\\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/)'\n for match in re.finditer(pattern, content):\n if match.group('comment'):\n comments.append(match.group('comment'))\n\n # HTML/模板类文件注释提取\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\n # 提取顶层 HTML 注释\n html_comments = re.findall(r'<!--[\\s\\S]*?-->', content)\n comments.extend(html_comments)\n \n # 提取 script 内部 JS 注释\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n js_pattern = r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)' \\\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/|//[^\\r\\n]*)'\n for script in scripts:\n for match in re.finditer(js_pattern, script):\n if match.group('comment'):\n comments.append(match.group('comment'))\n \n # 提取 style 内部 CSS 注释\n styles = re.findall(r'<style.*?>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n css_pattern = r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')' \\\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/)'\n for style in styles:\n for match in re.finditer(css_pattern, style):\n if match.group('comment'):\n comments.append(match.group('comment'))\n\n # Python文件注释提取\n elif ext == '.py':\n pattern = r'(?P<comment>\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|#[^\\r\\n]*)' \\\n r'|(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n for match in re.finditer(pattern, content):\n if match.group('comment'):\n comments.append(match.group('comment'))\n\n # 计算当前文件内所有的注释中包含 'CDN' 的总次数\n cdn_count = sum(comment.count('CDN') for comment in comments)\n \n # 若任意一份代码文件中 'CDN' 出现次数不为2，则直接不满足要求\n if cdn_count != 2:\n return False\n\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 定义常见代码文件的扩展名集合\n valid_extensions = {\n '.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', \n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rb', '.php', '.swift', '.json', '.sh'\n }\n \n has_code_file = False\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选出代码文件\n if ext in valid_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 忽略无法解析的特殊字符错误，以确保正确读取所有行\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 统计总行数（涵盖空行）\n num_lines = sum(1 for _ in f)\n \n # 判断行数是否为质数\n if not is_prime(num_lines):\n return False\n except Exception:\n # 如果打开或读取文件过程中发生未预期的异常，认定不满足要求\n return False\n \n # 确保至少存在一个代码文件，且均通过了质数行数验证\n return has_code_file","tags":["Quantity","Exact Value"]}]},{"round_id":5,"instruction":"我们还需要一个开发者控制台，这对于写 JS 太重要了。请在界面的最底部（或者预览区下方）添加一个可展开/折叠的“Console”面板。核心挑战来了：你需要拦截 iframe 内部运行时的 console.log、console.warn 和 console.error，将它们捕获并显示在这个“Console”面板中。不同级别的日志要有不同的文字颜色或背景色区分；如果输出的是对象或数组，要在面板里将其格式化为易于阅读的 JSON 字符串结构展示出来。另外请注意，后续所有轮次始终遵循以下要求：第一，原要求生成的每一份代码文件注释区域必须包含恰好2个关键词\"CDN\"的规则不再生效，替换为生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"CONSOLE\"；第二，后续执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的扩展名集合（涵盖Web开发及常见编程语言）\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', \n '.rs', '.rb', '.php', '.swift', '.kt', '.sh', \n '.json', '.xml', '.yaml', '.yml'\n }\n\n # 遍历工作区内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名\n _, ext = os.path.splitext(file)\n \n # 仅针对代码文件进行检查\n if ext.lower() in code_extensions:\n # 判断包含扩展名的完整文件名字符数是否小于 5\n if len(file) < 5:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"import os\nimport ast\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def remove_comments_and_strings_js(code: str) -> str:\n result = []\n i = 0\n n = len(code)\n in_string = False\n string_char = ''\n in_block_comment = False\n in_line_comment = False\n \n while i < n:\n if in_block_comment:\n if code[i:i+2] == '*/':\n in_block_comment = False\n i += 2\n continue\n i += 1\n continue\n \n if in_line_comment:\n if code[i] == '\\n':\n in_line_comment = False\n result.append('\\n') # 保留换行符以维持代码基础结构\n i += 1\n continue\n i += 1\n continue\n \n if in_string:\n if code[i] == '\\\\': # 跳过转义字符\n i += 2\n continue\n if code[i] == string_char:\n in_string = False\n i += 1\n continue\n i += 1\n continue\n \n if code[i:i+2] == '/*':\n in_block_comment = True\n i += 2\n continue\n \n if code[i:i+2] == '//':\n in_line_comment = True\n i += 2\n continue\n \n if code[i] in '\"\\'`':\n in_string = True\n string_char = code[i]\n i += 1\n continue\n \n result.append(code[i])\n i += 1\n \n return ''.join(result)\n\n def check_ternary_js(code: str) -> bool:\n stripped = remove_comments_and_strings_js(code)\n # 移除 JS/TS 中包含 '?' 但不是三元运算符的常见语法糖和类型定义\n stripped = stripped.replace('??', '')\n stripped = stripped.replace('?.', '')\n stripped = stripped.replace('?:', '')\n \n # 查找 '?' 后面跟着 ':' 的结构\n # 限制条件：两者之间不能出现分号 ';' (语句结束)，且跨越的换行符不能超过 3 个\n if re.search(r'\\?(?:[^;\\n]*\\n){0,3}[^;\\n]*:', stripped):\n return True\n return False\n\n def check_ternary_python(code: str) -> bool:\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n return True\n except SyntaxError:\n pass\n return False\n\n def check_ternary_html(code: str) -> bool:\n # 仅提取 HTML 中的 script 标签内容进行 JS 三元运算符检测\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', code, flags=re.IGNORECASE | re.DOTALL)\n for script in scripts:\n if check_ternary_js(script):\n return True\n return False\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 依据后缀判断文件类型并选用对应的三元运算符解析策略\n if ext == '.py':\n if check_ternary_python(content):\n return False # 发现三元运算符，违反了限制条件\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n if check_ternary_js(content):\n return False\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\n if check_ternary_html(content):\n return False\n except Exception:\n # 忽略读取失败的文件，避免非文本文件导致的系统级中断\n continue\n\n # 没有发现三元运算符，满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 辅助函数：判断数字是否为质数\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 辅助函数：移除代码中的注释以防止正则误判\n def remove_comments(code: str, ext: str) -> str:\n if ext == '.py':\n code = re.sub(r'#.*', '', code)\n code = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '', code)\n code = re.sub(r\"'''[\\s\\S]*?'''\", '', code)\n else:\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n code = re.sub(r'<!--[\\s\\S]*?-->', '', code)\n return code\n\n # 匹配 Python 函数\n py_func_pattern = re.compile(r\"def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(\")\n \n # 匹配 JS/TS 标准函数\n js_func_pattern = re.compile(r\"function\\s*\\*?\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(\")\n \n # 匹配 JS/TS 箭头函数赋值 (例如: const myFunc = () => )\n js_arrow_pattern = re.compile(r\"(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>\")\n \n # 匹配 JS/TS 匿名/具名函数赋值 (例如: const myFunc = function() )\n js_expr_pattern = re.compile(r\"(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?function\\s*\\*?(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(\")\n \n # 匹配 JS/TS 类方法 (例如: render() { 或 async fetch() {)\n js_method_pattern = re.compile(r\"^\\s*(?:async\\s+)?(?:get\\s+|set\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*(?::\\s*[^{]+)?\\{\", re.MULTILINE)\n\n # 排除 JS 中的常见控制流关键字（防止类方法正则发生误匹配）\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'function', \n 'return', 'yield', 'await', 'import', 'export', 'default'\n }\n\n if not os.path.isdir(workspace_path):\n return False\n\n valid_extensions = ('.js', '.ts', '.jsx', '.tsx', '.mjs', '.cjs', '.py', '.html', '.htm', '.vue', '.svelte')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n content = remove_comments(content, ext)\n names = []\n \n if ext == '.py':\n names.extend(py_func_pattern.findall(content))\n else:\n names.extend(js_func_pattern.findall(content))\n names.extend(js_arrow_pattern.findall(content))\n names.extend(js_expr_pattern.findall(content))\n \n methods = js_method_pattern.findall(content)\n for m in methods:\n if m not in keywords:\n names.append(m)\n \n # HTML/Vue 文件中可能内嵌了 Python（如通过特定模板/PyScript等），也作为备用检查一并包含\n if ext in ('.html', '.htm', '.vue', '.svelte'):\n names.extend(py_func_pattern.findall(content))\n\n # 遍历文件内的所有函数名，验证命名长度是否为质数\n for name in names:\n if not is_prime(len(name)):\n return False\n\n # 若没有任何函数违反规定（或没有定义任何函数），则返回True\n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"CONSOLE\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见代码文件类型拓展名集合\n valid_exts = {\n '.js', '.ts', '.jsx', '.tsx', \n '.css', '.scss', '.less', \n '.html', '.htm', \n '.py', \n '.java', '.c', '.cpp', '.cs', '.go', \n '.sh', '.rb'\n }\n \n def get_comments(file_path):\n ext = os.path.splitext(file_path)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return \"\"\n \n comments = []\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.java', '.c', '.cpp', '.cs', '.go']:\n # 采用正则表达式优先匹配字符串，如果匹配到单行/多行注释则收集\n pattern = r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`|/\\*[\\s\\S]*?\\*/|//.*)'\n tokens = re.findall(pattern, content)\n for token in tokens:\n if token.startswith('//') or token.startswith('/*'):\n comments.append(token)\n \n elif ext in ['.html', '.htm']:\n # HTML文件首先收集自带的 HTML 注释\n html_comments = re.findall(r'<!--[\\s\\S]*?-->', content)\n comments.extend(html_comments)\n \n # 然后分别提取 <script> 和 <style> 块中的代码进行 JS 和 CSS 的注释匹配\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n styles = re.findall(r'<style[^>]*>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n \n for script in scripts:\n pattern = r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`|/\\*[\\s\\S]*?\\*/|//.*)'\n tokens = re.findall(pattern, script)\n for token in tokens:\n if token.startswith('//') or token.startswith('/*'):\n comments.append(token)\n for style in styles:\n pattern = r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|/\\*[\\s\\S]*?\\*/|//.*)'\n tokens = re.findall(pattern, style)\n for token in tokens:\n if token.startswith('//') or token.startswith('/*'):\n comments.append(token)\n \n elif ext in ['.py']:\n # Python匹配包含多行字符串注释（Docstring）、单/双引号及 # 单行注释\n pattern = r'(\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|#.*)'\n tokens = re.findall(pattern, content)\n for token in tokens:\n if token.startswith('#') or token.startswith('\"\"\"') or token.startswith(\"'''\"):\n comments.append(token)\n \n elif ext in ['.sh', '.rb']:\n # Shell, Ruby等以 # 为注释\n pattern = r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|#.*)'\n tokens = re.findall(pattern, content)\n for token in tokens:\n if token.startswith('#'):\n comments.append(token)\n \n return \"\\n\".join(comments)\n\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除掉常见的非用户编写代码的文件夹\n if any(ignored in root for ignored in ['.git', 'node_modules', '__pycache__', '.idea', '.vscode']):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 提取出本文件中所有的注释内容\n comments_text = get_comments(file_path)\n \n # 统计关键词 \"CONSOLE\" 的出现次数\n count = comments_text.count(\"CONSOLE\")\n \n # 必须恰好为2个，否则验证失败\n if count != 2:\n return False\n \n # 如果没有扫描到任何代码文件，同样视为不符合要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n for i in range(2, int(n**0.5) + 1):\n if n % i == 0:\n return False\n return True\n\n # 针对当前前端 Web 项目推断可能涉及的代码文件扩展名\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符进行拆分，这样能够精准统计涵盖空行在内的IDE展示总行数\n lines_count = len(content.split('\\n'))\n \n # 检查总行数是否为质数\n if not is_prime(lines_count):\n return False\n \n has_code_file = True\n except Exception:\n # 如果由于编码等原因读取失败，则跳过\n continue\n\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":6,"instruction":"现在为了提升编码体验，我们来加入代码格式化功能。请通过 CDN 引入 Prettier 的核心库以及对应的 HTML/CSS/JS 解析插件。在顶部导航栏增加一个“格式化”按钮，当用户点击时，同时对当前三个编辑器里的代码进行美化排版，并将格式化后的代码回填到编辑器中。同时，请绑定键盘快捷键（Ctrl+S 或 Cmd+S），用户按下这个组合键时，不仅阻止浏览器的默认保存行为，也要触发这个代码格式化动作。另外请注意，后续所有轮次始终遵循以下要求：第一，原要求生成的每一份代码文件注释区域必须包含恰好2个关键词\"CONSOLE\"的规则不再生效，替换为生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"FORMAT\"；第二，请确保生成的代码文件中不包含多层if。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据多轮指令内容，推断可能的代码文件扩展名\n # 项目为Web代码编辑器，涉及 HTML, CSS, JS，且提到了Python的特定模块要求\n code_extensions = {'.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\n \n # 遍历代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n # 检查文件名长度是否至少包含 5 个字符\n if len(file) < 5:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def remove_comments_and_strings(text: str) -> str:\n # 正则表达式过滤掉常见的 JS/TS 字符串、单行注释及多行注释，避免对提取出来的 ? 和 : 产生干扰\n pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"|`(?:\\\\.|[^\\\\`])*`',\n re.DOTALL | re.MULTILINE\n )\n return pattern.sub('', text)\n\n def check_js(code: str) -> bool:\n clean_code = remove_comments_and_strings(code)\n # 寻找三元运算符的核心特征：不在注释和字符串中，由独立的 ? 和 : 组成，且中间没有出现分号或大括号\n # (?<!\\?)\\?(?![?.:]) 确保问号是单独作为运算符出现的，排除了 ??, ?., ?: 的情况\n if re.search(r'(?<!\\?)\\?(?![?.:])([^;{}]+?)(?<!:):(?!:)', clean_code):\n return False\n return True\n\n def check_python(code: str) -> bool:\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n # Python 语言中，三元运算符（If Expression）映射为 ast.IfExp 节点\n if isinstance(node, ast.IfExp):\n return False\n except Exception:\n # 解析失败意味着它大概率不是一个合法的 Python 代码（或含有语法错误），安全起见直接放行\n pass\n return True\n\n def check_html(code: str) -> bool:\n # 提取 HTML 文件中所有的 <script> 内容\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', code, re.IGNORECASE | re.DOTALL)\n for script in scripts:\n if not check_js(script):\n return False\n return True\n\n def check_general_code(code: str, lang: str) -> bool:\n lang = lang.lower()\n if lang in ['python', 'py']:\n return check_python(code)\n elif lang in ['javascript', 'js', 'typescript', 'ts', 'jsx', 'tsx', 'vue', 'svelte']:\n return check_js(code)\n elif lang in ['html', 'htm']:\n return check_html(code)\n else:\n # 遇到未知语言代码，双重检测以尽可能避免漏判\n if not check_python(code):\n return False\n if not check_js(code):\n return False\n return True\n\n # 1. 检查物理存在于 workspace_path 中的文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if ext == '.py':\n if not check_python(content):\n return False\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.svelte']:\n if not check_js(content):\n return False\n elif ext in ['.html', '.htm']:\n if not check_html(content):\n return False\n else:\n if not check_python(content):\n return False\n if not check_js(content):\n return False\n\n # 2. 检查 response 回复中携带的 Markdown 代码块\n code_blocks = re.findall(r'```([a-zA-Z]*)\\n(.*?)\\n```', response, re.DOTALL)\n for lang, code in code_blocks:\n if not check_general_code(code, lang.strip()):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"def check_function_name_prime_length(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n # 辅助函数：判断给定数字是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 解析 Python 代码中的函数\n def get_py_funcs(code: str):\n funcs = []\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n funcs.append(node.name)\n except Exception:\n pass\n return funcs\n\n # 解析 JavaScript/TypeScript 等代码中的函数\n def get_js_funcs(code: str):\n funcs = set()\n \n # 1. 常规函数：function foo(...)\n for m in re.finditer(r'\\bfunction\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\(', code):\n funcs.add(m.group(1))\n \n # 2. 赋值形式：foo = function(...) 或 foo : function(...)\n for m in re.finditer(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*[:=]\\s*function\\s*\\(', code):\n funcs.add(m.group(1))\n \n # 3. 箭头函数：foo = (...) => 或 foo : (...) => 或 foo = arg =>\n for m in re.finditer(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*[:=]\\s*(?:\\([^)]*\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=>', code):\n funcs.add(m.group(1))\n \n # 4. 类或对象方法：foo(...) {\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'constructor',\n 'super', 'return', 'import', 'export', 'typeof', 'new', 'get', 'set',\n 'do', 'with', 'yield', 'await', 'class', 'let', 'const', 'var', 'default'\n }\n for m in re.finditer(r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\([^)]*\\)\\s*\\{', code):\n name = m.group(1)\n if name not in keywords:\n funcs.add(name)\n \n return list(funcs)\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的隐藏文件夹及第三方依赖目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n funcs = []\n if ext == '.py':\n funcs = get_py_funcs(content)\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n funcs = get_js_funcs(content)\n elif ext in ['.html', '.htm']:\n # 提取 HTML 文件内的所有 <script> 标签包裹的代码内容\n script_matches = re.finditer(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for m in script_matches:\n funcs.extend(get_js_funcs(m.group(1)))\n \n # 若含有任何函数定义，立刻校验函数名长度是否为质数\n for func_name in funcs:\n if not is_prime(len(func_name)):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"FORMAT\"","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx'}\n code_files = []\n \n # 遍历项目目录，收集所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.lower().endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果项目中没有生成相关的代码文件，无法判断是否满足“每一份”的要求，返回 False\n if not code_files:\n return False\n \n # 综合正则表达式：\n # group 1, 2, 3 匹配双引号、单引号、反引号包裹的字符串（跳过字符串内部的转义符）\n # group 4 匹配 HTML 注释、多行注释、单行注释\n # 通过这种交替匹配的方式，可以安全地跳过位于字符串内部的伪注释符号\n combined_pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\")|'\n r'(\\'(?:\\\\.|[^\\'\\\\])*\\')|'\n r'(`(?:\\\\.|[^`\\\\])*`)|'\n r'(<!--[\\s\\S]*?-->|/\\*[\\s\\S]*?\\*/|//.*)'\n )\n \n # 检查每一份代码文件\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果文件无法按 UTF-8 读取，为保障评测顺利可选择跳过或视作不合规\n # 此处稳妥起见，无法读取视为异常情况，直接返回 False\n return False\n \n comments_content = []\n for match in combined_pattern.finditer(content):\n # 如果匹配到了 group 4，则说明是注释内容\n if match.group(4):\n comments_content.append(match.group(4))\n \n all_comments_text = '\\n'.join(comments_content)\n \n # 统计当前文件所有注释中 \"FORMAT\" 出现的次数\n format_count = all_comments_text.count(\"FORMAT\")\n \n # 如果不是恰好包含 2 个，则不满足要求\n if format_count != 2:\n return False\n \n # 所有代码文件均满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n i = 3\n while i * i <= n:\n if n % i == 0:\n return False\n i += 2\n return True\n\n # 定义涵盖前端与常见后端脚本的代码文件扩展名集合\n code_extensions = {\n '.py', '.js', '.html', '.css', '.ts', '.jsx', \n '.tsx', '.json', '.vue', '.sh', '.bash', \n '.java', '.c', '.cpp', '.h', '.go', '.rs'\n }\n \n found_any_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前代码项目工作空间\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按换行符切分以准确统计行数，涵盖空行逻辑\n line_count = len(content.split('\\n'))\n except Exception:\n # 对于非utf-8编码等读取异常的文件，跳过处理\n continue\n \n found_any_code_file = True\n # 若存在行数不为质数的代码文件，则不满足要求\n if not is_prime(line_count):\n return False\n \n # 如果没有找到任何代码文件，则无法验证是否满足要求，返回 False\n # 如果找到了代码文件并且所有文件行数都为质数，则返回 True\n return found_any_code_file","tags":["Quantity","Exact Value"]},{"constraint":"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保生成的代码文件中不包含多层if","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":7,"instruction":"最后一轮，我们要实现状态的持久化和 URL 分享功能。首先，用户的代码和外部资源配置需要在每次变动时自动存入 localStorage，刷新页面后能自动恢复上次编辑的状态。其次，点击顶部的“分享”按钮时，请将当前的 HTML、CSS、JS 代码以及外部资源列表，通过 Base64（或 LZString 压缩算法）序列化拼接到当前页面 URL 的 Hash (#) 中，然后弹出一个 toast 提示“链接已复制到剪贴板”。当别人通过带有 Hash 数据的 URL 访问我们的页面时，页面初始化要优先从 URL 解析并恢复这些代码和配置，而不是从 localStorage 中读取。另外请注意，后续所有轮次始终遵循以下要求：第一，原要求生成的每一份代码文件注释区域必须包含恰好2个关键词\"FORMAT\"的规则不再生效，替换为生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"SHARE\"；第二，生成的代码里，“=”号的前后必须各自包含两个空格。","instruction_following_checklist":[{"constraint":"你所输出的任何一个代码文件，其包含的注释都应当是全英文的","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的所有代码文件，其文件名必须至少包含 5 个字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n\n workspace = Path(workspace_path)\n if not workspace.exists():\n return False\n\n # 针对前端项目和常见的代码相关文件定义扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.py', '.md'}\n \n has_code_file = False\n \n # 递归遍历工作区下的所有文件\n for file_path in workspace.rglob('*'):\n if file_path.is_file() and file_path.suffix.lower() in code_extensions:\n has_code_file = True\n \n # file_path.name 会获取包含扩展名的文件名，例如 \"index.html\"\n if len(file_path.name) < 5:\n return False\n \n # 必须保证至少输出了一个代码文件，且全部通过了检查才能返回True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"限制条件：你生成的代码文件中不能使用三元运算符","validation_code":"import os\nimport ast\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def remove_comments_and_strings_js(code: str) -> str:\n \"\"\"\n 移除 JavaScript 代码中的注释和字符串内容，以避免在这些区域内误匹配 ? 和 :。\n \"\"\"\n result = []\n i = 0\n n = len(code)\n in_single_quote = False\n in_double_quote = False\n in_template_literal = False\n in_single_comment = False\n in_multi_comment = False\n \n while i < n:\n # 单行注释处理\n if in_single_comment:\n if code[i] == '\\n':\n in_single_comment = False\n i += 1\n continue\n # 多行注释处理\n if in_multi_comment:\n if i + 1 < n and code[i:i+2] == '*/':\n in_multi_comment = False\n i += 2\n else:\n i += 1\n continue\n # 单引号字符串处理\n if in_single_quote:\n if code[i] == '\\\\': # 转义跳过\n i += 2\n continue\n if code[i] == \"'\":\n in_single_quote = False\n i += 1\n continue\n # 双引号字符串处理\n if in_double_quote:\n if code[i] == '\\\\':\n i += 2\n continue\n if code[i] == '\"':\n in_double_quote = False\n i += 1\n continue\n # 模板字符串处理\n if in_template_literal:\n if code[i] == '\\\\':\n i += 2\n continue\n if code[i] == '`':\n in_template_literal = False\n i += 1\n continue\n \n # 检测进入注释\n if code[i] == '/' and i + 1 < n:\n if code[i+1] == '/':\n in_single_comment = True\n i += 2\n continue\n elif code[i+1] == '*':\n in_multi_comment = True\n i += 2\n continue\n \n # 检测进入字符串\n if code[i] == \"'\":\n in_single_quote = True\n i += 1\n continue\n if code[i] == '\"':\n in_double_quote = True\n i += 1\n continue\n if code[i] == '`':\n in_template_literal = True\n i += 1\n continue\n \n # 保存非字符串和非注释的代码本体内容\n result.append(code[i])\n i += 1\n \n return \"\".join(result)\n\n def check_js_ternary(code: str) -> bool:\n \"\"\"\n 检查去除了注释和字符串的 JS 纯代码中是否包含三元运算符。\n \"\"\"\n clean = remove_comments_and_strings_js(code)\n i = 0\n n = len(clean)\n while i < n:\n if clean[i] == '?':\n next_char = clean[i+1] if i+1 < n else ''\n \n # 排除可选链 ?.\n if next_char == '.':\n i += 2\n continue\n \n # 排除空值合并 ??\n if next_char == '?':\n i += 2\n continue\n \n # 排除 TypeScript 可选属性/参数 ?:\n if next_char == ':':\n i += 2\n continue\n \n # 在排除了其他语法后，如果在同一语句结束前找到了 :，则判定为三元运算符\n j = i + 1\n while j < n:\n if clean[j] == ':':\n return True\n if clean[j] == ';':\n break\n j += 1\n i += 1\n return False\n\n def extract_script_content(html_code: str) -> str:\n \"\"\"\n 提取 HTML 文件中的 <script> 标签包裹的 JavaScript 内容。\n \"\"\"\n pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n scripts = pattern.findall(html_code)\n return \"\\n\".join(scripts)\n\n # 遍历工作区下的所有代码文件进行检查\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n\n if ext == '.py':\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n return False # 发现了 Python 中的三元运算符 (if-else 表达式)\n except Exception:\n pass\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n if check_js_ternary(code):\n return False # 发现了 JS/TS 中的三元运算符\n elif ext in ['.html', '.htm', '.vue']:\n script_code = extract_script_content(code)\n if script_code and check_js_ternary(script_code):\n return False # 发现了 HTML 脚本中的三元运算符\n\n # 若所有代码文件都排除了包含三元运算符的可能性，则满足限制条件\n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数","validation_code":"import os\nimport ast\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断一个数字是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 用于匹配 JS/TS/HTML(script) 中的函数命名的正则表达式\n # 匹配普通函数定义：function foo(...) 或 function* foo(...)\n js_pattern1 = re.compile(r'\\bfunction\\s*\\*?\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n # 匹配变量赋值形式的函数：const foo = function(...) 或 const foo = (...) =>\n js_pattern2 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)')\n # 匹配对象内部的方法定义形式：foo: function(...)\n js_pattern3 = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?function\\b')\n\n # 遍历生成的代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 仅筛选常见的代码及 Web 文件类型\n if ext not in ['.py', '.js', '.ts', '.html', '.jsx', '.tsx', '.vue', '.svelte']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n # Python 文件：通过抽象语法树提取函数命名\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n func_name = node.name\n if not is_prime(len(func_name)):\n return False\n except Exception:\n pass\n else:\n # Web 文件：通过正则匹配提取函数命名\n names1 = js_pattern1.findall(content)\n names2 = js_pattern2.findall(content)\n names3 = js_pattern3.findall(content)\n \n # 遍历所有找到的函数名称\n for func_name in (names1 + names2 + names3):\n if not is_prime(len(func_name)):\n return False\n\n # 若所有函数命名长度均为质数（或没有定义函数），则满足要求\n return True","tags":["Style","Naming"]},{"constraint":"你输出的代码中禁止出现超过1层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件里面的函数，其圈复杂度不可高于2","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\"SHARE\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义前端与常见可能生成的代码文件的后缀集合\n code_extensions = {\n '.py', '.sh', \n '.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp',\n '.html', '.htm'\n }\n\n found_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无关目录，避免扫描测试用例生成的环境隔离文件\n for ignore_dir in ['.git', 'node_modules', '__pycache__', 'venv', '.venv', 'dist', 'build']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果生成的文件无法以普通的 UTF-8 文本读取，无法正确验证注释，判作失败\n return False\n \n comments = []\n \n # 针对 C-style 注释类型（如 js, css, java 等）\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp'}:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<comment>//[^\\n]*|/\\*[\\s\\S]*?\\*/)',\n re.DOTALL\n )\n for match in pattern.finditer(content):\n if match.group('comment'):\n comments.append(match.group('comment'))\n \n # 针对 Python 和 Shell 类型单行注释\n elif ext in {'.py', '.sh'}:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\')'\n r'|(?P<comment>#[^\\n]*)',\n re.DOTALL\n )\n for match in pattern.finditer(content):\n if match.group('comment'):\n comments.append(match.group('comment'))\n # Python中的多行字符串如果做文档字符串，通常也视为一种注释\n if ext == '.py' and match.group('string'):\n s = match.group('string')\n if s.startswith('\"\"\"') or s.startswith(\"'''\"):\n comments.append(s)\n \n # 针对 HTML 类型的多重嵌套注释\n elif ext in {'.html', '.htm'}:\n # 1. 提取 HTML 注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n # 2. 提取 script 标签中的 JS 注释\n script_pattern = re.compile(r'<script.*?>([\\s\\S]*?)</script>', re.IGNORECASE)\n for script_match in script_pattern.finditer(content):\n script_content = script_match.group(1)\n js_pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<comment>//[^\\n]*|/\\*[\\s\\S]*?\\*/)',\n re.DOTALL\n )\n for m in js_pattern.finditer(script_content):\n if m.group('comment'):\n comments.append(m.group('comment'))\n \n # 3. 提取 style 标签中的 CSS 注释\n style_pattern = re.compile(r'<style.*?>([\\s\\S]*?)</style>', re.IGNORECASE)\n for style_match in style_pattern.finditer(content):\n style_content = style_match.group(1)\n css_pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/)',\n re.DOTALL\n )\n for m in css_pattern.finditer(style_content):\n if m.group('comment'):\n comments.append(m.group('comment'))\n\n # 将该文件中抽出的所有有效注释组合并校验\n all_comments = \"\\n\".join(comments)\n share_count = all_comments.count(\"SHARE\")\n \n # 如果任意一份代码文件的注释中，包含\"SHARE\"的数量不恰好为2，验证均告失败\n if share_count != 2:\n return False\n\n # 若未找到任何代码文件，则不符合需求预期\n return found_file","tags":["Content","Keyword"]},{"constraint":"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个整数是否为质数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 定义可能生成的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.py', '.ts', '.jsx', \n '.tsx', '.json', '.sh', '.vue', '.svelte'\n }\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除通常不需要检测的第三方依赖或版本控制目录\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '__pycache__' in dirs:\n dirs.remove('__pycache__')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n num_lines = len(lines)\n \n # 判断行数是否为质数\n if not is_prime(num_lines):\n return False\n except Exception:\n # 如果由于编码等问题无法读取文件，认定为不符合要求\n return False\n \n # 确保至少检查过一个代码文件，防止空目录返回True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保生成的代码文件中不包含多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"生成的代码里，“=”号的前后必须各自包含两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def check_text_for_equals(text: str) -> bool:\n # 提前移除注释和字符串，以防其中的等号干扰判定\n # 1. 移除 HTML 注释\n text = re.sub(r'<!--[\\s\\S]*?-->', '', text)\n # 2. 移除 JS/CSS 块注释\n text = re.sub(r'/\\*[\\s\\S]*?\\*/', '', text)\n # 3. 移除 JS 单行注释\n text = re.sub(r'//.*', '', text)\n # 4. 移除双引号字符串\n text = re.sub(r'\"(?:[^\"\\\\]|\\\\.)*\"', '\"\"', text)\n # 5. 移除单引号字符串\n text = re.sub(r\"'(?:[^'\\\\]|\\\\.)*'\", \"''\", text)\n # 6. 移除反引号（模板字符串）\n text = re.sub(r'`(?:[^`\\\\]|\\\\.)*`', '``', text)\n \n # 查找不在特定操作符（如 ==, !=, >=, =>, += 等）中的孤立单等号 '='\n pattern = r'(?<![=!<>+\\-*/%&|^~$])=(?![=>])'\n \n for match in re.finditer(pattern, text):\n start = match.start()\n end = match.end()\n \n # 判断等号的前两个字符是否为两个连续空格\n if start < 2 or text[start-2:start] != ' ':\n return False\n # 判断等号的后两个字符是否为两个连续空格\n if end > len(text) - 2 or text[end:end+2] != ' ':\n return False\n \n return True\n\n # 1. 检查 response 中的 Markdown 代码块\n code_blocks = re.findall(r'```.*?\\n([\\s\\S]*?)```', response)\n for block in code_blocks:\n if not check_text_for_equals(block):\n return False\n\n # 2. 检查 workspace_path 下生成的代码文件\n if os.path.isdir(workspace_path):\n valid_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.json', '.vue', '.py', '.txt'}\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in valid_exts:\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_text_for_equals(content):\n return False\n except Exception:\n pass\n \n return True","tags":["Style","Layout"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"我们要从零搭建一个类似 CodePen 的 Web 代码编辑器。第一步，先来勾勒一下整体页面的基础骨架。请创建一个包含顶部导航和主界面的页面：顶部导航栏左侧放一个应用的 Logo（文字即可），右侧放置几个占位按钮（比如“运行”、“设置”、“分享”）。主界面采用经典的上下布局，上半部分等分为三个区域，分别加上“HTML”、“CSS”、“JS”的标题，并在里面各放一个普通的 <textarea> 作为临时的代码输入框；下半部分占据剩余空间，加上“预览区”的标题和一个空白占位用的 <div>。整体布局使用 Flexbox 或 Grid 撑满全屏，配色采用现代化的深色主题。请注意，后续所有轮次编写代码时都必须始终遵循以下要求：你输出的代码中禁止出现超过 2 层的循环嵌套，同时在生成的代码文件中必须保证所有函数的圈复杂度上限为 4。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx',\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp',\\n '.cs', '.go', '.rb', '.php', '.swift', '.kt', '.rs', \\n '.json', '.sh', '.bat'\\n }\\n \\n # 检查工作空间路径是否存在，如果不存在则代表没有违背该“针对生成文件”的约束\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 忽略隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n # 获取文件扩展名，并转换为小写以进行匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in code_extensions:\\n # 判断完整文件名（包含扩展名）的字符数是否达到 5 个\\n # 若文件名少于 5 个字符，说明该代码项目不满足给定要求\\n if len(file) < 5:\\n return False\\n \\n # 所有的代码文件均满足文件名至少包含 5 个字符，或没有代码文件违反此要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"def verify_no_ternary_operator(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n from pathlib import Path\\n\\n def remove_strings_and_comments(code: str) -> str:\\n # 使用正则去除所有格式的注释（单行、多行、HTML）与字符串字面量\\n pattern = re.compile(\\n r'//.*?$|' # 单行注释\\n r'/\\\\*.*?\\\\*/|' # 多行块注释\\n r'<!--.*?-->|' # HTML 注释\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|' # 双引号字符串\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\" # 单引号字符串\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', # 反引号字符串\\n re.DOTALL | re.MULTILINE\\n )\\n return pattern.sub('', code)\\n\\n def has_c_style_ternary(code: str) -> bool:\\n # 去除干扰内容\\n code = remove_strings_and_comments(code)\\n # 排除 TypeScript 中常见的非三元 '?' 结构\\n code = re.sub(r'\\\\?\\\\s*:', '', code) # 可选属性/参数，如 foo?: string\\n code = re.sub(r'\\\\?\\\\?', '', code) # 空值合并运算符，如 a ?? b\\n code = re.sub(r'\\\\?\\\\.', '', code) # 可选链，如 a?.b\\n \\n # 检查是否包含三元运算符，即 '?' 存在并在同一语句中跟随着 ':'\\n pattern = re.compile(r'\\\\?[^;]*?:')\\n return bool(pattern.search(code))\\n\\n def has_python_ternary(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n # 识别 Python 的三元表达式：a if condition else b\\n if isinstance(node, ast.IfExp):\\n return True\\n except Exception:\\n # 遇到解析错误则跳过\\n pass\\n return False\\n\\n workspace = Path(workspace_path)\\n if workspace.exists():\\n for root, _, files in os.walk(workspace):\\n for file in files:\\n file_path = Path(root) / file\\n ext = file_path.suffix.lower()\\n \\n # 筛选常见的受管辖代码文件扩展名\\n if ext not in ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.c', '.cpp', '.java', '.cs', '.php']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if has_python_ternary(content):\\n return False\\n elif ext in ['.html', '.htm']:\\n # 对于 HTML，抽取 script 标签内部的内容进行细致检测\\n script_pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n for script in script_pattern.findall(content):\\n if has_c_style_ternary(script):\\n return False\\n else:\\n if has_c_style_ternary(content):\\n return False\\n\\n # 兜底检查回复内容中包含的 Markdown 代码块\\n code_blocks = re.findall(r'```(\\\\w*)\\\\n(.*?)```', response, re.DOTALL)\\n for lang, code in code_blocks:\\n lang = lang.lower().strip()\\n if lang in ['python', 'py']:\\n if has_python_ternary(code):\\n return False\\n elif lang in ['html', 'htm']:\\n script_pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n for script in script_pattern.findall(code):\\n if has_c_style_ternary(script):\\n return False\\n elif lang in ['javascript', 'js', 'typescript', 'ts', 'jsx', 'tsx', 'c', 'cpp', 'java', 'cs', 'php']:\\n if has_c_style_ternary(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # JavaScript / TypeScript / HTML 脚本内部函数定义的正则匹配模式\\n js_func_patterns = [\\n # 匹配传统函数声明：function myFunc(...)\\n re.compile(r'function\\\\s+([a-zA-Z_$]\\\\w*)\\\\s*\\\\('),\\n # 匹配匿名函数 / 箭头函数的变量赋值：const myFunc = ...\\n re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z_$]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\s*\\\\(|\\\\(.*?\\\\)\\\\s*=>|[a-zA-Z_$]\\\\w*\\\\s*=>)')\\n ]\\n # 匹配对象字面量方法、类方法：methodName(...) {\\n js_method_pattern = re.compile(r'([a-zA-Z_$]\\\\w*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # 排除容易和 js_method_pattern 混淆的内置控制流/结构关键字\\n reserved_words = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', \\n 'return', 'else', 'constructor', 'super', 'with', 'await'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 解析 Web 相关代码文件中的函数\\n if ext in ['.js', '.ts', '.html', '.htm', '.jsx', '.tsx']:\\n funcs = set()\\n # 提取正常声明与赋值的函数名\\n for pattern in js_func_patterns:\\n funcs.update(pattern.findall(content))\\n \\n # 提取类方法与对象方法名\\n for m in js_method_pattern.findall(content):\\n if m not in reserved_words:\\n funcs.add(m)\\n \\n for func in funcs:\\n if not is_prime(len(func)):\\n return False\\n\\n # 解析 Python 代码文件中的函数\\n elif ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not is_prime(len(node.name)):\\n return False\\n except SyntaxError:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过 2 层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中必须保证所有函数的圈复杂度上限为 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"非常好，基础骨架有了。现在我们要把这三个简陋的 textarea 替换成真正的代码编辑器引擎。请通过 CDN 引入 Monaco Editor（或者 CodeMirror 6，看哪个你更方便实现）。然后在“HTML”、“CSS”和“JS”三个区域分别初始化一个编辑器实例，开启对应语言的语法高亮和基础行号显示。要求这三个编辑器实例的背景颜色与我们页面的深色主题融为一体，看起来像是一个统一的专业编码环境。另外请注意，后续所有轮次始终遵循以下要求：第一，你生成的代码文件里面的函数，其圈复杂度不可高于3；第二，操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def check_filename_length(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令上下文推断出的可能相关的代码文件类型\\n # 涉及 Web 代码编辑器、Python相关的处理（pathlib, logging, 单元测试等）\\n code_extensions = {'.html', '.css', '.js', '.py'}\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件后缀名\\n _, ext = os.path.splitext(file)\\n \\n # 如果是代码文件\\n if ext.lower() in code_extensions:\\n # 判断文件名的字符数（直接使用len计算字符数）是否至少包含 5 个字符\\n if len(file) < 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 1. 对于 Python 代码，通过 AST 进行精确判断\\n if ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n # 发现了 Python 的三元运算符\\n return False\\n except Exception:\\n pass\\n \\n # 2. 对于前端代码及类 C 语法代码，使用正则辅助判断\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.css', '.java', '.c', '.cpp', '.cs', '.php']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除多行注释和 HTML 注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 移除单行注释\\n content = re.sub(r'//.*', '', content)\\n \\n # 移除字符串字面量（避免字符串内部恰巧出现 \\\"? :\\\" 结构被误判）\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '', content)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", '', content)\\n content = re.sub(r\\\"`([^`\\\\\\\\]|\\\\\\\\.)*`\\\", '', content, flags=re.DOTALL)\\n \\n # 匹配经典的三元运算符结构模式\\n # 解释:\\n # \\\\s\\\\? : 匹配前面有空白符的问号\\n # (?!\\\\?|\\\\.|:) : 问号之后不能是问号、点或冒号，排除 ??, ?., ?:\\n # \\\\s : 问号之后有空白符\\n # [^;{}]*?: : 接着找冒号，期间不能跨越分号或花括号（确保在同一语句级内）\\n if re.search(r'\\\\s\\\\?(?!\\\\?|\\\\.|:)\\\\s[^;{}]*?:', content):\\n # 发现了类 C 语言的三元运算符\\n return False\\n except Exception:\\n pass\\n\\n # 若遍历项目内所有文件都没有找到三元运算符，则视作满足限制条件\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断数字是否为质数\\n def is_prime(n):\\n if n <= 1: \\n return False\\n if n <= 3: \\n return True\\n if n % 2 == 0 or n % 3 == 0: \\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0: \\n return False\\n i += 6\\n return True\\n\\n # 各种函数定义情况的正则匹配模式\\n py_pattern = re.compile(r'def\\\\s+([A-Za-z_]\\\\w*)\\\\s*\\\\(')\\n js_pattern1 = re.compile(r'function\\\\s+([A-Za-z_$][\\\\w$]*)\\\\s*\\\\(')\\n js_pattern2 = re.compile(r'([A-Za-z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[A-Za-z_$][\\\\w$]*\\\\s*=>)')\\n js_pattern3 = re.compile(r'\\\\b([A-Za-z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # 针对 js_pattern3 中可能产生的误判，建立关键字排除集合\\n js_keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', \\n 'super', 'with', 'function', 'async', 'await'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 考虑到类似 CodePen 包含 HTML/JS/CSS，也涵盖其可能含有的测试与后端 (.py)\\n if ext in {'.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm'}:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 简单剔除注释以防止正则误捕获\\n if ext == '.py':\\n content = re.sub(r'#.*', '', content)\\n else:\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n\\n funcs = []\\n # 提取 Python 格式的函数\\n if ext == '.py':\\n funcs.extend(py_pattern.findall(content))\\n # 提取 Web 前端（JS/TS/HTML等）格式的函数\\n else:\\n funcs.extend(js_pattern1.findall(content))\\n funcs.extend(js_pattern2.findall(content))\\n for match in js_pattern3.findall(content):\\n if match not in js_keywords:\\n funcs.append(match)\\n \\n # 验证每一个函数的命名长度是否恰好是质数\\n for func in funcs:\\n if not is_prime(len(func)):\\n return False\\n except Exception:\\n # 忽略无法正常读取/解析的非文本文件\\n pass\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过 2 层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于3\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来我们让代码真正“跑”起来，实现核心的实时预览功能。请将之前下方的空白“预览区”替换为一个 <iframe>。监听这三个编辑器的内容变化事件，当任何一个编辑器内容发生改变时，触发重新渲染。为了避免性能问题，请加上 500ms 的防抖（debounce）机制。渲染的具体做法是：将 HTML 内容放在 body 中，CSS 放在 <style> 标签里包裹进 head 中，JS 放在 <script> 标签里放在 body 末尾，然后动态拼接成完整的 HTML 字符串并写入 <iframe> 的 document 中。另外请注意，后续所有轮次始终遵循以下要求：第一，你生成的代码文件里面的函数，其圈复杂度不可高于2；第二，运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.scss', '.sass', '.less',\\n '.js', '.jsx', '.ts', '.tsx', '.py', '.java',\\n '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs',\\n '.php', '.rb', '.sh', '.json', '.xml', '.yaml', '.yml'\\n }\\n\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git, .vscode 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 忽略隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n _, ext = os.path.splitext(file)\\n # 判断是否为代码文件\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n # 检查文件名长度是否至少为 5\\n if len(file) < 5:\\n return False\\n\\n # 如果存在代码文件且均符合长度要求，或者暂未生成代码文件（无违反项），则判定为 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def check_python_ternary(content: str) -> bool:\\n \\\"\\\"\\\"检查 Python 代码中是否使用了三元运算符\\\"\\\"\\\"\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n return True\\n except Exception:\\n pass\\n return False\\n\\n def check_js_ts_html_ternary(content: str, ext: str) -> bool:\\n \\\"\\\"\\\"检查 JS/TS/HTML 等代码文件中是否使用了三元运算符\\\"\\\"\\\"\\n # 移除字符串字面量（兼容处理其中带转义符的情况）\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', '``', content)\\n \\n # 移除单行和多行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n # 如果是 HTML 文件，只抽取 <script> 标签中的内容进行检查\\n if ext == '.html':\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n js_content = \\\"\\\\n\\\".join(scripts)\\n else:\\n js_content = content\\n \\n if not js_content.strip():\\n return False\\n \\n # 查找独立的 '?' 以及其后与之相配对的 ':'\\n # 使用否定前瞻和否定后瞻过滤掉 '??', '?.', '?:' \\n # 同时要求 '?' 和 ':' 之间不能跨越分号 ';'\\n matches = re.finditer(r'(?<!\\\\?)\\\\?(?!\\\\?|\\\\.|:)([^;]*?):', js_content)\\n for _ in matches:\\n return True\\n \\n return False\\n\\n if not os.path.exists(workspace_path):\\n return True \\n\\n has_ternary = False\\n \\n # 待检测的前端文件扩展名范围\\n target_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', \\n '.html', '.vue', '.svelte', \\n '.mjs', '.cjs'\\n }\\n \\n # 遍历代码项目\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n filepath = os.path.join(root, file)\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 针对不同后缀使用对应解析策略\\n if ext == '.py':\\n if check_python_ternary(content):\\n has_ternary = True\\n break\\n elif ext in target_extensions:\\n if check_js_ts_html_ternary(content, ext):\\n has_ternary = True\\n break\\n except Exception:\\n continue\\n \\n if has_ternary:\\n break\\n \\n # 要求为：不能使用三元运算符。\\n # 所以如果存在三元运算符，则该要求未满足(False)，反之满足(True)。\\n return not has_ternary\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断给定的数字是否是质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 辅助函数：去除代码中的注释，减少正则误判\\n def remove_comments(code: str, ext: str) -> str:\\n if ext == '.py':\\n code = re.sub(r'#.*', '', code)\\n code = re.sub(r'\\\\'\\\\'\\\\'.*?\\\\'\\\\'\\\\'', '', code, flags=re.DOTALL)\\n code = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\".*?\\\\\\\"\\\\\\\"\\\\\\\"', '', code, flags=re.DOTALL)\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n elif ext == '.html':\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n return code\\n\\n # JavaScript中的一些关键字，防止类方法匹配正则误将其识别为函数名\\n js_keywords = {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'constructor', 'super', 'this'}\\n all_functions = []\\n\\n # 遍历项目文件\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 去除注释\\n content = remove_comments(content, ext)\\n\\n if ext == '.py':\\n # 匹配 Python 的函数定义\\n pattern = r'def\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\('\\n funcs = re.findall(pattern, content)\\n all_functions.extend(funcs)\\n \\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.html']:\\n # 1. 匹配普通的 function 声明\\n pattern1 = r'function\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\('\\n funcs1 = re.findall(pattern1, content)\\n all_functions.extend(funcs1)\\n\\n # 2. 匹配函数表达式与箭头函数赋值\\n pattern2 = r'(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)'\\n funcs2 = re.findall(pattern2, content)\\n all_functions.extend(funcs2)\\n\\n # 3. 粗略匹配类中的方法 (需排除关键字)\\n pattern3 = r'^\\\\s*(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+)?([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n funcs3 = re.findall(pattern3, content, re.MULTILINE)\\n for f_name in funcs3:\\n if f_name not in js_keywords:\\n all_functions.extend([f_name])\\n\\n # 如果没有任何函数，由于是全称量词（“定义的每一个函数”），可认为是恒成立的\\n if not all_functions:\\n return True\\n\\n # 检查每一个提取出的函数名，其长度是否为质数\\n for f in all_functions:\\n if not is_prime(len(f)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过 2 层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在的布局是固定等分的，对宽屏或特定场景不太友好。下一步，我们需要加入面板大小的拖拽调节功能。请在三个编辑器之间（垂直分割线），以及编辑器区域和下方预览区之间（水平分割线），添加可视化的拖拽手柄（Resizer）。当用户鼠标按下并拖动这些手柄时，能够动态改变对应面板的宽度或高度。注意拖拽时需要有视觉反馈（比如鼠标指针变成对应的拖拽箭头），并且要防止面板被拖得过小（设置一个最小宽高限制），拖拽结束后各个编辑器实例要能自适应重新计算布局大小（如调用 editor.layout()）。另外请注意，后续所有轮次始终遵循以下要求：第一，你输出的代码中禁止出现超过1层的循环嵌套；第二，生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"RESIZE\\\"。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs',\\n '.php', '.rb', '.sh', '.swift', '.kt', '.m', '.dart'\\n }\\n\\n # 遍历当前代码项目工作空间\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件的后缀名并统一转换为小写进行比对\\n _, ext = os.path.splitext(file)\\n ext = ext.lower()\\n \\n # 仅针对代码文件进行检查\\n if ext in code_extensions:\\n # 检查文件名（包含后缀）的字符数是否至少为 5 个字符\\n if len(file) < 5:\\n return False\\n \\n # 如果所有的代码文件名称均满足至少包含 5 个字符的条件，返回 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 辅助函数：专门用于剥离代码中的各类注释和字符串文本，防止干扰匹配\\n def clean_code(text: str, ext: str) -> str:\\n # 1. 移除多行注释 (针对类似JS/TS/Java等语言)\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.css', '.scss']:\\n text = re.sub(r'/\\\\*.*?\\\\*/', '', text, flags=re.DOTALL)\\n \\n # 2. 移除单行注释\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.css', '.scss']:\\n text = re.sub(r'//.*', '', text)\\n elif ext in ['.py', '.sh', '.rb', '.yml', '.yaml']:\\n text = re.sub(r'#.*', '', text)\\n \\n # 3. 移除各类字符串常量\\n if ext == '.py':\\n # Python独有的三引号字符串（允许换行）\\n text = re.sub(r'\\\"\\\"\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*?\\\"\\\"\\\"', '\\\"\\\"', text, flags=re.DOTALL)\\n text = re.sub(r\\\"'''(?:\\\\\\\\.|[^'\\\\\\\\])*?'''\\\", \\\"''\\\", text, flags=re.DOTALL)\\n \\n # 反引号字符串（主要用于JS/TS等语言，允许换行）\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*?`', '``', text, flags=re.DOTALL)\\n \\n # 普通单/双引号字符串（默认不跨行）\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*?\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\\\\\n])*?'\\\", \\\"''\\\", text)\\n \\n return text\\n\\n # 若项目目录不存在，默认无法验证或无生成代码\\n if not os.path.exists(workspace_path):\\n return False\\n\\n target_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.html', '.java', '.c', '.cpp', '.cs', '.css', '.scss'\\n }\\n\\n # 遍历当前工作空间所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 对当前文件代码进行“脱水”处理（清空注释与字符串内文本）\\n cleaned_content = clean_code(content, ext)\\n \\n if ext == '.py':\\n # 在Python语法中，同一行同时出现if和else基本属于三元表达式（或其生成器条件变体）\\n for line in cleaned_content.splitlines():\\n if re.search(r'\\\\bif\\\\b.*\\\\belse\\\\b', line):\\n return False\\n else:\\n # 对于JS/TS/Java/C/C++等语言，匹配类似 condition ? expr1 : expr2 的三元运算符\\n # 确保排除了 ?. (可选链) / ?? (空值合并) / ?: (类型声明及 Elvis 操作符变体)\\n # 限制匹配不能跨越语句分隔符 `;`\\n if re.search(r'(?<!\\\\?)\\\\?(?!\\\\?|\\\\.|:)[^;]*?:', cleaned_content):\\n return False\\n except Exception:\\n # 忽略个别因编码问题等无法读取的异常文件\\n continue\\n\\n # 若扫描遍布所有文件都没有匹配到三元运算符模式，说明已遵循限制要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 辅助判断函数：是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 提取 JS / TS 中的函数名\\n def get_js_functions(code: str) -> list:\\n functions = []\\n # 1. 匹配普通函数声明：function foo(...) 或 async function foo(...)\\n functions.extend(re.findall(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(', code))\\n \\n # 2. 匹配函数表达式或箭头函数：const/let/var foo = function(...) / (...) =>\\n pattern_expr = r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>)'\\n functions.extend(re.findall(pattern_expr, code))\\n \\n # 3. 匹配对象内方法：foo: function(...)\\n pattern_obj = r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?function\\\\b'\\n functions.extend(re.findall(pattern_obj, code))\\n \\n return functions\\n\\n # 提取 Python 中的函数名\\n def get_py_functions(code: str) -> list:\\n return re.findall(r'^[ \\\\t]*def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(', code, re.MULTILINE)\\n\\n # 提取 HTML 中的内联脚本并获取其函数名\\n def get_html_functions(code: str) -> list:\\n functions = []\\n script_blocks = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', code, re.IGNORECASE)\\n for block in script_blocks:\\n functions.extend(get_js_functions(block))\\n return functions\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n funcs = []\\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\\n funcs = get_js_functions(content)\\n elif ext == '.py':\\n funcs = get_py_functions(content)\\n elif ext in ['.html', '.htm']:\\n funcs = get_html_functions(content)\\n \\n # 判断每个函数名长度是否为质数\\n for func_name in funcs:\\n if not is_prime(len(func_name)):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"RESIZE\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前后端代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.ts', '.html', '.css', '.htm', \\n '.jsx', '.tsx', '.c', '.cpp', '.java', '.cs', '.go'\\n }\\n \\n # 正则表达式用于安全地提取注释：\\n # (?P<string>...) 优先匹配可能包含类似注释符号的各类字符串（三引号、双引号、单引号、反引号），避免被误认为注释\\n # (?P<comment>...) 匹配常见的各种代码注释格式\\n pattern = re.compile(\\n r'(?P<string>\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->|//.*?$|#.*?$)',\\n re.MULTILINE\\n )\\n \\n file_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取文件中的所有注释内容\\n comments_text = \\\"\\\"\\n for match in pattern.finditer(content):\\n if match.group('comment'):\\n comments_text += match.group('comment') + \\\"\\\\n\\\"\\n \\n # 统计提取的注释中 \\\"RESIZE\\\" 关键词出现的次数\\n count = comments_text.count(\\\"RESIZE\\\")\\n \\n # 若不是恰好等于 2 个，则判断为未满足要求\\n if count != 2:\\n return False\\n \\n # 如果项目中没有任何代码文件，或者存在不可读代码文件，默认当作未满足要求；\\n # 若所有读取成功的代码文件注释里都恰好包含了 2 个 \\\"RESIZE\\\"，则返回 True\\n return file_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"很多时候写代码需要引入第三方的库，比如 jQuery 或者 Bootstrap。现在我们来完善顶部的“设置”按钮。点击它后，需要弹出一个居中的模态框（Modal）。模态框里分为两个标签页（Tabs）：一个是“外部 CSS”，另一个是“外部 JS”。每个标签页里允许用户动态添加、编辑和删除多个外部资源的 CDN 链接（提供输入框和增删按钮）。当用户在模态框点击“保存并应用”后，你的实时预览生成逻辑需要更新，将这些外部链接分别转化为 <link> 和 <script> 标签，并注入到 iframe 渲染内容的对应位置。另外请注意，后续所有轮次始终遵循以下要求：第一，生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"CDN\\\"；第二，在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，说明未生成任何文件，未违反要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 结合构建类似 CodePen 网站的需求，定义常见的代码文件后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.json', '.py', '.md', '.sh'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录，避免检查如 .git 下的内部文件\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 忽略隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n # 提取文件的扩展名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断该文件是否为代码文件\\n if ext in code_extensions:\\n # 检查文件名的字符数（包含扩展名），如果小于5个字符，则不满足要求\\n if len(file) < 5:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def remove_strings_and_comments(code: str) -> str:\\n \\\"\\\"\\\"剥离 JS/TS 语言代码中的字符串字面量和注释\\\"\\\"\\\"\\n result = []\\n i = 0\\n n = len(code)\\n in_single_quote = False\\n in_double_quote = False\\n in_backtick = False\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < n:\\n if in_line_comment:\\n if code[i] == '\\\\n':\\n in_line_comment = False\\n result.append('\\\\n')\\n i += 1\\n continue\\n if in_block_comment:\\n if i < n - 1 and code[i:i+2] == '*/':\\n in_block_comment = False\\n i += 2\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n continue\\n \\n if in_single_quote:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n continue\\n if code[i] == \\\"'\\\":\\n in_single_quote = False\\n i += 1\\n continue\\n \\n if in_double_quote:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n continue\\n if code[i] == '\\\"':\\n in_double_quote = False\\n i += 1\\n continue\\n \\n if in_backtick:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n continue\\n if code[i] == '`':\\n in_backtick = False\\n i += 1\\n continue\\n \\n # 检查字符串或注释的起始位置\\n if code[i:i+2] == '//':\\n in_line_comment = True\\n i += 2\\n continue\\n if code[i:i+2] == '/*':\\n in_block_comment = True\\n i += 2\\n continue\\n if code[i] == \\\"'\\\":\\n in_single_quote = True\\n i += 1\\n continue\\n if code[i] == '\\\"':\\n in_double_quote = True\\n i += 1\\n continue\\n if code[i] == '`':\\n in_backtick = True\\n i += 1\\n continue\\n \\n result.append(code[i])\\n i += 1\\n \\n return ''.join(result)\\n\\n def has_js_ternary(code: str) -> bool:\\n \\\"\\\"\\\"检测纯代码中是否存在 JS/TS 的三元运算符\\\"\\\"\\\"\\n clean_code = remove_strings_and_comments(code)\\n # 移除 JS/TS 中常见的含 ? 其他合法语法的干扰\\n clean_code = clean_code.replace('??=', '')\\n clean_code = clean_code.replace('??', '')\\n clean_code = clean_code.replace('?.', '')\\n clean_code = clean_code.replace('?:', '')\\n \\n # 查找是否存在问号 ? 以及随后的冒号 : （同一语句内不允许跨分号）\\n if re.search(r'\\\\?[^;]*:', clean_code):\\n return True\\n return False\\n\\n def has_python_ternary(code: str) -> bool:\\n \\\"\\\"\\\"检测 Python 代码中是否存在三元运算符(IfExp)\\\"\\\"\\\"\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n return True\\n except Exception:\\n pass\\n return False\\n\\n def extract_scripts_from_html(html_code: str) -> str:\\n \\\"\\\"\\\"提取 HTML 内的 script 脚本块\\\"\\\"\\\"\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', html_code, re.IGNORECASE | re.DOTALL)\\n return '\\\\n'.join(scripts)\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n \\n # 分文件类型执行安全检测\\n if ext == '.py':\\n if has_python_ternary(content):\\n return False\\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if has_js_ternary(content):\\n return False\\n elif ext in ['.html', '.htm', '.vue']:\\n scripts_code = extract_scripts_from_html(content)\\n if has_js_ternary(scripts_code):\\n return False\\n\\n # 若没有任何文件包含三元运算符，则视作通过验证\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 清洗 JS 字符串和注释，防止正则表达式误匹配到字符串内容或注释里写的伪代码\\n def clean_js(content: str) -> str:\\n content = re.sub(r'([\\\"\\\\'`])(?:(?=(\\\\\\\\?))\\\\2.)*?\\\\1', '\\\"\\\"', content, flags=re.DOTALL)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n return content\\n\\n # 预设 JS 关键字黑名单，防止在匹配对象方法或类方法时将流程控制语句误当成函数定义\\n reserved = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', \\n 'class', 'import', 'export', 'constructor', 'typeof', 'instanceof', \\n 'void', 'delete', 'new', 'super', 'this', 'else', 'try', 'finally', \\n 'do', 'with', 'yield', 'await', 'eval', 'arguments', 'get', 'set'\\n }\\n\\n # 从 JS / TS 内容中提取所有函数名\\n def get_js_funcs(content: str) -> set:\\n content = clean_js(content)\\n names = set()\\n \\n # 1. 匹配函数声明：function fnName(...)\\n for match in re.finditer(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(', content):\\n names.add(match.group(1))\\n \\n # 2. 匹配赋值型和箭头函数：const fnName = function() 或者 const fnName = () => ...\\n pattern_assign = r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'\\n for match in re.finditer(pattern_assign, content):\\n names.add(match.group(1))\\n \\n # 3. 匹配对象或类中的简写方法：fnName() { ... }\\n for match in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', content):\\n name = match.group(1)\\n if name not in reserved:\\n names.add(name)\\n \\n # 4. 匹配对象属性式的函数定义：fnName: function(...)\\n for match in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?function\\\\b', content):\\n name = match.group(1)\\n if name not in reserved:\\n names.add(name)\\n \\n return names\\n\\n # 从 Python 内容中利用 AST 解析提取所有函数名\\n def get_py_funcs(content: str) -> set:\\n names = set()\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n names.add(node.name)\\n except Exception:\\n pass\\n return names\\n\\n # 如果工作目录不存在，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n names = set()\\n \\n # 根据后缀名进入对应的函数提取逻辑\\n if ext == '.py':\\n names = get_py_funcs(content)\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n names = get_js_funcs(content)\\n elif ext in ['.html', '.htm', '.vue']:\\n # 寻找所有的 script 脚本块并做二次提取\\n for match in re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE):\\n names.update(get_js_funcs(match.group(1)))\\n \\n # 对每一个提取到的函数名称判断长度是否为质数\\n for name in names:\\n if not is_prime(len(name)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"CDN\\\"\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_extensions = {\\n '.py', \\n '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs', \\n '.html', '.htm', '.vue', '.svelte', \\n '.css', '.scss', '.sass', '.less'\\n }\\n\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n \\n # JS/TS类文件注释提取\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs']:\\n pattern = r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' \\\\\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//[^\\\\r\\\\n]*)'\\n for match in re.finditer(pattern, content):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n \\n # CSS/样式类文件注释提取\\n elif ext in ['.css', '.scss', '.sass', '.less']:\\n if ext in ['.scss', '.sass', '.less']:\\n pattern = r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')' \\\\\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//[^\\\\r\\\\n]*)'\\n else:\\n pattern = r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')' \\\\\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n for match in re.finditer(pattern, content):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n\\n # HTML/模板类文件注释提取\\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\\n # 提取顶层 HTML 注释\\n html_comments = re.findall(r'<!--[\\\\s\\\\S]*?-->', content)\\n comments.extend(html_comments)\\n \\n # 提取 script 内部 JS 注释\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n js_pattern = r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' \\\\\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//[^\\\\r\\\\n]*)'\\n for script in scripts:\\n for match in re.finditer(js_pattern, script):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n \\n # 提取 style 内部 CSS 注释\\n styles = re.findall(r'<style.*?>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n css_pattern = r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')' \\\\\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n for style in styles:\\n for match in re.finditer(css_pattern, style):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n\\n # Python文件注释提取\\n elif ext == '.py':\\n pattern = r'(?P<comment>\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|#[^\\\\r\\\\n]*)' \\\\\\n r'|(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n for match in re.finditer(pattern, content):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n\\n # 计算当前文件内所有的注释中包含 'CDN' 的总次数\\n cdn_count = sum(comment.count('CDN') for comment in comments)\\n \\n # 若任意一份代码文件中 'CDN' 出现次数不为2，则直接不满足要求\\n if cdn_count != 2:\\n return False\\n\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 定义常见代码文件的扩展名集合\\n valid_extensions = {\\n '.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rb', '.php', '.swift', '.json', '.sh'\\n }\\n \\n has_code_file = False\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选出代码文件\\n if ext in valid_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 忽略无法解析的特殊字符错误，以确保正确读取所有行\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 统计总行数（涵盖空行）\\n num_lines = sum(1 for _ in f)\\n \\n # 判断行数是否为质数\\n if not is_prime(num_lines):\\n return False\\n except Exception:\\n # 如果打开或读取文件过程中发生未预期的异常，认定不满足要求\\n return False\\n \\n # 确保至少存在一个代码文件，且均通过了质数行数验证\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"我们还需要一个开发者控制台，这对于写 JS 太重要了。请在界面的最底部（或者预览区下方）添加一个可展开/折叠的“Console”面板。核心挑战来了：你需要拦截 iframe 内部运行时的 console.log、console.warn 和 console.error，将它们捕获并显示在这个“Console”面板中。不同级别的日志要有不同的文字颜色或背景色区分；如果输出的是对象或数组，要在面板里将其格式化为易于阅读的 JSON 字符串结构展示出来。另外请注意，后续所有轮次始终遵循以下要求：第一，原要求生成的每一份代码文件注释区域必须包含恰好2个关键词\\\"CDN\\\"的规则不再生效，替换为生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"CONSOLE\\\"；第二，后续执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的扩展名集合（涵盖Web开发及常见编程语言）\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', \\n '.rs', '.rb', '.php', '.swift', '.kt', '.sh', \\n '.json', '.xml', '.yaml', '.yml'\\n }\\n\\n # 遍历工作区内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名\\n _, ext = os.path.splitext(file)\\n \\n # 仅针对代码文件进行检查\\n if ext.lower() in code_extensions:\\n # 判断包含扩展名的完整文件名字符数是否小于 5\\n if len(file) < 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"import os\\nimport ast\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def remove_comments_and_strings_js(code: str) -> str:\\n result = []\\n i = 0\\n n = len(code)\\n in_string = False\\n string_char = ''\\n in_block_comment = False\\n in_line_comment = False\\n \\n while i < n:\\n if in_block_comment:\\n if code[i:i+2] == '*/':\\n in_block_comment = False\\n i += 2\\n continue\\n i += 1\\n continue\\n \\n if in_line_comment:\\n if code[i] == '\\\\n':\\n in_line_comment = False\\n result.append('\\\\n') # 保留换行符以维持代码基础结构\\n i += 1\\n continue\\n i += 1\\n continue\\n \\n if in_string:\\n if code[i] == '\\\\\\\\': # 跳过转义字符\\n i += 2\\n continue\\n if code[i] == string_char:\\n in_string = False\\n i += 1\\n continue\\n i += 1\\n continue\\n \\n if code[i:i+2] == '/*':\\n in_block_comment = True\\n i += 2\\n continue\\n \\n if code[i:i+2] == '//':\\n in_line_comment = True\\n i += 2\\n continue\\n \\n if code[i] in '\\\"\\\\'`':\\n in_string = True\\n string_char = code[i]\\n i += 1\\n continue\\n \\n result.append(code[i])\\n i += 1\\n \\n return ''.join(result)\\n\\n def check_ternary_js(code: str) -> bool:\\n stripped = remove_comments_and_strings_js(code)\\n # 移除 JS/TS 中包含 '?' 但不是三元运算符的常见语法糖和类型定义\\n stripped = stripped.replace('??', '')\\n stripped = stripped.replace('?.', '')\\n stripped = stripped.replace('?:', '')\\n \\n # 查找 '?' 后面跟着 ':' 的结构\\n # 限制条件：两者之间不能出现分号 ';' (语句结束)，且跨越的换行符不能超过 3 个\\n if re.search(r'\\\\?(?:[^;\\\\n]*\\\\n){0,3}[^;\\\\n]*:', stripped):\\n return True\\n return False\\n\\n def check_ternary_python(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n return True\\n except SyntaxError:\\n pass\\n return False\\n\\n def check_ternary_html(code: str) -> bool:\\n # 仅提取 HTML 中的 script 标签内容进行 JS 三元运算符检测\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', code, flags=re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n if check_ternary_js(script):\\n return True\\n return False\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 依据后缀判断文件类型并选用对应的三元运算符解析策略\\n if ext == '.py':\\n if check_ternary_python(content):\\n return False # 发现三元运算符，违反了限制条件\\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if check_ternary_js(content):\\n return False\\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\\n if check_ternary_html(content):\\n return False\\n except Exception:\\n # 忽略读取失败的文件，避免非文本文件导致的系统级中断\\n continue\\n\\n # 没有发现三元运算符，满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 辅助函数：判断数字是否为质数\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 辅助函数：移除代码中的注释以防止正则误判\\n def remove_comments(code: str, ext: str) -> str:\\n if ext == '.py':\\n code = re.sub(r'#.*', '', code)\\n code = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '', code)\\n code = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", '', code)\\n else:\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n code = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', code)\\n return code\\n\\n # 匹配 Python 函数\\n py_func_pattern = re.compile(r\\\"def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(\\\")\\n \\n # 匹配 JS/TS 标准函数\\n js_func_pattern = re.compile(r\\\"function\\\\s*\\\\*?\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(\\\")\\n \\n # 匹配 JS/TS 箭头函数赋值 (例如: const myFunc = () => )\\n js_arrow_pattern = re.compile(r\\\"(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>\\\")\\n \\n # 匹配 JS/TS 匿名/具名函数赋值 (例如: const myFunc = function() )\\n js_expr_pattern = re.compile(r\\\"(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\s*\\\\*?(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(\\\")\\n \\n # 匹配 JS/TS 类方法 (例如: render() { 或 async fetch() {)\\n js_method_pattern = re.compile(r\\\"^\\\\s*(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^{]+)?\\\\{\\\", re.MULTILINE)\\n\\n # 排除 JS 中的常见控制流关键字（防止类方法正则发生误匹配）\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'function', \\n 'return', 'yield', 'await', 'import', 'export', 'default'\\n }\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n valid_extensions = ('.js', '.ts', '.jsx', '.tsx', '.mjs', '.cjs', '.py', '.html', '.htm', '.vue', '.svelte')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n content = remove_comments(content, ext)\\n names = []\\n \\n if ext == '.py':\\n names.extend(py_func_pattern.findall(content))\\n else:\\n names.extend(js_func_pattern.findall(content))\\n names.extend(js_arrow_pattern.findall(content))\\n names.extend(js_expr_pattern.findall(content))\\n \\n methods = js_method_pattern.findall(content)\\n for m in methods:\\n if m not in keywords:\\n names.append(m)\\n \\n # HTML/Vue 文件中可能内嵌了 Python（如通过特定模板/PyScript等），也作为备用检查一并包含\\n if ext in ('.html', '.htm', '.vue', '.svelte'):\\n names.extend(py_func_pattern.findall(content))\\n\\n # 遍历文件内的所有函数名，验证命名长度是否为质数\\n for name in names:\\n if not is_prime(len(name)):\\n return False\\n\\n # 若没有任何函数违反规定（或没有定义任何函数），则返回True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"CONSOLE\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见代码文件类型拓展名集合\\n valid_exts = {\\n '.js', '.ts', '.jsx', '.tsx', \\n '.css', '.scss', '.less', \\n '.html', '.htm', \\n '.py', \\n '.java', '.c', '.cpp', '.cs', '.go', \\n '.sh', '.rb'\\n }\\n \\n def get_comments(file_path):\\n ext = os.path.splitext(file_path)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return \\\"\\\"\\n \\n comments = []\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', '.java', '.c', '.cpp', '.cs', '.go']:\\n # 采用正则表达式优先匹配字符串，如果匹配到单行/多行注释则收集\\n pattern = r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`|/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*)'\\n tokens = re.findall(pattern, content)\\n for token in tokens:\\n if token.startswith('//') or token.startswith('/*'):\\n comments.append(token)\\n \\n elif ext in ['.html', '.htm']:\\n # HTML文件首先收集自带的 HTML 注释\\n html_comments = re.findall(r'<!--[\\\\s\\\\S]*?-->', content)\\n comments.extend(html_comments)\\n \\n # 然后分别提取 <script> 和 <style> 块中的代码进行 JS 和 CSS 的注释匹配\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n styles = re.findall(r'<style[^>]*>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n \\n for script in scripts:\\n pattern = r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`|/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*)'\\n tokens = re.findall(pattern, script)\\n for token in tokens:\\n if token.startswith('//') or token.startswith('/*'):\\n comments.append(token)\\n for style in styles:\\n pattern = r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*)'\\n tokens = re.findall(pattern, style)\\n for token in tokens:\\n if token.startswith('//') or token.startswith('/*'):\\n comments.append(token)\\n \\n elif ext in ['.py']:\\n # Python匹配包含多行字符串注释（Docstring）、单/双引号及 # 单行注释\\n pattern = r'(\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|#.*)'\\n tokens = re.findall(pattern, content)\\n for token in tokens:\\n if token.startswith('#') or token.startswith('\\\"\\\"\\\"') or token.startswith(\\\"'''\\\"):\\n comments.append(token)\\n \\n elif ext in ['.sh', '.rb']:\\n # Shell, Ruby等以 # 为注释\\n pattern = r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|#.*)'\\n tokens = re.findall(pattern, content)\\n for token in tokens:\\n if token.startswith('#'):\\n comments.append(token)\\n \\n return \\\"\\\\n\\\".join(comments)\\n\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉常见的非用户编写代码的文件夹\\n if any(ignored in root for ignored in ['.git', 'node_modules', '__pycache__', '.idea', '.vscode']):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 提取出本文件中所有的注释内容\\n comments_text = get_comments(file_path)\\n \\n # 统计关键词 \\\"CONSOLE\\\" 的出现次数\\n count = comments_text.count(\\\"CONSOLE\\\")\\n \\n # 必须恰好为2个，否则验证失败\\n if count != 2:\\n return False\\n \\n # 如果没有扫描到任何代码文件，同样视为不符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n for i in range(2, int(n**0.5) + 1):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 针对当前前端 Web 项目推断可能涉及的代码文件扩展名\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符进行拆分，这样能够精准统计涵盖空行在内的IDE展示总行数\\n lines_count = len(content.split('\\\\n'))\\n \\n # 检查总行数是否为质数\\n if not is_prime(lines_count):\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果由于编码等原因读取失败，则跳过\\n continue\\n\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"现在为了提升编码体验，我们来加入代码格式化功能。请通过 CDN 引入 Prettier 的核心库以及对应的 HTML/CSS/JS 解析插件。在顶部导航栏增加一个“格式化”按钮，当用户点击时，同时对当前三个编辑器里的代码进行美化排版，并将格式化后的代码回填到编辑器中。同时，请绑定键盘快捷键（Ctrl+S 或 Cmd+S），用户按下这个组合键时，不仅阻止浏览器的默认保存行为，也要触发这个代码格式化动作。另外请注意，后续所有轮次始终遵循以下要求：第一，原要求生成的每一份代码文件注释区域必须包含恰好2个关键词\\\"CONSOLE\\\"的规则不再生效，替换为生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"FORMAT\\\"；第二，请确保生成的代码文件中不包含多层if。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据多轮指令内容，推断可能的代码文件扩展名\\n # 项目为Web代码编辑器，涉及 HTML, CSS, JS，且提到了Python的特定模块要求\\n code_extensions = {'.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\\n \\n # 遍历代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n # 检查文件名长度是否至少包含 5 个字符\\n if len(file) < 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def remove_comments_and_strings(text: str) -> str:\\n # 正则表达式过滤掉常见的 JS/TS 字符串、单行注释及多行注释，避免对提取出来的 ? 和 : 产生干扰\\n pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.DOTALL | re.MULTILINE\\n )\\n return pattern.sub('', text)\\n\\n def check_js(code: str) -> bool:\\n clean_code = remove_comments_and_strings(code)\\n # 寻找三元运算符的核心特征：不在注释和字符串中，由独立的 ? 和 : 组成，且中间没有出现分号或大括号\\n # (?<!\\\\?)\\\\?(?![?.:]) 确保问号是单独作为运算符出现的，排除了 ??, ?., ?: 的情况\\n if re.search(r'(?<!\\\\?)\\\\?(?![?.:])([^;{}]+?)(?<!:):(?!:)', clean_code):\\n return False\\n return True\\n\\n def check_python(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n # Python 语言中，三元运算符（If Expression）映射为 ast.IfExp 节点\\n if isinstance(node, ast.IfExp):\\n return False\\n except Exception:\\n # 解析失败意味着它大概率不是一个合法的 Python 代码（或含有语法错误），安全起见直接放行\\n pass\\n return True\\n\\n def check_html(code: str) -> bool:\\n # 提取 HTML 文件中所有的 <script> 内容\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', code, re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n if not check_js(script):\\n return False\\n return True\\n\\n def check_general_code(code: str, lang: str) -> bool:\\n lang = lang.lower()\\n if lang in ['python', 'py']:\\n return check_python(code)\\n elif lang in ['javascript', 'js', 'typescript', 'ts', 'jsx', 'tsx', 'vue', 'svelte']:\\n return check_js(code)\\n elif lang in ['html', 'htm']:\\n return check_html(code)\\n else:\\n # 遇到未知语言代码，双重检测以尽可能避免漏判\\n if not check_python(code):\\n return False\\n if not check_js(code):\\n return False\\n return True\\n\\n # 1. 检查物理存在于 workspace_path 中的文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if ext == '.py':\\n if not check_python(content):\\n return False\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.svelte']:\\n if not check_js(content):\\n return False\\n elif ext in ['.html', '.htm']:\\n if not check_html(content):\\n return False\\n else:\\n if not check_python(content):\\n return False\\n if not check_js(content):\\n return False\\n\\n # 2. 检查 response 回复中携带的 Markdown 代码块\\n code_blocks = re.findall(r'```([a-zA-Z]*)\\\\n(.*?)\\\\n```', response, re.DOTALL)\\n for lang, code in code_blocks:\\n if not check_general_code(code, lang.strip()):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"def check_function_name_prime_length(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n # 辅助函数：判断给定数字是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 解析 Python 代码中的函数\\n def get_py_funcs(code: str):\\n funcs = []\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n funcs.append(node.name)\\n except Exception:\\n pass\\n return funcs\\n\\n # 解析 JavaScript/TypeScript 等代码中的函数\\n def get_js_funcs(code: str):\\n funcs = set()\\n \\n # 1. 常规函数：function foo(...)\\n for m in re.finditer(r'\\\\bfunction\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\(', code):\\n funcs.add(m.group(1))\\n \\n # 2. 赋值形式：foo = function(...) 或 foo : function(...)\\n for m in re.finditer(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*[:=]\\\\s*function\\\\s*\\\\(', code):\\n funcs.add(m.group(1))\\n \\n # 3. 箭头函数：foo = (...) => 或 foo : (...) => 或 foo = arg =>\\n for m in re.finditer(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*[:=]\\\\s*(?:\\\\([^)]*\\\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=>', code):\\n funcs.add(m.group(1))\\n \\n # 4. 类或对象方法：foo(...) {\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'constructor',\\n 'super', 'return', 'import', 'export', 'typeof', 'new', 'get', 'set',\\n 'do', 'with', 'yield', 'await', 'class', 'let', 'const', 'var', 'default'\\n }\\n for m in re.finditer(r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', code):\\n name = m.group(1)\\n if name not in keywords:\\n funcs.add(name)\\n \\n return list(funcs)\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的隐藏文件夹及第三方依赖目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n funcs = []\\n if ext == '.py':\\n funcs = get_py_funcs(content)\\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n funcs = get_js_funcs(content)\\n elif ext in ['.html', '.htm']:\\n # 提取 HTML 文件内的所有 <script> 标签包裹的代码内容\\n script_matches = re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for m in script_matches:\\n funcs.extend(get_js_funcs(m.group(1)))\\n \\n # 若含有任何函数定义，立刻校验函数名长度是否为质数\\n for func_name in funcs:\\n if not is_prime(len(func_name)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"FORMAT\\\"\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx'}\\n code_files = []\\n \\n # 遍历项目目录，收集所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.lower().endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果项目中没有生成相关的代码文件，无法判断是否满足“每一份”的要求，返回 False\\n if not code_files:\\n return False\\n \\n # 综合正则表达式：\\n # group 1, 2, 3 匹配双引号、单引号、反引号包裹的字符串（跳过字符串内部的转义符）\\n # group 4 匹配 HTML 注释、多行注释、单行注释\\n # 通过这种交替匹配的方式，可以安全地跳过位于字符串内部的伪注释符号\\n combined_pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|'\\n r'(\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|'\\n r'(`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(<!--[\\\\s\\\\S]*?-->|/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*)'\\n )\\n \\n # 检查每一份代码文件\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果文件无法按 UTF-8 读取，为保障评测顺利可选择跳过或视作不合规\\n # 此处稳妥起见，无法读取视为异常情况，直接返回 False\\n return False\\n \\n comments_content = []\\n for match in combined_pattern.finditer(content):\\n # 如果匹配到了 group 4，则说明是注释内容\\n if match.group(4):\\n comments_content.append(match.group(4))\\n \\n all_comments_text = '\\\\n'.join(comments_content)\\n \\n # 统计当前文件所有注释中 \\\"FORMAT\\\" 出现的次数\\n format_count = all_comments_text.count(\\\"FORMAT\\\")\\n \\n # 如果不是恰好包含 2 个，则不满足要求\\n if format_count != 2:\\n return False\\n \\n # 所有代码文件均满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n i = 3\\n while i * i <= n:\\n if n % i == 0:\\n return False\\n i += 2\\n return True\\n\\n # 定义涵盖前端与常见后端脚本的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.js', '.html', '.css', '.ts', '.jsx', \\n '.tsx', '.json', '.vue', '.sh', '.bash', \\n '.java', '.c', '.cpp', '.h', '.go', '.rs'\\n }\\n \\n found_any_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前代码项目工作空间\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按换行符切分以准确统计行数，涵盖空行逻辑\\n line_count = len(content.split('\\\\n'))\\n except Exception:\\n # 对于非utf-8编码等读取异常的文件，跳过处理\\n continue\\n \\n found_any_code_file = True\\n # 若存在行数不为质数的代码文件，则不满足要求\\n if not is_prime(line_count):\\n return False\\n \\n # 如果没有找到任何代码文件，则无法验证是否满足要求，返回 False\\n # 如果找到了代码文件并且所有文件行数都为质数，则返回 True\\n return found_any_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中不包含多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"最后一轮，我们要实现状态的持久化和 URL 分享功能。首先，用户的代码和外部资源配置需要在每次变动时自动存入 localStorage，刷新页面后能自动恢复上次编辑的状态。其次，点击顶部的“分享”按钮时，请将当前的 HTML、CSS、JS 代码以及外部资源列表，通过 Base64（或 LZString 压缩算法）序列化拼接到当前页面 URL 的 Hash (#) 中，然后弹出一个 toast 提示“链接已复制到剪贴板”。当别人通过带有 Hash 数据的 URL 访问我们的页面时，页面初始化要优先从 URL 解析并恢复这些代码和配置，而不是从 localStorage 中读取。另外请注意，后续所有轮次始终遵循以下要求：第一，原要求生成的每一份代码文件注释区域必须包含恰好2个关键词\\\"FORMAT\\\"的规则不再生效，替换为生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"SHARE\\\"；第二，生成的代码里，“=”号的前后必须各自包含两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你所输出的任何一个代码文件，其包含的注释都应当是全英文的\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的所有代码文件，其文件名必须至少包含 5 个字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n\\n workspace = Path(workspace_path)\\n if not workspace.exists():\\n return False\\n\\n # 针对前端项目和常见的代码相关文件定义扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.py', '.md'}\\n \\n has_code_file = False\\n \\n # 递归遍历工作区下的所有文件\\n for file_path in workspace.rglob('*'):\\n if file_path.is_file() and file_path.suffix.lower() in code_extensions:\\n has_code_file = True\\n \\n # file_path.name 会获取包含扩展名的文件名，例如 \\\"index.html\\\"\\n if len(file_path.name) < 5:\\n return False\\n \\n # 必须保证至少输出了一个代码文件，且全部通过了检查才能返回True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"限制条件：你生成的代码文件中不能使用三元运算符\",\n \"validation_code\": \"import os\\nimport ast\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def remove_comments_and_strings_js(code: str) -> str:\\n \\\"\\\"\\\"\\n 移除 JavaScript 代码中的注释和字符串内容，以避免在这些区域内误匹配 ? 和 :。\\n \\\"\\\"\\\"\\n result = []\\n i = 0\\n n = len(code)\\n in_single_quote = False\\n in_double_quote = False\\n in_template_literal = False\\n in_single_comment = False\\n in_multi_comment = False\\n \\n while i < n:\\n # 单行注释处理\\n if in_single_comment:\\n if code[i] == '\\\\n':\\n in_single_comment = False\\n i += 1\\n continue\\n # 多行注释处理\\n if in_multi_comment:\\n if i + 1 < n and code[i:i+2] == '*/':\\n in_multi_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n # 单引号字符串处理\\n if in_single_quote:\\n if code[i] == '\\\\\\\\': # 转义跳过\\n i += 2\\n continue\\n if code[i] == \\\"'\\\":\\n in_single_quote = False\\n i += 1\\n continue\\n # 双引号字符串处理\\n if in_double_quote:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n continue\\n if code[i] == '\\\"':\\n in_double_quote = False\\n i += 1\\n continue\\n # 模板字符串处理\\n if in_template_literal:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n continue\\n if code[i] == '`':\\n in_template_literal = False\\n i += 1\\n continue\\n \\n # 检测进入注释\\n if code[i] == '/' and i + 1 < n:\\n if code[i+1] == '/':\\n in_single_comment = True\\n i += 2\\n continue\\n elif code[i+1] == '*':\\n in_multi_comment = True\\n i += 2\\n continue\\n \\n # 检测进入字符串\\n if code[i] == \\\"'\\\":\\n in_single_quote = True\\n i += 1\\n continue\\n if code[i] == '\\\"':\\n in_double_quote = True\\n i += 1\\n continue\\n if code[i] == '`':\\n in_template_literal = True\\n i += 1\\n continue\\n \\n # 保存非字符串和非注释的代码本体内容\\n result.append(code[i])\\n i += 1\\n \\n return \\\"\\\".join(result)\\n\\n def check_js_ternary(code: str) -> bool:\\n \\\"\\\"\\\"\\n 检查去除了注释和字符串的 JS 纯代码中是否包含三元运算符。\\n \\\"\\\"\\\"\\n clean = remove_comments_and_strings_js(code)\\n i = 0\\n n = len(clean)\\n while i < n:\\n if clean[i] == '?':\\n next_char = clean[i+1] if i+1 < n else ''\\n \\n # 排除可选链 ?.\\n if next_char == '.':\\n i += 2\\n continue\\n \\n # 排除空值合并 ??\\n if next_char == '?':\\n i += 2\\n continue\\n \\n # 排除 TypeScript 可选属性/参数 ?:\\n if next_char == ':':\\n i += 2\\n continue\\n \\n # 在排除了其他语法后，如果在同一语句结束前找到了 :，则判定为三元运算符\\n j = i + 1\\n while j < n:\\n if clean[j] == ':':\\n return True\\n if clean[j] == ';':\\n break\\n j += 1\\n i += 1\\n return False\\n\\n def extract_script_content(html_code: str) -> str:\\n \\\"\\\"\\\"\\n 提取 HTML 文件中的 <script> 标签包裹的 JavaScript 内容。\\n \\\"\\\"\\\"\\n pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n scripts = pattern.findall(html_code)\\n return \\\"\\\\n\\\".join(scripts)\\n\\n # 遍历工作区下的所有代码文件进行检查\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n\\n if ext == '.py':\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n return False # 发现了 Python 中的三元运算符 (if-else 表达式)\\n except Exception:\\n pass\\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if check_js_ternary(code):\\n return False # 发现了 JS/TS 中的三元运算符\\n elif ext in ['.html', '.htm', '.vue']:\\n script_code = extract_script_content(code)\\n if script_code and check_js_ternary(script_code):\\n return False # 发现了 HTML 脚本中的三元运算符\\n\\n # 若所有代码文件都排除了包含三元运算符的可能性，则满足限制条件\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"务必确保生成的代码在处理外部 I/O 操作、网络请求以及数据格式转换时带有异常捕获功能，如果出现异常则需返回错误信息\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成代码时，处理文件路径须用pathlib.Path，记录日志须用logging模块，不得使用os.path和print\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中定义的每一个函数，其命名长度都必须限制为恰好是质数\",\n \"validation_code\": \"import os\\nimport ast\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断一个数字是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 用于匹配 JS/TS/HTML(script) 中的函数命名的正则表达式\\n # 匹配普通函数定义：function foo(...) 或 function* foo(...)\\n js_pattern1 = re.compile(r'\\\\bfunction\\\\s*\\\\*?\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n # 匹配变量赋值形式的函数：const foo = function(...) 或 const foo = (...) =>\\n js_pattern2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)')\\n # 匹配对象内部的方法定义形式：foo: function(...)\\n js_pattern3 = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?function\\\\b')\\n\\n # 遍历生成的代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅筛选常见的代码及 Web 文件类型\\n if ext not in ['.py', '.js', '.ts', '.html', '.jsx', '.tsx', '.vue', '.svelte']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n # Python 文件：通过抽象语法树提取函数命名\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n func_name = node.name\\n if not is_prime(len(func_name)):\\n return False\\n except Exception:\\n pass\\n else:\\n # Web 文件：通过正则匹配提取函数命名\\n names1 = js_pattern1.findall(content)\\n names2 = js_pattern2.findall(content)\\n names3 = js_pattern3.findall(content)\\n \\n # 遍历所有找到的函数名称\\n for func_name in (names1 + names2 + names3):\\n if not is_prime(len(func_name)):\\n return False\\n\\n # 若所有函数命名长度均为质数（或没有定义函数），则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中禁止出现超过1层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件里面的函数，其圈复杂度不可高于2\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"操作文件时尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"运行诸如测试、安装或构建等 bash 命令时，务必在命令的最前面加上 `timeout 30s` 包装，以防进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"生成的每一份代码文件中，注释区域必须包含恰好2个关键词\\\"SHARE\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义前端与常见可能生成的代码文件的后缀集合\\n code_extensions = {\\n '.py', '.sh', \\n '.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp',\\n '.html', '.htm'\\n }\\n\\n found_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关目录，避免扫描测试用例生成的环境隔离文件\\n for ignore_dir in ['.git', 'node_modules', '__pycache__', 'venv', '.venv', 'dist', 'build']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果生成的文件无法以普通的 UTF-8 文本读取，无法正确验证注释，判作失败\\n return False\\n \\n comments = []\\n \\n # 针对 C-style 注释类型（如 js, css, java 等）\\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp'}:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<comment>//[^\\\\n]*|/\\\\*[\\\\s\\\\S]*?\\\\*/)',\\n re.DOTALL\\n )\\n for match in pattern.finditer(content):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n \\n # 针对 Python 和 Shell 类型单行注释\\n elif ext in {'.py', '.sh'}:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\')'\\n r'|(?P<comment>#[^\\\\n]*)',\\n re.DOTALL\\n )\\n for match in pattern.finditer(content):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n # Python中的多行字符串如果做文档字符串，通常也视为一种注释\\n if ext == '.py' and match.group('string'):\\n s = match.group('string')\\n if s.startswith('\\\"\\\"\\\"') or s.startswith(\\\"'''\\\"):\\n comments.append(s)\\n \\n # 针对 HTML 类型的多重嵌套注释\\n elif ext in {'.html', '.htm'}:\\n # 1. 提取 HTML 注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n # 2. 提取 script 标签中的 JS 注释\\n script_pattern = re.compile(r'<script.*?>([\\\\s\\\\S]*?)</script>', re.IGNORECASE)\\n for script_match in script_pattern.finditer(content):\\n script_content = script_match.group(1)\\n js_pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<comment>//[^\\\\n]*|/\\\\*[\\\\s\\\\S]*?\\\\*/)',\\n re.DOTALL\\n )\\n for m in js_pattern.finditer(script_content):\\n if m.group('comment'):\\n comments.append(m.group('comment'))\\n \\n # 3. 提取 style 标签中的 CSS 注释\\n style_pattern = re.compile(r'<style.*?>([\\\\s\\\\S]*?)</style>', re.IGNORECASE)\\n for style_match in style_pattern.finditer(content):\\n style_content = style_match.group(1)\\n css_pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)',\\n re.DOTALL\\n )\\n for m in css_pattern.finditer(style_content):\\n if m.group('comment'):\\n comments.append(m.group('comment'))\\n\\n # 将该文件中抽出的所有有效注释组合并校验\\n all_comments = \\\"\\\\n\\\".join(comments)\\n share_count = all_comments.count(\\\"SHARE\\\")\\n \\n # 如果任意一份代码文件的注释中，包含\\\"SHARE\\\"的数量不恰好为2，验证均告失败\\n if share_count != 2:\\n return False\\n\\n # 若未找到任何代码文件，则不符合需求预期\\n return found_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你生成的代码文件中，每一个文件的总行数（涵盖空行）须为质数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个整数是否为质数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 定义可能生成的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.py', '.ts', '.jsx', \\n '.tsx', '.json', '.sh', '.vue', '.svelte'\\n }\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除通常不需要检测的第三方依赖或版本控制目录\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '__pycache__' in dirs:\\n dirs.remove('__pycache__')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n num_lines = len(lines)\\n \\n # 判断行数是否为质数\\n if not is_prime(num_lines):\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取文件，认定为不符合要求\\n return False\\n \\n # 确保至少检查过一个代码文件，防止空目录返回True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"执行搜索操作必须依赖正则表达式，不可采用简单的字符串匹配方式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中不包含多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"生成的代码里，“=”号的前后必须各自包含两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def check_text_for_equals(text: str) -> bool:\\n # 提前移除注释和字符串，以防其中的等号干扰判定\\n # 1. 移除 HTML 注释\\n text = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', text)\\n # 2. 移除 JS/CSS 块注释\\n text = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', text)\\n # 3. 移除 JS 单行注释\\n text = re.sub(r'//.*', '', text)\\n # 4. 移除双引号字符串\\n text = re.sub(r'\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', text)\\n # 5. 移除单引号字符串\\n text = re.sub(r\\\"'(?:[^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", text)\\n # 6. 移除反引号（模板字符串）\\n text = re.sub(r'`(?:[^`\\\\\\\\]|\\\\\\\\.)*`', '``', text)\\n \\n # 查找不在特定操作符（如 ==, !=, >=, =>, += 等）中的孤立单等号 '='\\n pattern = r'(?<![=!<>+\\\\-*/%&|^~$])=(?![=>])'\\n \\n for match in re.finditer(pattern, text):\\n start = match.start()\\n end = match.end()\\n \\n # 判断等号的前两个字符是否为两个连续空格\\n if start < 2 or text[start-2:start] != ' ':\\n return False\\n # 判断等号的后两个字符是否为两个连续空格\\n if end > len(text) - 2 or text[end:end+2] != ' ':\\n return False\\n \\n return True\\n\\n # 1. 检查 response 中的 Markdown 代码块\\n code_blocks = re.findall(r'```.*?\\\\n([\\\\s\\\\S]*?)```', response)\\n for block in code_blocks:\\n if not check_text_for_equals(block):\\n return False\\n\\n # 2. 检查 workspace_path 下生成的代码文件\\n if os.path.isdir(workspace_path):\\n valid_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.json', '.vue', '.py', '.txt'}\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in valid_exts:\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_text_for_equals(content):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面整体布局和配色：确认页面采用深色主题；顶部导航栏左侧显示Logo文字，右侧有“运行”、“设置”、“分享”、“格式化”等按钮；主界面上半部分被分为带有“HTML”、“CSS”、“JS”标题的三个区域；下半部分显示为带有“预览区”标题的区域。","验证代码编辑器是否正确初始化：在 HTML、CSS 和 JS 三个编辑器内分别输入几行代码（例如 `<div>test</div>`, `.a{color:red;}`, `const a = 1;`），检查这三个输入区域左侧是否正常显示代码行号，且代码文本带有相应语言的语法高亮颜色。同时检查编辑器实例的背景颜色与我们页面的深色主题融为一体。","验证实时预览与防抖机制：先在 HTML 编辑器中输入 <h1 id=\"title\">Hello</h1>，停止输入后等待约500ms，检查预览区是否渲染出“Hello”文字；再在 CSS 编辑器中输入 #title { color: red; }，停止输入后等待约500ms，检查预览区中的文字是否自动更新为红色；最后在 JS 编辑器中输入 document.getElementById('title').textContent = 'Hello JS';，停止输入后等待约500ms，检查预览区中的文字是否再次自动更新为“Hello JS”。","验证垂直分割线拖拽功能：将鼠标指针移动到 HTML 和 CSS 编辑器之间的垂直分割线上，检查鼠标指针是否变成左右调整大小的箭头样式；按下鼠标向右拖拽一段距离后松开，截屏确认 HTML 编辑器的宽度明显变宽，CSS 编辑器的宽度变窄。同理分别检查CSS和JS之间的垂直分割线。","验证水平分割线拖拽功能：将鼠标指针移动到上半部分编辑器和下半部分预览区之间的水平分割线上，检查鼠标指针是否变成上下调整大小的箭头样式；按下鼠标向下拖拽一段距离后松开，截屏确认三大编辑器区域的高度明显增加，下方的预览区高度缩小。","验证面板拖拽的最小尺寸限制：将 HTML 和 CSS 编辑器之间的分割线一直向左拖动到屏幕边缘，松开鼠标，确认 HTML 编辑器没有完全消失，而是保留了一定最小宽度的可见区域。同理分别检查CSS和JS之间的最小尺寸限制。","验证“设置”模态框的UI结构：点击页面顶部导航栏的“设置”按钮，检查页面中央是否弹出一个模态框，模态框内包含“外部 CSS”和“外部 JS”两个标签页，并且提供了用于添加链接的输入框和增删按钮。","验证外部资源设置与注入功能：点击顶部“设置”按钮，检查页面中央是否弹出包含“外部 CSS”和“外部 JS”两个标签页的模态框，且每个标签页中都支持对多个外部资源链接进行添加、编辑和删除（提供输入框和增删按钮）。在“外部 CSS”中添加 Bootstrap CDN 后，HTML 编辑器内输入 <button class=\"btn btn-primary\">Button</button> 时，预览区应显示带有 Bootstrap 默认样式的按钮；同理，在“外部 JS”中添加可用脚本链接后，相关脚本能力也应能在预览区内正常生效。点击已经添加的链接后文本框变为编辑状态。选中已添加连接再点击删除后对应链接正常删除。","验证 Console 面板的展开与折叠：点击页面最下方（或预览区下方）的“Console”面板标题栏，检查面板内容区域是否可以正常展开显示，再次点击是否能够折叠收起。","验证 Console 级别拦截与颜色区分：在 JS 编辑器中依次输入 `console.log(\"info\"); console.warn(\"warning\"); console.error(\"error\");`。展开底部 Console 面板，确认这三条日志均被捕获显示，并且 warn 级别日志带有黄色背景/文字，error 级别日志带有红色背景/文字，实现了视觉上的区分。","验证 Console 对象数据格式化：在 JS 编辑器中输入 `console.log({ user: \"Alice\", age: 20 });`，在展开的 Console 面板中检查，确认该对象没有显示为 `[object Object]`，而是被格式化显示为清晰易读、包含花括号和键值对的 JSON 字符串结构。","验证按钮触发的代码格式化：在 HTML、CSS、JS 编辑器中故意输入缩进混乱、多余空格的代码，点击顶部导航栏的“格式化”按钮，检查编辑器中的代码文本是否被自动排版美化为规范的多行缩进形式。","验证快捷键触发的代码格式化与默认行为阻止：在编辑器中输入排版混乱的代码，按下键盘快捷键 Ctrl+S（Mac下为 Cmd+S），检查代码是否立即被格式化整理，同时确认浏览器没有弹出自带的“保存网页”系统对话框。","验证代码编辑状态的本地持久化：在 HTML 编辑器中输入特定的标识符代码（如 `<!-- PERSIST TEST -->`），然后点击浏览器的刷新按钮重新加载页面，页面加载完毕后，检查该段标识符代码是否依然自动显示在编辑器中而没有丢失；再点击顶部的“设置”按钮，查看之前添加或修改的外部CSS链接和JS链接是否还存在。","验证分享按钮触发的 URL 序列化及提示：点击顶部导航栏的“分享”按钮，检查页面上是否弹出了包含“链接已复制”字样的 Toast 提示框，同时观察浏览器的地址栏，确认当前的 URL 末尾自动追加了一段很长的数据字符串（带有 # 号的 Hash 值）。","验证基于 URL Hash 的环境恢复功能：复制刚才点击“分享”生成的、带有 Hash 数据的完整 URL，在无痕窗口、独立浏览器环境或已清除本站 localStorage 的页面中打开该链接，确认页面加载后会优先从 URL Hash 中解析并恢复分享时的 HTML、CSS、JS 代码以及外部资源配置，而不是显示为空白、默认内容或从 localStorage 中读取旧状态。"],"string":"[\n \"查看页面整体布局和配色：确认页面采用深色主题；顶部导航栏左侧显示Logo文字，右侧有“运行”、“设置”、“分享”、“格式化”等按钮；主界面上半部分被分为带有“HTML”、“CSS”、“JS”标题的三个区域；下半部分显示为带有“预览区”标题的区域。\",\n \"验证代码编辑器是否正确初始化：在 HTML、CSS 和 JS 三个编辑器内分别输入几行代码（例如 `<div>test</div>`, `.a{color:red;}`, `const a = 1;`），检查这三个输入区域左侧是否正常显示代码行号，且代码文本带有相应语言的语法高亮颜色。同时检查编辑器实例的背景颜色与我们页面的深色主题融为一体。\",\n \"验证实时预览与防抖机制：先在 HTML 编辑器中输入 <h1 id=\\\"title\\\">Hello</h1>，停止输入后等待约500ms，检查预览区是否渲染出“Hello”文字；再在 CSS 编辑器中输入 #title { color: red; }，停止输入后等待约500ms，检查预览区中的文字是否自动更新为红色；最后在 JS 编辑器中输入 document.getElementById('title').textContent = 'Hello JS';，停止输入后等待约500ms，检查预览区中的文字是否再次自动更新为“Hello JS”。\",\n \"验证垂直分割线拖拽功能：将鼠标指针移动到 HTML 和 CSS 编辑器之间的垂直分割线上，检查鼠标指针是否变成左右调整大小的箭头样式；按下鼠标向右拖拽一段距离后松开，截屏确认 HTML 编辑器的宽度明显变宽，CSS 编辑器的宽度变窄。同理分别检查CSS和JS之间的垂直分割线。\",\n \"验证水平分割线拖拽功能：将鼠标指针移动到上半部分编辑器和下半部分预览区之间的水平分割线上，检查鼠标指针是否变成上下调整大小的箭头样式；按下鼠标向下拖拽一段距离后松开，截屏确认三大编辑器区域的高度明显增加，下方的预览区高度缩小。\",\n \"验证面板拖拽的最小尺寸限制：将 HTML 和 CSS 编辑器之间的分割线一直向左拖动到屏幕边缘，松开鼠标，确认 HTML 编辑器没有完全消失，而是保留了一定最小宽度的可见区域。同理分别检查CSS和JS之间的最小尺寸限制。\",\n \"验证“设置”模态框的UI结构：点击页面顶部导航栏的“设置”按钮，检查页面中央是否弹出一个模态框，模态框内包含“外部 CSS”和“外部 JS”两个标签页，并且提供了用于添加链接的输入框和增删按钮。\",\n \"验证外部资源设置与注入功能：点击顶部“设置”按钮，检查页面中央是否弹出包含“外部 CSS”和“外部 JS”两个标签页的模态框，且每个标签页中都支持对多个外部资源链接进行添加、编辑和删除（提供输入框和增删按钮）。在“外部 CSS”中添加 Bootstrap CDN 后，HTML 编辑器内输入 <button class=\\\"btn btn-primary\\\">Button</button> 时，预览区应显示带有 Bootstrap 默认样式的按钮；同理，在“外部 JS”中添加可用脚本链接后，相关脚本能力也应能在预览区内正常生效。点击已经添加的链接后文本框变为编辑状态。选中已添加连接再点击删除后对应链接正常删除。\",\n \"验证 Console 面板的展开与折叠：点击页面最下方（或预览区下方）的“Console”面板标题栏，检查面板内容区域是否可以正常展开显示，再次点击是否能够折叠收起。\",\n \"验证 Console 级别拦截与颜色区分：在 JS 编辑器中依次输入 `console.log(\\\"info\\\"); console.warn(\\\"warning\\\"); console.error(\\\"error\\\");`。展开底部 Console 面板，确认这三条日志均被捕获显示，并且 warn 级别日志带有黄色背景/文字，error 级别日志带有红色背景/文字，实现了视觉上的区分。\",\n \"验证 Console 对象数据格式化：在 JS 编辑器中输入 `console.log({ user: \\\"Alice\\\", age: 20 });`，在展开的 Console 面板中检查，确认该对象没有显示为 `[object Object]`，而是被格式化显示为清晰易读、包含花括号和键值对的 JSON 字符串结构。\",\n \"验证按钮触发的代码格式化：在 HTML、CSS、JS 编辑器中故意输入缩进混乱、多余空格的代码，点击顶部导航栏的“格式化”按钮，检查编辑器中的代码文本是否被自动排版美化为规范的多行缩进形式。\",\n \"验证快捷键触发的代码格式化与默认行为阻止：在编辑器中输入排版混乱的代码，按下键盘快捷键 Ctrl+S（Mac下为 Cmd+S），检查代码是否立即被格式化整理，同时确认浏览器没有弹出自带的“保存网页”系统对话框。\",\n \"验证代码编辑状态的本地持久化：在 HTML 编辑器中输入特定的标识符代码（如 `<!-- PERSIST TEST -->`），然后点击浏览器的刷新按钮重新加载页面，页面加载完毕后，检查该段标识符代码是否依然自动显示在编辑器中而没有丢失；再点击顶部的“设置”按钮，查看之前添加或修改的外部CSS链接和JS链接是否还存在。\",\n \"验证分享按钮触发的 URL 序列化及提示：点击顶部导航栏的“分享”按钮，检查页面上是否弹出了包含“链接已复制”字样的 Toast 提示框，同时观察浏览器的地址栏，确认当前的 URL 末尾自动追加了一段很长的数据字符串（带有 # 号的 Hash 值）。\",\n \"验证基于 URL Hash 的环境恢复功能：复制刚才点击“分享”生成的、带有 Hash 数据的完整 URL，在无痕窗口、独立浏览器环境或已清除本站 localStorage 的页面中打开该链接，确认页面加载后会优先从 URL Hash 中解析并恢复分享时的 HTML、CSS、JS 代码以及外部资源配置，而不是显示为空白、默认内容或从 localStorage 中读取旧状态。\"\n]","isConversational":false}}},{"rowIdx":69,"cells":{"task_id":{"kind":"number","value":70,"string":"70"},"repository_policy":{"kind":"string","value":"请你在后续所有交互轮次中严格遵循以下约束完成各类用户需求：\n1. 进行任何文件处理之前，切勿使用`pwd`确定当前的目录。\n2. 你创建的代码文件无一例外，全都要用UTF-8进行编码。\n3. 你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径。\n4. 在生成针对用户的回复时，必须且只能一直采用中文。\n5. 对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层。\n6. 进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改。\n7. 请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次。\n8. 请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码、脚本、配置及标记语言文件的扩展名集合\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.c', '.cpp', '.cc', '.cxx', \n '.h', '.hpp', '.java', '.go', '.rs', '.rb', '.php', '.html', '.htm', \n '.css', '.scss', '.less', '.sh', '.bat', '.ps1', '.sql', '.json', \n '.xml', '.yml', '.yaml', '.md', '.txt', '.vue', '.svelte', '.cs', '.swift'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录（如 .git）以及常见的依赖/缓存目录，减少不必要的检查和误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证疑似代码/文本文件的编码格式\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明文件包含无法用 utf-8 解码的字符，不满足要求\n return False\n except Exception:\n # 忽略其他诸如权限问题或文件被锁定的异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"import os\nimport re\n\ndef check_todo_in_comments(response: str, workspace_path: str) -> bool:\n todo_count = 0\n # 定义常见的代码文件扩展名集合\n code_exts = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.h', '.hpp', \n '.java', '.cs', '.html', '.css', '.sh', '.rb', '.go', '.php', \n '.rs', '.swift', '.kt', '.sql', '.vue', '.m', '.scala', '.dart'\n }\n \n # 构造正则表达式，分为两个命名组：string 和 comment。\n # 利用\"|\"的前后匹配优先级，优先匹配出字符串部分，避免将其内部包含的//、#等当做真实注释。\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|'\n r'(?P<comment>\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|//.*|/\\*[\\s\\S]*?\\*/|#.*|<!--[\\s\\S]*?-->|--.*)'\n )\n \n if not os.path.isdir(workspace_path):\n return False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件由于特殊原因(比如非utf8编码导致)无法读取则跳过\n continue\n \n # 在代码内容中寻找匹配项\n for match in pattern.finditer(content):\n comment = match.group('comment')\n # 如果匹配到的是真正的注释组\n if comment:\n # 统计注释中出现大写 TODO 的次数\n todo_count += comment.count(\"TODO\")\n \n # 判断是否严格出现2次\n return todo_count == 2","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]}],"string":"[\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码、脚本、配置及标记语言文件的扩展名集合\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.c', '.cpp', '.cc', '.cxx', \\n '.h', '.hpp', '.java', '.go', '.rs', '.rb', '.php', '.html', '.htm', \\n '.css', '.scss', '.less', '.sh', '.bat', '.ps1', '.sql', '.json', \\n '.xml', '.yml', '.yaml', '.md', '.txt', '.vue', '.svelte', '.cs', '.swift'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录（如 .git）以及常见的依赖/缓存目录，减少不必要的检查和误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证疑似代码/文本文件的编码格式\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明文件包含无法用 utf-8 解码的字符，不满足要求\\n return False\\n except Exception:\\n # 忽略其他诸如权限问题或文件被锁定的异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_todo_in_comments(response: str, workspace_path: str) -> bool:\\n todo_count = 0\\n # 定义常见的代码文件扩展名集合\\n code_exts = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.h', '.hpp', \\n '.java', '.cs', '.html', '.css', '.sh', '.rb', '.go', '.php', \\n '.rs', '.swift', '.kt', '.sql', '.vue', '.m', '.scala', '.dart'\\n }\\n \\n # 构造正则表达式，分为两个命名组：string 和 comment。\\n # 利用\\\"|\\\"的前后匹配优先级，优先匹配出字符串部分，避免将其内部包含的//、#等当做真实注释。\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(?P<comment>\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/|#.*|<!--[\\\\s\\\\S]*?-->|--.*)'\\n )\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件由于特殊原因(比如非utf8编码导致)无法读取则跳过\\n continue\\n \\n # 在代码内容中寻找匹配项\\n for match in pattern.finditer(content):\\n comment = match.group('comment')\\n # 如果匹配到的是真正的注释组\\n if comment:\\n # 统计注释中出现大写 TODO 的次数\\n todo_count += comment.count(\\\"TODO\\\")\\n \\n # 判断是否严格出现2次\\n return todo_count == 2\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"你好，我需要开发一款集成路线规划和天气数据的交互式旅行地图应用。第一步，请帮我搭建整个项目的前端基础骨架。要求如下：1. 创建一个包含项目Logo和用户头像的顶部导航栏。2. 左侧创建一个占据一定宽度的可折叠侧边栏（里面先只放一个“行程规划”的标题占位）。3. 右侧剩余的屏幕空间作为地图主区域，目前用一个带网格背景的灰色div来占位即可。4. 在主区域右下角放置一个悬浮的“新建行程”按钮（FAB）。整体页面要求是自适应的现代简洁风格，先不要引入真实的地图库，只要把结构和样式搭好即可。\n另外请注意，后续所有轮次都必须严格遵循以下要求：请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-；对于产出的所有代码文件，务必在结尾以独立的一行写入注释，明确标示出：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码相关文件扩展名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rb', '.php', \n '.sh', '.json', '.xml', '.yaml', '.yml'\n }\n\n # 如果工作区路径不存在或不是目录，则认为没有违反规则\n if not os.path.isdir(workspace_path):\n return True\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选出代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是使用合法的 UTF-8 编码\n return False\n except Exception:\n # 如果因为其他权限等原因无法读取文件，选择跳过处理\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"import os\nimport re\n\ndef check_todo_in_comments(response: str, workspace_path: str) -> bool:\n # 定义需要检查的常见代码文件扩展名\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.vue',\n '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.sh', '.rb', '.go', '.rs'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见非代码和第三方依赖目录，避免误判和过多无效计算\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 读取错误则忽略该文件\n continue\n \n comments = []\n \n # 根据不同语言的文件后缀，采用恰当的正则提取注释部分\n # 使用 [\\s\\S]*? 来匹配跨行注释，避免使用全局 re.DOTALL 导致单行注释越界吞噬\n if ext in ['.html', '.htm', '.vue']:\n # 匹配 HTML注释, JS/CSS跨行注释, 以及排除如 https:// 这类带有冒号前缀的单行注释\n pattern = r'<!--[\\s\\S]*?-->|/\\*[\\s\\S]*?\\*/|(?<!:)//.*'\n comments = re.findall(pattern, content)\n \n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.go', '.rs']:\n pattern = r'/\\*[\\s\\S]*?\\*/|(?<!:)//.*'\n comments = re.findall(pattern, content)\n \n elif ext in ['.py', '.rb', '.sh']:\n pattern = r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"|#.*'\n comments = re.findall(pattern, content)\n \n # 统计所有注释片段中 \"TODO\" 出现的总次数\n todo_count = sum(c.count('TODO') for c in comments)\n \n # 严格控制关键词 \"TODO\" 出现且仅出现 2 次\n if todo_count != 2:\n return False\n \n # 如果项目中没有任何代码文件，也判定为不符合要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断前端项目可能生成的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的首行\n first_line = f.readline()\n \n # 检查首行是否包含规定的编码格式字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法使用utf-8读取或其他异常，同样认为不符合要求\n return False\n \n # 如果没有找到任何代码文件，无法证明其满足了\"每一个输出的代码文件...\"的要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"对于产出的所有代码文件，务必在结尾以独立的一行写入注释，明确标示出：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 涵盖前端基础骨架及常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n code_files_count = 0\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略版本控制和第三方模块目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 无法读取文件时视为不合规\n return False\n \n # 提取所有非空行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n # 如果代码文件全是空行，则显然不符合具有结尾注释的要求\n if not non_empty_lines:\n return False\n \n # 检查文件结尾（倒数 3 个非空行）是否包含目标注释内容\n # 此处容忍了块注释结束符可能单独占据倒数第一行的情况（如 html的 --> 符号）\n found_target = False\n for line in non_empty_lines[-3:]:\n if target_comment in line:\n found_target = True\n break\n \n # 只要有一个代码文件的结尾没有包含目标注释，就直接返回 False\n if not found_target:\n return False\n\n # 需保证至少验证到了一个代码文件\n return code_files_count > 0","tags":["Content","Format"]}]},{"round_id":1,"instruction":"基础骨架搭得很不错。现在我们要让右侧的地图区域活起来。请引入 Leaflet.js（或者类似的开源地图库）来替换掉之前的占位网格，加载默认的开源地图瓦片。接下来实现一个基础交互：允许用户在地图上任意位置点击，点击后在该处放置一个蓝色的“位置”标记（Marker）。当用户点击这个标记时，在标记上方弹出一个提示框（Popup），里面显示该点的具体经纬度坐标。确保地图支持鼠标拖拽平移和滚轮缩放功能。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的在代码文件结尾添加指定AI生成注释的规则调整为，在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，请确保你输出的每一个代码文件，包含空行在内的总行数均为偶数。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区目录不存在，可以直接视为未违反规则\n if not os.path.exists(workspace_path):\n return True\n\n # 推断可能的代码文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.json', '.py', '.vue', \n '.md', '.txt', '.xml'\n }\n\n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件后缀并统一转小写进行匹配\n ext = os.path.splitext(file)[1].lower()\n \n # 若是指定的代码文件，则进行编码检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 utf-8 规则进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 只要有一个文件出现解码异常，就说明不是全部采用 UTF-8 编码\n return False\n except Exception:\n # 忽略不可读等其他无关的文件系统异常\n pass\n\n # 若所有代码文件都能被正确解析为 UTF-8，则符合要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"def check_todo_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对 Python 文件的字符串与注释提取（含单/双/三引号，#注释）\n py_pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')|'\n r'(?P<comment>\"{3}[\\s\\S]*?\"{3}|\\'{3}[\\s\\S]*?\\'{3}|#.*)'\n )\n \n # 针对 JS/TS 等文件的字符串与注释提取（含反引号、双/单引号， //与/* */注释）\n js_pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|'\n r'(?P<comment>/\\*[\\s\\S]*?\\*/|//.*)'\n )\n \n # 针对 CSS 等文件的字符串与注释提取（/* */注释）\n css_pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')|'\n r'(?P<comment>/\\*[\\s\\S]*?\\*/)'\n )\n \n todo_count = 0\n\n # 辅助函数：给定正则与内容，统计有效注释中的 \"TODO\" 个数\n def count_in_pattern(content, pattern):\n count = 0\n for match in pattern.finditer(content):\n comment = match.group('comment')\n if comment:\n count += comment.count(\"TODO\")\n return count\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的外部无关目录，防止混入外部库中的代码产生干扰\n if any(part in root for part in ['.git', 'node_modules', '__pycache__', 'venv', 'env', '.idea', '.vscode']):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java',\n '.html', '.htm', '.vue',\n '.css', '.scss', '.less',\n '.py'}:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 根据不同扩展名执行对应的正则统计逻辑\n if ext in {'.py'}:\n todo_count += count_in_pattern(content, py_pattern)\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java'}:\n todo_count += count_in_pattern(content, js_pattern)\n elif ext in {'.css', '.scss', '.less'}:\n todo_count += count_in_pattern(content, css_pattern)\n elif ext in {'.html', '.htm', '.vue'}:\n # 1. HTML 原生注释： <!-- -->\n html_comments = re.findall(r'<!--[\\s\\S]*?-->', content)\n for comment in html_comments:\n todo_count += comment.count(\"TODO\")\n \n # 2. 提取 <script> 内容进行 JS 注释分析\n script_contents = re.findall(r'<script\\b[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for sc in script_contents:\n todo_count += count_in_pattern(sc, js_pattern)\n \n # 3. 提取 <style> 内容进行 CSS 注释分析\n style_contents = re.findall(r'<style\\b[^>]*>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n for st in style_contents:\n todo_count += count_in_pattern(st, css_pattern)\n \n # 验证是否严格控制并只出现了两次 TODO\n return todo_count == 2","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_first_line_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目描述推断主要的前端代码文件格式\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_checked_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方依赖、版本控制以及其他无关目录\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '.vscode', '.idea']]\n \n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n has_checked_file = True\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查文件是否为空，或者第一行中是否未包含指定的编码标识\n if not first_line or '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 出现无法读取文件等异常情况，也视作不满足要求\n return False\n \n # 确保至少验证过一个代码文件\n return has_checked_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 前端相关的常见代码文件后缀\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n has_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n # 排除无关的依赖和版本管理目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于编码等问题无法读取文件，直接视为未满足\n return False\n \n found_in_head = False\n # 检查文件的开头部分（设定检查前10行，保证兼容第一行编码声明的情况）\n for line in lines[:10]:\n if target_str in line:\n found_in_head = True\n break\n \n # 如果某个代码文件的开头没有该目标字符串，则直接判定为不满足要求\n if not found_in_head:\n return False\n \n # 如果所有的代码文件都满足要求，并且至少存在一个代码文件，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"请确保你输出的每一个代码文件，包含空行在内的总行数均为偶数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义该前端项目可能会用到的常见代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的版本控制和第三方依赖包目录，以防干扰判断\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符拆分，获取包含空行在内的总行数\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果有文件读取报错（比如不可读的非文本文件等）可以直接跳过\n continue\n \n return True","tags":["Quantity","Exact Value"]}]},{"round_id":2,"instruction":"地图已经可以交互了，接下来我们完善左侧边栏的路线规划输入区。请在侧边栏中添加“出发地”和“目的地”两个文本输入框，并在下方放置一个“规划路线”的按钮。为了跑通流程，当用户在输入框随意输入内容并点击按钮时，请在地图上随机或者基于固定坐标生成两个不同颜色（比如绿色代表起点，红色代表终点）的标记，并在它们之间绘制一条模拟的实线折线（Polyline）作为路线。绘制完成后，地图需要自动缩放和平移，以刚好完整显示出这条路线。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的输出代码文件包含空行在内总行数为偶数的规则调整为，请确保你输出的每一个代码文件，包含空行在内的总行数均为奇数；第二，后续所有轮次针对每一个生成的代码文件，注释的行数需精准控制在 4 行。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推测可能包含的代码相关文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.json', '.py', '.md', '.txt'\n }\n \n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制方式读取文件内容\n with open(file_path, 'rb') as f:\n content = f.read()\n \n # 尝试用 utf-8 解码，如果不抛出 UnicodeDecodeError，则说明符合 utf-8 编码要求\n content.decode('utf-8')\n \n except UnicodeDecodeError:\n # 一旦有文件无法用 utf-8 解码，直接返回 False\n return False\n except Exception:\n # 处理如文件无权限读取等意外情况（通常不应作为非 utf-8 判断依据，故跳过）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n total_todo_count = 0\n file_counts = []\n \n # 定义需要检查的常见代码文件扩展名\n code_exts = {'.html', '.htm', '.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.py'}\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的非业务代码或依赖目录，避免引入第三方代码的干扰\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'venv', 'env', 'dist', 'build'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_files = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n spans = []\n # 针对不同扩展名使用对应的正则提取注释的span起始位置\n if ext in {'.html', '.htm', '.vue'}:\n spans.extend([m.span() for m in re.finditer(r'<!--[\\s\\S]*?-->', code)])\n # HTML/Vue 文件内可能包含 script 和 style，同样提取其对应的注释片段\n spans.extend([m.span() for m in re.finditer(r'/\\*[\\s\\S]*?\\*/', code)])\n spans.extend([m.span() for m in re.finditer(r'//.*', code)])\n elif ext in {'.js', '.ts', '.css', '.jsx', '.tsx'}:\n spans.extend([m.span() for m in re.finditer(r'/\\*[\\s\\S]*?\\*/', code)])\n spans.extend([m.span() for m in re.finditer(r'//.*', code)])\n elif ext in {'.py'}:\n spans.extend([m.span() for m in re.finditer(r'#.*', code)])\n spans.extend([m.span() for m in re.finditer(r'\"\"\"[\\s\\S]*?\"\"\"', code)])\n spans.extend([m.span() for m in re.finditer(r\"'''[\\s\\S]*?'''\", code)])\n \n # 合并重叠的注释区间（防止不同的正则表达式重复捕获同一内容）\n spans.sort(key=lambda x: x[0])\n merged_spans = []\n for s in spans:\n if not merged_spans:\n merged_spans.append(s)\n else:\n last = merged_spans[-1]\n if s[0] <= last[1]:\n # 有重叠，合并区间\n merged_spans[-1] = (last[0], max(last[1], s[1]))\n else:\n merged_spans.append(s)\n \n # 统计当前代码文件注释中的 \"TODO\" 出现次数\n file_todo_count = 0\n for start, end in merged_spans:\n comment_text = code[start:end]\n file_todo_count += comment_text.count(\"TODO\")\n \n file_counts.append(file_todo_count)\n total_todo_count += file_todo_count\n\n if not has_code_files:\n return False\n \n # 满足以下任一条件均视为达标：\n # 1. 整个项目中所有生成的代码文件注释内 \"TODO\" 的总数为2\n # 2. 或者每个生成的代码文件注释内 \"TODO\" 的数量恰好均等于 2\n if total_todo_count == 2:\n return True\n \n if len(file_counts) > 0 and all(c == 2 for c in file_counts):\n return True\n \n return False","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.isdir(workspace_path):\n return False\n\n # 根据用户指令的前端项目特征，推断可能的代码文件扩展名\n code_exts = {'.html', '.css', '.js'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 如果存在第三方库等常见忽略文件夹，可跳过，但通常AI生成的纯代码目录下只有其编写的文件\n for file in files:\n if any(file.endswith(ext) for ext in code_exts):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含规定的编码格式字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于某种原因（如编码非UTF-8等）无法读取，则直接判定为不满足\n return False\n \n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 定义常见的前端/后端代码文件扩展名\n code_exts = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n \n file_count = 0\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_count += 1\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于编码或其他原因导致无法读取文件，认为要求未满足\n return False\n \n # 检查该文件的开头处（前 10 行）\n found = False\n for line in lines[:10]:\n if target_str in line:\n # 剔除目标字符串，留下周围的字符\n remainder = line.replace(target_str, '').strip()\n # 遍历并剔除各类常见编程语言的注释相关符号\n for sym in ['<!--', '-->', '/*', '*/', '//', '#', '{/*', '*/}', '--']:\n remainder = remainder.replace(sym, '')\n # 若剩余部分没有任何可见字符，则符合\"单独换行添加注释\"的要求\n if not remainder.strip():\n found = True\n break\n \n # 如果当前文件没有符合要求的注释，整体判定为不满足\n if not found:\n return False\n \n # 至少要生成了相关代码文件且全部满足要求\n if file_count == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"请确保你输出的每一个代码文件，包含空行在内的总行数均为奇数","validation_code":"import os\n\ndef check_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\n # 如果工作区路径不存在或是个文件而不是目录，则不满足要求\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）和 node_modules 依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 通过换行符拆分文本，统计包含空行的总行数\n lines_count = len(content.split('\\n'))\n \n # 若发现包含空行在内的总行数为偶数，则不满足要求\n if lines_count % 2 == 0:\n return False\n except Exception:\n # 如果读取过程中出现异常（例如权限不足），跳过处理\n continue\n \n # 如果找到了至少一个代码文件并且全为奇数行，则满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在 4 行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n def count_comments_in_code(code_str: str, file_ext: str) -> int:\n file_ext = file_ext.lower()\n \n # 针对不同语言扩展名构建模式匹配\n # 在正则中，将字符串字面量的匹配放在最高优先级。一旦被匹配为字符串并不会处理内部的注释符号，这就有效避免了误判。\n if file_ext in ['.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less']:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<multiline>/\\*.*?\\*/)'\n r'|(?P<singleline>//[^\\n]*)',\n re.DOTALL\n )\n elif file_ext in ['.html', '.htm', '.vue', '.svelte']:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<html_comment><!--.*?-->)'\n r'|(?P<multiline>/\\*.*?\\*/)'\n r'|(?P<singleline>//[^\\n]*)',\n re.DOTALL\n )\n elif file_ext in ['.py']:\n pattern = re.compile(\n r'(?P<string>\"\"\"(?:\\\\.|[^\"\\\\])*\"\"\"|\\'\\'\\'(?:\\\\.|[^\\'\\\\])*\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<singleline>#[^\\n]*)',\n re.DOTALL\n )\n else:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<html_comment><!--.*?-->)'\n r'|(?P<multiline>/\\*.*?\\*/)'\n r'|(?P<singleline>//[^\\n]*|#[^\\n]*)',\n re.DOTALL\n )\n\n comment_lines = set()\n \n # 预先找到所有的换行符位置用于二分查找获得对应行号\n line_starts = [0]\n for m in re.finditer(r'\\n', code_str):\n line_starts.append(m.start() + 1)\n \n def get_line(idx: int) -> int:\n return bisect.bisect_right(line_starts, idx)\n \n # 找出匹配块并计算被注释占据的所在行集合\n for m in pattern.finditer(code_str):\n if m.lastgroup in ['multiline', 'singleline', 'html_comment']:\n start_idx = m.start()\n end_idx = m.end()\n \n if m.lastgroup == 'singleline':\n line = get_line(start_idx)\n comment_lines.add(line)\n else:\n start_line = get_line(start_idx)\n # end_idx代表匹配块的终点后一个索引，如果是多行，实际在文本结束的前一个字符位置计算行号\n end_line = get_line(end_idx - 1) if end_idx > start_idx else start_line\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n \n return len(comment_lines)\n\n valid_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json', '.vue', '.svelte'}\n code_files_count = 0\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和第三方依赖目录以加快并确保检索准确\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for f in files:\n # 略过隐藏文件\n if f.startswith('.'):\n continue\n \n ext = os.path.splitext(f)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, f)\n try:\n with open(file_path, 'r', encoding='utf-8') as f_in:\n content = f_in.read()\n except UnicodeDecodeError:\n continue # 略过可能为二进制或编码非utf-8的文件\n \n comment_count = count_comments_in_code(content, ext)\n \n # 若任意一个代码文件的注释行数不恰好等于4，验证即为不通过\n if comment_count != 4:\n return False\n \n code_files_count += 1\n\n # 要求必须至少存在一个代码文件被验证且成功通过\n return code_files_count > 0","tags":["Quantity","Exact Value"]}]},{"round_id":3,"instruction":"真实的自驾游通常会有多个中途停靠点。请对刚才的路线表单进行复杂化升级：在起点和终点输入框之间增加一个“添加途经点”的按钮。用户点击后可以动态新增最多5个途经点输入框。最关键的是，请实现拖拽排序功能（Drag and Drop），让用户可以通过拖拽这些输入框来调整途经点的先后顺序。每一个途经点输入框旁还需要一个删除按钮。再次点击“规划路线”时，地图上的折线必须严格按照“起点 -> 排序后的途经点 -> 终点”的顺序重新连接绘制，并用带有数字1、2、3...的图标标出途经点的位置。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的每一个生成的代码文件注释行数需精准控制在4行的规则调整为，限制每个生成的代码文件中的注释数量必须刚好是5行；第二，你必须在代码写完后测试全部的代码，并确保测试成功通过，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见代码及文本文件的扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs',\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.scss', '.sass', '.less',\n '.json', '.xml', '.yaml', '.yml', '.sh', '.bat', '.ps1',\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.vue', '.svelte',\n '.md', '.txt'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 格式读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明并非合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如文件权限错误等其他异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n file_todo_counts = []\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的前端依赖、后端环境与隐藏目录\n if 'node_modules' in root or '.git' in root or 'venv' in root or '__pycache__' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 针对前端项目，推测可能的代码文件类型后缀\n if ext in ['.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.vue', '.py']:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n # 针对不同扩展名使用对应的正则表达式提取代码注释\n if ext in ['.html', '.htm', '.vue']:\n # 匹配 HTML 或 Vue 模板注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # 匹配内联脚本和样式中的 JS/CSS 注释 (注意 (?<!:) 用于避免误匹配到 http:// 中的双斜杠)\n comments.extend(re.findall(r'(?<!:)//.*?$|/\\*[\\s\\S]*?\\*/', content, re.MULTILINE))\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss']:\n comments.extend(re.findall(r'(?<!:)//.*?$|/\\*[\\s\\S]*?\\*/', content, re.MULTILINE))\n elif ext in ['.py']:\n # 匹配 Python 的单行和多行注释\n comments.extend(re.findall(r'#.*?$', content, re.MULTILINE))\n comments.extend(re.findall(r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"', content))\n \n # 统计该代码文件提取到的所有注释中 \"TODO\" 的具体出现次数\n todo_count = sum(c.count(\"TODO\") for c in comments)\n file_todo_counts.append(todo_count)\n \n if not has_code_file or not file_todo_counts:\n return False\n \n # 情况一：要求每个生成的代码文件自身的注释中，\"TODO\"均恰好出现 2 次\n if all(count == 2 for count in file_todo_counts):\n return True\n \n # 情况二：要求整体项目所有代码文件的注释总和中，\"TODO\"共计出现 2 次\n if sum(file_todo_counts) == 2:\n return True\n \n return False","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对前端及常规项目的常见代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰检查的无关目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含规定的编码声明字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果由于非utf-8编码等原因导致无法读取，说明不满足指定utf-8编码的要求\n return False\n\n # 必须至少生成了一个代码文件，并且全部通过检查才算满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据前端路线图应用开发的需求，推测可能产生的代码文件类型\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n comment_found = False\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n # 在文件开头处（前 5 行内）寻找指定注释\n for line in lines[:5]:\n if target_comment in line:\n comment_found = True\n break\n except Exception:\n pass\n \n # 如果某个代码文件未在开头找到指定的注释，或者无法读取，即视为不满足要求\n if not comment_found:\n return False\n \n return has_code_files","tags":["Content","Format"]},{"constraint":"请确保你输出的每一个代码文件，包含空行在内的总行数均为奇数","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义前端项目中可能的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能包含大量无关依赖包及版本控制信息的目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符切分以统计行数，此方法与主流IDE计算包含空行在内的总行数一致\n lines = content.split('\\n')\n total_lines = len(lines)\n \n # 检查包含空行在内的总行数是否为奇数\n if total_lines % 2 == 0:\n return False\n except Exception:\n # 如果遇到无法以UTF-8读取的文件，忽略该文件\n continue\n \n # 如果项目中没有任何代码文件，意味着并没有正确生成项目，返回 False\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"限制每个生成的代码文件中的注释数量必须刚好是5行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n from html.parser import HTMLParser\n\n # 定义目标代码文件的后缀名\n CODE_EXTS = {'.html', '.htm', '.vue', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp'}\n\n def get_line_starts(text):\n \"\"\"记录每行的起始下标，以便快速将字符索引转为行号\"\"\"\n line_starts = [0]\n for m in re.finditer(r'\\n', text):\n line_starts.append(m.end())\n return line_starts\n\n def get_line_num(line_starts, pos):\n \"\"\"利用二分查找定位 pos 所在的行号（基于1的索引）\"\"\"\n return bisect.bisect_right(line_starts, pos)\n\n files_checked = 0\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和第三方依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in CODE_EXTS:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n\n comment_lines = set()\n \n if ext in ('.html', '.htm', '.vue'):\n class Parser(HTMLParser):\n def __init__(self):\n super().__init__()\n self.comment_lines = set()\n self.embedded_code = []\n self.current_tag = None\n\n def handle_starttag(self, tag, attrs):\n self.current_tag = tag\n\n def handle_endtag(self, tag):\n self.current_tag = None\n\n def handle_data(self, data):\n if self.current_tag in ('script', 'style'):\n # 获取 <script> / <style> 数据块开头的行号\n start_line, _ = self.getpos()\n self.embedded_code.append((self.current_tag, data, start_line))\n\n def handle_comment(self, data):\n # 处理 HTML 原生注释\n start_line, _ = self.getpos()\n num_lines = data.count('\\n')\n for i in range(num_lines + 1):\n self.comment_lines.add(start_line + i)\n\n parser = Parser()\n try:\n parser.feed(code)\n comment_lines.update(parser.comment_lines)\n \n # 额外处理内嵌的 JS 或 CSS 代码中的注释\n for tag, data, start_line in parser.embedded_code:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<m_comment>/\\*[\\s\\S]*?\\*/)'\n r'|(?P<s_comment>//.*)',\n re.MULTILINE\n )\n l_starts = get_line_starts(data)\n for m in pattern.finditer(data):\n if m.group('m_comment'):\n s_pos = m.start('m_comment')\n e_pos = m.end('m_comment')\n s_line = get_line_num(l_starts, s_pos)\n e_line = get_line_num(l_starts, e_pos - 1)\n for l in range(s_line, e_line + 1):\n comment_lines.add(start_line + l - 1)\n elif m.group('s_comment'):\n s_pos = m.start('s_comment')\n s_line = get_line_num(l_starts, s_pos)\n comment_lines.add(start_line + s_line - 1)\n except Exception:\n # 若 HTML 解析器异常退出，降级使用基础正则以匹配 HTML 注释\n pattern = re.compile(r'<!--[\\s\\S]*?-->', re.MULTILINE)\n l_starts = get_line_starts(code)\n for m in pattern.finditer(code):\n s_line = get_line_num(l_starts, m.start())\n e_line = get_line_num(l_starts, m.end() - 1)\n for l in range(s_line, e_line + 1):\n comment_lines.add(l)\n \n elif ext in ('.js', '.jsx', '.ts', '.tsx', '.css', '.c', '.cpp', '.h', '.hpp', '.java'):\n # 处理 JS/TS/CSS 类文件的注释\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<m_comment>/\\*[\\s\\S]*?\\*/)'\n r'|(?P<s_comment>//.*)',\n re.MULTILINE\n )\n l_starts = get_line_starts(code)\n for m in pattern.finditer(code):\n if m.group('m_comment'):\n s_pos = m.start('m_comment')\n e_pos = m.end('m_comment')\n s_line = get_line_num(l_starts, s_pos)\n e_line = get_line_num(l_starts, e_pos - 1)\n for l in range(s_line, e_line + 1):\n comment_lines.add(l)\n elif m.group('s_comment'):\n s_pos = m.start('s_comment')\n s_line = get_line_num(l_starts, s_pos)\n comment_lines.add(s_line)\n \n elif ext in ('.py',):\n # 处理 Python 文件的注释（只抓取 # 注释行，多行字符串若未被变量接收暂视为普通字符串）\n pattern = re.compile(\n r'(?P<string>\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<s_comment>#.*)',\n re.MULTILINE\n )\n l_starts = get_line_starts(code)\n for m in pattern.finditer(code):\n if m.group('s_comment'):\n s_pos = m.start('s_comment')\n s_line = get_line_num(l_starts, s_pos)\n comment_lines.add(s_line)\n\n # 验证单一文件是否恰好为 5 行注释\n if len(comment_lines) != 5:\n return False\n \n files_checked += 1\n\n # 需确保项目下至少检查了一个代码文件\n return files_checked > 0","tags":["Quantity","Exact Value"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"路线画好后，用户需要知道具体的行程信息。请在侧边栏表单下方新增一个“行程详情”面板。每次生成路线后，在面板顶部展示模拟计算出的“总距离（如 350公里）”和“预计驾驶时长（如 4小时20分钟）”。在此下方，生成一个垂直的行程时间轴（Timeline）列表，依次展示“起点 -> 途经点1 -> ... -> 终点”。在时间轴的每一个路段节点之间，清晰地标出该特定路段的距离和预估耗时。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的输出代码文件包含空行在内总行数为奇数的规则调整为，你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数；第二，后续所有轮次请始终遵循，项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的前端与后端代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n \n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试用 UTF-8 编码格式读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出解码错误，说明该文件不是纯正的 UTF-8 编码\n return False\n except Exception:\n # 如果有其他异常(例如权限问题等)暂忽略，但主要检查点在于 UnicodeDecodeError\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.sh', '.bash'\n }\n \n code_files = []\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和第三方依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files.append(os.path.join(root, file))\n \n # 如果没有找到任何代码文件，则不满足条件\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n ext = os.path.splitext(file_path)[1].lower()\n patterns = []\n \n # 根据文件类型构造对应的注释匹配正则表达式\n if ext in {'.html', '.vue'}:\n patterns.append(r'<!--[\\s\\S]*?-->')\n patterns.append(r'/\\*[\\s\\S]*?\\*/')\n patterns.append(r'(?<!:)//.*')\n elif ext in {'.css'}:\n patterns.append(r'/\\*[\\s\\S]*?\\*/')\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.h', '.hpp'}:\n patterns.append(r'/\\*[\\s\\S]*?\\*/')\n patterns.append(r'(?<!:)//.*')\n elif ext in {'.py', '.sh', '.bash'}:\n patterns.append(r'#.*')\n if ext == '.py':\n patterns.append(r'\"\"\"[\\s\\S]*?\"\"\"')\n patterns.append(r\"'''[\\s\\S]*?'''\")\n \n if not patterns:\n continue\n \n # 使用逻辑“或”组合正则以同时匹配所有有效注释类型\n combined_pattern = '|'.join(f'({p})' for p in patterns)\n regex = re.compile(combined_pattern)\n \n todo_count = 0\n for match in regex.finditer(content):\n comment = match.group(0)\n # 统计注释内部的 \"TODO\" 关键词出现次数\n todo_count += comment.count('TODO')\n \n # 若任意一份代码文件中的 TODO 数量不为 2 次，即不满足要求\n if todo_count != 2:\n return False\n \n except Exception:\n # 根据要求，所有代码文件都必须为 UTF-8 编码，读文件失败视为错误\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见前端代码文件扩展名\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏文件夹及常见的第三方依赖目录，避免误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码格式声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件读取失败，则视为不符合要求\n return False\n \n # 如果项目中没有任何我们所关注的代码文件，视为未完成需求\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n file_count = 0\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏文件夹和无关的依赖、编译目录，避免误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'dist', 'build')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 检查开头部分（前10行），以容忍首行的文件编码说明（如 -*- coding: utf-8 -*-）及 HTML 的 doctype 等\n lines = [f.readline() for _ in range(10)]\n except Exception:\n # 若无法使用UTF-8正常读取，视同未按要求处理\n return False\n \n found = False\n for line in lines:\n if not line:\n continue\n # 检查目标注释文字是否包含在当前行中\n if target_text in line:\n found = True\n break\n \n # 只要存在任何一个代码文件不包含目标注释，即不满足要求\n if not found:\n return False\n \n file_count += 1\n \n # 至少要验证通过一个代码文件才算满足要求，避免空目录蒙混过关\n if file_count == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断的前端代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除掉不属于生成的第三方依赖及系统隐藏文件夹\n for exclude_dir in ['node_modules', '.git', '.idea', '.vscode']:\n if exclude_dir in dirs:\n dirs.remove(exclude_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 验证总行数（含空行）是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 如果有异常如编码不是UTF-8，尝试使用latin-1安全读取行数\n try:\n with open(file_path, 'r', encoding='latin-1') as f:\n lines = f.readlines()\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 极端异常无法读取情况可视为无法判定，直接忽略处理继续向下\n pass\n \n # 只要存在代码文件，并且没有触发奇数行的return False，就代表通过验证\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"限制每个生成的代码文件中的注释数量必须刚好是5行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n if not os.path.exists(workspace_path):\n return False\n \n # 定义需要检查的常见代码文件扩展名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less',\n '.html', '.htm', '.vue', '.svg', '.py', '.java', '.c', '.cpp',\n '.cs', '.go', '.php', '.rb', '.sh'\n }\n \n files_checked = 0\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏文件夹、前端依赖项和 Python 缓存\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n files_checked += 1\n \n # 记录所有换行符的位置以用来二分查找对应的行号\n line_starts = [0]\n for m in re.finditer(r'\\n', content):\n line_starts.append(m.end())\n \n def get_line_num(pos):\n return bisect.bisect_right(line_starts, pos)\n \n comment_lines = set()\n \n # C风格及前端家族文件（过滤引号内字符串和css中的url影响，分别捕获HTML、多行、单行注释）\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm', '.vue', '.svg', '.java', '.c', '.cpp', '.cs', '.go', '.php']:\n pattern = re.compile(\n r'(url\\([^)]*\\))' \n r'|(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)' \n r'|(<!--[\\s\\S]*?-->)' \n r'|(/\\*[\\s\\S]*?\\*/)' \n r'|((?<!:)//.*)' \n )\n for m in pattern.finditer(content):\n if m.group(1) or m.group(2):\n continue\n elif m.group(3):\n start_line = get_line_num(m.start(3))\n end_line = get_line_num(m.end(3) - 1)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n elif m.group(4):\n start_line = get_line_num(m.start(4))\n end_line = get_line_num(m.end(4) - 1)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n elif m.group(5):\n start_line = get_line_num(m.start(5))\n comment_lines.add(start_line)\n\n # Python/Shell 家族文件（以 # 为核心进行处理，同样需过滤字符串影响）\n elif ext in ['.py', '.rb', '.sh', '.yml', '.yaml']:\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\')' \n r'|(#.*)' \n )\n for m in pattern.finditer(content):\n if m.group(1):\n continue\n elif m.group(2):\n start_line = get_line_num(m.start(2))\n comment_lines.add(start_line)\n else:\n continue\n \n # 针对该特定文件，注释占据的总行数若不严格等于 5 行，则视为没有满足条件\n if len(comment_lines) != 5:\n return False\n \n # 如果没有找到任何可检查的代码文件，则不能算作满足要求\n return files_checked > 0","tags":["Quantity","Exact Value"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":5,"instruction":"太棒了！现在我们来引入这个应用的核心特色——天气数据。我们先使用模拟的天气数据。请修改“行程详情”面板的时间轴，在每一个站点（起点、所有途经点、终点）的右侧，加上天气信息的展示，包括一个直观的天气图标（晴、雨、多云等）、当前温度和降水概率。同时，联动一下地图：当用户在地图上点击这几个站点的标记时，弹出的 Popup 除了显示地名，还要同步展示刚才这套模拟的天气数据。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的你必须在代码写完后测试全部的代码并确保测试成功通过的规则调整为，当你把代码写完后，不得进行任何测试；第二，后续所有轮次生成的代码文件里，缩进必须严格保持为2个空格。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，说明没有生成文件，则不算违反规则\n if not os.path.exists(workspace_path):\n return True\n\n # 常见的前端项目代码文件及配置文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue', '.py'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试使用 utf-8 编码去读取整个文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果遇到 Unicode 解码错误，说明非 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如文件权限等其他不可预期的异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"import os\nimport re\n\ndef check_todo_count(response: str, workspace_path: str) -> bool:\n # 常见的代码文件后缀\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n all_files_todo_counts = []\n total_todo_count = 0\n \n # 遍历当前项目工作空间的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 根据不同扩展名赋予对应的正则表达式\n # 多行注释优先提取，单行注释后提取，避免出现重复计算的情况\n if ext in {'.html', '.vue'}:\n patterns = [r'<!--[\\s\\S]*?-->', r'/\\*[\\s\\S]*?\\*/', r'//.*']\n elif ext in {'.js', '.ts', '.jsx', '.tsx'}:\n patterns = [r'/\\*[\\s\\S]*?\\*/', r'//.*']\n elif ext == '.css':\n patterns = [r'/\\*[\\s\\S]*?\\*/']\n elif ext == '.py':\n patterns = [r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"', r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'', r'#.*']\n else:\n patterns = [r'<!--[\\s\\S]*?-->', r'/\\*[\\s\\S]*?\\*/', r'//.*', r'#.*']\n \n all_matches = []\n for p in patterns:\n matches = re.findall(p, content)\n # 将提取到的注释部分替换为空格，以免后续匹配互相干扰\n content = re.sub(p, ' ', content)\n all_matches.extend(matches)\n \n # 统计提取出的注释中 TODO 出现的次数\n todo_count = sum(comment.count(\"TODO\") for comment in all_matches)\n all_files_todo_counts.append(todo_count)\n total_todo_count += todo_count\n \n # 如果项目中没有扫描到任何代码文件，则不满足要求\n if not all_files_todo_counts:\n return False\n \n # 第一种理解：所有扫描的代码文件，每一个文件中的 TODO 数量均为 2 次\n if all(c == 2 for c in all_files_todo_counts):\n return True\n \n # 第二种理解：整个项目中所有代码文件的 TODO 数量总和为 2 次\n if total_todo_count == 2:\n return True\n \n return False","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 定义常见的前端与相关代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.vue'}\n # 定义需要忽略的目录，避免误检查非生成的代码文件\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查代码文件首行是否包含指定的编码格式声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果在打开或读取过程中遇到异常（如非文本文件或编码错误），认定为不符合要求\n return False\n \n # 如果所有代码文件的首行都包含指定内容，或没有扫描到需要处理的代码文件，则返回True\n return True","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 前端及通用代码文件的扩展名集合\n code_exts = {'.html', '.js', '.css', '.ts', '.vue', '.jsx', '.tsx', '.py'}\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 检查前5行是否包含目标字符串（考虑到第一行可能有编码声明等）\n found_in_head = False\n for line in lines[:5]:\n if target_str in line:\n found_in_head = True\n break\n \n if not found_in_head:\n return False\n except Exception:\n # 如果无法按utf-8读取，或出现其他异常，视为不符合要求\n return False\n \n # 如果找到了代码文件，并且全部通过了检查，则返回 True\n return found_code_file","tags":["Content","Format"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对前端项目的常见代码文件扩展名\n code_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_checked = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除掉无关目录，如 node_modules 及隐藏文件夹（保留 __tests__ 等）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按换行符分割，包括末尾的空行，能够真实反映文件包含空行的总行数\n lines = content.split('\\n')\n if len(lines) % 2 != 0:\n return False\n \n has_checked = True\n except Exception:\n # 如果由于非utf-8编码等原因无法读取，认为无法满足题目规定的有效文件读取，返回 False\n return False\n \n return has_checked","tags":["Quantity","Exact Value"]},{"constraint":"限制每个生成的代码文件中的注释数量必须刚好是5行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n CODE_EXTS = {'.js', '.css', '.html', '.htm', '.vue', '.py', '.ts', '.jsx', '.tsx'}\n exclude_dirs = {'node_modules', '.git', '__pycache__', 'venv', 'env', '.vscode', '.idea'}\n \n code_files = []\n # 获取所有的代码文件，排除常见依赖和缓存目录\n for root, dirs, files in os.walk(workspace_path):\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for f in files:\n ext = os.path.splitext(f)[1].lower()\n if ext in CODE_EXTS:\n code_files.append(os.path.join(root, f))\n \n if not code_files:\n return False\n \n for cf in code_files:\n ext = os.path.splitext(cf)[1].lower()\n try:\n with open(cf, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return False\n \n # 根据不同语言分别定义字符串和注释正则匹配模式\n if ext == '.py':\n # Python中处理f-string/r-string/b-string及常规三引号、单双引号内的字符串\n string_pattern = r'[frbFRB]?\"\"\"[\\s\\S]*?\"\"\"|[frbFRB]?\\'\\'\\'[\\s\\S]*?\\'\\'\\'|[frbFRB]?\"(?:\\\\.|[^\"\\\\\\n])*\"|[frbFRB]?\\'(?:\\\\.|[^\\'\\\\\\n])*\\''\n comment_pattern = r'(#.*)'\n else:\n # 前端代码(JS/CSS/HTML)中的单、双、反引号内的字符串\n string_pattern = r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n # 排除了 http/https/ws/wss 等常见非引用URL引起的假注释误判\n sl_comment = r'(?<!https:)(?<!http:)(?<!ws:)(?<!wss:)//.*'\n ml_comment = r'/\\*[\\s\\S]*?\\*/'\n html_comment = r'<!--[\\s\\S]*?-->'\n comment_pattern = f'({sl_comment}|{ml_comment}|{html_comment})'\n \n # 整体正则模式：将字符串优先匹配，以此保护不被当作注释\n regex = f'({string_pattern})|({comment_pattern})'\n matches = re.finditer(regex, content)\n \n # 通过预计算所有的换行符位置来帮助后续快速得出字符索引所在的行号\n line_starts = [0]\n for m in re.finditer(r'\\n', content):\n line_starts.append(m.start() + 1)\n \n def get_line_num(idx):\n return bisect.bisect_right(line_starts, idx)\n \n comment_lines = set()\n \n for match in matches:\n if match.group(2): # 如果第2个捕获组(即注释)命中了\n start_idx = match.start(2)\n end_idx = match.end(2)\n \n start_line = get_line_num(start_idx)\n # end_idx 是开区间，计算结束行时需减一\n if end_idx > start_idx:\n end_line = get_line_num(end_idx - 1)\n else:\n end_line = start_line\n \n # 将此条注释所跨越的所有行号加入去重集合\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n \n # 验证该文件的注释总行数是否为严格等于 5\n if len(comment_lines) != 5:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"当你把代码写完后，不得进行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码文件里，缩进必须严格保持为2个空格","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 推断可能的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.py', '.java', '.c', '.cpp', '.cs', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n has_code = False\n \n # 定义验证缩进的辅助函数\n def is_indentation_valid(lines) -> bool:\n for line in lines:\n if line.strip() == '':\n continue\n # 匹配每一行行首的空白字符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent = match.group(1)\n # 若包含制表符，直接判为不符合\n if '\\t' in indent:\n return False\n # 若仅由空格组成，空格数量应为 2 的倍数\n if len(indent) % 2 != 0:\n # 忽略以 * 开头的行（用于豁免常规的多行注释中间行格式，如 /* \\n * \\n */ ）\n if not line.strip().startswith('*'):\n return False\n return True\n\n # 1. 检查工作空间路径中生成的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n if not is_indentation_valid(lines):\n return False\n \n # 2. 检查助手最后的一轮 response 中的 Markdown 代码块\n code_blocks = re.findall(r'```(?:[a-zA-Z]*)\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n has_code = True\n lines = block.split('\\n')\n if not is_indentation_valid(lines):\n return False\n \n return has_code","tags":["Style","Layout"]}]},{"round_id":6,"instruction":"接下来这个挑战稍微有点难度：结合时间的动态天气预测。请在侧边栏顶部新增一个“出发时间”的日期时间选择器（默认当前时间）。当点击规划路线时，请根据每段路程的耗时，动态计算出到达每一个站点的“预计到达时间（ETA）”，并将该时间显示在时间轴和地图的 Popup 中；同时天气数据也要根据ETA进行变化（模拟不同时间段天气不同）。此外，在地图区域的底部叠加一个“行程时间轴滑块（Slider）”，用户拖动滑块可以模拟旅程的推进，地图上当前时间所处的路段折线需要用高亮颜色特殊标注出来。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的生成的代码文件缩进严格保持为2个空格的规则调整为，在生成的代码文件里，缩进必须严格保持为4个空格，该要求后续所有轮次始终遵循；第二，后续所有轮次请始终遵循，确保在你生成的每一个代码文件里，其注释部分包含关键词 \"FIXME\" 的次数正好是1次。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径为空或不存在，直接返回False\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令推断，这是一个前端旅行地图项目，可能包含以下类型代码文件\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json', '.py'}\n has_code_file = False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生解码错误，说明该文件不是使用UTF-8编码，验证失败\n return False\n except Exception:\n # 忽略其他如权限或文件系统引起的非编码相关异常\n pass\n \n # 如果至少存在一个代码文件，且均未抛出解码错误，则视为满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 根据项目推断可能的代码文件类型\n valid_exts = {'.html', '.htm', '.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.py', '.mjs', '.cjs'}\n files_checked = 0\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的非业务代码目录\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n todo_count = 0\n \n # 1. 处理 HTML 注释\n if ext in ['.html', '.htm']:\n html_comments = re.findall(r'<!--[\\s\\S]*?-->', code)\n for c in html_comments:\n todo_count += c.count(\"TODO\")\n # 匹配后移除，防止后续重复计算\n code = re.sub(r'<!--[\\s\\S]*?-->', '', code)\n \n # 2. 处理 多行注释 (JS/TS/CSS 等，以及 HTML 中可能嵌入的样式和脚本的多行注释)\n if ext in ['.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.html', '.htm', '.mjs', '.cjs']:\n ml_comments = re.findall(r'/\\*[\\s\\S]*?\\*/', code)\n for c in ml_comments:\n todo_count += c.count(\"TODO\")\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n \n # 处理 Python 多行注释 (Docstrings)\n if ext in ['.py']:\n ml_comments = re.findall(r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"', code)\n for c in ml_comments:\n todo_count += c.count(\"TODO\")\n code = re.sub(r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"', '', code)\n \n ml_comments = re.findall(r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'', code)\n for c in ml_comments:\n todo_count += c.count(\"TODO\")\n code = re.sub(r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'', '', code)\n\n # 3. 处理 单行注释\n if ext in ['.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.html', '.htm', '.mjs', '.cjs']:\n # 使用 (?<!:)//.* 避免误匹配 URL 中的双斜杠 (如 http://)\n sl_comments = re.findall(r'(?<!:)//.*', code)\n for c in sl_comments:\n todo_count += c.count(\"TODO\")\n \n if ext in ['.py']:\n sl_comments = re.findall(r'#.*', code)\n for c in sl_comments:\n todo_count += c.count(\"TODO\")\n \n # 严格控制在每一个生成的代码文件中，\"TODO\"必须出现且仅出现2次\n if todo_count != 2:\n return False\n files_checked += 1\n \n # 至少要验证并检查了一个代码文件\n return files_checked > 0","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断本项目可能生成的主要代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和不相关的依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的首行\n first_line = f.readline()\n \n # 检查特定字符串是否在首行中\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件不可读，稳妥起见视为不满足\n return False\n \n # 确保项目中至少生成了代码文件才算满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 给定要求中的特定注释目标字符串\n target_string = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据用户前端地图项目的指令内容，推断可能的代码文件类型\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\n \n found_any_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方依赖或版本控制目录以加快遍历并防止误判\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果遇到无法读取的文件，则跳过\n continue\n \n # 忽略空文件\n if not lines:\n continue\n \n found_any_code_file = True\n \n # 检查开头部分（前5行）是否存在该独立换行的目标注释\n is_valid = False\n for line in lines[:5]:\n if target_string in line:\n is_valid = True\n break\n \n # 如果任意一个代码文件的开头前5行内没有找到目标注释，则要求未满足\n if not is_valid:\n return False\n \n # 如果工作区内存在至少一个代码文件且全部代码文件均满足要求，则返回 True\n return found_any_code_file","tags":["Content","Format"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def verify_even_line_count(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据项目推测可能生成的前端及相关代码文件类型\n code_extensions = {\n '.html', '.css', '.js', '.json', '.ts', \n '.jsx', '.tsx', '.vue', '.py'\n }\n\n has_code_file = False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # readlines() 会保留空行并作为单独的元素，长度即为总行数\n lines = f.readlines()\n line_count = len(lines)\n \n # 如果某个代码文件行数不是偶数，则判定不满足要求\n if line_count % 2 != 0:\n return False\n except Exception:\n # 若文件编码错误或无法读取，为了容错可跳过\n continue\n\n # 如果至少检查了一个代码文件且都为偶数行，满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"限制每个生成的代码文件中的注释数量必须刚好是5行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n # 定义常见的代码文件扩展名\n code_extensions = {'.py', '.js', '.ts', '.css', '.html', '.htm', '.jsx', '.tsx', '.vue'}\n \n # 匹配 JS/HTML/CSS 的字符串和注释\n # (?P<name>...) 为命名捕获组，利用同一层级的正则匹配可防止字符串内的注释符号被误判\n js_html_pattern = re.compile(\n r'(?P<html_comment><!--[\\s\\S]*?-->)|'\n r'(?P<block_comment>/\\*[\\s\\S]*?\\*/)|'\n r'(?P<line_comment>//[^\\n]*)|'\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n )\n\n # 匹配 Python 的字符串、文档注释（多行字符串）和单行注释\n # 注意：在 Python 中，通常只有 # 被严格算作注释，文档字符串属于字符串范畴，因此分开捕获，后续只统计 py_comment\n py_pattern = re.compile(\n r'(?P<py_tq>\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\')|'\n r'(?P<py_comment>#[^\\n]*)|'\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n )\n\n file_checked = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 无法读取的文件跳过\n continue\n \n file_checked = True\n \n # 记录每一行的起始字符索引\n line_starts = [0]\n for i, char in enumerate(content):\n if char == '\\n':\n line_starts.append(i + 1)\n \n # 通过字符索引二分查找对应的行号（从0开始计数）\n def get_line_num(idx):\n return bisect.bisect_right(line_starts, idx) - 1\n \n # 用于标记某一行是否包含注释\n is_comment = [False] * len(line_starts)\n \n if ext == '.py':\n pattern = py_pattern\n else:\n pattern = js_html_pattern\n \n # 遍历所有匹配项\n for match in pattern.finditer(content):\n group_name = match.lastgroup\n # 如果匹配到的组名是注释\n if group_name in ('html_comment', 'block_comment', 'line_comment', 'py_comment'):\n start_line = get_line_num(match.start())\n # match.end() 是匹配结束后的下一个字符索引，减1代表该匹配结尾的实际字符位置\n end_line = get_line_num(match.end() - 1) if match.end() > match.start() else start_line\n \n # 将这段注释所跨越的所有行标记为 True\n for l in range(start_line, end_line + 1):\n is_comment[l] = True\n \n total_comments = sum(is_comment)\n \n # 只要有任何一个代码文件的注释总行数不等于 5，即不符合要求\n if total_comments != 5:\n return False\n \n # 确保至少检查了一个代码文件，并且全部文件均符合要求\n return file_checked","tags":["Quantity","Exact Value"]},{"constraint":"当你把代码写完后，不得进行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在生成的代码文件里，缩进必须严格保持为4个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目中常见的代码文件扩展名\n valid_exts = {'.js', '.css', '.html', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略依赖库、版本控制目录及测试相关目录\n for ignore_dir in ['node_modules', '.git', '__tests__']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果读取文件失败（如非文本文件等情况），跳过\n continue\n\n for line in lines:\n # 去掉行尾的换行符\n line = line.rstrip('\\r\\n')\n \n # 仅去掉行首的空格和Tab，以便精确获取缩进字符\n stripped_line = line.lstrip(' \\t')\n \n # 忽略纯空行\n if not stripped_line:\n continue\n \n # 截取该行的前导空白字符（即缩进部分）\n leading_whitespaces = line[:len(line) - len(stripped_line)]\n \n # 规则1：缩进中绝对不能包含制表符(\\t)\n if '\\t' in leading_whitespaces:\n return False\n \n # 规则2：缩进空格数必须严格是4的倍数\n if len(leading_whitespaces) % 4 != 0:\n # 豁免规则：多行注释的中间续行（通常以 * 开头，用于对齐）\n if stripped_line.startswith('*'):\n continue\n # 若不是多行注释且不满足4的倍数，则直接判定为不满足要求\n return False\n\n # 必须至少包含一个代码文件，且全部验证通过才算满足要求\n return has_code_files","tags":["Style","Layout"]},{"constraint":"确保在你生成的每一个代码文件里，其注释部分包含关键词 \"FIXME\" 的次数正好是1次","validation_code":"def check_fixme_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见的代码文件扩展名\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', \n '.vue', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rb', '.php', '.sh'\n }\n # 需排除的常见非开发目录\n exclude_dirs = {\n 'node_modules', 'venv', '.git', '__pycache__', 'dist', 'build', '.idea', '.vscode'\n }\n \n # 构造正则表达式同时匹配“注释”和“字符串”，使得可以忽略字符串内的干扰项\n # 注意: Python的多行字符串(\"\"\" 和 ''')通常用作Docstring，被优先归类到comment组\n pattern = re.compile(\n r'(?P<comment>'\n r'/\\*[\\s\\S]*?\\*/' # 多行注释 (C/C++/Java/JS/TS/CSS等)\n r'|//[^\\n]*' # 单行注释 (C/C++/Java/JS/TS等)\n r'|<!--[\\s\\S]*?-->' # HTML/XML 注释\n r'|\"\"\"[\\s\\S]*?\"\"\"' # Python 多行注释 (双引号)\n r'|\\'\\'\\'[\\s\\S]*?\\'\\'\\'' # Python 多行注释 (单引号)\n r'|#.*' # 单行注释 (Python/Shell/Ruby等)\n r')'\n r'|(?P<string>'\n r'\"(?:\\\\.|[^\"\\\\])*\"' # 双引号字符串\n r'|\\'(?:\\\\.|[^\\'\\\\])*\\'' # 单引号字符串\n r'|`(?:\\\\.|[^`\\\\])*`' # 反引号字符串 (JS/TS等)\n r')'\n )\n \n files_checked = 0\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 使用 errors='ignore' 保证即使存在不完全兼容 utf-8 字符的生成文件也能顺利读取\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n # 遍历所有匹配项，仅保留属于 comment 分组的匹配结果\n comments = []\n for match in pattern.finditer(content):\n if match.group('comment'):\n comments.append(match.group('comment'))\n \n comments_text = '\\n'.join(comments)\n \n # 统计注释中 \"FIXME\" 出现的次数\n fixme_count = comments_text.count(\"FIXME\")\n \n # 若次数不恰好为 1，则不符合要求\n if fixme_count != 1:\n return False\n \n files_checked += 1\n \n # 只要验证了至少一个文件且全部通过，即视为满足要求\n return files_checked > 0","tags":["Content","Keyword"]}]},{"round_id":7,"instruction":"功能越来越完善了。现在来丰富地图的图层和兴趣点（POI）控制。在地图的右上角添加一个半透明的地图控制面板，包含两个区域：“地图样式”和“沿途设施”。“地图样式”提供下拉菜单，允许在“标准视图”、“暗黑模式”和“地形图”之间切换底层瓦片（可以使用不同的免费瓦片URL模拟）。“沿途设施”提供三个复选框：“加油站”、“餐厅”、“酒店”。当勾选某项时，请在当前绘制的路线两侧一定范围内，随机散布几个对应类型的定制小图标；取消勾选则让它们从地图上消失。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的当你把代码写完后不得进行任何测试的规则调整为，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过，该要求后续所有轮次始终遵循；第二，后续所有轮次请始终遵循，在你输出的任何一个代码文件里，务必出现\"OPTIMIZE\"。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端及常见 Web 项目的代码文件扩展名\n code_extensions = {\n '.html', '.js', '.css', '.ts', '.jsx', '.tsx', \n '.json', '.vue', '.svelte', '.py', '.txt', '.md', '.sh'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的无关目录或依赖库目录\n if '.git' in root or 'node_modules' in root or '__pycache__' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件，过滤掉可能存在的图片等二进制资源文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试使用 UTF-8 进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果发生解码错误，说明该文件包含非 UTF-8 编码的内容\n return False\n except Exception:\n # 遇到其他读取异常也可视为验证失败\n return False\n\n # 必须至少存在一个代码文件，且全部验证通过才算满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"def check_todo_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n ignore_dirs = {'node_modules', '.git', '__pycache__', 'dist', 'build', '__tests__'}\n \n file_todo_counts = []\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除需要忽略的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 避免 http:// 和 https:// 中的双斜杠被误判为单行注释\n content_clean = content.replace(\"http://\", \"\").replace(\"https://\", \"\")\n \n comments = []\n \n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css']:\n block_comments = re.findall(r'/\\*.*?\\*/', content_clean, re.DOTALL)\n line_comments = re.findall(r'//.*$', content_clean, re.MULTILINE)\n comments.extend(block_comments)\n comments.extend(line_comments)\n elif ext in ['.html', '.vue']:\n html_comments = re.findall(r'<!--.*?-->', content_clean, re.DOTALL)\n block_comments = re.findall(r'/\\*.*?\\*/', content_clean, re.DOTALL)\n line_comments = re.findall(r'//.*$', content_clean, re.MULTILINE)\n comments.extend(html_comments)\n comments.extend(block_comments)\n comments.extend(line_comments)\n elif ext in ['.py']:\n block_comments = re.findall(r'\"\"\"(.*?)\"\"\"', content_clean, re.DOTALL)\n block_comments_2 = re.findall(r\"'''(.*?)'''\", content_clean, re.DOTALL)\n line_comments = re.findall(r'#.*$', content_clean, re.MULTILINE)\n comments.extend(block_comments)\n comments.extend(block_comments_2)\n comments.extend(line_comments)\n \n comment_text = \"\\n\".join(comments)\n todo_count = comment_text.count(\"TODO\")\n file_todo_counts.append(todo_count)\n except Exception:\n continue\n \n # 如果没有扫描到任何有效的代码文件，则不满足要求\n if not file_todo_counts:\n return False\n \n # 理解1: 每个代码文件的注释中均包含且仅包含 2 次 TODO\n every_file_has_two = all(count == 2 for count in file_todo_counts)\n # 理解2: 整个项目所有代码文件的注释中合计包含 2 次 TODO\n sum_is_two = sum(file_todo_counts) == 2\n \n return every_file_has_two or sum_is_two","tags":["Content","Keyword"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义推断可能的代码文件扩展名\n valid_extensions = {'.js', '.html', '.css', '.py', '.ts', '.jsx', '.tsx', '.vue'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的无关目录，避免扫描非AI生成的依赖文件\n for exclude_dir in ['node_modules', '.git', '__pycache__']:\n if exclude_dir in dirs:\n dirs.remove(exclude_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含指定的编码格式字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果遇到文件无法读取等异常情况，视作不满足要求\n return False\n \n # 如果扫描过至少一个代码文件且都满足要求，返回True，否则返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n # 推断本项目为前端地图应用，可能包含以下类型的代码文件\n # 不包括不支持标准注释的 .json 文件\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n expected_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取开头部分（前 15 行足够覆盖\"开头处\"的要求）\n lines = [f.readline() for _ in range(15)]\n except Exception:\n continue\n \n found = False\n for line in lines:\n if not line:\n break\n if expected_text in line:\n # 检查是否为单独换行的注释（前面不能有正常代码逻辑，只能是空白符或注释符号）\n prefix = line[:line.find(expected_text)].strip()\n # 常见的注释符号字符集合\n if not prefix or all(c in ' \\t#/!*-<>' for c in prefix):\n found = True\n break\n \n # 如果某个代码文件没有在开头处包含单独换行的特定注释，则验证失败\n if not found:\n return False\n \n # 如果没有生成任何代码文件，则也认为未能实现要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def check_even_lines_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断，前端可能用到的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n has_code_file = False\n\n # 遍历当前项目工作区的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 筛选出属于代码类型的文件\n if any(file_name.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 判断总行数（包含空行）是否为偶数\n if len(lines) % 2 != 0:\n return False\n except Exception:\n # 对于不可读或编码异常的文件，选择忽略或抛出，这里选择跳过继续检查\n pass\n \n # 如果找到了代码文件且无一奇数行，则返回True；若没找到任何代码文件返回False\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"限制每个生成的代码文件中的注释数量必须刚好是5行","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义通常被认为是代码文件的扩展名集合\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', \n '.html', '.css', '.vue', '.json',\n '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n \n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非项目主体逻辑的干扰目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if '__pycache__' in dirs:\n dirs.remove('__pycache__')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n count = 0\n in_block = False\n in_html_block = False\n \n for line in lines:\n s = line.strip()\n if not s:\n continue\n \n # 针对 Python / Ruby / Shell 等脚本语言使用 # 作为注释符\n if ext in {'.py', '.sh', '.yml', '.yaml', '.rb'}:\n if s.startswith('#'):\n count += 1\n \n # 针对 JS / TS / CSS / C系语言等使用 // 以及 /* */ 作为注释符\n elif ext in {'.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.swift', '.kt', '.php', '.json'}:\n if in_block:\n count += 1\n if '*/' in s:\n in_block = False\n else:\n if s.startswith('/*'):\n count += 1\n if '*/' not in s:\n in_block = True\n elif s.startswith('//'):\n count += 1\n \n # 针对 HTML / Vue 等标记类文件，兼容 HTML注释 与 其内嵌的 JS/CSS注释\n elif ext in {'.html', '.vue', '.xml', '.svg'}:\n if in_html_block:\n count += 1\n if '-->' in s:\n in_html_block = False\n elif in_block:\n count += 1\n if '*/' in s:\n in_block = False\n else:\n if s.startswith('<!--'):\n count += 1\n if '-->' not in s:\n in_html_block = True\n elif s.startswith('/*'):\n count += 1\n if '*/' not in s:\n in_block = True\n elif s.startswith('//'):\n count += 1\n \n files_checked += 1\n # 任何一个代码文件的注释总行数如果不为 5，判定为不满足条件\n if count != 5:\n return False\n \n # 确保项目中至少成功检查了一个代码文件\n return files_checked > 0","tags":["Quantity","Exact Value"]},{"constraint":"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在生成的代码文件里，缩进必须严格保持为4个空格","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件后缀\n valid_extensions = {'.js', '.css', '.html', '.py', '.ts', '.jsx', '.tsx', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略依赖库、虚拟环境、Git版本控制等无关目录\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'venv', 'env'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果有文件读取异常可以跳过，继续检查其他代码文件\n continue\n \n for line in lines:\n r_stripped = line.rstrip()\n # 跳过仅包含空白字符的空行\n if not r_stripped:\n continue\n \n stripped = line.lstrip()\n # 提取行首所有的空白缩进字符\n leading_spaces = line[:len(line) - len(stripped)]\n \n # 规则1：严禁使用 Tab (\\t) 进行缩进\n if '\\t' in leading_spaces:\n return False\n \n # 忽略以 '*' 开头的行（常见的块状注释内部，经常会有 1 个空格或者不对齐的特殊情况）\n if stripped.startswith('*'):\n continue\n \n # 规则2：首部缩进的空格数必须严格为 4 的整数倍\n if len(leading_spaces) % 4 != 0:\n return False\n \n # 只要存在至少一个被检查的代码文件且都通过了验证，则符合要求\n return has_code_file","tags":["Style","Layout"]},{"constraint":"确保在你生成的每一个代码文件里，其注释部分包含关键词 \"FIXME\" 的次数正好是1次","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义项目中可能涉及的代码文件扩展名（推断项目主要涉及前端HTML/JS/CSS，也可能带有少许配置文件或后端Python）\n extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_files = False\n \n # 预编译正则表达式提取各语言注释块：\n # 1. /\\*.*?\\*/ -> 匹配 JS/CSS 的多行注释\n # 2. <!--.*?--> -> 匹配 HTML/Vue 的多行/单行注释\n # 3. (?<!http:)(?<!https:)//[^\\n]* -> 匹配 JS 单行注释（巧妙排除了 http:// 和 https:// 这种网址中的双斜杠）\n # 4. #[^\\n]* -> 匹配 Python 单行注释\n comment_pattern = re.compile(\n r'/\\*.*?\\*/|<!--.*?-->|(?<!http:)(?<!https:)//[^\\n]*|#[^\\n]*',\n re.DOTALL\n )\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的依赖项与版本控制目录，保留存放代码和测试用例的目录（如 __tests__）\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__', 'dist', 'build')]\n \n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in extensions:\n has_files = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取文件中的所有注释部分内容\n comments = comment_pattern.findall(content)\n \n # 在所有注释内容中，统计关键词 \"FIXME\" 出现的总次数\n fixme_count = sum(comment.count(\"FIXME\") for comment in comments)\n \n # 要求每个文件的注释中 FIXME 次数必须正好是1次\n if fixme_count != 1:\n return False\n \n # 最终只要确保扫描过至少一个代码文件即可算作验证通过\n return has_files","tags":["Content","Keyword"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你输出的任何一个代码文件里，务必出现\"OPTIMIZE\"","validation_code":"def verify_optimization_keyword(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果该文件中没有出现 \"OPTIMIZE\"，则直接判定为不满足要求\n if \"OPTIMIZE\" not in content:\n return False\n except Exception:\n # 如果文件读取发生异常（如编码错误），保守起见视为未满足要求\n return False\n\n # 确保项目中确实存在代码文件，并且所有的代码文件都通过了检查\n return has_code_file","tags":["Content","Keyword"]}]},{"round_id":8,"instruction":"用户辛苦规划好的路线需要保存。请在侧边栏最底部添加一个“保存行程”按钮。点击后弹出一个对话框，让用户输入“行程名称”。确认后，请将当前的完整数据（行程名、出发时间、起终点名称、途经点列表）序列化并保存到浏览器的 localStorage 中。接着，利用第一轮我们在顶部导航栏留下的空间，新增一个“我的行程”入口。点击它会打开一个右侧滑出的抽屉（Drawer）或模态框，里面用列表展示所有已保存的历史行程。点击列表中的任何一个行程，都要能把它重新读取并渲染到侧边栏表单和地图上。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的每个生成的代码文件中的注释数量必须刚好是5行的规则调整为，确保你生成的每一个代码文件都精确包含 6 行注释；第二，后续所有轮次请始终遵循，在你的回复里，如需引用文件或代码的具体位置，请务必采用markdown链接语法，格式要求如下：[filename.ts](src/filename.ts)、[filename.ts:42](src/filename.ts#L42)、[filename.ts:42-51](src/filename.ts#L42-L51)。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令推断可能的代码文件类型（前端Web应用开发）\n code_extensions = ('.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue')\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 仅针对代码文件进行编码校验\n if file_name.lower().endswith(code_extensions):\n found_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n # 尝试用 UTF-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦捕获到解码异常，说明不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能导致无法读取的异常（例如权限问题）\n pass\n\n # 只有当发现至少一个代码文件并且全都没有引发 UnicodeDecodeError 时，才判定为满足要求\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"def check_todo_count_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义前端及常见项目的代码文件扩展名\n code_exts = {\n '.js', '.jsx', '.ts', '.tsx', '.css', '.html', \n '.py', '.java', '.cpp', '.c', '.h', '.hpp', \n '.cs', '.php', '.rb', '.go', '.rs', '.swift', \n '.kt', '.vue', '.svelte'\n }\n \n # 构造正则表达式以匹配主流语言的注释形式\n # 1. // ... 单行注释 (JS, TS, C++ 等)\n # 2. # ... 单行注释 (Python, Ruby 等)\n # 3. /* ... */ 多行注释 (JS, TS, CSS 等)\n # 4. <!-- ... --> 多行注释 (HTML)\n comment_pattern = re.compile(\n r'//.*?$|#.*?$|/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->',\n re.MULTILINE\n )\n \n files_checked = 0\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 找出文件中的所有注释文本块\n comments = comment_pattern.findall(content)\n \n # 统计这些注释块中出现 \"TODO\" 的总次数\n todo_count = sum(comment.count(\"TODO\") for comment in comments)\n \n # 根据要求，每个生成的代码文件的注释中，\"TODO\" 必须出现且仅出现 2 次\n if todo_count != 2:\n return False\n \n files_checked += 1\n except Exception:\n # 如果遇到文件编码等读取异常，可做静默跳过处理，不影响常规代码逻辑的判断\n pass\n \n # 如果项目路径中至少找到了一个代码文件并且全满足上述条件，则验证通过\n return files_checked > 0","tags":["Content","Keyword"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义该项目中可能出现的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果由于非utf-8编码等原因导致读取失败，则视为不满足\n return False\n\n # 确保项目中确实生成了代码文件并且全部通过检查\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的前端与通用代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n has_checked_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无关目录，避免扫描第三方库或者隐藏文件\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_checked_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = []\n # 读取文件的前5行检查，以容忍可能存在的编码声明、空行等\n for _ in range(5):\n line = f.readline()\n if not line:\n break\n lines.append(line)\n except Exception:\n # 如果由于权限或者编码问题无法读取文件内容，视为验证失败\n return False\n \n # 判断这几行内是否存在目标注释内容\n found = any(target_comment in line for line in lines)\n \n if not found:\n return False\n \n # 如果项目中没有任何代码文件，应当视作未满足要求，否则返回True\n return has_checked_file","tags":["Content","Format"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，主要为前端地图应用开发，包含以下常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略通常不由 AI 直接生成的大型依赖目录或版本控制目录\n path_parts = root.split(os.sep)\n if 'node_modules' in path_parts or '.git' in path_parts:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 连同空行一并计算总行数\n if not content:\n line_count = 0 # 空文件视为 0 行\n else:\n line_count = len(content.split('\\n'))\n \n # 判断是否为偶数行\n if line_count % 2 != 0:\n return False\n except Exception:\n # 如果由于编码异常或其他原因无法读取文件，无法证明满足要求，保守返回False\n return False\n \n # 如果找到了至少一个代码文件且全部为偶数行，则返回True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"确保你生成的每一个代码文件都精确包含 6 行注释","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_extensions = {'.js', '.ts', '.html', '.css', '.vue', '.jsx', '.tsx', '.py'}\n file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无关目录\n if 'node_modules' in root or '.git' in root:\n continue\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n comment_lines = set()\n\n # JavaScript, TypeScript, CSS, JSX, TSX\n if ext in ['.js', '.ts', '.css', '.jsx', '.tsx']:\n # 匹配字符串以避免误判，并匹配单行和多行注释\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|((?<!:)\\/\\/.*|/\\*[\\s\\S]*?\\*/)'\n )\n for match in pattern.finditer(content):\n if match.group(2): # 如果是注释部分\n start = match.start(2)\n end = match.end(2)\n start_line = content.count('\\n', 0, start)\n end_line = content.count('\\n', 0, end)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n\n # HTML, Vue\n elif ext in ['.html', '.vue']:\n # 匹配 HTML 注释\n pattern_html = re.compile(r'(<!--[\\s\\S]*?-->)')\n for match in pattern_html.finditer(content):\n start = match.start(1)\n end = match.end(1)\n start_line = content.count('\\n', 0, start)\n end_line = content.count('\\n', 0, end)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n\n # 同时匹配文件中可能包含的内部 JS/CSS 注释，同上忽略字符串中内容\n pattern_js = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|((?<!:)\\/\\/.*|/\\*[\\s\\S]*?\\*/)'\n )\n for match in pattern_js.finditer(content):\n if match.group(2):\n start = match.start(2)\n end = match.end(2)\n start_line = content.count('\\n', 0, start)\n end_line = content.count('\\n', 0, end)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n\n # Python\n elif ext == '.py':\n # 匹配多行和单行字符串，避免将内部的 # 当作注释\n pattern = re.compile(\n r'(\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')|(#.*)'\n )\n for match in pattern.finditer(content):\n if match.group(2):\n start = match.start(2)\n end = match.end(2)\n start_line = content.count('\\n', 0, start)\n end_line = content.count('\\n', 0, end)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n\n # 检查每个代码文件的注释行数是否精确包含 6 行\n if len(comment_lines) != 6:\n return False\n\n return file_found","tags":["Quantity","Exact Value"]},{"constraint":"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在生成的代码文件里，缩进必须严格保持为4个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目常见的代码文件扩展名\n code_extensions = {'.js', '.ts', '.html', '.css', '.json', '.jsx', '.tsx', '.py'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能包含第三方代码的依赖目录与隐藏目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__', 'venv', 'env')]\n \n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 如果是完全的空行，则跳过不作缩进检查\n if not line.strip():\n continue\n \n # 提取前导空白字符串\n leading_whitespace = \"\"\n for char in line:\n if char in (' ', '\\t'):\n leading_whitespace += char\n else:\n break\n \n # 要求中明确提出必须为“空格”，若混入制表符则不符合要求\n if '\\t' in leading_whitespace:\n return False\n \n # 前导空格的数量必须是4的倍数\n if len(leading_whitespace) % 4 != 0:\n stripped = line.lstrip()\n # 兼容 JSDoc 等多行注释排版（例如 ` * `），排除 `**` 以防误判\n if stripped.startswith('*') and not stripped.startswith('**'):\n continue\n return False\n except Exception:\n # 若文件无法以utf-8读取，直接跳过处理下一文件\n continue\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"确保在你生成的每一个代码文件里，其注释部分包含关键词 \"FIXME\" 的次数正好是1次","validation_code":"import os\nimport re\n\ndef check_fixme_in_comments(response: str, workspace_path: str) -> bool:\n # 根据用户前端交互式地图应用指令推断，可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n # 正则表达式：同时匹配字符串和各种注释，以防提取到字符串中包含的假注释符号\n # 捕获组 1 (group 1) 仅用于匹配真实注释内容，字符串字面量将在组 0 匹配而被过滤\n pattern = re.compile(\n r'(/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->|(?<!:)\\/\\/.*|#.*)|'\n r'\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n )\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除常规的非产物目录，保留 __tests__ 因为这是指令明确要求的测试代码目录\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', 'dist', 'build', '__pycache__')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n # 如果文件无法按 UTF-8 读取（违反了指令的全局 UTF-8 编码要求），则直接视为不满足\n return False\n \n # 提取出当前文件中所有的注释内容（正则表达式中 group(1) 成功匹配到的即为注释内容）\n comments = [m.group(1) for m in pattern.finditer(code) if m.group(1)]\n \n # 统计当前文件所有注释中关键词 \"FIXME\" 的总出现次数\n fixme_count = sum(comment.count(\"FIXME\") for comment in comments)\n \n # 若任意一个代码文件的注释中 \"FIXME\" 次数不等于 1，说明不满足条件\n if fixme_count != 1:\n return False\n \n # 如果找到了代码文件且均通过了测试，则返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你输出的任何一个代码文件里，务必出现\"OPTIMIZE\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的前端及后端代码文件扩展名集合\n code_exts = {\n '.html', '.htm', '.css', '.scss', '.less',\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs', '.php', '.rb', '.sh'\n }\n\n has_code_file = False\n \n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉无关目录以防止误判依赖库和编译产物\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'dist', 'build', 'out'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对属于代码文件的后缀进行内容读取与检查\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 确保不检查空文件\n if content.strip():\n has_code_file = True\n # 若有任何一个代码文件不包含\"OPTIMIZE\"，则直接返回False\n if \"OPTIMIZE\" not in content:\n return False\n except Exception:\n # 忽略无法正常读取或非utf-8编码的异常文件\n pass\n \n # 只要存在至少一个代码文件且都通过了检查，即返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"在你的回复里，如需引用文件或代码的具体位置，请务必采用markdown链接语法，格式要求如下：[filename.ts](src/filename.ts)、[filename.ts:42](src/filename.ts#L42)、[filename.ts:42-51](src/filename.ts#L42-L51)","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":9,"instruction":"最后一轮，我们需要开发一个便于分享和打印的“导出路书”沉浸模式。在侧边栏新增一个“生成路书”按钮。点击后，页面发生平滑过渡：隐藏顶部导航栏、侧边栏和所有的地图悬浮控件；地图区域变为占据屏幕上方40%的空间（自动居中显示整条路线）；屏幕下方60%的空间则渲染一个美观的“路书卡片”网格，将时间轴、每个站点的ETA和天气预测以大字号、精美排版的形式展示。在这个沉浸模式的右上角，提供一个“打印/保存为PDF”的占位按钮，以及一个“退出路书”按钮用于恢复到之前的编辑界面状态。\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的每个生成的代码文件精确包含6行注释的规则调整为，针对每一个生成的代码文件，注释的行数需精准控制在7行；第二，后续所有轮次请始终遵循，你输出的单元测试文件经运行，需确保目标代码获得至少85%的语句覆盖率。","instruction_following_checklist":[{"constraint":"进行任何文件处理之前，切勿使用`pwd`确定当前的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你创建的代码文件无一例外，全都要用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，该项目为前端页面应用，推测可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果报 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能因权限或系统原因导致的读取报错\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请在生成的所有代码文件的注释中，严格控制关键词 \"TODO\" 出现且仅出现 2 次","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义我们要检查的代码文件扩展名集合\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.vue', '.py'}\n found_code_file = False\n\n def get_comments_from_code(content: str, ext: str) -> str:\n comments = []\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.css']:\n # 该正则按从左到右匹配优先级，巧妙跳过字符串内包含的伪注释符号（如URL中的 //）\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\")|' # Group 1: 双引号字符串\n r\"('(?:\\\\.|[^'\\\\])*')|\" # Group 2: 单引号字符串\n r'(`(?:\\\\.|[^`\\\\])*`)|' # Group 3: 模板字符串（反引号）\n r'(<!--[\\s\\S]*?-->)|' # Group 4: HTML注释\n r'(/\\*[\\s\\S]*?\\*/)|' # Group 5: 多行块注释\n r'(//.*)' # Group 6: 单行注释\n )\n for match in pattern.finditer(content):\n # 如果匹配到了对应的注释组，则收集起来\n if match.group(4):\n comments.append(match.group(4))\n elif match.group(5):\n comments.append(match.group(5))\n elif match.group(6):\n comments.append(match.group(6))\n \n elif ext == '.py':\n pattern = re.compile(\n r'(\"\"\"[\\s\\S]*?\"\"\")|' # Group 1: 三双引号字符串/注释\n r\"('''[\\s\\S]*?''')|\" # Group 2: 三单引号字符串/注释\n r'(\"(?:\\\\.|[^\"\\\\])*\")|' # Group 3: 双引号\n r\"('(?:\\\\.|[^'\\\\])*')|\" # Group 4: 单引号\n r'(#.*)' # Group 5: 井号单行注释\n )\n for match in pattern.finditer(content):\n if match.group(1):\n comments.append(match.group(1))\n elif match.group(2):\n comments.append(match.group(2))\n elif match.group(5):\n comments.append(match.group(5))\n \n return \" \".join(comments)\n\n # 遍历工作目录\n for root, dirs, files in os.walk(workspace_path):\n # 忽略掉无关的依赖或版本管理文件夹，保留由AI生成的代码及__tests__测试代码\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n found_code_file = True\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取该文件中的所有注释内容\n comments_text = get_comments_from_code(content, ext)\n \n # 统计关键词 \"TODO\" 的出现次数\n todo_count = comments_text.count(\"TODO\")\n \n # 如果存在任何一个代码文件，其注释中的 TODO 数量不等于 2，即不满足要求\n if todo_count != 2:\n return False\n \n # 如果没有找到任何代码文件，则也视为失败\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 根据用户指令的前端项目特征，推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的依赖和版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 验证首行是否包含指定的编码格式声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n has_code_files = True\n except Exception:\n # 对于无法作为文本打开的异常文件可以选择跳过\n continue\n\n # 只有当至少验证过一个代码文件且全部通过时，才算满足要求\n return has_code_files","tags":["Content","Format"]},{"constraint":"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件后缀名\n code_extensions = {'.py', '.js', '.ts', '.html', '.css', '.jsx', '.tsx', '.vue'}\n # 目标注释声明\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制和常见依赖包目录\n if '.git' in root or 'node_modules' in root or '__pycache__' in root:\n continue\n \n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果遇到无法用UTF-8读取的异常情况，跳过或视为不合规，此处为避免非文本文件导致的错误选择跳过\n continue\n \n found_in_header = False\n # 检查文件的前15行（视为文件开头处）\n for line in lines[:15]:\n cleaned = line.strip()\n # 剥离各种可能的注释符号\n for comment_mark in ['//', '#', '/*', '*/', '<!--', '-->', '*']:\n cleaned = cleaned.replace(comment_mark, '')\n cleaned = cleaned.strip()\n \n # 验证剥离注释符后是否等于目标声明（以此保障它是单独一行的注释）\n if cleaned == target_text:\n found_in_header = True\n break\n \n # 若任意一个代码文件缺少此头部声明，则说明未完成要求\n if not found_in_header:\n return False\n \n # 确保项目中至少存在一个被检测的代码文件，并且全部通过了检查\n return has_code_file","tags":["Content","Format"]},{"constraint":"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数","validation_code":"import os\nimport re\n\ndef check_even_lines(response: str, workspace_path: str) -> bool:\n # 定义前端项目中常规会被生成的代码文件扩展名\n valid_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue'}\n # 忽略不需要检查的目录\n exclude_dirs = {'node_modules', '.git', '.idea', '.vscode', 'dist', 'build', '__pycache__'}\n \n files_checked = 0\n \n # 1. 优先检查 workspace_path 中的文件\n if workspace_path and os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按换行符分割以获取包含空行在内的总行数\n lines = content.split('\\n')\n if len(lines) % 2 != 0:\n return False\n \n files_checked += 1\n except Exception:\n pass\n \n # 如果在工作区中找到了代码文件并全都通过了检查，则认为满足条件\n if files_checked > 0:\n return True\n \n # 2. 如果工作区未找到符合要求的文件，退而检查 response 中的 Markdown 代码块\n # 匹配 ``` 开头且以 ``` 结尾的代码块内容\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n \n # 如果没有任何代码文件和代码块，认为未满足生成代码文件的要求\n if not code_blocks:\n return False\n \n for block in code_blocks:\n lines = block.split('\\n')\n if len(lines) % 2 != 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"针对每一个生成的代码文件，注释的行数需精准控制在7行","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n def count_comment_lines(file_path):\n ext = os.path.splitext(file_path)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except UnicodeDecodeError:\n return -1\n \n comment_lines = set()\n \n # 处理 C系、Web系语言\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.css', '.c', '.cpp', '.java', '.cs'}:\n pattern = re.compile(\n r'(?P<line_comment>//[^\\n]*)|(?P<block_comment>/\\*.*?\\*/)|(?P<string1>\\'(?:\\\\.|[^\\\\\\'])*\\')|(?P<string2>\"(?:\\\\.|[^\\\\\"])*\")|(?P<string3>`(?:\\\\.|[^\\\\`])*`)',\n re.DOTALL\n )\n for m in pattern.finditer(code):\n if m.group('line_comment') or m.group('block_comment'):\n if m.group('line_comment'):\n start_pos = m.start()\n if start_pos > 0 and code[start_pos-1] == ':':\n continue\n start_pos = m.start()\n end_pos = m.end()\n start_line = code.count('\\n', 0, start_pos)\n end_line = code.count('\\n', 0, end_pos)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n \n # 处理 脚本系语言\n elif ext in {'.py', '.sh', '.rb', '.yml', '.yaml'}:\n pattern = re.compile(\n r'(?P<line_comment>#[^\\n]*)|(?P<string3>\"\"\".*?\"\"\")|(?P<string4>\\'\\'\\'.*?\\'\\'\\')|(?P<string1>\\'(?:\\\\.|[^\\\\\\'])*\\')|(?P<string2>\"(?:\\\\.|[^\\\\\"])*\")',\n re.DOTALL\n )\n for m in pattern.finditer(code):\n if m.group('line_comment'):\n start_pos = m.start()\n end_pos = m.end()\n start_line = code.count('\\n', 0, start_pos)\n end_line = code.count('\\n', 0, end_pos)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n \n # 处理 标记系语言及可能混合的代码\n elif ext in {'.html', '.vue', '.xml'}:\n pattern = re.compile(\n r'(?P<html_comment><!--.*?-->)|(?P<line_comment>//[^\\n]*)|(?P<block_comment>/\\*.*?\\*/)|(?P<string1>\\'(?:\\\\.|[^\\\\\\'])*\\')|(?P<string2>\"(?:\\\\.|[^\\\\\"])*\")|(?P<string3>`(?:\\\\.|[^\\\\`])*`)',\n re.DOTALL\n )\n for m in pattern.finditer(code):\n if m.group('html_comment') or m.group('line_comment') or m.group('block_comment'):\n if m.group('line_comment'):\n start_pos = m.start()\n if start_pos > 0 and code[start_pos-1] == ':':\n continue\n start_pos = m.start()\n end_pos = m.end()\n start_line = code.count('\\n', 0, start_pos)\n end_line = code.count('\\n', 0, end_pos)\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n else:\n return -1\n\n return len(comment_lines)\n\n valid_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.css', '.c', '.cpp', '.java', '.cs',\n '.py', '.sh', '.rb', '.yml', '.yaml',\n '.html', '.vue', '.xml'\n }\n\n files_checked = 0\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', 'venv', 'env', '__pycache__', '.idea', '.vscode'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无关的构建与包目录，保留 __tests__ 等用户生成源码的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n cnt = count_comment_lines(file_path)\n if cnt != -1:\n files_checked += 1\n # 核心验证逻辑：注释行数必须精准控制在 7 行\n if cnt != 7:\n return False\n except Exception:\n pass\n\n return files_checked > 0","tags":["Quantity","Exact Value"]},{"constraint":"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在生成的代码文件里，缩进必须严格保持为4个空格","validation_code":"def verify_indentation_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端与后端代码文件扩展名\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue', '.json', '.py'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于某种原因无法读取文件，则跳过继续检查其他文件\n continue\n\n for line in lines:\n # 匹配行首的前导空白字符和第一个非空白字符及其后的内容\n match = re.match(r'^([ \\t]+)(\\S.*)', line)\n if match:\n leading_space = match.group(1)\n content = match.group(2)\n\n # 如果包含制表符，直接判定为不符合要求\n if '\\t' in leading_space:\n return False\n\n # 检查前导空格数是否严格为 4 的倍数\n if len(leading_space) % 4 != 0:\n # 豁免多行注释中以 '*' 开头的行（例如 /**\\n * 这种常规对齐格式）\n if not content.startswith('*'):\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"确保在你生成的每一个代码文件里，其注释部分包含关键词 \"FIXME\" 的次数正好是1次","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n # 定义前端应用通常会涉及到的代码文件扩展名\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.css', '.html', '.htm', '.vue', '.py'}\n has_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制和常见的依赖/缓存目录，不将它们作为生成的代码文件检查\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if '__pycache__' in dirs:\n dirs.remove('__pycache__')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n has_files = True\n filepath = os.path.join(root, file)\n \n try:\n # 按照用户指令要求，所有文件均为 UTF-8 编码\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件无法读取，视同验证失败\n return False\n \n comments = []\n \n # 使用正则表达式匹配，确保准确提取注释，忽略字符串中的干扰内容\n if ext in {'.html', '.htm', '.vue'}:\n # 匹配 HTML 注释、字符串、JS/CSS 多行注释、JS/CSS 单行注释\n pattern = re.compile(\n r'(<!--[\\s\\S]*?-->)' # Group 1: HTML 注释\n r'|(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)' # Group 2: 字符串（避免误判）\n r'|(/\\*[\\s\\S]*?\\*/)' # Group 3: 多行注释\n r'|(//[^\\n]*)' # Group 4: 单行注释\n )\n for match in pattern.finditer(content):\n if match.group(1):\n comments.append(match.group(1))\n elif match.group(3):\n comments.append(match.group(3))\n elif match.group(4):\n comments.append(match.group(4))\n \n elif ext in {'.py'}:\n # 匹配 Python 字符串（含多行字符串）和单行注释\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\')'\n r'|(#.*)'\n )\n for match in pattern.finditer(content):\n if match.group(2):\n comments.append(match.group(2))\n \n else:\n # 匹配 JS/TS/CSS 等文件的字符串、多行注释、单行注释\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(/\\*[\\s\\S]*?\\*/)'\n r'|(//[^\\n]*)'\n )\n for match in pattern.finditer(content):\n if match.group(2):\n comments.append(match.group(2))\n elif match.group(3):\n comments.append(match.group(3))\n \n # 统计当前代码文件中，所有注释里出现 \"FIXME\" 的总次数\n fixme_count = sum(c.count(\"FIXME\") for c in comments)\n \n # 如果不是恰好出现1次，直接判定失败\n if fixme_count != 1:\n return False\n \n return has_files","tags":["Content","Keyword"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你输出的任何一个代码文件里，务必出现\"OPTIMIZE\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见代码文件后缀集合\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的第三方依赖或版本控制文件夹\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若当前代码文件中不存在要求的关键词 \"OPTIMIZE\"，则直接返回 False\n if 'OPTIMIZE' not in content:\n return False\n except Exception:\n # 如果由于某种原因（如编码问题）无法读取该文件，则安全跳过\n pass\n \n # 如果至少检查了一个代码文件，并且所有代码文件都包含 \"OPTIMIZE\"，则满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"在你的回复里，如需引用文件或代码的具体位置，请务必采用markdown链接语法，格式要求如下：[filename.ts](src/filename.ts)、[filename.ts:42](src/filename.ts#L42)、[filename.ts:42-51](src/filename.ts#L42-L51)","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少85%的语句覆盖率","validation_code":"","tags":["Workflow","Testing"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"你好，我需要开发一款集成路线规划和天气数据的交互式旅行地图应用。第一步，请帮我搭建整个项目的前端基础骨架。要求如下：1. 创建一个包含项目Logo和用户头像的顶部导航栏。2. 左侧创建一个占据一定宽度的可折叠侧边栏（里面先只放一个“行程规划”的标题占位）。3. 右侧剩余的屏幕空间作为地图主区域，目前用一个带网格背景的灰色div来占位即可。4. 在主区域右下角放置一个悬浮的“新建行程”按钮（FAB）。整体页面要求是自适应的现代简洁风格，先不要引入真实的地图库，只要把结构和样式搭好即可。\\n另外请注意，后续所有轮次都必须严格遵循以下要求：请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-；对于产出的所有代码文件，务必在结尾以独立的一行写入注释，明确标示出：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码相关文件扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rb', '.php', \\n '.sh', '.json', '.xml', '.yaml', '.yml'\\n }\\n\\n # 如果工作区路径不存在或不是目录，则认为没有违反规则\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选出代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是使用合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 如果因为其他权限等原因无法读取文件，选择跳过处理\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_todo_in_comments(response: str, workspace_path: str) -> bool:\\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', '.vue',\\n '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.sh', '.rb', '.go', '.rs'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见非代码和第三方依赖目录，避免误判和过多无效计算\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 读取错误则忽略该文件\\n continue\\n \\n comments = []\\n \\n # 根据不同语言的文件后缀，采用恰当的正则提取注释部分\\n # 使用 [\\\\s\\\\S]*? 来匹配跨行注释，避免使用全局 re.DOTALL 导致单行注释越界吞噬\\n if ext in ['.html', '.htm', '.vue']:\\n # 匹配 HTML注释, JS/CSS跨行注释, 以及排除如 https:// 这类带有冒号前缀的单行注释\\n pattern = r'<!--[\\\\s\\\\S]*?-->|/\\\\*[\\\\s\\\\S]*?\\\\*/|(?<!:)//.*'\\n comments = re.findall(pattern, content)\\n \\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.go', '.rs']:\\n pattern = r'/\\\\*[\\\\s\\\\S]*?\\\\*/|(?<!:)//.*'\\n comments = re.findall(pattern, content)\\n \\n elif ext in ['.py', '.rb', '.sh']:\\n pattern = r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"|#.*'\\n comments = re.findall(pattern, content)\\n \\n # 统计所有注释片段中 \\\"TODO\\\" 出现的总次数\\n todo_count = sum(c.count('TODO') for c in comments)\\n \\n # 严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\\n if todo_count != 2:\\n return False\\n \\n # 如果项目中没有任何代码文件，也判定为不符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断前端项目可能生成的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的首行\\n first_line = f.readline()\\n \\n # 检查首行是否包含规定的编码格式字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法使用utf-8读取或其他异常，同样认为不符合要求\\n return False\\n \\n # 如果没有找到任何代码文件，无法证明其满足了\\\"每一个输出的代码文件...\\\"的要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"对于产出的所有代码文件，务必在结尾以独立的一行写入注释，明确标示出：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 涵盖前端基础骨架及常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n code_files_count = 0\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略版本控制和第三方模块目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 无法读取文件时视为不合规\\n return False\\n \\n # 提取所有非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n # 如果代码文件全是空行，则显然不符合具有结尾注释的要求\\n if not non_empty_lines:\\n return False\\n \\n # 检查文件结尾（倒数 3 个非空行）是否包含目标注释内容\\n # 此处容忍了块注释结束符可能单独占据倒数第一行的情况（如 html的 --> 符号）\\n found_target = False\\n for line in non_empty_lines[-3:]:\\n if target_comment in line:\\n found_target = True\\n break\\n \\n # 只要有一个代码文件的结尾没有包含目标注释，就直接返回 False\\n if not found_target:\\n return False\\n\\n # 需保证至少验证到了一个代码文件\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"基础骨架搭得很不错。现在我们要让右侧的地图区域活起来。请引入 Leaflet.js（或者类似的开源地图库）来替换掉之前的占位网格，加载默认的开源地图瓦片。接下来实现一个基础交互：允许用户在地图上任意位置点击，点击后在该处放置一个蓝色的“位置”标记（Marker）。当用户点击这个标记时，在标记上方弹出一个提示框（Popup），里面显示该点的具体经纬度坐标。确保地图支持鼠标拖拽平移和滚轮缩放功能。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的在代码文件结尾添加指定AI生成注释的规则调整为，在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，请确保你输出的每一个代码文件，包含空行在内的总行数均为偶数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区目录不存在，可以直接视为未违反规则\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 推断可能的代码文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.json', '.py', '.vue', \\n '.md', '.txt', '.xml'\\n }\\n\\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件后缀并统一转小写进行匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 若是指定的代码文件，则进行编码检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 utf-8 规则进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 只要有一个文件出现解码异常，就说明不是全部采用 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略不可读等其他无关的文件系统异常\\n pass\\n\\n # 若所有代码文件都能被正确解析为 UTF-8，则符合要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"def check_todo_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对 Python 文件的字符串与注释提取（含单/双/三引号，#注释）\\n py_pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|'\\n r'(?P<comment>\\\"{3}[\\\\s\\\\S]*?\\\"{3}|\\\\'{3}[\\\\s\\\\S]*?\\\\'{3}|#.*)'\\n )\\n \\n # 针对 JS/TS 等文件的字符串与注释提取（含反引号、双/单引号， //与/* */注释）\\n js_pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*)'\\n )\\n \\n # 针对 CSS 等文件的字符串与注释提取（/* */注释）\\n css_pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|'\\n r'(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n \\n todo_count = 0\\n\\n # 辅助函数：给定正则与内容，统计有效注释中的 \\\"TODO\\\" 个数\\n def count_in_pattern(content, pattern):\\n count = 0\\n for match in pattern.finditer(content):\\n comment = match.group('comment')\\n if comment:\\n count += comment.count(\\\"TODO\\\")\\n return count\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的外部无关目录，防止混入外部库中的代码产生干扰\\n if any(part in root for part in ['.git', 'node_modules', '__pycache__', 'venv', 'env', '.idea', '.vscode']):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java',\\n '.html', '.htm', '.vue',\\n '.css', '.scss', '.less',\\n '.py'}:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 根据不同扩展名执行对应的正则统计逻辑\\n if ext in {'.py'}:\\n todo_count += count_in_pattern(content, py_pattern)\\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java'}:\\n todo_count += count_in_pattern(content, js_pattern)\\n elif ext in {'.css', '.scss', '.less'}:\\n todo_count += count_in_pattern(content, css_pattern)\\n elif ext in {'.html', '.htm', '.vue'}:\\n # 1. HTML 原生注释： <!-- -->\\n html_comments = re.findall(r'<!--[\\\\s\\\\S]*?-->', content)\\n for comment in html_comments:\\n todo_count += comment.count(\\\"TODO\\\")\\n \\n # 2. 提取 <script> 内容进行 JS 注释分析\\n script_contents = re.findall(r'<script\\\\b[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for sc in script_contents:\\n todo_count += count_in_pattern(sc, js_pattern)\\n \\n # 3. 提取 <style> 内容进行 CSS 注释分析\\n style_contents = re.findall(r'<style\\\\b[^>]*>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n for st in style_contents:\\n todo_count += count_in_pattern(st, css_pattern)\\n \\n # 验证是否严格控制并只出现了两次 TODO\\n return todo_count == 2\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_first_line_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目描述推断主要的前端代码文件格式\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_checked_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方依赖、版本控制以及其他无关目录\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '.vscode', '.idea']]\\n \\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n has_checked_file = True\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查文件是否为空，或者第一行中是否未包含指定的编码标识\\n if not first_line or '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 出现无法读取文件等异常情况，也视作不满足要求\\n return False\\n \\n # 确保至少验证过一个代码文件\\n return has_checked_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 前端相关的常见代码文件后缀\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n has_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关的依赖和版本管理目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于编码等问题无法读取文件，直接视为未满足\\n return False\\n \\n found_in_head = False\\n # 检查文件的开头部分（设定检查前10行，保证兼容第一行编码声明的情况）\\n for line in lines[:10]:\\n if target_str in line:\\n found_in_head = True\\n break\\n \\n # 如果某个代码文件的开头没有该目标字符串，则直接判定为不满足要求\\n if not found_in_head:\\n return False\\n \\n # 如果所有的代码文件都满足要求，并且至少存在一个代码文件，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请确保你输出的每一个代码文件，包含空行在内的总行数均为偶数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义该前端项目可能会用到的常见代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的版本控制和第三方依赖包目录，以防干扰判断\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符拆分，获取包含空行在内的总行数\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果有文件读取报错（比如不可读的非文本文件等）可以直接跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"地图已经可以交互了，接下来我们完善左侧边栏的路线规划输入区。请在侧边栏中添加“出发地”和“目的地”两个文本输入框，并在下方放置一个“规划路线”的按钮。为了跑通流程，当用户在输入框随意输入内容并点击按钮时，请在地图上随机或者基于固定坐标生成两个不同颜色（比如绿色代表起点，红色代表终点）的标记，并在它们之间绘制一条模拟的实线折线（Polyline）作为路线。绘制完成后，地图需要自动缩放和平移，以刚好完整显示出这条路线。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的输出代码文件包含空行在内总行数为偶数的规则调整为，请确保你输出的每一个代码文件，包含空行在内的总行数均为奇数；第二，后续所有轮次针对每一个生成的代码文件，注释的行数需精准控制在 4 行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推测可能包含的代码相关文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.json', '.py', '.md', '.txt'\\n }\\n \\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制方式读取文件内容\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n \\n # 尝试用 utf-8 解码，如果不抛出 UnicodeDecodeError，则说明符合 utf-8 编码要求\\n content.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 一旦有文件无法用 utf-8 解码，直接返回 False\\n return False\\n except Exception:\\n # 处理如文件无权限读取等意外情况（通常不应作为非 utf-8 判断依据，故跳过）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n total_todo_count = 0\\n file_counts = []\\n \\n # 定义需要检查的常见代码文件扩展名\\n code_exts = {'.html', '.htm', '.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.py'}\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的非业务代码或依赖目录，避免引入第三方代码的干扰\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'venv', 'env', 'dist', 'build'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_files = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n spans = []\\n # 针对不同扩展名使用对应的正则提取注释的span起始位置\\n if ext in {'.html', '.htm', '.vue'}:\\n spans.extend([m.span() for m in re.finditer(r'<!--[\\\\s\\\\S]*?-->', code)])\\n # HTML/Vue 文件内可能包含 script 和 style，同样提取其对应的注释片段\\n spans.extend([m.span() for m in re.finditer(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', code)])\\n spans.extend([m.span() for m in re.finditer(r'//.*', code)])\\n elif ext in {'.js', '.ts', '.css', '.jsx', '.tsx'}:\\n spans.extend([m.span() for m in re.finditer(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', code)])\\n spans.extend([m.span() for m in re.finditer(r'//.*', code)])\\n elif ext in {'.py'}:\\n spans.extend([m.span() for m in re.finditer(r'#.*', code)])\\n spans.extend([m.span() for m in re.finditer(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', code)])\\n spans.extend([m.span() for m in re.finditer(r\\\"'''[\\\\s\\\\S]*?'''\\\", code)])\\n \\n # 合并重叠的注释区间（防止不同的正则表达式重复捕获同一内容）\\n spans.sort(key=lambda x: x[0])\\n merged_spans = []\\n for s in spans:\\n if not merged_spans:\\n merged_spans.append(s)\\n else:\\n last = merged_spans[-1]\\n if s[0] <= last[1]:\\n # 有重叠，合并区间\\n merged_spans[-1] = (last[0], max(last[1], s[1]))\\n else:\\n merged_spans.append(s)\\n \\n # 统计当前代码文件注释中的 \\\"TODO\\\" 出现次数\\n file_todo_count = 0\\n for start, end in merged_spans:\\n comment_text = code[start:end]\\n file_todo_count += comment_text.count(\\\"TODO\\\")\\n \\n file_counts.append(file_todo_count)\\n total_todo_count += file_todo_count\\n\\n if not has_code_files:\\n return False\\n \\n # 满足以下任一条件均视为达标：\\n # 1. 整个项目中所有生成的代码文件注释内 \\\"TODO\\\" 的总数为2\\n # 2. 或者每个生成的代码文件注释内 \\\"TODO\\\" 的数量恰好均等于 2\\n if total_todo_count == 2:\\n return True\\n \\n if len(file_counts) > 0 and all(c == 2 for c in file_counts):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据用户指令的前端项目特征，推断可能的代码文件扩展名\\n code_exts = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 如果存在第三方库等常见忽略文件夹，可跳过，但通常AI生成的纯代码目录下只有其编写的文件\\n for file in files:\\n if any(file.endswith(ext) for ext in code_exts):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含规定的编码格式字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于某种原因（如编码非UTF-8等）无法读取，则直接判定为不满足\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 定义常见的前端/后端代码文件扩展名\\n code_exts = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n \\n file_count = 0\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_count += 1\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于编码或其他原因导致无法读取文件，认为要求未满足\\n return False\\n \\n # 检查该文件的开头处（前 10 行）\\n found = False\\n for line in lines[:10]:\\n if target_str in line:\\n # 剔除目标字符串，留下周围的字符\\n remainder = line.replace(target_str, '').strip()\\n # 遍历并剔除各类常见编程语言的注释相关符号\\n for sym in ['<!--', '-->', '/*', '*/', '//', '#', '{/*', '*/}', '--']:\\n remainder = remainder.replace(sym, '')\\n # 若剩余部分没有任何可见字符，则符合\\\"单独换行添加注释\\\"的要求\\n if not remainder.strip():\\n found = True\\n break\\n \\n # 如果当前文件没有符合要求的注释，整体判定为不满足\\n if not found:\\n return False\\n \\n # 至少要生成了相关代码文件且全部满足要求\\n if file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请确保你输出的每一个代码文件，包含空行在内的总行数均为奇数\",\n \"validation_code\": \"import os\\n\\ndef check_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\\n # 如果工作区路径不存在或是个文件而不是目录，则不满足要求\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）和 node_modules 依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 通过换行符拆分文本，统计包含空行的总行数\\n lines_count = len(content.split('\\\\n'))\\n \\n # 若发现包含空行在内的总行数为偶数，则不满足要求\\n if lines_count % 2 == 0:\\n return False\\n except Exception:\\n # 如果读取过程中出现异常（例如权限不足），跳过处理\\n continue\\n \\n # 如果找到了至少一个代码文件并且全为奇数行，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在 4 行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n def count_comments_in_code(code_str: str, file_ext: str) -> int:\\n file_ext = file_ext.lower()\\n \\n # 针对不同语言扩展名构建模式匹配\\n # 在正则中，将字符串字面量的匹配放在最高优先级。一旦被匹配为字符串并不会处理内部的注释符号，这就有效避免了误判。\\n if file_ext in ['.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less']:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<multiline>/\\\\*.*?\\\\*/)'\\n r'|(?P<singleline>//[^\\\\n]*)',\\n re.DOTALL\\n )\\n elif file_ext in ['.html', '.htm', '.vue', '.svelte']:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<html_comment><!--.*?-->)'\\n r'|(?P<multiline>/\\\\*.*?\\\\*/)'\\n r'|(?P<singleline>//[^\\\\n]*)',\\n re.DOTALL\\n )\\n elif file_ext in ['.py']:\\n pattern = re.compile(\\n r'(?P<string>\\\"\\\"\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"\\\"\\\"|\\\\'\\\\'\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<singleline>#[^\\\\n]*)',\\n re.DOTALL\\n )\\n else:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<html_comment><!--.*?-->)'\\n r'|(?P<multiline>/\\\\*.*?\\\\*/)'\\n r'|(?P<singleline>//[^\\\\n]*|#[^\\\\n]*)',\\n re.DOTALL\\n )\\n\\n comment_lines = set()\\n \\n # 预先找到所有的换行符位置用于二分查找获得对应行号\\n line_starts = [0]\\n for m in re.finditer(r'\\\\n', code_str):\\n line_starts.append(m.start() + 1)\\n \\n def get_line(idx: int) -> int:\\n return bisect.bisect_right(line_starts, idx)\\n \\n # 找出匹配块并计算被注释占据的所在行集合\\n for m in pattern.finditer(code_str):\\n if m.lastgroup in ['multiline', 'singleline', 'html_comment']:\\n start_idx = m.start()\\n end_idx = m.end()\\n \\n if m.lastgroup == 'singleline':\\n line = get_line(start_idx)\\n comment_lines.add(line)\\n else:\\n start_line = get_line(start_idx)\\n # end_idx代表匹配块的终点后一个索引，如果是多行，实际在文本结束的前一个字符位置计算行号\\n end_line = get_line(end_idx - 1) if end_idx > start_idx else start_line\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n \\n return len(comment_lines)\\n\\n valid_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.json', '.vue', '.svelte'}\\n code_files_count = 0\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和第三方依赖目录以加快并确保检索准确\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for f in files:\\n # 略过隐藏文件\\n if f.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(f)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, f)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f_in:\\n content = f_in.read()\\n except UnicodeDecodeError:\\n continue # 略过可能为二进制或编码非utf-8的文件\\n \\n comment_count = count_comments_in_code(content, ext)\\n \\n # 若任意一个代码文件的注释行数不恰好等于4，验证即为不通过\\n if comment_count != 4:\\n return False\\n \\n code_files_count += 1\\n\\n # 要求必须至少存在一个代码文件被验证且成功通过\\n return code_files_count > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"真实的自驾游通常会有多个中途停靠点。请对刚才的路线表单进行复杂化升级：在起点和终点输入框之间增加一个“添加途经点”的按钮。用户点击后可以动态新增最多5个途经点输入框。最关键的是，请实现拖拽排序功能（Drag and Drop），让用户可以通过拖拽这些输入框来调整途经点的先后顺序。每一个途经点输入框旁还需要一个删除按钮。再次点击“规划路线”时，地图上的折线必须严格按照“起点 -> 排序后的途经点 -> 终点”的顺序重新连接绘制，并用带有数字1、2、3...的图标标出途经点的位置。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的每一个生成的代码文件注释行数需精准控制在4行的规则调整为，限制每个生成的代码文件中的注释数量必须刚好是5行；第二，你必须在代码写完后测试全部的代码，并确保测试成功通过，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见代码及文本文件的扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs',\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.scss', '.sass', '.less',\\n '.json', '.xml', '.yaml', '.yml', '.sh', '.bat', '.ps1',\\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', '.vue', '.svelte',\\n '.md', '.txt'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 格式读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明并非合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如文件权限错误等其他异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n file_todo_counts = []\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的前端依赖、后端环境与隐藏目录\\n if 'node_modules' in root or '.git' in root or 'venv' in root or '__pycache__' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 针对前端项目，推测可能的代码文件类型后缀\\n if ext in ['.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.vue', '.py']:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n # 针对不同扩展名使用对应的正则表达式提取代码注释\\n if ext in ['.html', '.htm', '.vue']:\\n # 匹配 HTML 或 Vue 模板注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # 匹配内联脚本和样式中的 JS/CSS 注释 (注意 (?<!:) 用于避免误匹配到 http:// 中的双斜杠)\\n comments.extend(re.findall(r'(?<!:)//.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/', content, re.MULTILINE))\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss']:\\n comments.extend(re.findall(r'(?<!:)//.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/', content, re.MULTILINE))\\n elif ext in ['.py']:\\n # 匹配 Python 的单行和多行注释\\n comments.extend(re.findall(r'#.*?$', content, re.MULTILINE))\\n comments.extend(re.findall(r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"', content))\\n \\n # 统计该代码文件提取到的所有注释中 \\\"TODO\\\" 的具体出现次数\\n todo_count = sum(c.count(\\\"TODO\\\") for c in comments)\\n file_todo_counts.append(todo_count)\\n \\n if not has_code_file or not file_todo_counts:\\n return False\\n \\n # 情况一：要求每个生成的代码文件自身的注释中，\\\"TODO\\\"均恰好出现 2 次\\n if all(count == 2 for count in file_todo_counts):\\n return True\\n \\n # 情况二：要求整体项目所有代码文件的注释总和中，\\\"TODO\\\"共计出现 2 次\\n if sum(file_todo_counts) == 2:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对前端及常规项目的常见代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰检查的无关目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含规定的编码声明字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果由于非utf-8编码等原因导致无法读取，说明不满足指定utf-8编码的要求\\n return False\\n\\n # 必须至少生成了一个代码文件，并且全部通过检查才算满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据前端路线图应用开发的需求，推测可能产生的代码文件类型\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n comment_found = False\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n # 在文件开头处（前 5 行内）寻找指定注释\\n for line in lines[:5]:\\n if target_comment in line:\\n comment_found = True\\n break\\n except Exception:\\n pass\\n \\n # 如果某个代码文件未在开头找到指定的注释，或者无法读取，即视为不满足要求\\n if not comment_found:\\n return False\\n \\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"请确保你输出的每一个代码文件，包含空行在内的总行数均为奇数\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义前端项目中可能的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能包含大量无关依赖包及版本控制信息的目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符切分以统计行数，此方法与主流IDE计算包含空行在内的总行数一致\\n lines = content.split('\\\\n')\\n total_lines = len(lines)\\n \\n # 检查包含空行在内的总行数是否为奇数\\n if total_lines % 2 == 0:\\n return False\\n except Exception:\\n # 如果遇到无法以UTF-8读取的文件，忽略该文件\\n continue\\n \\n # 如果项目中没有任何代码文件，意味着并没有正确生成项目，返回 False\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"限制每个生成的代码文件中的注释数量必须刚好是5行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n from html.parser import HTMLParser\\n\\n # 定义目标代码文件的后缀名\\n CODE_EXTS = {'.html', '.htm', '.vue', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp'}\\n\\n def get_line_starts(text):\\n \\\"\\\"\\\"记录每行的起始下标，以便快速将字符索引转为行号\\\"\\\"\\\"\\n line_starts = [0]\\n for m in re.finditer(r'\\\\n', text):\\n line_starts.append(m.end())\\n return line_starts\\n\\n def get_line_num(line_starts, pos):\\n \\\"\\\"\\\"利用二分查找定位 pos 所在的行号（基于1的索引）\\\"\\\"\\\"\\n return bisect.bisect_right(line_starts, pos)\\n\\n files_checked = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和第三方依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in CODE_EXTS:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n\\n comment_lines = set()\\n \\n if ext in ('.html', '.htm', '.vue'):\\n class Parser(HTMLParser):\\n def __init__(self):\\n super().__init__()\\n self.comment_lines = set()\\n self.embedded_code = []\\n self.current_tag = None\\n\\n def handle_starttag(self, tag, attrs):\\n self.current_tag = tag\\n\\n def handle_endtag(self, tag):\\n self.current_tag = None\\n\\n def handle_data(self, data):\\n if self.current_tag in ('script', 'style'):\\n # 获取 <script> / <style> 数据块开头的行号\\n start_line, _ = self.getpos()\\n self.embedded_code.append((self.current_tag, data, start_line))\\n\\n def handle_comment(self, data):\\n # 处理 HTML 原生注释\\n start_line, _ = self.getpos()\\n num_lines = data.count('\\\\n')\\n for i in range(num_lines + 1):\\n self.comment_lines.add(start_line + i)\\n\\n parser = Parser()\\n try:\\n parser.feed(code)\\n comment_lines.update(parser.comment_lines)\\n \\n # 额外处理内嵌的 JS 或 CSS 代码中的注释\\n for tag, data, start_line in parser.embedded_code:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<m_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n r'|(?P<s_comment>//.*)',\\n re.MULTILINE\\n )\\n l_starts = get_line_starts(data)\\n for m in pattern.finditer(data):\\n if m.group('m_comment'):\\n s_pos = m.start('m_comment')\\n e_pos = m.end('m_comment')\\n s_line = get_line_num(l_starts, s_pos)\\n e_line = get_line_num(l_starts, e_pos - 1)\\n for l in range(s_line, e_line + 1):\\n comment_lines.add(start_line + l - 1)\\n elif m.group('s_comment'):\\n s_pos = m.start('s_comment')\\n s_line = get_line_num(l_starts, s_pos)\\n comment_lines.add(start_line + s_line - 1)\\n except Exception:\\n # 若 HTML 解析器异常退出，降级使用基础正则以匹配 HTML 注释\\n pattern = re.compile(r'<!--[\\\\s\\\\S]*?-->', re.MULTILINE)\\n l_starts = get_line_starts(code)\\n for m in pattern.finditer(code):\\n s_line = get_line_num(l_starts, m.start())\\n e_line = get_line_num(l_starts, m.end() - 1)\\n for l in range(s_line, e_line + 1):\\n comment_lines.add(l)\\n \\n elif ext in ('.js', '.jsx', '.ts', '.tsx', '.css', '.c', '.cpp', '.h', '.hpp', '.java'):\\n # 处理 JS/TS/CSS 类文件的注释\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<m_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n r'|(?P<s_comment>//.*)',\\n re.MULTILINE\\n )\\n l_starts = get_line_starts(code)\\n for m in pattern.finditer(code):\\n if m.group('m_comment'):\\n s_pos = m.start('m_comment')\\n e_pos = m.end('m_comment')\\n s_line = get_line_num(l_starts, s_pos)\\n e_line = get_line_num(l_starts, e_pos - 1)\\n for l in range(s_line, e_line + 1):\\n comment_lines.add(l)\\n elif m.group('s_comment'):\\n s_pos = m.start('s_comment')\\n s_line = get_line_num(l_starts, s_pos)\\n comment_lines.add(s_line)\\n \\n elif ext in ('.py',):\\n # 处理 Python 文件的注释（只抓取 # 注释行，多行字符串若未被变量接收暂视为普通字符串）\\n pattern = re.compile(\\n r'(?P<string>\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<s_comment>#.*)',\\n re.MULTILINE\\n )\\n l_starts = get_line_starts(code)\\n for m in pattern.finditer(code):\\n if m.group('s_comment'):\\n s_pos = m.start('s_comment')\\n s_line = get_line_num(l_starts, s_pos)\\n comment_lines.add(s_line)\\n\\n # 验证单一文件是否恰好为 5 行注释\\n if len(comment_lines) != 5:\\n return False\\n \\n files_checked += 1\\n\\n # 需确保项目下至少检查了一个代码文件\\n return files_checked > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"路线画好后，用户需要知道具体的行程信息。请在侧边栏表单下方新增一个“行程详情”面板。每次生成路线后，在面板顶部展示模拟计算出的“总距离（如 350公里）”和“预计驾驶时长（如 4小时20分钟）”。在此下方，生成一个垂直的行程时间轴（Timeline）列表，依次展示“起点 -> 途经点1 -> ... -> 终点”。在时间轴的每一个路段节点之间，清晰地标出该特定路段的距离和预估耗时。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的输出代码文件包含空行在内总行数为奇数的规则调整为，你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数；第二，后续所有轮次请始终遵循，项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端与后端代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试用 UTF-8 编码格式读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出解码错误，说明该文件不是纯正的 UTF-8 编码\\n return False\\n except Exception:\\n # 如果有其他异常(例如权限问题等)暂忽略，但主要检查点在于 UnicodeDecodeError\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.sh', '.bash'\\n }\\n \\n code_files = []\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和第三方依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何代码文件，则不满足条件\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n ext = os.path.splitext(file_path)[1].lower()\\n patterns = []\\n \\n # 根据文件类型构造对应的注释匹配正则表达式\\n if ext in {'.html', '.vue'}:\\n patterns.append(r'<!--[\\\\s\\\\S]*?-->')\\n patterns.append(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n patterns.append(r'(?<!:)//.*')\\n elif ext in {'.css'}:\\n patterns.append(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.h', '.hpp'}:\\n patterns.append(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n patterns.append(r'(?<!:)//.*')\\n elif ext in {'.py', '.sh', '.bash'}:\\n patterns.append(r'#.*')\\n if ext == '.py':\\n patterns.append(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"')\\n patterns.append(r\\\"'''[\\\\s\\\\S]*?'''\\\")\\n \\n if not patterns:\\n continue\\n \\n # 使用逻辑“或”组合正则以同时匹配所有有效注释类型\\n combined_pattern = '|'.join(f'({p})' for p in patterns)\\n regex = re.compile(combined_pattern)\\n \\n todo_count = 0\\n for match in regex.finditer(content):\\n comment = match.group(0)\\n # 统计注释内部的 \\\"TODO\\\" 关键词出现次数\\n todo_count += comment.count('TODO')\\n \\n # 若任意一份代码文件中的 TODO 数量不为 2 次，即不满足要求\\n if todo_count != 2:\\n return False\\n \\n except Exception:\\n # 根据要求，所有代码文件都必须为 UTF-8 编码，读文件失败视为错误\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见前端代码文件扩展名\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏文件夹及常见的第三方依赖目录，避免误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码格式声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取失败，则视为不符合要求\\n return False\\n \\n # 如果项目中没有任何我们所关注的代码文件，视为未完成需求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n file_count = 0\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏文件夹和无关的依赖、编译目录，避免误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'dist', 'build')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 检查开头部分（前10行），以容忍首行的文件编码说明（如 -*- coding: utf-8 -*-）及 HTML 的 doctype 等\\n lines = [f.readline() for _ in range(10)]\\n except Exception:\\n # 若无法使用UTF-8正常读取，视同未按要求处理\\n return False\\n \\n found = False\\n for line in lines:\\n if not line:\\n continue\\n # 检查目标注释文字是否包含在当前行中\\n if target_text in line:\\n found = True\\n break\\n \\n # 只要存在任何一个代码文件不包含目标注释，即不满足要求\\n if not found:\\n return False\\n \\n file_count += 1\\n \\n # 至少要验证通过一个代码文件才算满足要求，避免空目录蒙混过关\\n if file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断的前端代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉不属于生成的第三方依赖及系统隐藏文件夹\\n for exclude_dir in ['node_modules', '.git', '.idea', '.vscode']:\\n if exclude_dir in dirs:\\n dirs.remove(exclude_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 验证总行数（含空行）是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 如果有异常如编码不是UTF-8，尝试使用latin-1安全读取行数\\n try:\\n with open(file_path, 'r', encoding='latin-1') as f:\\n lines = f.readlines()\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 极端异常无法读取情况可视为无法判定，直接忽略处理继续向下\\n pass\\n \\n # 只要存在代码文件，并且没有触发奇数行的return False，就代表通过验证\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"限制每个生成的代码文件中的注释数量必须刚好是5行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less',\\n '.html', '.htm', '.vue', '.svg', '.py', '.java', '.c', '.cpp',\\n '.cs', '.go', '.php', '.rb', '.sh'\\n }\\n \\n files_checked = 0\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏文件夹、前端依赖项和 Python 缓存\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n files_checked += 1\\n \\n # 记录所有换行符的位置以用来二分查找对应的行号\\n line_starts = [0]\\n for m in re.finditer(r'\\\\n', content):\\n line_starts.append(m.end())\\n \\n def get_line_num(pos):\\n return bisect.bisect_right(line_starts, pos)\\n \\n comment_lines = set()\\n \\n # C风格及前端家族文件（过滤引号内字符串和css中的url影响，分别捕获HTML、多行、单行注释）\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css', '.scss', '.less', '.html', '.htm', '.vue', '.svg', '.java', '.c', '.cpp', '.cs', '.go', '.php']:\\n pattern = re.compile(\\n r'(url\\\\([^)]*\\\\))' \\n r'|(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' \\n r'|(<!--[\\\\s\\\\S]*?-->)' \\n r'|(/\\\\*[\\\\s\\\\S]*?\\\\*/)' \\n r'|((?<!:)//.*)' \\n )\\n for m in pattern.finditer(content):\\n if m.group(1) or m.group(2):\\n continue\\n elif m.group(3):\\n start_line = get_line_num(m.start(3))\\n end_line = get_line_num(m.end(3) - 1)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n elif m.group(4):\\n start_line = get_line_num(m.start(4))\\n end_line = get_line_num(m.end(4) - 1)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n elif m.group(5):\\n start_line = get_line_num(m.start(5))\\n comment_lines.add(start_line)\\n\\n # Python/Shell 家族文件（以 # 为核心进行处理，同样需过滤字符串影响）\\n elif ext in ['.py', '.rb', '.sh', '.yml', '.yaml']:\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\')' \\n r'|(#.*)' \\n )\\n for m in pattern.finditer(content):\\n if m.group(1):\\n continue\\n elif m.group(2):\\n start_line = get_line_num(m.start(2))\\n comment_lines.add(start_line)\\n else:\\n continue\\n \\n # 针对该特定文件，注释占据的总行数若不严格等于 5 行，则视为没有满足条件\\n if len(comment_lines) != 5:\\n return False\\n \\n # 如果没有找到任何可检查的代码文件，则不能算作满足要求\\n return files_checked > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"太棒了！现在我们来引入这个应用的核心特色——天气数据。我们先使用模拟的天气数据。请修改“行程详情”面板的时间轴，在每一个站点（起点、所有途经点、终点）的右侧，加上天气信息的展示，包括一个直观的天气图标（晴、雨、多云等）、当前温度和降水概率。同时，联动一下地图：当用户在地图上点击这几个站点的标记时，弹出的 Popup 除了显示地名，还要同步展示刚才这套模拟的天气数据。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的你必须在代码写完后测试全部的代码并确保测试成功通过的规则调整为，当你把代码写完后，不得进行任何测试；第二，后续所有轮次生成的代码文件里，缩进必须严格保持为2个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，说明没有生成文件，则不算违反规则\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 常见的前端项目代码文件及配置文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue', '.py'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试使用 utf-8 编码去读取整个文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果遇到 Unicode 解码错误，说明非 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如文件权限等其他不可预期的异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_todo_count(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件后缀\\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n all_files_todo_counts = []\\n total_todo_count = 0\\n \\n # 遍历当前项目工作空间的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 根据不同扩展名赋予对应的正则表达式\\n # 多行注释优先提取，单行注释后提取，避免出现重复计算的情况\\n if ext in {'.html', '.vue'}:\\n patterns = [r'<!--[\\\\s\\\\S]*?-->', r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'//.*']\\n elif ext in {'.js', '.ts', '.jsx', '.tsx'}:\\n patterns = [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'//.*']\\n elif ext == '.css':\\n patterns = [r'/\\\\*[\\\\s\\\\S]*?\\\\*/']\\n elif ext == '.py':\\n patterns = [r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"', r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'', r'#.*']\\n else:\\n patterns = [r'<!--[\\\\s\\\\S]*?-->', r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'//.*', r'#.*']\\n \\n all_matches = []\\n for p in patterns:\\n matches = re.findall(p, content)\\n # 将提取到的注释部分替换为空格，以免后续匹配互相干扰\\n content = re.sub(p, ' ', content)\\n all_matches.extend(matches)\\n \\n # 统计提取出的注释中 TODO 出现的次数\\n todo_count = sum(comment.count(\\\"TODO\\\") for comment in all_matches)\\n all_files_todo_counts.append(todo_count)\\n total_todo_count += todo_count\\n \\n # 如果项目中没有扫描到任何代码文件，则不满足要求\\n if not all_files_todo_counts:\\n return False\\n \\n # 第一种理解：所有扫描的代码文件，每一个文件中的 TODO 数量均为 2 次\\n if all(c == 2 for c in all_files_todo_counts):\\n return True\\n \\n # 第二种理解：整个项目中所有代码文件的 TODO 数量总和为 2 次\\n if total_todo_count == 2:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端与相关代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.vue'}\\n # 定义需要忽略的目录，避免误检查非生成的代码文件\\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'dist', 'build', '.idea', '.vscode'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查代码文件首行是否包含指定的编码格式声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果在打开或读取过程中遇到异常（如非文本文件或编码错误），认定为不符合要求\\n return False\\n \\n # 如果所有代码文件的首行都包含指定内容，或没有扫描到需要处理的代码文件，则返回True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 前端及通用代码文件的扩展名集合\\n code_exts = {'.html', '.js', '.css', '.ts', '.vue', '.jsx', '.tsx', '.py'}\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 检查前5行是否包含目标字符串（考虑到第一行可能有编码声明等）\\n found_in_head = False\\n for line in lines[:5]:\\n if target_str in line:\\n found_in_head = True\\n break\\n \\n if not found_in_head:\\n return False\\n except Exception:\\n # 如果无法按utf-8读取，或出现其他异常，视为不符合要求\\n return False\\n \\n # 如果找到了代码文件，并且全部通过了检查，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对前端项目的常见代码文件扩展名\\n code_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_checked = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉无关目录，如 node_modules 及隐藏文件夹（保留 __tests__ 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按换行符分割，包括末尾的空行，能够真实反映文件包含空行的总行数\\n lines = content.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n \\n has_checked = True\\n except Exception:\\n # 如果由于非utf-8编码等原因无法读取，认为无法满足题目规定的有效文件读取，返回 False\\n return False\\n \\n return has_checked\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"限制每个生成的代码文件中的注释数量必须刚好是5行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n CODE_EXTS = {'.js', '.css', '.html', '.htm', '.vue', '.py', '.ts', '.jsx', '.tsx'}\\n exclude_dirs = {'node_modules', '.git', '__pycache__', 'venv', 'env', '.vscode', '.idea'}\\n \\n code_files = []\\n # 获取所有的代码文件，排除常见依赖和缓存目录\\n for root, dirs, files in os.walk(workspace_path):\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for f in files:\\n ext = os.path.splitext(f)[1].lower()\\n if ext in CODE_EXTS:\\n code_files.append(os.path.join(root, f))\\n \\n if not code_files:\\n return False\\n \\n for cf in code_files:\\n ext = os.path.splitext(cf)[1].lower()\\n try:\\n with open(cf, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return False\\n \\n # 根据不同语言分别定义字符串和注释正则匹配模式\\n if ext == '.py':\\n # Python中处理f-string/r-string/b-string及常规三引号、单双引号内的字符串\\n string_pattern = r'[frbFRB]?\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|[frbFRB]?\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|[frbFRB]?\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\"|[frbFRB]?\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\\\\\n])*\\\\''\\n comment_pattern = r'(#.*)'\\n else:\\n # 前端代码(JS/CSS/HTML)中的单、双、反引号内的字符串\\n string_pattern = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n # 排除了 http/https/ws/wss 等常见非引用URL引起的假注释误判\\n sl_comment = r'(?<!https:)(?<!http:)(?<!ws:)(?<!wss:)//.*'\\n ml_comment = r'/\\\\*[\\\\s\\\\S]*?\\\\*/'\\n html_comment = r'<!--[\\\\s\\\\S]*?-->'\\n comment_pattern = f'({sl_comment}|{ml_comment}|{html_comment})'\\n \\n # 整体正则模式：将字符串优先匹配，以此保护不被当作注释\\n regex = f'({string_pattern})|({comment_pattern})'\\n matches = re.finditer(regex, content)\\n \\n # 通过预计算所有的换行符位置来帮助后续快速得出字符索引所在的行号\\n line_starts = [0]\\n for m in re.finditer(r'\\\\n', content):\\n line_starts.append(m.start() + 1)\\n \\n def get_line_num(idx):\\n return bisect.bisect_right(line_starts, idx)\\n \\n comment_lines = set()\\n \\n for match in matches:\\n if match.group(2): # 如果第2个捕获组(即注释)命中了\\n start_idx = match.start(2)\\n end_idx = match.end(2)\\n \\n start_line = get_line_num(start_idx)\\n # end_idx 是开区间，计算结束行时需减一\\n if end_idx > start_idx:\\n end_line = get_line_num(end_idx - 1)\\n else:\\n end_line = start_line\\n \\n # 将此条注释所跨越的所有行号加入去重集合\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n \\n # 验证该文件的注释总行数是否为严格等于 5\\n if len(comment_lines) != 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"当你把代码写完后，不得进行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件里，缩进必须严格保持为2个空格\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 推断可能的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.py', '.java', '.c', '.cpp', '.cs', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n has_code = False\\n \\n # 定义验证缩进的辅助函数\\n def is_indentation_valid(lines) -> bool:\\n for line in lines:\\n if line.strip() == '':\\n continue\\n # 匹配每一行行首的空白字符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent = match.group(1)\\n # 若包含制表符，直接判为不符合\\n if '\\\\t' in indent:\\n return False\\n # 若仅由空格组成，空格数量应为 2 的倍数\\n if len(indent) % 2 != 0:\\n # 忽略以 * 开头的行（用于豁免常规的多行注释中间行格式，如 /* \\\\n * \\\\n */ ）\\n if not line.strip().startswith('*'):\\n return False\\n return True\\n\\n # 1. 检查工作空间路径中生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n if not is_indentation_valid(lines):\\n return False\\n \\n # 2. 检查助手最后的一轮 response 中的 Markdown 代码块\\n code_blocks = re.findall(r'```(?:[a-zA-Z]*)\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n has_code = True\\n lines = block.split('\\\\n')\\n if not is_indentation_valid(lines):\\n return False\\n \\n return has_code\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"接下来这个挑战稍微有点难度：结合时间的动态天气预测。请在侧边栏顶部新增一个“出发时间”的日期时间选择器（默认当前时间）。当点击规划路线时，请根据每段路程的耗时，动态计算出到达每一个站点的“预计到达时间（ETA）”，并将该时间显示在时间轴和地图的 Popup 中；同时天气数据也要根据ETA进行变化（模拟不同时间段天气不同）。此外，在地图区域的底部叠加一个“行程时间轴滑块（Slider）”，用户拖动滑块可以模拟旅程的推进，地图上当前时间所处的路段折线需要用高亮颜色特殊标注出来。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的生成的代码文件缩进严格保持为2个空格的规则调整为，在生成的代码文件里，缩进必须严格保持为4个空格，该要求后续所有轮次始终遵循；第二，后续所有轮次请始终遵循，确保在你生成的每一个代码文件里，其注释部分包含关键词 \\\"FIXME\\\" 的次数正好是1次。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径为空或不存在，直接返回False\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令推断，这是一个前端旅行地图项目，可能包含以下类型代码文件\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json', '.py'}\\n has_code_file = False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明该文件不是使用UTF-8编码，验证失败\\n return False\\n except Exception:\\n # 忽略其他如权限或文件系统引起的非编码相关异常\\n pass\\n \\n # 如果至少存在一个代码文件，且均未抛出解码错误，则视为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 根据项目推断可能的代码文件类型\\n valid_exts = {'.html', '.htm', '.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.py', '.mjs', '.cjs'}\\n files_checked = 0\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的非业务代码目录\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n todo_count = 0\\n \\n # 1. 处理 HTML 注释\\n if ext in ['.html', '.htm']:\\n html_comments = re.findall(r'<!--[\\\\s\\\\S]*?-->', code)\\n for c in html_comments:\\n todo_count += c.count(\\\"TODO\\\")\\n # 匹配后移除，防止后续重复计算\\n code = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', code)\\n \\n # 2. 处理 多行注释 (JS/TS/CSS 等，以及 HTML 中可能嵌入的样式和脚本的多行注释)\\n if ext in ['.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.html', '.htm', '.mjs', '.cjs']:\\n ml_comments = re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', code)\\n for c in ml_comments:\\n todo_count += c.count(\\\"TODO\\\")\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n \\n # 处理 Python 多行注释 (Docstrings)\\n if ext in ['.py']:\\n ml_comments = re.findall(r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"', code)\\n for c in ml_comments:\\n todo_count += c.count(\\\"TODO\\\")\\n code = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"', '', code)\\n \\n ml_comments = re.findall(r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'', code)\\n for c in ml_comments:\\n todo_count += c.count(\\\"TODO\\\")\\n code = re.sub(r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'', '', code)\\n\\n # 3. 处理 单行注释\\n if ext in ['.js', '.ts', '.css', '.vue', '.jsx', '.tsx', '.html', '.htm', '.mjs', '.cjs']:\\n # 使用 (?<!:)//.* 避免误匹配 URL 中的双斜杠 (如 http://)\\n sl_comments = re.findall(r'(?<!:)//.*', code)\\n for c in sl_comments:\\n todo_count += c.count(\\\"TODO\\\")\\n \\n if ext in ['.py']:\\n sl_comments = re.findall(r'#.*', code)\\n for c in sl_comments:\\n todo_count += c.count(\\\"TODO\\\")\\n \\n # 严格控制在每一个生成的代码文件中，\\\"TODO\\\"必须出现且仅出现2次\\n if todo_count != 2:\\n return False\\n files_checked += 1\\n \\n # 至少要验证并检查了一个代码文件\\n return files_checked > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断本项目可能生成的主要代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和不相关的依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的首行\\n first_line = f.readline()\\n \\n # 检查特定字符串是否在首行中\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件不可读，稳妥起见视为不满足\\n return False\\n \\n # 确保项目中至少生成了代码文件才算满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 给定要求中的特定注释目标字符串\\n target_string = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据用户前端地图项目的指令内容，推断可能的代码文件类型\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n found_any_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方依赖或版本控制目录以加快遍历并防止误判\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果遇到无法读取的文件，则跳过\\n continue\\n \\n # 忽略空文件\\n if not lines:\\n continue\\n \\n found_any_code_file = True\\n \\n # 检查开头部分（前5行）是否存在该独立换行的目标注释\\n is_valid = False\\n for line in lines[:5]:\\n if target_string in line:\\n is_valid = True\\n break\\n \\n # 如果任意一个代码文件的开头前5行内没有找到目标注释，则要求未满足\\n if not is_valid:\\n return False\\n \\n # 如果工作区内存在至少一个代码文件且全部代码文件均满足要求，则返回 True\\n return found_any_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def verify_even_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据项目推测可能生成的前端及相关代码文件类型\\n code_extensions = {\\n '.html', '.css', '.js', '.json', '.ts', \\n '.jsx', '.tsx', '.vue', '.py'\\n }\\n\\n has_code_file = False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # readlines() 会保留空行并作为单独的元素，长度即为总行数\\n lines = f.readlines()\\n line_count = len(lines)\\n \\n # 如果某个代码文件行数不是偶数，则判定不满足要求\\n if line_count % 2 != 0:\\n return False\\n except Exception:\\n # 若文件编码错误或无法读取，为了容错可跳过\\n continue\\n\\n # 如果至少检查了一个代码文件且都为偶数行，满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"限制每个生成的代码文件中的注释数量必须刚好是5行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n # 定义常见的代码文件扩展名\\n code_extensions = {'.py', '.js', '.ts', '.css', '.html', '.htm', '.jsx', '.tsx', '.vue'}\\n \\n # 匹配 JS/HTML/CSS 的字符串和注释\\n # (?P<name>...) 为命名捕获组，利用同一层级的正则匹配可防止字符串内的注释符号被误判\\n js_html_pattern = re.compile(\\n r'(?P<html_comment><!--[\\\\s\\\\S]*?-->)|'\\n r'(?P<block_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)|'\\n r'(?P<line_comment>//[^\\\\n]*)|'\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n )\\n\\n # 匹配 Python 的字符串、文档注释（多行字符串）和单行注释\\n # 注意：在 Python 中，通常只有 # 被严格算作注释，文档字符串属于字符串范畴，因此分开捕获，后续只统计 py_comment\\n py_pattern = re.compile(\\n r'(?P<py_tq>\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\')|'\\n r'(?P<py_comment>#[^\\\\n]*)|'\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n )\\n\\n file_checked = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 无法读取的文件跳过\\n continue\\n \\n file_checked = True\\n \\n # 记录每一行的起始字符索引\\n line_starts = [0]\\n for i, char in enumerate(content):\\n if char == '\\\\n':\\n line_starts.append(i + 1)\\n \\n # 通过字符索引二分查找对应的行号（从0开始计数）\\n def get_line_num(idx):\\n return bisect.bisect_right(line_starts, idx) - 1\\n \\n # 用于标记某一行是否包含注释\\n is_comment = [False] * len(line_starts)\\n \\n if ext == '.py':\\n pattern = py_pattern\\n else:\\n pattern = js_html_pattern\\n \\n # 遍历所有匹配项\\n for match in pattern.finditer(content):\\n group_name = match.lastgroup\\n # 如果匹配到的组名是注释\\n if group_name in ('html_comment', 'block_comment', 'line_comment', 'py_comment'):\\n start_line = get_line_num(match.start())\\n # match.end() 是匹配结束后的下一个字符索引，减1代表该匹配结尾的实际字符位置\\n end_line = get_line_num(match.end() - 1) if match.end() > match.start() else start_line\\n \\n # 将这段注释所跨越的所有行标记为 True\\n for l in range(start_line, end_line + 1):\\n is_comment[l] = True\\n \\n total_comments = sum(is_comment)\\n \\n # 只要有任何一个代码文件的注释总行数不等于 5，即不符合要求\\n if total_comments != 5:\\n return False\\n \\n # 确保至少检查了一个代码文件，并且全部文件均符合要求\\n return file_checked\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"当你把代码写完后，不得进行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件里，缩进必须严格保持为4个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目中常见的代码文件扩展名\\n valid_exts = {'.js', '.css', '.html', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略依赖库、版本控制目录及测试相关目录\\n for ignore_dir in ['node_modules', '.git', '__tests__']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果读取文件失败（如非文本文件等情况），跳过\\n continue\\n\\n for line in lines:\\n # 去掉行尾的换行符\\n line = line.rstrip('\\\\r\\\\n')\\n \\n # 仅去掉行首的空格和Tab，以便精确获取缩进字符\\n stripped_line = line.lstrip(' \\\\t')\\n \\n # 忽略纯空行\\n if not stripped_line:\\n continue\\n \\n # 截取该行的前导空白字符（即缩进部分）\\n leading_whitespaces = line[:len(line) - len(stripped_line)]\\n \\n # 规则1：缩进中绝对不能包含制表符(\\\\t)\\n if '\\\\t' in leading_whitespaces:\\n return False\\n \\n # 规则2：缩进空格数必须严格是4的倍数\\n if len(leading_whitespaces) % 4 != 0:\\n # 豁免规则：多行注释的中间续行（通常以 * 开头，用于对齐）\\n if stripped_line.startswith('*'):\\n continue\\n # 若不是多行注释且不满足4的倍数，则直接判定为不满足要求\\n return False\\n\\n # 必须至少包含一个代码文件，且全部验证通过才算满足要求\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的每一个代码文件里，其注释部分包含关键词 \\\"FIXME\\\" 的次数正好是1次\",\n \"validation_code\": \"def check_fixme_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见的代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', \\n '.vue', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rb', '.php', '.sh'\\n }\\n # 需排除的常见非开发目录\\n exclude_dirs = {\\n 'node_modules', 'venv', '.git', '__pycache__', 'dist', 'build', '.idea', '.vscode'\\n }\\n \\n # 构造正则表达式同时匹配“注释”和“字符串”，使得可以忽略字符串内的干扰项\\n # 注意: Python的多行字符串(\\\"\\\"\\\" 和 ''')通常用作Docstring，被优先归类到comment组\\n pattern = re.compile(\\n r'(?P<comment>'\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/' # 多行注释 (C/C++/Java/JS/TS/CSS等)\\n r'|//[^\\\\n]*' # 单行注释 (C/C++/Java/JS/TS等)\\n r'|<!--[\\\\s\\\\S]*?-->' # HTML/XML 注释\\n r'|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"' # Python 多行注释 (双引号)\\n r'|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'' # Python 多行注释 (单引号)\\n r'|#.*' # 单行注释 (Python/Shell/Ruby等)\\n r')'\\n r'|(?P<string>'\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"' # 双引号字符串\\n r'|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'' # 单引号字符串\\n r'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`' # 反引号字符串 (JS/TS等)\\n r')'\\n )\\n \\n files_checked = 0\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 errors='ignore' 保证即使存在不完全兼容 utf-8 字符的生成文件也能顺利读取\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 遍历所有匹配项，仅保留属于 comment 分组的匹配结果\\n comments = []\\n for match in pattern.finditer(content):\\n if match.group('comment'):\\n comments.append(match.group('comment'))\\n \\n comments_text = '\\\\n'.join(comments)\\n \\n # 统计注释中 \\\"FIXME\\\" 出现的次数\\n fixme_count = comments_text.count(\\\"FIXME\\\")\\n \\n # 若次数不恰好为 1，则不符合要求\\n if fixme_count != 1:\\n return False\\n \\n files_checked += 1\\n \\n # 只要验证了至少一个文件且全部通过，即视为满足要求\\n return files_checked > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"功能越来越完善了。现在来丰富地图的图层和兴趣点（POI）控制。在地图的右上角添加一个半透明的地图控制面板，包含两个区域：“地图样式”和“沿途设施”。“地图样式”提供下拉菜单，允许在“标准视图”、“暗黑模式”和“地形图”之间切换底层瓦片（可以使用不同的免费瓦片URL模拟）。“沿途设施”提供三个复选框：“加油站”、“餐厅”、“酒店”。当勾选某项时，请在当前绘制的路线两侧一定范围内，随机散布几个对应类型的定制小图标；取消勾选则让它们从地图上消失。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的当你把代码写完后不得进行任何测试的规则调整为，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过，该要求后续所有轮次始终遵循；第二，后续所有轮次请始终遵循，在你输出的任何一个代码文件里，务必出现\\\"OPTIMIZE\\\"。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端及常见 Web 项目的代码文件扩展名\\n code_extensions = {\\n '.html', '.js', '.css', '.ts', '.jsx', '.tsx', \\n '.json', '.vue', '.svelte', '.py', '.txt', '.md', '.sh'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的无关目录或依赖库目录\\n if '.git' in root or 'node_modules' in root or '__pycache__' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件，过滤掉可能存在的图片等二进制资源文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试使用 UTF-8 进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明该文件包含非 UTF-8 编码的内容\\n return False\\n except Exception:\\n # 遇到其他读取异常也可视为验证失败\\n return False\\n\\n # 必须至少存在一个代码文件，且全部验证通过才算满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"def check_todo_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n ignore_dirs = {'node_modules', '.git', '__pycache__', 'dist', 'build', '__tests__'}\\n \\n file_todo_counts = []\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除需要忽略的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 避免 http:// 和 https:// 中的双斜杠被误判为单行注释\\n content_clean = content.replace(\\\"http://\\\", \\\"\\\").replace(\\\"https://\\\", \\\"\\\")\\n \\n comments = []\\n \\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.css']:\\n block_comments = re.findall(r'/\\\\*.*?\\\\*/', content_clean, re.DOTALL)\\n line_comments = re.findall(r'//.*$', content_clean, re.MULTILINE)\\n comments.extend(block_comments)\\n comments.extend(line_comments)\\n elif ext in ['.html', '.vue']:\\n html_comments = re.findall(r'<!--.*?-->', content_clean, re.DOTALL)\\n block_comments = re.findall(r'/\\\\*.*?\\\\*/', content_clean, re.DOTALL)\\n line_comments = re.findall(r'//.*$', content_clean, re.MULTILINE)\\n comments.extend(html_comments)\\n comments.extend(block_comments)\\n comments.extend(line_comments)\\n elif ext in ['.py']:\\n block_comments = re.findall(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', content_clean, re.DOTALL)\\n block_comments_2 = re.findall(r\\\"'''(.*?)'''\\\", content_clean, re.DOTALL)\\n line_comments = re.findall(r'#.*$', content_clean, re.MULTILINE)\\n comments.extend(block_comments)\\n comments.extend(block_comments_2)\\n comments.extend(line_comments)\\n \\n comment_text = \\\"\\\\n\\\".join(comments)\\n todo_count = comment_text.count(\\\"TODO\\\")\\n file_todo_counts.append(todo_count)\\n except Exception:\\n continue\\n \\n # 如果没有扫描到任何有效的代码文件，则不满足要求\\n if not file_todo_counts:\\n return False\\n \\n # 理解1: 每个代码文件的注释中均包含且仅包含 2 次 TODO\\n every_file_has_two = all(count == 2 for count in file_todo_counts)\\n # 理解2: 整个项目所有代码文件的注释中合计包含 2 次 TODO\\n sum_is_two = sum(file_todo_counts) == 2\\n \\n return every_file_has_two or sum_is_two\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义推断可能的代码文件扩展名\\n valid_extensions = {'.js', '.html', '.css', '.py', '.ts', '.jsx', '.tsx', '.vue'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的无关目录，避免扫描非AI生成的依赖文件\\n for exclude_dir in ['node_modules', '.git', '__pycache__']:\\n if exclude_dir in dirs:\\n dirs.remove(exclude_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含指定的编码格式字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果遇到文件无法读取等异常情况，视作不满足要求\\n return False\\n \\n # 如果扫描过至少一个代码文件且都满足要求，返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 推断本项目为前端地图应用，可能包含以下类型的代码文件\\n # 不包括不支持标准注释的 .json 文件\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n expected_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取开头部分（前 15 行足够覆盖\\\"开头处\\\"的要求）\\n lines = [f.readline() for _ in range(15)]\\n except Exception:\\n continue\\n \\n found = False\\n for line in lines:\\n if not line:\\n break\\n if expected_text in line:\\n # 检查是否为单独换行的注释（前面不能有正常代码逻辑，只能是空白符或注释符号）\\n prefix = line[:line.find(expected_text)].strip()\\n # 常见的注释符号字符集合\\n if not prefix or all(c in ' \\\\t#/!*-<>' for c in prefix):\\n found = True\\n break\\n \\n # 如果某个代码文件没有在开头处包含单独换行的特定注释，则验证失败\\n if not found:\\n return False\\n \\n # 如果没有生成任何代码文件，则也认为未能实现要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def check_even_lines_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断，前端可能用到的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n has_code_file = False\\n\\n # 遍历当前项目工作区的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 筛选出属于代码类型的文件\\n if any(file_name.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 判断总行数（包含空行）是否为偶数\\n if len(lines) % 2 != 0:\\n return False\\n except Exception:\\n # 对于不可读或编码异常的文件，选择忽略或抛出，这里选择跳过继续检查\\n pass\\n \\n # 如果找到了代码文件且无一奇数行，则返回True；若没找到任何代码文件返回False\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"限制每个生成的代码文件中的注释数量必须刚好是5行\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义通常被认为是代码文件的扩展名集合\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', \\n '.html', '.css', '.vue', '.json',\\n '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n \\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非项目主体逻辑的干扰目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if '__pycache__' in dirs:\\n dirs.remove('__pycache__')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n count = 0\\n in_block = False\\n in_html_block = False\\n \\n for line in lines:\\n s = line.strip()\\n if not s:\\n continue\\n \\n # 针对 Python / Ruby / Shell 等脚本语言使用 # 作为注释符\\n if ext in {'.py', '.sh', '.yml', '.yaml', '.rb'}:\\n if s.startswith('#'):\\n count += 1\\n \\n # 针对 JS / TS / CSS / C系语言等使用 // 以及 /* */ 作为注释符\\n elif ext in {'.js', '.jsx', '.ts', '.tsx', '.css', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.swift', '.kt', '.php', '.json'}:\\n if in_block:\\n count += 1\\n if '*/' in s:\\n in_block = False\\n else:\\n if s.startswith('/*'):\\n count += 1\\n if '*/' not in s:\\n in_block = True\\n elif s.startswith('//'):\\n count += 1\\n \\n # 针对 HTML / Vue 等标记类文件，兼容 HTML注释 与 其内嵌的 JS/CSS注释\\n elif ext in {'.html', '.vue', '.xml', '.svg'}:\\n if in_html_block:\\n count += 1\\n if '-->' in s:\\n in_html_block = False\\n elif in_block:\\n count += 1\\n if '*/' in s:\\n in_block = False\\n else:\\n if s.startswith('<!--'):\\n count += 1\\n if '-->' not in s:\\n in_html_block = True\\n elif s.startswith('/*'):\\n count += 1\\n if '*/' not in s:\\n in_block = True\\n elif s.startswith('//'):\\n count += 1\\n \\n files_checked += 1\\n # 任何一个代码文件的注释总行数如果不为 5，判定为不满足条件\\n if count != 5:\\n return False\\n \\n # 确保项目中至少成功检查了一个代码文件\\n return files_checked > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件里，缩进必须严格保持为4个空格\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件后缀\\n valid_extensions = {'.js', '.css', '.html', '.py', '.ts', '.jsx', '.tsx', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略依赖库、虚拟环境、Git版本控制等无关目录\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'venv', 'env'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果有文件读取异常可以跳过，继续检查其他代码文件\\n continue\\n \\n for line in lines:\\n r_stripped = line.rstrip()\\n # 跳过仅包含空白字符的空行\\n if not r_stripped:\\n continue\\n \\n stripped = line.lstrip()\\n # 提取行首所有的空白缩进字符\\n leading_spaces = line[:len(line) - len(stripped)]\\n \\n # 规则1：严禁使用 Tab (\\\\t) 进行缩进\\n if '\\\\t' in leading_spaces:\\n return False\\n \\n # 忽略以 '*' 开头的行（常见的块状注释内部，经常会有 1 个空格或者不对齐的特殊情况）\\n if stripped.startswith('*'):\\n continue\\n \\n # 规则2：首部缩进的空格数必须严格为 4 的整数倍\\n if len(leading_spaces) % 4 != 0:\\n return False\\n \\n # 只要存在至少一个被检查的代码文件且都通过了验证，则符合要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的每一个代码文件里，其注释部分包含关键词 \\\"FIXME\\\" 的次数正好是1次\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义项目中可能涉及的代码文件扩展名（推断项目主要涉及前端HTML/JS/CSS，也可能带有少许配置文件或后端Python）\\n extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_files = False\\n \\n # 预编译正则表达式提取各语言注释块：\\n # 1. /\\\\*.*?\\\\*/ -> 匹配 JS/CSS 的多行注释\\n # 2. <!--.*?--> -> 匹配 HTML/Vue 的多行/单行注释\\n # 3. (?<!http:)(?<!https:)//[^\\\\n]* -> 匹配 JS 单行注释（巧妙排除了 http:// 和 https:// 这种网址中的双斜杠）\\n # 4. #[^\\\\n]* -> 匹配 Python 单行注释\\n comment_pattern = re.compile(\\n r'/\\\\*.*?\\\\*/|<!--.*?-->|(?<!http:)(?<!https:)//[^\\\\n]*|#[^\\\\n]*',\\n re.DOTALL\\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的依赖项与版本控制目录，保留存放代码和测试用例的目录（如 __tests__）\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__', 'dist', 'build')]\\n \\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in extensions:\\n has_files = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取文件中的所有注释部分内容\\n comments = comment_pattern.findall(content)\\n \\n # 在所有注释内容中，统计关键词 \\\"FIXME\\\" 出现的总次数\\n fixme_count = sum(comment.count(\\\"FIXME\\\") for comment in comments)\\n \\n # 要求每个文件的注释中 FIXME 次数必须正好是1次\\n if fixme_count != 1:\\n return False\\n \\n # 最终只要确保扫描过至少一个代码文件即可算作验证通过\\n return has_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现\\\"OPTIMIZE\\\"\",\n \"validation_code\": \"def verify_optimization_keyword(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果该文件中没有出现 \\\"OPTIMIZE\\\"，则直接判定为不满足要求\\n if \\\"OPTIMIZE\\\" not in content:\\n return False\\n except Exception:\\n # 如果文件读取发生异常（如编码错误），保守起见视为未满足要求\\n return False\\n\\n # 确保项目中确实存在代码文件，并且所有的代码文件都通过了检查\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"用户辛苦规划好的路线需要保存。请在侧边栏最底部添加一个“保存行程”按钮。点击后弹出一个对话框，让用户输入“行程名称”。确认后，请将当前的完整数据（行程名、出发时间、起终点名称、途经点列表）序列化并保存到浏览器的 localStorage 中。接着，利用第一轮我们在顶部导航栏留下的空间，新增一个“我的行程”入口。点击它会打开一个右侧滑出的抽屉（Drawer）或模态框，里面用列表展示所有已保存的历史行程。点击列表中的任何一个行程，都要能把它重新读取并渲染到侧边栏表单和地图上。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的每个生成的代码文件中的注释数量必须刚好是5行的规则调整为，确保你生成的每一个代码文件都精确包含 6 行注释；第二，后续所有轮次请始终遵循，在你的回复里，如需引用文件或代码的具体位置，请务必采用markdown链接语法，格式要求如下：[filename.ts](src/filename.ts)、[filename.ts:42](src/filename.ts#L42)、[filename.ts:42-51](src/filename.ts#L42-L51)。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令推断可能的代码文件类型（前端Web应用开发）\\n code_extensions = ('.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue')\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 仅针对代码文件进行编码校验\\n if file_name.lower().endswith(code_extensions):\\n found_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n # 尝试用 UTF-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦捕获到解码异常，说明不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能导致无法读取的异常（例如权限问题）\\n pass\\n\\n # 只有当发现至少一个代码文件并且全都没有引发 UnicodeDecodeError 时，才判定为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"def check_todo_count_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义前端及常见项目的代码文件扩展名\\n code_exts = {\\n '.js', '.jsx', '.ts', '.tsx', '.css', '.html', \\n '.py', '.java', '.cpp', '.c', '.h', '.hpp', \\n '.cs', '.php', '.rb', '.go', '.rs', '.swift', \\n '.kt', '.vue', '.svelte'\\n }\\n \\n # 构造正则表达式以匹配主流语言的注释形式\\n # 1. // ... 单行注释 (JS, TS, C++ 等)\\n # 2. # ... 单行注释 (Python, Ruby 等)\\n # 3. /* ... */ 多行注释 (JS, TS, CSS 等)\\n # 4. <!-- ... --> 多行注释 (HTML)\\n comment_pattern = re.compile(\\n r'//.*?$|#.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->',\\n re.MULTILINE\\n )\\n \\n files_checked = 0\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 找出文件中的所有注释文本块\\n comments = comment_pattern.findall(content)\\n \\n # 统计这些注释块中出现 \\\"TODO\\\" 的总次数\\n todo_count = sum(comment.count(\\\"TODO\\\") for comment in comments)\\n \\n # 根据要求，每个生成的代码文件的注释中，\\\"TODO\\\" 必须出现且仅出现 2 次\\n if todo_count != 2:\\n return False\\n \\n files_checked += 1\\n except Exception:\\n # 如果遇到文件编码等读取异常，可做静默跳过处理，不影响常规代码逻辑的判断\\n pass\\n \\n # 如果项目路径中至少找到了一个代码文件并且全满足上述条件，则验证通过\\n return files_checked > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义该项目中可能出现的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果由于非utf-8编码等原因导致读取失败，则视为不满足\\n return False\\n\\n # 确保项目中确实生成了代码文件并且全部通过检查\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的前端与通用代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n has_checked_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关目录，避免扫描第三方库或者隐藏文件\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_checked_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = []\\n # 读取文件的前5行检查，以容忍可能存在的编码声明、空行等\\n for _ in range(5):\\n line = f.readline()\\n if not line:\\n break\\n lines.append(line)\\n except Exception:\\n # 如果由于权限或者编码问题无法读取文件内容，视为验证失败\\n return False\\n \\n # 判断这几行内是否存在目标注释内容\\n found = any(target_comment in line for line in lines)\\n \\n if not found:\\n return False\\n \\n # 如果项目中没有任何代码文件，应当视作未满足要求，否则返回True\\n return has_checked_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，主要为前端地图应用开发，包含以下常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略通常不由 AI 直接生成的大型依赖目录或版本控制目录\\n path_parts = root.split(os.sep)\\n if 'node_modules' in path_parts or '.git' in path_parts:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 连同空行一并计算总行数\\n if not content:\\n line_count = 0 # 空文件视为 0 行\\n else:\\n line_count = len(content.split('\\\\n'))\\n \\n # 判断是否为偶数行\\n if line_count % 2 != 0:\\n return False\\n except Exception:\\n # 如果由于编码异常或其他原因无法读取文件，无法证明满足要求，保守返回False\\n return False\\n \\n # 如果找到了至少一个代码文件且全部为偶数行，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 6 行注释\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_extensions = {'.js', '.ts', '.html', '.css', '.vue', '.jsx', '.tsx', '.py'}\\n file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n comment_lines = set()\\n\\n # JavaScript, TypeScript, CSS, JSX, TSX\\n if ext in ['.js', '.ts', '.css', '.jsx', '.tsx']:\\n # 匹配字符串以避免误判，并匹配单行和多行注释\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|((?<!:)\\\\/\\\\/.*|/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n for match in pattern.finditer(content):\\n if match.group(2): # 如果是注释部分\\n start = match.start(2)\\n end = match.end(2)\\n start_line = content.count('\\\\n', 0, start)\\n end_line = content.count('\\\\n', 0, end)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n\\n # HTML, Vue\\n elif ext in ['.html', '.vue']:\\n # 匹配 HTML 注释\\n pattern_html = re.compile(r'(<!--[\\\\s\\\\S]*?-->)')\\n for match in pattern_html.finditer(content):\\n start = match.start(1)\\n end = match.end(1)\\n start_line = content.count('\\\\n', 0, start)\\n end_line = content.count('\\\\n', 0, end)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n\\n # 同时匹配文件中可能包含的内部 JS/CSS 注释，同上忽略字符串中内容\\n pattern_js = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|((?<!:)\\\\/\\\\/.*|/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n for match in pattern_js.finditer(content):\\n if match.group(2):\\n start = match.start(2)\\n end = match.end(2)\\n start_line = content.count('\\\\n', 0, start)\\n end_line = content.count('\\\\n', 0, end)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n\\n # Python\\n elif ext == '.py':\\n # 匹配多行和单行字符串，避免将内部的 # 当作注释\\n pattern = re.compile(\\n r'(\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|(#.*)'\\n )\\n for match in pattern.finditer(content):\\n if match.group(2):\\n start = match.start(2)\\n end = match.end(2)\\n start_line = content.count('\\\\n', 0, start)\\n end_line = content.count('\\\\n', 0, end)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n\\n # 检查每个代码文件的注释行数是否精确包含 6 行\\n if len(comment_lines) != 6:\\n return False\\n\\n return file_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件里，缩进必须严格保持为4个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目常见的代码文件扩展名\\n code_extensions = {'.js', '.ts', '.html', '.css', '.json', '.jsx', '.tsx', '.py'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能包含第三方代码的依赖目录与隐藏目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__', 'venv', 'env')]\\n \\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 如果是完全的空行，则跳过不作缩进检查\\n if not line.strip():\\n continue\\n \\n # 提取前导空白字符串\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_whitespace += char\\n else:\\n break\\n \\n # 要求中明确提出必须为“空格”，若混入制表符则不符合要求\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 前导空格的数量必须是4的倍数\\n if len(leading_whitespace) % 4 != 0:\\n stripped = line.lstrip()\\n # 兼容 JSDoc 等多行注释排版（例如 ` * `），排除 `**` 以防误判\\n if stripped.startswith('*') and not stripped.startswith('**'):\\n continue\\n return False\\n except Exception:\\n # 若文件无法以utf-8读取，直接跳过处理下一文件\\n continue\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的每一个代码文件里，其注释部分包含关键词 \\\"FIXME\\\" 的次数正好是1次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_fixme_in_comments(response: str, workspace_path: str) -> bool:\\n # 根据用户前端交互式地图应用指令推断，可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 正则表达式：同时匹配字符串和各种注释，以防提取到字符串中包含的假注释符号\\n # 捕获组 1 (group 1) 仅用于匹配真实注释内容，字符串字面量将在组 0 匹配而被过滤\\n pattern = re.compile(\\n r'(/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->|(?<!:)\\\\/\\\\/.*|#.*)|'\\n r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n )\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常规的非产物目录，保留 __tests__ 因为这是指令明确要求的测试代码目录\\n dirs[:] = [d for d in dirs if d not in ('.git', 'node_modules', 'dist', 'build', '__pycache__')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n # 如果文件无法按 UTF-8 读取（违反了指令的全局 UTF-8 编码要求），则直接视为不满足\\n return False\\n \\n # 提取出当前文件中所有的注释内容（正则表达式中 group(1) 成功匹配到的即为注释内容）\\n comments = [m.group(1) for m in pattern.finditer(code) if m.group(1)]\\n \\n # 统计当前文件所有注释中关键词 \\\"FIXME\\\" 的总出现次数\\n fixme_count = sum(comment.count(\\\"FIXME\\\") for comment in comments)\\n \\n # 若任意一个代码文件的注释中 \\\"FIXME\\\" 次数不等于 1，说明不满足条件\\n if fixme_count != 1:\\n return False\\n \\n # 如果找到了代码文件且均通过了测试，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现\\\"OPTIMIZE\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的前端及后端代码文件扩展名集合\\n code_exts = {\\n '.html', '.htm', '.css', '.scss', '.less',\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte',\\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n\\n has_code_file = False\\n \\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉无关目录以防止误判依赖库和编译产物\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', 'dist', 'build', 'out'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对属于代码文件的后缀进行内容读取与检查\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 确保不检查空文件\\n if content.strip():\\n has_code_file = True\\n # 若有任何一个代码文件不包含\\\"OPTIMIZE\\\"，则直接返回False\\n if \\\"OPTIMIZE\\\" not in content:\\n return False\\n except Exception:\\n # 忽略无法正常读取或非utf-8编码的异常文件\\n pass\\n \\n # 只要存在至少一个代码文件且都通过了检查，即返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你的回复里，如需引用文件或代码的具体位置，请务必采用markdown链接语法，格式要求如下：[filename.ts](src/filename.ts)、[filename.ts:42](src/filename.ts#L42)、[filename.ts:42-51](src/filename.ts#L42-L51)\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 9,\n \"instruction\": \"最后一轮，我们需要开发一个便于分享和打印的“导出路书”沉浸模式。在侧边栏新增一个“生成路书”按钮。点击后，页面发生平滑过渡：隐藏顶部导航栏、侧边栏和所有的地图悬浮控件；地图区域变为占据屏幕上方40%的空间（自动居中显示整条路线）；屏幕下方60%的空间则渲染一个美观的“路书卡片”网格，将时间轴、每个站点的ETA和天气预测以大字号、精美排版的形式展示。在这个沉浸模式的右上角，提供一个“打印/保存为PDF”的占位按钮，以及一个“退出路书”按钮用于恢复到之前的编辑界面状态。\\n另外请注意，后续所有轮次都必须严格遵循以下更新要求：第一，此前要求的每个生成的代码文件精确包含6行注释的规则调整为，针对每一个生成的代码文件，注释的行数需精准控制在7行；第二，后续所有轮次请始终遵循，你输出的单元测试文件经运行，需确保目标代码获得至少85%的语句覆盖率。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行任何文件处理之前，切勿使用`pwd`确定当前的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你创建的代码文件无一例外，全都要用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，该项目为前端页面应用，推测可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json'}\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果报 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能因权限或系统原因导致的读取报错\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 3 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件的注释中，严格控制关键词 \\\"TODO\\\" 出现且仅出现 2 次\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义我们要检查的代码文件扩展名集合\\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.css', '.vue', '.py'}\\n found_code_file = False\\n\\n def get_comments_from_code(content: str, ext: str) -> str:\\n comments = []\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.css']:\\n # 该正则按从左到右匹配优先级，巧妙跳过字符串内包含的伪注释符号（如URL中的 //）\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|' # Group 1: 双引号字符串\\n r\\\"('(?:\\\\\\\\.|[^'\\\\\\\\])*')|\\\" # Group 2: 单引号字符串\\n r'(`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|' # Group 3: 模板字符串（反引号）\\n r'(<!--[\\\\s\\\\S]*?-->)|' # Group 4: HTML注释\\n r'(/\\\\*[\\\\s\\\\S]*?\\\\*/)|' # Group 5: 多行块注释\\n r'(//.*)' # Group 6: 单行注释\\n )\\n for match in pattern.finditer(content):\\n # 如果匹配到了对应的注释组，则收集起来\\n if match.group(4):\\n comments.append(match.group(4))\\n elif match.group(5):\\n comments.append(match.group(5))\\n elif match.group(6):\\n comments.append(match.group(6))\\n \\n elif ext == '.py':\\n pattern = re.compile(\\n r'(\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\")|' # Group 1: 三双引号字符串/注释\\n r\\\"('''[\\\\s\\\\S]*?''')|\\\" # Group 2: 三单引号字符串/注释\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|' # Group 3: 双引号\\n r\\\"('(?:\\\\\\\\.|[^'\\\\\\\\])*')|\\\" # Group 4: 单引号\\n r'(#.*)' # Group 5: 井号单行注释\\n )\\n for match in pattern.finditer(content):\\n if match.group(1):\\n comments.append(match.group(1))\\n elif match.group(2):\\n comments.append(match.group(2))\\n elif match.group(5):\\n comments.append(match.group(5))\\n \\n return \\\" \\\".join(comments)\\n\\n # 遍历工作目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略掉无关的依赖或版本管理文件夹，保留由AI生成的代码及__tests__测试代码\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n found_code_file = True\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取该文件中的所有注释内容\\n comments_text = get_comments_from_code(content, ext)\\n \\n # 统计关键词 \\\"TODO\\\" 的出现次数\\n todo_count = comments_text.count(\\\"TODO\\\")\\n \\n # 如果存在任何一个代码文件，其注释中的 TODO 数量不等于 2，即不满足要求\\n if todo_count != 2:\\n return False\\n \\n # 如果没有找到任何代码文件，则也视为失败\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请注意，每一个输出的代码文件首行都需要指定编码格式，即添加：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 根据用户指令的前端项目特征，推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的依赖和版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 验证首行是否包含指定的编码格式声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n has_code_files = True\\n except Exception:\\n # 对于无法作为文本打开的异常文件可以选择跳过\\n continue\\n\\n # 只有当至少验证过一个代码文件且全部通过时，才算满足要求\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件中，请于开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件后缀名\\n code_extensions = {'.py', '.js', '.ts', '.html', '.css', '.jsx', '.tsx', '.vue'}\\n # 目标注释声明\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制和常见依赖包目录\\n if '.git' in root or 'node_modules' in root or '__pycache__' in root:\\n continue\\n \\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果遇到无法用UTF-8读取的异常情况，跳过或视为不合规，此处为避免非文本文件导致的错误选择跳过\\n continue\\n \\n found_in_header = False\\n # 检查文件的前15行（视为文件开头处）\\n for line in lines[:15]:\\n cleaned = line.strip()\\n # 剥离各种可能的注释符号\\n for comment_mark in ['//', '#', '/*', '*/', '<!--', '-->', '*']:\\n cleaned = cleaned.replace(comment_mark, '')\\n cleaned = cleaned.strip()\\n \\n # 验证剥离注释符后是否等于目标声明（以此保障它是单独一行的注释）\\n if cleaned == target_text:\\n found_in_header = True\\n break\\n \\n # 若任意一个代码文件缺少此头部声明，则说明未完成要求\\n if not found_in_header:\\n return False\\n \\n # 确保项目中至少存在一个被检测的代码文件，并且全部通过了检查\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件需满足一项条件：每个文件的总行数（连同空行一并计算）必须为偶数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_even_lines(response: str, workspace_path: str) -> bool:\\n # 定义前端项目中常规会被生成的代码文件扩展名\\n valid_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.vue'}\\n # 忽略不需要检查的目录\\n exclude_dirs = {'node_modules', '.git', '.idea', '.vscode', 'dist', 'build', '__pycache__'}\\n \\n files_checked = 0\\n \\n # 1. 优先检查 workspace_path 中的文件\\n if workspace_path and os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按换行符分割以获取包含空行在内的总行数\\n lines = content.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n \\n files_checked += 1\\n except Exception:\\n pass\\n \\n # 如果在工作区中找到了代码文件并全都通过了检查，则认为满足条件\\n if files_checked > 0:\\n return True\\n \\n # 2. 如果工作区未找到符合要求的文件，退而检查 response 中的 Markdown 代码块\\n # 匹配 ``` 开头且以 ``` 结尾的代码块内容\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n \\n # 如果没有任何代码文件和代码块，认为未满足生成代码文件的要求\\n if not code_blocks:\\n return False\\n \\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n if len(lines) % 2 != 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"针对每一个生成的代码文件，注释的行数需精准控制在7行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n def count_comment_lines(file_path):\\n ext = os.path.splitext(file_path)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except UnicodeDecodeError:\\n return -1\\n \\n comment_lines = set()\\n \\n # 处理 C系、Web系语言\\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.css', '.c', '.cpp', '.java', '.cs'}:\\n pattern = re.compile(\\n r'(?P<line_comment>//[^\\\\n]*)|(?P<block_comment>/\\\\*.*?\\\\*/)|(?P<string1>\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\')|(?P<string2>\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\")|(?P<string3>`(?:\\\\\\\\.|[^\\\\\\\\`])*`)',\\n re.DOTALL\\n )\\n for m in pattern.finditer(code):\\n if m.group('line_comment') or m.group('block_comment'):\\n if m.group('line_comment'):\\n start_pos = m.start()\\n if start_pos > 0 and code[start_pos-1] == ':':\\n continue\\n start_pos = m.start()\\n end_pos = m.end()\\n start_line = code.count('\\\\n', 0, start_pos)\\n end_line = code.count('\\\\n', 0, end_pos)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n \\n # 处理 脚本系语言\\n elif ext in {'.py', '.sh', '.rb', '.yml', '.yaml'}:\\n pattern = re.compile(\\n r'(?P<line_comment>#[^\\\\n]*)|(?P<string3>\\\"\\\"\\\".*?\\\"\\\"\\\")|(?P<string4>\\\\'\\\\'\\\\'.*?\\\\'\\\\'\\\\')|(?P<string1>\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\')|(?P<string2>\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\")',\\n re.DOTALL\\n )\\n for m in pattern.finditer(code):\\n if m.group('line_comment'):\\n start_pos = m.start()\\n end_pos = m.end()\\n start_line = code.count('\\\\n', 0, start_pos)\\n end_line = code.count('\\\\n', 0, end_pos)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n \\n # 处理 标记系语言及可能混合的代码\\n elif ext in {'.html', '.vue', '.xml'}:\\n pattern = re.compile(\\n r'(?P<html_comment><!--.*?-->)|(?P<line_comment>//[^\\\\n]*)|(?P<block_comment>/\\\\*.*?\\\\*/)|(?P<string1>\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\')|(?P<string2>\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\")|(?P<string3>`(?:\\\\\\\\.|[^\\\\\\\\`])*`)',\\n re.DOTALL\\n )\\n for m in pattern.finditer(code):\\n if m.group('html_comment') or m.group('line_comment') or m.group('block_comment'):\\n if m.group('line_comment'):\\n start_pos = m.start()\\n if start_pos > 0 and code[start_pos-1] == ':':\\n continue\\n start_pos = m.start()\\n end_pos = m.end()\\n start_line = code.count('\\\\n', 0, start_pos)\\n end_line = code.count('\\\\n', 0, end_pos)\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n else:\\n return -1\\n\\n return len(comment_lines)\\n\\n valid_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.css', '.c', '.cpp', '.java', '.cs',\\n '.py', '.sh', '.rb', '.yml', '.yaml',\\n '.html', '.vue', '.xml'\\n }\\n\\n files_checked = 0\\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', 'venv', 'env', '__pycache__', '.idea', '.vscode'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关的构建与包目录，保留 __tests__ 等用户生成源码的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n cnt = count_comment_lines(file_path)\\n if cnt != -1:\\n files_checked += 1\\n # 核心验证逻辑：注释行数必须精准控制在 7 行\\n if cnt != 7:\\n return False\\n except Exception:\\n pass\\n\\n return files_checked > 0\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"项目目录的__tests__子目录应当用于存放并保留你对生成代码进行的所有中间测试代码\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件里，缩进必须严格保持为4个空格\",\n \"validation_code\": \"def verify_indentation_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端与后端代码文件扩展名\\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue', '.json', '.py'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于某种原因无法读取文件，则跳过继续检查其他文件\\n continue\\n\\n for line in lines:\\n # 匹配行首的前导空白字符和第一个非空白字符及其后的内容\\n match = re.match(r'^([ \\\\t]+)(\\\\S.*)', line)\\n if match:\\n leading_space = match.group(1)\\n content = match.group(2)\\n\\n # 如果包含制表符，直接判定为不符合要求\\n if '\\\\t' in leading_space:\\n return False\\n\\n # 检查前导空格数是否严格为 4 的倍数\\n if len(leading_space) % 4 != 0:\\n # 豁免多行注释中以 '*' 开头的行（例如 /**\\\\n * 这种常规对齐格式）\\n if not content.startswith('*'):\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在你生成的每一个代码文件里，其注释部分包含关键词 \\\"FIXME\\\" 的次数正好是1次\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义前端应用通常会涉及到的代码文件扩展名\\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.css', '.html', '.htm', '.vue', '.py'}\\n has_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制和常见的依赖/缓存目录，不将它们作为生成的代码文件检查\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if '__pycache__' in dirs:\\n dirs.remove('__pycache__')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n has_files = True\\n filepath = os.path.join(root, file)\\n \\n try:\\n # 按照用户指令要求，所有文件均为 UTF-8 编码\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件无法读取，视同验证失败\\n return False\\n \\n comments = []\\n \\n # 使用正则表达式匹配，确保准确提取注释，忽略字符串中的干扰内容\\n if ext in {'.html', '.htm', '.vue'}:\\n # 匹配 HTML 注释、字符串、JS/CSS 多行注释、JS/CSS 单行注释\\n pattern = re.compile(\\n r'(<!--[\\\\s\\\\S]*?-->)' # Group 1: HTML 注释\\n r'|(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' # Group 2: 字符串（避免误判）\\n r'|(/\\\\*[\\\\s\\\\S]*?\\\\*/)' # Group 3: 多行注释\\n r'|(//[^\\\\n]*)' # Group 4: 单行注释\\n )\\n for match in pattern.finditer(content):\\n if match.group(1):\\n comments.append(match.group(1))\\n elif match.group(3):\\n comments.append(match.group(3))\\n elif match.group(4):\\n comments.append(match.group(4))\\n \\n elif ext in {'.py'}:\\n # 匹配 Python 字符串（含多行字符串）和单行注释\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\')'\\n r'|(#.*)'\\n )\\n for match in pattern.finditer(content):\\n if match.group(2):\\n comments.append(match.group(2))\\n \\n else:\\n # 匹配 JS/TS/CSS 等文件的字符串、多行注释、单行注释\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n r'|(//[^\\\\n]*)'\\n )\\n for match in pattern.finditer(content):\\n if match.group(2):\\n comments.append(match.group(2))\\n elif match.group(3):\\n comments.append(match.group(3))\\n \\n # 统计当前代码文件中，所有注释里出现 \\\"FIXME\\\" 的总次数\\n fixme_count = sum(c.count(\\\"FIXME\\\") for c in comments)\\n \\n # 如果不是恰好出现1次，直接判定失败\\n if fixme_count != 1:\\n return False\\n \\n return has_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现\\\"OPTIMIZE\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见代码文件后缀集合\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的第三方依赖或版本控制文件夹\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若当前代码文件中不存在要求的关键词 \\\"OPTIMIZE\\\"，则直接返回 False\\n if 'OPTIMIZE' not in content:\\n return False\\n except Exception:\\n # 如果由于某种原因（如编码问题）无法读取该文件，则安全跳过\\n pass\\n \\n # 如果至少检查了一个代码文件，并且所有代码文件都包含 \\\"OPTIMIZE\\\"，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你的回复里，如需引用文件或代码的具体位置，请务必采用markdown链接语法，格式要求如下：[filename.ts](src/filename.ts)、[filename.ts:42](src/filename.ts#L42)、[filename.ts:42-51](src/filename.ts#L42-L51)\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少85%的语句覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["观察页面顶部，确认存在一个顶部导航栏，且内部包含项目的Logo图标和用户的头像元素。","观察页面左侧，确认存在一个包含'行程规划'标题占位的侧边栏；点击侧边栏的折叠/展开按钮（或边缘区域），确认侧边栏能够正常进行折叠隐藏和展开操作。","观察页面右下角，确认放置了一个悬浮的'新建行程'按钮（FAB）。","在右侧主地图区域按下鼠标并拖拽，确认地图视图可以跟随鼠标平移；滚动鼠标滚轮，确认地图能够正常进行放大和缩小操作。","在地图的任意空白位置点击鼠标，确认点击的位置立即出现了一个蓝色的位置定位标记（Marker）。","鼠标点击上一步放置的蓝色位置标记，确认标记上方弹出一个提示框（Popup），且内部展示了该标记点的准确经纬度坐标信息。","在侧边栏中找到'出发地'和'目的地'输入框，分别输入测试地点后点击'规划路线'按钮，截取地图图片，确认地图上生成了两个不同颜色（如绿色和红色）的标记以区分起终点。","接着上一步操作，观察地图区域，确认在起点和终点标记之间，绘制出了一条实线的折线（Polyline）用于代表生成的路线。","点击'规划路线'按钮时不做任何地图拖拽，确认地图视口会自动进行缩放和平移，使得生成的起点、终点和它们之间的路线刚好完整居中展示在当前可视范围内。","在侧边栏中点击'添加途经点'按钮，确认每点击一次会新增一个途经点输入框；连续点击验证最多只能成功新增5个输入框（超过后按钮禁用或不生效）。","观察新增的途经点输入框，确认每个输入框旁边都有一个'删除'按钮；点击其中一个删除按钮，确认该对应的途经点输入框从侧边栏中消失。","在侧边栏添加至少两个途经点输入框，使用鼠标拖拽其中一个途经点输入框至另一个输入框的上方或下方，确认松开鼠标后途经点的先后顺序成功发生改变（Drag and Drop 排序成功）。","在侧边栏输入多个途经点并打乱顺序，点击'规划路线'按钮，截取地图图片，确认折线严格按照'起点 -> 侧边栏排序后的途经点 -> 终点'的先后顺序依次连接，且地图上的途经点图标标有数字序号（如1、2、3...）。","规划路线完成后，观察侧边栏表单下方新增的'行程详情'面板，确认面板顶部显示了模拟计算得到的'总距离'和'预计驾驶时长'文本数据。","在'行程详情'面板中向下浏览，确认展示了一个垂直的行程时间轴列表，列表内的站点名称严格按照'起点 -> 各个途经点 -> 终点'的顺序从上到下排列。","观察垂直行程时间轴列表，确认在相邻的任意两个路段节点（如起点和途经点1）之间，明确标注了该特定路段的'距离'和'预估耗时'信息。","观察行程时间轴列表中的每一个站点节点（起终点和途经点），确认在站点的右侧展示了模拟的天气数据，包括直观的天气图标（晴/雨/多云等）、当前温度和降水概率。","在地图上点击起点、终点或途经点标记，截取弹出的提示框（Popup），确认除了显示地名外，还同步展示了与时间轴上完全一致的该站点模拟天气数据（天气图标、温度、降水概率）。","在侧边栏顶部的'出发时间'日期时间选择器中设置一个自定义时间，点击'规划路线'，确认时间轴列表中每一个站点都基于耗时计算并显示了具体的'预计到达时间（ETA）'。","设定出发时间并规划路线后，点击地图上的某个站点标记，确认其弹出的提示框中同样包含了对应的'预计到达时间（ETA）'，且天气数据也产生动态关联变化。","在地图区域的底部找到'行程时间轴滑块（Slider）'，使用鼠标左右拖动滑块模拟旅程推进，截取地图图片，确认当前进度所处路段的折线变为高亮颜色，且高亮路段会随滑块的移动而动态改变。","在地图右上角的半透明地图控制面板中，点击'地图样式'下拉菜单，选择与当前不同的选项（如'暗黑模式'或'地形图'），确认地图底层的瓦片背景立刻切换为对应的样式外观。","在地图控制面板的'沿途设施'区域，勾选'加油站'复选框，截取地图图片，确认在路线沿途的两侧随机散布显示了多个代表加油站的定制小图标。","在上一步的基础上，取消勾选'加油站'复选框，确认地图上原先散布的加油站定制小图标全部消失不见。","在侧边栏最底部点击'保存行程'按钮，确认弹出一个要求输入'行程名称'的对话框，输入内容并点击确认，对话框正常关闭。","点击顶部导航栏中的'我的行程'入口，确认页面右侧平滑滑出一个抽屉（Drawer）或模态框，且列表内正确展示了刚才输入的已保存'历史行程名称'。","刷新整个页面以清空当前状态，再次点击'我的行程'并选择历史列表中的某一行程，确认侧边栏的表单输入框数据和地图上的路线被重新渲染恢复到了保存时的状态。","在侧边栏点击'生成路书'按钮进入沉浸模式，截取全屏图片，确认页面顶部的导航栏、左侧的侧边栏以及地图上的悬浮控制面板等非路书元素全被隐藏不见。","在沉浸模式下截取全屏图片，确认屏幕上方约40%的空间被地图占据（且包含完整的居中路线），屏幕下方约60%的空间渲染为美观的'路书卡片'网格，大字号展示了时间轴、ETA和天气预测等信息。","在沉浸模式的右上角，找到'打印/保存为PDF'按钮和'退出路书'按钮；点击'退出路书'按钮，确认页面发生平滑过渡，恢复到正常显示侧边栏和导航栏的编辑界面初始状态。","在地图控制面板的'沿途设施'区域，分别勾选'餐厅'和'酒店'复选框，截取地图图片，确认在路线沿途散布显示了对应类型的定制小图标；随后取消勾选，确认对应图标从地图上消失。"],"string":"[\n \"观察页面顶部，确认存在一个顶部导航栏，且内部包含项目的Logo图标和用户的头像元素。\",\n \"观察页面左侧，确认存在一个包含'行程规划'标题占位的侧边栏；点击侧边栏的折叠/展开按钮（或边缘区域），确认侧边栏能够正常进行折叠隐藏和展开操作。\",\n \"观察页面右下角，确认放置了一个悬浮的'新建行程'按钮（FAB）。\",\n \"在右侧主地图区域按下鼠标并拖拽，确认地图视图可以跟随鼠标平移；滚动鼠标滚轮，确认地图能够正常进行放大和缩小操作。\",\n \"在地图的任意空白位置点击鼠标，确认点击的位置立即出现了一个蓝色的位置定位标记（Marker）。\",\n \"鼠标点击上一步放置的蓝色位置标记，确认标记上方弹出一个提示框（Popup），且内部展示了该标记点的准确经纬度坐标信息。\",\n \"在侧边栏中找到'出发地'和'目的地'输入框，分别输入测试地点后点击'规划路线'按钮，截取地图图片，确认地图上生成了两个不同颜色（如绿色和红色）的标记以区分起终点。\",\n \"接着上一步操作，观察地图区域，确认在起点和终点标记之间，绘制出了一条实线的折线（Polyline）用于代表生成的路线。\",\n \"点击'规划路线'按钮时不做任何地图拖拽，确认地图视口会自动进行缩放和平移，使得生成的起点、终点和它们之间的路线刚好完整居中展示在当前可视范围内。\",\n \"在侧边栏中点击'添加途经点'按钮，确认每点击一次会新增一个途经点输入框；连续点击验证最多只能成功新增5个输入框（超过后按钮禁用或不生效）。\",\n \"观察新增的途经点输入框，确认每个输入框旁边都有一个'删除'按钮；点击其中一个删除按钮，确认该对应的途经点输入框从侧边栏中消失。\",\n \"在侧边栏添加至少两个途经点输入框，使用鼠标拖拽其中一个途经点输入框至另一个输入框的上方或下方，确认松开鼠标后途经点的先后顺序成功发生改变（Drag and Drop 排序成功）。\",\n \"在侧边栏输入多个途经点并打乱顺序，点击'规划路线'按钮，截取地图图片，确认折线严格按照'起点 -> 侧边栏排序后的途经点 -> 终点'的先后顺序依次连接，且地图上的途经点图标标有数字序号（如1、2、3...）。\",\n \"规划路线完成后，观察侧边栏表单下方新增的'行程详情'面板，确认面板顶部显示了模拟计算得到的'总距离'和'预计驾驶时长'文本数据。\",\n \"在'行程详情'面板中向下浏览，确认展示了一个垂直的行程时间轴列表，列表内的站点名称严格按照'起点 -> 各个途经点 -> 终点'的顺序从上到下排列。\",\n \"观察垂直行程时间轴列表，确认在相邻的任意两个路段节点（如起点和途经点1）之间，明确标注了该特定路段的'距离'和'预估耗时'信息。\",\n \"观察行程时间轴列表中的每一个站点节点（起终点和途经点），确认在站点的右侧展示了模拟的天气数据，包括直观的天气图标（晴/雨/多云等）、当前温度和降水概率。\",\n \"在地图上点击起点、终点或途经点标记，截取弹出的提示框（Popup），确认除了显示地名外，还同步展示了与时间轴上完全一致的该站点模拟天气数据（天气图标、温度、降水概率）。\",\n \"在侧边栏顶部的'出发时间'日期时间选择器中设置一个自定义时间，点击'规划路线'，确认时间轴列表中每一个站点都基于耗时计算并显示了具体的'预计到达时间（ETA）'。\",\n \"设定出发时间并规划路线后，点击地图上的某个站点标记，确认其弹出的提示框中同样包含了对应的'预计到达时间（ETA）'，且天气数据也产生动态关联变化。\",\n \"在地图区域的底部找到'行程时间轴滑块（Slider）'，使用鼠标左右拖动滑块模拟旅程推进，截取地图图片，确认当前进度所处路段的折线变为高亮颜色，且高亮路段会随滑块的移动而动态改变。\",\n \"在地图右上角的半透明地图控制面板中，点击'地图样式'下拉菜单，选择与当前不同的选项（如'暗黑模式'或'地形图'），确认地图底层的瓦片背景立刻切换为对应的样式外观。\",\n \"在地图控制面板的'沿途设施'区域，勾选'加油站'复选框，截取地图图片，确认在路线沿途的两侧随机散布显示了多个代表加油站的定制小图标。\",\n \"在上一步的基础上，取消勾选'加油站'复选框，确认地图上原先散布的加油站定制小图标全部消失不见。\",\n \"在侧边栏最底部点击'保存行程'按钮，确认弹出一个要求输入'行程名称'的对话框，输入内容并点击确认，对话框正常关闭。\",\n \"点击顶部导航栏中的'我的行程'入口，确认页面右侧平滑滑出一个抽屉（Drawer）或模态框，且列表内正确展示了刚才输入的已保存'历史行程名称'。\",\n \"刷新整个页面以清空当前状态，再次点击'我的行程'并选择历史列表中的某一行程，确认侧边栏的表单输入框数据和地图上的路线被重新渲染恢复到了保存时的状态。\",\n \"在侧边栏点击'生成路书'按钮进入沉浸模式，截取全屏图片，确认页面顶部的导航栏、左侧的侧边栏以及地图上的悬浮控制面板等非路书元素全被隐藏不见。\",\n \"在沉浸模式下截取全屏图片，确认屏幕上方约40%的空间被地图占据（且包含完整的居中路线），屏幕下方约60%的空间渲染为美观的'路书卡片'网格，大字号展示了时间轴、ETA和天气预测等信息。\",\n \"在沉浸模式的右上角，找到'打印/保存为PDF'按钮和'退出路书'按钮；点击'退出路书'按钮，确认页面发生平滑过渡，恢复到正常显示侧边栏和导航栏的编辑界面初始状态。\",\n \"在地图控制面板的'沿途设施'区域，分别勾选'餐厅'和'酒店'复选框，截取地图图片，确认在路线沿途散布显示了对应类型的定制小图标；随后取消勾选，确认对应图标从地图上消失。\"\n]","isConversational":false}}},{"rowIdx":70,"cells":{"task_id":{"kind":"number","value":71,"string":"71"},"repository_policy":{"kind":"string","value":"你需要严格遵循以下所有规则，且后续所有对话轮次都必须始终遵守，不得违反：\n1. 无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词。\n2. 你的所有回复都必须在结尾处加上\"呜~\"语气词。\n3. 禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"。\n4. 在你输出的每一个代码文件中，务必保证所有的注释均为中文。\n5. 在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具。\n6. 在每次编辑已有文件之前，必须先借助对应工具审阅该文件。\n7. 为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装。\n8. 只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复的最开头是否带有\"汪~\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除可能存在的前导空白字符，然后判断最开头是否为\"汪~\"\n return response.lstrip().startswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 去除回复内容两端的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 检查清理后的回复内容是否以\"呜~\"结尾\n if cleaned_response.endswith(\"呜~\"):\n return True\n else:\n return False","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含被禁用的称呼词“你”\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]}],"string":"[\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复的最开头是否带有\\\"汪~\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除可能存在的前导空白字符，然后判断最开头是否为\\\"汪~\\\"\\n return response.lstrip().startswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 去除回复内容两端的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 检查清理后的回复内容是否以\\\"呜~\\\"结尾\\n if cleaned_response.endswith(\\\"呜~\\\"):\\n return True\\n else:\\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含被禁用的称呼词“你”\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我开发一个基于Web的“自由落体物理模拟实验工具”。第一步，请先搭建好基础的页面骨架。需求如下：1. 页面分为左右两栏布局，左侧是占据主要屏幕空间的“实验舱”，请使用 Canvas 或带边框的大块 DOM 区域，底部画一条粗线代表“地面”，实验舱靠上的位置放置一个静止的小球作为实验对象。2. 右侧是“控制面板”，包含三个结构清晰的卡片占位：实验参数设置区、播放控制区（放置开始、暂停、重置三个按钮，暂不实现功能）、实时数据看板。3. 整体界面风格保持现代、极简。目前阶段不需要编写任何物理运动计算逻辑，只要把静态排版和基本UI元素渲染出来即可。\n另外，请你在后续所有轮次的开发过程中都严格遵循以下要求：第一，生成的每一个文件行数均在100-200行之间；第二，输出的代码文件里不允许出现行数超过30行的函数。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除回复首部的空白字符后，判断是否以\"汪~\"开头\n return response.lstrip().startswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 匹配末尾的\"呜~\"（兼容半角和全角波浪号），允许其后跟随空白字符及常见标点符号\n pattern = r'呜[~～][\\s。，！、？；：\\.,!;:\\'\\\"”’]*$'\n \n return bool(re.search(pattern, response))","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复内容中是否包含\"你\"字\n if \"你\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的每一个文件行数均在100-200行之间","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n file_count = 0\n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 忽略隐藏文件\n if file.startswith('.'):\n continue\n \n # 根据用户指令推断，Web项目代码文件通常为 .html, .css, .js\n if not file.endswith(('.html', '.css', '.js')):\n continue\n \n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n file_count += 1\n \n # 判断文件行数是否在100至200之间\n if len(lines) < 100 or len(lines) > 200:\n return False\n except Exception:\n # 读取异常可能因为非纯文本或是编码问题，此处作为无效生成情况返回False\n return False\n \n # 必须保证至少成功检查了一个以上的目标文件\n return file_count > 0","tags":["Quantity","Range"]},{"constraint":"输出的代码文件里不允许出现行数超过30行的函数","validation_code":"def check_function_lines_limit(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def clean_code(code: str) -> str:\n \"\"\"净化代码，将注释、字符串内容替换为空格，但严格保留所有换行符以保持行号对应\"\"\"\n state = 'NORMAL'\n result = []\n i = 0\n n = len(code)\n while i < n:\n char = code[i]\n if state == 'NORMAL':\n if char == '/' and i + 1 < n and code[i+1] == '/':\n state = 'LINE_COMMENT'\n result.append(' ')\n i += 1\n elif char == '/' and i + 1 < n and code[i+1] == '*':\n state = 'BLOCK_COMMENT'\n result.append(' ')\n i += 1\n elif char == '\"':\n state = 'STRING_DOUBLE'\n result.append(char)\n elif char == \"'\":\n state = 'STRING_SINGLE'\n result.append(char)\n elif char == '`':\n state = 'STRING_BACKTICK'\n result.append(char)\n else:\n result.append(char)\n elif state == 'LINE_COMMENT':\n if char == '\\n':\n state = 'NORMAL'\n result.append(char)\n else:\n result.append(' ')\n elif state == 'BLOCK_COMMENT':\n if char == '*' and i + 1 < n and code[i+1] == '/':\n state = 'NORMAL'\n result.append(' ')\n i += 1\n elif char == '\\n':\n result.append(char)\n else:\n result.append(' ')\n elif state == 'STRING_DOUBLE':\n if char == '\\\\':\n result.append(' ')\n i += 1\n if i < n:\n result.append(code[i] if code[i] == '\\n' else ' ')\n elif char == '\"':\n state = 'NORMAL'\n result.append(char)\n elif char == '\\n':\n result.append(char)\n else:\n result.append(' ')\n elif state == 'STRING_SINGLE':\n if char == '\\\\':\n result.append(' ')\n i += 1\n if i < n:\n result.append(code[i] if code[i] == '\\n' else ' ')\n elif char == \"'\":\n state = 'NORMAL'\n result.append(char)\n elif char == '\\n':\n result.append(char)\n else:\n result.append(' ')\n elif state == 'STRING_BACKTICK':\n if char == '\\\\':\n result.append(' ')\n i += 1\n if i < n:\n result.append(code[i] if code[i] == '\\n' else ' ')\n elif char == '`':\n state = 'NORMAL'\n result.append(char)\n elif char == '\\n':\n result.append(char)\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n\n def check_if_function_prefix(prefix: str) -> bool:\n \"\"\"检查 '{' 前方的代码前缀，判断该括号是否属于一个函数的开头部位\"\"\"\n prefix = prefix.rstrip()\n if not prefix:\n return False\n \n # 匹配箭头函数特征\n if prefix.endswith('=>'):\n return True\n \n i = len(prefix) - 1\n while i >= 0 and prefix[i].isspace():\n i -= 1\n \n if i < 0:\n return False\n \n # 应对带有返回类型注解的 TS 语法 (例如 : void { ) 寻找上一层 ')'\n if prefix[i] != ')':\n valid_type_chars = set(\"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_$:<>,[]|&. \\t\\n\")\n found_colon = False\n while i >= 0:\n if prefix[i] == ':':\n found_colon = True\n i -= 1\n break\n if prefix[i] not in valid_type_chars:\n return False\n i -= 1\n \n if not found_colon:\n return False\n \n while i >= 0 and prefix[i].isspace():\n i -= 1\n if i < 0 or prefix[i] != ')':\n return False\n\n # 进行括号向左匹配寻找对应的 '('\n paren_count = 1\n i -= 1\n while i >= 0:\n if prefix[i] == ')':\n paren_count += 1\n elif prefix[i] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n i -= 1\n \n if paren_count == 0 and i >= 0:\n j = i - 1\n while j >= 0 and prefix[j].isspace():\n j -= 1\n \n token_end = j\n while j >= 0 and (prefix[j].isalnum() or prefix[j] in '_$'):\n j -= 1\n token_start = j + 1\n token = prefix[token_start:token_end+1]\n \n # 排除控制流和特定关键字\n control_keywords = {'if', 'for', 'while', 'catch', 'switch', 'with'}\n if token in control_keywords:\n return False\n return True\n\n return False\n\n def check_jsts_content(content: str) -> bool:\n \"\"\"解析并检查 JS/TS 等类文件的内容行数是否超标\"\"\"\n cleaned = clean_code(content)\n line_num = 1\n stack = [] \n \n for idx, char in enumerate(cleaned):\n if char == '\\n':\n line_num += 1\n elif char == '{':\n start_prefix = max(0, idx - 500)\n prefix = cleaned[start_prefix:idx]\n is_func = check_if_function_prefix(prefix)\n stack.append((line_num, is_func))\n elif char == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n # 检查总行数（结束行-开始行+1）是否超过 30 行\n if line_num - start_line + 1 > 30:\n return False\n return True\n\n def check_python_file(content: str) -> bool:\n \"\"\"使用 ast 模块解析并检查 Python 文件行数是否超标\"\"\"\n try:\n tree = ast.parse(content)\n except:\n return True\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n if node.end_lineno - node.lineno + 1 > 30:\n return False\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except:\n continue\n \n if file.endswith('.py'):\n if not check_python_file(content):\n return False\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.mjs', '.cjs')):\n if not check_jsts_content(content):\n return False\n elif file.endswith(('.html', '.htm', '.vue')):\n # 针对 HTML/Vue 文件内嵌脚本的处理\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script_content in scripts:\n if not check_jsts_content(script_content):\n return False\n\n return True","tags":["Quantity","Range"]}]},{"round_id":1,"instruction":"很好，骨架搭出来了。接下来我们实现核心的物理引擎和基础控制逻辑。请为左侧实验舱建立一个“像素与物理现实”的映射关系（例如设定 50 像素等于 1 米，并在界面角落标出比例尺）。给小球赋予固定的重力加速度（g=9.8m/s²）。实现控制面板里的“开始”、“暂停”和“重置”按钮逻辑：点击开始后，小球开始受重力影响做加速下落运动；触碰到地面时，小球绝对不能穿模，并且需要以 0.7 的弹性系数向上反弹，反复弹跳直到速度趋近于零最终静止；重置按钮能清空状态并让小球回到页面顶部的初始位置。另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，生成的每一个文件行数调整为需在120-180行之间；第二，请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复最开头是否带有\"汪~\"语气词。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，代表该要求是否满足。\n \"\"\"\n if not response:\n return False\n \n # 移除可能存在的前导空白字符后，判断是否以\"汪~\"开头\n cleaned_response = response.lstrip()\n return cleaned_response.startswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import re\n if not response:\n return False\n \n # 将回复内容末尾的空白字符以及常见的标点符号去除，以防止误判\n clean_response = re.sub(r'[。，！？\\.\\!\\?\\s]+$', '', response)\n \n # 判断清理后的字符串是否以\"呜~\"结尾\n return clean_response.endswith(\"呜~\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否没有使用\"你\"字。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n # 如果回复中包含\"你\"字，则认为违反了禁止用\"你\"称呼用户的规定\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的每一个文件行数调整为需在120-180行之间","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义前端Web开发中常见的代码文件后缀名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n found_code_file = False\n \n # 遍历生成的工作区目录\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀并转换为小写\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件并获取其行数\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n line_count = len(lines)\n \n # 检查行数是否在120-180行之间\n if not (120 <= line_count <= 180):\n return False\n except Exception:\n # 遇到读取异常情况（如编码错误等），判定验证失败\n return False\n \n # 如果找到了至少一个代码文件且全都符合要求，则返回 True\n return found_code_file","tags":["Quantity","Range"]},{"constraint":"输出的代码文件里不允许出现行数超过30行的函数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区不存在，则视作没有产生违规代码，满足要求\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n \n def strip_comments_and_strings(code: str) -> str:\n \"\"\"\n 去除代码中的所有注释和字符串的内容（替换为空格），但保留换行符。\n 从而使得过滤后的文本和原始文本在行号结构上完全一致。\n \"\"\"\n result = []\n i = 0\n state = 'NORMAL'\n length = len(code)\n while i < length:\n c = code[i]\n if state == 'NORMAL':\n if c == '/' and i + 1 < length and code[i+1] == '/':\n state = 'LINE_COMMENT'\n result.extend([' ', ' '])\n i += 2\n continue\n elif c == '/' and i + 1 < length and code[i+1] == '*':\n state = 'BLOCK_COMMENT'\n result.extend([' ', ' '])\n i += 2\n continue\n elif c == '\"':\n state = 'DOUBLE_QUOTE'\n result.append('\"')\n elif c == \"'\":\n state = 'SINGLE_QUOTE'\n result.append(\"'\")\n elif c == '`':\n state = 'BACKTICK'\n result.append('`')\n else:\n result.append(c)\n i += 1\n elif state == 'LINE_COMMENT':\n if c == '\\n':\n state = 'NORMAL'\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'BLOCK_COMMENT':\n if c == '*' and i + 1 < length and code[i+1] == '/':\n state = 'NORMAL'\n result.extend([' ', ' '])\n i += 2\n continue\n elif c == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state in ('DOUBLE_QUOTE', 'SINGLE_QUOTE', 'BACKTICK'):\n # 应对字符串内部的反斜杠转义\n if c == '\\\\':\n if i + 1 < length and code[i+1] == '\\n':\n result.extend([' ', '\\n'])\n i += 2\n continue\n else:\n result.extend([' ', ' '])\n i += 2\n continue\n \n target = '\"' if state == 'DOUBLE_QUOTE' else (\"'\" if state == 'SINGLE_QUOTE' else '`')\n if c == target:\n state = 'NORMAL'\n result.append(target)\n elif c == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n\n # 支持扫描的主流前端/后端文件后缀\n target_extensions = ('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte', '.c', '.cpp', '.cs', '.java', '.h', '.hpp')\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见依赖包目录与版本控制目录\n if 'node_modules' in root.split(os.sep) or '.git' in root.split(os.sep):\n continue\n \n for file in files:\n filepath = os.path.join(root, file)\n if not os.path.isfile(filepath):\n continue\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 针对 Python 文件做 AST 解析验证\n if file.endswith('.py'):\n import ast\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n # 包含函数定义到结束的总行数\n if node.end_lineno - node.lineno + 1 > 30:\n return False\n except Exception:\n pass\n \n # 针对前端与类C语言通过启发式语法标记和括号匹配进行验证\n elif file.endswith(target_extensions):\n cleaned = strip_comments_and_strings(code)\n stack = []\n blocks = []\n \n # 记录所有大括号闭合位置\n for i, char in enumerate(cleaned):\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start_i = stack.pop()\n blocks.append((start_i, i))\n \n for start_i, end_i in blocks:\n prefix = \"\"\n # 向前寻找 `{` 之前的部分，止于上一个代码结构分界\n for j in range(start_i - 1, -1, -1):\n if cleaned[j] in ';}{':\n break\n prefix = cleaned[j] + prefix\n \n prefix = prefix.strip()\n if not prefix:\n continue\n \n is_function = False\n \n if '=>' in prefix:\n is_function = True\n elif 'function' in prefix:\n is_function = True\n elif prefix.endswith(')'):\n idx = prefix.rfind('(')\n if idx != -1:\n before_paren = prefix[:idx].strip()\n words = re.findall(r'[a-zA-Z0-9_]+', before_paren)\n if words:\n last_word = words[-1]\n # 排除常见的控制流关键字\n if last_word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n is_function = True\n else:\n is_function = True\n \n if is_function:\n # 查找这个函数的准确起始行号\n if 'function' in prefix:\n func_pos = prefix.find('function')\n start_line = cleaned[:start_i - len(prefix) + func_pos].count('\\n') + 1\n else:\n first_non_ws = len(prefix) - len(prefix.lstrip())\n start_line = cleaned[:start_i - len(prefix) + first_non_ws].count('\\n') + 1\n \n end_line = cleaned[:end_i].count('\\n') + 1\n \n if end_line - start_line + 1 > 30:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":2,"instruction":"现在来完善参数控制和数据监控。在右侧的“实验参数设置区”中，添加三个可交互的输入框，允许用户自定义：初始高度（范围 1-100 米）、重力加速度（允许修改，以便模拟月球等不同星体环境）、小球质量（千克）。同时，在“实时数据看板”中，请使用清晰的字体实时更新并显示小球在运动过程中的核心物理量：已用时间（秒，保留两位小数）、当前速度（米/秒）和离地高度（米）。要求：当用户在运行前拖动滑块或修改“初始高度”参数时，左侧实验舱里小球的垂直位置需要实时同步调整。\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过20行的函数；第二，只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复是否以\"汪~\"语气词开头。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，表示是否满足要求。\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容左侧可能的空白字符（包括空格、换行等），再判断是否以\"汪~\"开头\n return response.lstrip().startswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 检查回复末尾是否为“呜~”或“呜～”，允许其后跟随任意标点符号或空白字符\n is_match = bool(re.search(r'呜[~～][\\s\\W]*$', response))\n \n return is_match","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否禁止使用了“你”来称呼用户。\n 只要回复中出现了“你”字，即视为没有将其替换为“主人”，返回False。\n \"\"\"\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的每一个文件行数调整为需在120-180行之间","validation_code":"def check_file_line_count(response: str, workspace_path: str) -> bool:\n import os\n \n file_checked = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，防止读取不必要的系统或版本控制文件\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 UTF-8 编码读取文本文件\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n line_count = len(lines)\n \n # 判定当前文件的行数是否在120至180之间\n if line_count < 120 or line_count > 180:\n return False\n \n file_checked = True\n \n except UnicodeDecodeError:\n # 遇到非文本文件解码错误时直接跳过\n continue\n except Exception:\n # 其他异常同样跳过\n continue\n \n # 至少需要检查过一个有效的文件且全部符合要求\n return file_checked","tags":["Quantity","Range"]},{"constraint":"输出的代码文件里不允许出现行数超过20行的函数","validation_code":"import os\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def remove_comments_and_strings(js_code: str) -> str:\n NORMAL = 0\n IN_SINGLE_QUOTE = 1\n IN_DOUBLE_QUOTE = 2\n IN_BACKTICK = 3\n IN_LINE_COMMENT = 4\n IN_BLOCK_COMMENT = 5\n \n state = NORMAL\n clean_code = []\n i = 0\n n = len(js_code)\n \n while i < n:\n c = js_code[i]\n if state == NORMAL:\n if c == '/' and i + 1 < n and js_code[i+1] == '/':\n state = IN_LINE_COMMENT\n clean_code.append(' ')\n i += 1\n elif c == '/' and i + 1 < n and js_code[i+1] == '*':\n state = IN_BLOCK_COMMENT\n clean_code.append(' ')\n i += 1\n elif c == \"'\":\n state = IN_SINGLE_QUOTE\n clean_code.append(' ')\n elif c == '\"':\n state = IN_DOUBLE_QUOTE\n clean_code.append(' ')\n elif c == \"`\":\n state = IN_BACKTICK\n clean_code.append(' ')\n else:\n clean_code.append(c)\n elif state == IN_LINE_COMMENT:\n if c == '\\n':\n state = NORMAL\n clean_code.append(c)\n elif state == IN_BLOCK_COMMENT:\n if c == '*' and i + 1 < n and js_code[i+1] == '/':\n state = NORMAL\n i += 1\n elif c == '\\n':\n clean_code.append(c)\n elif state == IN_SINGLE_QUOTE:\n if c == '\\\\':\n if i + 1 < n and js_code[i+1] == '\\n':\n clean_code.append('\\n')\n i += 1\n elif c == \"'\":\n state = NORMAL\n elif c == '\\n':\n clean_code.append(c)\n elif state == IN_DOUBLE_QUOTE:\n if c == '\\\\':\n if i + 1 < n and js_code[i+1] == '\\n':\n clean_code.append('\\n')\n i += 1\n elif c == '\"':\n state = NORMAL\n elif c == '\\n':\n clean_code.append(c)\n elif state == IN_BACKTICK:\n if c == '\\\\':\n if i + 1 < n and js_code[i+1] == '\\n':\n clean_code.append('\\n')\n i += 1\n elif c == \"`\":\n state = NORMAL\n elif c == '\\n':\n clean_code.append(c)\n i += 1\n \n return \"\".join(clean_code)\n\n def get_js_function_blocks(js_code: str, offset_line: int = 0) -> list:\n clean_code = remove_comments_and_strings(js_code)\n \n line_starts = [0]\n for i, c in enumerate(clean_code):\n if c == '\\n':\n line_starts.append(i + 1)\n \n def get_line_num(idx):\n l = 0\n r = len(line_starts) - 1\n while l <= r:\n mid = (l + r) // 2\n if line_starts[mid] <= idx:\n if mid == len(line_starts) - 1 or line_starts[mid+1] > idx:\n return mid + 1\n else:\n l = mid + 1\n else:\n r = mid - 1\n return 1\n\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with', 'return', 'try', 'do'}\n blocks = []\n i = 0\n n = len(clean_code)\n \n while i < n:\n if clean_code[i] == '{':\n j = i - 1\n while j >= 0 and clean_code[j].isspace():\n j -= 1\n if j >= 0:\n stmt_start = j\n while stmt_start >= 0 and clean_code[stmt_start] not in {';', '{', '}'}:\n stmt_start -= 1\n stmt_start += 1\n pre_str = clean_code[stmt_start:j+1]\n \n is_func = False\n keyword_idx = -1\n \n temp_j = len(pre_str) - 1\n while temp_j >= 0 and pre_str[temp_j].isspace():\n temp_j -= 1\n \n if temp_j >= 1 and pre_str[temp_j-1:temp_j+1] == '=>':\n is_func = True\n k = temp_j - 2\n while k >= 0 and pre_str[k].isspace():\n k -= 1\n paren_count = 0\n while k >= 0:\n if pre_str[k] == ')':\n paren_count += 1\n elif pre_str[k] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n k -= 1\n if k >= 0:\n keyword_idx = k\n else:\n keyword_idx = temp_j - 1\n else:\n k = temp_j\n while k >= 0:\n if pre_str[k] == ')':\n break\n if pre_str[k] in {';', '{', '}'}:\n break\n k -= 1\n \n if k >= 0 and pre_str[k] == ')':\n paren_count = 0\n p = k\n while p >= 0:\n if pre_str[p] == ')':\n paren_count += 1\n elif pre_str[p] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n p -= 1\n \n if p >= 0:\n w = p - 1\n while w >= 0 and pre_str[w].isspace():\n w -= 1\n \n if w >= 0 and pre_str[w] == '>':\n angle_count = 0\n while w >= 0:\n if pre_str[w] == '>':\n angle_count += 1\n elif pre_str[w] == '<':\n angle_count -= 1\n if angle_count == 0:\n w -= 1\n break\n w -= 1\n while w >= 0 and pre_str[w].isspace():\n w -= 1\n \n word_end = w\n while w >= 0 and (pre_str[w].isalnum() or pre_str[w] in '_$'):\n w -= 1\n word = pre_str[w+1:word_end+1]\n \n if word == 'function':\n is_func = True\n keyword_idx = w + 1\n elif word not in keywords and word != '':\n w2 = w\n while w2 >= 0 and pre_str[w2].isspace():\n w2 -= 1\n word2_end = w2\n while w2 >= 0 and (pre_str[w2].isalnum() or pre_str[w2] in '_$'):\n w2 -= 1\n word2 = pre_str[w2+1:word2_end+1]\n if word2 == 'function':\n is_func = True\n keyword_idx = w2 + 1\n elif word not in keywords:\n is_func = True\n keyword_idx = w + 1\n elif word == '':\n w2 = w\n while w2 >= 0 and pre_str[w2].isspace():\n w2 -= 1\n word2_end = w2\n while w2 >= 0 and (pre_str[w2].isalnum() or pre_str[w2] in '_$'):\n w2 -= 1\n word2 = pre_str[w2+1:word2_end+1]\n if word2 == 'function':\n is_func = True\n keyword_idx = w2 + 1\n\n if is_func:\n real_start_idx = stmt_start + keyword_idx if keyword_idx != -1 else i\n start_line = get_line_num(real_start_idx)\n \n brace_count = 1\n k = i + 1\n while k < n:\n if clean_code[k] == '{':\n brace_count += 1\n elif clean_code[k] == '}':\n brace_count -= 1\n if brace_count == 0:\n break\n k += 1\n \n if k < n:\n end_line = get_line_num(k)\n blocks.append((start_line + offset_line, end_line + offset_line))\n i += 1\n return blocks\n\n def extract_script_tags(html_code: str) -> list:\n script_blocks = []\n lower_html = html_code.lower()\n idx = 0\n while True:\n start = lower_html.find('<script', idx)\n if start == -1:\n break\n end_start = lower_html.find('>', start)\n if end_start == -1:\n break\n end_tag = lower_html.find('</script>', end_start)\n if end_tag == -1:\n break\n script_blocks.append((end_start + 1, end_tag))\n idx = end_tag + 9\n return script_blocks\n\n def get_py_function_blocks(py_code: str) -> list:\n blocks = []\n try:\n tree = ast.parse(py_code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n end = getattr(node, 'end_lineno', node.lineno)\n blocks.append((start, end))\n except:\n pass\n return blocks\n\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte', '.py', '.cjs', '.mjs'}\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_extensions:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except:\n continue\n \n blocks = []\n if ext == '.py':\n blocks = get_py_function_blocks(code)\n elif ext in {'.html', '.htm', '.vue', '.svelte'}:\n scripts = extract_script_tags(code)\n for start, end in scripts:\n script_code = code[start:end]\n offset = code[:start].count('\\n')\n blocks.extend(get_js_function_blocks(script_code, offset))\n else:\n blocks = get_js_function_blocks(code, 0)\n \n for start_line, end_line in blocks:\n if end_line - start_line + 1 > 20:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":3,"instruction":"为了让物理模拟更真实且具备探究价值，下一步我们需要打破理想环境，引入“空气阻力”模型。请在参数设置区新增一个“材质与形状”的下拉菜单（选项包含：实心铁球、木球、带降落伞的沙袋）。选择不同的选项时，系统在后台自动赋予小球不同的风阻系数(C)和迎风面积(A)。在小球的下落运动演算中加入空气阻力公式（F = 0.5 * ρ * v² * C * A，空气密度ρ设为常数）。由于阻力与速度平方成正比，小球最终应能达到终端速度。另外，请在实时数据看板中新增一项，实时显示当前小球受到的空气阻力大小（牛顿）。\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过15行的函数；第二，生成的代码文件中必须仅用4个空格实现缩进。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复是否在最开头带有“汪~”语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除由于排版可能产生的开头空白字符后，判断是否以“汪~”开头\n return response.lstrip().startswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复文本末尾可能存在的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 判断清理后的回复是否以\"呜~\"结尾\n return cleaned_response.endswith(\"呜~\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 判断最后一轮回复中是否包含字符\"你\"\n # 如果包含，则说明违反了“禁止用'你'称呼用户”的要求\n if '你' in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的每一个文件行数调整为需在120-180行之间","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 校验工作区路径是否存在且为有效目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令（Web应用及progress.md文档）推断可能的目标文件类型\n target_extensions = {'.html', '.css', '.js', '.md', '.txt', '.json'}\n has_target_file = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n # 忽略所有隐藏目录（例如 .git, .vscode 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 忽略所有隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n line_count = len(lines)\n \n # 检查当前文件的行数是否在限定的 120-180 之间\n if line_count < 120 or line_count > 180:\n return False\n has_target_file = True\n except Exception:\n # 如果遇到文件读取异常（如编码错误或权限问题），跳过该文件\n continue\n \n # 只要存在至少一个目标文件且所有读取的文件都符合行数区间，即判定通过\n return has_target_file","tags":["Quantity","Range"]},{"constraint":"输出的代码文件里不允许出现行数超过15行的函数","validation_code":"import os\nimport re\nimport ast\n\ndef clean_code(code_str: str) -> str:\n \"\"\"清理 JavaScript / TypeScript 代码，将注释和字符串替换为空格以方便分析，保证行号一致。\"\"\"\n result = []\n i = 0\n length = len(code_str)\n in_single = False\n in_double = False\n in_backtick = False\n in_single_comment = False\n in_multi_comment = False\n \n while i < length:\n char = code_str[i]\n \n # 1. 单行注释处理\n if in_single_comment:\n if char == '\\n':\n in_single_comment = False\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n continue\n \n # 2. 多行注释处理\n if in_multi_comment:\n if char == '\\n':\n result.append('\\n')\n i += 1\n elif char == '*' and i + 1 < length and code_str[i+1] == '/':\n in_multi_comment = False\n result.append(' ')\n i += 2\n else:\n result.append(' ')\n i += 1\n continue\n \n # 3. 字符串（单引号、双引号、反引号）处理\n if in_single:\n if char == '\\n':\n result.append('\\n')\n i += 1\n elif char == '\\\\' and i + 1 < length:\n if code_str[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n elif char == \"'\":\n in_single = False\n result.append(' ')\n i += 1\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_double:\n if char == '\\n':\n result.append('\\n')\n i += 1\n elif char == '\\\\' and i + 1 < length:\n if code_str[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n elif char == '\"':\n in_double = False\n result.append(' ')\n i += 1\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_backtick:\n if char == '\\n':\n result.append('\\n')\n i += 1\n elif char == '\\\\' and i + 1 < length:\n if code_str[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n elif char == '`':\n in_backtick = False\n result.append(' ')\n i += 1\n else:\n result.append(' ')\n i += 1\n continue\n \n # 进入字符串与注释判断\n if char == \"'\":\n in_single = True\n result.append(' ')\n i += 1\n continue\n if char == '\"':\n in_double = True\n result.append(' ')\n i += 1\n continue\n if char == '`':\n in_backtick = True\n result.append(' ')\n i += 1\n continue\n if char == '/' and i + 1 < length:\n if code_str[i+1] == '/':\n in_single_comment = True\n result.append(' ')\n i += 2\n continue\n elif code_str[i+1] == '*':\n in_multi_comment = True\n result.append(' ')\n i += 2\n continue\n \n result.append(char)\n i += 1\n \n return \"\".join(result)\n\ndef check_functions_length_in_js(code_str: str) -> bool:\n \"\"\"提取JS中的大括号深度并分析函数长度。\"\"\"\n cleaned = clean_code(code_str)\n \n # 记录每行的起始索引以支持二分法查找行号\n line_starts = [0]\n for i, char in enumerate(cleaned):\n if char == '\\n':\n line_starts.append(i + 1)\n \n def get_line_number(index: int) -> int:\n left, right = 0, len(line_starts) - 1\n while left <= right:\n mid = (left + right) // 2\n if line_starts[mid] == index:\n return mid + 1\n elif line_starts[mid] < index:\n left = mid + 1\n else:\n right = mid - 1\n return right + 1\n\n brace_stack = []\n active_funcs = []\n method_pattern = re.compile(r'\\b(?!if|for|while|switch|catch)\\w+\\s*\\([^)]*\\)\\s*$')\n last_block_end = 0\n \n for i, char in enumerate(cleaned):\n if char in ('{', '}', ';'):\n if char == '{':\n raw_prefix = cleaned[last_block_end:i]\n is_func = False\n real_start_index = -1\n \n # 判断当前的大括号是否属于一个函数的开启\n func_matches = list(re.finditer(r'\\bfunction\\b', raw_prefix))\n if func_matches:\n is_func = True\n real_start_index = last_block_end + func_matches[-1].start()\n elif '=>' in raw_prefix:\n is_func = True\n real_start_index = last_block_end + raw_prefix.rfind('=>')\n else:\n match = method_pattern.search(raw_prefix)\n if match:\n is_func = True\n real_start_index = last_block_end + match.start()\n \n if is_func:\n start_line = get_line_number(real_start_index)\n active_funcs.append((len(brace_stack), start_line))\n \n brace_stack.append('{')\n elif char == '}':\n if brace_stack:\n brace_stack.pop()\n for j in range(len(active_funcs)-1, -1, -1):\n depth, start_line = active_funcs[j]\n if depth == len(brace_stack):\n # 函数闭合，计算行数\n end_line = get_line_number(i)\n func_lines = end_line - start_line + 1\n if func_lines > 15:\n return False\n active_funcs.pop(j)\n last_block_end = i + 1\n \n return True\n\ndef check_functions_length_in_html(html_str: str) -> bool:\n \"\"\"提取HTML内联的脚本区块并检查。\"\"\"\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', html_str, re.DOTALL | re.IGNORECASE)\n for script_code in scripts:\n if script_code.strip() and not check_functions_length_in_js(script_code):\n return False\n return True\n\ndef check_functions_length_in_py(code_str: str) -> bool:\n \"\"\"借助AST直接解析并校验Python代码文件。\"\"\"\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n end = node.end_lineno\n if end is not None and (end - start + 1) > 15:\n return False\n except Exception:\n pass\n return True\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"验证工作区中的所有代码文件内的函数，是否有任何函数行数超过15行。\"\"\"\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_functions_length_in_py(content):\n return False\n elif ext in ('.js', '.ts', '.jsx', '.tsx'):\n if not check_functions_length_in_js(content):\n return False\n elif ext in ('.html', '.htm', '.vue', '.svelte'):\n if not check_functions_length_in_html(content):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"生成的代码文件中必须仅用4个空格实现缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件扩展名集合\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n in_c_comment = False\n in_html_comment = False\n \n for line in lines:\n # 忽略空行\n if not line.strip():\n continue\n \n # 提取该行的前导空白字符\n leading_spaces = \"\"\n for char in line:\n if char in (' ', '\\t'):\n leading_spaces += char\n else:\n break\n \n # 若使用制表符（Tab）进行缩进，判定为不符合要求\n if '\\t' in leading_spaces:\n return False\n \n # 若该行无任何缩进，则继续检查下一行\n if not leading_spaces:\n continue\n \n stripped_line = line.strip()\n \n # 过滤常见的块注释（C风格），避免注释内部随意的空格对齐被误判\n if stripped_line.startswith('/*'):\n in_c_comment = True\n if in_c_comment:\n if '*/' in stripped_line:\n in_c_comment = False\n continue\n if stripped_line.startswith('*'): # 用于兼容多行注释中间行的前导星号\n continue\n \n # 过滤HTML块注释\n if stripped_line.startswith('<!--'):\n in_html_comment = True\n if in_html_comment:\n if '-->' in stripped_line:\n in_html_comment = False\n continue\n \n # 判断前导空格的长度是否严格为4的倍数\n if len(leading_spaces) % 4 != 0:\n return False\n\n # 若至少遍历到一个代码文件且所有文件缩进均合规，则返回 True\n return has_code_file","tags":["Style","Layout"]}]},{"round_id":4,"instruction":"做得不错！现在我想增强工具的视觉表现和横向对比能力。请在实验舱增加“频闪打点”功能：小球下落时，每隔 0.2 秒在它当前所在位置留下一个半透明的残影（轨迹点），并在残影旁边用微小字体标注出那一刻的时间戳。此外，请将系统升级为“双球对比”模式：在实验舱里再放入第二个不同颜色的小球（球B），控制面板里的参数区进行左右分栏，允许用户分别对球A和球B独立设置不同的材质、质量或初始高度。点击开始时，双球在同一个场景下同时下落，让用户可以直观观察到空气阻力或质量差异对轨迹的影响。\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过12行的函数；第二，所有由你创建的代码文件，其编码必须统一为UTF-8。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查回复的最开头是否带有\"汪~\"语气词。\n \"\"\"\n return response.lstrip().startswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n if not response:\n return False\n \n # 使用正则表达式匹配回复末尾是否为\"呜~\"，允许其后有空格、换行符以及常见的句末标点符号\n pattern = r'呜~[\\s。！!\\.?？]*$'\n match = re.search(pattern, response)\n \n return match is not None","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复中是否禁止使用了“你”或“您”来称呼用户。\n \"\"\"\n if \"你\" in response or \"您\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的每一个文件行数调整为需在120-180行之间","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否有效\n if not os.path.isdir(workspace_path):\n return False\n \n has_valid_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，如 .git, .vscode 等，避免误将非生成的环境文件计入\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n # 尝试作为文本文件读取内容并计算行数\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n has_valid_file = True\n line_count = len(lines)\n \n # 检查文件的行数是否在 [120, 180] 范围内\n if line_count < 120 or line_count > 180:\n return False\n except UnicodeDecodeError:\n # 无法用 UTF-8 读取的文件当做二进制/媒体等非代码文件跳过\n continue\n except Exception:\n # 其他读取异常跳过\n continue\n \n # 如果找到了合法的代码文件并且全部通过检查，则返回 True\n return has_valid_file","tags":["Quantity","Range"]},{"constraint":"输出的代码文件里不允许出现行数超过12行的函数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def get_python_function_lines(code):\n try:\n tree = ast.parse(code)\n except Exception:\n return []\n lengths = []\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n lengths.append(node.end_lineno - node.lineno + 1)\n return lengths\n\n def extract_js_functions(code):\n # 稳健剔除 JS 中的字符串（包括模板字符串）和各类注释，同时保留空行以对齐绝对行号\n pattern = re.compile(r'/\\*.*?\\*/|//.*|\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\'|`(?:[^`\\\\]|\\\\.)*`', re.DOTALL)\n def replacer(match):\n m = match.group(0)\n if m.startswith('/*'): \n return '\\n' * m.count('\\n')\n elif m.startswith('//'): \n return ''\n else: \n return '\"\"' + '\\n' * m.count('\\n')\n\n clean_code = pattern.sub(replacer, code)\n \n line_num = 1\n char_lines = []\n for char in clean_code:\n char_lines.append(line_num)\n if char == '\\n':\n line_num += 1\n \n tokens = []\n token_pattern = re.compile(r'[a-zA-Z_$][a-zA-Z0-9_$]*|=>|[{}()\\[\\];=,.]|\\s+')\n pos = 0\n while pos < len(clean_code):\n m = token_pattern.match(clean_code, pos)\n if m:\n tok = m.group(0)\n if not tok.isspace():\n tokens.append({'val': tok, 'pos': pos, 'line': char_lines[pos]})\n pos = m.end()\n else:\n pos += 1\n\n function_starts = []\n for i, tok in enumerate(tokens):\n if tok['val'] == '{':\n is_func = False\n # 检查箭头函数： => {\n if i >= 1 and tokens[i-1]['val'] == '=>':\n is_func = True\n # 检查标准/匿名/类成员函数：) {\n elif i >= 1 and tokens[i-1]['val'] == ')':\n paren_count = 1\n j = i - 2\n while j >= 0 and paren_count > 0:\n if tokens[j]['val'] == ')':\n paren_count += 1\n elif tokens[j]['val'] == '(':\n paren_count -= 1\n j -= 1\n if paren_count == 0:\n if j >= 0:\n prev_tok = tokens[j]['val']\n # 剔除常见的控制流关键字\n if prev_tok not in ['if', 'for', 'while', 'catch', 'switch', 'with']:\n # 剔除类的继承特征（如 class A extends B(...) { ）\n k = j - 1\n is_class = False\n while k >= 0:\n if tokens[k]['val'] in ['{', '}', ';']:\n break\n if tokens[k]['val'] == 'class':\n is_class = True\n break\n k -= 1\n if not is_class:\n is_func = True\n else:\n is_func = True\n \n if is_func:\n function_starts.append(i)\n\n function_lines = []\n for start_idx in function_starts:\n brace_count = 1\n start_line = tokens[start_idx]['line']\n end_line = start_line\n for j in range(start_idx + 1, len(tokens)):\n if tokens[j]['val'] == '{':\n brace_count += 1\n elif tokens[j]['val'] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_line = tokens[j]['line']\n break\n else:\n # 若未闭合，则认为跨度直至文件末尾\n if len(tokens) > start_idx:\n end_line = tokens[-1]['line']\n \n function_lines.append(end_line - start_line + 1)\n \n return function_lines\n\n def check_file(file_path):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 读写异常、无法解析的非文本文件则跳过\n return True\n\n ext = os.path.splitext(file_path)[1].lower()\n \n if ext == '.py':\n lines = get_python_function_lines(content)\n for l in lines:\n if l > 12:\n return False\n \n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\n lines = extract_js_functions(content)\n for l in lines:\n if l > 12:\n return False\n \n elif ext in ['.html', '.htm', '.vue', '.svelte']:\n # 提取文件内嵌入的 <script>...</script> 代码块\n script_pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\n scripts = script_pattern.findall(content)\n for script in scripts:\n lines = extract_js_functions(script)\n for l in lines:\n if l > 12:\n return False\n \n return True\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略构建和版本控制目录，优化解析效率\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n file_path = os.path.join(root, file)\n # 仅校验特定的代码文件后缀，如若其中有函数长于12行立刻失败\n ext = os.path.splitext(file_path)[1].lower()\n if ext in ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.svelte']:\n if not check_file(file_path):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"生成的代码文件中必须仅用4个空格实现缩进","validation_code":"import os\n\ndef check_indentation(response: str, workspace_path: str) -> bool:\n # 针对基于 Web 的应用及常见的相关后端语言，定义需要检查的代码文件后缀\n code_extensions = {'.html', '.js', '.css', '.ts', '.py', '.java', '.cpp', '.c', '.cs'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果文件无法按 UTF-8 读取则跳过\n continue\n \n for line in lines:\n # 忽略空行或仅包含空白字符的行\n if not line.strip():\n continue\n \n # 提取该行的前导空白字符\n leading_space = \"\"\n for ch in line:\n if ch in (' ', '\\t'):\n leading_space += ch\n else:\n break\n \n # 如果行首没有空白字符，说明未缩进，属于正常现象，直接跳过\n if not leading_space:\n continue\n \n # 规则1：必须仅用空格实现缩进，包含制表符即违规\n if '\\t' in leading_space:\n return False\n \n # 规则2：使用的空格数必须是4的倍数\n if len(leading_space) % 4 != 0:\n stripped_line = line.lstrip()\n # 特殊容错：忽略以 '*' 开头的多行注释行（常见于 JS / CSS / Java 等）\n # 此类注释的对齐通常会引入额外的空格，属于正常代码规范\n if stripped_line.startswith('*'):\n continue\n \n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查代码项目路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令推断的可能文件类型：Web前端文件 + progress.md\n target_extensions = {'.html', '.css', '.js', '.md'}\n has_target_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录（如 .git）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_target_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制模式读取文件内容\n with open(file_path, 'rb') as f:\n content = f.read()\n \n # 尝试用 utf-8 解码，若文件中包含非UTF-8字节序列（如GBK编码的中文字符），将抛出异常\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 解码失败，说明不是合法的UTF-8编码\n return False\n except Exception:\n # 发生其他文件读取错误，视为未满足要求\n return False\n \n # 如果没有找到任何目标代码文件，无法证明满足要求，返回False；否则返回True\n return has_target_file","tags":["Language","File Encoding"]}]},{"round_id":5,"instruction":"最后一轮，我们加入能量可视化分析和实验导出功能来提升工具的专业度。请在页面底部新增一个横向展开的面板，使用原生 DOM 元素的高度动态变化来模拟柱状图。为每个小球提供三根并排的动态柱子，实时显示其：动能(Ek)、重力势能(Ep)和总机械能(E)。用户可以直观看到由于空气阻力和地面反弹做功导致的机械能衰减过程。最后，在控制面板最下方增加一个“导出实验报告”按钮，点击后能将整个下落过程（以 0.1 秒为采样间隔）中两个球的时间、高度、速度和总机械能数据拼接成一份完整的 CSV 文件，并自动触发浏览器下载。\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过10行的函数；第二，请确保在改动代码前使用 TodoWrite 工具规划任务。对于任意任务，只要其相关的代码修改执行完毕，下一个工具调用则强制要求为 TodoComplete 以标记任务完结，期间不可穿插其他任务的代码修改。","instruction_following_checklist":[{"constraint":"无论回复什么内容，回复的最开头都必须带有\"汪~\"语气词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n # 去除回复内容最前面的空白字符（包括空格、换行符等），确保准确判断首字符\n cleaned_response = response.lstrip()\n \n # 判断是否以“汪~”或全角波浪号的“汪～”开头\n if cleaned_response.startswith(\"汪~\") or cleaned_response.startswith(\"汪～\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"你的所有回复都必须在结尾处加上\"呜~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复末尾的空白字符以及可能存在的常见标点符号，防止误判\n cleaned_response = response.rstrip(' \\n\\r\\t。！!.')\n \n # 判断清理后的回复是否以“呜~”结尾\n return cleaned_response.endswith(\"呜~\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"主人\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查AI的回复中是否不包含“你”字，以确保其没有使用“你”称呼用户。\n \"\"\"\n # 如果回复中存在“你”，则说明未满足替换要求\n if \"你\" in response:\n return False\n \n # 未出现“你”则视为满足要求\n return True","tags":["Content","Persona"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的每一个文件行数调整为需在120-180行之间","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目性质推断可能生成的代码及文档文件后缀\n valid_exts = {'.html', '.css', '.js', '.md'}\n file_checked = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和可能存在的常见的依赖项目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'vendor')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n line_count = len(lines)\n # 检查当前文件行数是否落在120-180范围内\n if line_count < 120 or line_count > 180:\n return False\n \n file_checked = True\n except Exception:\n # 遇到非文本或者读取异常的情况直接跳过\n continue\n\n # 至少需要存在被成功检查的符合类型的文件，才能判定满足该要求\n return file_checked","tags":["Quantity","Range"]},{"constraint":"输出的代码文件里不允许出现行数超过10行的函数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n \n def check_py_functions(code_str: str) -> bool:\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n if node.end_lineno - node.lineno + 1 > 10:\n return False\n return True\n except Exception:\n return True\n \n def check_js_functions(code_str: str) -> bool:\n # 匹配字符串、多行注释和单行注释，用于剔除这些会干扰大括号嵌套解析的内容\n pattern = r'(?:\"(?:\\\\.|[^\"\\\\])*\")|(?:\\'(?:\\\\.|[^\\'\\\\])*\\')|(?:`(?:\\\\.|[^`\\\\])*`)|(?:/\\*.*?\\*/)|(?://[^\\n]*)'\n \n def replacer(match):\n s = match.group(0)\n # 保证原本的换行符不丢失，只将非换行字符替换为空格，以确保行号和索引保持不变\n return ''.join('\\n' if c == '\\n' else ' ' for c in s)\n \n try:\n cleaned = re.sub(pattern, replacer, code_str, flags=re.DOTALL)\n except Exception:\n cleaned = code_str\n \n function_starts = set()\n \n # 1. 寻找普通函数: function 关键字，到 { 之间不出现 ; 的内容\n for m in re.finditer(r'\\bfunction\\b[^{;]*\\{', cleaned):\n brace_idx = m.end() - 1\n function_starts.add(brace_idx)\n \n # 2. 寻找箭头函数: => {\n for m in re.finditer(r'=>\\s*\\{', cleaned):\n brace_idx = m.end() - 1\n function_starts.add(brace_idx)\n \n # 3. 寻找类/对象内的方法简写形式: name(args) { \n keywords = {\"if\", \"for\", \"while\", \"switch\", \"catch\", \"function\", \"return\", \"with\"}\n for m in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^;{()]*\\)\\s*\\{', cleaned):\n name = m.group(1)\n if name not in keywords:\n brace_idx = m.end() - 1\n function_starts.add(brace_idx)\n \n # 验证每一个函数块的行数\n for brace_idx in function_starts:\n depth = 1\n i = brace_idx + 1\n while i < len(cleaned):\n if cleaned[i] == '{':\n depth += 1\n elif cleaned[i] == '}':\n depth -= 1\n if depth == 0:\n start_line = cleaned.count('\\n', 0, brace_idx) + 1\n end_line = cleaned.count('\\n', 0, i) + 1\n if end_line - start_line + 1 > 10:\n return False\n break\n i += 1\n return True\n\n def check_html_functions(code_str: str) -> bool:\n # 提取所有的 <script> 标签中的内容进行校验\n pattern = r'<script\\b[^>]*>(.*?)</script>'\n for m in re.finditer(pattern, code_str, flags=re.DOTALL | re.IGNORECASE):\n script_content = m.group(1)\n start_pos = m.start(1)\n prefix = code_str[:start_pos]\n # 为了维持行号准确，将提取出来的JS内容前填充对应数量的换行符\n newlines = '\\n' * prefix.count('\\n')\n padded_script = newlines + script_content\n if not check_js_functions(padded_script):\n return False\n return True\n\n # 遍历项目工作区中生成的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_str = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_py_functions(code_str):\n return False\n elif ext in ['.html', '.htm', '.vue']:\n if not check_html_functions(code_str):\n return False\n else:\n if not check_js_functions(code_str):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"生成的代码文件中必须仅用4个空格实现缩进","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 预估可能出现的常见代码文件扩展名\n code_exts = {\n '.py', '.js', '.ts', '.html', '.css', '.vue', '.jsx', \n '.tsx', '.json', '.java', '.c', '.cpp', '.h', '.cs', \n '.go', '.rb', '.php', '.swift', '.kt'\n }\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果因权限或非UTF-8编码导致无法读取，安全跳过\n continue\n \n indent_counts = []\n for line in lines:\n # 仅匹配行首的空白字符，且要求行内紧接着存在非空白字符\n match = re.match(r'^([ \\t]+)(?=\\S)', line)\n if match:\n indent = match.group(1)\n content = line[len(indent):]\n \n # 1. 严格检查：只要前导空白中包含 tab 制表符，即判定不合规\n if '\\t' in indent:\n return False\n \n # 2. 排除多行注释中常见的星号对齐行带来的统计干扰\n if content.startswith('*'):\n continue\n \n indent_counts.append(len(indent))\n \n # 若文件内实际包含有效缩进记录，则分析其空格分布\n if indent_counts:\n # 统计缩进空格数量恰为 4 的倍数的行数\n mod_4_count = sum(1 for count in indent_counts if count % 4 == 0)\n \n # 兼容部分特殊场景（如长参数折行对齐等），要求 70% 满足条件即认为合规。\n # 如果比例过低，则很大可能使用的是 2 个空格或杂乱规范的缩进。\n if mod_4_count / len(indent_counts) < 0.7:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，说明没有创建文件，默认未违背规则\n if not os.path.exists(workspace_path):\n return True\n \n # 根据用户指令，推测Web实验工具项目可能的代码文件类型\n code_extensions = {'.html', '.htm', '.css', '.js', '.md', '.json'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制模式读取文件\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 验证文件是否可以通过 utf-8 解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明包含非 UTF-8 编码的内容\n return False\n except Exception:\n # 忽略其他可能导致无法正常读取的系统层面错误\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"请确保在改动代码前使用 TodoWrite 工具规划任务。对于任意任务，只要其相关的代码修改执行完毕，下一个工具调用则强制要求为 TodoComplete 以标记任务完结，期间不可穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我开发一个基于Web的“自由落体物理模拟实验工具”。第一步，请先搭建好基础的页面骨架。需求如下：1. 页面分为左右两栏布局，左侧是占据主要屏幕空间的“实验舱”，请使用 Canvas 或带边框的大块 DOM 区域，底部画一条粗线代表“地面”，实验舱靠上的位置放置一个静止的小球作为实验对象。2. 右侧是“控制面板”，包含三个结构清晰的卡片占位：实验参数设置区、播放控制区（放置开始、暂停、重置三个按钮，暂不实现功能）、实时数据看板。3. 整体界面风格保持现代、极简。目前阶段不需要编写任何物理运动计算逻辑，只要把静态排版和基本UI元素渲染出来即可。\\n另外，请你在后续所有轮次的开发过程中都严格遵循以下要求：第一，生成的每一个文件行数均在100-200行之间；第二，输出的代码文件里不允许出现行数超过30行的函数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除回复首部的空白字符后，判断是否以\\\"汪~\\\"开头\\n return response.lstrip().startswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 匹配末尾的\\\"呜~\\\"（兼容半角和全角波浪号），允许其后跟随空白字符及常见标点符号\\n pattern = r'呜[~～][\\\\s。，！、？；：\\\\.,!;:\\\\'\\\\\\\"”’]*$'\\n \\n return bool(re.search(pattern, response))\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复内容中是否包含\\\"你\\\"字\\n if \\\"你\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的每一个文件行数均在100-200行之间\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n file_count = 0\\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 忽略隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n # 根据用户指令推断，Web项目代码文件通常为 .html, .css, .js\\n if not file.endswith(('.html', '.css', '.js')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n file_count += 1\\n \\n # 判断文件行数是否在100至200之间\\n if len(lines) < 100 or len(lines) > 200:\\n return False\\n except Exception:\\n # 读取异常可能因为非纯文本或是编码问题，此处作为无效生成情况返回False\\n return False\\n \\n # 必须保证至少成功检查了一个以上的目标文件\\n return file_count > 0\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"输出的代码文件里不允许出现行数超过30行的函数\",\n \"validation_code\": \"def check_function_lines_limit(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def clean_code(code: str) -> str:\\n \\\"\\\"\\\"净化代码，将注释、字符串内容替换为空格，但严格保留所有换行符以保持行号对应\\\"\\\"\\\"\\n state = 'NORMAL'\\n result = []\\n i = 0\\n n = len(code)\\n while i < n:\\n char = code[i]\\n if state == 'NORMAL':\\n if char == '/' and i + 1 < n and code[i+1] == '/':\\n state = 'LINE_COMMENT'\\n result.append(' ')\\n i += 1\\n elif char == '/' and i + 1 < n and code[i+1] == '*':\\n state = 'BLOCK_COMMENT'\\n result.append(' ')\\n i += 1\\n elif char == '\\\"':\\n state = 'STRING_DOUBLE'\\n result.append(char)\\n elif char == \\\"'\\\":\\n state = 'STRING_SINGLE'\\n result.append(char)\\n elif char == '`':\\n state = 'STRING_BACKTICK'\\n result.append(char)\\n else:\\n result.append(char)\\n elif state == 'LINE_COMMENT':\\n if char == '\\\\n':\\n state = 'NORMAL'\\n result.append(char)\\n else:\\n result.append(' ')\\n elif state == 'BLOCK_COMMENT':\\n if char == '*' and i + 1 < n and code[i+1] == '/':\\n state = 'NORMAL'\\n result.append(' ')\\n i += 1\\n elif char == '\\\\n':\\n result.append(char)\\n else:\\n result.append(' ')\\n elif state == 'STRING_DOUBLE':\\n if char == '\\\\\\\\':\\n result.append(' ')\\n i += 1\\n if i < n:\\n result.append(code[i] if code[i] == '\\\\n' else ' ')\\n elif char == '\\\"':\\n state = 'NORMAL'\\n result.append(char)\\n elif char == '\\\\n':\\n result.append(char)\\n else:\\n result.append(' ')\\n elif state == 'STRING_SINGLE':\\n if char == '\\\\\\\\':\\n result.append(' ')\\n i += 1\\n if i < n:\\n result.append(code[i] if code[i] == '\\\\n' else ' ')\\n elif char == \\\"'\\\":\\n state = 'NORMAL'\\n result.append(char)\\n elif char == '\\\\n':\\n result.append(char)\\n else:\\n result.append(' ')\\n elif state == 'STRING_BACKTICK':\\n if char == '\\\\\\\\':\\n result.append(' ')\\n i += 1\\n if i < n:\\n result.append(code[i] if code[i] == '\\\\n' else ' ')\\n elif char == '`':\\n state = 'NORMAL'\\n result.append(char)\\n elif char == '\\\\n':\\n result.append(char)\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def check_if_function_prefix(prefix: str) -> bool:\\n \\\"\\\"\\\"检查 '{' 前方的代码前缀，判断该括号是否属于一个函数的开头部位\\\"\\\"\\\"\\n prefix = prefix.rstrip()\\n if not prefix:\\n return False\\n \\n # 匹配箭头函数特征\\n if prefix.endswith('=>'):\\n return True\\n \\n i = len(prefix) - 1\\n while i >= 0 and prefix[i].isspace():\\n i -= 1\\n \\n if i < 0:\\n return False\\n \\n # 应对带有返回类型注解的 TS 语法 (例如 : void { ) 寻找上一层 ')'\\n if prefix[i] != ')':\\n valid_type_chars = set(\\\"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_$:<>,[]|&. \\\\t\\\\n\\\")\\n found_colon = False\\n while i >= 0:\\n if prefix[i] == ':':\\n found_colon = True\\n i -= 1\\n break\\n if prefix[i] not in valid_type_chars:\\n return False\\n i -= 1\\n \\n if not found_colon:\\n return False\\n \\n while i >= 0 and prefix[i].isspace():\\n i -= 1\\n if i < 0 or prefix[i] != ')':\\n return False\\n\\n # 进行括号向左匹配寻找对应的 '('\\n paren_count = 1\\n i -= 1\\n while i >= 0:\\n if prefix[i] == ')':\\n paren_count += 1\\n elif prefix[i] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n i -= 1\\n \\n if paren_count == 0 and i >= 0:\\n j = i - 1\\n while j >= 0 and prefix[j].isspace():\\n j -= 1\\n \\n token_end = j\\n while j >= 0 and (prefix[j].isalnum() or prefix[j] in '_$'):\\n j -= 1\\n token_start = j + 1\\n token = prefix[token_start:token_end+1]\\n \\n # 排除控制流和特定关键字\\n control_keywords = {'if', 'for', 'while', 'catch', 'switch', 'with'}\\n if token in control_keywords:\\n return False\\n return True\\n\\n return False\\n\\n def check_jsts_content(content: str) -> bool:\\n \\\"\\\"\\\"解析并检查 JS/TS 等类文件的内容行数是否超标\\\"\\\"\\\"\\n cleaned = clean_code(content)\\n line_num = 1\\n stack = [] \\n \\n for idx, char in enumerate(cleaned):\\n if char == '\\\\n':\\n line_num += 1\\n elif char == '{':\\n start_prefix = max(0, idx - 500)\\n prefix = cleaned[start_prefix:idx]\\n is_func = check_if_function_prefix(prefix)\\n stack.append((line_num, is_func))\\n elif char == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n # 检查总行数（结束行-开始行+1）是否超过 30 行\\n if line_num - start_line + 1 > 30:\\n return False\\n return True\\n\\n def check_python_file(content: str) -> bool:\\n \\\"\\\"\\\"使用 ast 模块解析并检查 Python 文件行数是否超标\\\"\\\"\\\"\\n try:\\n tree = ast.parse(content)\\n except:\\n return True\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n if node.end_lineno - node.lineno + 1 > 30:\\n return False\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except:\\n continue\\n \\n if file.endswith('.py'):\\n if not check_python_file(content):\\n return False\\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.mjs', '.cjs')):\\n if not check_jsts_content(content):\\n return False\\n elif file.endswith(('.html', '.htm', '.vue')):\\n # 针对 HTML/Vue 文件内嵌脚本的处理\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script_content in scripts:\\n if not check_jsts_content(script_content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"很好，骨架搭出来了。接下来我们实现核心的物理引擎和基础控制逻辑。请为左侧实验舱建立一个“像素与物理现实”的映射关系（例如设定 50 像素等于 1 米，并在界面角落标出比例尺）。给小球赋予固定的重力加速度（g=9.8m/s²）。实现控制面板里的“开始”、“暂停”和“重置”按钮逻辑：点击开始后，小球开始受重力影响做加速下落运动；触碰到地面时，小球绝对不能穿模，并且需要以 0.7 的弹性系数向上反弹，反复弹跳直到速度趋近于零最终静止；重置按钮能清空状态并让小球回到页面顶部的初始位置。另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，生成的每一个文件行数调整为需在120-180行之间；第二，请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复最开头是否带有\\\"汪~\\\"语气词。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，代表该要求是否满足。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 移除可能存在的前导空白字符后，判断是否以\\\"汪~\\\"开头\\n cleaned_response = response.lstrip()\\n return cleaned_response.startswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import re\\n if not response:\\n return False\\n \\n # 将回复内容末尾的空白字符以及常见的标点符号去除，以防止误判\\n clean_response = re.sub(r'[。，！？\\\\.\\\\!\\\\?\\\\s]+$', '', response)\\n \\n # 判断清理后的字符串是否以\\\"呜~\\\"结尾\\n return clean_response.endswith(\\\"呜~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否没有使用\\\"你\\\"字。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n # 如果回复中包含\\\"你\\\"字，则认为违反了禁止用\\\"你\\\"称呼用户的规定\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的每一个文件行数调整为需在120-180行之间\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义前端Web开发中常见的代码文件后缀名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n found_code_file = False\\n \\n # 遍历生成的工作区目录\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀并转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件并获取其行数\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n line_count = len(lines)\\n \\n # 检查行数是否在120-180行之间\\n if not (120 <= line_count <= 180):\\n return False\\n except Exception:\\n # 遇到读取异常情况（如编码错误等），判定验证失败\\n return False\\n \\n # 如果找到了至少一个代码文件且全都符合要求，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"输出的代码文件里不允许出现行数超过30行的函数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区不存在，则视作没有产生违规代码，满足要求\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n \\n def strip_comments_and_strings(code: str) -> str:\\n \\\"\\\"\\\"\\n 去除代码中的所有注释和字符串的内容（替换为空格），但保留换行符。\\n 从而使得过滤后的文本和原始文本在行号结构上完全一致。\\n \\\"\\\"\\\"\\n result = []\\n i = 0\\n state = 'NORMAL'\\n length = len(code)\\n while i < length:\\n c = code[i]\\n if state == 'NORMAL':\\n if c == '/' and i + 1 < length and code[i+1] == '/':\\n state = 'LINE_COMMENT'\\n result.extend([' ', ' '])\\n i += 2\\n continue\\n elif c == '/' and i + 1 < length and code[i+1] == '*':\\n state = 'BLOCK_COMMENT'\\n result.extend([' ', ' '])\\n i += 2\\n continue\\n elif c == '\\\"':\\n state = 'DOUBLE_QUOTE'\\n result.append('\\\"')\\n elif c == \\\"'\\\":\\n state = 'SINGLE_QUOTE'\\n result.append(\\\"'\\\")\\n elif c == '`':\\n state = 'BACKTICK'\\n result.append('`')\\n else:\\n result.append(c)\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if c == '\\\\n':\\n state = 'NORMAL'\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if c == '*' and i + 1 < length and code[i+1] == '/':\\n state = 'NORMAL'\\n result.extend([' ', ' '])\\n i += 2\\n continue\\n elif c == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state in ('DOUBLE_QUOTE', 'SINGLE_QUOTE', 'BACKTICK'):\\n # 应对字符串内部的反斜杠转义\\n if c == '\\\\\\\\':\\n if i + 1 < length and code[i+1] == '\\\\n':\\n result.extend([' ', '\\\\n'])\\n i += 2\\n continue\\n else:\\n result.extend([' ', ' '])\\n i += 2\\n continue\\n \\n target = '\\\"' if state == 'DOUBLE_QUOTE' else (\\\"'\\\" if state == 'SINGLE_QUOTE' else '`')\\n if c == target:\\n state = 'NORMAL'\\n result.append(target)\\n elif c == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n\\n # 支持扫描的主流前端/后端文件后缀\\n target_extensions = ('.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte', '.c', '.cpp', '.cs', '.java', '.h', '.hpp')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见依赖包目录与版本控制目录\\n if 'node_modules' in root.split(os.sep) or '.git' in root.split(os.sep):\\n continue\\n \\n for file in files:\\n filepath = os.path.join(root, file)\\n if not os.path.isfile(filepath):\\n continue\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 针对 Python 文件做 AST 解析验证\\n if file.endswith('.py'):\\n import ast\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n # 包含函数定义到结束的总行数\\n if node.end_lineno - node.lineno + 1 > 30:\\n return False\\n except Exception:\\n pass\\n \\n # 针对前端与类C语言通过启发式语法标记和括号匹配进行验证\\n elif file.endswith(target_extensions):\\n cleaned = strip_comments_and_strings(code)\\n stack = []\\n blocks = []\\n \\n # 记录所有大括号闭合位置\\n for i, char in enumerate(cleaned):\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start_i = stack.pop()\\n blocks.append((start_i, i))\\n \\n for start_i, end_i in blocks:\\n prefix = \\\"\\\"\\n # 向前寻找 `{` 之前的部分，止于上一个代码结构分界\\n for j in range(start_i - 1, -1, -1):\\n if cleaned[j] in ';}{':\\n break\\n prefix = cleaned[j] + prefix\\n \\n prefix = prefix.strip()\\n if not prefix:\\n continue\\n \\n is_function = False\\n \\n if '=>' in prefix:\\n is_function = True\\n elif 'function' in prefix:\\n is_function = True\\n elif prefix.endswith(')'):\\n idx = prefix.rfind('(')\\n if idx != -1:\\n before_paren = prefix[:idx].strip()\\n words = re.findall(r'[a-zA-Z0-9_]+', before_paren)\\n if words:\\n last_word = words[-1]\\n # 排除常见的控制流关键字\\n if last_word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n is_function = True\\n else:\\n is_function = True\\n \\n if is_function:\\n # 查找这个函数的准确起始行号\\n if 'function' in prefix:\\n func_pos = prefix.find('function')\\n start_line = cleaned[:start_i - len(prefix) + func_pos].count('\\\\n') + 1\\n else:\\n first_non_ws = len(prefix) - len(prefix.lstrip())\\n start_line = cleaned[:start_i - len(prefix) + first_non_ws].count('\\\\n') + 1\\n \\n end_line = cleaned[:end_i].count('\\\\n') + 1\\n \\n if end_line - start_line + 1 > 30:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来完善参数控制和数据监控。在右侧的“实验参数设置区”中，添加三个可交互的输入框，允许用户自定义：初始高度（范围 1-100 米）、重力加速度（允许修改，以便模拟月球等不同星体环境）、小球质量（千克）。同时，在“实时数据看板”中，请使用清晰的字体实时更新并显示小球在运动过程中的核心物理量：已用时间（秒，保留两位小数）、当前速度（米/秒）和离地高度（米）。要求：当用户在运行前拖动滑块或修改“初始高度”参数时，左侧实验舱里小球的垂直位置需要实时同步调整。\\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过20行的函数；第二，只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复是否以\\\"汪~\\\"语气词开头。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，表示是否满足要求。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容左侧可能的空白字符（包括空格、换行等），再判断是否以\\\"汪~\\\"开头\\n return response.lstrip().startswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 检查回复末尾是否为“呜~”或“呜～”，允许其后跟随任意标点符号或空白字符\\n is_match = bool(re.search(r'呜[~～][\\\\s\\\\W]*$', response))\\n \\n return is_match\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否禁止使用了“你”来称呼用户。\\n 只要回复中出现了“你”字，即视为没有将其替换为“主人”，返回False。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的每一个文件行数调整为需在120-180行之间\",\n \"validation_code\": \"def check_file_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n \\n file_checked = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，防止读取不必要的系统或版本控制文件\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 UTF-8 编码读取文本文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n line_count = len(lines)\\n \\n # 判定当前文件的行数是否在120至180之间\\n if line_count < 120 or line_count > 180:\\n return False\\n \\n file_checked = True\\n \\n except UnicodeDecodeError:\\n # 遇到非文本文件解码错误时直接跳过\\n continue\\n except Exception:\\n # 其他异常同样跳过\\n continue\\n \\n # 至少需要检查过一个有效的文件且全部符合要求\\n return file_checked\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"输出的代码文件里不允许出现行数超过20行的函数\",\n \"validation_code\": \"import os\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def remove_comments_and_strings(js_code: str) -> str:\\n NORMAL = 0\\n IN_SINGLE_QUOTE = 1\\n IN_DOUBLE_QUOTE = 2\\n IN_BACKTICK = 3\\n IN_LINE_COMMENT = 4\\n IN_BLOCK_COMMENT = 5\\n \\n state = NORMAL\\n clean_code = []\\n i = 0\\n n = len(js_code)\\n \\n while i < n:\\n c = js_code[i]\\n if state == NORMAL:\\n if c == '/' and i + 1 < n and js_code[i+1] == '/':\\n state = IN_LINE_COMMENT\\n clean_code.append(' ')\\n i += 1\\n elif c == '/' and i + 1 < n and js_code[i+1] == '*':\\n state = IN_BLOCK_COMMENT\\n clean_code.append(' ')\\n i += 1\\n elif c == \\\"'\\\":\\n state = IN_SINGLE_QUOTE\\n clean_code.append(' ')\\n elif c == '\\\"':\\n state = IN_DOUBLE_QUOTE\\n clean_code.append(' ')\\n elif c == \\\"`\\\":\\n state = IN_BACKTICK\\n clean_code.append(' ')\\n else:\\n clean_code.append(c)\\n elif state == IN_LINE_COMMENT:\\n if c == '\\\\n':\\n state = NORMAL\\n clean_code.append(c)\\n elif state == IN_BLOCK_COMMENT:\\n if c == '*' and i + 1 < n and js_code[i+1] == '/':\\n state = NORMAL\\n i += 1\\n elif c == '\\\\n':\\n clean_code.append(c)\\n elif state == IN_SINGLE_QUOTE:\\n if c == '\\\\\\\\':\\n if i + 1 < n and js_code[i+1] == '\\\\n':\\n clean_code.append('\\\\n')\\n i += 1\\n elif c == \\\"'\\\":\\n state = NORMAL\\n elif c == '\\\\n':\\n clean_code.append(c)\\n elif state == IN_DOUBLE_QUOTE:\\n if c == '\\\\\\\\':\\n if i + 1 < n and js_code[i+1] == '\\\\n':\\n clean_code.append('\\\\n')\\n i += 1\\n elif c == '\\\"':\\n state = NORMAL\\n elif c == '\\\\n':\\n clean_code.append(c)\\n elif state == IN_BACKTICK:\\n if c == '\\\\\\\\':\\n if i + 1 < n and js_code[i+1] == '\\\\n':\\n clean_code.append('\\\\n')\\n i += 1\\n elif c == \\\"`\\\":\\n state = NORMAL\\n elif c == '\\\\n':\\n clean_code.append(c)\\n i += 1\\n \\n return \\\"\\\".join(clean_code)\\n\\n def get_js_function_blocks(js_code: str, offset_line: int = 0) -> list:\\n clean_code = remove_comments_and_strings(js_code)\\n \\n line_starts = [0]\\n for i, c in enumerate(clean_code):\\n if c == '\\\\n':\\n line_starts.append(i + 1)\\n \\n def get_line_num(idx):\\n l = 0\\n r = len(line_starts) - 1\\n while l <= r:\\n mid = (l + r) // 2\\n if line_starts[mid] <= idx:\\n if mid == len(line_starts) - 1 or line_starts[mid+1] > idx:\\n return mid + 1\\n else:\\n l = mid + 1\\n else:\\n r = mid - 1\\n return 1\\n\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with', 'return', 'try', 'do'}\\n blocks = []\\n i = 0\\n n = len(clean_code)\\n \\n while i < n:\\n if clean_code[i] == '{':\\n j = i - 1\\n while j >= 0 and clean_code[j].isspace():\\n j -= 1\\n if j >= 0:\\n stmt_start = j\\n while stmt_start >= 0 and clean_code[stmt_start] not in {';', '{', '}'}:\\n stmt_start -= 1\\n stmt_start += 1\\n pre_str = clean_code[stmt_start:j+1]\\n \\n is_func = False\\n keyword_idx = -1\\n \\n temp_j = len(pre_str) - 1\\n while temp_j >= 0 and pre_str[temp_j].isspace():\\n temp_j -= 1\\n \\n if temp_j >= 1 and pre_str[temp_j-1:temp_j+1] == '=>':\\n is_func = True\\n k = temp_j - 2\\n while k >= 0 and pre_str[k].isspace():\\n k -= 1\\n paren_count = 0\\n while k >= 0:\\n if pre_str[k] == ')':\\n paren_count += 1\\n elif pre_str[k] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n k -= 1\\n if k >= 0:\\n keyword_idx = k\\n else:\\n keyword_idx = temp_j - 1\\n else:\\n k = temp_j\\n while k >= 0:\\n if pre_str[k] == ')':\\n break\\n if pre_str[k] in {';', '{', '}'}:\\n break\\n k -= 1\\n \\n if k >= 0 and pre_str[k] == ')':\\n paren_count = 0\\n p = k\\n while p >= 0:\\n if pre_str[p] == ')':\\n paren_count += 1\\n elif pre_str[p] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n p -= 1\\n \\n if p >= 0:\\n w = p - 1\\n while w >= 0 and pre_str[w].isspace():\\n w -= 1\\n \\n if w >= 0 and pre_str[w] == '>':\\n angle_count = 0\\n while w >= 0:\\n if pre_str[w] == '>':\\n angle_count += 1\\n elif pre_str[w] == '<':\\n angle_count -= 1\\n if angle_count == 0:\\n w -= 1\\n break\\n w -= 1\\n while w >= 0 and pre_str[w].isspace():\\n w -= 1\\n \\n word_end = w\\n while w >= 0 and (pre_str[w].isalnum() or pre_str[w] in '_$'):\\n w -= 1\\n word = pre_str[w+1:word_end+1]\\n \\n if word == 'function':\\n is_func = True\\n keyword_idx = w + 1\\n elif word not in keywords and word != '':\\n w2 = w\\n while w2 >= 0 and pre_str[w2].isspace():\\n w2 -= 1\\n word2_end = w2\\n while w2 >= 0 and (pre_str[w2].isalnum() or pre_str[w2] in '_$'):\\n w2 -= 1\\n word2 = pre_str[w2+1:word2_end+1]\\n if word2 == 'function':\\n is_func = True\\n keyword_idx = w2 + 1\\n elif word not in keywords:\\n is_func = True\\n keyword_idx = w + 1\\n elif word == '':\\n w2 = w\\n while w2 >= 0 and pre_str[w2].isspace():\\n w2 -= 1\\n word2_end = w2\\n while w2 >= 0 and (pre_str[w2].isalnum() or pre_str[w2] in '_$'):\\n w2 -= 1\\n word2 = pre_str[w2+1:word2_end+1]\\n if word2 == 'function':\\n is_func = True\\n keyword_idx = w2 + 1\\n\\n if is_func:\\n real_start_idx = stmt_start + keyword_idx if keyword_idx != -1 else i\\n start_line = get_line_num(real_start_idx)\\n \\n brace_count = 1\\n k = i + 1\\n while k < n:\\n if clean_code[k] == '{':\\n brace_count += 1\\n elif clean_code[k] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n break\\n k += 1\\n \\n if k < n:\\n end_line = get_line_num(k)\\n blocks.append((start_line + offset_line, end_line + offset_line))\\n i += 1\\n return blocks\\n\\n def extract_script_tags(html_code: str) -> list:\\n script_blocks = []\\n lower_html = html_code.lower()\\n idx = 0\\n while True:\\n start = lower_html.find('<script', idx)\\n if start == -1:\\n break\\n end_start = lower_html.find('>', start)\\n if end_start == -1:\\n break\\n end_tag = lower_html.find('</script>', end_start)\\n if end_tag == -1:\\n break\\n script_blocks.append((end_start + 1, end_tag))\\n idx = end_tag + 9\\n return script_blocks\\n\\n def get_py_function_blocks(py_code: str) -> list:\\n blocks = []\\n try:\\n tree = ast.parse(py_code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n end = getattr(node, 'end_lineno', node.lineno)\\n blocks.append((start, end))\\n except:\\n pass\\n return blocks\\n\\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte', '.py', '.cjs', '.mjs'}\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except:\\n continue\\n \\n blocks = []\\n if ext == '.py':\\n blocks = get_py_function_blocks(code)\\n elif ext in {'.html', '.htm', '.vue', '.svelte'}:\\n scripts = extract_script_tags(code)\\n for start, end in scripts:\\n script_code = code[start:end]\\n offset = code[:start].count('\\\\n')\\n blocks.extend(get_js_function_blocks(script_code, offset))\\n else:\\n blocks = get_js_function_blocks(code, 0)\\n \\n for start_line, end_line in blocks:\\n if end_line - start_line + 1 > 20:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"为了让物理模拟更真实且具备探究价值，下一步我们需要打破理想环境，引入“空气阻力”模型。请在参数设置区新增一个“材质与形状”的下拉菜单（选项包含：实心铁球、木球、带降落伞的沙袋）。选择不同的选项时，系统在后台自动赋予小球不同的风阻系数(C)和迎风面积(A)。在小球的下落运动演算中加入空气阻力公式（F = 0.5 * ρ * v² * C * A，空气密度ρ设为常数）。由于阻力与速度平方成正比，小球最终应能达到终端速度。另外，请在实时数据看板中新增一项，实时显示当前小球受到的空气阻力大小（牛顿）。\\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过15行的函数；第二，生成的代码文件中必须仅用4个空格实现缩进。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复是否在最开头带有“汪~”语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除由于排版可能产生的开头空白字符后，判断是否以“汪~”开头\\n return response.lstrip().startswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复文本末尾可能存在的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 判断清理后的回复是否以\\\"呜~\\\"结尾\\n return cleaned_response.endswith(\\\"呜~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 判断最后一轮回复中是否包含字符\\\"你\\\"\\n # 如果包含，则说明违反了“禁止用'你'称呼用户”的要求\\n if '你' in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的每一个文件行数调整为需在120-180行之间\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 校验工作区路径是否存在且为有效目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令（Web应用及progress.md文档）推断可能的目标文件类型\\n target_extensions = {'.html', '.css', '.js', '.md', '.txt', '.json'}\\n has_target_file = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略所有隐藏目录（例如 .git, .vscode 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 忽略所有隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n line_count = len(lines)\\n \\n # 检查当前文件的行数是否在限定的 120-180 之间\\n if line_count < 120 or line_count > 180:\\n return False\\n has_target_file = True\\n except Exception:\\n # 如果遇到文件读取异常（如编码错误或权限问题），跳过该文件\\n continue\\n \\n # 只要存在至少一个目标文件且所有读取的文件都符合行数区间，即判定通过\\n return has_target_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"输出的代码文件里不允许出现行数超过15行的函数\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef clean_code(code_str: str) -> str:\\n \\\"\\\"\\\"清理 JavaScript / TypeScript 代码，将注释和字符串替换为空格以方便分析，保证行号一致。\\\"\\\"\\\"\\n result = []\\n i = 0\\n length = len(code_str)\\n in_single = False\\n in_double = False\\n in_backtick = False\\n in_single_comment = False\\n in_multi_comment = False\\n \\n while i < length:\\n char = code_str[i]\\n \\n # 1. 单行注释处理\\n if in_single_comment:\\n if char == '\\\\n':\\n in_single_comment = False\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n # 2. 多行注释处理\\n if in_multi_comment:\\n if char == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n elif char == '*' and i + 1 < length and code_str[i+1] == '/':\\n in_multi_comment = False\\n result.append(' ')\\n i += 2\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n # 3. 字符串（单引号、双引号、反引号）处理\\n if in_single:\\n if char == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n elif char == '\\\\\\\\' and i + 1 < length:\\n if code_str[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif char == \\\"'\\\":\\n in_single = False\\n result.append(' ')\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_double:\\n if char == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n elif char == '\\\\\\\\' and i + 1 < length:\\n if code_str[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif char == '\\\"':\\n in_double = False\\n result.append(' ')\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_backtick:\\n if char == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n elif char == '\\\\\\\\' and i + 1 < length:\\n if code_str[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif char == '`':\\n in_backtick = False\\n result.append(' ')\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n # 进入字符串与注释判断\\n if char == \\\"'\\\":\\n in_single = True\\n result.append(' ')\\n i += 1\\n continue\\n if char == '\\\"':\\n in_double = True\\n result.append(' ')\\n i += 1\\n continue\\n if char == '`':\\n in_backtick = True\\n result.append(' ')\\n i += 1\\n continue\\n if char == '/' and i + 1 < length:\\n if code_str[i+1] == '/':\\n in_single_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n elif code_str[i+1] == '*':\\n in_multi_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n \\n result.append(char)\\n i += 1\\n \\n return \\\"\\\".join(result)\\n\\ndef check_functions_length_in_js(code_str: str) -> bool:\\n \\\"\\\"\\\"提取JS中的大括号深度并分析函数长度。\\\"\\\"\\\"\\n cleaned = clean_code(code_str)\\n \\n # 记录每行的起始索引以支持二分法查找行号\\n line_starts = [0]\\n for i, char in enumerate(cleaned):\\n if char == '\\\\n':\\n line_starts.append(i + 1)\\n \\n def get_line_number(index: int) -> int:\\n left, right = 0, len(line_starts) - 1\\n while left <= right:\\n mid = (left + right) // 2\\n if line_starts[mid] == index:\\n return mid + 1\\n elif line_starts[mid] < index:\\n left = mid + 1\\n else:\\n right = mid - 1\\n return right + 1\\n\\n brace_stack = []\\n active_funcs = []\\n method_pattern = re.compile(r'\\\\b(?!if|for|while|switch|catch)\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*$')\\n last_block_end = 0\\n \\n for i, char in enumerate(cleaned):\\n if char in ('{', '}', ';'):\\n if char == '{':\\n raw_prefix = cleaned[last_block_end:i]\\n is_func = False\\n real_start_index = -1\\n \\n # 判断当前的大括号是否属于一个函数的开启\\n func_matches = list(re.finditer(r'\\\\bfunction\\\\b', raw_prefix))\\n if func_matches:\\n is_func = True\\n real_start_index = last_block_end + func_matches[-1].start()\\n elif '=>' in raw_prefix:\\n is_func = True\\n real_start_index = last_block_end + raw_prefix.rfind('=>')\\n else:\\n match = method_pattern.search(raw_prefix)\\n if match:\\n is_func = True\\n real_start_index = last_block_end + match.start()\\n \\n if is_func:\\n start_line = get_line_number(real_start_index)\\n active_funcs.append((len(brace_stack), start_line))\\n \\n brace_stack.append('{')\\n elif char == '}':\\n if brace_stack:\\n brace_stack.pop()\\n for j in range(len(active_funcs)-1, -1, -1):\\n depth, start_line = active_funcs[j]\\n if depth == len(brace_stack):\\n # 函数闭合，计算行数\\n end_line = get_line_number(i)\\n func_lines = end_line - start_line + 1\\n if func_lines > 15:\\n return False\\n active_funcs.pop(j)\\n last_block_end = i + 1\\n \\n return True\\n\\ndef check_functions_length_in_html(html_str: str) -> bool:\\n \\\"\\\"\\\"提取HTML内联的脚本区块并检查。\\\"\\\"\\\"\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', html_str, re.DOTALL | re.IGNORECASE)\\n for script_code in scripts:\\n if script_code.strip() and not check_functions_length_in_js(script_code):\\n return False\\n return True\\n\\ndef check_functions_length_in_py(code_str: str) -> bool:\\n \\\"\\\"\\\"借助AST直接解析并校验Python代码文件。\\\"\\\"\\\"\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n end = node.end_lineno\\n if end is not None and (end - start + 1) > 15:\\n return False\\n except Exception:\\n pass\\n return True\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"验证工作区中的所有代码文件内的函数，是否有任何函数行数超过15行。\\\"\\\"\\\"\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_functions_length_in_py(content):\\n return False\\n elif ext in ('.js', '.ts', '.jsx', '.tsx'):\\n if not check_functions_length_in_js(content):\\n return False\\n elif ext in ('.html', '.htm', '.vue', '.svelte'):\\n if not check_functions_length_in_html(content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中必须仅用4个空格实现缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件扩展名集合\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n in_c_comment = False\\n in_html_comment = False\\n \\n for line in lines:\\n # 忽略空行\\n if not line.strip():\\n continue\\n \\n # 提取该行的前导空白字符\\n leading_spaces = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_spaces += char\\n else:\\n break\\n \\n # 若使用制表符（Tab）进行缩进，判定为不符合要求\\n if '\\\\t' in leading_spaces:\\n return False\\n \\n # 若该行无任何缩进，则继续检查下一行\\n if not leading_spaces:\\n continue\\n \\n stripped_line = line.strip()\\n \\n # 过滤常见的块注释（C风格），避免注释内部随意的空格对齐被误判\\n if stripped_line.startswith('/*'):\\n in_c_comment = True\\n if in_c_comment:\\n if '*/' in stripped_line:\\n in_c_comment = False\\n continue\\n if stripped_line.startswith('*'): # 用于兼容多行注释中间行的前导星号\\n continue\\n \\n # 过滤HTML块注释\\n if stripped_line.startswith('<!--'):\\n in_html_comment = True\\n if in_html_comment:\\n if '-->' in stripped_line:\\n in_html_comment = False\\n continue\\n \\n # 判断前导空格的长度是否严格为4的倍数\\n if len(leading_spaces) % 4 != 0:\\n return False\\n\\n # 若至少遍历到一个代码文件且所有文件缩进均合规，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"做得不错！现在我想增强工具的视觉表现和横向对比能力。请在实验舱增加“频闪打点”功能：小球下落时，每隔 0.2 秒在它当前所在位置留下一个半透明的残影（轨迹点），并在残影旁边用微小字体标注出那一刻的时间戳。此外，请将系统升级为“双球对比”模式：在实验舱里再放入第二个不同颜色的小球（球B），控制面板里的参数区进行左右分栏，允许用户分别对球A和球B独立设置不同的材质、质量或初始高度。点击开始时，双球在同一个场景下同时下落，让用户可以直观观察到空气阻力或质量差异对轨迹的影响。\\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过12行的函数；第二，所有由你创建的代码文件，其编码必须统一为UTF-8。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查回复的最开头是否带有\\\"汪~\\\"语气词。\\n \\\"\\\"\\\"\\n return response.lstrip().startswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n if not response:\\n return False\\n \\n # 使用正则表达式匹配回复末尾是否为\\\"呜~\\\"，允许其后有空格、换行符以及常见的句末标点符号\\n pattern = r'呜~[\\\\s。！!\\\\.?？]*$'\\n match = re.search(pattern, response)\\n \\n return match is not None\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复中是否禁止使用了“你”或“您”来称呼用户。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response or \\\"您\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的每一个文件行数调整为需在120-180行之间\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否有效\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n has_valid_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，如 .git, .vscode 等，避免误将非生成的环境文件计入\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n # 尝试作为文本文件读取内容并计算行数\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n has_valid_file = True\\n line_count = len(lines)\\n \\n # 检查文件的行数是否在 [120, 180] 范围内\\n if line_count < 120 or line_count > 180:\\n return False\\n except UnicodeDecodeError:\\n # 无法用 UTF-8 读取的文件当做二进制/媒体等非代码文件跳过\\n continue\\n except Exception:\\n # 其他读取异常跳过\\n continue\\n \\n # 如果找到了合法的代码文件并且全部通过检查，则返回 True\\n return has_valid_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"输出的代码文件里不允许出现行数超过12行的函数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def get_python_function_lines(code):\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return []\\n lengths = []\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n lengths.append(node.end_lineno - node.lineno + 1)\\n return lengths\\n\\n def extract_js_functions(code):\\n # 稳健剔除 JS 中的字符串（包括模板字符串）和各类注释，同时保留空行以对齐绝对行号\\n pattern = re.compile(r'/\\\\*.*?\\\\*/|//.*|\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`(?:[^`\\\\\\\\]|\\\\\\\\.)*`', re.DOTALL)\\n def replacer(match):\\n m = match.group(0)\\n if m.startswith('/*'): \\n return '\\\\n' * m.count('\\\\n')\\n elif m.startswith('//'): \\n return ''\\n else: \\n return '\\\"\\\"' + '\\\\n' * m.count('\\\\n')\\n\\n clean_code = pattern.sub(replacer, code)\\n \\n line_num = 1\\n char_lines = []\\n for char in clean_code:\\n char_lines.append(line_num)\\n if char == '\\\\n':\\n line_num += 1\\n \\n tokens = []\\n token_pattern = re.compile(r'[a-zA-Z_$][a-zA-Z0-9_$]*|=>|[{}()\\\\[\\\\];=,.]|\\\\s+')\\n pos = 0\\n while pos < len(clean_code):\\n m = token_pattern.match(clean_code, pos)\\n if m:\\n tok = m.group(0)\\n if not tok.isspace():\\n tokens.append({'val': tok, 'pos': pos, 'line': char_lines[pos]})\\n pos = m.end()\\n else:\\n pos += 1\\n\\n function_starts = []\\n for i, tok in enumerate(tokens):\\n if tok['val'] == '{':\\n is_func = False\\n # 检查箭头函数： => {\\n if i >= 1 and tokens[i-1]['val'] == '=>':\\n is_func = True\\n # 检查标准/匿名/类成员函数：) {\\n elif i >= 1 and tokens[i-1]['val'] == ')':\\n paren_count = 1\\n j = i - 2\\n while j >= 0 and paren_count > 0:\\n if tokens[j]['val'] == ')':\\n paren_count += 1\\n elif tokens[j]['val'] == '(':\\n paren_count -= 1\\n j -= 1\\n if paren_count == 0:\\n if j >= 0:\\n prev_tok = tokens[j]['val']\\n # 剔除常见的控制流关键字\\n if prev_tok not in ['if', 'for', 'while', 'catch', 'switch', 'with']:\\n # 剔除类的继承特征（如 class A extends B(...) { ）\\n k = j - 1\\n is_class = False\\n while k >= 0:\\n if tokens[k]['val'] in ['{', '}', ';']:\\n break\\n if tokens[k]['val'] == 'class':\\n is_class = True\\n break\\n k -= 1\\n if not is_class:\\n is_func = True\\n else:\\n is_func = True\\n \\n if is_func:\\n function_starts.append(i)\\n\\n function_lines = []\\n for start_idx in function_starts:\\n brace_count = 1\\n start_line = tokens[start_idx]['line']\\n end_line = start_line\\n for j in range(start_idx + 1, len(tokens)):\\n if tokens[j]['val'] == '{':\\n brace_count += 1\\n elif tokens[j]['val'] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_line = tokens[j]['line']\\n break\\n else:\\n # 若未闭合，则认为跨度直至文件末尾\\n if len(tokens) > start_idx:\\n end_line = tokens[-1]['line']\\n \\n function_lines.append(end_line - start_line + 1)\\n \\n return function_lines\\n\\n def check_file(file_path):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 读写异常、无法解析的非文本文件则跳过\\n return True\\n\\n ext = os.path.splitext(file_path)[1].lower()\\n \\n if ext == '.py':\\n lines = get_python_function_lines(content)\\n for l in lines:\\n if l > 12:\\n return False\\n \\n elif ext in ['.js', '.jsx', '.ts', '.tsx']:\\n lines = extract_js_functions(content)\\n for l in lines:\\n if l > 12:\\n return False\\n \\n elif ext in ['.html', '.htm', '.vue', '.svelte']:\\n # 提取文件内嵌入的 <script>...</script> 代码块\\n script_pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\\n scripts = script_pattern.findall(content)\\n for script in scripts:\\n lines = extract_js_functions(script)\\n for l in lines:\\n if l > 12:\\n return False\\n \\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略构建和版本控制目录，优化解析效率\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n # 仅校验特定的代码文件后缀，如若其中有函数长于12行立刻失败\\n ext = os.path.splitext(file_path)[1].lower()\\n if ext in ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue', '.svelte']:\\n if not check_file(file_path):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中必须仅用4个空格实现缩进\",\n \"validation_code\": \"import os\\n\\ndef check_indentation(response: str, workspace_path: str) -> bool:\\n # 针对基于 Web 的应用及常见的相关后端语言，定义需要检查的代码文件后缀\\n code_extensions = {'.html', '.js', '.css', '.ts', '.py', '.java', '.cpp', '.c', '.cs'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果文件无法按 UTF-8 读取则跳过\\n continue\\n \\n for line in lines:\\n # 忽略空行或仅包含空白字符的行\\n if not line.strip():\\n continue\\n \\n # 提取该行的前导空白字符\\n leading_space = \\\"\\\"\\n for ch in line:\\n if ch in (' ', '\\\\t'):\\n leading_space += ch\\n else:\\n break\\n \\n # 如果行首没有空白字符，说明未缩进，属于正常现象，直接跳过\\n if not leading_space:\\n continue\\n \\n # 规则1：必须仅用空格实现缩进，包含制表符即违规\\n if '\\\\t' in leading_space:\\n return False\\n \\n # 规则2：使用的空格数必须是4的倍数\\n if len(leading_space) % 4 != 0:\\n stripped_line = line.lstrip()\\n # 特殊容错：忽略以 '*' 开头的多行注释行（常见于 JS / CSS / Java 等）\\n # 此类注释的对齐通常会引入额外的空格，属于正常代码规范\\n if stripped_line.startswith('*'):\\n continue\\n \\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查代码项目路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令推断的可能文件类型：Web前端文件 + progress.md\\n target_extensions = {'.html', '.css', '.js', '.md'}\\n has_target_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录（如 .git）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_target_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制模式读取文件内容\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n \\n # 尝试用 utf-8 解码，若文件中包含非UTF-8字节序列（如GBK编码的中文字符），将抛出异常\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 解码失败，说明不是合法的UTF-8编码\\n return False\\n except Exception:\\n # 发生其他文件读取错误，视为未满足要求\\n return False\\n \\n # 如果没有找到任何目标代码文件，无法证明满足要求，返回False；否则返回True\\n return has_target_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后一轮，我们加入能量可视化分析和实验导出功能来提升工具的专业度。请在页面底部新增一个横向展开的面板，使用原生 DOM 元素的高度动态变化来模拟柱状图。为每个小球提供三根并排的动态柱子，实时显示其：动能(Ek)、重力势能(Ep)和总机械能(E)。用户可以直观看到由于空气阻力和地面反弹做功导致的机械能衰减过程。最后，在控制面板最下方增加一个“导出实验报告”按钮，点击后能将整个下落过程（以 0.1 秒为采样间隔）中两个球的时间、高度、速度和总机械能数据拼接成一份完整的 CSV 文件，并自动触发浏览器下载。\\n另外请你注意，后续所有轮次的开发过程中都需要遵守两项更新的要求：第一，输出的代码文件里不允许出现行数超过10行的函数；第二，请确保在改动代码前使用 TodoWrite 工具规划任务。对于任意任务，只要其相关的代码修改执行完毕，下一个工具调用则强制要求为 TodoComplete 以标记任务完结，期间不可穿插其他任务的代码修改。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n # 去除回复内容最前面的空白字符（包括空格、换行符等），确保准确判断首字符\\n cleaned_response = response.lstrip()\\n \\n # 判断是否以“汪~”或全角波浪号的“汪～”开头\\n if cleaned_response.startswith(\\\"汪~\\\") or cleaned_response.startswith(\\\"汪～\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的所有回复都必须在结尾处加上\\\"呜~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复末尾的空白字符以及可能存在的常见标点符号，防止误判\\n cleaned_response = response.rstrip(' \\\\n\\\\r\\\\t。！!.')\\n \\n # 判断清理后的回复是否以“呜~”结尾\\n return cleaned_response.endswith(\\\"呜~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"主人\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查AI的回复中是否不包含“你”字，以确保其没有使用“你”称呼用户。\\n \\\"\\\"\\\"\\n # 如果回复中存在“你”，则说明未满足替换要求\\n if \\\"你\\\" in response:\\n return False\\n \\n # 未出现“你”则视为满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在探索代码时，请最大程度地依赖 Task 工具，切勿直接使用 Grep 或 Glob 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"为杜绝进程挂起，当运行 bash 相关的构建、安装或测试命令时，一律需要在命令的最前端加上 `timeout 60s` 包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的每一个文件行数调整为需在120-180行之间\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目性质推断可能生成的代码及文档文件后缀\\n valid_exts = {'.html', '.css', '.js', '.md'}\\n file_checked = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和可能存在的常见的依赖项目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'vendor')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n line_count = len(lines)\\n # 检查当前文件行数是否落在120-180范围内\\n if line_count < 120 or line_count > 180:\\n return False\\n \\n file_checked = True\\n except Exception:\\n # 遇到非文本或者读取异常的情况直接跳过\\n continue\\n\\n # 至少需要存在被成功检查的符合类型的文件，才能判定满足该要求\\n return file_checked\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"输出的代码文件里不允许出现行数超过10行的函数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n \\n def check_py_functions(code_str: str) -> bool:\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n if node.end_lineno - node.lineno + 1 > 10:\\n return False\\n return True\\n except Exception:\\n return True\\n \\n def check_js_functions(code_str: str) -> bool:\\n # 匹配字符串、多行注释和单行注释，用于剔除这些会干扰大括号嵌套解析的内容\\n pattern = r'(?:\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|(?:\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|(?:`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|(?:/\\\\*.*?\\\\*/)|(?://[^\\\\n]*)'\\n \\n def replacer(match):\\n s = match.group(0)\\n # 保证原本的换行符不丢失，只将非换行字符替换为空格，以确保行号和索引保持不变\\n return ''.join('\\\\n' if c == '\\\\n' else ' ' for c in s)\\n \\n try:\\n cleaned = re.sub(pattern, replacer, code_str, flags=re.DOTALL)\\n except Exception:\\n cleaned = code_str\\n \\n function_starts = set()\\n \\n # 1. 寻找普通函数: function 关键字，到 { 之间不出现 ; 的内容\\n for m in re.finditer(r'\\\\bfunction\\\\b[^{;]*\\\\{', cleaned):\\n brace_idx = m.end() - 1\\n function_starts.add(brace_idx)\\n \\n # 2. 寻找箭头函数: => {\\n for m in re.finditer(r'=>\\\\s*\\\\{', cleaned):\\n brace_idx = m.end() - 1\\n function_starts.add(brace_idx)\\n \\n # 3. 寻找类/对象内的方法简写形式: name(args) { \\n keywords = {\\\"if\\\", \\\"for\\\", \\\"while\\\", \\\"switch\\\", \\\"catch\\\", \\\"function\\\", \\\"return\\\", \\\"with\\\"}\\n for m in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^;{()]*\\\\)\\\\s*\\\\{', cleaned):\\n name = m.group(1)\\n if name not in keywords:\\n brace_idx = m.end() - 1\\n function_starts.add(brace_idx)\\n \\n # 验证每一个函数块的行数\\n for brace_idx in function_starts:\\n depth = 1\\n i = brace_idx + 1\\n while i < len(cleaned):\\n if cleaned[i] == '{':\\n depth += 1\\n elif cleaned[i] == '}':\\n depth -= 1\\n if depth == 0:\\n start_line = cleaned.count('\\\\n', 0, brace_idx) + 1\\n end_line = cleaned.count('\\\\n', 0, i) + 1\\n if end_line - start_line + 1 > 10:\\n return False\\n break\\n i += 1\\n return True\\n\\n def check_html_functions(code_str: str) -> bool:\\n # 提取所有的 <script> 标签中的内容进行校验\\n pattern = r'<script\\\\b[^>]*>(.*?)</script>'\\n for m in re.finditer(pattern, code_str, flags=re.DOTALL | re.IGNORECASE):\\n script_content = m.group(1)\\n start_pos = m.start(1)\\n prefix = code_str[:start_pos]\\n # 为了维持行号准确，将提取出来的JS内容前填充对应数量的换行符\\n newlines = '\\\\n' * prefix.count('\\\\n')\\n padded_script = newlines + script_content\\n if not check_js_functions(padded_script):\\n return False\\n return True\\n\\n # 遍历项目工作区中生成的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_py_functions(code_str):\\n return False\\n elif ext in ['.html', '.htm', '.vue']:\\n if not check_html_functions(code_str):\\n return False\\n else:\\n if not check_js_functions(code_str):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在 3 或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到progress.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中必须仅用4个空格实现缩进\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 预估可能出现的常见代码文件扩展名\\n code_exts = {\\n '.py', '.js', '.ts', '.html', '.css', '.vue', '.jsx', \\n '.tsx', '.json', '.java', '.c', '.cpp', '.h', '.cs', \\n '.go', '.rb', '.php', '.swift', '.kt'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果因权限或非UTF-8编码导致无法读取，安全跳过\\n continue\\n \\n indent_counts = []\\n for line in lines:\\n # 仅匹配行首的空白字符，且要求行内紧接着存在非空白字符\\n match = re.match(r'^([ \\\\t]+)(?=\\\\S)', line)\\n if match:\\n indent = match.group(1)\\n content = line[len(indent):]\\n \\n # 1. 严格检查：只要前导空白中包含 tab 制表符，即判定不合规\\n if '\\\\t' in indent:\\n return False\\n \\n # 2. 排除多行注释中常见的星号对齐行带来的统计干扰\\n if content.startswith('*'):\\n continue\\n \\n indent_counts.append(len(indent))\\n \\n # 若文件内实际包含有效缩进记录，则分析其空格分布\\n if indent_counts:\\n # 统计缩进空格数量恰为 4 的倍数的行数\\n mod_4_count = sum(1 for count in indent_counts if count % 4 == 0)\\n \\n # 兼容部分特殊场景（如长参数折行对齐等），要求 70% 满足条件即认为合规。\\n # 如果比例过低，则很大可能使用的是 2 个空格或杂乱规范的缩进。\\n if mod_4_count / len(indent_counts) < 0.7:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，说明没有创建文件，默认未违背规则\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 根据用户指令，推测Web实验工具项目可能的代码文件类型\\n code_extensions = {'.html', '.htm', '.css', '.js', '.md', '.json'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制模式读取文件\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 验证文件是否可以通过 utf-8 解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明包含非 UTF-8 编码的内容\\n return False\\n except Exception:\\n # 忽略其他可能导致无法正常读取的系统层面错误\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请确保在改动代码前使用 TodoWrite 工具规划任务。对于任意任务，只要其相关的代码修改执行完毕，下一个工具调用则强制要求为 TodoComplete 以标记任务完结，期间不可穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面整体静态排版，确认左侧存在主要区域“实验舱”（含底部代表地面的粗线和界面角落的比例尺），实验舱内有两颗颜色不同的小球；右侧有控制面板；页面底部包含能量分析的柱状图面板。","查看右侧“实验参数设置区”，确认其采用了分栏布局以分别控制两个小球，且每个小球独立具备：初始高度（米）、重力加速度（m/s²）、质量（千克）的输入框或滑块，以及“材质与形状”（如实心铁球、木球、带降落伞的沙袋）下拉菜单。","在未启动实验（静止状态）下，调整任一小球的“初始高度”数值，确认其取值范围被限制在1-100米之间，且左侧实验舱中该小球的垂直位置会实时上下移动并与设定的高度对应。","将球A的“重力加速度”设为极大值（如20或更高），球B保持9.8，点击“开始”按钮，观察下落过程，确认重力加速度大的球下落速度明显更快。","将球A材质选为“实心铁球”，球B材质选为“带降落伞的沙袋”，高度相同，点击“开始”按钮。观察下落过程，确认引入空气阻力后，球B的下落速度由于空气阻力较大而明显慢于球A。","点击“开始”让小球下落，在半空中点击“暂停”按钮，确认所有小球瞬间悬停在空中不再移动；再次点击“开始”，小球从原悬停位置继续加速下落。","点击“开始”按钮，紧盯小球触底的瞬间，确认小球没有穿透底部代表地面的粗线（不穿模），而是向上发生反弹，且每次反弹的最高点逐渐降低，最后完全静止在地面粗线上。","在小球正在下落或已静止在底部的状态下，点击“重置”按钮，确认左侧实验舱的小球立刻返回页面顶部的初始设定高度位置，且处于静止状态。","点击“开始”按钮让小球下落，观察左侧实验舱，确认小球在下落路线会持续留下半透明的残影轨迹点，且每个残影旁边都有微小字体标注的具体时间戳数值，验证任意两个相邻时间戳数值之差转化为“秒”后为0.2秒。","点击“开始”按钮后，观察右侧“实时数据看板”，确认面板中的数字在小球运动期间持续快速刷新，且明确显示带单位的物理量：已用时间（秒，保留两位小数）、当前速度（米/秒）、离地高度（米）、空气阻力大小（牛顿）。","点击“开始”按钮后，观察页面底部的能量柱状图面板，确认每个小球对应的三根柱子（动能、重力势能、总机械能）的高度随下落实时动态变化，且能在下落过程中观察到重力势能柱缩短、动能柱长高的现象。","点击控制面板最下方的“导出实验报告”按钮，检查浏览器是否成功触发文件下载弹窗，下载的文件后缀为 .csv 格式；打开CSV文件确认下落过程数据是以0.1秒为采样间隔记录的，且包含两个球的时间、高度、速度和总机械能数据。"],"string":"[\n \"查看页面整体静态排版，确认左侧存在主要区域“实验舱”（含底部代表地面的粗线和界面角落的比例尺），实验舱内有两颗颜色不同的小球；右侧有控制面板；页面底部包含能量分析的柱状图面板。\",\n \"查看右侧“实验参数设置区”，确认其采用了分栏布局以分别控制两个小球，且每个小球独立具备：初始高度（米）、重力加速度（m/s²）、质量（千克）的输入框或滑块，以及“材质与形状”（如实心铁球、木球、带降落伞的沙袋）下拉菜单。\",\n \"在未启动实验（静止状态）下，调整任一小球的“初始高度”数值，确认其取值范围被限制在1-100米之间，且左侧实验舱中该小球的垂直位置会实时上下移动并与设定的高度对应。\",\n \"将球A的“重力加速度”设为极大值（如20或更高），球B保持9.8，点击“开始”按钮，观察下落过程，确认重力加速度大的球下落速度明显更快。\",\n \"将球A材质选为“实心铁球”，球B材质选为“带降落伞的沙袋”，高度相同，点击“开始”按钮。观察下落过程，确认引入空气阻力后，球B的下落速度由于空气阻力较大而明显慢于球A。\",\n \"点击“开始”让小球下落，在半空中点击“暂停”按钮，确认所有小球瞬间悬停在空中不再移动；再次点击“开始”，小球从原悬停位置继续加速下落。\",\n \"点击“开始”按钮，紧盯小球触底的瞬间，确认小球没有穿透底部代表地面的粗线（不穿模），而是向上发生反弹，且每次反弹的最高点逐渐降低，最后完全静止在地面粗线上。\",\n \"在小球正在下落或已静止在底部的状态下，点击“重置”按钮，确认左侧实验舱的小球立刻返回页面顶部的初始设定高度位置，且处于静止状态。\",\n \"点击“开始”按钮让小球下落，观察左侧实验舱，确认小球在下落路线会持续留下半透明的残影轨迹点，且每个残影旁边都有微小字体标注的具体时间戳数值，验证任意两个相邻时间戳数值之差转化为“秒”后为0.2秒。\",\n \"点击“开始”按钮后，观察右侧“实时数据看板”，确认面板中的数字在小球运动期间持续快速刷新，且明确显示带单位的物理量：已用时间（秒，保留两位小数）、当前速度（米/秒）、离地高度（米）、空气阻力大小（牛顿）。\",\n \"点击“开始”按钮后，观察页面底部的能量柱状图面板，确认每个小球对应的三根柱子（动能、重力势能、总机械能）的高度随下落实时动态变化，且能在下落过程中观察到重力势能柱缩短、动能柱长高的现象。\",\n \"点击控制面板最下方的“导出实验报告”按钮，检查浏览器是否成功触发文件下载弹窗，下载的文件后缀为 .csv 格式；打开CSV文件确认下落过程数据是以0.1秒为采样间隔记录的，且包含两个球的时间、高度、速度和总机械能数据。\"\n]","isConversational":false}}},{"rowIdx":71,"cells":{"task_id":{"kind":"number","value":72,"string":"72"},"repository_policy":{"kind":"string","value":"后续所有轮次的交互都必须严格遵循以下约束要求：\n1. 凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json。\n2. 每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容。\n3. 无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码。\n4. 记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n5. 生成的各个代码文件中，均需含有 \"——\"。\n6. 请保证你的所有源文件在注释区域均不涉及 emoji表情符号。\n7. 请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的编程语言及前端代码源码文件后缀\n code_extensions = (\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', \n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rb', \n '.php', '.sh', '.vue', '.swift', '.kt', '.rs', '.m'\n )\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和通常无需检查的依赖/编译目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env', 'dist', 'build')]\n \n for file in files:\n if file.endswith(code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n except Exception:\n # 若文件读取发生异常，无法验证是否包含特定声明，按照严格要求视为失败\n return False\n \n # 检查该文件的起始第一行中是否包含目标编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 常见的代码文件扩展名集合\n code_exts = {\n '.py', '.js', '.ts', '.java', '.c', '.cpp', '.h', '.hpp', \n '.cs', '.go', '.rb', '.php', '.sh', '.html', '.css', \n '.jsx', '.tsx', '.vue', '.swift', '.kt', '.m'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的前10行以涵盖“开头处另起一行”的范围\n head_content = \"\"\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n head_content += line\n \n # 检查目标注释字符串是否出现在文件开头的内容中\n if target_comment not in head_content:\n return False\n except Exception:\n # 如果出现编码或读取错误，通常说明生成的文件存在问题或格式不符合预期\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件扩展名\n code_extensions = {\n '.py', '.js', '.ts', '.java', '.cpp', '.c', '.h', '.cs', \n '.go', '.rb', '.php', '.swift', '.sh', '.html', '.css'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查代码文件中是否含有 \"——\"\n if \"——\" not in content:\n return False\n except Exception:\n # 如果读取遇到问题（如非 utf-8 编码或无权限），视为无法满足要求\n return False\n \n # 如果项目中没有任何代码文件，则无法判定为已满足生成要求，返回False；反之返回True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import tokenize\n import io\n\n if not os.path.exists(workspace_path):\n return True\n\n # 定义常见的源文件扩展名及其对应的注释风格\n source_extensions = {\n '.py': 'python',\n '.java': 'c_style',\n '.c': 'c_style',\n '.cpp': 'c_style',\n '.cc': 'c_style',\n '.h': 'c_style',\n '.hpp': 'c_style',\n '.cs': 'c_style',\n '.js': 'c_style',\n '.ts': 'c_style',\n '.go': 'c_style',\n '.rs': 'c_style',\n '.php': 'c_style',\n '.swift': 'c_style',\n '.kt': 'c_style',\n '.rb': 'hash_style',\n '.sh': 'hash_style'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于编码等问题无法读取文件，则跳过\n continue\n \n style = source_extensions[ext]\n \n if style == 'python':\n try:\n # 尝试使用 tokenize 精准解析 Python 源码\n tokens = tokenize.generate_tokens(io.StringIO(content).readline)\n for tok_type, tok_string, _, _, _ in tokens:\n if tok_type == tokenize.COMMENT:\n for char in tok_string:\n if is_emoji(char):\n return False\n elif tok_type == tokenize.STRING:\n # 检查是否为多行字符串（通常用作函数/类的文档注释 docstring）\n if tok_string.startswith('\"\"\"') or tok_string.startswith(\"'''\"):\n for char in tok_string:\n if is_emoji(char):\n return False\n except Exception:\n # 降级方案：由于语法不完整等导致解析失败时，使用正则表达式匹配\n comments = re.findall(r'#.*', content)\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n elif style == 'c_style':\n # 匹配 C/C++/Java/JS 等语言的单行和多行注释\n comments = re.findall(r'//.*', content)\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n elif style == 'hash_style':\n # 匹配 Ruby/Shell 等语言的单行注释\n comments = re.findall(r'#.*', content)\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]}],"string":"[\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的编程语言及前端代码源码文件后缀\\n code_extensions = (\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rb', \\n '.php', '.sh', '.vue', '.swift', '.kt', '.rs', '.m'\\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和通常无需检查的依赖/编译目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env', 'dist', 'build')]\\n \\n for file in files:\\n if file.endswith(code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n except Exception:\\n # 若文件读取发生异常，无法验证是否包含特定声明，按照严格要求视为失败\\n return False\\n \\n # 检查该文件的起始第一行中是否包含目标编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 常见的代码文件扩展名集合\\n code_exts = {\\n '.py', '.js', '.ts', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.cs', '.go', '.rb', '.php', '.sh', '.html', '.css', \\n '.jsx', '.tsx', '.vue', '.swift', '.kt', '.m'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的前10行以涵盖“开头处另起一行”的范围\\n head_content = \\\"\\\"\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n head_content += line\\n \\n # 检查目标注释字符串是否出现在文件开头的内容中\\n if target_comment not in head_content:\\n return False\\n except Exception:\\n # 如果出现编码或读取错误，通常说明生成的文件存在问题或格式不符合预期\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.ts', '.java', '.cpp', '.c', '.h', '.cs', \\n '.go', '.rb', '.php', '.swift', '.sh', '.html', '.css'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查代码文件中是否含有 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 如果读取遇到问题（如非 utf-8 编码或无权限），视为无法满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，则无法判定为已满足生成要求，返回False；反之返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import tokenize\\n import io\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 定义常见的源文件扩展名及其对应的注释风格\\n source_extensions = {\\n '.py': 'python',\\n '.java': 'c_style',\\n '.c': 'c_style',\\n '.cpp': 'c_style',\\n '.cc': 'c_style',\\n '.h': 'c_style',\\n '.hpp': 'c_style',\\n '.cs': 'c_style',\\n '.js': 'c_style',\\n '.ts': 'c_style',\\n '.go': 'c_style',\\n '.rs': 'c_style',\\n '.php': 'c_style',\\n '.swift': 'c_style',\\n '.kt': 'c_style',\\n '.rb': 'hash_style',\\n '.sh': 'hash_style'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于编码等问题无法读取文件，则跳过\\n continue\\n \\n style = source_extensions[ext]\\n \\n if style == 'python':\\n try:\\n # 尝试使用 tokenize 精准解析 Python 源码\\n tokens = tokenize.generate_tokens(io.StringIO(content).readline)\\n for tok_type, tok_string, _, _, _ in tokens:\\n if tok_type == tokenize.COMMENT:\\n for char in tok_string:\\n if is_emoji(char):\\n return False\\n elif tok_type == tokenize.STRING:\\n # 检查是否为多行字符串（通常用作函数/类的文档注释 docstring）\\n if tok_string.startswith('\\\"\\\"\\\"') or tok_string.startswith(\\\"'''\\\"):\\n for char in tok_string:\\n if is_emoji(char):\\n return False\\n except Exception:\\n # 降级方案：由于语法不完整等导致解析失败时，使用正则表达式匹配\\n comments = re.findall(r'#.*', content)\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n elif style == 'c_style':\\n # 匹配 C/C++/Java/JS 等语言的单行和多行注释\\n comments = re.findall(r'//.*', content)\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n elif style == 'hash_style':\\n # 匹配 Ruby/Shell 等语言的单行注释\\n comments = re.findall(r'#.*', content)\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"UI/UX Optimization"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我设计一个SVG格式的logo。首先，创建一个800x800的画布，背景色设为深邃的藏青色或深黑色。在画布正中间，用简单的粗线条或者单一颜色的多边形画一个大写字母'Z'。这只是个基础骨架，后面我们会一步步把它做得非常酷炫。另外有两个后续所有轮次都必须严格遵循的要求：第一，不要让你生成的任何代码文件里存在空行；第二，在你输出的每一段给用户的回复前，务必先加上\"呜哇～\"这个语气词。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件后缀，并包含用户指令中明确要生成的 .svg 格式\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', \n '.hpp', '.go', '.rs', '.rb', '.php', '.sh', '.bash', \n '.svg', '.xml', '.vue', '.svelte'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法用 utf-8 读取或其他异常（包含空文件无内容的情况），判定为未满足\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见代码文件的扩展名集合，根据指令推断至少可能涉及.svg\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', \n '.java', '.cpp', '.c', '.h', '.hpp', '.svg', '.go', \n '.rs', '.php', '.rb', '.sh', '.xml'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 仅需读取前5行即可验证是否在文件\"开头处\"\n lines = [f.readline() for _ in range(5)]\n except Exception:\n # 如果由于编码或其他原因无法读取文件内容，默认返回 False\n return False\n \n # 检查前几行中是否包含目标注释字符串\n found = False\n for line in lines:\n if not line:\n break\n if target_text in line:\n found = True\n break\n \n # 如果某个代码文件没有包含所需的注释，则总体判定为不满足要求\n if not found:\n return False\n \n # 如果项目中没有任何代码文件，说明没有可验证对象，这里将其视为未满足生成要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名，根据上下文包含.svg并排除.json/.md等非典型或不可注释的格式\n code_exts = {\n '.py', '.svg', '.js', '.ts', '.html', '.css', \n '.java', '.cpp', '.c', '.h', '.cs', '.go', \n '.rs', '.php', '.rb', '.sh'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n \n # 遍历当前工作区路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 判断内容是否包含 \"——\"\n if \"——\" not in content:\n return False\n except Exception:\n # 无法读取或解码文件，视为无法验证该文件包含指定字符串\n return False\n \n # 如果找到了至少一个代码文件且全都包含 \"——\"，返回True，否则返回False\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n \n # 1. 匹配脚本语言文件中的注释（如 Python, Shell, Ruby等）\n if ext in {'.py', '.rb', '.sh'}:\n comments.extend(re.findall(r'#.*', content))\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n \n # 2. 匹配标记语言文件中的注释（包含 SVG, HTML, XML等）\n elif ext in {'.html', '.htm', '.svg', '.xml'}:\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n # 3. 匹配 C风格及前端主流语言文件中的注释（如 C/C++, Java, Go, JS, TS, PHP, CSS等）\n elif ext in {'.c', '.cpp', '.h', '.hpp', '.java', '.go', '.js', '.ts', '.php', '.css', '.scss', '.less'}:\n comments.extend(re.findall(r'//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n else:\n # 非源文件跳过检查\n continue\n \n # 检查提取到的所有注释中是否包含 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推测相关的代码文件后缀\n code_extensions = ['.svg', '.json', '.md']\n has_code_file = False\n \n # 如果工作区路径不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 筛选符合推测后缀的代码文件\n if any(file_name.lower().endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 判定是否包含空行（该行去掉空白符后为空）\n if line.strip() == \"\":\n return False\n except Exception:\n # 若遇文件读取或编码错误，则安全跳过继续检查其他文件\n pass\n \n # 如果至少存在一个代码文件且全都经受住了空行检查，则符合要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"呜哇～\"这个语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断AI助手的回复开头是否包含“呜哇～”这个语气词。\n \"\"\"\n if not response:\n return False\n \n return response.lstrip().startswith(\"呜哇～\")","tags":["Content","Persona"]}]},{"round_id":1,"instruction":"不错。现在我们要让这个'Z'具有“几何切割”的风格。请把单调的'Z'拆分成至少四五个紧密贴合的三角形或四边形拼接而成。每个几何块先用不同灰度的白色/浅灰色填充，让它初步具备被切割成多块的视觉感受。另外后续所有轮次都必须严格遵循两个要求：第一，在你输出的每一段给用户的回复前，务必先加上\"哇呜～\"这个语气词；第二，你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能生成的代码文件扩展名\n code_extensions = {\n '.svg', '.json', '.md', '.py', '.js', '.ts', \n '.html', '.css', '.xml', '.txt', '.sh', '.yaml', '.yml'\n }\n file_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果无法按 utf-8 读取或者发生其他异常，视为不符合要求\n return False\n \n # 确保工作区中至少检查过一个代码文件\n return file_found","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_sentence = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 定义常见的代码文件扩展名，涵盖用户要求中可能生成的 .svg, .html, .py 等\n code_extensions = {\n \".py\", \".java\", \".c\", \".cpp\", \".cs\", \".js\", \".ts\", \n \".html\", \".css\", \".svg\", \".sh\", \".xml\", \".go\", \n \".rs\", \".php\", \".rb\"\n }\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对代码文件进行检查\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n found = False\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 检查文件开头处，宽松地设定为前20行内\n for _ in range(20):\n line = f.readline()\n if not line: # 到达文件末尾\n break\n if target_sentence in line:\n found = True\n break\n except Exception:\n # 如果发生读取错误（如非utf-8编码等情况），认为是不满足要求\n return False\n \n # 如果发现某一份代码文件开头没有这句注释，立即返回 False\n if not found:\n return False\n \n # 必须至少存在一份代码文件，且所有代码文件都满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，主要文件格式包括 svg, json, md，同时加入常见的代码文件格式\n target_extensions = {'.svg', '.json', '.md', '.py', '.html', '.css', '.js', '.xml'}\n file_count = 0\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，如 .git\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果读取失败（例如非文本格式或编码不符），则跳过\n continue\n \n file_count += 1\n # 检查文件中是否包含指定的字符串 \"——\"\n if \"——\" not in content:\n return False\n \n # 如果找到了至少一个代码文件，并且所有的文件都满足条件，返回 True\n return file_count > 0","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 根据用户指令及一般编程常识，圈定常见的源代码文件扩展名（重点关注本项目生成的 .svg）\n source_exts = {\n '.svg', '.py', '.js', '.ts', '.html', '.htm', \n '.css', '.c', '.cpp', '.h', '.hpp', '.java', \n '.go', '.rs', '.sh', '.rb', '.php'\n }\n \n # 编译各类型注释的正则表达式\n xml_comment_re = re.compile(r'<!--(.*?)-->', re.DOTALL)\n c_block_comment_re = re.compile(r'/\\*(.*?)\\*/', re.DOTALL)\n c_line_comment_re = re.compile(r'//(.*?)$', re.MULTILINE)\n hash_comment_re = re.compile(r'#(.*?)$', re.MULTILINE)\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n # 根据不同扩展名提取相应格式的注释文本\n if ext in {'.svg', '.html', '.htm'}:\n comments.extend(xml_comment_re.findall(content))\n elif ext in {'.js', '.ts', '.css', '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', '.php'}:\n comments.extend(c_block_comment_re.findall(content))\n comments.extend(c_line_comment_re.findall(content))\n # PHP可能还混用#注释\n if ext == '.php':\n comments.extend(hash_comment_re.findall(content))\n elif ext in {'.py', '.sh', '.rb'}:\n comments.extend(hash_comment_re.findall(content))\n \n # 遍历所提取出注释中的每一个字符，判断是否存在emoji\n for comment_text in comments:\n for char in comment_text:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件及文本文件扩展名\n code_extensions = {\n '.svg', '.json', '.md', '.py', '.js', '.html', '.css', '.xml', '.txt',\n '.sh', '.yaml', '.yml', '.jsx', '.ts', '.tsx', '.vue', '.java', '.c',\n '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 忽略隐藏文件，例如 .git 目录下的文件或操作系统生成的隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 逐行读取文件内容\n for line in f:\n # 判定空行：若剥离空格、制表符和换行符后为空，即为存在空行\n if line.strip() == '':\n return False\n except Exception:\n # 如果遇到无法以 utf-8 解码的二进制文件等异常情况，跳过即可\n continue\n \n # 如果遍历完毕没有发现任何空行，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"哇呜～\"这个语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复首尾的空白字符和换行符\n cleaned_response = response.strip()\n \n # 检查回复是否以指定语气词开头\n return cleaned_response.startswith(\"哇呜～\")","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，主要的代码文件应当是 .svg，同时加入其他常见的代码文件后缀\n code_extensions = {\n '.svg', '.xml', '.html', '.htm', '.py', '.js', '.ts', \n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.php', \n '.go', '.rs', '.swift', '.kt', '.css', '.sh'\n }\n \n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n content = f.read()\n except Exception:\n # 如果存在无法以常规编码读取的二进制文件或异常情况，安全跳过\n continue\n except Exception:\n continue\n \n # 检查要求的目标字符串是否在代码文件中\n if \"<!-- 几何切割Z标识 -->\" not in content:\n return False\n \n # 如果项目中没有任何代码文件，通常也不能算作满足要求\n return has_code_file","tags":["Content","Keyword"]}]},{"round_id":2,"instruction":"好，接下来给这些几何碎片上色。把灰白填充替换成赛博朋克风格的渐变色（比如亮青色到深蓝色的渐变，或者蓝紫渐变）。要求每一块多边形的渐变方向或角度稍微错开一点，利用色彩的高光和暗部，营造出强烈的立体感和水晶切割的感觉。另外后续所有轮次都必须严格遵循两个要求：第一，在你输出的每一段给用户的回复前，务必先加上\"酷喔～\"这个语气词；第二，请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令中提到的\"SVG格式的logo\"，推断核心代码文件为.svg\n # 同时囊括可能出现的关联网页或脚本语言代码类型\n code_exts = {'.svg', '.py', '.js', '.html', '.css', '.ts', '.sh'}\n has_code_file = False\n \n # 遍历当前工作区所有生成的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理推断出属于代码类型的文件\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否严格添加了要求中的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果发生文件读取错误，视为不满足\n return False\n \n # 如果找到了至少一个代码文件并且全满足要求则返回True，否则没有生成代码文件返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据用户指令中涉及的文件，设定可能被视为代码/文本文件的后缀\n valid_exts = {'.svg', '.json', '.md', '.py', '.html', '.css', '.js', '.xml'}\n \n checked_any_file = False\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in valid_exts:\n checked_any_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n lines = []\n with open(file_path, 'r', encoding='utf-8') as f:\n for _ in range(10): # 提取前10行作为“开头处”\n line = f.readline()\n if not line:\n break\n lines.append(line)\n except Exception:\n # 如果有文件无法读取或者不是文本文件，安全跳过\n continue\n \n # 检查该文件前几行中是否包含目标注释文字\n found_target = any(target_text in line for line in lines)\n \n if not found_target:\n return False\n \n # 如果扫描且验证了符合扩展名的文件，并且全部都包含该注释，则满足要求\n return checked_any_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.isdir(workspace_path):\n return False\n\n # 根据用户指令内容推断，代码项目涉及的文件可能是 svg、json 和 md 文件\n target_extensions = {'.svg', '.json', '.md'}\n checked_files = 0\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果是备份文件（比如 .bak 结尾），如果原文件名的扩展名在目标内也可考虑，此处以当前扩展名为准\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 验证文件内容中是否含有要求字符串 \"——\"\n if \"——\" not in content:\n return False\n \n checked_files += 1\n except Exception:\n # 如果无法读取（极少情况的编码等异常），为严谨保守起见，视为未满足条件\n return False\n\n # 确保有相关文件生成且都被检查通过\n return checked_files > 0","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对不同源文件扩展名定义对应的注释匹配正则表达式\n ext_to_patterns = {\n '.svg': [r'<!--[\\s\\S]*?-->'],\n '.html': [r'<!--[\\s\\S]*?-->'],\n '.xml': [r'<!--[\\s\\S]*?-->'],\n '.py': [r'#.*', r'\\'\\'\\'[\\s\\S]*?\\'\\'\\'', r'\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"'],\n '.rb': [r'#.*'],\n '.sh': [r'#.*'],\n '.js': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.ts': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.java': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.c': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.cpp': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.cs': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.go': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*'],\n '.php': [r'/\\*[\\s\\S]*?\\*/', r'(?<![:\\w])//.*', r'#.*'],\n '.css': [r'/\\*[\\s\\S]*?\\*/']\n }\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为我们关注的源文件\n if ext in ext_to_patterns:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 组合当前类型源文件支持的注释提取正则\n patterns = ext_to_patterns[ext]\n combined_pattern = re.compile('|'.join(patterns))\n \n # 提取该文件中的所有注释区域\n comments = combined_pattern.findall(content)\n \n # 检查所有的注释内容中是否包含 emoji\n for comment in comments:\n for char in comment:\n # is_emoji是提供的专家函数，代表判断某一字符是否为emoji\n if is_emoji(char):\n return False\n except Exception:\n # 如果遇到文件无法以utf-8读取或其它系统级别的访问异常情况，直接忽略继续排查即可\n pass\n\n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 包含用户指令提及和常用的代码文件扩展名\n code_exts = {'.svg', '.json', '.md', '.py', '.xml', '.html', '.css', '.js', '.bak'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n \n # 判断是否为代码文件或是备份文件\n if ext in code_exts or \"backup\" in file_name.lower() or \"bak\" in file_name.lower():\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # splitlines() 按行切分，安全过滤了末尾正常换行带来的假空行问题\n lines = f.read().splitlines()\n \n for line in lines:\n # 如果去掉行内空白字符后什么都不剩，则视为存在空行\n if line.strip() == \"\":\n return False\n \n except Exception:\n # 忽略读取异常或非 utf-8 编码的无关二进制文件\n continue\n \n # 如果所有被检查的文件均不存在空行，或者没有任何文件，则返回 True\n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"酷喔～\"这个语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response or not response.strip():\n return False\n \n lines = response.split('\\n')\n paragraphs = []\n current_para = []\n in_code_block = False\n \n for line in lines:\n stripped = line.strip()\n # 遇到代码块标记，切换是否在代码块内的状态\n if stripped.startswith('```'):\n if in_code_block:\n in_code_block = False\n else:\n if current_para:\n paragraphs.append('\\n'.join(current_para))\n current_para = []\n in_code_block = True\n continue\n \n # 如果当前不在代码块内，提取常规回复段落\n if not in_code_block:\n if not stripped: # 空行作为段落的分界线\n if current_para:\n paragraphs.append('\\n'.join(current_para))\n current_para = []\n else:\n current_para.append(stripped)\n \n if current_para:\n paragraphs.append('\\n'.join(current_para))\n \n # 如果整个回复除了代码没有任何普通的正文段落，则退化为检查全文中是否有目标词汇\n if not paragraphs:\n return \"酷喔～\" in response\n \n # 检查所有提取出的回复正文段落是否均以\"酷喔～\"开头\n for para in paragraphs:\n if not para.startswith(\"酷喔～\"):\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令内容推断可能的代码文件类型：SVG (图形代码)、JSON (配置文件)，以及相关的XML和HTML\n code_extensions = {'.svg', '.json', '.xml', '.html'}\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若有任意一个代码文件不包含该标识，则立即返回False\n if \"<!-- 几何切割Z标识 -->\" not in content:\n return False\n except Exception:\n # 如果文件无法正常按utf-8读取，说明可能存在异常，保守视为验证失败\n return False\n\n # 若找到了代码文件并且都通过了校验，则满足要求\n return has_code_files","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":3,"instruction":"看着有内味了。为了让它在深色背景上更加抢眼，我想给整个'Z'加上发光效果。请使用SVG的滤镜（如 feGaussianBlur 或 drop-shadow），在构成字母的几何块背后加一层亮青色或紫色的霓虹发光晕影，让它充满科技感。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\"炸场～\"这个语气词；第二，新增要求，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能生成的代码文件扩展名，尤其是 SVG 文件\n code_extensions = {\n '.svg', '.py', '.html', '.htm', '.js', '.css', \n '.ts', '.jsx', '.tsx', '.xml', '.json'\n }\n \n found_code_file = False\n \n if not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n found_code_file = True\n try:\n # 读取文件时忽略编码报错，避免因文件原本编码错误导致抛出异常\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 检查起始第一行是否包含指定的编码声明字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法读取，则认为未满足要求\n return False\n \n return found_code_file","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 若工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n \n # 定义目标注释字符串\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 典型代码文件的扩展名集合，其中主要包含本任务涉及到的 .svg\n code_exts = {'.svg', '.py', '.js', '.html', '.css', '.java', '.cpp', '.c', '.h', '.ts', '.go', '.rs'}\n \n found_code_file = False\n \n # 遍历工作目录寻找代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的前15行进行检查，以覆盖\"开头处\"的要求\n lines = [f.readline() for _ in range(15)]\n except Exception:\n # 如果读取异常（如编码不支持等），稳妥起见跳过该文件（极小概率）\n continue\n \n # 检查这些行中是否包含目标注释字符串\n has_comment = False\n for line in lines:\n if line and target_comment in line:\n has_comment = True\n break\n \n # 只要有一份代码文件未包含该注释，则判定要求未满足\n if not has_comment:\n return False\n \n # 如果找到了至少一份代码文件且所有的代码文件都符合要求，则返回 True\n return found_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令及一般情况，推断可能的代码文件扩展名\n code_extensions = {'.svg', '.json', '.md', '.py', '.html', '.css', '.js', '.ts', '.xml'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果只要有一个代码文件不包含 \"——\"，就认为不满足要求\n if \"——\" not in content:\n return False\n except Exception:\n # 如果文件不可读，也视为失败\n return False\n \n # 如果找到了至少一个代码文件且都通过了验证，则返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见的源文件扩展名集合\n source_exts = {\n '.py', '.svg', '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.go', '.rs', '.sh', \n '.rb', '.php', '.xml', '.md', '.json'\n }\n \n # 定义提取各语言常见注释的正则表达式\n comment_patterns = [\n re.compile(r'<!--.*?-->', re.DOTALL), # XML / HTML / SVG / MD 格式注释\n re.compile(r'/\\*.*?\\*/', re.DOTALL), # 多行注释 (/* ... */)\n re.compile(r'(?<![\"\\':])//.*'), # 单行注释 (// ...)，简单排除 URL 或字符串情况\n re.compile(r'(?<![\"\\'])#.*') # 单行注释 (# ...)，简单排除字符串情况\n ]\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果文件读取失败（例如非 UTF-8 编码），跳过该文件\n continue\n \n # 在当前文件内容中寻找所有匹配的注释\n for pattern in comment_patterns:\n comments = pattern.findall(content)\n for comment in comments:\n # 检查这段注释中是否存在 emoji\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的代码文件类型（包含源文件、JSON配置、MD文档以及可能的备份变体如.svg.bak）\n target_exts = ['.svg', '.json', '.md']\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 只要文件名包含目标扩展名，就认为是需要检查的代码或相关文本文件\n if any(ext in file for ext in target_exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # splitlines() 能兼容各类系统的换行符，且避免文件末尾正常的换行符被误判为空行\n lines = content.splitlines()\n for line in lines:\n # 如果一行全部由空白字符组成，或者完全没有字符，strip() 返回空字符串，即认为是空行\n if not line.strip():\n return False\n except Exception:\n # 忽略在异常情况或者编码不符时导致的文件读取失败\n pass\n \n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"炸场～\"这个语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 清除回复内容开头和结尾可能存在的空白字符\n cleaned_response = response.strip()\n \n # 检查回复的起始位置是否以要求的语气词开头，兼容全角和半角波浪号\n if cleaned_response.startswith(\"炸场～\") or cleaned_response.startswith(\"炸场~\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_str = \"<!-- 几何切割Z标识 -->\"\n # 根据用户指令内容推断可能生成的代码文件后缀类型\n code_extensions = {'.svg', '.json', '.md'}\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 检查文件是否是以推断出的代码文件后缀结尾的\n if any(file.lower().endswith(ext) for ext in code_extensions):\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 若发现某个被认定的代码文件未包含该标识，立即判定为不满足要求\n if target_str not in content:\n return False\n except Exception:\n # 如果由于编码等原因无法读取此文件，则选择忽略异常继续遍历\n pass\n \n # 如果工作区内找不到任何相关的代码文件，视作未满足要求\n if not found_code_file:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"现在来丰富一下周围的背景空间。请在'Z'的周围随机散落一些细小的几何碎片（比如很小的三角形、倾斜的短线或十字星）。这些小元素也用相应的渐变色或半透明颜色，看起来像是主体字母在碎裂时崩出的能量火花。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\"炫爆～\"这个语气词；第二，新增要求，生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件扩展名，特别关注针对当前任务的 .svg 格式\n code_exts = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', \n '.svg', '.xml', '.java', '.c', '.cpp', '.cs', '.go', \n '.rs', '.php', '.rb', '.sh'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理判定为代码格式的文件\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 以避免可能存在的 BOM 干扰第一行的读取\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline()\n \n # 检查第一行是否含有指定的编码声明\n if not first_line or '-*- coding: utf-8 -*-' not in first_line:\n return False\n \n has_code_file = True\n except Exception:\n # 如果由于编码或其他异常原因无法读取判定为代码的文件，视为验证失败\n return False\n \n # 如果遍历完成没有发现违规文件，且至少找到并验证了一个代码文件，则视为通过\n return has_code_file","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户指令内容，推断涉及的代码文件扩展名主要包含 SVG, JSON 和 Markdown\n target_exts = {'.svg', '.json', '.md'}\n checked_files_count = 0\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_exts:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 截取前10行作为“开头处”的检测范围\n lines = [f.readline() for _ in range(10)]\n except Exception:\n # 如果无法按utf-8读取文本，跳过该文件\n continue\n \n checked_files_count += 1\n \n # 检查前10行内是否包含指定的目标注释文本\n found = any(target_comment in line for line in lines if line)\n \n if not found:\n return False\n\n # 至少要有一份目标代码文件被生成并验证通过，才能视为满足要求\n return checked_files_count > 0","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断可能的代码文件后缀及备份文件\n code_extensions = {'.svg', '.json', '.md', '.html', '.xml', '.css', '.js', '.py', '.txt', '.bak'}\n file_found = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选推断为代码文件、配置文件或其对应备份的文件\n if ext in code_extensions or 'backup' in file.lower() or 'bak' in file.lower():\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查文件内容是否包含 \"——\"\n if \"——\" not in content:\n return False\n except Exception:\n # 遇到非文本或者无法用UTF-8解码的文件读取异常，视作不满足要求\n return False\n \n # 只要存在至少一个代码文件，且上述循环没有因为缺少 \"——\" 触发 return False，则满足要求\n return file_found","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断常见的代码源文件后缀名\n source_exts = {\n '.svg', '.html', '.htm', '.xml', \n '.py', '.sh', \n '.js', '.ts', '.css', '.java', '.c', '.cpp', '.h', '.cs'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于编码等原因无法读取，则跳过\n continue\n \n # 提取不同类型源文件内的全部注释文本\n comments = []\n if ext in ['.svg', '.html', '.htm', '.xml']:\n # HTML/SVG/XML 风格注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n elif ext in ['.py', '.sh']:\n # 单行注释 #\n comments.extend(re.findall(r'#.*', content))\n if ext == '.py':\n # Python 多行注释/文档字符串\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n elif ext in ['.js', '.ts', '.css', '.java', '.c', '.cpp', '.h', '.cs']:\n # C/JS 风格单行和多行注释\n comments.extend(re.findall(r'//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n # 遍历所提取的每个注释字符串\n for comment in comments:\n for char in comment:\n # 使用人类专家给定的 emoji 校验函数\n if is_emoji(char):\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 结合项目背景推断代码文件，并列举主流的代码文件类型。\n # 排除 .md、.txt 等用于撰写文档且排版离不开空行的文本文档。\n code_extensions = {\n '.svg', '.json', '.xml', '.html', '.css', '.js', '.ts',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go',\n '.rs', '.php', '.rb', '.sh', '.bat', '.jsx', '.tsx'\n }\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 判定该行如果只包含空白字符或者什么也没有，即为空行\n if line.strip() == \"\":\n return False\n except Exception:\n # 若文件遇到读取错误或无法用 utf-8 解析，则跳过\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"炫爆～\"这个语气词","validation_code":"import re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI的回复中，是否在每一段给用户的回复前都加上了\"炫爆～\"语气词。\n \"\"\"\n # 移除 markdown 格式的代码块（代码块内部不应被视为“给用户的回复段落”进行语气词检查）\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\n \n # 按照双换行符进行段落切分（Markdown标准段落切分方式）\n paragraphs = [p.strip() for p in text_without_code.split('\\n\\n') if p.strip()]\n \n # 如果回复中不存在任何非代码块的普通文本段落，则判定为不满足\n if not paragraphs:\n return False\n \n # 检查每一个段落是否均以\"炫爆～\"开头\n for p in paragraphs:\n if not p.startswith(\"炫爆～\"):\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_string = \"<!-- 几何切割Z标识 -->\"\n \n # 根据任务和常见的编程约定，定义典型的代码文件后缀\n code_extensions = {\n '.svg', '.html', '.htm', '.xml', '.css', '.js', '.jsx', '.ts', '.tsx',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \n '.rs', '.php', '.rb', '.sh'\n }\n \n found_code_file = False\n \n # 遍历当前工作区的所有目录及文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证，忽略 .json, .md 以及图片等非代码格式\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查目标注释字符串是否包含在该文件内容中\n if target_string not in content:\n return False\n except Exception:\n # 若文件存在编码读取异常（例如并非纯文本等意外情况），视为验证不通过\n return False\n \n # 必须至少生成了一个代码文件，且全部验证通过才返回 True\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推测可能生成的代码文件后缀，涵盖了svg、json、md等以及常见的代码格式\n code_extensions = {\n '.py', '.js', '.ts', '.html', '.css', '.json', \n '.xml', '.svg', '.md', '.java', '.cpp', '.c', \n '.h', '.sh'\n }\n\n # 确保给定的项目路径确实存在\n if not os.path.isdir(workspace_path):\n return False\n\n # 遍历当前项目地址下所有的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 提取出每一行开头的空白字符（即缩进部分）\n leading_whitespace = \"\"\n for char in line:\n if char == ' ' or char == '\\t':\n leading_whitespace += char\n else:\n break\n \n # 检查缩进部分中是否包含了制表符\n if '\\t' in leading_whitespace:\n return False\n except Exception:\n # 若文件存在编码等问题无法读取，予以跳过继续检查下一个\n continue\n \n # 所有代码文件的所有行均通过检查，没有发现使用制表符缩进\n return True","tags":["Style","Layout"]}]},{"round_id":5,"instruction":"下一步，给深色的背景底增加一点质感。请利用SVG的 `<pattern>` 标签，在背景的最底层铺上一层隐约可见的细网格线或者点阵图案，颜色非常暗淡且带有较高透明度，营造出一种赛博空间的数字感。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\"飒炸～\"这个语气词；第二，新增要求，你提供的每个代码文件里，注释内容须100%使用中文撰写，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件拓展名（由于指令核心是SVG，着重于Web和主流语言相关后缀，不包含纯文档与配置文件如.md、.json）\n code_exts = {'.svg', '.xml', '.py', '.js', '.html', '.css', '.c', '.cpp', '.java', '.go', '.rs', '.sh', '.php'}\n has_checked_code = False\n\n # 1. 检查工作目录中实际生成的代码文件\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n # 排除此前多轮对话中可能生成的备份文件\n if file.endswith('.bak') or '.bak.' in file:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 无论文件是否为空，只要是生成的代码文件，首行就应当具有该声明\n if not first_line:\n return False\n \n has_checked_code = True\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n pass\n\n # 2. 检查回复(response)中可能输出的Markdown代码块\n # 使用正则匹配所有代码块\n blocks = re.findall(r'```([a-zA-Z0-9]*)\\s*\\n(.*?)```', response, re.DOTALL)\n for lang, code_content in blocks:\n lang = lang.strip().lower()\n # 筛选特定语言或未显式声明语言但是疑似SVG/XML标签的代码块\n if lang in {'svg', 'xml', 'python', 'py', 'javascript', 'js', 'html', 'css', 'c', 'cpp', 'java'} or (lang == '' and '<svg' in code_content):\n lines = code_content.split('\\n')\n # 过滤纯空代码块的情况\n if not lines or (len(lines) == 1 and not lines[0].strip()):\n continue\n \n has_checked_code = True\n # 获取代码块首行内容进行校验\n first_line = lines[0]\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n\n # 只要生成了代码并全部检查通过即返回True；若没生成任何代码则默认要求未被满足\n return has_checked_code","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_string = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 定义常见代码文件的后缀，针对该任务重点是 .svg 格式\n code_extensions = {'.svg', '.py', '.html', '.css', '.js', '.ts', '.c', '.cpp', '.java', '.go'}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n found_in_file = False\n # 检查目标字符串是否在文件开头处（取前15行给予合理余量）\n for line in lines[:15]:\n if target_string in line:\n found_in_file = True\n break\n \n # 如果当前代码文件开头处没有该句注释，则未满足要求\n if not found_in_file:\n return False\n except Exception:\n # 如果遇到文件无法以utf-8读取等异常，视为验证不通过\n return False\n \n # 如果项目中没有任何代码文件，则也视为不符合该任务场景的期望\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断的可能的代码文件扩展名\n code_extensions = ['.svg', '.html', '.xml']\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 若文件内容中不包含指定的字符串 \"——\"，则直接判定为不满足要求\n if \"——\" not in content:\n return False\n except Exception:\n # 出现无法读取等异常时保守判定为不通过\n return False\n \n # 如果找到了至少一个代码文件，且均未在前面返回False，则表示满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return True\n \n # 定义不同类型源文件对应的注释提取正则表达式\n patterns = {\n 'html_xml_svg_md': [\n re.compile(r'<!--(.*?)-->', re.DOTALL)\n ],\n 'py_sh': [\n re.compile(r'#(.*?)$', re.MULTILINE), \n re.compile(r'\"\"\"(.*?)\"\"\"', re.DOTALL), \n re.compile(r\"'''(.*?)'''\", re.DOTALL)\n ],\n 'c_style': [\n re.compile(r'/\\*(.*?)\\*/', re.DOTALL), \n re.compile(r'//(.*?)$', re.MULTILINE)\n ]\n }\n\n # 依据指令推断及常规情况给出的常见源代码文件扩展名映射\n ext_mapping = {\n '.svg': 'html_xml_svg_md',\n '.html': 'html_xml_svg_md',\n '.xml': 'html_xml_svg_md',\n '.md': 'html_xml_svg_md',\n '.py': 'py_sh',\n '.sh': 'py_sh',\n '.js': 'c_style',\n '.ts': 'c_style',\n '.css': 'c_style',\n '.java': 'c_style',\n '.c': 'c_style',\n '.cpp': 'c_style',\n '.h': 'c_style',\n '.json': 'c_style' # JSON通常无注释，但部分环境支持C风格注释，故以c_style检查\n }\n\n # 遍历当前工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in ext_mapping:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 对于无法读取或无法由 utf-8 解码的非标准源文件直接跳过\n continue\n \n # 获取当前代码文件适用的正则提取器集合\n applicable_patterns = patterns[ext_mapping[ext]]\n \n for ptn in applicable_patterns:\n comments = ptn.findall(content)\n for comment in comments:\n # 遍历抽取的每段注释内容中的各个字符\n for char in comment:\n # is_emoji为预置函数，判断单个字符是否为emoji\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 定义需要检查的常见代码文件扩展名\n code_extensions = {'.svg', '.json', '.py', '.js', '.html', '.css', '.xml', '.sh', '.bat'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.read().splitlines()\n \n has_code_file = True\n # 遍历判断是否存在空行（或仅包含空白字符的行）\n for line in lines:\n if line.strip() == '':\n return False\n except Exception:\n # 忽略读取异常的文件（例如编码不对齐等异常情况）\n continue\n\n # 只要确实检查到了代码文件，且并未在任何代码文件里发现空行，即视为满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"飒炸～\"这个语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否以指定语气词\"飒炸～\"开头。\n \"\"\"\n if not response:\n return False\n \n # 去除首尾的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 检查回复是否以\"飒炸～\"开头\n return cleaned_response.startswith(\"飒炸～\")","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_str = \"<!-- 几何切割Z标识 -->\"\n \n # 结合题目上下文推断，定义涵盖SVG设计及配置管理可能的代码/文本文件后缀\n code_extensions = {\n '.svg', '.xml', '.html', '.htm', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.py', '.json', '.md', '.txt', '.sh'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n has_code_file = True\n # 如果任何一个代码文件没有包含该标识，则返回 False\n if target_str not in content:\n return False\n except Exception:\n # 如果文件无法用 utf-8 读取，或者出现其他IO异常，直接跳过\n continue\n \n # 如果找到了至少一个代码文件，并且上面的循环没有因为缺失标识而返回 False，则代表全部包含\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进","validation_code":"import os\nimport re\n\ndef check_indentation_without_tabs(response: str, workspace_path: str) -> bool:\n # 匹配行首空白字符（缩进）的正则表达式\n indent_pattern = re.compile(r'^[ \\t]+')\n \n # 常见的代码和文本文件后缀名集合\n code_extensions = {\n '.py', '.java', '.js', '.ts', '.html', '.css', '.c', '.cpp', '.hpp', '.h', \n '.cs', '.php', '.rb', '.go', '.rs', '.json', '.svg', '.md', '.txt', '.sh', '.xml'\n }\n \n # 1. 检查 workspace_path 中的文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n match = indent_pattern.match(line)\n # 如果行首的空白字符(缩进)中包含了制表符，则违反要求\n if match and '\\t' in match.group():\n return False\n except Exception:\n pass\n \n # 2. 检查 response 中的代码块\n # 匹配 markdown 格式的代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n for line in block.split('\\n'):\n match = indent_pattern.match(line)\n # 同样判断行首缩进是否包含制表符\n if match and '\\t' in match.group():\n return False\n \n # 如果全都没有使用制表符缩进，则满足要求\n return True","tags":["Style","Layout"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":6,"instruction":"太静态了，我们加点动画效果吧。请使用 CSS 或 SVG 内置的 `<animate>` 标签，让中心的主体字母'Z'整体产生一个缓慢的、上下悬浮的呼吸效果。同时，让周围散落的那些小能量碎片也进行轻微的旋转或透明度闪烁。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\"燃爆～\"这个语气词；第二，新增要求，每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，说明未生成任何项目文件\n if not os.path.exists(workspace_path):\n return False\n\n # 常见代码文件扩展名，特别包含本任务的主体代码格式 .svg\n # 排除 .json 和 .md，以防它们被当做非代码文件处理而引起误判\n code_extensions = {\n \".svg\", \".py\", \".js\", \".ts\", \".html\", \".css\", \".xml\",\n \".java\", \".cpp\", \".c\", \".h\", \".cs\", \".go\", \".rs\", \n \".php\", \".rb\", \".sh\"\n }\n\n has_checked_any_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，防止遍历 .git 等文件夹内的文件\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 如果文件为空或者第一行没有所需的编码声明\n if not first_line or \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n \n has_checked_any_code_file = True\n except Exception:\n # 如果读取出现异常，保守返回 False\n return False\n\n # 确保至少检查到了一个代码文件且都符合要求\n return has_checked_any_code_file","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 用户要求在代码文件开头处另起一行添加的明确注释文本\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 推断可能的代码文件类型，涵盖前端相关文件和常见编程语言扩展名\n code_exts = {\n '.svg', '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx',\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果UTF-8解码失败，尝试使用忽略错误的模式读取\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n return False\n\n # 过滤掉空行，以便准确判断文本是否在“文件开头处”\n lines = [line.strip() for line in lines if line.strip()]\n \n # 检查前10行非空行内是否存在指定的注释声明\n found_comment = False\n for line in lines[:10]:\n if target_comment in line:\n found_comment = True\n break\n \n # 如果某一个代码文件未包含该注释，或者不在开头处，立即返回 False\n if not found_comment:\n return False\n \n # 只要存在至少一个代码文件且全都满足条件即为 True；如果整个项目无代码文件则判断为 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀集合，包含 SVG, HTML, CSS 等任务可能产生的文件类型\n # 排除了 .md, .json 等文档及配置文件\n code_exts = {\n '.svg', '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', \n '.php', '.rb', '.sh', '.xml', '.vue', '.svelte'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n \n # 遍历工作目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名并转换为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查内容是否包含要求中的中划线\n if \"——\" not in content:\n return False\n except Exception:\n # 如果无法读取文件或编码不符合，则视为检查不通过\n return False\n \n # 如果找到了至少一个代码文件，并且全包含 \"——\"，返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码源文件后缀名\n source_extensions = {\n '.svg', '.html', '.htm', '.xml', \n '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', '.php'\n }\n \n # 匹配各类常见注释的正则表达式\n comment_patterns = [\n r'<!--[\\s\\S]*?-->', # XML / HTML / SVG 注释\n r'/\\*[\\s\\S]*?\\*/', # 多行注释 (C, C++, Java, JS, CSS)\n r'//.*', # 单行注释 (C, C++, Java, JS)\n r'#.*' # 单行注释 (Python, Shell 等)\n ]\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 在文件内容中寻找所有匹配的注释\n for pattern in comment_patterns:\n comments = re.findall(pattern, content)\n for comment in comments:\n # 遍历注释字符串中的每一个字符，检查是否为 emoji\n for char in comment:\n if is_emoji(char):\n return False\n except Exception:\n # 忽略读取异常的文件\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n checked_any_file = False\n \n # 优先检查工作区生成的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n checked_any_file = True\n \n # 逐行检查是否包含空行\n for line in lines:\n if not line.strip():\n return False\n except Exception:\n # 忽略二进制文件、非UTF-8文件等不可读取的文件\n continue\n \n # 如果工作区未能读取到任何代码文件，则降级从模型回复中提取代码块检查\n if not checked_any_file:\n code_blocks = re.findall(r'```[a-zA-Z0-9]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n # 剥离首尾多余换行符防止分割时误判，然后再切分行\n lines = block.strip('\\n').split('\\n')\n for line in lines:\n if not line.strip():\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"燃爆～\"这个语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否在最前面加上了\"燃爆～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容两端的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 判断回复是否以\"燃爆～\"开头\n return cleaned_response.startswith(\"燃爆～\")","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 涵盖当前任务场景可能的输出类型，以及常用代码文件的扩展名集合\n code_extensions = {\n '.svg', '.xml', '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx',\n '.py', '.json', '.md', '.txt', '.sh', '.java', '.c', '.cpp', '.h', \n '.cs', '.php', '.rb', '.go', '.rs', '.swift'\n }\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n found_code_file = True\n # 如果有任何一个代码文件不包含该标识，则立即判定为未满足要求\n if \"<!-- 几何切割Z标识 -->\" not in content:\n return False\n except Exception:\n # 忽略不可读或非UTF-8编码的文件\n continue\n \n # 如果找到了至少一个代码文件且都包含标识，则返回True，否则返回False\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 检查 response 中的 Markdown 代码块\n code_blocks = re.findall(r'```.*?\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n for line in block.split('\\n'):\n # 提取行首空白字符部分（缩进部分）\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n if '\\t' in leading_whitespace:\n return False\n\n # 检查工作区中的代码文件\n # 根据用户指令上下文，涉及文件类型主要为 .svg, .json, .md 以及潜在的相关其他文件\n code_extensions = {'.svg', '.json', '.md', '.css', '.html', '.xml', '.js', '.py', '.txt'}\n \n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 提取行首空白字符部分（缩进部分）\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n if '\\t' in leading_whitespace:\n return False\n except Exception:\n # 如遇编码等读取问题则跳过当前文件\n pass\n \n return True","tags":["Style","Layout"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":7,"instruction":"继续增加高级的交互性。我想加一个鼠标悬停（hover）的动画效果：当鼠标悬停到主体'Z'所在的区域时，构成'Z'的那些几何切割块会向外微微散开（产生几像素的位移，表现出解体感），同时背后的发光效果变得更亮；当鼠标移开时，碎片平滑地拼合回原样。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\"超炫～\"这个语气词；第二，新增要求，务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令中涉及的文件，推断可能的代码文件/文本文件后缀名\n code_extensions = {'.svg', '.json', '.md', '.log', '.xml', '.html', '.css', '.js', '.py'}\n \n has_checked_files = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n # 筛选出属于代码类型的文件\n if ext in code_extensions:\n has_checked_files = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 读取起始第一行\n first_line = f.readline()\n \n # 检查第一行是否包含声明编码的特定字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于某种原因无法读取文件，认为不符合要求\n return False\n \n # 如果项目下没有生成任何相关的代码文件，也不视为满足任务目标\n return has_checked_files","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 目标注释字符串（不包含可能被系统指令引入的中文双引号）\n target_string = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常见的代码文件后缀集合，包括本项目强相关的 SVG 和常见的源文件格式\n code_extensions = {'.svg', '.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java'}\n \n code_file_found = False\n \n # 确保给定的项目路径存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n try:\n # 忽略可能遇到的一些非常规字符导致的编码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n \n # 检查前 10 行，以兼容第一行的编码声明等内容\n found_in_header = False\n for line in lines[:10]:\n if target_string in line:\n found_in_header = True\n break\n \n # 只要有一个代码文件的头部没有包含指定的注释文字，即不符合条件\n if not found_in_header:\n return False\n except Exception:\n # 如果由于任何系统原因文件读取异常，为保证严谨视作验证失败\n return False\n \n # 必须要至少生成了一份或以上的代码文件，全部通过验证后方能判定满足\n return code_file_found","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断是否为代码文件或其备份文件\n def is_target_code_file(filename: str) -> bool:\n code_exts = {'.svg', '.html', '.css', '.js', '.ts', '.py'}\n filename = filename.lower()\n for ext in code_exts:\n if filename.endswith(ext):\n return True\n # 处理备份文件可能带有的后缀\n if filename.endswith(ext + '.bak') or filename.endswith(ext + '.backup'):\n return True\n # 处理形如 logo_backup.svg 或类似包含备份字眼且包含代码扩展名的情况\n if ext in filename and ('bak' in filename or 'backup' in filename):\n return True\n return False\n\n file_count = 0\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n if is_target_code_file(file):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n file_count += 1\n # 验证是否含有 \"——\"\n if \"——\" not in content:\n return False\n except Exception:\n # 如果因权限或编码问题导致无法读取，则安全跳过该文件\n pass\n \n # 必须至少有一份符合条件的代码文件存在，且全都满足包含 \"——\" 的要求\n return file_count > 0","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的源代码文件后缀集合\n source_extensions = {\n '.svg', '.xml', '.html', '.htm',\n '.js', '.ts', '.jsx', '.tsx', \n '.css', '.scss', '.less',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.sh', '.rb'\n }\n \n # 遍历代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取源文件内容，忽略无法解码的字符以增加鲁棒性\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n comments = []\n # 针对不同扩展名进行针对性的注释提取\n if ext in {'.svg', '.xml', '.html', '.htm'}:\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # 提取内部的 <style> 和 <script> 并读取其中潜藏的注释\n styles = re.findall(r'<style[^>]*>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n for style in styles:\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', style))\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for script in scripts:\n comments.extend(re.findall(r'(?<!:)//.*', script))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', script))\n \n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', \n '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs'}:\n # 使用 (?<!:) 避免将 url 中的 // 误提取为注释\n comments.extend(re.findall(r'(?<!:)//.*', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n elif ext in {'.py', '.sh', '.rb'}:\n comments.extend(re.findall(r'#.*', content))\n \n # 遍历提取到的所有注释内容\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n except Exception:\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区不存在，或者没有任何文件生成，也算不违背该特定要求\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤掉系统隐藏文件，例如 .git 文件夹内文件或 .DS_Store 等\n if file.startswith('.'):\n continue\n \n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 UTF-8 格式读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n # 若抛出解码异常，说明可能是图片(.png)等二进制文件，跳过即可\n continue\n except Exception:\n continue\n \n lines = content.split('\\n')\n for i, line in enumerate(lines):\n if line.strip() == '':\n # 允许因文件末尾带有单个正常换行符而导致 split 后列表最后一个元素为空字符串\n if i == len(lines) - 1 and line == '':\n continue\n # 发现任何一个不应存在的空行，则判定为没有满足要求\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"超炫～\"这个语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import re\n \n response_stripped = response.strip()\n if not response_stripped:\n return False\n \n # 如果整体回复中根本就没有出现该语气词，直接判定失败\n if \"超炫～\" not in response:\n return False\n \n # 移除回复中的所有的代码块内容\n text_no_code = re.sub(r'```.*?```', '\\n\\n', response, flags=re.DOTALL)\n \n # 按照双换行划分段落\n raw_paragraphs = [p.strip() for p in text_no_code.split('\\n\\n') if p.strip()]\n \n normal_paragraphs = []\n for p in raw_paragraphs:\n # 跳过首字符为 '<' 的 XML/HTML 格式的段落\n if p.startswith('<'):\n continue\n \n # 跳过由于前面轮次要求的 Review.md 格式输出的相关段落\n if 'Code Quality Review:' in p or 'Strengths:' in p or 'Issues:' in p or 'Recommendation:' in p:\n continue\n \n lines = [line.strip() for line in p.split('\\n') if line.strip()]\n if not lines:\n continue\n \n # 检查是否是纯列表形式组成的段落，如果是则跳过\n if all(re.match(r'^([\\-\\*\\+]|\\d+\\.)\\s', line) for line in lines):\n continue\n \n normal_paragraphs.append(p)\n \n # 如果没有任何普通的文本段落，只有特殊格式时，此时只要含有目标词汇即可认为满足\n if not normal_paragraphs:\n return True\n \n # 所有保留下来的普通文本段落，都必须以 \"超炫～\" 开头\n for p in normal_paragraphs:\n # 清除段落开头可能带有的 Markdown 修饰符或空格\n clean_p = re.sub(r'^[\\#\\*\\-\\>\\s]+', '', p)\n # 检查原始段落或清理后的段落是否以规定词汇开头\n if not (clean_p.startswith(\"超炫～\") or p.startswith(\"超炫～\")):\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见代码文件的扩展名集合（排除 .md, .json, .log 等文档与数据文件）\n code_exts = {\n '.svg', '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.xml', '.sh', '.bat'\n }\n \n target_str = \"<!-- 几何切割Z标识 -->\"\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 如果有任意一个代码文件不包含目标字符串，直接返回 False\n if target_str not in content:\n return False\n except Exception:\n # 如果出现读取异常（如编码错误），无法验证包含该标识，亦视为失败\n return False\n \n # 确保至少检查过一个代码文件，否则返回 False 避免空项目的假阳性通过\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证代码项目中生成的代码文件是否完全使用纯空格缩进，不包含任何制表符(\\t)。\n \"\"\"\n # 根据多轮指令，推断代码文件可能包含 svg, json 等，涵盖常见的代码扩展名\n code_extensions = {'.svg', '.html', '.css', '.js', '.ts', '.json', '.xml', '.py', '.md'}\n has_code_files = False\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 获取每一行开头的空白字符（即缩进部分）\n # line.lstrip() 去掉左侧所有空白符，原长度减去 lstrip 后的长度即为左侧空白符的长度\n leading_whitespaces = line[:len(line) - len(line.lstrip())]\n \n # 检查缩进部分是否包含制表符 \\t\n if '\\t' in leading_whitespaces:\n return False\n except UnicodeDecodeError:\n # 忽略非UTF-8编码的可能引发读取异常的文件\n continue\n except Exception:\n # 忽略其他读取异常\n continue\n \n # 确保至少扫描到了代码文件，并且没有触发制表符缩进的检查\n return has_code_files","tags":["Style","Layout"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":8,"instruction":"最后一步，来做收尾和完善。在主体字母的正下方增加一行无衬线风格的纯文本品牌名“ZENITH”，要求字距拉宽（letter-spacing），文字颜色与 logo 整体的渐变色相呼应。同时，去除原先固定长宽的绝对像素尺寸，改为利用 `viewBox` 实现完美的响应式，使其能在任何尺寸的网页容器中等比缩放。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\"封神～\"这个语气词；第二，新增要求，确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项，该要求后续所有轮次始终遵循。","instruction_following_checklist":[{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义预期可能生成的代码文件扩展名集合\n # 主要针对 SVG 图像设计及配套的 Web 页面、样式、脚本和常见程序代码\n code_extensions = {'.svg', '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含指定的编码声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法按 utf-8 读取或发生了其他读取异常，直接判定为失败\n return False\n \n # 如果找到了至少一个代码文件且全都验证通过，则返回 True；否则返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n target_sentence = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户指令推断，可能的代码文件扩展名\n code_extensions = {\n '.svg', '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs'\n }\n \n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的前10行（确保覆盖“开头处，另起一行”的要求）\n lines = [f.readline() for _ in range(10)]\n \n found = False\n for line in lines:\n if not line: # 遇到文件末尾(EOF)\n break\n if target_sentence in line:\n found = True\n break\n \n # 如果该代码文件开头处不包含目标注释，则要求未满足\n if not found:\n return False\n except Exception:\n # 如果文件读取出现异常，保守返回 False\n return False\n \n # 确保项目中确实存在生成的代码文件，且均满足了条件\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的各个代码文件中，均需含有 \"——\"","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断可能的代码文件类型（例如 SVG、HTML、CSS、JS 等）\n # 排除 .md、.log 等通常不被视作核心源文件的文档或日志文件\n code_extensions = {\n '.svg', '.html', '.xml', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.cs'\n }\n \n found_code_files = False\n \n # 首先检查工作区中生成的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查各个代码文件是否含有 \"——\"\n if \"——\" not in content:\n return False\n except Exception:\n # 如果存在无法以 utf-8 读取的文件，可能非文本源文件，暂且忽略\n pass\n \n # 如果工作区中没有发现实际保存的代码文件，备选检查 response 中可能生成的代码块\n if not found_code_files:\n # 匹配 markdown 代码块，格式：```语言\\n代码内容\\n```\n code_blocks = re.findall(r'```([a-zA-Z0-9]*)\\n(.*?)```', response, re.DOTALL)\n if not code_blocks:\n # 既没有实际文件，也没有代码块，无法证明满足要求\n return False \n \n for lang, code_content in code_blocks:\n lang = lang.strip().lower()\n # 排除非代码性质的内容块\n if lang in {'markdown', 'md', 'json', 'log', 'text', 'yaml', 'yml', 'bash', 'sh'}:\n continue\n \n found_code_files = True\n if \"——\" not in code_content:\n return False\n \n # 确保只要发现了代码文件，且它们都顺利通过了包含 \"——\" 的检查，即可返回 True\n return found_code_files","tags":["Content","Keyword"]},{"constraint":"请保证你的所有源文件在注释区域均不涉及 emoji表情符号","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义我们要检查的源文件后缀（排除了 .md, .json, .log 等非代码源文件）\n source_extensions = {\n '.svg', '.xml', '.html', '.htm', \n '.css', '.js', '.ts', \n '.py', '.java', '.c', '.cpp', '.cs', '.go'\n }\n\n # 预编译各种常见代码注释的正则表达式（尽量覆盖各种注释格式）\n # XML/HTML 风格: <!-- ... -->\n xml_comment_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\n # C/CSS/JS 风格多行注释: /* ... */\n c_multi_pattern = re.compile(r'/\\*(.*?)\\*/', re.DOTALL)\n # C/JS 风格单行注释: // ...\n c_single_pattern = re.compile(r'//(.*?)$', re.MULTILINE)\n # Python/Shell 风格单行注释: # ...\n hash_pattern = re.compile(r'#(.*?)$', re.MULTILINE)\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in source_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取源文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略无法通过 utf-8 读取的文件\n continue\n \n comments = []\n \n # 根据文件类型提取对应格式的注释内容\n if ext in {'.svg', '.xml', '.html', '.htm'}:\n comments.extend(xml_comment_pattern.findall(content))\n # SVG中也可能包含 <style> 标签和 <script> 标签，因此顺带提取 C 风格注释\n comments.extend(c_multi_pattern.findall(content))\n comments.extend(c_single_pattern.findall(content))\n elif ext in {'.css'}:\n comments.extend(c_multi_pattern.findall(content))\n elif ext in {'.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go'}:\n comments.extend(c_multi_pattern.findall(content))\n comments.extend(c_single_pattern.findall(content))\n elif ext in {'.py'}:\n comments.extend(hash_pattern.findall(content))\n \n # 遍历提取到的每一个注释段落\n for comment in comments:\n # 遍历注释内的每个字符，调用 is_emoji() 进行判断\n for char in comment:\n if is_emoji(char):\n return False # 如果在注释区域发现任何emoji，则返回不满足\n\n # 所有源文件的注释区域均未发现 emoji，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 确保给定的项目路径存在\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n\n # 遍历代码项目的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # 尝试以 utf-8 格式读取文件内容\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except UnicodeDecodeError:\n # 如果引发解码错误，说明可能是二进制文件（如图片等），直接跳过\n continue\n except Exception:\n continue\n \n # 逐行检查是否包含空行\n for line in lines:\n # strip() 可以同时去除纯换行符和带有空格/制表符的空行\n if line.strip() == \"\":\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"在你输出的每一段给用户的回复前，务必先加上\"封神～\"这个语气词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import re\n \n if not response or not response.strip():\n return False\n \n lines = response.split('\\n')\n in_code_block = False\n text_lines = []\n \n # 逐行遍历，过滤掉代码块内的内容\n for line in lines:\n if line.strip().startswith('```'):\n in_code_block = not in_code_block\n # 插入一个空行标记，防止代码块前后的自然语言段落粘连\n text_lines.append(\"\")\n continue\n \n if not in_code_block:\n text_lines.append(line)\n \n text_content = '\\n'.join(text_lines)\n \n # 利用正则表达式按连续的空行（允许包含空格等空白字符）分隔段落\n paragraphs = re.split(r'\\n\\s*\\n', text_content)\n \n # 清理段落两端空白，并剔除空段落\n valid_paragraphs = [p.strip() for p in paragraphs if p.strip()]\n \n # 如果除了代码块没有任何自然语言段落，无法确认是否满足要求\n if not valid_paragraphs:\n return False\n \n # 验证每个自然语言段落是否都以 \"封神～\" 开头\n for p in valid_paragraphs:\n if not p.startswith(\"封神～\"):\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你生成的所有代码文件中都必须包含\"<!-- 几何切割Z标识 -->\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀，排除 .md, .log, .json 等文档及记录类文件\n code_extensions = {\n '.svg', '.html', '.htm', '.xml', '.css', '.scss', \n '.js', '.jsx', '.ts', '.tsx', '.py', '.java', \n '.c', '.cpp', '.h', '.hpp', '.sh', '.bash'\n }\n\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n # 排除常见的版本控制或依赖等非人工编写/生成的文件夹\n if '.git' in file_path or 'node_modules' in file_path:\n continue\n \n found_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if \"<!-- 几何切割Z标识 -->\" not in content:\n return False\n except UnicodeDecodeError:\n # 如果无法以 utf-8 解码，认为其不是期望的文本代码文件\n pass\n\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 1. 检查 workspace_path 下生成的代码文件\n # 根据用户指令中提到的SVG logo设计、JSON文件记录及前端特效，框定相关的代码文件类型\n code_extensions = {'.svg', '.json', '.html', '.css', '.js', '.xml'}\n \n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 匹配行首的前导空白字符（空格和制表符）\n match = re.match(r'^[ \\t]+', line)\n if match:\n leading_whitespace = match.group(0)\n # 如果前导缩进中包含制表符，返回 False\n if '\\t' in leading_whitespace:\n return False\n except Exception:\n continue\n\n # 2. 检查 response 回复文本中输出的代码块\n # 匹配各类代码块（含带语言标识如 ```svg, ```json 和不带标识的常规代码块）\n pattern = r'```[a-zA-Z0-9]*\\s*\\n(.*?)```'\n code_blocks = re.findall(pattern, response, flags=re.DOTALL)\n \n for block in code_blocks:\n for line in block.split('\\n'):\n match = re.match(r'^[ \\t]+', line)\n if match:\n leading_whitespace = match.group(0)\n if '\\t' in leading_whitespace:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我设计一个SVG格式的logo。首先，创建一个800x800的画布，背景色设为深邃的藏青色或深黑色。在画布正中间，用简单的粗线条或者单一颜色的多边形画一个大写字母'Z'。这只是个基础骨架，后面我们会一步步把它做得非常酷炫。另外有两个后续所有轮次都必须严格遵循的要求：第一，不要让你生成的任何代码文件里存在空行；第二，在你输出的每一段给用户的回复前，务必先加上\\\"呜哇～\\\"这个语气词。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件后缀，并包含用户指令中明确要生成的 .svg 格式\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', \\n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', \\n '.hpp', '.go', '.rs', '.rb', '.php', '.sh', '.bash', \\n '.svg', '.xml', '.vue', '.svelte'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法用 utf-8 读取或其他异常（包含空文件无内容的情况），判定为未满足\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义常见代码文件的扩展名集合，根据指令推断至少可能涉及.svg\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', \\n '.java', '.cpp', '.c', '.h', '.hpp', '.svg', '.go', \\n '.rs', '.php', '.rb', '.sh', '.xml'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 仅需读取前5行即可验证是否在文件\\\"开头处\\\"\\n lines = [f.readline() for _ in range(5)]\\n except Exception:\\n # 如果由于编码或其他原因无法读取文件内容，默认返回 False\\n return False\\n \\n # 检查前几行中是否包含目标注释字符串\\n found = False\\n for line in lines:\\n if not line:\\n break\\n if target_text in line:\\n found = True\\n break\\n \\n # 如果某个代码文件没有包含所需的注释，则总体判定为不满足要求\\n if not found:\\n return False\\n \\n # 如果项目中没有任何代码文件，说明没有可验证对象，这里将其视为未满足生成要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名，根据上下文包含.svg并排除.json/.md等非典型或不可注释的格式\\n code_exts = {\\n '.py', '.svg', '.js', '.ts', '.html', '.css', \\n '.java', '.cpp', '.c', '.h', '.cs', '.go', \\n '.rs', '.php', '.rb', '.sh'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n \\n # 遍历当前工作区路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 判断内容是否包含 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 无法读取或解码文件，视为无法验证该文件包含指定字符串\\n return False\\n \\n # 如果找到了至少一个代码文件且全都包含 \\\"——\\\"，返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n \\n # 1. 匹配脚本语言文件中的注释（如 Python, Shell, Ruby等）\\n if ext in {'.py', '.rb', '.sh'}:\\n comments.extend(re.findall(r'#.*', content))\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n \\n # 2. 匹配标记语言文件中的注释（包含 SVG, HTML, XML等）\\n elif ext in {'.html', '.htm', '.svg', '.xml'}:\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n # 3. 匹配 C风格及前端主流语言文件中的注释（如 C/C++, Java, Go, JS, TS, PHP, CSS等）\\n elif ext in {'.c', '.cpp', '.h', '.hpp', '.java', '.go', '.js', '.ts', '.php', '.css', '.scss', '.less'}:\\n comments.extend(re.findall(r'//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n else:\\n # 非源文件跳过检查\\n continue\\n \\n # 检查提取到的所有注释中是否包含 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推测相关的代码文件后缀\\n code_extensions = ['.svg', '.json', '.md']\\n has_code_file = False\\n \\n # 如果工作区路径不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 筛选符合推测后缀的代码文件\\n if any(file_name.lower().endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 判定是否包含空行（该行去掉空白符后为空）\\n if line.strip() == \\\"\\\":\\n return False\\n except Exception:\\n # 若遇文件读取或编码错误，则安全跳过继续检查其他文件\\n pass\\n \\n # 如果至少存在一个代码文件且全都经受住了空行检查，则符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"呜哇～\\\"这个语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断AI助手的回复开头是否包含“呜哇～”这个语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n return response.lstrip().startswith(\\\"呜哇～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错。现在我们要让这个'Z'具有“几何切割”的风格。请把单调的'Z'拆分成至少四五个紧密贴合的三角形或四边形拼接而成。每个几何块先用不同灰度的白色/浅灰色填充，让它初步具备被切割成多块的视觉感受。另外后续所有轮次都必须严格遵循两个要求：第一，在你输出的每一段给用户的回复前，务必先加上\\\"哇呜～\\\"这个语气词；第二，你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能生成的代码文件扩展名\\n code_extensions = {\\n '.svg', '.json', '.md', '.py', '.js', '.ts', \\n '.html', '.css', '.xml', '.txt', '.sh', '.yaml', '.yml'\\n }\\n file_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果无法按 utf-8 读取或者发生其他异常，视为不符合要求\\n return False\\n \\n # 确保工作区中至少检查过一个代码文件\\n return file_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_sentence = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 定义常见的代码文件扩展名，涵盖用户要求中可能生成的 .svg, .html, .py 等\\n code_extensions = {\\n \\\".py\\\", \\\".java\\\", \\\".c\\\", \\\".cpp\\\", \\\".cs\\\", \\\".js\\\", \\\".ts\\\", \\n \\\".html\\\", \\\".css\\\", \\\".svg\\\", \\\".sh\\\", \\\".xml\\\", \\\".go\\\", \\n \\\".rs\\\", \\\".php\\\", \\\".rb\\\"\\n }\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对代码文件进行检查\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n found = False\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 检查文件开头处，宽松地设定为前20行内\\n for _ in range(20):\\n line = f.readline()\\n if not line: # 到达文件末尾\\n break\\n if target_sentence in line:\\n found = True\\n break\\n except Exception:\\n # 如果发生读取错误（如非utf-8编码等情况），认为是不满足要求\\n return False\\n \\n # 如果发现某一份代码文件开头没有这句注释，立即返回 False\\n if not found:\\n return False\\n \\n # 必须至少存在一份代码文件，且所有代码文件都满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，主要文件格式包括 svg, json, md，同时加入常见的代码文件格式\\n target_extensions = {'.svg', '.json', '.md', '.py', '.html', '.css', '.js', '.xml'}\\n file_count = 0\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，如 .git\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果读取失败（例如非文本格式或编码不符），则跳过\\n continue\\n \\n file_count += 1\\n # 检查文件中是否包含指定的字符串 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n \\n # 如果找到了至少一个代码文件，并且所有的文件都满足条件，返回 True\\n return file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 根据用户指令及一般编程常识，圈定常见的源代码文件扩展名（重点关注本项目生成的 .svg）\\n source_exts = {\\n '.svg', '.py', '.js', '.ts', '.html', '.htm', \\n '.css', '.c', '.cpp', '.h', '.hpp', '.java', \\n '.go', '.rs', '.sh', '.rb', '.php'\\n }\\n \\n # 编译各类型注释的正则表达式\\n xml_comment_re = re.compile(r'<!--(.*?)-->', re.DOTALL)\\n c_block_comment_re = re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL)\\n c_line_comment_re = re.compile(r'//(.*?)$', re.MULTILINE)\\n hash_comment_re = re.compile(r'#(.*?)$', re.MULTILINE)\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n # 根据不同扩展名提取相应格式的注释文本\\n if ext in {'.svg', '.html', '.htm'}:\\n comments.extend(xml_comment_re.findall(content))\\n elif ext in {'.js', '.ts', '.css', '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', '.php'}:\\n comments.extend(c_block_comment_re.findall(content))\\n comments.extend(c_line_comment_re.findall(content))\\n # PHP可能还混用#注释\\n if ext == '.php':\\n comments.extend(hash_comment_re.findall(content))\\n elif ext in {'.py', '.sh', '.rb'}:\\n comments.extend(hash_comment_re.findall(content))\\n \\n # 遍历所提取出注释中的每一个字符，判断是否存在emoji\\n for comment_text in comments:\\n for char in comment_text:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件及文本文件扩展名\\n code_extensions = {\\n '.svg', '.json', '.md', '.py', '.js', '.html', '.css', '.xml', '.txt',\\n '.sh', '.yaml', '.yml', '.jsx', '.ts', '.tsx', '.vue', '.java', '.c',\\n '.cpp', '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 忽略隐藏文件，例如 .git 目录下的文件或操作系统生成的隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 逐行读取文件内容\\n for line in f:\\n # 判定空行：若剥离空格、制表符和换行符后为空，即为存在空行\\n if line.strip() == '':\\n return False\\n except Exception:\\n # 如果遇到无法以 utf-8 解码的二进制文件等异常情况，跳过即可\\n continue\\n \\n # 如果遍历完毕没有发现任何空行，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"哇呜～\\\"这个语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复首尾的空白字符和换行符\\n cleaned_response = response.strip()\\n \\n # 检查回复是否以指定语气词开头\\n return cleaned_response.startswith(\\\"哇呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，主要的代码文件应当是 .svg，同时加入其他常见的代码文件后缀\\n code_extensions = {\\n '.svg', '.xml', '.html', '.htm', '.py', '.js', '.ts', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.php', \\n '.go', '.rs', '.swift', '.kt', '.css', '.sh'\\n }\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n content = f.read()\\n except Exception:\\n # 如果存在无法以常规编码读取的二进制文件或异常情况，安全跳过\\n continue\\n except Exception:\\n continue\\n \\n # 检查要求的目标字符串是否在代码文件中\\n if \\\"<!-- 几何切割Z标识 -->\\\" not in content:\\n return False\\n \\n # 如果项目中没有任何代码文件，通常也不能算作满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，接下来给这些几何碎片上色。把灰白填充替换成赛博朋克风格的渐变色（比如亮青色到深蓝色的渐变，或者蓝紫渐变）。要求每一块多边形的渐变方向或角度稍微错开一点，利用色彩的高光和暗部，营造出强烈的立体感和水晶切割的感觉。另外后续所有轮次都必须严格遵循两个要求：第一，在你输出的每一段给用户的回复前，务必先加上\\\"酷喔～\\\"这个语气词；第二，请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令中提到的\\\"SVG格式的logo\\\"，推断核心代码文件为.svg\\n # 同时囊括可能出现的关联网页或脚本语言代码类型\\n code_exts = {'.svg', '.py', '.js', '.html', '.css', '.ts', '.sh'}\\n has_code_file = False\\n \\n # 遍历当前工作区所有生成的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理推断出属于代码类型的文件\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否严格添加了要求中的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果发生文件读取错误，视为不满足\\n return False\\n \\n # 如果找到了至少一个代码文件并且全满足要求则返回True，否则没有生成代码文件返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据用户指令中涉及的文件，设定可能被视为代码/文本文件的后缀\\n valid_exts = {'.svg', '.json', '.md', '.py', '.html', '.css', '.js', '.xml'}\\n \\n checked_any_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in valid_exts:\\n checked_any_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n lines = []\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for _ in range(10): # 提取前10行作为“开头处”\\n line = f.readline()\\n if not line:\\n break\\n lines.append(line)\\n except Exception:\\n # 如果有文件无法读取或者不是文本文件，安全跳过\\n continue\\n \\n # 检查该文件前几行中是否包含目标注释文字\\n found_target = any(target_text in line for line in lines)\\n \\n if not found_target:\\n return False\\n \\n # 如果扫描且验证了符合扩展名的文件，并且全部都包含该注释，则满足要求\\n return checked_any_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据用户指令内容推断，代码项目涉及的文件可能是 svg、json 和 md 文件\\n target_extensions = {'.svg', '.json', '.md'}\\n checked_files = 0\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果是备份文件（比如 .bak 结尾），如果原文件名的扩展名在目标内也可考虑，此处以当前扩展名为准\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 验证文件内容中是否含有要求字符串 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n \\n checked_files += 1\\n except Exception:\\n # 如果无法读取（极少情况的编码等异常），为严谨保守起见，视为未满足条件\\n return False\\n\\n # 确保有相关文件生成且都被检查通过\\n return checked_files > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对不同源文件扩展名定义对应的注释匹配正则表达式\\n ext_to_patterns = {\\n '.svg': [r'<!--[\\\\s\\\\S]*?-->'],\\n '.html': [r'<!--[\\\\s\\\\S]*?-->'],\\n '.xml': [r'<!--[\\\\s\\\\S]*?-->'],\\n '.py': [r'#.*', r'\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'', r'\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"'],\\n '.rb': [r'#.*'],\\n '.sh': [r'#.*'],\\n '.js': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.ts': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.java': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.c': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.cpp': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.cs': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.go': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*'],\\n '.php': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/', r'(?<![:\\\\w])//.*', r'#.*'],\\n '.css': [r'/\\\\*[\\\\s\\\\S]*?\\\\*/']\\n }\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为我们关注的源文件\\n if ext in ext_to_patterns:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 组合当前类型源文件支持的注释提取正则\\n patterns = ext_to_patterns[ext]\\n combined_pattern = re.compile('|'.join(patterns))\\n \\n # 提取该文件中的所有注释区域\\n comments = combined_pattern.findall(content)\\n \\n # 检查所有的注释内容中是否包含 emoji\\n for comment in comments:\\n for char in comment:\\n # is_emoji是提供的专家函数，代表判断某一字符是否为emoji\\n if is_emoji(char):\\n return False\\n except Exception:\\n # 如果遇到文件无法以utf-8读取或其它系统级别的访问异常情况，直接忽略继续排查即可\\n pass\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 包含用户指令提及和常用的代码文件扩展名\\n code_exts = {'.svg', '.json', '.md', '.py', '.xml', '.html', '.css', '.js', '.bak'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 判断是否为代码文件或是备份文件\\n if ext in code_exts or \\\"backup\\\" in file_name.lower() or \\\"bak\\\" in file_name.lower():\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # splitlines() 按行切分，安全过滤了末尾正常换行带来的假空行问题\\n lines = f.read().splitlines()\\n \\n for line in lines:\\n # 如果去掉行内空白字符后什么都不剩，则视为存在空行\\n if line.strip() == \\\"\\\":\\n return False\\n \\n except Exception:\\n # 忽略读取异常或非 utf-8 编码的无关二进制文件\\n continue\\n \\n # 如果所有被检查的文件均不存在空行，或者没有任何文件，则返回 True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"酷喔～\\\"这个语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response or not response.strip():\\n return False\\n \\n lines = response.split('\\\\n')\\n paragraphs = []\\n current_para = []\\n in_code_block = False\\n \\n for line in lines:\\n stripped = line.strip()\\n # 遇到代码块标记，切换是否在代码块内的状态\\n if stripped.startswith('```'):\\n if in_code_block:\\n in_code_block = False\\n else:\\n if current_para:\\n paragraphs.append('\\\\n'.join(current_para))\\n current_para = []\\n in_code_block = True\\n continue\\n \\n # 如果当前不在代码块内，提取常规回复段落\\n if not in_code_block:\\n if not stripped: # 空行作为段落的分界线\\n if current_para:\\n paragraphs.append('\\\\n'.join(current_para))\\n current_para = []\\n else:\\n current_para.append(stripped)\\n \\n if current_para:\\n paragraphs.append('\\\\n'.join(current_para))\\n \\n # 如果整个回复除了代码没有任何普通的正文段落，则退化为检查全文中是否有目标词汇\\n if not paragraphs:\\n return \\\"酷喔～\\\" in response\\n \\n # 检查所有提取出的回复正文段落是否均以\\\"酷喔～\\\"开头\\n for para in paragraphs:\\n if not para.startswith(\\\"酷喔～\\\"):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令内容推断可能的代码文件类型：SVG (图形代码)、JSON (配置文件)，以及相关的XML和HTML\\n code_extensions = {'.svg', '.json', '.xml', '.html'}\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若有任意一个代码文件不包含该标识，则立即返回False\\n if \\\"<!-- 几何切割Z标识 -->\\\" not in content:\\n return False\\n except Exception:\\n # 如果文件无法正常按utf-8读取，说明可能存在异常，保守视为验证失败\\n return False\\n\\n # 若找到了代码文件并且都通过了校验，则满足要求\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"看着有内味了。为了让它在深色背景上更加抢眼，我想给整个'Z'加上发光效果。请使用SVG的滤镜（如 feGaussianBlur 或 drop-shadow），在构成字母的几何块背后加一层亮青色或紫色的霓虹发光晕影，让它充满科技感。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\\\"炸场～\\\"这个语气词；第二，新增要求，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能生成的代码文件扩展名，尤其是 SVG 文件\\n code_extensions = {\\n '.svg', '.py', '.html', '.htm', '.js', '.css', \\n '.ts', '.jsx', '.tsx', '.xml', '.json'\\n }\\n \\n found_code_file = False\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n found_code_file = True\\n try:\\n # 读取文件时忽略编码报错，避免因文件原本编码错误导致抛出异常\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 检查起始第一行是否包含指定的编码声明字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法读取，则认为未满足要求\\n return False\\n \\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 若工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义目标注释字符串\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 典型代码文件的扩展名集合，其中主要包含本任务涉及到的 .svg\\n code_exts = {'.svg', '.py', '.js', '.html', '.css', '.java', '.cpp', '.c', '.h', '.ts', '.go', '.rs'}\\n \\n found_code_file = False\\n \\n # 遍历工作目录寻找代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的前15行进行检查，以覆盖\\\"开头处\\\"的要求\\n lines = [f.readline() for _ in range(15)]\\n except Exception:\\n # 如果读取异常（如编码不支持等），稳妥起见跳过该文件（极小概率）\\n continue\\n \\n # 检查这些行中是否包含目标注释字符串\\n has_comment = False\\n for line in lines:\\n if line and target_comment in line:\\n has_comment = True\\n break\\n \\n # 只要有一份代码文件未包含该注释，则判定要求未满足\\n if not has_comment:\\n return False\\n \\n # 如果找到了至少一份代码文件且所有的代码文件都符合要求，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令及一般情况，推断可能的代码文件扩展名\\n code_extensions = {'.svg', '.json', '.md', '.py', '.html', '.css', '.js', '.ts', '.xml'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果只要有一个代码文件不包含 \\\"——\\\"，就认为不满足要求\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 如果文件不可读，也视为失败\\n return False\\n \\n # 如果找到了至少一个代码文件且都通过了验证，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见的源文件扩展名集合\\n source_exts = {\\n '.py', '.svg', '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.c', '.cpp', '.h', '.hpp', '.java', '.cs', '.go', '.rs', '.sh', \\n '.rb', '.php', '.xml', '.md', '.json'\\n }\\n \\n # 定义提取各语言常见注释的正则表达式\\n comment_patterns = [\\n re.compile(r'<!--.*?-->', re.DOTALL), # XML / HTML / SVG / MD 格式注释\\n re.compile(r'/\\\\*.*?\\\\*/', re.DOTALL), # 多行注释 (/* ... */)\\n re.compile(r'(?<![\\\"\\\\':])//.*'), # 单行注释 (// ...)，简单排除 URL 或字符串情况\\n re.compile(r'(?<![\\\"\\\\'])#.*') # 单行注释 (# ...)，简单排除字符串情况\\n ]\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果文件读取失败（例如非 UTF-8 编码），跳过该文件\\n continue\\n \\n # 在当前文件内容中寻找所有匹配的注释\\n for pattern in comment_patterns:\\n comments = pattern.findall(content)\\n for comment in comments:\\n # 检查这段注释中是否存在 emoji\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的代码文件类型（包含源文件、JSON配置、MD文档以及可能的备份变体如.svg.bak）\\n target_exts = ['.svg', '.json', '.md']\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 只要文件名包含目标扩展名，就认为是需要检查的代码或相关文本文件\\n if any(ext in file for ext in target_exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # splitlines() 能兼容各类系统的换行符，且避免文件末尾正常的换行符被误判为空行\\n lines = content.splitlines()\\n for line in lines:\\n # 如果一行全部由空白字符组成，或者完全没有字符，strip() 返回空字符串，即认为是空行\\n if not line.strip():\\n return False\\n except Exception:\\n # 忽略在异常情况或者编码不符时导致的文件读取失败\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"炸场～\\\"这个语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 清除回复内容开头和结尾可能存在的空白字符\\n cleaned_response = response.strip()\\n \\n # 检查回复的起始位置是否以要求的语气词开头，兼容全角和半角波浪号\\n if cleaned_response.startswith(\\\"炸场～\\\") or cleaned_response.startswith(\\\"炸场~\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_str = \\\"<!-- 几何切割Z标识 -->\\\"\\n # 根据用户指令内容推断可能生成的代码文件后缀类型\\n code_extensions = {'.svg', '.json', '.md'}\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 检查文件是否是以推断出的代码文件后缀结尾的\\n if any(file.lower().endswith(ext) for ext in code_extensions):\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 若发现某个被认定的代码文件未包含该标识，立即判定为不满足要求\\n if target_str not in content:\\n return False\\n except Exception:\\n # 如果由于编码等原因无法读取此文件，则选择忽略异常继续遍历\\n pass\\n \\n # 如果工作区内找不到任何相关的代码文件，视作未满足要求\\n if not found_code_file:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在来丰富一下周围的背景空间。请在'Z'的周围随机散落一些细小的几何碎片（比如很小的三角形、倾斜的短线或十字星）。这些小元素也用相应的渐变色或半透明颜色，看起来像是主体字母在碎裂时崩出的能量火花。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\\\"炫爆～\\\"这个语气词；第二，新增要求，生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件扩展名，特别关注针对当前任务的 .svg 格式\\n code_exts = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', \\n '.svg', '.xml', '.java', '.c', '.cpp', '.cs', '.go', \\n '.rs', '.php', '.rb', '.sh'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理判定为代码格式的文件\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 以避免可能存在的 BOM 干扰第一行的读取\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline()\\n \\n # 检查第一行是否含有指定的编码声明\\n if not first_line or '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 如果由于编码或其他异常原因无法读取判定为代码的文件，视为验证失败\\n return False\\n \\n # 如果遍历完成没有发现违规文件，且至少找到并验证了一个代码文件，则视为通过\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户指令内容，推断涉及的代码文件扩展名主要包含 SVG, JSON 和 Markdown\\n target_exts = {'.svg', '.json', '.md'}\\n checked_files_count = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_exts:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 截取前10行作为“开头处”的检测范围\\n lines = [f.readline() for _ in range(10)]\\n except Exception:\\n # 如果无法按utf-8读取文本，跳过该文件\\n continue\\n \\n checked_files_count += 1\\n \\n # 检查前10行内是否包含指定的目标注释文本\\n found = any(target_comment in line for line in lines if line)\\n \\n if not found:\\n return False\\n\\n # 至少要有一份目标代码文件被生成并验证通过，才能视为满足要求\\n return checked_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断可能的代码文件后缀及备份文件\\n code_extensions = {'.svg', '.json', '.md', '.html', '.xml', '.css', '.js', '.py', '.txt', '.bak'}\\n file_found = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选推断为代码文件、配置文件或其对应备份的文件\\n if ext in code_extensions or 'backup' in file.lower() or 'bak' in file.lower():\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查文件内容是否包含 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 遇到非文本或者无法用UTF-8解码的文件读取异常，视作不满足要求\\n return False\\n \\n # 只要存在至少一个代码文件，且上述循环没有因为缺少 \\\"——\\\" 触发 return False，则满足要求\\n return file_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断常见的代码源文件后缀名\\n source_exts = {\\n '.svg', '.html', '.htm', '.xml', \\n '.py', '.sh', \\n '.js', '.ts', '.css', '.java', '.c', '.cpp', '.h', '.cs'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于编码等原因无法读取，则跳过\\n continue\\n \\n # 提取不同类型源文件内的全部注释文本\\n comments = []\\n if ext in ['.svg', '.html', '.htm', '.xml']:\\n # HTML/SVG/XML 风格注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n elif ext in ['.py', '.sh']:\\n # 单行注释 #\\n comments.extend(re.findall(r'#.*', content))\\n if ext == '.py':\\n # Python 多行注释/文档字符串\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n elif ext in ['.js', '.ts', '.css', '.java', '.c', '.cpp', '.h', '.cs']:\\n # C/JS 风格单行和多行注释\\n comments.extend(re.findall(r'//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n # 遍历所提取的每个注释字符串\\n for comment in comments:\\n for char in comment:\\n # 使用人类专家给定的 emoji 校验函数\\n if is_emoji(char):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 结合项目背景推断代码文件，并列举主流的代码文件类型。\\n # 排除 .md、.txt 等用于撰写文档且排版离不开空行的文本文档。\\n code_extensions = {\\n '.svg', '.json', '.xml', '.html', '.css', '.js', '.ts',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go',\\n '.rs', '.php', '.rb', '.sh', '.bat', '.jsx', '.tsx'\\n }\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 判定该行如果只包含空白字符或者什么也没有，即为空行\\n if line.strip() == \\\"\\\":\\n return False\\n except Exception:\\n # 若文件遇到读取错误或无法用 utf-8 解析，则跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"炫爆～\\\"这个语气词\",\n \"validation_code\": \"import re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI的回复中，是否在每一段给用户的回复前都加上了\\\"炫爆～\\\"语气词。\\n \\\"\\\"\\\"\\n # 移除 markdown 格式的代码块（代码块内部不应被视为“给用户的回复段落”进行语气词检查）\\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\\n \\n # 按照双换行符进行段落切分（Markdown标准段落切分方式）\\n paragraphs = [p.strip() for p in text_without_code.split('\\\\n\\\\n') if p.strip()]\\n \\n # 如果回复中不存在任何非代码块的普通文本段落，则判定为不满足\\n if not paragraphs:\\n return False\\n \\n # 检查每一个段落是否均以\\\"炫爆～\\\"开头\\n for p in paragraphs:\\n if not p.startswith(\\\"炫爆～\\\"):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_string = \\\"<!-- 几何切割Z标识 -->\\\"\\n \\n # 根据任务和常见的编程约定，定义典型的代码文件后缀\\n code_extensions = {\\n '.svg', '.html', '.htm', '.xml', '.css', '.js', '.jsx', '.ts', '.tsx',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \\n '.rs', '.php', '.rb', '.sh'\\n }\\n \\n found_code_file = False\\n \\n # 遍历当前工作区的所有目录及文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证，忽略 .json, .md 以及图片等非代码格式\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查目标注释字符串是否包含在该文件内容中\\n if target_string not in content:\\n return False\\n except Exception:\\n # 若文件存在编码读取异常（例如并非纯文本等意外情况），视为验证不通过\\n return False\\n \\n # 必须至少生成了一个代码文件，且全部验证通过才返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推测可能生成的代码文件后缀，涵盖了svg、json、md等以及常见的代码格式\\n code_extensions = {\\n '.py', '.js', '.ts', '.html', '.css', '.json', \\n '.xml', '.svg', '.md', '.java', '.cpp', '.c', \\n '.h', '.sh'\\n }\\n\\n # 确保给定的项目路径确实存在\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历当前项目地址下所有的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 提取出每一行开头的空白字符（即缩进部分）\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char == ' ' or char == '\\\\t':\\n leading_whitespace += char\\n else:\\n break\\n \\n # 检查缩进部分中是否包含了制表符\\n if '\\\\t' in leading_whitespace:\\n return False\\n except Exception:\\n # 若文件存在编码等问题无法读取，予以跳过继续检查下一个\\n continue\\n \\n # 所有代码文件的所有行均通过检查，没有发现使用制表符缩进\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"下一步，给深色的背景底增加一点质感。请利用SVG的 `<pattern>` 标签，在背景的最底层铺上一层隐约可见的细网格线或者点阵图案，颜色非常暗淡且带有较高透明度，营造出一种赛博空间的数字感。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\\\"飒炸～\\\"这个语气词；第二，新增要求，你提供的每个代码文件里，注释内容须100%使用中文撰写，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件拓展名（由于指令核心是SVG，着重于Web和主流语言相关后缀，不包含纯文档与配置文件如.md、.json）\\n code_exts = {'.svg', '.xml', '.py', '.js', '.html', '.css', '.c', '.cpp', '.java', '.go', '.rs', '.sh', '.php'}\\n has_checked_code = False\\n\\n # 1. 检查工作目录中实际生成的代码文件\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n # 排除此前多轮对话中可能生成的备份文件\\n if file.endswith('.bak') or '.bak.' in file:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 无论文件是否为空，只要是生成的代码文件，首行就应当具有该声明\\n if not first_line:\\n return False\\n \\n has_checked_code = True\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n pass\\n\\n # 2. 检查回复(response)中可能输出的Markdown代码块\\n # 使用正则匹配所有代码块\\n blocks = re.findall(r'```([a-zA-Z0-9]*)\\\\s*\\\\n(.*?)```', response, re.DOTALL)\\n for lang, code_content in blocks:\\n lang = lang.strip().lower()\\n # 筛选特定语言或未显式声明语言但是疑似SVG/XML标签的代码块\\n if lang in {'svg', 'xml', 'python', 'py', 'javascript', 'js', 'html', 'css', 'c', 'cpp', 'java'} or (lang == '' and '<svg' in code_content):\\n lines = code_content.split('\\\\n')\\n # 过滤纯空代码块的情况\\n if not lines or (len(lines) == 1 and not lines[0].strip()):\\n continue\\n \\n has_checked_code = True\\n # 获取代码块首行内容进行校验\\n first_line = lines[0]\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n\\n # 只要生成了代码并全部检查通过即返回True；若没生成任何代码则默认要求未被满足\\n return has_checked_code\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_string = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 定义常见代码文件的后缀，针对该任务重点是 .svg 格式\\n code_extensions = {'.svg', '.py', '.html', '.css', '.js', '.ts', '.c', '.cpp', '.java', '.go'}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n found_in_file = False\\n # 检查目标字符串是否在文件开头处（取前15行给予合理余量）\\n for line in lines[:15]:\\n if target_string in line:\\n found_in_file = True\\n break\\n \\n # 如果当前代码文件开头处没有该句注释，则未满足要求\\n if not found_in_file:\\n return False\\n except Exception:\\n # 如果遇到文件无法以utf-8读取等异常，视为验证不通过\\n return False\\n \\n # 如果项目中没有任何代码文件，则也视为不符合该任务场景的期望\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断的可能的代码文件扩展名\\n code_extensions = ['.svg', '.html', '.xml']\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 若文件内容中不包含指定的字符串 \\\"——\\\"，则直接判定为不满足要求\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 出现无法读取等异常时保守判定为不通过\\n return False\\n \\n # 如果找到了至少一个代码文件，且均未在前面返回False，则表示满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义不同类型源文件对应的注释提取正则表达式\\n patterns = {\\n 'html_xml_svg_md': [\\n re.compile(r'<!--(.*?)-->', re.DOTALL)\\n ],\\n 'py_sh': [\\n re.compile(r'#(.*?)$', re.MULTILINE), \\n re.compile(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', re.DOTALL), \\n re.compile(r\\\"'''(.*?)'''\\\", re.DOTALL)\\n ],\\n 'c_style': [\\n re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL), \\n re.compile(r'//(.*?)$', re.MULTILINE)\\n ]\\n }\\n\\n # 依据指令推断及常规情况给出的常见源代码文件扩展名映射\\n ext_mapping = {\\n '.svg': 'html_xml_svg_md',\\n '.html': 'html_xml_svg_md',\\n '.xml': 'html_xml_svg_md',\\n '.md': 'html_xml_svg_md',\\n '.py': 'py_sh',\\n '.sh': 'py_sh',\\n '.js': 'c_style',\\n '.ts': 'c_style',\\n '.css': 'c_style',\\n '.java': 'c_style',\\n '.c': 'c_style',\\n '.cpp': 'c_style',\\n '.h': 'c_style',\\n '.json': 'c_style' # JSON通常无注释，但部分环境支持C风格注释，故以c_style检查\\n }\\n\\n # 遍历当前工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in ext_mapping:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 对于无法读取或无法由 utf-8 解码的非标准源文件直接跳过\\n continue\\n \\n # 获取当前代码文件适用的正则提取器集合\\n applicable_patterns = patterns[ext_mapping[ext]]\\n \\n for ptn in applicable_patterns:\\n comments = ptn.findall(content)\\n for comment in comments:\\n # 遍历抽取的每段注释内容中的各个字符\\n for char in comment:\\n # is_emoji为预置函数，判断单个字符是否为emoji\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义需要检查的常见代码文件扩展名\\n code_extensions = {'.svg', '.json', '.py', '.js', '.html', '.css', '.xml', '.sh', '.bat'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.read().splitlines()\\n \\n has_code_file = True\\n # 遍历判断是否存在空行（或仅包含空白字符的行）\\n for line in lines:\\n if line.strip() == '':\\n return False\\n except Exception:\\n # 忽略读取异常的文件（例如编码不对齐等异常情况）\\n continue\\n\\n # 只要确实检查到了代码文件，且并未在任何代码文件里发现空行，即视为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"飒炸～\\\"这个语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否以指定语气词\\\"飒炸～\\\"开头。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 检查回复是否以\\\"飒炸～\\\"开头\\n return cleaned_response.startswith(\\\"飒炸～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_str = \\\"<!-- 几何切割Z标识 -->\\\"\\n \\n # 结合题目上下文推断，定义涵盖SVG设计及配置管理可能的代码/文本文件后缀\\n code_extensions = {\\n '.svg', '.xml', '.html', '.htm', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.py', '.json', '.md', '.txt', '.sh'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n has_code_file = True\\n # 如果任何一个代码文件没有包含该标识，则返回 False\\n if target_str not in content:\\n return False\\n except Exception:\\n # 如果文件无法用 utf-8 读取，或者出现其他IO异常，直接跳过\\n continue\\n \\n # 如果找到了至少一个代码文件，并且上面的循环没有因为缺失标识而返回 False，则代表全部包含\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_indentation_without_tabs(response: str, workspace_path: str) -> bool:\\n # 匹配行首空白字符（缩进）的正则表达式\\n indent_pattern = re.compile(r'^[ \\\\t]+')\\n \\n # 常见的代码和文本文件后缀名集合\\n code_extensions = {\\n '.py', '.java', '.js', '.ts', '.html', '.css', '.c', '.cpp', '.hpp', '.h', \\n '.cs', '.php', '.rb', '.go', '.rs', '.json', '.svg', '.md', '.txt', '.sh', '.xml'\\n }\\n \\n # 1. 检查 workspace_path 中的文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n match = indent_pattern.match(line)\\n # 如果行首的空白字符(缩进)中包含了制表符，则违反要求\\n if match and '\\\\t' in match.group():\\n return False\\n except Exception:\\n pass\\n \\n # 2. 检查 response 中的代码块\\n # 匹配 markdown 格式的代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n for line in block.split('\\\\n'):\\n match = indent_pattern.match(line)\\n # 同样判断行首缩进是否包含制表符\\n if match and '\\\\t' in match.group():\\n return False\\n \\n # 如果全都没有使用制表符缩进，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"太静态了，我们加点动画效果吧。请使用 CSS 或 SVG 内置的 `<animate>` 标签，让中心的主体字母'Z'整体产生一个缓慢的、上下悬浮的呼吸效果。同时，让周围散落的那些小能量碎片也进行轻微的旋转或透明度闪烁。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\\\"燃爆～\\\"这个语气词；第二，新增要求，每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，说明未生成任何项目文件\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 常见代码文件扩展名，特别包含本任务的主体代码格式 .svg\\n # 排除 .json 和 .md，以防它们被当做非代码文件处理而引起误判\\n code_extensions = {\\n \\\".svg\\\", \\\".py\\\", \\\".js\\\", \\\".ts\\\", \\\".html\\\", \\\".css\\\", \\\".xml\\\",\\n \\\".java\\\", \\\".cpp\\\", \\\".c\\\", \\\".h\\\", \\\".cs\\\", \\\".go\\\", \\\".rs\\\", \\n \\\".php\\\", \\\".rb\\\", \\\".sh\\\"\\n }\\n\\n has_checked_any_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，防止遍历 .git 等文件夹内的文件\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 如果文件为空或者第一行没有所需的编码声明\\n if not first_line or \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n \\n has_checked_any_code_file = True\\n except Exception:\\n # 如果读取出现异常，保守返回 False\\n return False\\n\\n # 确保至少检查到了一个代码文件且都符合要求\\n return has_checked_any_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 用户要求在代码文件开头处另起一行添加的明确注释文本\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 推断可能的代码文件类型，涵盖前端相关文件和常见编程语言扩展名\\n code_exts = {\\n '.svg', '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx',\\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果UTF-8解码失败，尝试使用忽略错误的模式读取\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n return False\\n\\n # 过滤掉空行，以便准确判断文本是否在“文件开头处”\\n lines = [line.strip() for line in lines if line.strip()]\\n \\n # 检查前10行非空行内是否存在指定的注释声明\\n found_comment = False\\n for line in lines[:10]:\\n if target_comment in line:\\n found_comment = True\\n break\\n \\n # 如果某一个代码文件未包含该注释，或者不在开头处，立即返回 False\\n if not found_comment:\\n return False\\n \\n # 只要存在至少一个代码文件且全都满足条件即为 True；如果整个项目无代码文件则判断为 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀集合，包含 SVG, HTML, CSS 等任务可能产生的文件类型\\n # 排除了 .md, .json 等文档及配置文件\\n code_exts = {\\n '.svg', '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', \\n '.php', '.rb', '.sh', '.xml', '.vue', '.svelte'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n \\n # 遍历工作目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名并转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查内容是否包含要求中的中划线\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 如果无法读取文件或编码不符合，则视为检查不通过\\n return False\\n \\n # 如果找到了至少一个代码文件，并且全包含 \\\"——\\\"，返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码源文件后缀名\\n source_extensions = {\\n '.svg', '.html', '.htm', '.xml', \\n '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', '.php'\\n }\\n \\n # 匹配各类常见注释的正则表达式\\n comment_patterns = [\\n r'<!--[\\\\s\\\\S]*?-->', # XML / HTML / SVG 注释\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/', # 多行注释 (C, C++, Java, JS, CSS)\\n r'//.*', # 单行注释 (C, C++, Java, JS)\\n r'#.*' # 单行注释 (Python, Shell 等)\\n ]\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 在文件内容中寻找所有匹配的注释\\n for pattern in comment_patterns:\\n comments = re.findall(pattern, content)\\n for comment in comments:\\n # 遍历注释字符串中的每一个字符，检查是否为 emoji\\n for char in comment:\\n if is_emoji(char):\\n return False\\n except Exception:\\n # 忽略读取异常的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n checked_any_file = False\\n \\n # 优先检查工作区生成的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n checked_any_file = True\\n \\n # 逐行检查是否包含空行\\n for line in lines:\\n if not line.strip():\\n return False\\n except Exception:\\n # 忽略二进制文件、非UTF-8文件等不可读取的文件\\n continue\\n \\n # 如果工作区未能读取到任何代码文件，则降级从模型回复中提取代码块检查\\n if not checked_any_file:\\n code_blocks = re.findall(r'```[a-zA-Z0-9]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n # 剥离首尾多余换行符防止分割时误判，然后再切分行\\n lines = block.strip('\\\\n').split('\\\\n')\\n for line in lines:\\n if not line.strip():\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"燃爆～\\\"这个语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否在最前面加上了\\\"燃爆～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容两端的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 判断回复是否以\\\"燃爆～\\\"开头\\n return cleaned_response.startswith(\\\"燃爆～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 涵盖当前任务场景可能的输出类型，以及常用代码文件的扩展名集合\\n code_extensions = {\\n '.svg', '.xml', '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx',\\n '.py', '.json', '.md', '.txt', '.sh', '.java', '.c', '.cpp', '.h', \\n '.cs', '.php', '.rb', '.go', '.rs', '.swift'\\n }\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n found_code_file = True\\n # 如果有任何一个代码文件不包含该标识，则立即判定为未满足要求\\n if \\\"<!-- 几何切割Z标识 -->\\\" not in content:\\n return False\\n except Exception:\\n # 忽略不可读或非UTF-8编码的文件\\n continue\\n \\n # 如果找到了至少一个代码文件且都包含标识，则返回True，否则返回False\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 检查 response 中的 Markdown 代码块\\n code_blocks = re.findall(r'```.*?\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n for line in block.split('\\\\n'):\\n # 提取行首空白字符部分（缩进部分）\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n if '\\\\t' in leading_whitespace:\\n return False\\n\\n # 检查工作区中的代码文件\\n # 根据用户指令上下文，涉及文件类型主要为 .svg, .json, .md 以及潜在的相关其他文件\\n code_extensions = {'.svg', '.json', '.md', '.css', '.html', '.xml', '.js', '.py', '.txt'}\\n \\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 提取行首空白字符部分（缩进部分）\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n if '\\\\t' in leading_whitespace:\\n return False\\n except Exception:\\n # 如遇编码等读取问题则跳过当前文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"继续增加高级的交互性。我想加一个鼠标悬停（hover）的动画效果：当鼠标悬停到主体'Z'所在的区域时，构成'Z'的那些几何切割块会向外微微散开（产生几像素的位移，表现出解体感），同时背后的发光效果变得更亮；当鼠标移开时，碎片平滑地拼合回原样。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\\\"超炫～\\\"这个语气词；第二，新增要求，务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令中涉及的文件，推断可能的代码文件/文本文件后缀名\\n code_extensions = {'.svg', '.json', '.md', '.log', '.xml', '.html', '.css', '.js', '.py'}\\n \\n has_checked_files = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n # 筛选出属于代码类型的文件\\n if ext in code_extensions:\\n has_checked_files = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 读取起始第一行\\n first_line = f.readline()\\n \\n # 检查第一行是否包含声明编码的特定字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于某种原因无法读取文件，认为不符合要求\\n return False\\n \\n # 如果项目下没有生成任何相关的代码文件，也不视为满足任务目标\\n return has_checked_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 目标注释字符串（不包含可能被系统指令引入的中文双引号）\\n target_string = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常见的代码文件后缀集合，包括本项目强相关的 SVG 和常见的源文件格式\\n code_extensions = {'.svg', '.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.java'}\\n \\n code_file_found = False\\n \\n # 确保给定的项目路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n # 忽略可能遇到的一些非常规字符导致的编码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n \\n # 检查前 10 行，以兼容第一行的编码声明等内容\\n found_in_header = False\\n for line in lines[:10]:\\n if target_string in line:\\n found_in_header = True\\n break\\n \\n # 只要有一个代码文件的头部没有包含指定的注释文字，即不符合条件\\n if not found_in_header:\\n return False\\n except Exception:\\n # 如果由于任何系统原因文件读取异常，为保证严谨视作验证失败\\n return False\\n \\n # 必须要至少生成了一份或以上的代码文件，全部通过验证后方能判定满足\\n return code_file_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断是否为代码文件或其备份文件\\n def is_target_code_file(filename: str) -> bool:\\n code_exts = {'.svg', '.html', '.css', '.js', '.ts', '.py'}\\n filename = filename.lower()\\n for ext in code_exts:\\n if filename.endswith(ext):\\n return True\\n # 处理备份文件可能带有的后缀\\n if filename.endswith(ext + '.bak') or filename.endswith(ext + '.backup'):\\n return True\\n # 处理形如 logo_backup.svg 或类似包含备份字眼且包含代码扩展名的情况\\n if ext in filename and ('bak' in filename or 'backup' in filename):\\n return True\\n return False\\n\\n file_count = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n if is_target_code_file(file):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n file_count += 1\\n # 验证是否含有 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 如果因权限或编码问题导致无法读取，则安全跳过该文件\\n pass\\n \\n # 必须至少有一份符合条件的代码文件存在，且全都满足包含 \\\"——\\\" 的要求\\n return file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的源代码文件后缀集合\\n source_extensions = {\\n '.svg', '.xml', '.html', '.htm',\\n '.js', '.ts', '.jsx', '.tsx', \\n '.css', '.scss', '.less',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.sh', '.rb'\\n }\\n \\n # 遍历代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取源文件内容，忽略无法解码的字符以增加鲁棒性\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n comments = []\\n # 针对不同扩展名进行针对性的注释提取\\n if ext in {'.svg', '.xml', '.html', '.htm'}:\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # 提取内部的 <style> 和 <script> 并读取其中潜藏的注释\\n styles = re.findall(r'<style[^>]*>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n for style in styles:\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', style))\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for script in scripts:\\n comments.extend(re.findall(r'(?<!:)//.*', script))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', script))\\n \\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.css', '.scss', '.less', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs'}:\\n # 使用 (?<!:) 避免将 url 中的 // 误提取为注释\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n elif ext in {'.py', '.sh', '.rb'}:\\n comments.extend(re.findall(r'#.*', content))\\n \\n # 遍历提取到的所有注释内容\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n except Exception:\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区不存在，或者没有任何文件生成，也算不违背该特定要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤掉系统隐藏文件，例如 .git 文件夹内文件或 .DS_Store 等\\n if file.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 UTF-8 格式读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n # 若抛出解码异常，说明可能是图片(.png)等二进制文件，跳过即可\\n continue\\n except Exception:\\n continue\\n \\n lines = content.split('\\\\n')\\n for i, line in enumerate(lines):\\n if line.strip() == '':\\n # 允许因文件末尾带有单个正常换行符而导致 split 后列表最后一个元素为空字符串\\n if i == len(lines) - 1 and line == '':\\n continue\\n # 发现任何一个不应存在的空行，则判定为没有满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"超炫～\\\"这个语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import re\\n \\n response_stripped = response.strip()\\n if not response_stripped:\\n return False\\n \\n # 如果整体回复中根本就没有出现该语气词，直接判定失败\\n if \\\"超炫～\\\" not in response:\\n return False\\n \\n # 移除回复中的所有的代码块内容\\n text_no_code = re.sub(r'```.*?```', '\\\\n\\\\n', response, flags=re.DOTALL)\\n \\n # 按照双换行划分段落\\n raw_paragraphs = [p.strip() for p in text_no_code.split('\\\\n\\\\n') if p.strip()]\\n \\n normal_paragraphs = []\\n for p in raw_paragraphs:\\n # 跳过首字符为 '<' 的 XML/HTML 格式的段落\\n if p.startswith('<'):\\n continue\\n \\n # 跳过由于前面轮次要求的 Review.md 格式输出的相关段落\\n if 'Code Quality Review:' in p or 'Strengths:' in p or 'Issues:' in p or 'Recommendation:' in p:\\n continue\\n \\n lines = [line.strip() for line in p.split('\\\\n') if line.strip()]\\n if not lines:\\n continue\\n \\n # 检查是否是纯列表形式组成的段落，如果是则跳过\\n if all(re.match(r'^([\\\\-\\\\*\\\\+]|\\\\d+\\\\.)\\\\s', line) for line in lines):\\n continue\\n \\n normal_paragraphs.append(p)\\n \\n # 如果没有任何普通的文本段落，只有特殊格式时，此时只要含有目标词汇即可认为满足\\n if not normal_paragraphs:\\n return True\\n \\n # 所有保留下来的普通文本段落，都必须以 \\\"超炫～\\\" 开头\\n for p in normal_paragraphs:\\n # 清除段落开头可能带有的 Markdown 修饰符或空格\\n clean_p = re.sub(r'^[\\\\#\\\\*\\\\-\\\\>\\\\s]+', '', p)\\n # 检查原始段落或清理后的段落是否以规定词汇开头\\n if not (clean_p.startswith(\\\"超炫～\\\") or p.startswith(\\\"超炫～\\\")):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见代码文件的扩展名集合（排除 .md, .json, .log 等文档与数据文件）\\n code_exts = {\\n '.svg', '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.xml', '.sh', '.bat'\\n }\\n \\n target_str = \\\"<!-- 几何切割Z标识 -->\\\"\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 如果有任意一个代码文件不包含目标字符串，直接返回 False\\n if target_str not in content:\\n return False\\n except Exception:\\n # 如果出现读取异常（如编码错误），无法验证包含该标识，亦视为失败\\n return False\\n \\n # 确保至少检查过一个代码文件，否则返回 False 避免空项目的假阳性通过\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证代码项目中生成的代码文件是否完全使用纯空格缩进，不包含任何制表符(\\\\t)。\\n \\\"\\\"\\\"\\n # 根据多轮指令，推断代码文件可能包含 svg, json 等，涵盖常见的代码扩展名\\n code_extensions = {'.svg', '.html', '.css', '.js', '.ts', '.json', '.xml', '.py', '.md'}\\n has_code_files = False\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 获取每一行开头的空白字符（即缩进部分）\\n # line.lstrip() 去掉左侧所有空白符，原长度减去 lstrip 后的长度即为左侧空白符的长度\\n leading_whitespaces = line[:len(line) - len(line.lstrip())]\\n \\n # 检查缩进部分是否包含制表符 \\\\t\\n if '\\\\t' in leading_whitespaces:\\n return False\\n except UnicodeDecodeError:\\n # 忽略非UTF-8编码的可能引发读取异常的文件\\n continue\\n except Exception:\\n # 忽略其他读取异常\\n continue\\n \\n # 确保至少扫描到了代码文件，并且没有触发制表符缩进的检查\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"最后一步，来做收尾和完善。在主体字母的正下方增加一行无衬线风格的纯文本品牌名“ZENITH”，要求字距拉宽（letter-spacing），文字颜色与 logo 整体的渐变色相呼应。同时，去除原先固定长宽的绝对像素尺寸，改为利用 `viewBox` 实现完美的响应式，使其能在任何尺寸的网页容器中等比缩放。另外后续所有轮次都必须严格遵循两个要求：第一，替换此前的语气词要求，在你输出的每一段给用户的回复前，务必先加上\\\"封神～\\\"这个语气词；第二，新增要求，确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项，该要求后续所有轮次始终遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进implementation_plan.json\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每次完成项目的修改后，都必须在 progress.md 文件中记录这一轮已完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义预期可能生成的代码文件扩展名集合\\n # 主要针对 SVG 图像设计及配套的 Web 页面、样式、脚本和常见程序代码\\n code_extensions = {'.svg', '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含指定的编码声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法按 utf-8 读取或发生了其他读取异常，直接判定为失败\\n return False\\n \\n # 如果找到了至少一个代码文件且全都验证通过，则返回 True；否则返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"记得在你输出的每一份代码文件里的开头处，另起一行加上这句清晰的注释：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_sentence = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户指令推断，可能的代码文件扩展名\\n code_extensions = {\\n '.svg', '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs'\\n }\\n \\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的前10行（确保覆盖“开头处，另起一行”的要求）\\n lines = [f.readline() for _ in range(10)]\\n \\n found = False\\n for line in lines:\\n if not line: # 遇到文件末尾(EOF)\\n break\\n if target_sentence in line:\\n found = True\\n break\\n \\n # 如果该代码文件开头处不包含目标注释，则要求未满足\\n if not found:\\n return False\\n except Exception:\\n # 如果文件读取出现异常，保守返回 False\\n return False\\n \\n # 确保项目中确实存在生成的代码文件，且均满足了条件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件中，均需含有 \\\"——\\\"\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断可能的代码文件类型（例如 SVG、HTML、CSS、JS 等）\\n # 排除 .md、.log 等通常不被视作核心源文件的文档或日志文件\\n code_extensions = {\\n '.svg', '.html', '.xml', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.cs'\\n }\\n \\n found_code_files = False\\n \\n # 首先检查工作区中生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查各个代码文件是否含有 \\\"——\\\"\\n if \\\"——\\\" not in content:\\n return False\\n except Exception:\\n # 如果存在无法以 utf-8 读取的文件，可能非文本源文件，暂且忽略\\n pass\\n \\n # 如果工作区中没有发现实际保存的代码文件，备选检查 response 中可能生成的代码块\\n if not found_code_files:\\n # 匹配 markdown 代码块，格式：```语言\\\\n代码内容\\\\n```\\n code_blocks = re.findall(r'```([a-zA-Z0-9]*)\\\\n(.*?)```', response, re.DOTALL)\\n if not code_blocks:\\n # 既没有实际文件，也没有代码块，无法证明满足要求\\n return False \\n \\n for lang, code_content in code_blocks:\\n lang = lang.strip().lower()\\n # 排除非代码性质的内容块\\n if lang in {'markdown', 'md', 'json', 'log', 'text', 'yaml', 'yml', 'bash', 'sh'}:\\n continue\\n \\n found_code_files = True\\n if \\\"——\\\" not in code_content:\\n return False\\n \\n # 确保只要发现了代码文件，且它们都顺利通过了包含 \\\"——\\\" 的检查，即可返回 True\\n return found_code_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请保证你的所有源文件在注释区域均不涉及 emoji表情符号\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义我们要检查的源文件后缀（排除了 .md, .json, .log 等非代码源文件）\\n source_extensions = {\\n '.svg', '.xml', '.html', '.htm', \\n '.css', '.js', '.ts', \\n '.py', '.java', '.c', '.cpp', '.cs', '.go'\\n }\\n\\n # 预编译各种常见代码注释的正则表达式（尽量覆盖各种注释格式）\\n # XML/HTML 风格: <!-- ... -->\\n xml_comment_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\\n # C/CSS/JS 风格多行注释: /* ... */\\n c_multi_pattern = re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL)\\n # C/JS 风格单行注释: // ...\\n c_single_pattern = re.compile(r'//(.*?)$', re.MULTILINE)\\n # Python/Shell 风格单行注释: # ...\\n hash_pattern = re.compile(r'#(.*?)$', re.MULTILINE)\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in source_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取源文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略无法通过 utf-8 读取的文件\\n continue\\n \\n comments = []\\n \\n # 根据文件类型提取对应格式的注释内容\\n if ext in {'.svg', '.xml', '.html', '.htm'}:\\n comments.extend(xml_comment_pattern.findall(content))\\n # SVG中也可能包含 <style> 标签和 <script> 标签，因此顺带提取 C 风格注释\\n comments.extend(c_multi_pattern.findall(content))\\n comments.extend(c_single_pattern.findall(content))\\n elif ext in {'.css'}:\\n comments.extend(c_multi_pattern.findall(content))\\n elif ext in {'.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go'}:\\n comments.extend(c_multi_pattern.findall(content))\\n comments.extend(c_single_pattern.findall(content))\\n elif ext in {'.py'}:\\n comments.extend(hash_pattern.findall(content))\\n \\n # 遍历提取到的每一个注释段落\\n for comment in comments:\\n # 遍历注释内的每个字符，调用 is_emoji() 进行判断\\n for char in comment:\\n if is_emoji(char):\\n return False # 如果在注释区域发现任何emoji，则返回不满足\\n\\n # 所有源文件的注释区域均未发现 emoji，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次测试生成的代码完毕后，回复中严格根据如下XML格式逐项输出执行结果及描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 确保给定的项目路径存在\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历代码项目的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # 尝试以 utf-8 格式读取文件内容\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except UnicodeDecodeError:\\n # 如果引发解码错误，说明可能是二进制文件（如图片等），直接跳过\\n continue\\n except Exception:\\n continue\\n \\n # 逐行检查是否包含空行\\n for line in lines:\\n # strip() 可以同时去除纯换行符和带有空格/制表符的空行\\n if line.strip() == \\\"\\\":\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一段给用户的回复前，务必先加上\\\"封神～\\\"这个语气词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response or not response.strip():\\n return False\\n \\n lines = response.split('\\\\n')\\n in_code_block = False\\n text_lines = []\\n \\n # 逐行遍历，过滤掉代码块内的内容\\n for line in lines:\\n if line.strip().startswith('```'):\\n in_code_block = not in_code_block\\n # 插入一个空行标记，防止代码块前后的自然语言段落粘连\\n text_lines.append(\\\"\\\")\\n continue\\n \\n if not in_code_block:\\n text_lines.append(line)\\n \\n text_content = '\\\\n'.join(text_lines)\\n \\n # 利用正则表达式按连续的空行（允许包含空格等空白字符）分隔段落\\n paragraphs = re.split(r'\\\\n\\\\s*\\\\n', text_content)\\n \\n # 清理段落两端空白，并剔除空段落\\n valid_paragraphs = [p.strip() for p in paragraphs if p.strip()]\\n \\n # 如果除了代码块没有任何自然语言段落，无法确认是否满足要求\\n if not valid_paragraphs:\\n return False\\n \\n # 验证每个自然语言段落是否都以 \\\"封神～\\\" 开头\\n for p in valid_paragraphs:\\n if not p.startswith(\\\"封神～\\\"):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中都必须包含\\\"<!-- 几何切割Z标识 -->\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀，排除 .md, .log, .json 等文档及记录类文件\\n code_extensions = {\\n '.svg', '.html', '.htm', '.xml', '.css', '.scss', \\n '.js', '.jsx', '.ts', '.tsx', '.py', '.java', \\n '.c', '.cpp', '.h', '.hpp', '.sh', '.bash'\\n }\\n\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n # 排除常见的版本控制或依赖等非人工编写/生成的文件夹\\n if '.git' in file_path or 'node_modules' in file_path:\\n continue\\n \\n found_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if \\\"<!-- 几何切割Z标识 -->\\\" not in content:\\n return False\\n except UnicodeDecodeError:\\n # 如果无法以 utf-8 解码，认为其不是期望的文本代码文件\\n pass\\n\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在每次变动已有文件之前，都在其所在的同级目录中创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"生成的代码中不能用制表符来实现缩进，只能使用纯空格进行缩进\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 1. 检查 workspace_path 下生成的代码文件\\n # 根据用户指令中提到的SVG logo设计、JSON文件记录及前端特效，框定相关的代码文件类型\\n code_extensions = {'.svg', '.json', '.html', '.css', '.js', '.xml'}\\n \\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 匹配行首的前导空白字符（空格和制表符）\\n match = re.match(r'^[ \\\\t]+', line)\\n if match:\\n leading_whitespace = match.group(0)\\n # 如果前导缩进中包含制表符，返回 False\\n if '\\\\t' in leading_whitespace:\\n return False\\n except Exception:\\n continue\\n\\n # 2. 检查 response 回复文本中输出的代码块\\n # 匹配各类代码块（含带语言标识如 ```svg, ```json 和不带标识的常规代码块）\\n pattern = r'```[a-zA-Z0-9]*\\\\s*\\\\n(.*?)```'\\n code_blocks = re.findall(pattern, response, flags=re.DOTALL)\\n \\n for block in code_blocks:\\n for line in block.split('\\\\n'):\\n match = re.match(r'^[ \\\\t]+', line)\\n if match:\\n leading_whitespace = match.group(0)\\n if '\\\\t' in leading_whitespace:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"每当你更新了新内容，请按如下格式更新 Review.md 文件，以便对已存在的代码进行Review：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["改变浏览器窗口宽度，截图验证SVG画面是否完整显示且等比例缩放，确认其利用了viewBox特性实现了响应式尺寸适配，没有被固定的绝对像素长宽限制。","截取网页整体画面，确认页面背景的主色调为深邃的藏青色或深黑色。","仔细查看背景的最底层，确认背景上铺设了一层颜色暗淡且具有较高透明度的细网格线或点阵图案。","查看页面正中央的主体图案，确认其整体外形呈现出一个大写字母'Z'的形态。","查看字母'Z'的内部结构，确认该字母已经被拆分，且是由至少四个及以上的三角形或四边形几何碎块紧密贴合拼接而成的。","检查构成'Z'的各个几何碎块，确认它们均使用了赛博朋克风格（如亮青色到深蓝色的渐变，或蓝紫渐变）进行着色填充。","观察相邻几何碎块的色彩细节，确认每一块的渐变方向或渐变角度有微微的错开，利用色彩的明暗对比营造出了立体感和水晶切割的视觉效果。","查看'Z'字母主体的轮廓后方，确认底层加有一层亮青色或紫色的霓虹发光晕影滤镜效果，让字母在深色背景中显得发光。","查看'Z'字母周围的空间，确认随机散落有多个细小的几何碎片（如极小的三角形、倾斜短线或十字星等元素）。","不操作鼠标静静观察并连续截图，对比画面，确认中心主体字母'Z'整体正在产生缓慢且持续的上下移动（悬浮呼吸）动画效果。","不操作鼠标静静观察并连续截图，对比画面，确认周围散落的小碎片元素正在持续进行轻微的旋转或者透明度闪烁变化的动画效果。","模拟鼠标悬停到主体字母'Z'所在的区域，截图确认在悬停状态下，构成'Z'的各个几何切割碎块向外产生了位移，表现出微微散开（解体）的视觉效果。","模拟鼠标悬停到主体字母'Z'所在的区域，对比悬停前后的发光状态，确认鼠标悬停时，字母背后的霓虹发光晕影变得比之前更加明亮。","将鼠标从主体'Z'区域完全移开，截图对比，确认鼠标移出后，向外散开的几何碎块产生了平滑的过渡动画，重新拼合回原先紧密相连的'Z'字形态。","查看主体字母'Z'的正下方，确认显示了一行纯文本内容为“ZENITH”。","检查“ZENITH”文本的排版和样式，确认其采用了无衬线字体，字间距被明显拉宽（letter-spacing增加），且文字的颜色能够与logo主体的渐变风格相呼应。"],"string":"[\n \"改变浏览器窗口宽度，截图验证SVG画面是否完整显示且等比例缩放，确认其利用了viewBox特性实现了响应式尺寸适配，没有被固定的绝对像素长宽限制。\",\n \"截取网页整体画面，确认页面背景的主色调为深邃的藏青色或深黑色。\",\n \"仔细查看背景的最底层，确认背景上铺设了一层颜色暗淡且具有较高透明度的细网格线或点阵图案。\",\n \"查看页面正中央的主体图案，确认其整体外形呈现出一个大写字母'Z'的形态。\",\n \"查看字母'Z'的内部结构，确认该字母已经被拆分，且是由至少四个及以上的三角形或四边形几何碎块紧密贴合拼接而成的。\",\n \"检查构成'Z'的各个几何碎块，确认它们均使用了赛博朋克风格（如亮青色到深蓝色的渐变，或蓝紫渐变）进行着色填充。\",\n \"观察相邻几何碎块的色彩细节，确认每一块的渐变方向或渐变角度有微微的错开，利用色彩的明暗对比营造出了立体感和水晶切割的视觉效果。\",\n \"查看'Z'字母主体的轮廓后方，确认底层加有一层亮青色或紫色的霓虹发光晕影滤镜效果，让字母在深色背景中显得发光。\",\n \"查看'Z'字母周围的空间，确认随机散落有多个细小的几何碎片（如极小的三角形、倾斜短线或十字星等元素）。\",\n \"不操作鼠标静静观察并连续截图，对比画面，确认中心主体字母'Z'整体正在产生缓慢且持续的上下移动（悬浮呼吸）动画效果。\",\n \"不操作鼠标静静观察并连续截图，对比画面，确认周围散落的小碎片元素正在持续进行轻微的旋转或者透明度闪烁变化的动画效果。\",\n \"模拟鼠标悬停到主体字母'Z'所在的区域，截图确认在悬停状态下，构成'Z'的各个几何切割碎块向外产生了位移，表现出微微散开（解体）的视觉效果。\",\n \"模拟鼠标悬停到主体字母'Z'所在的区域，对比悬停前后的发光状态，确认鼠标悬停时，字母背后的霓虹发光晕影变得比之前更加明亮。\",\n \"将鼠标从主体'Z'区域完全移开，截图对比，确认鼠标移出后，向外散开的几何碎块产生了平滑的过渡动画，重新拼合回原先紧密相连的'Z'字形态。\",\n \"查看主体字母'Z'的正下方，确认显示了一行纯文本内容为“ZENITH”。\",\n \"检查“ZENITH”文本的排版和样式，确认其采用了无衬线字体，字间距被明显拉宽（letter-spacing增加），且文字的颜色能够与logo主体的渐变风格相呼应。\"\n]","isConversational":false}}},{"rowIdx":72,"cells":{"task_id":{"kind":"number","value":73,"string":"73"},"repository_policy":{"kind":"string","value":"你作为代码智能体，需完成后续所有代码开发相关需求，以下约束条件在后续所有交互轮次必须始终严格遵循：\n1. 你提供的每个代码文件都严格要求使用UTF-8作为编码。\n2. 你构建的代码工程目录层级必须小于或等于2层。\n3. 针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母。\n4. 确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项。\n5. 保持你生成的代码文件里每个函数的圈复杂度均在3以内。\n6. 请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```。\n7. 在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md。\n8. 你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码文件及项目文本文件后缀名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp',\n '.cs', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css',\n '.scss', '.less', '.json', '.xml', '.md', '.sh', '.bash',\n '.yaml', '.yml', '.rb', '.go', '.rs', '.php', '.swift', '.kt', '.sql'\n }\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀并统一转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 如果是代码文件，尝试以 UTF-8 编码读取\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果引发 UnicodeDecodeError，说明非严格 UTF-8 编码\n return False\n except Exception:\n # 忽略其他读取权限等非编码导致的异常\n pass\n \n # 如果遍历检查完所有代码文件均无 UnicodeDecodeError，则认定满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n \n # 检查工作区路径是否存在以及是否为有效目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n workspace = Path(workspace_path).resolve()\n \n # 遍历当前项目的所有目录\n for root, dirs, files in os.walk(workspace_path):\n current_path = Path(root).resolve()\n \n # 计算当前目录相对于工作区的相对路径\n try:\n rel_path = current_path.relative_to(workspace)\n except ValueError:\n # 理论上不会发生，防御性处理\n continue\n \n # rel_path.parts 返回路径的组成部分，例如：\n # - 当 current_path 为工作区根目录时，parts 为空 ()\n # - 当为一级子目录时，parts 长度为 1，如 ('src',)\n # - 当为二级子目录时，parts 长度为 2，如 ('src', 'utils')\n # 因此，目录所在层级 = len(rel_path.parts) + 1\n level = len(rel_path.parts) + 1\n \n # 按照要求，工程目录层级必须小于或等于 2 层\n if level > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 常见代码文件后缀\n valid_exts = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', \n '.c', '.cpp', '.cs', '.go', '.php', '.rb', \n '.html', '.vue', '.css'\n }\n \n # 常用编程语言关键字，用于排除误匹配的非变量保留字\n keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'return', 'break', \n 'continue', 'class', 'struct', 'enum', 'interface', 'implements', 'extends', \n 'public', 'private', 'protected', 'static', 'final', 'void', 'new', 'this', \n 'super', 'true', 'false', 'null', 'undefined', 'NaN', 'try', 'catch', \n 'finally', 'throw', 'throws', 'import', 'export', 'default', 'function', \n 'typeof', 'instanceof', 'in', 'of', 'let', 'var', 'const', 'int', 'float', \n 'double', 'char', 'String', 'boolean', 'bool', 'long', 'short', 'byte',\n 'True', 'False', 'None', 'and', 'or', 'not', 'is', 'as', 'assert', 'async', \n 'await', 'def', 'del', 'elif', 'except', 'global', 'lambda', 'nonlocal', \n 'pass', 'raise', 'with', 'yield'\n }\n\n # 检查单个变量名是否符合要求\n def check_var(name: str) -> bool:\n has_digit = any(c.isdigit() for c in name)\n has_lower = any(c.islower() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_underscore = '_' in name\n return has_digit and has_lower and has_upper and has_underscore\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 优先使用 AST 精确解析 Python 文件\n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if node.id not in keywords and not check_var(node.id):\n return False\n elif isinstance(node, ast.arg):\n if node.arg not in keywords and not check_var(node.arg):\n return False\n continue # Python文件解析且验证成功后跳过后续的正则检查\n except Exception:\n # 如果由于语法不完整等原因解析失败，回退到正则表达式检查\n pass\n \n # 使用正则表达式来提取其他语言的变量（或处理解析失败的 Python 文件）\n # 1. 简单的正则移除注释和字符串字面量，减少干扰\n clean_content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n clean_content = re.sub(r'//.*', '', clean_content)\n clean_content = re.sub(r'#.*', '', clean_content)\n clean_content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '', clean_content)\n clean_content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", '', clean_content)\n \n # 2. 匹配声明语句的变量名，并用否定前瞻 `(?!\\s*\\()` 排除函数定义\n decl_pattern = r'\\b(?:let|var|const|int|float|double|char|String|boolean|bool|long|short|byte)\\s+([a-zA-Z_]\\w*)\\b(?!\\s*\\()'\n \n # 3. 匹配赋值语句左侧的变量名，支持复合赋值运算符，通过 `(?!=|>)` 排除 `==`, `===`, `=>` 等\n assign_pattern = r'\\b([a-zA-Z_]\\w*)\\s*(?:\\+|-|\\*|/|%|&|\\||\\^|<<|>>)?=(?!=|>)'\n \n matches = re.findall(decl_pattern, clean_content) + re.findall(assign_pattern, clean_content)\n \n for var_name in matches:\n if var_name not in keywords and not check_var(var_name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 罗列常见的代码文件后缀名，用于过滤掉诸如 .md 的普通文本文件\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp',\n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.go', '.rs',\n '.php', '.rb', '.swift', '.kt', '.sh', '.bash', '.sql', '.cs',\n '.m', '.scala', '.dart', '.lua', '.vue', '.svelte'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git, .idea, .vscode 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n name, ext = os.path.splitext(file)\n # 仅检查识别范围内的代码文件\n if ext.lower() in code_extensions:\n # 检查主文件名是否仅由小写字母和下划线组成\n if not bool(re.match(r'^[a-z_]+$', name)):\n return False\n \n return True","tags":["Style","Naming"]}],"string":"[\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码文件及项目文本文件后缀名\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp',\\n '.cs', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css',\\n '.scss', '.less', '.json', '.xml', '.md', '.sh', '.bash',\\n '.yaml', '.yml', '.rb', '.go', '.rs', '.php', '.swift', '.kt', '.sql'\\n }\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀并统一转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果是代码文件，尝试以 UTF-8 编码读取\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果引发 UnicodeDecodeError，说明非严格 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他读取权限等非编码导致的异常\\n pass\\n \\n # 如果遍历检查完所有代码文件均无 UnicodeDecodeError，则认定满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n \\n # 检查工作区路径是否存在以及是否为有效目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n workspace = Path(workspace_path).resolve()\\n \\n # 遍历当前项目的所有目录\\n for root, dirs, files in os.walk(workspace_path):\\n current_path = Path(root).resolve()\\n \\n # 计算当前目录相对于工作区的相对路径\\n try:\\n rel_path = current_path.relative_to(workspace)\\n except ValueError:\\n # 理论上不会发生，防御性处理\\n continue\\n \\n # rel_path.parts 返回路径的组成部分，例如：\\n # - 当 current_path 为工作区根目录时，parts 为空 ()\\n # - 当为一级子目录时，parts 长度为 1，如 ('src',)\\n # - 当为二级子目录时，parts 长度为 2，如 ('src', 'utils')\\n # 因此，目录所在层级 = len(rel_path.parts) + 1\\n level = len(rel_path.parts) + 1\\n \\n # 按照要求，工程目录层级必须小于或等于 2 层\\n if level > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 常见代码文件后缀\\n valid_exts = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', \\n '.c', '.cpp', '.cs', '.go', '.php', '.rb', \\n '.html', '.vue', '.css'\\n }\\n \\n # 常用编程语言关键字，用于排除误匹配的非变量保留字\\n keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'return', 'break', \\n 'continue', 'class', 'struct', 'enum', 'interface', 'implements', 'extends', \\n 'public', 'private', 'protected', 'static', 'final', 'void', 'new', 'this', \\n 'super', 'true', 'false', 'null', 'undefined', 'NaN', 'try', 'catch', \\n 'finally', 'throw', 'throws', 'import', 'export', 'default', 'function', \\n 'typeof', 'instanceof', 'in', 'of', 'let', 'var', 'const', 'int', 'float', \\n 'double', 'char', 'String', 'boolean', 'bool', 'long', 'short', 'byte',\\n 'True', 'False', 'None', 'and', 'or', 'not', 'is', 'as', 'assert', 'async', \\n 'await', 'def', 'del', 'elif', 'except', 'global', 'lambda', 'nonlocal', \\n 'pass', 'raise', 'with', 'yield'\\n }\\n\\n # 检查单个变量名是否符合要求\\n def check_var(name: str) -> bool:\\n has_digit = any(c.isdigit() for c in name)\\n has_lower = any(c.islower() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_underscore = '_' in name\\n return has_digit and has_lower and has_upper and has_underscore\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 优先使用 AST 精确解析 Python 文件\\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if node.id not in keywords and not check_var(node.id):\\n return False\\n elif isinstance(node, ast.arg):\\n if node.arg not in keywords and not check_var(node.arg):\\n return False\\n continue # Python文件解析且验证成功后跳过后续的正则检查\\n except Exception:\\n # 如果由于语法不完整等原因解析失败，回退到正则表达式检查\\n pass\\n \\n # 使用正则表达式来提取其他语言的变量（或处理解析失败的 Python 文件）\\n # 1. 简单的正则移除注释和字符串字面量，减少干扰\\n clean_content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n clean_content = re.sub(r'//.*', '', clean_content)\\n clean_content = re.sub(r'#.*', '', clean_content)\\n clean_content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '', clean_content)\\n clean_content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", '', clean_content)\\n \\n # 2. 匹配声明语句的变量名，并用否定前瞻 `(?!\\\\s*\\\\()` 排除函数定义\\n decl_pattern = r'\\\\b(?:let|var|const|int|float|double|char|String|boolean|bool|long|short|byte)\\\\s+([a-zA-Z_]\\\\w*)\\\\b(?!\\\\s*\\\\()'\\n \\n # 3. 匹配赋值语句左侧的变量名，支持复合赋值运算符，通过 `(?!=|>)` 排除 `==`, `===`, `=>` 等\\n assign_pattern = r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*(?:\\\\+|-|\\\\*|/|%|&|\\\\||\\\\^|<<|>>)?=(?!=|>)'\\n \\n matches = re.findall(decl_pattern, clean_content) + re.findall(assign_pattern, clean_content)\\n \\n for var_name in matches:\\n if var_name not in keywords and not check_var(var_name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 罗列常见的代码文件后缀名，用于过滤掉诸如 .md 的普通文本文件\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp',\\n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.go', '.rs',\\n '.php', '.rb', '.swift', '.kt', '.sh', '.bash', '.sql', '.cs',\\n '.m', '.scala', '.dart', '.lua', '.vue', '.svelte'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git, .idea, .vscode 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n name, ext = os.path.splitext(file)\\n # 仅检查识别范围内的代码文件\\n if ext.lower() in code_extensions:\\n # 检查主文件名是否仅由小写字母和下划线组成\\n if not bool(re.match(r'^[a-z_]+$', name)):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用HTML、CSS和JS做一个七巧板益智游戏的雏形。先搭建基础界面的骨架：左侧是一个大的游戏操作区（带明显边界的画布或容器），右侧是控制面板（包含一个标题‘七巧板拼图’和几个占位按钮）。在游戏区中间画一个大正方形，代表未拆解的完整七巧板，现在不需要拆解它，也不需要复杂的交互，先把UI框架和基本的页面响应式布局搭好。请你在后续所有轮次的开发中都严格遵循两个要求：无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用；生成的代码文件中应优先以正斜杠 / 作为路径的分割符。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n\n # 遍历当前项目的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果是代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制形式读取文件内容\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试用 utf-8 进行解码，如果不抛出 UnicodeDecodeError 则说明是合法的 utf-8 编码格式\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码失败，说明文件不是 UTF-8 编码\n return False\n except Exception:\n # 遇到其他读取异常无法判断时，视为未满足要求\n return False\n\n # 如果至少存在一个代码文件，且全部验证通过，则返回 True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在或者不是一个目录，则无法进行判断\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 计算当前遍历到的目录相对于根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 根目录本身的相对路径通常为 '.'，代表第1层\n if rel_path == '.':\n depth = 1\n else:\n # 统一路径分隔符为 '/' 以兼容不同操作系统，通过分割路径来计算目录深度\n # 例如相对路径 'src' 长度为1，表示为根目录下的直接子目录，总层级为第2层\n # 例如相对路径 'src/components' 长度为2，总层级为第3层\n normalized_rel_path = rel_path.replace('\\\\', '/')\n depth = len(normalized_rel_path.split('/')) + 1\n \n # 若某目录所处深度大于2层，说明不符合要求，直接返回 False\n if depth > 2:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 校验变量名是否同时包含：数字、下划线、大写字母、小写字母\n def check_var_name(name: str) -> bool:\n has_digit = any(c.isdigit() for c in name)\n has_underscore = '_' in name\n has_upper = any(c.isupper() for c in name)\n has_lower = any(c.islower() for c in name)\n return has_digit and has_underscore and has_upper and has_lower\n\n variables = set()\n\n # 从 JavaScript 文本中提取可能的变量标识符\n def extract_from_js(js_code: str):\n # 去除单行和多行注释\n js_code = re.sub(r'//.*', '', js_code)\n js_code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', js_code)\n # 去除字符串字面量（避免匹配到字符串内部的内容）\n js_code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', js_code)\n js_code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", js_code)\n js_code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', js_code)\n \n # 1. 提取 var, let, const, function, class 紧接的标识符\n decl_pattern = r'\\b(?:var|let|const|function|class)\\s+([a-zA-Z_]\\w*)'\n for match in re.finditer(decl_pattern, js_code):\n variables.add(match.group(1))\n \n # 2. 提取普通函数参数：function [name](param1, param2)\n func_pattern = r'\\bfunction\\s*(?:[a-zA-Z_]\\w*)?\\s*\\(([^)]*)\\)'\n for match in re.finditer(func_pattern, js_code):\n params_str = match.group(1)\n for p in params_str.split(','):\n p = p.split('=')[0].strip() # 剔除默认值\n if p and re.match(r'^[a-zA-Z_]\\w*$', p):\n variables.add(p)\n \n # 3. 提取括号箭头函数参数：(param1, param2) =>\n arrow_pattern = r'\\(([^)]*)\\)\\s*=>'\n for match in re.finditer(arrow_pattern, js_code):\n params_str = match.group(1)\n for p in params_str.split(','):\n p = p.split('=')[0].strip()\n if p and re.match(r'^[a-zA-Z_]\\w*$', p):\n variables.add(p)\n \n # 4. 提取单参数箭头函数参数：param =>\n single_arrow_pattern = r'\\b([a-zA-Z_]\\w*)\\s*=>'\n for match in re.finditer(single_arrow_pattern, js_code):\n variables.add(match.group(1))\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目目录收集代码文件中的所有变量\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n extract_from_js(f.read())\n except Exception:\n pass\n \n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 提取 HTML 内联 JS\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for script in scripts:\n extract_from_js(script)\n \n # 提取 HTML 内联 CSS 变量\n styles = re.findall(r'<style.*?>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n for style in styles:\n style = re.sub(r'/\\*[\\s\\S]*?\\*/', '', style)\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', style)\n for css_var in css_vars:\n variables.add(css_var)\n except Exception:\n pass\n \n elif file.endswith('.css'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n # 提取 CSS 变量名\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', content)\n for css_var in css_vars:\n variables.add(css_var)\n except Exception:\n pass\n\n # 排除意外被正则捕获的常见JS语言关键字\n keywords = {'if', 'else', 'for', 'while', 'return', 'switch', 'case', 'catch', 'true', 'false', 'null', 'undefined'}\n variables = {v for v in variables if v not in keywords}\n\n if not variables:\n return True\n\n # 遍历所有捕捉到的变量，验证要求\n for var in variables:\n if not check_var_name(var):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断可能存在的代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n name, ext = os.path.splitext(file)\n \n # 判断该文件是否属于我们要检查的代码文件\n if ext.lower() in code_extensions:\n \n # 检查1：文件名主体只能包含小写字母和下划线。\n # 不能包含大写字母、数字，以及任何符号（如果存在多个点如 \"script.min.js\"，name 会是 \"script.min\"，也会被判定出含符号并返回False）\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n # 检查2：因为要求不能出现大写字母，所以扩展名也必须是全小写\n # （如 “.HTML” 虽属于上述文件类型，但由于包含大写字母而不被允许）\n if ext != ext.lower():\n return False\n \n # 所有的代码文件均符合要求，返回 True\n return True","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中应优先以正斜杠 / 作为路径的分割符","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":1,"instruction":"好，接下来我们需要把那个大正方形拆解成真正的七巧板。请在代码中定义标准的七巧板的7个基本图形（2个大直角三角形、1个中直角三角形、2个小直角三角形、1个正方形、1个平行四边形），并给它们分别填充7种不同的鲜明颜色。初始状态下，不要拼在一起，而是将这7个图形散落在游戏区底部或者控制面板下方的‘备选区’中。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，在更新任何已有文件之前，必须在同一级别目录下生成一份该文件被更改前的备份；第二，每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 依据用户指令推测产生的代码和文本文件后缀类型\n target_extensions = {'.html', '.css', '.js', '.md', '.json'}\n \n # 若工作区不存在，则无法验证，返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检测相关代码/文本文件\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果触发解码错误，说明存在文件不是使用 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能导致读取失败的文件操作级别错误\n pass\n \n # 所有文件均能以 UTF-8 正确读取，说明符合编码要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n\n base = Path(workspace_path).resolve()\n \n # 如果工作区目录不存在，可以视为没有创建深层目录，满足不大于2层的限制\n if not base.exists() or not base.is_dir():\n return True\n \n for item in base.rglob('*'):\n rel = item.relative_to(base)\n \n # 忽略以 '.' 开头的隐藏文件或文件夹，如 .git 等，避免误判\n if any(part.startswith('.') for part in rel.parts):\n continue\n \n if item.is_dir():\n # base作为第1层，直接子文件夹为第2层（len == 1）\n # 如果目录的相对层级 > 1，意味着出现了孙子目录（第3层或以上），违反规则\n if len(rel.parts) > 1:\n return False\n \n elif item.is_file():\n # 直接在base下的文件（len == 1），在第2层子目录下的文件（len == 2）\n # 如果文件的相对层级 > 2，意味着其处在超过两层的目录中，违反规则\n if len(rel.parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 捕获 var, let, const, function, class 后直接跟随的常规标识符作为自定义变量名\n var_pattern = re.compile(r'\\b(?:var|let|const|function|class)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 根据用户指令推断主要代码文件为 html 和 js 格式\n if ext in ['.js', '.html']:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 1. 移除多行注释和单行注释\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n content = re.sub(r'//.*', '', content)\n \n # 2. 移除 HTML 注释\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n \n # 3. 移除字符串内部的内容，防止将字符串内的话误认为代码变量声明\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n variables = var_pattern.findall(content)\n for var in variables:\n # 检查要求：同时包含数字、下划线、大写字母和小写字母\n has_num = bool(re.search(r'[0-9]', var))\n has_under = bool(re.search(r'_', var))\n has_upper = bool(re.search(r'[A-Z]', var))\n has_lower = bool(re.search(r'[a-z]', var))\n \n if not (has_num and has_under and has_upper and has_lower):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 明确代码文件的扩展名类型（基于用户多轮指令中对前端项目的要求）\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历当前项目空间中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件的后缀名\n _, ext = os.path.splitext(file)\n \n # 仅对代码文件进行检查\n if ext.lower() in code_extensions:\n has_code_file = True\n \n # 检查文件名是否仅包含小写字母、下划线以及唯一的一个点(作为扩展名的分隔)\n # ^[a-z_]+ : 匹配文件主名，仅限小写字母和下划线\n # \\. : 匹配必须存在的一个点（扩展名分隔符）\n # [a-z_]+$ : 匹配扩展名部分，同样仅限小写字母和下划线\n if not re.match(r'^[a-z_]+\\.[a-z_]+$', file):\n return False\n \n # 如果找到了代码文件且均通过检查，则返回 True；否则返回 False\n return has_code_file","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中应优先以正斜杠 / 作为路径的分割符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在更新任何已有文件之前，必须在同一级别目录下生成一份该文件被更改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":2,"instruction":"现在来增加核心交互：自由拖拽功能。请实现使用鼠标或触摸屏能够选中并自由拖动这7个板块的功能。要求：拖动过程中，当前被拖动的板块层级（z-index）必须处于最顶层；另外要限制拖动边界，板块不能被完全拖出游戏操作区之外；松开鼠标时板块能停留在放下的位置。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令；第二，原有的修改文件前需在同级目录创建备份的要求作废，更新为任何时候修改现有文件前，都不得在同级目录下创建该文件原始状态的备份。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码和文本文件的扩展名集合\n # 涵盖了用户指令中提到的 html, css, js, json, md，以及其他常规代码类型\n code_extensions = {\n '.html', '.css', '.js', '.json', '.md', \n '.py', '.java', '.c', '.cpp', '.h', \n '.ts', '.jsx', '.tsx', '.txt', '.xml', \n '.sh', '.bat'\n }\n \n # 如果工作区路径不存在，则说明无任何代码文件生成，尚未违反编码限制\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试将文件内容的字节流以utf-8进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦抛出解码错误，说明存在文件不是UTF-8编码，要求不满足\n return False\n except Exception:\n # 忽略其他可能的文件读取权限等非编码相关的系统异常\n pass\n \n # 所有代码文件全部成功用 utf-8 验证通过，或未发现任何代码文件\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n rel_path = os.path.relpath(root, workspace_path)\n if rel_path == '.':\n continue\n \n # 统一使用正斜杠处理相对路径，兼容不同操作系统\n rel_path = rel_path.replace('\\\\', '/')\n parts = [p for p in rel_path.split('/') if p]\n \n # 忽略以 \".\" 或 \"__\" 开头的系统、缓存及版本控制特殊目录\n if any(p.startswith('.') or p.startswith('__') for p in parts):\n continue\n \n # 根目录视为第1层，一级子目录（parts长度为1）视为第2层\n # 若存在 parts 长度大于等于 2 的目录，说明出现了第3层或更深层级的目录，违反条件\n if len(parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断变量名是否满足条件\n def check_var_name(name: str) -> bool:\n if not name:\n return True\n has_digit = any(c.isdigit() for c in name)\n has_lower = any(c.islower() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_under = '_' in name\n return has_digit and has_lower and has_upper and has_under\n\n # 辅助函数：检查 JavaScript / TypeScript 代码中的变量名\n def check_js_code(code: str) -> bool:\n # 去除单行和多行注释\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n \n # 移除字符串字面量（避免字符串内部存在疑似关键字的内容干扰匹配）\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', code, flags=re.DOTALL)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', code, flags=re.DOTALL)\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', code, flags=re.DOTALL)\n\n # 1. 匹配 let, var, const 后声明的变量\n matches = re.finditer(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', code)\n for m in matches:\n if not check_var_name(m.group(1)):\n return False\n \n # 2. 匹配函数声明的名称\n matches = re.finditer(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', code)\n for m in matches:\n if not check_var_name(m.group(1)):\n return False\n \n # 3. 匹配普通函数的参数名\n matches = re.finditer(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', code)\n for m in matches:\n params_str = m.group(1)\n if params_str.strip():\n for param in params_str.split(','):\n param = param.split('=')[0].strip()\n # 粗略跳过解构赋值的复杂参数，主要校验常规参数\n if param and '{' not in param and '[' not in param:\n if not check_var_name(param):\n return False\n \n # 4. 匹配类的名称\n matches = re.finditer(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', code)\n for m in matches:\n if not check_var_name(m.group(1)):\n return False\n \n # 5. 匹配箭头函数的参数名\n matches = re.finditer(r'(?:\\(([^)]*)\\)|([a-zA-Z_$][a-zA-Z0-9_$]*))\\s*=>', code)\n for m in matches:\n if m.group(1):\n params_str = m.group(1)\n for param in params_str.split(','):\n param = param.split('=')[0].strip()\n if param and '{' not in param and '[' not in param:\n if not check_var_name(param):\n return False\n elif m.group(2):\n param = m.group(2).strip()\n if param and not check_var_name(param):\n return False\n \n return True\n\n # 辅助函数：利用 AST 检查 Python 代码中的变量名\n def check_py_code(code: str) -> bool:\n try:\n tree = ast.parse(code)\n except SyntaxError:\n # 如果语法错误解析失败，则跳过\n return True\n \n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if not check_var_name(node.id):\n return False\n elif isinstance(node, ast.arg):\n if not check_var_name(node.arg):\n return False\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):\n if not check_var_name(node.name):\n return False\n return True\n\n # 遍历项目文件，提取相关代码\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理前端常见代码文件及可能的 Python 后端文件\n if ext in ['.js', '.ts', '.py', '.html']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext in ['.js', '.ts']:\n if not check_js_code(content):\n return False\n elif ext == '.py':\n if not check_py_code(content):\n return False\n elif ext == '.html':\n # 提取 HTML 内部的 JS 代码\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_js_code(script):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令，推断只检查 HTML, CSS 和 JS 作为“代码文件”\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历当前代码项目下所有的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 分离出文件名和扩展名\n basename, ext = os.path.splitext(file)\n \n # 判断是否为代码文件\n if ext.lower() in code_extensions:\n has_code_file = True\n \n # 检查主文件名是否仅由小写字母和下划线组成\n # 如果包含数字、大写字母或其他符号，均返回 False\n if not re.match(r'^[a-z_]+$', basename):\n return False\n \n # 如果没生成任何相关代码文件则返回False，否则所有代码文件均通过测试则返回True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中应优先以正斜杠 / 作为路径的分割符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"任何时候修改现有文件前，都不得在同级目录下创建该文件原始状态的备份","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":3,"instruction":"下一步是实现板块的旋转和翻转。请为板块添加旋转机制：选中或悬浮某个板块时，可以通过鼠标滚轮或者提供一个悬浮的“旋转手柄”，让板块能够顺时针或逆时针旋转（每次旋转45度）。特别注意的是，对于“平行四边形”，普通的平面旋转无法覆盖所有拼图情况，请额外增加一个翻转功能（比如双击该平行四边形使其水平或垂直镜像翻转）。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的修改文件前不得在同级目录创建备份的要求作废，更新为每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份；第二，在输出的所有代码文件里，不允许出现行数超过40行的函数。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义当前项目可能涉及到的代码文件及文本文件扩展名\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', \n '.json', '.md', '.txt', '.xml', '.svg',\n '.py', '.java', '.c', '.cpp', '.sh'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码检测\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试用 UTF-8 进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果不能以 UTF-8 解码，说明存在非 UTF-8 编码的文件，返回 False\n return False\n except Exception:\n # 忽略因权限等导致的其他系统级读取异常\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 获取工作空间的绝对路径\n workspace = Path(workspace_path).resolve()\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的无关隐藏目录（如.git），避免影响对“代码工程”真实层级的评判\n if '.git' in dirs:\n dirs.remove('.git')\n \n root_path = Path(root).resolve()\n \n try:\n # 计算当前目录相对工作空间根目录的相对路径\n rel_path = root_path.relative_to(workspace)\n except ValueError:\n # 万一遇到非子路径异常（通常不会发生），安全起见跳过\n continue\n \n # 工作空间根目录算第1层 (parts为空，len=0，深度=1)\n # 直接子目录算第2层 (parts如 ('js',)，len=1，深度=2)\n # 更深子目录算第3+层 (parts如 ('js', 'utils')，len=2，深度=3)\n depth = 1 + len(rel_path.parts)\n \n # 目录层级必须小于或等于2层\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 检查单个变量名是否满足“同时包含数字、下划线、大写字母和小写字母”的条件\n def check_var_name(name: str) -> bool:\n has_digit = bool(re.search(r'\\d', name))\n has_underscore = bool(re.search(r'_', name))\n has_upper = bool(re.search(r'[A-Z]', name))\n has_lower = bool(re.search(r'[a-z]', name))\n return has_digit and has_underscore and has_upper and has_lower\n\n # 从 JavaScript 文本中提取变量和参数名\n def extract_variables(code: str) -> set:\n variables = set()\n \n # 清除注释避免干扰\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n code = re.sub(r'//.*', '', code)\n \n # 清除字符串避免文本内部出现 let/var 等单词干扰提取\n code = re.sub(r\"'[^']*'\", \"''\", code)\n code = re.sub(r'\"[^\"]*\"', '\"\"', code)\n code = re.sub(r'`[^`]*`', '``', code)\n \n # 1. 匹配 let / var / const 声明后的首个变量名\n decl_matches = re.findall(r'\\b(?:let|var|const)\\s+([a-zA-Z_]\\w*)', code)\n variables.update(decl_matches)\n \n # 2. 匹配 function 声明时的参数列表\n func_matches = re.findall(r'\\bfunction(?:\\s+[a-zA-Z_]\\w*)?\\s*\\(([^)]*)\\)', code)\n for param_str in func_matches:\n # 去除可能存在的解构大括号\n param_str = re.sub(r'[{}]', '', param_str)\n for p in param_str.split(','):\n p = p.strip()\n if p:\n m = re.match(r'^[a-zA-Z_]\\w*', p)\n if m:\n variables.add(m.group(0))\n \n # 3. 匹配常规箭头函数参数列表 (param1, param2) =>\n arrow_matches = re.findall(r'\\(([^)]*)\\)\\s*=>', code)\n for param_str in arrow_matches:\n param_str = re.sub(r'[{}]', '', param_str)\n for p in param_str.split(','):\n p = p.strip()\n if p:\n m = re.match(r'^[a-zA-Z_]\\w*', p)\n if m:\n variables.add(m.group(0))\n \n # 4. 匹配单参数箭头函数参数 param =>\n arrow_single = re.findall(r'\\b([a-zA-Z_]\\w*)\\s*=>', code)\n variables.update(arrow_single)\n \n return variables\n\n # 从 CSS 文本中提取自定义属性（CSS变量）\n def extract_css_variables(code: str) -> set:\n variables = set()\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # CSS 变量声明示例: --var-name: value;\n matches = re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', code)\n variables.update(matches)\n return variables\n\n all_vars = set()\n has_code_files = False\n\n # 验证工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作区，处理 HTML, JS, CSS 文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n if file.endswith('.js'):\n has_code_files = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n all_vars.update(extract_variables(code))\n except Exception:\n pass\n \n elif file.endswith('.css'):\n has_code_files = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n all_vars.update(extract_css_variables(code))\n except Exception:\n pass\n \n elif file.endswith('.html'):\n has_code_files = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 解析内联 script 块\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n for script in scripts:\n all_vars.update(extract_variables(script))\n \n # 解析内联 style 块\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.DOTALL | re.IGNORECASE)\n for style in styles:\n all_vars.update(extract_css_variables(style))\n except Exception:\n pass\n\n # 若没有任何相关代码文件或未提取出任何变量（七巧板游戏必然存在变量，除非未按要求生成）\n if not has_code_files or not all_vars:\n return False\n\n # 校验发现的所有变量名\n for var in all_vars:\n if not check_var_name(var):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check_filename_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义项目中可能涉及到的前端代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并转为小写进行判断\n ext = os.path.splitext(file)[1].lower()\n \n # 判断当前文件是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n \n # 检查文件名是否严格符合要求：仅包含小写字母、下划线\n # 允许且仅允许最后带有由一个点和纯小写字母组成的常规扩展名\n # 大写字母、数字、连字符或其他任何多余符号都会导致正则匹配失败\n if not re.match(r'^[a-z_]+(\\.[a-z]+)?$', file):\n return False\n \n # 如果项目中没有任何代码文件返回 False，否则说明所有代码文件名均合规并返回 True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中应优先以正斜杠 / 作为路径的分割符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在输出的所有代码文件里，不允许出现行数超过40行的函数","validation_code":"def is_valid(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n import ast\n\n def remove_strings_and_comments(code: str) -> str:\n def replacer(match):\n # 将匹配到的字符串或注释替换为相应数量的换行符，以保证代码总行数和对应的行号不变\n return '\\n' * match.group(0).count('\\n')\n \n # 匹配单行/多行注释、双引号/单引号/反引号包含的字符串字面量\n pattern = re.compile(\n r'//.*|/\\*[\\s\\S]*?\\*/|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n )\n return pattern.sub(replacer, code)\n\n def check_js_code(code: str) -> bool:\n cleaned = remove_strings_and_comments(code)\n \n # 预计算所有换行符的位置，实现 O(log N) 的行号查询\n newline_indices = [i for i, char in enumerate(cleaned) if char == '\\n']\n def get_line_number(index: int) -> int:\n return bisect.bisect_right(newline_indices, index) + 1\n\n func_brace_indices = set()\n \n # 1. 处理由 function 和 => 引导的函数定义\n for m in re.finditer(r'\\bfunction\\b|=>', cleaned):\n paren_level = 0\n idx = m.end()\n while idx < len(cleaned):\n c = cleaned[idx]\n if c == '(':\n paren_level += 1\n elif c == ')':\n paren_level -= 1\n elif c == '{':\n if paren_level <= 0:\n func_brace_indices.add(idx)\n break\n elif c == '}':\n if paren_level <= 0:\n break\n elif c == ';':\n if paren_level <= 0:\n break\n idx += 1\n\n # 2. 处理对象中的方法简写 (例如 methodName(...) { )\n method_pattern = re.compile(r'\\b(?!(?:if|for|while|switch|catch|function)\\b)([a-zA-Z_]\\w*)\\s*\\(')\n for m in method_pattern.finditer(cleaned):\n paren_level = 0\n idx = m.end() - 1 # idx 落在此处的 '(' 上\n while idx < len(cleaned):\n c = cleaned[idx]\n if c == '(':\n paren_level += 1\n elif c == ')':\n paren_level -= 1\n elif c == '{':\n if paren_level <= 0:\n func_brace_indices.add(idx)\n break\n elif c == '}':\n if paren_level <= 0:\n break\n elif c == ';':\n if paren_level <= 0:\n break\n idx += 1\n\n stack = []\n for match in re.finditer(r'\\{|\\}', cleaned):\n char = match.group(0)\n idx = match.start()\n if char == '{':\n is_func = (idx in func_brace_indices)\n stack.append((is_func, idx))\n else:\n if stack:\n is_func, start_idx = stack.pop()\n if is_func:\n start_line = get_line_number(start_idx)\n end_line = get_line_number(idx)\n if end_line - start_line + 1 > 40:\n return False\n return True\n\n def check_html_code(code: str) -> bool:\n script_pattern = re.compile(r'<script\\b[^>]*>([\\s\\S]*?)</script>', re.IGNORECASE)\n for match in script_pattern.finditer(code):\n script_content = match.group(1)\n # 添加前缀换行符保持代码片段的行号映射准确度\n prefix_newlines = '\\n' * code.count('\\n', 0, match.start(1))\n if not check_js_code(prefix_newlines + script_content):\n return False\n return True\n\n def check_py_code(code: str) -> bool:\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n if node.end_lineno - node.lineno + 1 > 40:\n return False\n except Exception:\n pass\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 针对常见的 Web 及相关代码文件类型做判定\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_js_code(code):\n return False\n except Exception:\n pass\n elif file.endswith(('.html', '.htm')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_html_code(code):\n return False\n except Exception:\n pass\n elif file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_py_code(code):\n return False\n except Exception:\n pass\n \n return True","tags":["Quantity","Range"]}]},{"round_id":4,"instruction":"为了让玩家拼图时对齐更方便，我们需要加入“吸附（Snapping）”效果。当玩家拖动一个板块靠近其他已经放置好的板块的顶点或边缘时，如果距离小于设定的阈值（比如15像素），自动让它吸附并对齐。这样可以避免手动微调的麻烦，提升游戏体验。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的输出的所有代码文件里不允许出现行数超过40行的函数的要求作废，更新为在输出的所有代码文件里，不允许出现行数超过30行的函数；第二，强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-，该要求后续所有轮次均需严格遵循。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断的可能出现的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py'}\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并转为小写匹配\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 其他不可预见的读取错误（如权限问题）暂时忽略，避免误判\n continue\n\n # 若所有代码文件都能成功以 UTF-8 读取（或者没有代码文件），则认为符合该要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"from pathlib import Path\n\ndef check(response: str, workspace_path: str) -> bool:\n workspace = Path(workspace_path)\n \n # 确保给定的 workspace_path 存在且为目录\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n # 递归遍历项目目录中的所有文件和文件夹\n for path in workspace.rglob('*'):\n # 获取当前路径相对于根目录的各部分\n rel_parts = path.relative_to(workspace).parts\n \n # 排除包含隐藏文件夹（以 . 开头）或特殊环境文件夹的路径，避免干扰判定\n if any(part.startswith('.') for part in rel_parts):\n continue\n if 'node_modules' in rel_parts or '__pycache__' in rel_parts:\n continue\n \n # 若某目录相对根目录的分段数 >= 2 (如 src/utils)\n # 意味着它是第二级或更深层级的子目录，即目录工程层级达到了 3 层及以上\n if path.is_dir() and len(rel_parts) >= 2:\n return False\n \n # 若某文件相对根目录的分段数 >= 3 (如 src/utils/main.py)\n # 同样意味着它深于或等于第 3 层的目录之中\n if path.is_file() and len(rel_parts) >= 3:\n return False\n \n # 如果全项目均未超出层级限制，返回 True\n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 辅助函数：判断变量名是否同时包含小写、大写、数字和下划线\n def has_all_required_chars(name: str) -> bool:\n has_lower = any(c.islower() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_digit = any(c.isdigit() for c in name)\n has_under = '_' in name\n return has_lower and has_upper and has_digit and has_under\n\n # 辅助函数：移除JS/CSS代码中的注释，防止正则误匹配\n def remove_js_comments(text: str) -> str:\n # 移除多行注释 /* ... */\n text = re.sub(r'/\\*.*?\\*/', '', text, flags=re.DOTALL)\n # 移除单行注释 // ...\n text = re.sub(r'//.*', '', text)\n return text\n\n variables_found = []\n\n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # 兼容处理可能的Python文件\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n variables_found.append(node.id)\n elif isinstance(node, ast.arg):\n variables_found.append(node.arg)\n except Exception:\n pass\n\n # 处理JavaScript文件\n elif file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n content = remove_js_comments(content)\n \n # 匹配 var, let, const 声明的变量名\n decls = re.findall(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content)\n variables_found.extend(decls)\n \n # 匹配函数的参数变量\n func_params = re.findall(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', content)\n for param_str in func_params:\n params = [p.strip() for p in param_str.split(',') if p.strip()]\n for p in params:\n p_name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\n variables_found.append(p_name)\n except Exception:\n pass\n\n # 处理HTML文件中的嵌入式JS代码\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 移除HTML注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 提取所有script标签中的内容\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n script = remove_js_comments(script)\n # 匹配 var, let, const 声明的变量名\n decls = re.findall(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', script)\n variables_found.extend(decls)\n \n # 匹配函数的参数变量\n func_params = re.findall(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', script)\n for param_str in func_params:\n params = [p.strip() for p in param_str.split(',') if p.strip()]\n for p in params:\n p_name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\n variables_found.append(p_name)\n except Exception:\n pass\n\n # 处理CSS文件中的CSS变量（Custom Properties）\n elif file.endswith('.css'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n content = remove_js_comments(content)\n \n # 匹配CSS变量定义，例如 --main-color: ...\n css_vars = re.findall(r'(--[a-zA-Z0-9_-]+)\\s*:', content)\n variables_found.extend(css_vars)\n except Exception:\n pass\n\n # 过滤掉可能的空值\n variables_found = [v for v in variables_found if v]\n\n # 若未找到任何变量，默认算作满足（通常代码中肯定存在变量）\n if not variables_found:\n return True\n\n # 判断是否每一个变量都符合要求\n for var in variables_found:\n if not has_all_required_chars(var):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check_code_filenames(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', \n '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', \n '.rb', '.php', '.sh', '.swift', '.kt'\n }\n \n # 遍历当前项目工作空间下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件的扩展名\n _, ext = os.path.splitext(file)\n \n # 判断是否为代码文件（使用 lower() 以覆盖可能含有大写后缀的情况并进行识别）\n if ext.lower() in code_extensions:\n # 检查文件名是否严格符合要求：\n # 除了唯一的一个扩展名圆点外，文件基础名及扩展名中只允许存在小写字母(a-z)和下划线(_)\n # 不能包含数字、大写字母、多个圆点或其他任何特殊符号\n if not re.fullmatch(r'[a-z_]+\\.[a-z_]+', file):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中应优先以正斜杠 / 作为路径的分割符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在输出的所有代码文件里，不允许出现行数超过30行的函数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def replacer(match):\n s = match.group(0)\n # 替换为空格，但保留原有的换行符，以便维持正确的行号对应关系不变\n return '\\n' * s.count('\\n') + ' ' * (len(s) - s.count('\\n'))\n\n def clean_js_code(code_str: str) -> str:\n # 正则匹配 JS 中的字符串字面量(支持转义符)、块注释、行注释\n pattern = r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)' \\\n r'|(?P<bcomment>/\\*[\\s\\S]*?\\*/)' \\\n r'|(?P<lcomment>//[^\\n]*)'\n return re.sub(pattern, replacer, code_str)\n\n def get_js_func_ranges(code_str: str) -> list:\n cleaned = clean_js_code(code_str)\n func_ranges = []\n start_indices = []\n \n # 1. 匹配 function 关键字\n for m in re.finditer(r'\\bfunction\\b', cleaned):\n brace_idx = cleaned.find('{', m.end())\n if brace_idx != -1:\n start_indices.append((m.start(), brace_idx))\n \n # 2. 匹配箭头函数 => {\n for m in re.finditer(r'=>\\s*\\{', cleaned):\n brace_idx = cleaned.find('{', m.start())\n if brace_idx != -1:\n start_indices.append((m.start(), brace_idx))\n \n # 3. 匹配可能的对象方法或类方法的语法（排除 if/for 等控制流关键字）\n pattern = r'\\b(?!(?:if|for|while|switch|catch|with|function)\\b)([a-zA-Z_$][a-zA-Z_$0-9]*)\\s*\\([^)]*\\)\\s*\\{'\n for m in re.finditer(pattern, cleaned):\n brace_idx = cleaned.find('{', m.start())\n if brace_idx != -1:\n start_indices.append((m.start(), brace_idx))\n \n # 坐标去重，防止相同位置被多种规则重复捕捉\n start_indices = list(set(start_indices))\n \n for start_idx, brace_idx in start_indices:\n stack = 0\n end_idx = -1\n # 利用栈进行花括号配对，寻找函数结束位置\n for i in range(brace_idx, len(cleaned)):\n if cleaned[i] == '{':\n stack += 1\n elif cleaned[i] == '}':\n stack -= 1\n if stack == 0:\n end_idx = i\n break\n \n # 若配对成功，则根据 \\n 计算函数所跨越的行数\n if end_idx != -1:\n start_line = code_str.count('\\n', 0, start_idx) + 1\n end_line = code_str.count('\\n', 0, end_idx) + 1\n func_ranges.append((start_line, end_line))\n \n return func_ranges\n\n def check_py_code(code_str: str) -> bool:\n try:\n tree = ast.parse(code_str)\n except SyntaxError:\n return True # 语法错误无法解析时，略过不将其判定为包含长函数\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno:\n lines = node.end_lineno - node.lineno + 1\n if lines > 30:\n return False\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n continue\n \n # 针对 JavaScript / TypeScript 类型文件\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\n ranges = get_js_func_ranges(content)\n for start, end in ranges:\n if (end - start + 1) > 30:\n return False\n \n # 针对包含内嵌脚本的 HTML 文件\n elif ext in ['.html', '.htm']:\n # 提取 <script> ... </script> 块内容\n for script_match in re.finditer(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE):\n script_content = script_match.group(1)\n # 补齐脚本前面的换行符，以便维持在原文件中的正确行号\n start_pos = script_match.start(1)\n newlines_before = content.count('\\n', 0, start_pos)\n padded_script = '\\n' * newlines_before + script_content\n \n ranges = get_js_func_ranges(padded_script)\n for start, end in ranges:\n if (end - start + 1) > 30:\n return False\n \n # 针对可能存在的 Python 代码文件\n elif ext == '.py':\n if not check_py_code(content):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_coding_declaration(response: str, workspace_path: str) -> bool:\n # 根据用户指令内容推断，可能的代码文件扩展名为HTML, CSS, JS\n code_extensions = {'.html', '.css', '.js'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取顶行内容\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件读取发生异常（如编码错误或权限问题），判定为不满足要求\n return False\n \n # 如果所有的代码文件均满足要求，则返回 True\n return True","tags":["Content","Format"]}]},{"round_id":5,"instruction":"基础操作都有了，现在引入真正的游戏关卡。请在右侧控制面板增加3个关卡按钮（例如：狐狸、帆船、天鹅）。点击任意一个关卡，在游戏区中央渲染出一个由7块七巧板特定角度组合而成的黑色阴影轮廓作为“目标图案”。这个轮廓没有内部线条，供玩家参照去拼凑。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的输出的所有代码文件里不允许出现行数超过30行的函数的要求作废，更新为在输出的所有代码文件里，不允许出现行数超过25行的函数；第二，项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起，该要求后续所有轮次均需严格遵循。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令推断可能存在的代码和配置文件后缀\n code_extensions = {\n '.html', '.css', '.js', '.json', \n '.md', '.log', '.txt', '.py', \n '.ts', '.jsx', '.tsx', '.sh'\n }\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，则说明包含了非 UTF-8 编码的字符\n return False\n except Exception:\n # 忽略其他非编码相关的偶发异常（如权限异常等）\n pass\n \n # 如果所有代码文件都能正常被 UTF-8 成功解码，或没有发现异常，说明满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 如果工作区路径不存在，则无法判断\n if not os.path.exists(workspace_path):\n return False\n \n workspace = Path(workspace_path).resolve()\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉以 '.' 开头的隐藏文件夹（如 .git, .idea等），防止对层级判断造成干扰\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n root_path = Path(root).resolve()\n try:\n # 计算当前遍历到的目录相对于根目录的相对路径\n rel_path = root_path.relative_to(workspace)\n except ValueError:\n continue\n \n # 对于根目录，rel_path.parts 为 ()，长度为 0（代表第1层级）\n # 对于一级子目录，rel_path.parts 为 ('dir1',)，长度为 1（代表第2层级）\n # 对于二级子目录，rel_path.parts 为 ('dir1', 'dir2')，长度为 2（代表第3层级）\n # 若总层级要求小于等于2层，则相对路径的层级部分长度不得大于1\n if len(rel_path.parts) > 1:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def check_name(name: str) -> bool:\n has_upper = bool(re.search(r'[A-Z]', name))\n has_lower = bool(re.search(r'[a-z]', name))\n has_digit = bool(re.search(r'[0-9]', name))\n has_under = '_' in name\n return has_upper and has_lower and has_digit and has_under\n\n js_var_pattern = re.compile(r'\\b(?:let|var|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理预期中可能生成的代码文件\n if ext not in ['.js', '.html', '.css', '.py']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if not check_name(node.id):\n return False\n elif isinstance(node, ast.arg):\n if not check_name(node.arg):\n return False\n except Exception:\n pass\n \n elif ext in ['.js', '.html']:\n text_to_search = content\n if ext == '.html':\n # 针对HTML，仅提取 script 标签内容进行检验\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n text_to_search = \"\\n\".join(scripts)\n \n # 预处理：清除单行、多行注释以及反引号、双引号、单引号包围的字符串字面量，避免被正则误认为变量声明\n text_to_search = re.sub(r'//.*', '', text_to_search)\n text_to_search = re.sub(r'/\\*.*?\\*/', '', text_to_search, flags=re.DOTALL)\n text_to_search = re.sub(r'`.*?`', '``', text_to_search, flags=re.DOTALL)\n text_to_search = re.sub(r'\".*?(?<!\\\\)\"', '\"\"', text_to_search)\n text_to_search = re.sub(r\"'.*?(?<!\\\\)'\", \"''\", text_to_search)\n \n # 开始匹配通过 let / var / const 声明的变量\n for match in js_var_pattern.findall(text_to_search):\n if not check_name(match):\n return False\n \n elif ext == '.css':\n # 预处理：清除多行注释\n text_to_search = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 开始匹配通过 -- 声明的 CSS 自定义属性（CSS变量）\n for match in re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', text_to_search):\n # 遇到类似 --my-var: 的形式，取出 my-var 进行验证\n # 考虑到 CSS 变量常使用破折号连字符，在要求强制使用下划线、大小写、数字的情况下仍应进行统一验证\n if not check_name(match):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证代码文件名是否只包含小写字母和下划线（除了扩展名的点以外）。\n \"\"\"\n # 常见代码文件扩展名集合\n code_exts = {\n '.js', '.html', '.css', '.py', '.ts', '.jsx', '.tsx', \n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt',\n '.vue', '.svelte', '.scss', '.sass', '.less'\n }\n \n # 确保当前代码项目地址存在\n if not os.path.exists(workspace_path):\n return False\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 忽略隐藏文件如 .DS_Store 或 .gitignore\n if file_name.startswith('.'):\n continue\n \n # 提取文件扩展名并判断是否属于代码文件\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_exts:\n found_code_file = True\n \n # 剔除文件名中的点符号\n name_without_dots = file_name.replace('.', '')\n \n # 正则匹配：除了点符号外，文件名应当仅仅由小写字母和下划线组成\n if not re.fullmatch(r'[a-z_]+', name_without_dots):\n return False\n \n # 如果没有找到任何代码文件，说明生成失败，返回 False\n return found_code_file","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件中应优先以正斜杠 / 作为路径的分割符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在输出的所有代码文件里，不允许出现行数超过25行的函数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def remove_comments_and_strings(code: str) -> str:\n # 替换匹配到的字符串及注释为等长的空格，同时保留所有的换行符，以便不影响行数计算\n def replacer(match):\n s = match.group(0)\n return ''.join('\\n' if c == '\\n' else ' ' for c in s)\n \n pattern = re.compile(\n r'<!--.*?-->|' # HTML 注释\n r'//.*?$|' # 单行注释\n r'/\\*.*?\\*/|' # 多行注释\n r\"'(?:\\\\.|[^'\\\\])*'|\" # 单引号字符串\n r'\"(?:\\\\.|[^\"\\\\])*\"|' # 双引号字符串\n r'`(?:\\\\.|[^`\\\\])*`', # 模板字符串\n re.DOTALL | re.MULTILINE\n )\n return pattern.sub(replacer, code)\n\n def get_function_blocks(clean_code: str):\n # 使用字典去重：以 { 的索引为键，保存函数声明能够匹配到的最早的索引为值\n blocks = {}\n \n # 1. 匹配 function 关键字声明的函数\n for match in re.finditer(r'\\bfunction\\b[^{]*\\{', clean_code):\n b_idx = match.end() - 1\n blocks[b_idx] = min(blocks.get(b_idx, len(clean_code)), match.start())\n \n # 2. 匹配箭头函数\n for match in re.finditer(r'=>\\s*\\{', clean_code):\n b_idx = match.end() - 1\n blocks[b_idx] = min(blocks.get(b_idx, len(clean_code)), match.start())\n \n # 3. 匹配常规的类方法或对象方法\n for match in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{', clean_code):\n name = match.group(1)\n # 排除由于具有括号和花括号特征而被误判的控制流和内置关键字\n if name not in {'if', 'for', 'while', 'switch', 'catch', 'with', 'return', 'typeof', 'new', 'else'}:\n b_idx = match.end() - 1\n blocks[b_idx] = min(blocks.get(b_idx, len(clean_code)), match.start())\n \n return [(v, k) for k, v in blocks.items()]\n\n def check_js_html_lines(code: str, max_lines: int = 25) -> bool:\n clean_code = remove_comments_and_strings(code)\n blocks = get_function_blocks(clean_code)\n \n for decl_start_idx, brace_start_idx in blocks:\n count = 1\n i = brace_start_idx + 1\n while i < len(clean_code):\n if clean_code[i] == '{':\n count += 1\n elif clean_code[i] == '}':\n count -= 1\n if count == 0:\n end_idx = i\n # 截取函数定义到结束的片段，计算行数\n func_str = clean_code[decl_start_idx:end_idx+1]\n lines = func_str.count('\\n') + 1\n if lines > max_lines:\n return False\n break\n i += 1\n return True\n\n def check_python_lines(code: str, max_lines: int = 25) -> bool:\n try:\n tree = ast.parse(code)\n except Exception:\n # 存在语法错误的情况暂且放过\n return True\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n lines = node.end_lineno - node.lineno + 1\n if lines > max_lines:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n has_files = False\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.py', '.html'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n has_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_python_lines(code, 25):\n return False\n elif ext == '.html':\n # HTML文件需提取所有 script 内容作为 JS 进行分析\n scripts = []\n pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\n for match in pattern.finditer(code):\n scripts.append(match.group(1))\n js_code = '\\n'.join(scripts)\n if not check_js_html_lines(js_code, 25):\n return False\n else:\n if not check_js_html_lines(code, 25):\n return False\n \n return True if has_files else False","tags":["Quantity","Range"]},{"constraint":"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义项目中可能出现的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的第一行\n first_line = f.readline()\n \n # 判断首行是否包含特定的编码声明内容\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法打开或发生编码读取错误，则判定为不满足要求\n return False\n \n # 如果项目中没有代码文件，则说明没有满足生成代码的要求，返回 False；反之返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":6,"instruction":"有了目标之后，需要加上通关检测逻辑。当玩家将所有7个板块都拖入黑色阴影轮廓内，并且每块的坐标和旋转角度与目标数据匹配（允许一定的容差值）时，判定为拼图成功。此时请弹出一个“恭喜通关”的模态框，并在屏幕上播放一些简单的彩纸飘落或烟花庆祝动画。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的生成的代码文件中应优先以正斜杠 / 作为路径的分割符的要求作废，更新为在编写的代码文件中，务必优先挑选反斜杠 \\ 来表示路径分隔；第二，原有的输出的所有代码文件里不允许出现行数超过25行的函数的要求作废，更新为生成代码时，请将各函数的行数严格控制在 20 行及以内。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码与相关配置文件扩展名集合\n code_exts = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \n '.go', '.rs', '.php', '.rb', '.swift', '.kt',\n '.json', '.md', '.log', '.txt', '.yaml', '.yml', \n '.xml', '.sh', '.bat'\n }\n \n # 若工作区路径不存在，视作未生成符合条件的代码工程\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 格式读取文件全部内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，说明文件非 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能的文件读取异常（如权限问题等）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_directory_depth(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 如果工作空间路径不存在或不是一个目录，直接判定为不满足\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n workspace = Path(workspace_path).resolve()\n\n for root, dirs, files in os.walk(workspace):\n # 过滤掉以 '.' 开头的隐藏文件夹（如 .git, .vscode 等），防止对判定产生干扰\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 检查当前目录下的所有文件和文件夹\n for name in dirs + files:\n # 忽略隐藏文件\n if name.startswith('.'):\n continue\n \n item_path = Path(root) / name\n \n try:\n # 获取相对于工程根目录的相对路径\n rel_path = item_path.relative_to(workspace)\n \n # rel_path.parts 可以获取相对路径层级切分后的元组\n # 例如 'js/main.js' 会被切分为 ('js', 'main.js')，长度为 2\n if len(rel_path.parts) > 2:\n return False\n except ValueError:\n pass\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_var_name(name: str) -> bool:\n \"\"\"检查变量名是否同时包含数字、下划线、大写字母和小写字母\"\"\"\n has_digit = any(c.isdigit() for c in name)\n has_under = '_' in name\n has_upper = any(c.isupper() for c in name)\n has_lower = any(c.islower() for c in name)\n return has_digit and has_under and has_upper and has_lower\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 针对 Python 文件的变量判断\n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 判断赋值变量名\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n if not check_var_name(node.id):\n return False\n # 判断函数参数变量名\n elif isinstance(node, ast.arg):\n if not check_var_name(node.arg):\n return False\n except SyntaxError:\n pass\n \n # 针对 JS/TS 及 HTML 里的 JS 代码部分做变量判断\n if ext in ['.js', '.ts', '.html']:\n if ext == '.html':\n # 提取 <script> 标签内容\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\n js_content = '\\n'.join(scripts)\n else:\n js_content = content\n \n if js_content.strip():\n # 粗略地去除字符串和注释，减少匹配误判\n js_content = re.sub(r'([\"\\'`]).*?\\1', '\"\"', js_content, flags=re.DOTALL)\n js_content = re.sub(r'/\\*.*?\\*/', '', js_content, flags=re.DOTALL)\n js_content = re.sub(r'//.*', '', js_content)\n \n # 匹配 var, let, const 声明的变量\n statements = re.findall(r'\\b(?:var|let|const)\\s+([^;\\n]+)', js_content)\n for stmt in statements:\n parts = stmt.split(',')\n for part in parts:\n var_name = part.split('=')[0].strip()\n # 提取开头的合法标识符\n m = re.match(r'^([a-zA-Z_$][0-9a-zA-Z_$]*)', var_name)\n if m:\n name = m.group(1)\n if name not in ['typeof', 'yield', 'await']:\n if not check_var_name(name):\n return False\n \n # 匹配函数参数\n funcs = re.findall(r'\\bfunction\\s*(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\s*\\(([^)]*)\\)', js_content)\n for params_str in funcs:\n params = params_str.split(',')\n for p in params:\n p = p.strip()\n m = re.match(r'^([a-zA-Z_$][0-9a-zA-Z_$]*)', p)\n if m:\n name = m.group(1)\n if not check_var_name(name):\n return False\n\n # 针对 CSS 及 HTML 里的 CSS 代码做变量判断（检测CSS自定义属性/变量）\n if ext in ['.css', '.html']:\n # 寻找形如 --my-var 的CSS变量\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\s*:', content)\n for css_var in css_vars:\n if not check_var_name(css_var):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check_code_filename(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见代码文件的扩展名集合\n code_exts = {\n '.html', '.htm', '.css', '.js', '.ts', \n '.jsx', '.tsx', '.py', '.java', '.c', \n '.cpp', '.h', '.hpp', '.cs', '.go', \n '.rb', '.php', '.swift', '.kt', '.rs', \n '.sh', '.bash'\n }\n \n has_code_file = False\n \n # 遍历当前项目的工作区文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件的扩展名并转小写进行归类匹配\n ext = os.path.splitext(file)[1].lower()\n \n # 仅针对代码文件进行校验，跳过文档和配置文件 (.md, .json, .log 等)\n if ext in code_exts:\n has_code_file = True\n \n # 规则1：除了作为扩展名分隔符的点之外，不得出现其他任何点符号（例如不允许出现多重后缀 app.min.js）\n if file.count('.') > 1:\n return False\n \n # 规则2：整个文件名中（去掉唯一的扩展名分隔符 `.` 后），只能包含小写字母和下划线\n # 任何大写字母、数字、连字符等其他符号均会导致正则匹配失败\n name_without_dot = file.replace('.', '')\n if not re.fullmatch(r'[a-z_]+', name_without_dot):\n return False\n \n # 如果项目中成功生成了至少一个符合规范的代码文件，且没有发现任何违规项，则要求被满足\n return has_code_file","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写的代码文件中，务必优先挑选反斜杠 \\ 来表示路径分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在 20 行及以内","validation_code":"import os\nimport ast\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \n def check_js_code(code: str) -> bool:\n \"\"\"\n 验证JavaScript或HTML script标签内的函数长度\n \"\"\"\n # 第一步：清洗代码内的注释与字符串，保留换行符使代码行号不变\n cleaned = []\n i = 0\n n = len(code)\n state = 'NORMAL'\n while i < n:\n c = code[i]\n if state == 'NORMAL':\n if i < n - 1 and code[i:i+2] == '//':\n state = 'COM_S'\n cleaned.append(' ')\n i += 2\n elif i < n - 1 and code[i:i+2] == '/*':\n state = 'COM_M'\n cleaned.append(' ')\n i += 2\n elif c == \"'\":\n state = 'STR_S'\n cleaned.append(' ')\n i += 1\n elif c == '\"':\n state = 'STR_D'\n cleaned.append(' ')\n i += 1\n elif c == '`':\n state = 'STR_B'\n cleaned.append(' ')\n i += 1\n else:\n cleaned.append(c)\n i += 1\n elif state == 'COM_S':\n if c == '\\n':\n state = 'NORMAL'\n cleaned.append('\\n')\n else:\n cleaned.append(' ')\n i += 1\n elif state == 'COM_M':\n if i < n - 1 and code[i:i+2] == '*/':\n state = 'NORMAL'\n cleaned.append(' ')\n i += 2\n else:\n cleaned.append('\\n' if c == '\\n' else ' ')\n i += 1\n elif state in ('STR_S', 'STR_D', 'STR_B'):\n if c == '\\\\':\n cleaned.append(' ')\n i += 1\n if i < n:\n c2 = code[i]\n cleaned.append('\\n' if c2 == '\\n' else ' ')\n i += 1\n continue\n elif (state == 'STR_S' and c == \"'\") or \\\n (state == 'STR_D' and c == '\"') or \\\n (state == 'STR_B' and c == '`'):\n state = 'NORMAL'\n cleaned.append(' ')\n i += 1\n else:\n cleaned.append('\\n' if c == '\\n' else ' ')\n i += 1\n\n cleaned_code = \"\".join(cleaned)\n \n # 第二步：利用正则提取清理过后代码里的函数定义开头部分\n p1 = r'\\bfunction(?:\\s+\\w+)?\\s*\\([^)]*\\)\\s*\\{'\n p2 = r'(?:\\([^)]*\\)|\\b\\w+)\\s*=>\\s*\\{'\n p3 = r'\\b(?!(?:if|for|while|catch|switch)\\b)\\w+\\s*\\([^)]*\\)\\s*\\{'\n pattern = re.compile(f'({p1}|{p2}|{p3})')\n \n # 第三步：跟踪匹配并计算每个函数的真实行数\n for match in pattern.finditer(cleaned_code):\n start_idx = match.start()\n brace_idx = match.end() - 1\n \n brace_count = 1\n end_idx = -1\n for j in range(brace_idx + 1, len(cleaned_code)):\n if cleaned_code[j] == '{':\n brace_count += 1\n elif cleaned_code[j] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_idx = j\n break\n \n if end_idx != -1:\n # 行数为对应字符串中换行符数加一\n lines_count = code[start_idx:end_idx+1].count('\\n') + 1\n if lines_count > 20:\n return False\n \n return True\n\n def check_py_code(file_path: str) -> bool:\n \"\"\"\n 验证Python文件内的函数长度\n \"\"\"\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n code = f.read()\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n lines_count = node.end_lineno - node.lineno + 1\n if lines_count > 20:\n return False\n return True\n except Exception:\n return True # 解析异常跳过\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目文件树\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n if ext == '.py':\n if not check_py_code(file_path):\n return False\n elif ext == '.js':\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n code = f.read()\n if not check_js_code(code):\n return False\n elif ext in ('.html', '.htm'):\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 寻找html中的script块代码进行分析\n scripts = re.findall(r'(?i)<script[^>]*>(.*?)</script>', content, re.DOTALL)\n for script in scripts:\n if script.strip():\n if not check_js_code(script):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名集合，专门用于过滤出代码文件\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.h', '.java'}\n code_files_found = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验，跳过 .md, .json, .log 等非代码文件\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 读取可以自动跳过不可见的 BOM 头信息\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若文件无法读取（例如空文件或编码异常读错），则视为不满足要求\n return False\n\n # 若遍历完毕且中途没有返回 False，需确保至少验证了一个代码文件\n return code_files_found","tags":["Content","Format"]},{"constraint":"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":7,"instruction":"功能越来越完善了。现在来优化游戏辅助功能：加入计时器和重置功能。当玩家在选中某个关卡并开始拖动第一块七巧板时，计时器开始走秒（格式 mm:ss），显示在控制面板上。同时增加一个“重置”按钮，点击后将当前的7个板块瞬间带动画飞回到初始的备选区位置，同时计时器归零重置。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的生成代码时，请将各函数的行数严格控制在20行及以内的要求作废，更新为生成代码时，请将各函数的行数严格控制在15行及以内；第二，后续所有轮次生成的每个代码文件，算上所有的空行的总行数必须保持为奇数，该要求需严格遵循。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令内容，推断可能生成的代码文件与相关文本文件类型\n code_extensions = {'.html', '.css', '.js', '.py', '.json', '.md', '.log'}\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，则说明该文件不是使用 UTF-8 编码的\n return False\n except Exception:\n # 忽略其他非编码相关的异常（如权限问题等）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n from pathlib import Path\n \n workspace = Path(workspace_path)\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for p in workspace.rglob('*'):\n rel_path = p.relative_to(workspace)\n \n # 忽略隐藏文件夹（如 .git 或 .vscode 等），防止对构建代码层级判断产生干扰\n if any(part.startswith('.') for part in rel_path.parts):\n continue\n \n # 仅针对文件夹计算其目录深度\n if p.is_dir():\n depth = len(rel_path.parts)\n # 若任意合法目录的层级大于2层，则判定不满足要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_var_name(name):\n has_digit = any(c.isdigit() for c in name)\n has_lower = any(c.islower() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_under = '_' in name\n return has_digit and has_lower and has_upper and has_under\n\n def clean_js_content(content):\n # 移除转义的字符串以防止误匹配\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content, flags=re.DOTALL)\n # 移除单行与多行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n return content\n\n def extract_js_vars(content):\n vars_to_check = set()\n \n # 匹配 `let`、`var`、`const` 声明的基础变量名\n for match in re.finditer(r'\\b(?:let|var|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', content):\n vars_to_check.add(match.group(1))\n \n # 匹配普通赋值操作（排除对象属性的赋值如 object.property = value）\n for match in re.finditer(r'(?<!\\.)\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=(?!=)', content):\n vars_to_check.add(match.group(1))\n \n # 匹配函数参数\n for match in re.finditer(r'\\bfunction\\s+[a-zA-Z0-9_$]*\\s*\\((.*?)\\)', content):\n params = match.group(1).split(',')\n for p in params:\n p = p.strip()\n if p and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p):\n vars_to_check.add(p)\n \n return vars_to_check\n\n html_script_pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n \n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n if ext in ['.py', '.js', '.html', '.css']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n vars_to_check = set()\n \n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n # 排除内置双下划线命名的变量，如 __name__\n if not (node.id.startswith('__') and node.id.endswith('__')):\n vars_to_check.add(node.id)\n elif isinstance(node, ast.arg):\n vars_to_check.add(node.arg)\n except Exception:\n pass\n \n elif ext == '.js':\n clean_content = clean_js_content(content)\n vars_to_check.update(extract_js_vars(clean_content))\n \n elif ext == '.html':\n for script_match in html_script_pattern.finditer(content):\n script_content = script_match.group(1)\n clean_content = clean_js_content(script_content)\n vars_to_check.update(extract_js_vars(clean_content))\n \n elif ext == '.css':\n # 匹配 CSS 原生变量\n for match in re.finditer(r'--([a-zA-Z0-9_-]+)\\s*:', content):\n vars_to_check.add(match.group(1))\n \n for var in vars_to_check:\n # 排除不小心抓取到的 JS 内置常量词汇\n if var in ['true', 'false', 'null', 'undefined', 'NaN', 'Infinity']:\n continue\n \n # 若发现任意一个提取出的变量名未同时具备要求中的四种元素，则判定失败\n if not check_var_name(var):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码文件扩展名集合（涵盖题目指令中提及的 HTML, CSS, JS 及其他常见代码类型）\n code_extensions = {\n '.html', '.htm',\n '.css', '.scss', '.sass', '.less',\n '.js', '.ts', '.jsx', '.tsx',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp',\n '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n \n # 如果工作区路径不存在，无法进行验证\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 拆分文件名主体和扩展名\n name, ext = os.path.splitext(file)\n \n # 我们仅对代码文件进行限制检查\n if ext.lower() in code_extensions:\n # 检查文件名主体（不含最后的扩展名部分）是否仅由小写字母和下划线组成\n # 若包含数字、大写字母或其他符号（如中划线 -、额外的点 . 等），均不合规并返回 False\n if not re.fullmatch(r'[a-z_]+', name):\n return False\n \n # 检查扩展名部分（去掉分隔点 `.` 后）是否也完全仅由小写字母和下划线组成\n # 该检测能够有效防止文件带有大写扩展名（例如 .JS, .Html 等）\n if not re.fullmatch(r'[a-z_]+', ext[1:]):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写的代码文件中，务必优先挑选反斜杠 \\ 来表示路径分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在 15 行及以内","validation_code":"import os\nimport re\nimport ast\n\ndef check(response: str, workspace_path: str) -> bool:\n \n def get_py_function_lengths(code: str):\n lengths = []\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\n lengths.append(node.end_lineno - node.lineno + 1)\n except Exception:\n pass\n return lengths\n\n def get_js_function_lengths(code: str):\n # 消除注释和字符串的干扰，但是保留原有的换行符，确保行数匹配不出错\n pattern = r'(/\\*.*?\\*/|//[^\\n]*|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n def replacer(match):\n text = match.group(0)\n return '\\n' * text.count('\\n') + ' ' * (len(text) - text.count('\\n'))\n \n cleaned_code = re.sub(pattern, replacer, code, flags=re.DOTALL)\n lengths = []\n \n # 常见的三种JS函数定义的正则表达式探测\n p1 = r'\\bfunction\\b[^\\{;}]*\\{'\n p2 = r'=>\\s*\\{'\n p3 = r'\\b(?!(?:if|for|while|switch|catch|with)\\b)\\w+\\s*\\([^)]*\\)\\s*\\{'\n func_pattern = f'(?:{p1}|{p2}|{p3})'\n\n # 找到所有可能是函数起始的 '{' 的索引\n starts = [m.end() - 1 for m in re.finditer(func_pattern, cleaned_code)]\n \n for start_idx in starts:\n if cleaned_code[start_idx] != '{':\n continue\n brace_level = 1\n end_idx = -1\n # 从 '{' 的下一位开始进行括号配对\n for j in range(start_idx + 1, len(cleaned_code)):\n if cleaned_code[j] == '{':\n brace_level += 1\n elif cleaned_code[j] == '}':\n brace_level -= 1\n if brace_level == 0:\n end_idx = j\n break\n # 成功配对到函数闭合的 '}'\n if end_idx != -1:\n start_line = cleaned_code.count('\\n', 0, start_idx)\n end_line = cleaned_code.count('\\n', 0, end_idx)\n lengths.append(end_line - start_line + 1)\n \n return lengths\n\n def extract_scripts_from_html(html_code: str):\n scripts = []\n for match in re.finditer(r'<script[^>]*>(.*?)</script>', html_code, flags=re.IGNORECASE | re.DOTALL):\n scripts.append(match.group(1))\n return scripts\n\n # ================= 1. 检查工作区生成的代码项目 =================\n for root, dirs, files in os.walk(workspace_path):\n if any(ignored in root for ignored in ['node_modules', '.git', 'venv']):\n continue\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n code = f.read()\n except Exception:\n continue\n\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\n lengths = get_js_function_lengths(code)\n if any(l > 15 for l in lengths):\n return False\n elif ext == '.py':\n lengths = get_py_function_lengths(code)\n if any(l > 15 for l in lengths):\n return False\n elif ext in ['.html', '.htm']:\n scripts = extract_scripts_from_html(code)\n for script in scripts:\n lengths = get_js_function_lengths(script)\n if any(l > 15 for l in lengths):\n return False\n\n # ================= 2. 检查助手最后一轮回复中的代码块 =================\n code_blocks = re.findall(r'```([a-zA-Z]*)\\n(.*?)```', response, flags=re.DOTALL)\n for lang, code in code_blocks:\n lang = lang.lower().strip()\n if lang in ('js', 'javascript', 'ts', 'typescript', ''):\n lengths = get_js_function_lengths(code)\n if any(l > 15 for l in lengths):\n return False\n elif lang in ('py', 'python'):\n lengths = get_py_function_lengths(code)\n if any(l > 15 for l in lengths):\n return False\n elif lang in ('html', 'htm'):\n scripts = extract_scripts_from_html(code)\n for script in scripts:\n lengths = get_js_function_lengths(script)\n if any(l > 15 for l in lengths):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能被视为“代码文件”的常见扩展名集合，排除.md, .json, .log等非代码文件\n code_extensions = {\n '.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', \n '.java', '.c', '.cpp', '.h', '.cs', '.go', '.php', '.sh'\n }\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n code_file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含目标编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果打开或读取发生异常（如编码非utf-8），则视作不满足要求\n return False\n\n # 若未找到任何代码文件，通常意味着未按要求生成项目代码，返回False；否则返回True\n return code_file_found","tags":["Content","Format"]},{"constraint":"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"后续所有轮次生成的每个代码文件，算上所有的空行的总行数必须保持为奇数","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 提取符合前序交互轮次预期的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 若文件完全为空，则行数为 0，0 是偶数，不符合奇数要求\n if not content:\n return False\n \n # 按照换行符切分以统计所有行（包括文件末尾可能由于回车产生的空行）\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果文件因编码等异常无法作为文本正常读取，可视为未能满足代码规范要求\n return False\n \n # 如果项目中存在代码文件且所有代码文件的总行数均为奇数，返回 True\n return has_code_file","tags":["Quantity","Exact Value"]}]},{"round_id":8,"instruction":"为了照顾不同水平的玩家，给控制面板增加一个“难度模式”的单选切换开关：分为“普通模式”和“困难模式”。普通模式下，游戏区中央的目标黑色阴影会以浅灰色线条勾勒出每个板块的内部辅助线；而在困难模式下，就是纯粹的黑色剪影。玩家在拼图过程中可以随时切换难度，目标图案的渲染需即时响应。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的生成代码时，请将各函数的行数严格控制在15行及以内的要求作废，更新为生成代码时，请将各函数的行数严格控制在 10 行及以内；第二，使用ESLint对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终ESLint的检查结果没有任何错误，该要求后续所有轮次均需严格遵循。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能的代码文件及其相关文本文件的扩展名\n target_extensions = {'.html', '.css', '.js', '.md', '.json', '.log'}\n\n # 如果工作区路径不存在，说明没有生成任何文件，不违反UTF-8编码要求\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历工作区所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出符合目标后缀的文件\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出异常，说明文件不是有效的 UTF-8 编码，不符合要求\n return False\n except Exception:\n # 如果有其他读取错误（如权限问题等）暂不作为编码错误处理\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n workspace = Path(workspace_path).resolve()\n \n # 检查工作空间路径是否存在且为目录\n if not workspace.exists() or not workspace.is_dir():\n return False\n \n for root, dirs, files in os.walk(workspace):\n # 忽略隐藏目录和常见的 Python 缓存目录，避免不必要的误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n rel_path = Path(root).resolve().relative_to(workspace)\n \n # 将相对路径分解为各级组件\n parts = [p for p in rel_path.parts if p and p != '.']\n \n # 若存在相对路径层级 >= 2 的目录（即子目录下的子目录），\n # 则说明总目录层级 >= 3，不满足小于等于2层的要求\n if len(parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"import os\nimport re\n\ndef check_var_name(name: str) -> bool:\n has_digit = any(c.isdigit() for c in name)\n has_upper = any(c.isupper() for c in name)\n has_lower = any(c.islower() for c in name)\n has_under = '_' in name\n return has_digit and has_upper and has_lower and has_under\n\ndef extract_variables(text: str) -> list:\n vars_set = set()\n \n # 1. 提取标准 let, const, var\n matches = re.findall(r'\\b(?:let|const|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', text)\n vars_set.update(matches)\n \n # 2. 提取解构赋值中的变量\n destruct_matches = re.findall(r'\\b(?:let|const|var)\\s*([\\[\\{][^\\]\\}]+[\\]\\}])', text)\n for d_str in destruct_matches:\n inner = d_str[1:-1]\n for p in inner.split(','):\n p = p.strip()\n if not p: continue\n if ':' in p: p = p.split(':')[1].strip()\n if '=' in p: p = p.split('=')[0].strip()\n m = re.match(r'^[\\[\\{]?\\s*(?:\\.\\.\\.)?\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\n if m: vars_set.add(m.group(1))\n\n # 3. 提取函数名和类名\n funcs = re.findall(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', text)\n vars_set.update(funcs)\n classes = re.findall(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', text)\n vars_set.update(classes)\n \n # 4. 提取常规函数参数\n func_params = re.findall(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', text)\n for param_str in func_params:\n for p in param_str.split(','):\n p = p.strip()\n if not p: continue\n if ':' in p: p = p.split(':')[1].strip()\n if '=' in p: p = p.split('=')[0].strip()\n m = re.match(r'^[\\[\\{]?\\s*(?:\\.\\.\\.)?\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\n if m: vars_set.add(m.group(1))\n \n # 5. 提取带括号的箭头函数参数\n arrow_params = re.findall(r'\\(([^)]*)\\)\\s*=>', text)\n for param_str in arrow_params:\n for p in param_str.split(','):\n p = p.strip()\n if not p: continue\n if ':' in p: p = p.split(':')[1].strip()\n if '=' in p: p = p.split('=')[0].strip()\n m = re.match(r'^[\\[\\{]?\\s*(?:\\.\\.\\.)?\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\n if m: vars_set.add(m.group(1))\n \n # 6. 提取单参数箭头函数的参数\n single_arrows = re.findall(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', text)\n vars_set.update(single_arrows)\n \n # 7. 提取 catch 块中的错误对象\n catch_blocks = re.findall(r'\\bcatch\\s*\\(\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\)', text)\n vars_set.update(catch_blocks)\n \n return list(vars_set)\n\n\ndef verify(response: str, workspace_path: str) -> bool:\n all_vars = set()\n has_files = False\n \n # 预编译正则，用于清除多行注释、单行注释及各种字符串\n clean_pattern = re.compile(\n r'/\\*.*?\\*/' \n r'|//[^\\n]*' \n r'|\"(?:\\\\.|[^\"\\\\])*\"' \n r'|\\'(?:\\\\.|[^\\'\\\\])*\\'' \n r'|`(?:\\\\.|[^`\\\\])*`', \n re.DOTALL\n )\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html') or file.endswith('.css'):\n has_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file.endswith('.html'):\n # 提取并解析 JS 代码块\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL)\n for script in scripts:\n script = clean_pattern.sub('', script)\n all_vars.update(extract_variables(script))\n \n # 提取并解析 CSS 代码块中的变量\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, re.IGNORECASE | re.DOTALL)\n for style in styles:\n style = clean_pattern.sub('', style)\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', style)\n all_vars.update(css_vars)\n \n elif file.endswith('.js'):\n content = clean_pattern.sub('', content)\n all_vars.update(extract_variables(content))\n \n elif file.endswith('.css'):\n content = clean_pattern.sub('', content)\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', content)\n all_vars.update(css_vars)\n \n except Exception:\n continue\n \n # 如果项目中没有任何代码文件，或者没有声明任何变量，则未满足约束。\n if not has_files or not all_vars:\n return False\n \n # 检查所有被提取出的变量名\n for var in all_vars:\n if not check_var_name(var):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义属于“代码文件”的常规扩展名集合（排除 .md, .json, .log 等非代码文件）\n code_exts = {\n '.js', '.mjs', '.cjs', '.ts', '.tsx', '.jsx', \n '.html', '.htm', '.css', '.scss', '.less', '.sass',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 提取文件的扩展名\n _, ext = os.path.splitext(file_name)\n \n # 判断是否为代码文件\n if ext.lower() in code_exts:\n found_code_file = True\n \n # 去除扩展名的 \".\" 后，对整个文件名进行检查\n # 这保证了即使扩展名中含有大写（如 .JS）或主文件命中含有数字（如 index1），也会被成功拦截\n name_check = file_name.replace('.', '')\n \n # 检查是否仅由小写字母和下划线构成\n if not re.match(r'^[a-z_]+$', name_check):\n return False\n \n return found_code_file","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写的代码文件中，务必优先挑选反斜杠 \\ 来表示路径分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在 10 行及以内","validation_code":"import os\nimport re\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n def remove_comments_and_strings(code):\n state = 'NORMAL'\n result = []\n i = 0\n while i < len(code):\n if state == 'NORMAL':\n if code[i:i+2] == '//':\n state = 'LINE_COMMENT'\n result.append(' ')\n i += 2\n elif code[i:i+2] == '/*':\n state = 'BLOCK_COMMENT'\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n state = 'DQUOTE'\n result.append(' ')\n i += 1\n elif code[i] == \"'\":\n state = 'SQUOTE'\n result.append(' ')\n i += 1\n elif code[i] == '`':\n state = 'BQUOTE'\n result.append(' ')\n i += 1\n else:\n result.append(code[i])\n i += 1\n elif state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state = 'NORMAL'\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'BLOCK_COMMENT':\n if code[i:i+2] == '*/':\n state = 'NORMAL'\n result.append(' ')\n i += 2\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'DQUOTE':\n if code[i] == '\\\\':\n result.append(' ')\n if i + 1 < len(code):\n if code[i+1] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n state = 'NORMAL'\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'SQUOTE':\n if code[i] == '\\\\':\n result.append(' ')\n if i + 1 < len(code):\n if code[i+1] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 2\n elif code[i] == \"'\":\n state = 'NORMAL'\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'BQUOTE':\n if code[i] == '\\\\':\n result.append(' ')\n if i + 1 < len(code):\n if code[i+1] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 2\n elif code[i] == '`':\n state = 'NORMAL'\n result.append(' ')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n\n def check_script_functions(code_str):\n clean_code = remove_comments_and_strings(code_str)\n lines = clean_code.split('\\n')\n \n i = 0\n while i < len(lines):\n line = lines[i]\n \n is_function = False\n # 匹配常规 function\n if re.search(r'\\bfunction\\b', line):\n is_function = True\n # 匹配箭头函数\n elif re.search(r'=>\\s*\\{', line):\n is_function = True\n # 匹配类方法或对象方法 (忽略控制流关键字)\n elif re.search(r'^\\s*(?:async\\s+)?(?!(?:if|for|while|switch|catch|return)\\b)[a-zA-Z_$][\\w$]*\\s*\\([^)]*\\)\\s*\\{', line):\n is_function = True\n \n if is_function:\n brace_count = 0\n started = False\n start_line = i\n end_line = i\n \n for j in range(i, len(lines)):\n for char in lines[j]:\n if char == '{':\n brace_count += 1\n started = True\n elif char == '}':\n if started:\n brace_count -= 1\n if brace_count <= 0:\n break\n if started and brace_count <= 0:\n end_line = j\n break\n \n # 如果成功匹配闭合大括号，校验函数行数\n if started and brace_count == 0:\n func_length = end_line - start_line + 1\n if func_length > 10:\n return False\n i += 1\n return True\n\n def check_python_functions(code_str):\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno'):\n length = node.end_lineno - node.lineno + 1\n if length > 10:\n return False\n except Exception:\n pass\n return True\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_script_functions(content):\n return False\n except Exception:\n pass\n \n elif file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_python_functions(content):\n return False\n except Exception:\n pass\n \n elif file.endswith('.html') or file.endswith('.htm'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 提取内联 script 脚本进行验证\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_script_functions(script):\n return False\n except Exception:\n pass\n\n return True","tags":["Quantity","Range"]},{"constraint":"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n\n code_extensions = ('.html', '.css', '.js')\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅推断 HTML、CSS、JS 作为必须检查的代码文件类型\n if file.endswith(code_extensions):\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 指令中要求使用 UTF-8 编码，因此以该编码读取\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果无法读取文件（例如非法字符或权限错误），则视为不满足\n return False\n\n return found_code_file","tags":["Content","Format"]},{"constraint":"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"后续所有轮次生成的每个代码文件，算上所有的空行的总行数必须保持为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对前端七巧板项目推测代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 按换行符进行分割，可以准确统计出算上空行在内的文本总行数\n lines = content.split('\\n')\n \n # 判断总行数是否为偶数\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果遇到文件读取异常（例如不可读或非规范的UTF-8），保守处理返回False\n return False\n \n # 如果至少存在一个代码文件且遍历结束无任何文件违背奇数行数原则，返回 True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"使用ESLint对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终ESLint的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":9,"instruction":"最后一轮，我们来提升UI的质感和物理拟真视觉体验。请用CSS渐变或SVG滤镜给这7块七巧板加上逼真的木纹纹理，并为它们添加投影（box-shadow/drop-shadow）。当板块被鼠标按住拖拽时，投影变得更模糊且偏移变大，模拟被从桌面上“提起来”的3D视觉效果；松开放下时，投影恢复紧贴桌面的状态。让整体游戏看起来像一个高级的实体桌面玩具。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的要求生成的每个代码文件算上所有空行总行数必须为奇数的规则作废，更新为每个生成的代码文件总行数（含空行）必须是偶数，该要求需严格遵循；第二，在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 25s` 包装，这样可以防范进程挂起，该要求后续所有轮次均需严格遵循。","instruction_following_checklist":[{"constraint":"你提供的每个代码文件都严格要求使用UTF-8作为编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推测代码文件及相关文本文件可能存在的后缀名\n valid_extensions = {'.html', '.css', '.js', '.json', '.md', '.log', '.sh'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否是推断的代码或文本文件类型，并额外适配可能的 ESLint 配置文件\n if ext in valid_extensions or file == '.eslintrc':\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用 UTF-8 解码二进制内容\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 处理诸如文件无法读取等其他异常（非要求重点，为防止执行崩溃均算作校验失败）\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你构建的代码工程目录层级必须小于或等于2层","validation_code":"def check_directory_level(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作空间不存在，自然无法满足构建代码工程的前提\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的隐藏目录（如 .git, .vscode 等），防止对判定造成干扰\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n rel_path = os.path.relpath(root, workspace_path)\n # 如果是根目录本身，跳过检查\n if rel_path == '.':\n continue\n \n # 统一不同操作系统下的路径分隔符\n rel_path = rel_path.replace('\\\\', '/')\n \n # 将路径以 '/' 切分并过滤掉空字符串，从而获取相对目录层级深度\n parts = [p for p in rel_path.split('/') if p]\n \n # 如果包含的层数大于 2，说明目录层级超出了要求\n if len(parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_var_name(name):\n has_digit = bool(re.search(r'\\d', name))\n has_upper = bool(re.search(r'[A-Z]', name))\n has_lower = bool(re.search(r'[a-z]', name))\n has_underscore = '_' in name\n return has_digit and has_upper and has_lower and has_underscore\n\n def extract_js_vars_regex(content):\n # 移除字符串（处理包含转义字符的情况，防止字符串内部文本扰乱正则匹配）\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content, flags=re.DOTALL)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content, flags=re.DOTALL)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content, flags=re.DOTALL)\n \n # 移除单行及多行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n vars_found = set()\n \n # 1. 提取 let, const, var 声明的变量（状态机匹配以应对多变量声明等复杂情况）\n for match in re.finditer(r'\\b(let|const|var)\\b', content):\n pos = match.end()\n state = 'expect_var'\n current_var = ''\n while pos < len(content):\n char = content[pos]\n if state == 'expect_var':\n if char.isspace():\n pass\n elif char.isalpha() or char in '_$':\n current_var = char\n state = 'in_var'\n elif char in '{[':\n pass\n else:\n break\n elif state == 'in_var':\n if char.isalnum() or char in '_$':\n current_var += char\n else:\n vars_found.add(current_var)\n current_var = ''\n if char.isspace():\n state = 'after_var'\n elif char == '=':\n state = 'in_value'\n elif char == ',':\n state = 'expect_var'\n elif char in '}]':\n state = 'after_var'\n elif char == ';':\n break\n else:\n break\n elif state == 'after_var':\n if char.isspace():\n pass\n elif char == '=':\n state = 'in_value'\n elif char == ',':\n state = 'expect_var'\n elif char in '}]':\n pass\n elif char == ';':\n break\n else:\n break\n elif state == 'in_value':\n if char == ';':\n break\n pos += 1\n\n # 2. 提取函数声明中的函数名及参数\n func_pattern = re.compile(r'\\bfunction\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)')\n for match in func_pattern.finditer(content):\n if match.group(1):\n vars_found.add(match.group(1))\n args = match.group(2).split(',')\n for arg in args:\n arg = arg.strip().split('=')[0].strip()\n if arg:\n words = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg)\n if words:\n vars_found.add(words[0])\n \n # 3. 提取箭头函数参数\n arrow_pattern = re.compile(r'(\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>')\n for match in arrow_pattern.finditer(content):\n args_str = match.group(1)\n if args_str.startswith('('):\n args_str = args_str[1:-1]\n args = args_str.split(',')\n for arg in args:\n arg = arg.strip().split('=')[0].strip()\n if arg:\n words = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg)\n if words:\n vars_found.add(words[0])\n \n # 4. 提取 catch 捕获的异常变量名\n catch_pattern = re.compile(r'\\bcatch\\s*\\(\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\)')\n for match in catch_pattern.finditer(content):\n vars_found.add(match.group(1))\n \n # 5. 提取 class 类名\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n for match in class_pattern.finditer(content):\n vars_found.add(match.group(1))\n\n return list(vars_found)\n\n def extract_css_vars(content):\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n vars_found = set()\n for match in re.finditer(r'--([a-zA-Z0-9_-]+)', content):\n vars_found.add(match.group(0))\n return list(vars_found)\n\n found_any = False\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 跳过历史备份文件\n if 'backup' in file.lower() or file.endswith('.bak'):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if file.endswith('.js'):\n vars_found = extract_js_vars_regex(content)\n for v in vars_found:\n found_any = True\n if not check_var_name(v):\n return False\n \n elif file.endswith('.py'):\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n found_any = True\n if not check_var_name(node.id):\n return False\n elif isinstance(node, ast.arg):\n found_any = True\n if not check_var_name(node.arg):\n return False\n except Exception:\n pass\n \n elif file.endswith('.html'):\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\n for script in scripts:\n vars_found = extract_js_vars_regex(script)\n for v in vars_found:\n found_any = True\n if not check_var_name(v):\n return False\n \n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.IGNORECASE | re.DOTALL)\n for style in styles:\n vars_found = extract_css_vars(style)\n for v in vars_found:\n found_any = True\n if not check_var_name(v):\n return False\n \n elif file.endswith('.css'):\n vars_found = extract_css_vars(content)\n for v in vars_found:\n found_any = True\n if not check_var_name(v):\n return False\n\n # 若未找到任何变量定义视作暂未生成代码主体，或者不满足条件\n return True if found_any else False","tags":["Style","Naming"]},{"constraint":"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项","validation_code":"","tags":["Environment","Logging"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在3以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义属于“代码文件”的常见扩展名集合（针对前端七巧板项目以及其他常见开发语言）\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.mjs', '.cjs', '.ts', '.jsx', '.tsx', \n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', \n '.cs', '.go', '.rs', '.rb', '.php', '.swift', '.kt', '.m', '.sh'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）和常见的包管理器依赖目录（如 node_modules）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for file in files:\n # 忽略隐藏文件\n if file.startswith('.'):\n continue\n \n # 获取文件扩展名并转为小写进行匹配\n ext = os.path.splitext(file)[1].lower()\n \n # 判断该文件是否属于代码文件\n if ext in code_extensions:\n # 剔除文件名中的点号 (.) \n name_without_dots = file.replace('.', '')\n \n # 验证剩余部分是否完全由小写字母和下划线组成\n # 如果包含数字、大写字母或其他符号，均会匹配失败并返回 False\n if not re.fullmatch(r'[a-z_]+', name_without_dots):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写的代码文件中，务必优先挑选反斜杠 \\ 来表示路径分隔","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\"pass\"、\"fail\"或\"error\"这三者之一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在 10 行及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_python_functions(code_str: str) -> bool:\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 判断当前环境是否支持 end_lineno (Python 3.8+)\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n line_count = node.end_lineno - node.lineno + 1\n if line_count > 10:\n return False\n return True\n except Exception:\n return True\n\n def clean_js_code(code_str: str) -> str:\n \"\"\"\n 清空 JS 代码里的注释和字符串，但是严格保留换行符。\n 这样可以在没有干扰的前提下正确匹配括号，并依靠换行符准确统计行号。\n \"\"\"\n result = []\n i = 0\n length = len(code_str)\n in_block_comment = False\n in_line_comment = False\n in_string = False\n string_char = ''\n \n while i < length:\n c = code_str[i]\n if in_block_comment:\n if c == '*' and i + 1 < length and code_str[i+1] == '/':\n in_block_comment = False\n result.append(' ')\n i += 2\n continue\n else:\n result.append('\\n' if c == '\\n' else ' ')\n i += 1\n continue\n elif in_line_comment:\n if c == '\\n':\n in_line_comment = False\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n continue\n elif in_string:\n if c == '\\\\':\n result.append(' ')\n i += 2\n continue\n elif c == string_char:\n in_string = False\n result.append(' ')\n else:\n result.append('\\n' if c == '\\n' else ' ')\n i += 1\n continue\n \n if c == '/' and i + 1 < length and code_str[i+1] == '*':\n in_block_comment = True\n result.append(' ')\n i += 2\n continue\n elif c == '/' and i + 1 < length and code_str[i+1] == '/':\n in_line_comment = True\n result.append(' ')\n i += 2\n continue\n elif c in (\"'\", '\"', '`'):\n in_string = True\n string_char = c\n result.append(' ')\n i += 1\n continue\n \n result.append(c)\n i += 1\n \n return \"\".join(result)\n\n def check_js_functions(code_str: str) -> bool:\n clean_code = clean_js_code(code_str)\n \n brace_pairs = {}\n stack = []\n # 提取全部 { } 匹配对\n for i, c in enumerate(clean_code):\n if c == '{':\n stack.append(i)\n elif c == '}':\n if stack:\n start = stack.pop()\n brace_pairs[start] = i\n \n for start_idx, end_idx in brace_pairs.items():\n # 往回找语句终止符\n search_start = start_idx - 1\n while search_start >= 0 and clean_code[search_start] not in \"{};\":\n search_start -= 1\n \n prefix = clean_code[search_start + 1 : start_idx].strip()\n \n # 如果包含控制流等非函数的关键字，直接跳过\n if re.search(r'\\b(?:if|for|while|switch|catch|class|try|else)\\b', prefix):\n continue\n \n is_func = False\n # 常见函数关键字\n if 'function' in prefix or '=>' in prefix:\n is_func = True\n # 对象/类内部方法声明形式检测（如: `methodName(args) {`）\n elif re.search(r'\\b[\\w$]+\\s*\\([^)]*\\)$', prefix):\n is_func = True\n \n if is_func:\n abs_start_idx = start_idx\n \n feature_idx_func = prefix.rfind('function')\n feature_idx_arrow = prefix.rfind('=>')\n feature_idx_paren = prefix.rfind(')')\n \n # 寻找更精确的声明起点，防止漏算或将上一行代码错算入\n if feature_idx_func != -1:\n abs_start_idx = search_start + 1 + feature_idx_func\n \n if feature_idx_paren != -1:\n abs_paren_idx = search_start + 1 + feature_idx_paren\n paren_depth = 0\n # 匹配函数参数对应的 \"(\"，以将其设为起点的基准\n for j in range(abs_paren_idx, search_start, -1):\n if clean_code[j] == ')':\n paren_depth += 1\n elif clean_code[j] == '(':\n paren_depth -= 1\n if paren_depth == 0:\n current_abs = j\n if feature_idx_func != -1 and (search_start + 1 + feature_idx_func) < j:\n current_abs = search_start + 1 + feature_idx_func\n if current_abs < abs_start_idx:\n abs_start_idx = current_abs\n break\n else:\n if (search_start + 1 + feature_idx_paren) < abs_start_idx:\n abs_start_idx = search_start + 1 + feature_idx_paren\n \n if feature_idx_arrow != -1 and feature_idx_paren == -1:\n if (search_start + 1 + feature_idx_arrow) < abs_start_idx:\n abs_start_idx = search_start + 1 + feature_idx_arrow\n \n start_lineno = clean_code.count('\\n', 0, abs_start_idx) + 1\n end_lineno = clean_code.count('\\n', 0, end_idx) + 1\n \n if end_lineno - start_lineno + 1 > 10:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 校验 Python 文件\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_str = f.read()\n if not check_python_functions(code_str):\n return False\n except Exception:\n pass\n # 校验 JavaScript / TypeScript 文件\n elif file.endswith(('.js', '.jsx', '.ts', '.tsx')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_str = f.read()\n if not check_js_functions(code_str):\n return False\n except Exception:\n pass\n # 校验 HTML 文件中被包裹在 <script> 中的 JS 代码\n elif file.endswith(('.html', '.htm')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n scripts = re.findall(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for script in scripts:\n if not check_js_functions(script):\n return False\n except Exception:\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码文件扩展名，排除文档、配置和日志等非代码文件\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法读取或其他异常情况，视作未满足要求\n return False\n \n # 如果遍历完毕没有发现任何代码文件，说明未生成有效的代码工程\n return has_code_file","tags":["Content","Format"]},{"constraint":"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每个生成的代码文件总行数（含空行）必须是偶数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名，根据上下文主要包括前端等代码类型\n code_extensions = {\n '.js', '.css', '.html', '.py', '.ts', '.jsx', '.tsx',\n '.vue', '.java', '.c', '.cpp', '.h', '.cs', '.go', \n '.rs', '.rb', '.php', '.swift', '.kt'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照 '\\n' 分割可以准确计算出编辑器中的物理行数（包括末尾独立空行）\n lines_count = len(content.split('\\n'))\n \n # 若存在奇数行数的文件，直接判定不满足要求\n if lines_count % 2 != 0:\n return False\n except Exception:\n # 如果由于某种原因（如非预期的二进制混入）无法用 UTF-8 读取则容错跳过\n pass\n \n # 确保存在至少一个代码文件并且所有的代码文件行数皆为偶数\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"使用ESLint对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终ESLint的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 25s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用HTML、CSS和JS做一个七巧板益智游戏的雏形。先搭建基础界面的骨架：左侧是一个大的游戏操作区（带明显边界的画布或容器），右侧是控制面板（包含一个标题‘七巧板拼图’和几个占位按钮）。在游戏区中间画一个大正方形，代表未拆解的完整七巧板，现在不需要拆解它，也不需要复杂的交互，先把UI框架和基本的页面响应式布局搭好。请你在后续所有轮次的开发中都严格遵循两个要求：无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用；生成的代码文件中应优先以正斜杠 / 作为路径的分割符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n\\n # 遍历当前项目的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果是代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制形式读取文件内容\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试用 utf-8 进行解码，如果不抛出 UnicodeDecodeError 则说明是合法的 utf-8 编码格式\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码失败，说明文件不是 UTF-8 编码\\n return False\\n except Exception:\\n # 遇到其他读取异常无法判断时，视为未满足要求\\n return False\\n\\n # 如果至少存在一个代码文件，且全部验证通过，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在或者不是一个目录，则无法进行判断\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 计算当前遍历到的目录相对于根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 根目录本身的相对路径通常为 '.'，代表第1层\\n if rel_path == '.':\\n depth = 1\\n else:\\n # 统一路径分隔符为 '/' 以兼容不同操作系统，通过分割路径来计算目录深度\\n # 例如相对路径 'src' 长度为1，表示为根目录下的直接子目录，总层级为第2层\\n # 例如相对路径 'src/components' 长度为2，总层级为第3层\\n normalized_rel_path = rel_path.replace('\\\\\\\\', '/')\\n depth = len(normalized_rel_path.split('/')) + 1\\n \\n # 若某目录所处深度大于2层，说明不符合要求，直接返回 False\\n if depth > 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 校验变量名是否同时包含：数字、下划线、大写字母、小写字母\\n def check_var_name(name: str) -> bool:\\n has_digit = any(c.isdigit() for c in name)\\n has_underscore = '_' in name\\n has_upper = any(c.isupper() for c in name)\\n has_lower = any(c.islower() for c in name)\\n return has_digit and has_underscore and has_upper and has_lower\\n\\n variables = set()\\n\\n # 从 JavaScript 文本中提取可能的变量标识符\\n def extract_from_js(js_code: str):\\n # 去除单行和多行注释\\n js_code = re.sub(r'//.*', '', js_code)\\n js_code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', js_code)\\n # 去除字符串字面量（避免匹配到字符串内部的内容）\\n js_code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', js_code)\\n js_code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", js_code)\\n js_code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', js_code)\\n \\n # 1. 提取 var, let, const, function, class 紧接的标识符\\n decl_pattern = r'\\\\b(?:var|let|const|function|class)\\\\s+([a-zA-Z_]\\\\w*)'\\n for match in re.finditer(decl_pattern, js_code):\\n variables.add(match.group(1))\\n \\n # 2. 提取普通函数参数：function [name](param1, param2)\\n func_pattern = r'\\\\bfunction\\\\s*(?:[a-zA-Z_]\\\\w*)?\\\\s*\\\\(([^)]*)\\\\)'\\n for match in re.finditer(func_pattern, js_code):\\n params_str = match.group(1)\\n for p in params_str.split(','):\\n p = p.split('=')[0].strip() # 剔除默认值\\n if p and re.match(r'^[a-zA-Z_]\\\\w*$', p):\\n variables.add(p)\\n \\n # 3. 提取括号箭头函数参数：(param1, param2) =>\\n arrow_pattern = r'\\\\(([^)]*)\\\\)\\\\s*=>'\\n for match in re.finditer(arrow_pattern, js_code):\\n params_str = match.group(1)\\n for p in params_str.split(','):\\n p = p.split('=')[0].strip()\\n if p and re.match(r'^[a-zA-Z_]\\\\w*$', p):\\n variables.add(p)\\n \\n # 4. 提取单参数箭头函数参数：param =>\\n single_arrow_pattern = r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*=>'\\n for match in re.finditer(single_arrow_pattern, js_code):\\n variables.add(match.group(1))\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目目录收集代码文件中的所有变量\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n extract_from_js(f.read())\\n except Exception:\\n pass\\n \\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 提取 HTML 内联 JS\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for script in scripts:\\n extract_from_js(script)\\n \\n # 提取 HTML 内联 CSS 变量\\n styles = re.findall(r'<style.*?>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n for style in styles:\\n style = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', style)\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', style)\\n for css_var in css_vars:\\n variables.add(css_var)\\n except Exception:\\n pass\\n \\n elif file.endswith('.css'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n # 提取 CSS 变量名\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', content)\\n for css_var in css_vars:\\n variables.add(css_var)\\n except Exception:\\n pass\\n\\n # 排除意外被正则捕获的常见JS语言关键字\\n keywords = {'if', 'else', 'for', 'while', 'return', 'switch', 'case', 'catch', 'true', 'false', 'null', 'undefined'}\\n variables = {v for v in variables if v not in keywords}\\n\\n if not variables:\\n return True\\n\\n # 遍历所有捕捉到的变量，验证要求\\n for var in variables:\\n if not check_var_name(var):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断可能存在的代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n name, ext = os.path.splitext(file)\\n \\n # 判断该文件是否属于我们要检查的代码文件\\n if ext.lower() in code_extensions:\\n \\n # 检查1：文件名主体只能包含小写字母和下划线。\\n # 不能包含大写字母、数字，以及任何符号（如果存在多个点如 \\\"script.min.js\\\"，name 会是 \\\"script.min\\\"，也会被判定出含符号并返回False）\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n # 检查2：因为要求不能出现大写字母，所以扩展名也必须是全小写\\n # （如 “.HTML” 虽属于上述文件类型，但由于包含大写字母而不被允许）\\n if ext != ext.lower():\\n return False\\n \\n # 所有的代码文件均符合要求，返回 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中应优先以正斜杠 / 作为路径的分割符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来我们需要把那个大正方形拆解成真正的七巧板。请在代码中定义标准的七巧板的7个基本图形（2个大直角三角形、1个中直角三角形、2个小直角三角形、1个正方形、1个平行四边形），并给它们分别填充7种不同的鲜明颜色。初始状态下，不要拼在一起，而是将这7个图形散落在游戏区底部或者控制面板下方的‘备选区’中。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，在更新任何已有文件之前，必须在同一级别目录下生成一份该文件被更改前的备份；第二，每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 依据用户指令推测产生的代码和文本文件后缀类型\\n target_extensions = {'.html', '.css', '.js', '.md', '.json'}\\n \\n # 若工作区不存在，则无法验证，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检测相关代码/文本文件\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果触发解码错误，说明存在文件不是使用 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能导致读取失败的文件操作级别错误\\n pass\\n \\n # 所有文件均能以 UTF-8 正确读取，说明符合编码要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n\\n base = Path(workspace_path).resolve()\\n \\n # 如果工作区目录不存在，可以视为没有创建深层目录，满足不大于2层的限制\\n if not base.exists() or not base.is_dir():\\n return True\\n \\n for item in base.rglob('*'):\\n rel = item.relative_to(base)\\n \\n # 忽略以 '.' 开头的隐藏文件或文件夹，如 .git 等，避免误判\\n if any(part.startswith('.') for part in rel.parts):\\n continue\\n \\n if item.is_dir():\\n # base作为第1层，直接子文件夹为第2层（len == 1）\\n # 如果目录的相对层级 > 1，意味着出现了孙子目录（第3层或以上），违反规则\\n if len(rel.parts) > 1:\\n return False\\n \\n elif item.is_file():\\n # 直接在base下的文件（len == 1），在第2层子目录下的文件（len == 2）\\n # 如果文件的相对层级 > 2，意味着其处在超过两层的目录中，违反规则\\n if len(rel.parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 捕获 var, let, const, function, class 后直接跟随的常规标识符作为自定义变量名\\n var_pattern = re.compile(r'\\\\b(?:var|let|const|function|class)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 根据用户指令推断主要代码文件为 html 和 js 格式\\n if ext in ['.js', '.html']:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 1. 移除多行注释和单行注释\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n content = re.sub(r'//.*', '', content)\\n \\n # 2. 移除 HTML 注释\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n \\n # 3. 移除字符串内部的内容，防止将字符串内的话误认为代码变量声明\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n variables = var_pattern.findall(content)\\n for var in variables:\\n # 检查要求：同时包含数字、下划线、大写字母和小写字母\\n has_num = bool(re.search(r'[0-9]', var))\\n has_under = bool(re.search(r'_', var))\\n has_upper = bool(re.search(r'[A-Z]', var))\\n has_lower = bool(re.search(r'[a-z]', var))\\n \\n if not (has_num and has_under and has_upper and has_lower):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 明确代码文件的扩展名类型（基于用户多轮指令中对前端项目的要求）\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历当前项目空间中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件的后缀名\\n _, ext = os.path.splitext(file)\\n \\n # 仅对代码文件进行检查\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n \\n # 检查文件名是否仅包含小写字母、下划线以及唯一的一个点(作为扩展名的分隔)\\n # ^[a-z_]+ : 匹配文件主名，仅限小写字母和下划线\\n # \\\\. : 匹配必须存在的一个点（扩展名分隔符）\\n # [a-z_]+$ : 匹配扩展名部分，同样仅限小写字母和下划线\\n if not re.match(r'^[a-z_]+\\\\.[a-z_]+$', file):\\n return False\\n \\n # 如果找到了代码文件且均通过检查，则返回 True；否则返回 False\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中应优先以正斜杠 / 作为路径的分割符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在更新任何已有文件之前，必须在同一级别目录下生成一份该文件被更改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来增加核心交互：自由拖拽功能。请实现使用鼠标或触摸屏能够选中并自由拖动这7个板块的功能。要求：拖动过程中，当前被拖动的板块层级（z-index）必须处于最顶层；另外要限制拖动边界，板块不能被完全拖出游戏操作区之外；松开鼠标时板块能停留在放下的位置。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令；第二，原有的修改文件前需在同级目录创建备份的要求作废，更新为任何时候修改现有文件前，都不得在同级目录下创建该文件原始状态的备份。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码和文本文件的扩展名集合\\n # 涵盖了用户指令中提到的 html, css, js, json, md，以及其他常规代码类型\\n code_extensions = {\\n '.html', '.css', '.js', '.json', '.md', \\n '.py', '.java', '.c', '.cpp', '.h', \\n '.ts', '.jsx', '.tsx', '.txt', '.xml', \\n '.sh', '.bat'\\n }\\n \\n # 如果工作区路径不存在，则说明无任何代码文件生成，尚未违反编码限制\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试将文件内容的字节流以utf-8进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦抛出解码错误，说明存在文件不是UTF-8编码，要求不满足\\n return False\\n except Exception:\\n # 忽略其他可能的文件读取权限等非编码相关的系统异常\\n pass\\n \\n # 所有代码文件全部成功用 utf-8 验证通过，或未发现任何代码文件\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n rel_path = os.path.relpath(root, workspace_path)\\n if rel_path == '.':\\n continue\\n \\n # 统一使用正斜杠处理相对路径，兼容不同操作系统\\n rel_path = rel_path.replace('\\\\\\\\', '/')\\n parts = [p for p in rel_path.split('/') if p]\\n \\n # 忽略以 \\\".\\\" 或 \\\"__\\\" 开头的系统、缓存及版本控制特殊目录\\n if any(p.startswith('.') or p.startswith('__') for p in parts):\\n continue\\n \\n # 根目录视为第1层，一级子目录（parts长度为1）视为第2层\\n # 若存在 parts 长度大于等于 2 的目录，说明出现了第3层或更深层级的目录，违反条件\\n if len(parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断变量名是否满足条件\\n def check_var_name(name: str) -> bool:\\n if not name:\\n return True\\n has_digit = any(c.isdigit() for c in name)\\n has_lower = any(c.islower() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_under = '_' in name\\n return has_digit and has_lower and has_upper and has_under\\n\\n # 辅助函数：检查 JavaScript / TypeScript 代码中的变量名\\n def check_js_code(code: str) -> bool:\\n # 去除单行和多行注释\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n \\n # 移除字符串字面量（避免字符串内部存在疑似关键字的内容干扰匹配）\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', code, flags=re.DOTALL)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', code, flags=re.DOTALL)\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', code, flags=re.DOTALL)\\n\\n # 1. 匹配 let, var, const 后声明的变量\\n matches = re.finditer(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', code)\\n for m in matches:\\n if not check_var_name(m.group(1)):\\n return False\\n \\n # 2. 匹配函数声明的名称\\n matches = re.finditer(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', code)\\n for m in matches:\\n if not check_var_name(m.group(1)):\\n return False\\n \\n # 3. 匹配普通函数的参数名\\n matches = re.finditer(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', code)\\n for m in matches:\\n params_str = m.group(1)\\n if params_str.strip():\\n for param in params_str.split(','):\\n param = param.split('=')[0].strip()\\n # 粗略跳过解构赋值的复杂参数，主要校验常规参数\\n if param and '{' not in param and '[' not in param:\\n if not check_var_name(param):\\n return False\\n \\n # 4. 匹配类的名称\\n matches = re.finditer(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)', code)\\n for m in matches:\\n if not check_var_name(m.group(1)):\\n return False\\n \\n # 5. 匹配箭头函数的参数名\\n matches = re.finditer(r'(?:\\\\(([^)]*)\\\\)|([a-zA-Z_$][a-zA-Z0-9_$]*))\\\\s*=>', code)\\n for m in matches:\\n if m.group(1):\\n params_str = m.group(1)\\n for param in params_str.split(','):\\n param = param.split('=')[0].strip()\\n if param and '{' not in param and '[' not in param:\\n if not check_var_name(param):\\n return False\\n elif m.group(2):\\n param = m.group(2).strip()\\n if param and not check_var_name(param):\\n return False\\n \\n return True\\n\\n # 辅助函数：利用 AST 检查 Python 代码中的变量名\\n def check_py_code(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n except SyntaxError:\\n # 如果语法错误解析失败，则跳过\\n return True\\n \\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if not check_var_name(node.id):\\n return False\\n elif isinstance(node, ast.arg):\\n if not check_var_name(node.arg):\\n return False\\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):\\n if not check_var_name(node.name):\\n return False\\n return True\\n\\n # 遍历项目文件，提取相关代码\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理前端常见代码文件及可能的 Python 后端文件\\n if ext in ['.js', '.ts', '.py', '.html']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext in ['.js', '.ts']:\\n if not check_js_code(content):\\n return False\\n elif ext == '.py':\\n if not check_py_code(content):\\n return False\\n elif ext == '.html':\\n # 提取 HTML 内部的 JS 代码\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_js_code(script):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令，推断只检查 HTML, CSS 和 JS 作为“代码文件”\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历当前代码项目下所有的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 分离出文件名和扩展名\\n basename, ext = os.path.splitext(file)\\n \\n # 判断是否为代码文件\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n \\n # 检查主文件名是否仅由小写字母和下划线组成\\n # 如果包含数字、大写字母或其他符号，均返回 False\\n if not re.match(r'^[a-z_]+$', basename):\\n return False\\n \\n # 如果没生成任何相关代码文件则返回False，否则所有代码文件均通过测试则返回True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中应优先以正斜杠 / 作为路径的分割符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"任何时候修改现有文件前，都不得在同级目录下创建该文件原始状态的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步是实现板块的旋转和翻转。请为板块添加旋转机制：选中或悬浮某个板块时，可以通过鼠标滚轮或者提供一个悬浮的“旋转手柄”，让板块能够顺时针或逆时针旋转（每次旋转45度）。特别注意的是，对于“平行四边形”，普通的平面旋转无法覆盖所有拼图情况，请额外增加一个翻转功能（比如双击该平行四边形使其水平或垂直镜像翻转）。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的修改文件前不得在同级目录创建备份的要求作废，更新为每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份；第二，在输出的所有代码文件里，不允许出现行数超过40行的函数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义当前项目可能涉及到的代码文件及文本文件扩展名\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', \\n '.json', '.md', '.txt', '.xml', '.svg',\\n '.py', '.java', '.c', '.cpp', '.sh'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码检测\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试用 UTF-8 进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果不能以 UTF-8 解码，说明存在非 UTF-8 编码的文件，返回 False\\n return False\\n except Exception:\\n # 忽略因权限等导致的其他系统级读取异常\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 获取工作空间的绝对路径\\n workspace = Path(workspace_path).resolve()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的无关隐藏目录（如.git），避免影响对“代码工程”真实层级的评判\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n root_path = Path(root).resolve()\\n \\n try:\\n # 计算当前目录相对工作空间根目录的相对路径\\n rel_path = root_path.relative_to(workspace)\\n except ValueError:\\n # 万一遇到非子路径异常（通常不会发生），安全起见跳过\\n continue\\n \\n # 工作空间根目录算第1层 (parts为空，len=0，深度=1)\\n # 直接子目录算第2层 (parts如 ('js',)，len=1，深度=2)\\n # 更深子目录算第3+层 (parts如 ('js', 'utils')，len=2，深度=3)\\n depth = 1 + len(rel_path.parts)\\n \\n # 目录层级必须小于或等于2层\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 检查单个变量名是否满足“同时包含数字、下划线、大写字母和小写字母”的条件\\n def check_var_name(name: str) -> bool:\\n has_digit = bool(re.search(r'\\\\d', name))\\n has_underscore = bool(re.search(r'_', name))\\n has_upper = bool(re.search(r'[A-Z]', name))\\n has_lower = bool(re.search(r'[a-z]', name))\\n return has_digit and has_underscore and has_upper and has_lower\\n\\n # 从 JavaScript 文本中提取变量和参数名\\n def extract_variables(code: str) -> set:\\n variables = set()\\n \\n # 清除注释避免干扰\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n code = re.sub(r'//.*', '', code)\\n \\n # 清除字符串避免文本内部出现 let/var 等单词干扰提取\\n code = re.sub(r\\\"'[^']*'\\\", \\\"''\\\", code)\\n code = re.sub(r'\\\"[^\\\"]*\\\"', '\\\"\\\"', code)\\n code = re.sub(r'`[^`]*`', '``', code)\\n \\n # 1. 匹配 let / var / const 声明后的首个变量名\\n decl_matches = re.findall(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_]\\\\w*)', code)\\n variables.update(decl_matches)\\n \\n # 2. 匹配 function 声明时的参数列表\\n func_matches = re.findall(r'\\\\bfunction(?:\\\\s+[a-zA-Z_]\\\\w*)?\\\\s*\\\\(([^)]*)\\\\)', code)\\n for param_str in func_matches:\\n # 去除可能存在的解构大括号\\n param_str = re.sub(r'[{}]', '', param_str)\\n for p in param_str.split(','):\\n p = p.strip()\\n if p:\\n m = re.match(r'^[a-zA-Z_]\\\\w*', p)\\n if m:\\n variables.add(m.group(0))\\n \\n # 3. 匹配常规箭头函数参数列表 (param1, param2) =>\\n arrow_matches = re.findall(r'\\\\(([^)]*)\\\\)\\\\s*=>', code)\\n for param_str in arrow_matches:\\n param_str = re.sub(r'[{}]', '', param_str)\\n for p in param_str.split(','):\\n p = p.strip()\\n if p:\\n m = re.match(r'^[a-zA-Z_]\\\\w*', p)\\n if m:\\n variables.add(m.group(0))\\n \\n # 4. 匹配单参数箭头函数参数 param =>\\n arrow_single = re.findall(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*=>', code)\\n variables.update(arrow_single)\\n \\n return variables\\n\\n # 从 CSS 文本中提取自定义属性（CSS变量）\\n def extract_css_variables(code: str) -> set:\\n variables = set()\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # CSS 变量声明示例: --var-name: value;\\n matches = re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', code)\\n variables.update(matches)\\n return variables\\n\\n all_vars = set()\\n has_code_files = False\\n\\n # 验证工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作区，处理 HTML, JS, CSS 文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n if file.endswith('.js'):\\n has_code_files = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n all_vars.update(extract_variables(code))\\n except Exception:\\n pass\\n \\n elif file.endswith('.css'):\\n has_code_files = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n all_vars.update(extract_css_variables(code))\\n except Exception:\\n pass\\n \\n elif file.endswith('.html'):\\n has_code_files = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 解析内联 script 块\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n all_vars.update(extract_variables(script))\\n \\n # 解析内联 style 块\\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.DOTALL | re.IGNORECASE)\\n for style in styles:\\n all_vars.update(extract_css_variables(style))\\n except Exception:\\n pass\\n\\n # 若没有任何相关代码文件或未提取出任何变量（七巧板游戏必然存在变量，除非未按要求生成）\\n if not has_code_files or not all_vars:\\n return False\\n\\n # 校验发现的所有变量名\\n for var in all_vars:\\n if not check_var_name(var):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check_filename_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义项目中可能涉及到的前端代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并转为小写进行判断\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断当前文件是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n \\n # 检查文件名是否严格符合要求：仅包含小写字母、下划线\\n # 允许且仅允许最后带有由一个点和纯小写字母组成的常规扩展名\\n # 大写字母、数字、连字符或其他任何多余符号都会导致正则匹配失败\\n if not re.match(r'^[a-z_]+(\\\\.[a-z]+)?$', file):\\n return False\\n \\n # 如果项目中没有任何代码文件返回 False，否则说明所有代码文件名均合规并返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中应优先以正斜杠 / 作为路径的分割符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在输出的所有代码文件里，不允许出现行数超过40行的函数\",\n \"validation_code\": \"def is_valid(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n import ast\\n\\n def remove_strings_and_comments(code: str) -> str:\\n def replacer(match):\\n # 将匹配到的字符串或注释替换为相应数量的换行符，以保证代码总行数和对应的行号不变\\n return '\\\\n' * match.group(0).count('\\\\n')\\n \\n # 匹配单行/多行注释、双引号/单引号/反引号包含的字符串字面量\\n pattern = re.compile(\\n r'//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n )\\n return pattern.sub(replacer, code)\\n\\n def check_js_code(code: str) -> bool:\\n cleaned = remove_strings_and_comments(code)\\n \\n # 预计算所有换行符的位置，实现 O(log N) 的行号查询\\n newline_indices = [i for i, char in enumerate(cleaned) if char == '\\\\n']\\n def get_line_number(index: int) -> int:\\n return bisect.bisect_right(newline_indices, index) + 1\\n\\n func_brace_indices = set()\\n \\n # 1. 处理由 function 和 => 引导的函数定义\\n for m in re.finditer(r'\\\\bfunction\\\\b|=>', cleaned):\\n paren_level = 0\\n idx = m.end()\\n while idx < len(cleaned):\\n c = cleaned[idx]\\n if c == '(':\\n paren_level += 1\\n elif c == ')':\\n paren_level -= 1\\n elif c == '{':\\n if paren_level <= 0:\\n func_brace_indices.add(idx)\\n break\\n elif c == '}':\\n if paren_level <= 0:\\n break\\n elif c == ';':\\n if paren_level <= 0:\\n break\\n idx += 1\\n\\n # 2. 处理对象中的方法简写 (例如 methodName(...) { )\\n method_pattern = re.compile(r'\\\\b(?!(?:if|for|while|switch|catch|function)\\\\b)([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n for m in method_pattern.finditer(cleaned):\\n paren_level = 0\\n idx = m.end() - 1 # idx 落在此处的 '(' 上\\n while idx < len(cleaned):\\n c = cleaned[idx]\\n if c == '(':\\n paren_level += 1\\n elif c == ')':\\n paren_level -= 1\\n elif c == '{':\\n if paren_level <= 0:\\n func_brace_indices.add(idx)\\n break\\n elif c == '}':\\n if paren_level <= 0:\\n break\\n elif c == ';':\\n if paren_level <= 0:\\n break\\n idx += 1\\n\\n stack = []\\n for match in re.finditer(r'\\\\{|\\\\}', cleaned):\\n char = match.group(0)\\n idx = match.start()\\n if char == '{':\\n is_func = (idx in func_brace_indices)\\n stack.append((is_func, idx))\\n else:\\n if stack:\\n is_func, start_idx = stack.pop()\\n if is_func:\\n start_line = get_line_number(start_idx)\\n end_line = get_line_number(idx)\\n if end_line - start_line + 1 > 40:\\n return False\\n return True\\n\\n def check_html_code(code: str) -> bool:\\n script_pattern = re.compile(r'<script\\\\b[^>]*>([\\\\s\\\\S]*?)</script>', re.IGNORECASE)\\n for match in script_pattern.finditer(code):\\n script_content = match.group(1)\\n # 添加前缀换行符保持代码片段的行号映射准确度\\n prefix_newlines = '\\\\n' * code.count('\\\\n', 0, match.start(1))\\n if not check_js_code(prefix_newlines + script_content):\\n return False\\n return True\\n\\n def check_py_code(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n if node.end_lineno - node.lineno + 1 > 40:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 针对常见的 Web 及相关代码文件类型做判定\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_js_code(code):\\n return False\\n except Exception:\\n pass\\n elif file.endswith(('.html', '.htm')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_html_code(code):\\n return False\\n except Exception:\\n pass\\n elif file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_py_code(code):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"为了让玩家拼图时对齐更方便，我们需要加入“吸附（Snapping）”效果。当玩家拖动一个板块靠近其他已经放置好的板块的顶点或边缘时，如果距离小于设定的阈值（比如15像素），自动让它吸附并对齐。这样可以避免手动微调的麻烦，提升游戏体验。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的输出的所有代码文件里不允许出现行数超过40行的函数的要求作废，更新为在输出的所有代码文件里，不允许出现行数超过30行的函数；第二，强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-，该要求后续所有轮次均需严格遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断的可能出现的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py'}\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并转为小写匹配\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 其他不可预见的读取错误（如权限问题）暂时忽略，避免误判\\n continue\\n\\n # 若所有代码文件都能成功以 UTF-8 读取（或者没有代码文件），则认为符合该要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"from pathlib import Path\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n workspace = Path(workspace_path)\\n \\n # 确保给定的 workspace_path 存在且为目录\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n # 递归遍历项目目录中的所有文件和文件夹\\n for path in workspace.rglob('*'):\\n # 获取当前路径相对于根目录的各部分\\n rel_parts = path.relative_to(workspace).parts\\n \\n # 排除包含隐藏文件夹（以 . 开头）或特殊环境文件夹的路径，避免干扰判定\\n if any(part.startswith('.') for part in rel_parts):\\n continue\\n if 'node_modules' in rel_parts or '__pycache__' in rel_parts:\\n continue\\n \\n # 若某目录相对根目录的分段数 >= 2 (如 src/utils)\\n # 意味着它是第二级或更深层级的子目录，即目录工程层级达到了 3 层及以上\\n if path.is_dir() and len(rel_parts) >= 2:\\n return False\\n \\n # 若某文件相对根目录的分段数 >= 3 (如 src/utils/main.py)\\n # 同样意味着它深于或等于第 3 层的目录之中\\n if path.is_file() and len(rel_parts) >= 3:\\n return False\\n \\n # 如果全项目均未超出层级限制，返回 True\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 辅助函数：判断变量名是否同时包含小写、大写、数字和下划线\\n def has_all_required_chars(name: str) -> bool:\\n has_lower = any(c.islower() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_digit = any(c.isdigit() for c in name)\\n has_under = '_' in name\\n return has_lower and has_upper and has_digit and has_under\\n\\n # 辅助函数：移除JS/CSS代码中的注释，防止正则误匹配\\n def remove_js_comments(text: str) -> str:\\n # 移除多行注释 /* ... */\\n text = re.sub(r'/\\\\*.*?\\\\*/', '', text, flags=re.DOTALL)\\n # 移除单行注释 // ...\\n text = re.sub(r'//.*', '', text)\\n return text\\n\\n variables_found = []\\n\\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # 兼容处理可能的Python文件\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n variables_found.append(node.id)\\n elif isinstance(node, ast.arg):\\n variables_found.append(node.arg)\\n except Exception:\\n pass\\n\\n # 处理JavaScript文件\\n elif file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n content = remove_js_comments(content)\\n \\n # 匹配 var, let, const 声明的变量名\\n decls = re.findall(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content)\\n variables_found.extend(decls)\\n \\n # 匹配函数的参数变量\\n func_params = re.findall(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', content)\\n for param_str in func_params:\\n params = [p.strip() for p in param_str.split(',') if p.strip()]\\n for p in params:\\n p_name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\\n variables_found.append(p_name)\\n except Exception:\\n pass\\n\\n # 处理HTML文件中的嵌入式JS代码\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 移除HTML注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 提取所有script标签中的内容\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n script = remove_js_comments(script)\\n # 匹配 var, let, const 声明的变量名\\n decls = re.findall(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', script)\\n variables_found.extend(decls)\\n \\n # 匹配函数的参数变量\\n func_params = re.findall(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', script)\\n for param_str in func_params:\\n params = [p.strip() for p in param_str.split(',') if p.strip()]\\n for p in params:\\n p_name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p_name):\\n variables_found.append(p_name)\\n except Exception:\\n pass\\n\\n # 处理CSS文件中的CSS变量（Custom Properties）\\n elif file.endswith('.css'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n content = remove_js_comments(content)\\n \\n # 匹配CSS变量定义，例如 --main-color: ...\\n css_vars = re.findall(r'(--[a-zA-Z0-9_-]+)\\\\s*:', content)\\n variables_found.extend(css_vars)\\n except Exception:\\n pass\\n\\n # 过滤掉可能的空值\\n variables_found = [v for v in variables_found if v]\\n\\n # 若未找到任何变量，默认算作满足（通常代码中肯定存在变量）\\n if not variables_found:\\n return True\\n\\n # 判断是否每一个变量都符合要求\\n for var in variables_found:\\n if not has_all_required_chars(var):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check_code_filenames(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', \\n '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', \\n '.rb', '.php', '.sh', '.swift', '.kt'\\n }\\n \\n # 遍历当前项目工作空间下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件的扩展名\\n _, ext = os.path.splitext(file)\\n \\n # 判断是否为代码文件（使用 lower() 以覆盖可能含有大写后缀的情况并进行识别）\\n if ext.lower() in code_extensions:\\n # 检查文件名是否严格符合要求：\\n # 除了唯一的一个扩展名圆点外，文件基础名及扩展名中只允许存在小写字母(a-z)和下划线(_)\\n # 不能包含数字、大写字母、多个圆点或其他任何特殊符号\\n if not re.fullmatch(r'[a-z_]+\\\\.[a-z_]+', file):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中应优先以正斜杠 / 作为路径的分割符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在输出的所有代码文件里，不允许出现行数超过30行的函数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def replacer(match):\\n s = match.group(0)\\n # 替换为空格，但保留原有的换行符，以便维持正确的行号对应关系不变\\n return '\\\\n' * s.count('\\\\n') + ' ' * (len(s) - s.count('\\\\n'))\\n\\n def clean_js_code(code_str: str) -> str:\\n # 正则匹配 JS 中的字符串字面量(支持转义符)、块注释、行注释\\n pattern = r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)' \\\\\\n r'|(?P<bcomment>/\\\\*[\\\\s\\\\S]*?\\\\*/)' \\\\\\n r'|(?P<lcomment>//[^\\\\n]*)'\\n return re.sub(pattern, replacer, code_str)\\n\\n def get_js_func_ranges(code_str: str) -> list:\\n cleaned = clean_js_code(code_str)\\n func_ranges = []\\n start_indices = []\\n \\n # 1. 匹配 function 关键字\\n for m in re.finditer(r'\\\\bfunction\\\\b', cleaned):\\n brace_idx = cleaned.find('{', m.end())\\n if brace_idx != -1:\\n start_indices.append((m.start(), brace_idx))\\n \\n # 2. 匹配箭头函数 => {\\n for m in re.finditer(r'=>\\\\s*\\\\{', cleaned):\\n brace_idx = cleaned.find('{', m.start())\\n if brace_idx != -1:\\n start_indices.append((m.start(), brace_idx))\\n \\n # 3. 匹配可能的对象方法或类方法的语法（排除 if/for 等控制流关键字）\\n pattern = r'\\\\b(?!(?:if|for|while|switch|catch|with|function)\\\\b)([a-zA-Z_$][a-zA-Z_$0-9]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n for m in re.finditer(pattern, cleaned):\\n brace_idx = cleaned.find('{', m.start())\\n if brace_idx != -1:\\n start_indices.append((m.start(), brace_idx))\\n \\n # 坐标去重，防止相同位置被多种规则重复捕捉\\n start_indices = list(set(start_indices))\\n \\n for start_idx, brace_idx in start_indices:\\n stack = 0\\n end_idx = -1\\n # 利用栈进行花括号配对，寻找函数结束位置\\n for i in range(brace_idx, len(cleaned)):\\n if cleaned[i] == '{':\\n stack += 1\\n elif cleaned[i] == '}':\\n stack -= 1\\n if stack == 0:\\n end_idx = i\\n break\\n \\n # 若配对成功，则根据 \\\\n 计算函数所跨越的行数\\n if end_idx != -1:\\n start_line = code_str.count('\\\\n', 0, start_idx) + 1\\n end_line = code_str.count('\\\\n', 0, end_idx) + 1\\n func_ranges.append((start_line, end_line))\\n \\n return func_ranges\\n\\n def check_py_code(code_str: str) -> bool:\\n try:\\n tree = ast.parse(code_str)\\n except SyntaxError:\\n return True # 语法错误无法解析时，略过不将其判定为包含长函数\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno:\\n lines = node.end_lineno - node.lineno + 1\\n if lines > 30:\\n return False\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n continue\\n \\n # 针对 JavaScript / TypeScript 类型文件\\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\\n ranges = get_js_func_ranges(content)\\n for start, end in ranges:\\n if (end - start + 1) > 30:\\n return False\\n \\n # 针对包含内嵌脚本的 HTML 文件\\n elif ext in ['.html', '.htm']:\\n # 提取 <script> ... </script> 块内容\\n for script_match in re.finditer(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE):\\n script_content = script_match.group(1)\\n # 补齐脚本前面的换行符，以便维持在原文件中的正确行号\\n start_pos = script_match.start(1)\\n newlines_before = content.count('\\\\n', 0, start_pos)\\n padded_script = '\\\\n' * newlines_before + script_content\\n \\n ranges = get_js_func_ranges(padded_script)\\n for start, end in ranges:\\n if (end - start + 1) > 30:\\n return False\\n \\n # 针对可能存在的 Python 代码文件\\n elif ext == '.py':\\n if not check_py_code(content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_coding_declaration(response: str, workspace_path: str) -> bool:\\n # 根据用户指令内容推断，可能的代码文件扩展名为HTML, CSS, JS\\n code_extensions = {'.html', '.css', '.js'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取顶行内容\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取发生异常（如编码错误或权限问题），判定为不满足要求\\n return False\\n \\n # 如果所有的代码文件均满足要求，则返回 True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"基础操作都有了，现在引入真正的游戏关卡。请在右侧控制面板增加3个关卡按钮（例如：狐狸、帆船、天鹅）。点击任意一个关卡，在游戏区中央渲染出一个由7块七巧板特定角度组合而成的黑色阴影轮廓作为“目标图案”。这个轮廓没有内部线条，供玩家参照去拼凑。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的输出的所有代码文件里不允许出现行数超过30行的函数的要求作废，更新为在输出的所有代码文件里，不允许出现行数超过25行的函数；第二，项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起，该要求后续所有轮次均需严格遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令推断可能存在的代码和配置文件后缀\\n code_extensions = {\\n '.html', '.css', '.js', '.json', \\n '.md', '.log', '.txt', '.py', \\n '.ts', '.jsx', '.tsx', '.sh'\\n }\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，则说明包含了非 UTF-8 编码的字符\\n return False\\n except Exception:\\n # 忽略其他非编码相关的偶发异常（如权限异常等）\\n pass\\n \\n # 如果所有代码文件都能正常被 UTF-8 成功解码，或没有发现异常，说明满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 如果工作区路径不存在，则无法判断\\n if not os.path.exists(workspace_path):\\n return False\\n \\n workspace = Path(workspace_path).resolve()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉以 '.' 开头的隐藏文件夹（如 .git, .idea等），防止对层级判断造成干扰\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n root_path = Path(root).resolve()\\n try:\\n # 计算当前遍历到的目录相对于根目录的相对路径\\n rel_path = root_path.relative_to(workspace)\\n except ValueError:\\n continue\\n \\n # 对于根目录，rel_path.parts 为 ()，长度为 0（代表第1层级）\\n # 对于一级子目录，rel_path.parts 为 ('dir1',)，长度为 1（代表第2层级）\\n # 对于二级子目录，rel_path.parts 为 ('dir1', 'dir2')，长度为 2（代表第3层级）\\n # 若总层级要求小于等于2层，则相对路径的层级部分长度不得大于1\\n if len(rel_path.parts) > 1:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def check_name(name: str) -> bool:\\n has_upper = bool(re.search(r'[A-Z]', name))\\n has_lower = bool(re.search(r'[a-z]', name))\\n has_digit = bool(re.search(r'[0-9]', name))\\n has_under = '_' in name\\n return has_upper and has_lower and has_digit and has_under\\n\\n js_var_pattern = re.compile(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理预期中可能生成的代码文件\\n if ext not in ['.js', '.html', '.css', '.py']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if not check_name(node.id):\\n return False\\n elif isinstance(node, ast.arg):\\n if not check_name(node.arg):\\n return False\\n except Exception:\\n pass\\n \\n elif ext in ['.js', '.html']:\\n text_to_search = content\\n if ext == '.html':\\n # 针对HTML，仅提取 script 标签内容进行检验\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n text_to_search = \\\"\\\\n\\\".join(scripts)\\n \\n # 预处理：清除单行、多行注释以及反引号、双引号、单引号包围的字符串字面量，避免被正则误认为变量声明\\n text_to_search = re.sub(r'//.*', '', text_to_search)\\n text_to_search = re.sub(r'/\\\\*.*?\\\\*/', '', text_to_search, flags=re.DOTALL)\\n text_to_search = re.sub(r'`.*?`', '``', text_to_search, flags=re.DOTALL)\\n text_to_search = re.sub(r'\\\".*?(?<!\\\\\\\\)\\\"', '\\\"\\\"', text_to_search)\\n text_to_search = re.sub(r\\\"'.*?(?<!\\\\\\\\)'\\\", \\\"''\\\", text_to_search)\\n \\n # 开始匹配通过 let / var / const 声明的变量\\n for match in js_var_pattern.findall(text_to_search):\\n if not check_name(match):\\n return False\\n \\n elif ext == '.css':\\n # 预处理：清除多行注释\\n text_to_search = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 开始匹配通过 -- 声明的 CSS 自定义属性（CSS变量）\\n for match in re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', text_to_search):\\n # 遇到类似 --my-var: 的形式，取出 my-var 进行验证\\n # 考虑到 CSS 变量常使用破折号连字符，在要求强制使用下划线、大小写、数字的情况下仍应进行统一验证\\n if not check_name(match):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证代码文件名是否只包含小写字母和下划线（除了扩展名的点以外）。\\n \\\"\\\"\\\"\\n # 常见代码文件扩展名集合\\n code_exts = {\\n '.js', '.html', '.css', '.py', '.ts', '.jsx', '.tsx', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt',\\n '.vue', '.svelte', '.scss', '.sass', '.less'\\n }\\n \\n # 确保当前代码项目地址存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 忽略隐藏文件如 .DS_Store 或 .gitignore\\n if file_name.startswith('.'):\\n continue\\n \\n # 提取文件扩展名并判断是否属于代码文件\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n \\n # 剔除文件名中的点符号\\n name_without_dots = file_name.replace('.', '')\\n \\n # 正则匹配：除了点符号外，文件名应当仅仅由小写字母和下划线组成\\n if not re.fullmatch(r'[a-z_]+', name_without_dots):\\n return False\\n \\n # 如果没有找到任何代码文件，说明生成失败，返回 False\\n return found_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中应优先以正斜杠 / 作为路径的分割符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在输出的所有代码文件里，不允许出现行数超过25行的函数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def remove_comments_and_strings(code: str) -> str:\\n # 替换匹配到的字符串及注释为等长的空格，同时保留所有的换行符，以便不影响行数计算\\n def replacer(match):\\n s = match.group(0)\\n return ''.join('\\\\n' if c == '\\\\n' else ' ' for c in s)\\n \\n pattern = re.compile(\\n r'<!--.*?-->|' # HTML 注释\\n r'//.*?$|' # 单行注释\\n r'/\\\\*.*?\\\\*/|' # 多行注释\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\" # 单引号字符串\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|' # 双引号字符串\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', # 模板字符串\\n re.DOTALL | re.MULTILINE\\n )\\n return pattern.sub(replacer, code)\\n\\n def get_function_blocks(clean_code: str):\\n # 使用字典去重：以 { 的索引为键，保存函数声明能够匹配到的最早的索引为值\\n blocks = {}\\n \\n # 1. 匹配 function 关键字声明的函数\\n for match in re.finditer(r'\\\\bfunction\\\\b[^{]*\\\\{', clean_code):\\n b_idx = match.end() - 1\\n blocks[b_idx] = min(blocks.get(b_idx, len(clean_code)), match.start())\\n \\n # 2. 匹配箭头函数\\n for match in re.finditer(r'=>\\\\s*\\\\{', clean_code):\\n b_idx = match.end() - 1\\n blocks[b_idx] = min(blocks.get(b_idx, len(clean_code)), match.start())\\n \\n # 3. 匹配常规的类方法或对象方法\\n for match in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', clean_code):\\n name = match.group(1)\\n # 排除由于具有括号和花括号特征而被误判的控制流和内置关键字\\n if name not in {'if', 'for', 'while', 'switch', 'catch', 'with', 'return', 'typeof', 'new', 'else'}:\\n b_idx = match.end() - 1\\n blocks[b_idx] = min(blocks.get(b_idx, len(clean_code)), match.start())\\n \\n return [(v, k) for k, v in blocks.items()]\\n\\n def check_js_html_lines(code: str, max_lines: int = 25) -> bool:\\n clean_code = remove_comments_and_strings(code)\\n blocks = get_function_blocks(clean_code)\\n \\n for decl_start_idx, brace_start_idx in blocks:\\n count = 1\\n i = brace_start_idx + 1\\n while i < len(clean_code):\\n if clean_code[i] == '{':\\n count += 1\\n elif clean_code[i] == '}':\\n count -= 1\\n if count == 0:\\n end_idx = i\\n # 截取函数定义到结束的片段，计算行数\\n func_str = clean_code[decl_start_idx:end_idx+1]\\n lines = func_str.count('\\\\n') + 1\\n if lines > max_lines:\\n return False\\n break\\n i += 1\\n return True\\n\\n def check_python_lines(code: str, max_lines: int = 25) -> bool:\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n # 存在语法错误的情况暂且放过\\n return True\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n lines = node.end_lineno - node.lineno + 1\\n if lines > max_lines:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_files = False\\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.py', '.html'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n has_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_python_lines(code, 25):\\n return False\\n elif ext == '.html':\\n # HTML文件需提取所有 script 内容作为 JS 进行分析\\n scripts = []\\n pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\\n for match in pattern.finditer(code):\\n scripts.append(match.group(1))\\n js_code = '\\\\n'.join(scripts)\\n if not check_js_html_lines(js_code, 25):\\n return False\\n else:\\n if not check_js_html_lines(code, 25):\\n return False\\n \\n return True if has_files else False\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义项目中可能出现的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的第一行\\n first_line = f.readline()\\n \\n # 判断首行是否包含特定的编码声明内容\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法打开或发生编码读取错误，则判定为不满足要求\\n return False\\n \\n # 如果项目中没有代码文件，则说明没有满足生成代码的要求，返回 False；反之返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"有了目标之后，需要加上通关检测逻辑。当玩家将所有7个板块都拖入黑色阴影轮廓内，并且每块的坐标和旋转角度与目标数据匹配（允许一定的容差值）时，判定为拼图成功。此时请弹出一个“恭喜通关”的模态框，并在屏幕上播放一些简单的彩纸飘落或烟花庆祝动画。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的生成的代码文件中应优先以正斜杠 / 作为路径的分割符的要求作废，更新为在编写的代码文件中，务必优先挑选反斜杠 \\\\ 来表示路径分隔；第二，原有的输出的所有代码文件里不允许出现行数超过25行的函数的要求作废，更新为生成代码时，请将各函数的行数严格控制在 20 行及以内。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码与相关配置文件扩展名集合\\n code_exts = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.go', '.rs', '.php', '.rb', '.swift', '.kt',\\n '.json', '.md', '.log', '.txt', '.yaml', '.yml', \\n '.xml', '.sh', '.bat'\\n }\\n \\n # 若工作区路径不存在，视作未生成符合条件的代码工程\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 格式读取文件全部内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，说明文件非 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能的文件读取异常（如权限问题等）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_directory_depth(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 如果工作空间路径不存在或不是一个目录，直接判定为不满足\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n workspace = Path(workspace_path).resolve()\\n\\n for root, dirs, files in os.walk(workspace):\\n # 过滤掉以 '.' 开头的隐藏文件夹（如 .git, .vscode 等），防止对判定产生干扰\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 检查当前目录下的所有文件和文件夹\\n for name in dirs + files:\\n # 忽略隐藏文件\\n if name.startswith('.'):\\n continue\\n \\n item_path = Path(root) / name\\n \\n try:\\n # 获取相对于工程根目录的相对路径\\n rel_path = item_path.relative_to(workspace)\\n \\n # rel_path.parts 可以获取相对路径层级切分后的元组\\n # 例如 'js/main.js' 会被切分为 ('js', 'main.js')，长度为 2\\n if len(rel_path.parts) > 2:\\n return False\\n except ValueError:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_var_name(name: str) -> bool:\\n \\\"\\\"\\\"检查变量名是否同时包含数字、下划线、大写字母和小写字母\\\"\\\"\\\"\\n has_digit = any(c.isdigit() for c in name)\\n has_under = '_' in name\\n has_upper = any(c.isupper() for c in name)\\n has_lower = any(c.islower() for c in name)\\n return has_digit and has_under and has_upper and has_lower\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 针对 Python 文件的变量判断\\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 判断赋值变量名\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n if not check_var_name(node.id):\\n return False\\n # 判断函数参数变量名\\n elif isinstance(node, ast.arg):\\n if not check_var_name(node.arg):\\n return False\\n except SyntaxError:\\n pass\\n \\n # 针对 JS/TS 及 HTML 里的 JS 代码部分做变量判断\\n if ext in ['.js', '.ts', '.html']:\\n if ext == '.html':\\n # 提取 <script> 标签内容\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\\n js_content = '\\\\n'.join(scripts)\\n else:\\n js_content = content\\n \\n if js_content.strip():\\n # 粗略地去除字符串和注释，减少匹配误判\\n js_content = re.sub(r'([\\\"\\\\'`]).*?\\\\1', '\\\"\\\"', js_content, flags=re.DOTALL)\\n js_content = re.sub(r'/\\\\*.*?\\\\*/', '', js_content, flags=re.DOTALL)\\n js_content = re.sub(r'//.*', '', js_content)\\n \\n # 匹配 var, let, const 声明的变量\\n statements = re.findall(r'\\\\b(?:var|let|const)\\\\s+([^;\\\\n]+)', js_content)\\n for stmt in statements:\\n parts = stmt.split(',')\\n for part in parts:\\n var_name = part.split('=')[0].strip()\\n # 提取开头的合法标识符\\n m = re.match(r'^([a-zA-Z_$][0-9a-zA-Z_$]*)', var_name)\\n if m:\\n name = m.group(1)\\n if name not in ['typeof', 'yield', 'await']:\\n if not check_var_name(name):\\n return False\\n \\n # 匹配函数参数\\n funcs = re.findall(r'\\\\bfunction\\\\s*(?:[a-zA-Z_$][0-9a-zA-Z_$]*)?\\\\s*\\\\(([^)]*)\\\\)', js_content)\\n for params_str in funcs:\\n params = params_str.split(',')\\n for p in params:\\n p = p.strip()\\n m = re.match(r'^([a-zA-Z_$][0-9a-zA-Z_$]*)', p)\\n if m:\\n name = m.group(1)\\n if not check_var_name(name):\\n return False\\n\\n # 针对 CSS 及 HTML 里的 CSS 代码做变量判断（检测CSS自定义属性/变量）\\n if ext in ['.css', '.html']:\\n # 寻找形如 --my-var 的CSS变量\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)\\\\s*:', content)\\n for css_var in css_vars:\\n if not check_var_name(css_var):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check_code_filename(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见代码文件的扩展名集合\\n code_exts = {\\n '.html', '.htm', '.css', '.js', '.ts', \\n '.jsx', '.tsx', '.py', '.java', '.c', \\n '.cpp', '.h', '.hpp', '.cs', '.go', \\n '.rb', '.php', '.swift', '.kt', '.rs', \\n '.sh', '.bash'\\n }\\n \\n has_code_file = False\\n \\n # 遍历当前项目的工作区文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件的扩展名并转小写进行归类匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅针对代码文件进行校验，跳过文档和配置文件 (.md, .json, .log 等)\\n if ext in code_exts:\\n has_code_file = True\\n \\n # 规则1：除了作为扩展名分隔符的点之外，不得出现其他任何点符号（例如不允许出现多重后缀 app.min.js）\\n if file.count('.') > 1:\\n return False\\n \\n # 规则2：整个文件名中（去掉唯一的扩展名分隔符 `.` 后），只能包含小写字母和下划线\\n # 任何大写字母、数字、连字符等其他符号均会导致正则匹配失败\\n name_without_dot = file.replace('.', '')\\n if not re.fullmatch(r'[a-z_]+', name_without_dot):\\n return False\\n \\n # 如果项目中成功生成了至少一个符合规范的代码文件，且没有发现任何违规项，则要求被满足\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写的代码文件中，务必优先挑选反斜杠 \\\\ 来表示路径分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在 20 行及以内\",\n \"validation_code\": \"import os\\nimport ast\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\n def check_js_code(code: str) -> bool:\\n \\\"\\\"\\\"\\n 验证JavaScript或HTML script标签内的函数长度\\n \\\"\\\"\\\"\\n # 第一步：清洗代码内的注释与字符串，保留换行符使代码行号不变\\n cleaned = []\\n i = 0\\n n = len(code)\\n state = 'NORMAL'\\n while i < n:\\n c = code[i]\\n if state == 'NORMAL':\\n if i < n - 1 and code[i:i+2] == '//':\\n state = 'COM_S'\\n cleaned.append(' ')\\n i += 2\\n elif i < n - 1 and code[i:i+2] == '/*':\\n state = 'COM_M'\\n cleaned.append(' ')\\n i += 2\\n elif c == \\\"'\\\":\\n state = 'STR_S'\\n cleaned.append(' ')\\n i += 1\\n elif c == '\\\"':\\n state = 'STR_D'\\n cleaned.append(' ')\\n i += 1\\n elif c == '`':\\n state = 'STR_B'\\n cleaned.append(' ')\\n i += 1\\n else:\\n cleaned.append(c)\\n i += 1\\n elif state == 'COM_S':\\n if c == '\\\\n':\\n state = 'NORMAL'\\n cleaned.append('\\\\n')\\n else:\\n cleaned.append(' ')\\n i += 1\\n elif state == 'COM_M':\\n if i < n - 1 and code[i:i+2] == '*/':\\n state = 'NORMAL'\\n cleaned.append(' ')\\n i += 2\\n else:\\n cleaned.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n elif state in ('STR_S', 'STR_D', 'STR_B'):\\n if c == '\\\\\\\\':\\n cleaned.append(' ')\\n i += 1\\n if i < n:\\n c2 = code[i]\\n cleaned.append('\\\\n' if c2 == '\\\\n' else ' ')\\n i += 1\\n continue\\n elif (state == 'STR_S' and c == \\\"'\\\") or \\\\\\n (state == 'STR_D' and c == '\\\"') or \\\\\\n (state == 'STR_B' and c == '`'):\\n state = 'NORMAL'\\n cleaned.append(' ')\\n i += 1\\n else:\\n cleaned.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n\\n cleaned_code = \\\"\\\".join(cleaned)\\n \\n # 第二步：利用正则提取清理过后代码里的函数定义开头部分\\n p1 = r'\\\\bfunction(?:\\\\s+\\\\w+)?\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n p2 = r'(?:\\\\([^)]*\\\\)|\\\\b\\\\w+)\\\\s*=>\\\\s*\\\\{'\\n p3 = r'\\\\b(?!(?:if|for|while|catch|switch)\\\\b)\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n pattern = re.compile(f'({p1}|{p2}|{p3})')\\n \\n # 第三步：跟踪匹配并计算每个函数的真实行数\\n for match in pattern.finditer(cleaned_code):\\n start_idx = match.start()\\n brace_idx = match.end() - 1\\n \\n brace_count = 1\\n end_idx = -1\\n for j in range(brace_idx + 1, len(cleaned_code)):\\n if cleaned_code[j] == '{':\\n brace_count += 1\\n elif cleaned_code[j] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_idx = j\\n break\\n \\n if end_idx != -1:\\n # 行数为对应字符串中换行符数加一\\n lines_count = code[start_idx:end_idx+1].count('\\\\n') + 1\\n if lines_count > 20:\\n return False\\n \\n return True\\n\\n def check_py_code(file_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证Python文件内的函数长度\\n \\\"\\\"\\\"\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n code = f.read()\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n lines_count = node.end_lineno - node.lineno + 1\\n if lines_count > 20:\\n return False\\n return True\\n except Exception:\\n return True # 解析异常跳过\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目文件树\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n if ext == '.py':\\n if not check_py_code(file_path):\\n return False\\n elif ext == '.js':\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n code = f.read()\\n if not check_js_code(code):\\n return False\\n elif ext in ('.html', '.htm'):\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 寻找html中的script块代码进行分析\\n scripts = re.findall(r'(?i)<script[^>]*>(.*?)</script>', content, re.DOTALL)\\n for script in scripts:\\n if script.strip():\\n if not check_js_code(script):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名集合，专门用于过滤出代码文件\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.h', '.java'}\\n code_files_found = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验，跳过 .md, .json, .log 等非代码文件\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 读取可以自动跳过不可见的 BOM 头信息\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若文件无法读取（例如空文件或编码异常读错），则视为不满足要求\\n return False\\n\\n # 若遍历完毕且中途没有返回 False，需确保至少验证了一个代码文件\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"功能越来越完善了。现在来优化游戏辅助功能：加入计时器和重置功能。当玩家在选中某个关卡并开始拖动第一块七巧板时，计时器开始走秒（格式 mm:ss），显示在控制面板上。同时增加一个“重置”按钮，点击后将当前的7个板块瞬间带动画飞回到初始的备选区位置，同时计时器归零重置。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的生成代码时，请将各函数的行数严格控制在20行及以内的要求作废，更新为生成代码时，请将各函数的行数严格控制在15行及以内；第二，后续所有轮次生成的每个代码文件，算上所有的空行的总行数必须保持为奇数，该要求需严格遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令内容，推断可能生成的代码文件与相关文本文件类型\\n code_extensions = {'.html', '.css', '.js', '.py', '.json', '.md', '.log'}\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，则说明该文件不是使用 UTF-8 编码的\\n return False\\n except Exception:\\n # 忽略其他非编码相关的异常（如权限问题等）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n from pathlib import Path\\n \\n workspace = Path(workspace_path)\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for p in workspace.rglob('*'):\\n rel_path = p.relative_to(workspace)\\n \\n # 忽略隐藏文件夹（如 .git 或 .vscode 等），防止对构建代码层级判断产生干扰\\n if any(part.startswith('.') for part in rel_path.parts):\\n continue\\n \\n # 仅针对文件夹计算其目录深度\\n if p.is_dir():\\n depth = len(rel_path.parts)\\n # 若任意合法目录的层级大于2层，则判定不满足要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_var_name(name):\\n has_digit = any(c.isdigit() for c in name)\\n has_lower = any(c.islower() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_under = '_' in name\\n return has_digit and has_lower and has_upper and has_under\\n\\n def clean_js_content(content):\\n # 移除转义的字符串以防止误匹配\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content, flags=re.DOTALL)\\n # 移除单行与多行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n return content\\n\\n def extract_js_vars(content):\\n vars_to_check = set()\\n \\n # 匹配 `let`、`var`、`const` 声明的基础变量名\\n for match in re.finditer(r'\\\\b(?:let|var|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', content):\\n vars_to_check.add(match.group(1))\\n \\n # 匹配普通赋值操作（排除对象属性的赋值如 object.property = value）\\n for match in re.finditer(r'(?<!\\\\.)\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=(?!=)', content):\\n vars_to_check.add(match.group(1))\\n \\n # 匹配函数参数\\n for match in re.finditer(r'\\\\bfunction\\\\s+[a-zA-Z0-9_$]*\\\\s*\\\\((.*?)\\\\)', content):\\n params = match.group(1).split(',')\\n for p in params:\\n p = p.strip()\\n if p and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', p):\\n vars_to_check.add(p)\\n \\n return vars_to_check\\n\\n html_script_pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in ['.py', '.js', '.html', '.css']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n vars_to_check = set()\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n # 排除内置双下划线命名的变量，如 __name__\\n if not (node.id.startswith('__') and node.id.endswith('__')):\\n vars_to_check.add(node.id)\\n elif isinstance(node, ast.arg):\\n vars_to_check.add(node.arg)\\n except Exception:\\n pass\\n \\n elif ext == '.js':\\n clean_content = clean_js_content(content)\\n vars_to_check.update(extract_js_vars(clean_content))\\n \\n elif ext == '.html':\\n for script_match in html_script_pattern.finditer(content):\\n script_content = script_match.group(1)\\n clean_content = clean_js_content(script_content)\\n vars_to_check.update(extract_js_vars(clean_content))\\n \\n elif ext == '.css':\\n # 匹配 CSS 原生变量\\n for match in re.finditer(r'--([a-zA-Z0-9_-]+)\\\\s*:', content):\\n vars_to_check.add(match.group(1))\\n \\n for var in vars_to_check:\\n # 排除不小心抓取到的 JS 内置常量词汇\\n if var in ['true', 'false', 'null', 'undefined', 'NaN', 'Infinity']:\\n continue\\n \\n # 若发现任意一个提取出的变量名未同时具备要求中的四种元素，则判定失败\\n if not check_var_name(var):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码文件扩展名集合（涵盖题目指令中提及的 HTML, CSS, JS 及其他常见代码类型）\\n code_extensions = {\\n '.html', '.htm',\\n '.css', '.scss', '.sass', '.less',\\n '.js', '.ts', '.jsx', '.tsx',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp',\\n '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n \\n # 如果工作区路径不存在，无法进行验证\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 拆分文件名主体和扩展名\\n name, ext = os.path.splitext(file)\\n \\n # 我们仅对代码文件进行限制检查\\n if ext.lower() in code_extensions:\\n # 检查文件名主体（不含最后的扩展名部分）是否仅由小写字母和下划线组成\\n # 若包含数字、大写字母或其他符号（如中划线 -、额外的点 . 等），均不合规并返回 False\\n if not re.fullmatch(r'[a-z_]+', name):\\n return False\\n \\n # 检查扩展名部分（去掉分隔点 `.` 后）是否也完全仅由小写字母和下划线组成\\n # 该检测能够有效防止文件带有大写扩展名（例如 .JS, .Html 等）\\n if not re.fullmatch(r'[a-z_]+', ext[1:]):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写的代码文件中，务必优先挑选反斜杠 \\\\ 来表示路径分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在 15 行及以内\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n \\n def get_py_function_lengths(code: str):\\n lengths = []\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\\n lengths.append(node.end_lineno - node.lineno + 1)\\n except Exception:\\n pass\\n return lengths\\n\\n def get_js_function_lengths(code: str):\\n # 消除注释和字符串的干扰，但是保留原有的换行符，确保行数匹配不出错\\n pattern = r'(/\\\\*.*?\\\\*/|//[^\\\\n]*|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n def replacer(match):\\n text = match.group(0)\\n return '\\\\n' * text.count('\\\\n') + ' ' * (len(text) - text.count('\\\\n'))\\n \\n cleaned_code = re.sub(pattern, replacer, code, flags=re.DOTALL)\\n lengths = []\\n \\n # 常见的三种JS函数定义的正则表达式探测\\n p1 = r'\\\\bfunction\\\\b[^\\\\{;}]*\\\\{'\\n p2 = r'=>\\\\s*\\\\{'\\n p3 = r'\\\\b(?!(?:if|for|while|switch|catch|with)\\\\b)\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n func_pattern = f'(?:{p1}|{p2}|{p3})'\\n\\n # 找到所有可能是函数起始的 '{' 的索引\\n starts = [m.end() - 1 for m in re.finditer(func_pattern, cleaned_code)]\\n \\n for start_idx in starts:\\n if cleaned_code[start_idx] != '{':\\n continue\\n brace_level = 1\\n end_idx = -1\\n # 从 '{' 的下一位开始进行括号配对\\n for j in range(start_idx + 1, len(cleaned_code)):\\n if cleaned_code[j] == '{':\\n brace_level += 1\\n elif cleaned_code[j] == '}':\\n brace_level -= 1\\n if brace_level == 0:\\n end_idx = j\\n break\\n # 成功配对到函数闭合的 '}'\\n if end_idx != -1:\\n start_line = cleaned_code.count('\\\\n', 0, start_idx)\\n end_line = cleaned_code.count('\\\\n', 0, end_idx)\\n lengths.append(end_line - start_line + 1)\\n \\n return lengths\\n\\n def extract_scripts_from_html(html_code: str):\\n scripts = []\\n for match in re.finditer(r'<script[^>]*>(.*?)</script>', html_code, flags=re.IGNORECASE | re.DOTALL):\\n scripts.append(match.group(1))\\n return scripts\\n\\n # ================= 1. 检查工作区生成的代码项目 =================\\n for root, dirs, files in os.walk(workspace_path):\\n if any(ignored in root for ignored in ['node_modules', '.git', 'venv']):\\n continue\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n code = f.read()\\n except Exception:\\n continue\\n\\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\\n lengths = get_js_function_lengths(code)\\n if any(l > 15 for l in lengths):\\n return False\\n elif ext == '.py':\\n lengths = get_py_function_lengths(code)\\n if any(l > 15 for l in lengths):\\n return False\\n elif ext in ['.html', '.htm']:\\n scripts = extract_scripts_from_html(code)\\n for script in scripts:\\n lengths = get_js_function_lengths(script)\\n if any(l > 15 for l in lengths):\\n return False\\n\\n # ================= 2. 检查助手最后一轮回复中的代码块 =================\\n code_blocks = re.findall(r'```([a-zA-Z]*)\\\\n(.*?)```', response, flags=re.DOTALL)\\n for lang, code in code_blocks:\\n lang = lang.lower().strip()\\n if lang in ('js', 'javascript', 'ts', 'typescript', ''):\\n lengths = get_js_function_lengths(code)\\n if any(l > 15 for l in lengths):\\n return False\\n elif lang in ('py', 'python'):\\n lengths = get_py_function_lengths(code)\\n if any(l > 15 for l in lengths):\\n return False\\n elif lang in ('html', 'htm'):\\n scripts = extract_scripts_from_html(code)\\n for script in scripts:\\n lengths = get_js_function_lengths(script)\\n if any(l > 15 for l in lengths):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能被视为“代码文件”的常见扩展名集合，排除.md, .json, .log等非代码文件\\n code_extensions = {\\n '.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', \\n '.java', '.c', '.cpp', '.h', '.cs', '.go', '.php', '.sh'\\n }\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n code_file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含目标编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果打开或读取发生异常（如编码非utf-8），则视作不满足要求\\n return False\\n\\n # 若未找到任何代码文件，通常意味着未按要求生成项目代码，返回False；否则返回True\\n return code_file_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"后续所有轮次生成的每个代码文件，算上所有的空行的总行数必须保持为奇数\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 提取符合前序交互轮次预期的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 若文件完全为空，则行数为 0，0 是偶数，不符合奇数要求\\n if not content:\\n return False\\n \\n # 按照换行符切分以统计所有行（包括文件末尾可能由于回车产生的空行）\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果文件因编码等异常无法作为文本正常读取，可视为未能满足代码规范要求\\n return False\\n \\n # 如果项目中存在代码文件且所有代码文件的总行数均为奇数，返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"为了照顾不同水平的玩家，给控制面板增加一个“难度模式”的单选切换开关：分为“普通模式”和“困难模式”。普通模式下，游戏区中央的目标黑色阴影会以浅灰色线条勾勒出每个板块的内部辅助线；而在困难模式下，就是纯粹的黑色剪影。玩家在拼图过程中可以随时切换难度，目标图案的渲染需即时响应。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的生成代码时，请将各函数的行数严格控制在15行及以内的要求作废，更新为生成代码时，请将各函数的行数严格控制在 10 行及以内；第二，使用ESLint对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终ESLint的检查结果没有任何错误，该要求后续所有轮次均需严格遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能的代码文件及其相关文本文件的扩展名\\n target_extensions = {'.html', '.css', '.js', '.md', '.json', '.log'}\\n\\n # 如果工作区路径不存在，说明没有生成任何文件，不违反UTF-8编码要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历工作区所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出符合目标后缀的文件\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出异常，说明文件不是有效的 UTF-8 编码，不符合要求\\n return False\\n except Exception:\\n # 如果有其他读取错误（如权限问题等）暂不作为编码错误处理\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n workspace = Path(workspace_path).resolve()\\n \\n # 检查工作空间路径是否存在且为目录\\n if not workspace.exists() or not workspace.is_dir():\\n return False\\n \\n for root, dirs, files in os.walk(workspace):\\n # 忽略隐藏目录和常见的 Python 缓存目录，避免不必要的误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n rel_path = Path(root).resolve().relative_to(workspace)\\n \\n # 将相对路径分解为各级组件\\n parts = [p for p in rel_path.parts if p and p != '.']\\n \\n # 若存在相对路径层级 >= 2 的目录（即子目录下的子目录），\\n # 则说明总目录层级 >= 3，不满足小于等于2层的要求\\n if len(parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_var_name(name: str) -> bool:\\n has_digit = any(c.isdigit() for c in name)\\n has_upper = any(c.isupper() for c in name)\\n has_lower = any(c.islower() for c in name)\\n has_under = '_' in name\\n return has_digit and has_upper and has_lower and has_under\\n\\ndef extract_variables(text: str) -> list:\\n vars_set = set()\\n \\n # 1. 提取标准 let, const, var\\n matches = re.findall(r'\\\\b(?:let|const|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', text)\\n vars_set.update(matches)\\n \\n # 2. 提取解构赋值中的变量\\n destruct_matches = re.findall(r'\\\\b(?:let|const|var)\\\\s*([\\\\[\\\\{][^\\\\]\\\\}]+[\\\\]\\\\}])', text)\\n for d_str in destruct_matches:\\n inner = d_str[1:-1]\\n for p in inner.split(','):\\n p = p.strip()\\n if not p: continue\\n if ':' in p: p = p.split(':')[1].strip()\\n if '=' in p: p = p.split('=')[0].strip()\\n m = re.match(r'^[\\\\[\\\\{]?\\\\s*(?:\\\\.\\\\.\\\\.)?\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\\n if m: vars_set.add(m.group(1))\\n\\n # 3. 提取函数名和类名\\n funcs = re.findall(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', text)\\n vars_set.update(funcs)\\n classes = re.findall(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', text)\\n vars_set.update(classes)\\n \\n # 4. 提取常规函数参数\\n func_params = re.findall(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', text)\\n for param_str in func_params:\\n for p in param_str.split(','):\\n p = p.strip()\\n if not p: continue\\n if ':' in p: p = p.split(':')[1].strip()\\n if '=' in p: p = p.split('=')[0].strip()\\n m = re.match(r'^[\\\\[\\\\{]?\\\\s*(?:\\\\.\\\\.\\\\.)?\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\\n if m: vars_set.add(m.group(1))\\n \\n # 5. 提取带括号的箭头函数参数\\n arrow_params = re.findall(r'\\\\(([^)]*)\\\\)\\\\s*=>', text)\\n for param_str in arrow_params:\\n for p in param_str.split(','):\\n p = p.strip()\\n if not p: continue\\n if ':' in p: p = p.split(':')[1].strip()\\n if '=' in p: p = p.split('=')[0].strip()\\n m = re.match(r'^[\\\\[\\\\{]?\\\\s*(?:\\\\.\\\\.\\\\.)?\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)', p)\\n if m: vars_set.add(m.group(1))\\n \\n # 6. 提取单参数箭头函数的参数\\n single_arrows = re.findall(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', text)\\n vars_set.update(single_arrows)\\n \\n # 7. 提取 catch 块中的错误对象\\n catch_blocks = re.findall(r'\\\\bcatch\\\\s*\\\\(\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\)', text)\\n vars_set.update(catch_blocks)\\n \\n return list(vars_set)\\n\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n all_vars = set()\\n has_files = False\\n \\n # 预编译正则，用于清除多行注释、单行注释及各种字符串\\n clean_pattern = re.compile(\\n r'/\\\\*.*?\\\\*/' \\n r'|//[^\\\\n]*' \\n r'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"' \\n r'|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'' \\n r'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`', \\n re.DOTALL\\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html') or file.endswith('.css'):\\n has_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file.endswith('.html'):\\n # 提取并解析 JS 代码块\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n script = clean_pattern.sub('', script)\\n all_vars.update(extract_variables(script))\\n \\n # 提取并解析 CSS 代码块中的变量\\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, re.IGNORECASE | re.DOTALL)\\n for style in styles:\\n style = clean_pattern.sub('', style)\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', style)\\n all_vars.update(css_vars)\\n \\n elif file.endswith('.js'):\\n content = clean_pattern.sub('', content)\\n all_vars.update(extract_variables(content))\\n \\n elif file.endswith('.css'):\\n content = clean_pattern.sub('', content)\\n css_vars = re.findall(r'--([a-zA-Z0-9_-]+)', content)\\n all_vars.update(css_vars)\\n \\n except Exception:\\n continue\\n \\n # 如果项目中没有任何代码文件，或者没有声明任何变量，则未满足约束。\\n if not has_files or not all_vars:\\n return False\\n \\n # 检查所有被提取出的变量名\\n for var in all_vars:\\n if not check_var_name(var):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义属于“代码文件”的常规扩展名集合（排除 .md, .json, .log 等非代码文件）\\n code_exts = {\\n '.js', '.mjs', '.cjs', '.ts', '.tsx', '.jsx', \\n '.html', '.htm', '.css', '.scss', '.less', '.sass',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 提取文件的扩展名\\n _, ext = os.path.splitext(file_name)\\n \\n # 判断是否为代码文件\\n if ext.lower() in code_exts:\\n found_code_file = True\\n \\n # 去除扩展名的 \\\".\\\" 后，对整个文件名进行检查\\n # 这保证了即使扩展名中含有大写（如 .JS）或主文件命中含有数字（如 index1），也会被成功拦截\\n name_check = file_name.replace('.', '')\\n \\n # 检查是否仅由小写字母和下划线构成\\n if not re.match(r'^[a-z_]+$', name_check):\\n return False\\n \\n return found_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写的代码文件中，务必优先挑选反斜杠 \\\\ 来表示路径分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在 10 行及以内\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def remove_comments_and_strings(code):\\n state = 'NORMAL'\\n result = []\\n i = 0\\n while i < len(code):\\n if state == 'NORMAL':\\n if code[i:i+2] == '//':\\n state = 'LINE_COMMENT'\\n result.append(' ')\\n i += 2\\n elif code[i:i+2] == '/*':\\n state = 'BLOCK_COMMENT'\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n state = 'DQUOTE'\\n result.append(' ')\\n i += 1\\n elif code[i] == \\\"'\\\":\\n state = 'SQUOTE'\\n result.append(' ')\\n i += 1\\n elif code[i] == '`':\\n state = 'BQUOTE'\\n result.append(' ')\\n i += 1\\n else:\\n result.append(code[i])\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state = 'NORMAL'\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if code[i:i+2] == '*/':\\n state = 'NORMAL'\\n result.append(' ')\\n i += 2\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'DQUOTE':\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n if i + 1 < len(code):\\n if code[i+1] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n state = 'NORMAL'\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'SQUOTE':\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n if i + 1 < len(code):\\n if code[i+1] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif code[i] == \\\"'\\\":\\n state = 'NORMAL'\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'BQUOTE':\\n if code[i] == '\\\\\\\\':\\n result.append(' ')\\n if i + 1 < len(code):\\n if code[i+1] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif code[i] == '`':\\n state = 'NORMAL'\\n result.append(' ')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def check_script_functions(code_str):\\n clean_code = remove_comments_and_strings(code_str)\\n lines = clean_code.split('\\\\n')\\n \\n i = 0\\n while i < len(lines):\\n line = lines[i]\\n \\n is_function = False\\n # 匹配常规 function\\n if re.search(r'\\\\bfunction\\\\b', line):\\n is_function = True\\n # 匹配箭头函数\\n elif re.search(r'=>\\\\s*\\\\{', line):\\n is_function = True\\n # 匹配类方法或对象方法 (忽略控制流关键字)\\n elif re.search(r'^\\\\s*(?:async\\\\s+)?(?!(?:if|for|while|switch|catch|return)\\\\b)[a-zA-Z_$][\\\\w$]*\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', line):\\n is_function = True\\n \\n if is_function:\\n brace_count = 0\\n started = False\\n start_line = i\\n end_line = i\\n \\n for j in range(i, len(lines)):\\n for char in lines[j]:\\n if char == '{':\\n brace_count += 1\\n started = True\\n elif char == '}':\\n if started:\\n brace_count -= 1\\n if brace_count <= 0:\\n break\\n if started and brace_count <= 0:\\n end_line = j\\n break\\n \\n # 如果成功匹配闭合大括号，校验函数行数\\n if started and brace_count == 0:\\n func_length = end_line - start_line + 1\\n if func_length > 10:\\n return False\\n i += 1\\n return True\\n\\n def check_python_functions(code_str):\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno'):\\n length = node.end_lineno - node.lineno + 1\\n if length > 10:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_script_functions(content):\\n return False\\n except Exception:\\n pass\\n \\n elif file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_python_functions(content):\\n return False\\n except Exception:\\n pass\\n \\n elif file.endswith('.html') or file.endswith('.htm'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 提取内联 script 脚本进行验证\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_script_functions(script):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n\\n code_extensions = ('.html', '.css', '.js')\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅推断 HTML、CSS、JS 作为必须检查的代码文件类型\\n if file.endswith(code_extensions):\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 指令中要求使用 UTF-8 编码，因此以该编码读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取文件（例如非法字符或权限错误），则视为不满足\\n return False\\n\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"后续所有轮次生成的每个代码文件，算上所有的空行的总行数必须保持为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对前端七巧板项目推测代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 按换行符进行分割，可以准确统计出算上空行在内的文本总行数\\n lines = content.split('\\\\n')\\n \\n # 判断总行数是否为偶数\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果遇到文件读取异常（例如不可读或非规范的UTF-8），保守处理返回False\\n return False\\n \\n # 如果至少存在一个代码文件且遍历结束无任何文件违背奇数行数原则，返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"使用ESLint对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终ESLint的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 9,\n \"instruction\": \"最后一轮，我们来提升UI的质感和物理拟真视觉体验。请用CSS渐变或SVG滤镜给这7块七巧板加上逼真的木纹纹理，并为它们添加投影（box-shadow/drop-shadow）。当板块被鼠标按住拖拽时，投影变得更模糊且偏移变大，模拟被从桌面上“提起来”的3D视觉效果；松开放下时，投影恢复紧贴桌面的状态。让整体游戏看起来像一个高级的实体桌面玩具。另外请你注意，后续所有轮次的开发都需要遵循两个新要求：第一，原有的要求生成的每个代码文件算上所有空行总行数必须为奇数的规则作废，更新为每个生成的代码文件总行数（含空行）必须是偶数，该要求需严格遵循；第二，在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 25s` 包装，这样可以防范进程挂起，该要求后续所有轮次均需严格遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你提供的每个代码文件都严格要求使用UTF-8作为编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推测代码文件及相关文本文件可能存在的后缀名\\n valid_extensions = {'.html', '.css', '.js', '.json', '.md', '.log', '.sh'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否是推断的代码或文本文件类型，并额外适配可能的 ESLint 配置文件\\n if ext in valid_extensions or file == '.eslintrc':\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用 UTF-8 解码二进制内容\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 处理诸如文件无法读取等其他异常（非要求重点，为防止执行崩溃均算作校验失败）\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你构建的代码工程目录层级必须小于或等于2层\",\n \"validation_code\": \"def check_directory_level(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作空间不存在，自然无法满足构建代码工程的前提\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的隐藏目录（如 .git, .vscode 等），防止对判定造成干扰\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n rel_path = os.path.relpath(root, workspace_path)\\n # 如果是根目录本身，跳过检查\\n if rel_path == '.':\\n continue\\n \\n # 统一不同操作系统下的路径分隔符\\n rel_path = rel_path.replace('\\\\\\\\', '/')\\n \\n # 将路径以 '/' 切分并过滤掉空字符串，从而获取相对目录层级深度\\n parts = [p for p in rel_path.split('/') if p]\\n \\n # 如果包含的层数大于 2，说明目录层级超出了要求\\n if len(parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"针对生成的代码文件，要求其中的每个变量名都必须同时包含数字、下划线、大写字母和小写字母\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_var_name(name):\\n has_digit = bool(re.search(r'\\\\d', name))\\n has_upper = bool(re.search(r'[A-Z]', name))\\n has_lower = bool(re.search(r'[a-z]', name))\\n has_underscore = '_' in name\\n return has_digit and has_upper and has_lower and has_underscore\\n\\n def extract_js_vars_regex(content):\\n # 移除字符串（处理包含转义字符的情况，防止字符串内部文本扰乱正则匹配）\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content, flags=re.DOTALL)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content, flags=re.DOTALL)\\n \\n # 移除单行及多行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n vars_found = set()\\n \\n # 1. 提取 let, const, var 声明的变量（状态机匹配以应对多变量声明等复杂情况）\\n for match in re.finditer(r'\\\\b(let|const|var)\\\\b', content):\\n pos = match.end()\\n state = 'expect_var'\\n current_var = ''\\n while pos < len(content):\\n char = content[pos]\\n if state == 'expect_var':\\n if char.isspace():\\n pass\\n elif char.isalpha() or char in '_$':\\n current_var = char\\n state = 'in_var'\\n elif char in '{[':\\n pass\\n else:\\n break\\n elif state == 'in_var':\\n if char.isalnum() or char in '_$':\\n current_var += char\\n else:\\n vars_found.add(current_var)\\n current_var = ''\\n if char.isspace():\\n state = 'after_var'\\n elif char == '=':\\n state = 'in_value'\\n elif char == ',':\\n state = 'expect_var'\\n elif char in '}]':\\n state = 'after_var'\\n elif char == ';':\\n break\\n else:\\n break\\n elif state == 'after_var':\\n if char.isspace():\\n pass\\n elif char == '=':\\n state = 'in_value'\\n elif char == ',':\\n state = 'expect_var'\\n elif char in '}]':\\n pass\\n elif char == ';':\\n break\\n else:\\n break\\n elif state == 'in_value':\\n if char == ';':\\n break\\n pos += 1\\n\\n # 2. 提取函数声明中的函数名及参数\\n func_pattern = re.compile(r'\\\\bfunction\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)')\\n for match in func_pattern.finditer(content):\\n if match.group(1):\\n vars_found.add(match.group(1))\\n args = match.group(2).split(',')\\n for arg in args:\\n arg = arg.strip().split('=')[0].strip()\\n if arg:\\n words = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg)\\n if words:\\n vars_found.add(words[0])\\n \\n # 3. 提取箭头函数参数\\n arrow_pattern = re.compile(r'(\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>')\\n for match in arrow_pattern.finditer(content):\\n args_str = match.group(1)\\n if args_str.startswith('('):\\n args_str = args_str[1:-1]\\n args = args_str.split(',')\\n for arg in args:\\n arg = arg.strip().split('=')[0].strip()\\n if arg:\\n words = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', arg)\\n if words:\\n vars_found.add(words[0])\\n \\n # 4. 提取 catch 捕获的异常变量名\\n catch_pattern = re.compile(r'\\\\bcatch\\\\s*\\\\(\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\)')\\n for match in catch_pattern.finditer(content):\\n vars_found.add(match.group(1))\\n \\n # 5. 提取 class 类名\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n for match in class_pattern.finditer(content):\\n vars_found.add(match.group(1))\\n\\n return list(vars_found)\\n\\n def extract_css_vars(content):\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n vars_found = set()\\n for match in re.finditer(r'--([a-zA-Z0-9_-]+)', content):\\n vars_found.add(match.group(0))\\n return list(vars_found)\\n\\n found_any = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 跳过历史备份文件\\n if 'backup' in file.lower() or file.endswith('.bak'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if file.endswith('.js'):\\n vars_found = extract_js_vars_regex(content)\\n for v in vars_found:\\n found_any = True\\n if not check_var_name(v):\\n return False\\n \\n elif file.endswith('.py'):\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n found_any = True\\n if not check_var_name(node.id):\\n return False\\n elif isinstance(node, ast.arg):\\n found_any = True\\n if not check_var_name(node.arg):\\n return False\\n except Exception:\\n pass\\n \\n elif file.endswith('.html'):\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n vars_found = extract_js_vars_regex(script)\\n for v in vars_found:\\n found_any = True\\n if not check_var_name(v):\\n return False\\n \\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.IGNORECASE | re.DOTALL)\\n for style in styles:\\n vars_found = extract_css_vars(style)\\n for v in vars_found:\\n found_any = True\\n if not check_var_name(v):\\n return False\\n \\n elif file.endswith('.css'):\\n vars_found = extract_css_vars(content)\\n for v in vars_found:\\n found_any = True\\n if not check_var_name(v):\\n return False\\n\\n # 若未找到任何变量定义视作暂未生成代码主体，或者不满足条件\\n return True if found_any else False\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"确保在每次修改项目之后，都对dev-progress.md文件进行更新，记录本次完成的事项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在3以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保在每次加入新内容后，更新 code_review.md 文件来对现有代码进行代码审查（Review），格式要求如下：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你真正动手更改代码之前，务必把准备进行的修改计划写入change_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件名中，除了小写字母和下划线外，不得出现任何其他符号或大写字母\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义属于“代码文件”的常见扩展名集合（针对前端七巧板项目以及其他常见开发语言）\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.mjs', '.cjs', '.ts', '.jsx', '.tsx', \\n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', \\n '.cs', '.go', '.rs', '.rb', '.php', '.swift', '.kt', '.m', '.sh'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）和常见的包管理器依赖目录（如 node_modules）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for file in files:\\n # 忽略隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n # 获取文件扩展名并转为小写进行匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断该文件是否属于代码文件\\n if ext in code_extensions:\\n # 剔除文件名中的点号 (.) \\n name_without_dots = file.replace('.', '')\\n \\n # 验证剩余部分是否完全由小写字母和下划线组成\\n # 如果包含数字、大写字母或其他符号，均会匹配失败并返回 False\\n if not re.fullmatch(r'[a-z_]+', name_without_dots):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"无论何种情况，生成的代码文件里的文件引用都得是相对路径，杜绝绝对路径的使用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写的代码文件中，务必优先挑选反斜杠 \\\\ 来表示路径分隔\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每次结束对生成代码的测试之后，必须对test_results.json文件作出更新，更新时需遵循List[Dict]的文件格式要求，每个元素所含的两个键值为：description（内容为本次测试内容），以及status（必须是\\\"pass\\\"、\\\"fail\\\"或\\\"error\\\"这三者之一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在 10 行及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_python_functions(code_str: str) -> bool:\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 判断当前环境是否支持 end_lineno (Python 3.8+)\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n line_count = node.end_lineno - node.lineno + 1\\n if line_count > 10:\\n return False\\n return True\\n except Exception:\\n return True\\n\\n def clean_js_code(code_str: str) -> str:\\n \\\"\\\"\\\"\\n 清空 JS 代码里的注释和字符串，但是严格保留换行符。\\n 这样可以在没有干扰的前提下正确匹配括号，并依靠换行符准确统计行号。\\n \\\"\\\"\\\"\\n result = []\\n i = 0\\n length = len(code_str)\\n in_block_comment = False\\n in_line_comment = False\\n in_string = False\\n string_char = ''\\n \\n while i < length:\\n c = code_str[i]\\n if in_block_comment:\\n if c == '*' and i + 1 < length and code_str[i+1] == '/':\\n in_block_comment = False\\n result.append(' ')\\n i += 2\\n continue\\n else:\\n result.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n continue\\n elif in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n elif in_string:\\n if c == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n continue\\n elif c == string_char:\\n in_string = False\\n result.append(' ')\\n else:\\n result.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n continue\\n \\n if c == '/' and i + 1 < length and code_str[i+1] == '*':\\n in_block_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n elif c == '/' and i + 1 < length and code_str[i+1] == '/':\\n in_line_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n elif c in (\\\"'\\\", '\\\"', '`'):\\n in_string = True\\n string_char = c\\n result.append(' ')\\n i += 1\\n continue\\n \\n result.append(c)\\n i += 1\\n \\n return \\\"\\\".join(result)\\n\\n def check_js_functions(code_str: str) -> bool:\\n clean_code = clean_js_code(code_str)\\n \\n brace_pairs = {}\\n stack = []\\n # 提取全部 { } 匹配对\\n for i, c in enumerate(clean_code):\\n if c == '{':\\n stack.append(i)\\n elif c == '}':\\n if stack:\\n start = stack.pop()\\n brace_pairs[start] = i\\n \\n for start_idx, end_idx in brace_pairs.items():\\n # 往回找语句终止符\\n search_start = start_idx - 1\\n while search_start >= 0 and clean_code[search_start] not in \\\"{};\\\":\\n search_start -= 1\\n \\n prefix = clean_code[search_start + 1 : start_idx].strip()\\n \\n # 如果包含控制流等非函数的关键字，直接跳过\\n if re.search(r'\\\\b(?:if|for|while|switch|catch|class|try|else)\\\\b', prefix):\\n continue\\n \\n is_func = False\\n # 常见函数关键字\\n if 'function' in prefix or '=>' in prefix:\\n is_func = True\\n # 对象/类内部方法声明形式检测（如: `methodName(args) {`）\\n elif re.search(r'\\\\b[\\\\w$]+\\\\s*\\\\([^)]*\\\\)$', prefix):\\n is_func = True\\n \\n if is_func:\\n abs_start_idx = start_idx\\n \\n feature_idx_func = prefix.rfind('function')\\n feature_idx_arrow = prefix.rfind('=>')\\n feature_idx_paren = prefix.rfind(')')\\n \\n # 寻找更精确的声明起点，防止漏算或将上一行代码错算入\\n if feature_idx_func != -1:\\n abs_start_idx = search_start + 1 + feature_idx_func\\n \\n if feature_idx_paren != -1:\\n abs_paren_idx = search_start + 1 + feature_idx_paren\\n paren_depth = 0\\n # 匹配函数参数对应的 \\\"(\\\"，以将其设为起点的基准\\n for j in range(abs_paren_idx, search_start, -1):\\n if clean_code[j] == ')':\\n paren_depth += 1\\n elif clean_code[j] == '(':\\n paren_depth -= 1\\n if paren_depth == 0:\\n current_abs = j\\n if feature_idx_func != -1 and (search_start + 1 + feature_idx_func) < j:\\n current_abs = search_start + 1 + feature_idx_func\\n if current_abs < abs_start_idx:\\n abs_start_idx = current_abs\\n break\\n else:\\n if (search_start + 1 + feature_idx_paren) < abs_start_idx:\\n abs_start_idx = search_start + 1 + feature_idx_paren\\n \\n if feature_idx_arrow != -1 and feature_idx_paren == -1:\\n if (search_start + 1 + feature_idx_arrow) < abs_start_idx:\\n abs_start_idx = search_start + 1 + feature_idx_arrow\\n \\n start_lineno = clean_code.count('\\\\n', 0, abs_start_idx) + 1\\n end_lineno = clean_code.count('\\\\n', 0, end_idx) + 1\\n \\n if end_lineno - start_lineno + 1 > 10:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 校验 Python 文件\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n if not check_python_functions(code_str):\\n return False\\n except Exception:\\n pass\\n # 校验 JavaScript / TypeScript 文件\\n elif file.endswith(('.js', '.jsx', '.ts', '.tsx')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_str = f.read()\\n if not check_js_functions(code_str):\\n return False\\n except Exception:\\n pass\\n # 校验 HTML 文件中被包裹在 <script> 中的 JS 代码\\n elif file.endswith(('.html', '.htm')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n scripts = re.findall(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for script in scripts:\\n if not check_js_functions(script):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"强制要求所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码文件扩展名，排除文档、配置和日志等非代码文件\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法读取或其他异常情况，视作未满足要求\\n return False\\n \\n # 如果遍历完毕没有发现任何代码文件，说明未生成有效的代码工程\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"项目每经历一次改动，都需要将其记录在新建的文件 change_log_X.log 中，编号 X 表示改动次数且从 0 算起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每个生成的代码文件总行数（含空行）必须是偶数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名，根据上下文主要包括前端等代码类型\\n code_extensions = {\\n '.js', '.css', '.html', '.py', '.ts', '.jsx', '.tsx',\\n '.vue', '.java', '.c', '.cpp', '.h', '.cs', '.go', \\n '.rs', '.rb', '.php', '.swift', '.kt'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照 '\\\\n' 分割可以准确计算出编辑器中的物理行数（包括末尾独立空行）\\n lines_count = len(content.split('\\\\n'))\\n \\n # 若存在奇数行数的文件，直接判定不满足要求\\n if lines_count % 2 != 0:\\n return False\\n except Exception:\\n # 如果由于某种原因（如非预期的二进制混入）无法用 UTF-8 读取则容错跳过\\n pass\\n \\n # 确保存在至少一个代码文件并且所有的代码文件行数皆为偶数\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"使用ESLint对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终ESLint的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 25s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面初始渲染，确认左侧为带明显边界的大的游戏操作区，右侧为控制面板，控制面板中包含标题“七巧板拼图”；并在底部或侧边的备选区，能够看到散落的7个带有鲜明不同颜色的七巧板图形（2个大直角三角、1个中直角三角、2个小直角三角、1个正方形、1个平行四边形）。","模拟鼠标或触摸屏按住并拖动任意七巧板，确认在拖动过程中该板块处于最顶层（不会被其它板块遮挡）；将其拖到游戏区内的新位置后松开，确认板块能停留在放下的位置；再尝试将其向页面边缘拖拽，验证板块受到边界限制，无法完全拖出游戏操作区。","使用鼠标悬浮于任意七巧板并滚动滚轮（或点击可能存在的旋转手柄），验证该板块会发生旋转，且每次旋转角度固定为45度（顺时针或逆时针均可）。","通过页面提供的翻转方式（例如双击）操作平行四边形板块，验证该板块发生水平或垂直方向的翻转变化（镜像翻转），以满足平面旋转无法覆盖的拼图形态。","拖拽一个七巧板缓慢靠近另一个静止的七巧板，当两者的顶点或边缘距离很近（大约15px内）时，验证拖拽的七巧板会自动发生位置吸附并紧密对齐贴合到另一板块上。","点击右侧控制面板上的关卡按钮（如狐狸、帆船或天鹅），验证游戏区正中间能够即时渲染出对应形状的黑色目标图案阴影轮廓。","在控制面板中将难度切换开关置于'普通模式'，验证游戏区中央的黑色目标阴影图案上出现浅灰色的内部辅助线条（勾勒出每块的轮廓）；将其切换至'困难模式'，验证辅助线消失，图案变回无内部线条的纯黑剪影。","刷新页面并点击任意关卡，观察控制面板上的计时器应显示为'00:00'。用鼠标选中并开始拖动第一块七巧板，验证拖动行为触发后，计时器立即开始正常的走秒计时（显示格式如mm:ss）。","在板块被移动至各处且计时器正在走秒的情况下，点击控制面板上的'重置'按钮，验证所有的七巧板带有过渡动画飞回到初始备选区位置，同时计时器立即归零重置为'00:00'。","观察七巧板表面，确认其应用了木纹质感纹理以及基础的下拉阴影。用鼠标按下并拖拽其中一块，验证在按住拖拽的状态下该板块的阴影偏移量变大且变模糊（模拟被从桌面提起的3D视觉），松开鼠标放置后阴影立即恢复成紧贴桌面的初始原状。","进行完整通关模拟，将7个板块依次拖入中央目标黑色阴影轮廓内，调整位置和角度直至完全匹配。验证在所有板块均放置正确的情况下，页面会弹出'恭喜通关'的模态框，并且屏幕上自动播放彩纸飘落或烟花的庆祝动画。","调整浏览器窗口宽度或在不同屏幕尺寸下查看页面，确认游戏操作区、控制面板和备选区能够保持基本可用的响应式布局，不出现主要内容严重溢出或不可操作的问题。"],"string":"[\n \"查看页面初始渲染，确认左侧为带明显边界的大的游戏操作区，右侧为控制面板，控制面板中包含标题“七巧板拼图”；并在底部或侧边的备选区，能够看到散落的7个带有鲜明不同颜色的七巧板图形（2个大直角三角、1个中直角三角、2个小直角三角、1个正方形、1个平行四边形）。\",\n \"模拟鼠标或触摸屏按住并拖动任意七巧板，确认在拖动过程中该板块处于最顶层（不会被其它板块遮挡）；将其拖到游戏区内的新位置后松开，确认板块能停留在放下的位置；再尝试将其向页面边缘拖拽，验证板块受到边界限制，无法完全拖出游戏操作区。\",\n \"使用鼠标悬浮于任意七巧板并滚动滚轮（或点击可能存在的旋转手柄），验证该板块会发生旋转，且每次旋转角度固定为45度（顺时针或逆时针均可）。\",\n \"通过页面提供的翻转方式（例如双击）操作平行四边形板块，验证该板块发生水平或垂直方向的翻转变化（镜像翻转），以满足平面旋转无法覆盖的拼图形态。\",\n \"拖拽一个七巧板缓慢靠近另一个静止的七巧板，当两者的顶点或边缘距离很近（大约15px内）时，验证拖拽的七巧板会自动发生位置吸附并紧密对齐贴合到另一板块上。\",\n \"点击右侧控制面板上的关卡按钮（如狐狸、帆船或天鹅），验证游戏区正中间能够即时渲染出对应形状的黑色目标图案阴影轮廓。\",\n \"在控制面板中将难度切换开关置于'普通模式'，验证游戏区中央的黑色目标阴影图案上出现浅灰色的内部辅助线条（勾勒出每块的轮廓）；将其切换至'困难模式'，验证辅助线消失，图案变回无内部线条的纯黑剪影。\",\n \"刷新页面并点击任意关卡，观察控制面板上的计时器应显示为'00:00'。用鼠标选中并开始拖动第一块七巧板，验证拖动行为触发后，计时器立即开始正常的走秒计时（显示格式如mm:ss）。\",\n \"在板块被移动至各处且计时器正在走秒的情况下，点击控制面板上的'重置'按钮，验证所有的七巧板带有过渡动画飞回到初始备选区位置，同时计时器立即归零重置为'00:00'。\",\n \"观察七巧板表面，确认其应用了木纹质感纹理以及基础的下拉阴影。用鼠标按下并拖拽其中一块，验证在按住拖拽的状态下该板块的阴影偏移量变大且变模糊（模拟被从桌面提起的3D视觉），松开鼠标放置后阴影立即恢复成紧贴桌面的初始原状。\",\n \"进行完整通关模拟，将7个板块依次拖入中央目标黑色阴影轮廓内，调整位置和角度直至完全匹配。验证在所有板块均放置正确的情况下，页面会弹出'恭喜通关'的模态框，并且屏幕上自动播放彩纸飘落或烟花的庆祝动画。\",\n \"调整浏览器窗口宽度或在不同屏幕尺寸下查看页面，确认游戏操作区、控制面板和备选区能够保持基本可用的响应式布局，不出现主要内容严重溢出或不可操作的问题。\"\n]","isConversational":false}}},{"rowIdx":73,"cells":{"task_id":{"kind":"number","value":74,"string":"74"},"repository_policy":{"kind":"string","value":"你在后续所有轮次的交互中必须始终严格遵循以下约束要求：\n1. 严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具。\n2. 在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令。\n3. 在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令。\n4. 对于你输出的每一个代码文件，请将其中的注释全部写成中文。\n5. 所有由你创建的代码文件，其编码必须统一为UTF-8。\n6. 生成的代码文件应全部使用 4 个空格来控制缩进。\n7. 请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求。\n8. 必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于你输出的每一个代码文件，请将其中的注释全部写成中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts',\n '.html', '.htm', '.css', '.php', '.rb', '.go', '.rs', '.swift', \n '.sh', '.bash', '.json', '.xml', '.yaml', '.yml', '.sql', '.kt', \n '.scala', '.m', '.vue', '.jsx', '.tsx'\n }\n\n if not os.path.exists(workspace_path):\n # 如果工作区不存在，说明并未违背编码统一的要求\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用 UTF-8 解码，若非 UTF-8 编码且含有非 ASCII 字符将引发 UnicodeDecodeError\n content.decode('utf-8')\n except UnicodeDecodeError:\n return False\n except Exception:\n # 遇到文件无法读取等其他异常，可跳过，继续检测其他文件\n continue\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义需要检查的常见代码文件扩展名\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', \n '.rs', '.swift', '.kt', '.m', '.sh', '.bash', '.html', '.css', '.vue'\n }\n \n has_code_file = False\n \n # 1. 遍历工作区内所有的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n tab_count = 0\n valid_space = 0\n invalid_space = 0\n \n for line in lines:\n # 匹配每行前导的空格或制表符，以及紧随其后的第一个非空字符\n m = re.match(r'^([ \\t]+)(\\S)', line)\n if m:\n indent = m.group(1)\n first_char = m.group(2)\n \n # 忽略可能是多行注释对齐的行\n if first_char == '*':\n continue\n \n if '\\t' in indent:\n tab_count += 1\n else:\n if len(indent) % 4 == 0:\n valid_space += 1\n else:\n invalid_space += 1\n \n # 一旦包含 tab 作为缩进，直接违反\n if tab_count > 0:\n return False\n \n # 判断非4倍数空格缩进的行数比例，如果过高，说明没有严格使用4空格缩进\n total_space_indents = valid_space + invalid_space\n if total_space_indents > 0:\n if invalid_space / total_space_indents > 0.3:\n return False\n\n # 2. 如果工作区没有生成任何代码文件，则回退检查 Response 中的代码块\n if not has_code_file:\n code_blocks = re.findall(r'```[ \\t]*(\\w*)\\n(.*?)```', response, re.DOTALL)\n # 允许的代码块语言标识\n allowed_langs = [\n 'python', 'py', 'java', 'c', 'cpp', 'c++', 'javascript', 'js', \n 'typescript', 'ts', 'go', 'ruby', 'rb', 'php', 'csharp', 'cs', \n 'html', 'css', 'bash', 'sh'\n ]\n \n for lang, code in code_blocks:\n lang = lang.lower()\n if lang in allowed_langs:\n tab_count = 0\n valid_space = 0\n invalid_space = 0\n \n for line in code.split('\\n'):\n m = re.match(r'^([ \\t]+)(\\S)', line)\n if m:\n indent = m.group(1)\n first_char = m.group(2)\n \n if first_char == '*':\n continue\n \n if '\\t' in indent:\n tab_count += 1\n else:\n if len(indent) % 4 == 0:\n valid_space += 1\n else:\n invalid_space += 1\n \n if tab_count > 0:\n return False\n \n total_space_indents = valid_space + invalid_space\n if total_space_indents > 0:\n if invalid_space / total_space_indents > 0.3:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n \n def is_pascal_case(name: str) -> bool:\n # Python 内置的特殊方法/魔法方法（如 __init__）受语言规范强制，应予以放行\n if name.startswith('__') and name.endswith('__'):\n return True\n # 常见程序入口方法（如 C/C++/Java/Go 中的 main）予以放行\n if name == 'main':\n return True\n # 大驼峰规则检查：首字母大写，仅包含字母数字，无下划线\n return bool(re.match(r'^[A-Z][a-zA-Z0-9]*$', name))\n \n if not os.path.exists(workspace_path):\n return False\n \n # 需要过滤的控制流关键词，防止在 Java/C/C++ 正则中将其错误捕获为函数名\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'return'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n funcs_to_check = []\n \n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n funcs_to_check.append(node.name)\n except Exception:\n pass\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n # 匹配 function 关键字声明的函数\n funcs = re.findall(r'\\bfunction\\s+([a-zA-Z0-9_]+)\\s*\\(', content)\n funcs_to_check.extend(funcs)\n # 匹配 ES6 箭头函数声明的变量名\n arrow_funcs = re.findall(r'(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z0-9_]+)\\s*=>', content)\n funcs_to_check.extend(arrow_funcs)\n elif ext in ['.java', '.cpp', '.c', '.h', '.hpp', '.cs']:\n # 匹配 C/C++/Java/C# 静态或常规类型的函数\n funcs = re.findall(r'^[ \\t]*(?:(?:public|private|protected|static|virtual|inline|const|explicit|override|async)\\s+)*[A-Za-z0-9_<>\\[\\]]+\\s+([a-zA-Z0-9_]+)\\s*\\([^)]*\\)\\s*(?:const|override)?\\s*\\{', content, re.MULTILINE)\n funcs_to_check.extend([f for f in funcs if f not in keywords])\n elif ext in ['.go']:\n # 匹配 Go 语言的函数\n funcs = re.findall(r'\\bfunc\\s+(?:\\[[^\\]]*\\]\\s+)?(?:\\([^)]+\\)\\s+)?([a-zA-Z0-9_]+)\\s*\\(', content)\n funcs_to_check.extend(funcs)\n elif ext in ['.sh', '.bash'] or (not ext and content.startswith('#!/bin/')):\n # 匹配 Bash 脚本中的函数\n funcs = re.findall(r'(?:^|\\n)[ \\t]*(?:function\\s+([a-zA-Z0-9_]+)|([a-zA-Z0-9_]+)\\s*\\(\\))', content)\n for f1, f2 in funcs:\n funcs_to_check.append(f1 if f1 else f2)\n \n # 对所有成功提取的该代码文件中的函数名进行大驼峰检查\n for func in funcs_to_check:\n if not is_pascal_case(func):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]}],"string":"[\n {\n \"constraint\": \"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请将其中的注释全部写成中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts',\\n '.html', '.htm', '.css', '.php', '.rb', '.go', '.rs', '.swift', \\n '.sh', '.bash', '.json', '.xml', '.yaml', '.yml', '.sql', '.kt', \\n '.scala', '.m', '.vue', '.jsx', '.tsx'\\n }\\n\\n if not os.path.exists(workspace_path):\\n # 如果工作区不存在，说明并未违背编码统一的要求\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用 UTF-8 解码，若非 UTF-8 编码且含有非 ASCII 字符将引发 UnicodeDecodeError\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n return False\\n except Exception:\\n # 遇到文件无法读取等其他异常，可跳过，继续检测其他文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义需要检查的常见代码文件扩展名\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', \\n '.rs', '.swift', '.kt', '.m', '.sh', '.bash', '.html', '.css', '.vue'\\n }\\n \\n has_code_file = False\\n \\n # 1. 遍历工作区内所有的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n tab_count = 0\\n valid_space = 0\\n invalid_space = 0\\n \\n for line in lines:\\n # 匹配每行前导的空格或制表符，以及紧随其后的第一个非空字符\\n m = re.match(r'^([ \\\\t]+)(\\\\S)', line)\\n if m:\\n indent = m.group(1)\\n first_char = m.group(2)\\n \\n # 忽略可能是多行注释对齐的行\\n if first_char == '*':\\n continue\\n \\n if '\\\\t' in indent:\\n tab_count += 1\\n else:\\n if len(indent) % 4 == 0:\\n valid_space += 1\\n else:\\n invalid_space += 1\\n \\n # 一旦包含 tab 作为缩进，直接违反\\n if tab_count > 0:\\n return False\\n \\n # 判断非4倍数空格缩进的行数比例，如果过高，说明没有严格使用4空格缩进\\n total_space_indents = valid_space + invalid_space\\n if total_space_indents > 0:\\n if invalid_space / total_space_indents > 0.3:\\n return False\\n\\n # 2. 如果工作区没有生成任何代码文件，则回退检查 Response 中的代码块\\n if not has_code_file:\\n code_blocks = re.findall(r'```[ \\\\t]*(\\\\w*)\\\\n(.*?)```', response, re.DOTALL)\\n # 允许的代码块语言标识\\n allowed_langs = [\\n 'python', 'py', 'java', 'c', 'cpp', 'c++', 'javascript', 'js', \\n 'typescript', 'ts', 'go', 'ruby', 'rb', 'php', 'csharp', 'cs', \\n 'html', 'css', 'bash', 'sh'\\n ]\\n \\n for lang, code in code_blocks:\\n lang = lang.lower()\\n if lang in allowed_langs:\\n tab_count = 0\\n valid_space = 0\\n invalid_space = 0\\n \\n for line in code.split('\\\\n'):\\n m = re.match(r'^([ \\\\t]+)(\\\\S)', line)\\n if m:\\n indent = m.group(1)\\n first_char = m.group(2)\\n \\n if first_char == '*':\\n continue\\n \\n if '\\\\t' in indent:\\n tab_count += 1\\n else:\\n if len(indent) % 4 == 0:\\n valid_space += 1\\n else:\\n invalid_space += 1\\n \\n if tab_count > 0:\\n return False\\n \\n total_space_indents = valid_space + invalid_space\\n if total_space_indents > 0:\\n if invalid_space / total_space_indents > 0.3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n \\n def is_pascal_case(name: str) -> bool:\\n # Python 内置的特殊方法/魔法方法（如 __init__）受语言规范强制，应予以放行\\n if name.startswith('__') and name.endswith('__'):\\n return True\\n # 常见程序入口方法（如 C/C++/Java/Go 中的 main）予以放行\\n if name == 'main':\\n return True\\n # 大驼峰规则检查：首字母大写，仅包含字母数字，无下划线\\n return bool(re.match(r'^[A-Z][a-zA-Z0-9]*$', name))\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 需要过滤的控制流关键词，防止在 Java/C/C++ 正则中将其错误捕获为函数名\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'return'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n funcs_to_check = []\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n funcs_to_check.append(node.name)\\n except Exception:\\n pass\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n # 匹配 function 关键字声明的函数\\n funcs = re.findall(r'\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(', content)\\n funcs_to_check.extend(funcs)\\n # 匹配 ES6 箭头函数声明的变量名\\n arrow_funcs = re.findall(r'(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z0-9_]+)\\\\s*=>', content)\\n funcs_to_check.extend(arrow_funcs)\\n elif ext in ['.java', '.cpp', '.c', '.h', '.hpp', '.cs']:\\n # 匹配 C/C++/Java/C# 静态或常规类型的函数\\n funcs = re.findall(r'^[ \\\\t]*(?:(?:public|private|protected|static|virtual|inline|const|explicit|override|async)\\\\s+)*[A-Za-z0-9_<>\\\\[\\\\]]+\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*(?:const|override)?\\\\s*\\\\{', content, re.MULTILINE)\\n funcs_to_check.extend([f for f in funcs if f not in keywords])\\n elif ext in ['.go']:\\n # 匹配 Go 语言的函数\\n funcs = re.findall(r'\\\\bfunc\\\\s+(?:\\\\[[^\\\\]]*\\\\]\\\\s+)?(?:\\\\([^)]+\\\\)\\\\s+)?([a-zA-Z0-9_]+)\\\\s*\\\\(', content)\\n funcs_to_check.extend(funcs)\\n elif ext in ['.sh', '.bash'] or (not ext and content.startswith('#!/bin/')):\\n # 匹配 Bash 脚本中的函数\\n funcs = re.findall(r'(?:^|\\\\n)[ \\\\t]*(?:function\\\\s+([a-zA-Z0-9_]+)|([a-zA-Z0-9_]+)\\\\s*\\\\(\\\\))', content)\\n for f1, f2 in funcs:\\n funcs_to_check.append(f1 if f1 else f2)\\n \\n # 对所有成功提取的该代码文件中的函数名进行大驼峰检查\\n for func in funcs_to_check:\\n if not is_pascal_case(func):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个基于游戏段位的行为数据表格的基础页面框架。行需要表示不同的游戏段位（比如青铜、白银、黄金、铂金、钻石、大师、王者），列表示不同的行为数据指标（比如胜率、KDA、场均输出、参团率、视野得分）。现阶段只需要把表格的静态结构搭出来，页面顶部加一个大标题即可。请用一些模拟数据填充表格内容，暂时不需要任何特殊颜色或图表，只要求表格样式干净整洁，并留好后续加功能的代码结构扩展点。请你后续所有操作始终遵循以下要求：你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过；执行你编写的单元测试文件时，目标代码的语句覆盖率要求达到90%或以上。","instruction_following_checklist":[{"constraint":"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于你输出的每一个代码文件，请将其中的注释全部写成中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据前端项目的常见特性，定义可能会创建的代码文件后缀名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'}\n\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的第三方包目录和版本控制目录\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'coverage'}\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明存在非 utf-8 编码（例如包含非ASCII字符的GBK编码文件等）\n return False\n except Exception:\n # 其他读取异常，暂时视作不合规文件\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能的前端及配置相关代码文件扩展名\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less', '.json'}\n exclude_dirs = {'node_modules', '.git', 'dist', 'build'}\n has_code_file = False\n \n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略第三方库及构建产物等不属于“生成的代码文件”的目录\n root_parts = set(root.replace('\\\\', '/').split('/'))\n if root_parts & exclude_dirs:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 若文件因编码等问题无法读取，则跳过\n continue\n \n has_code_file = True\n valid_indent_lines = 0\n invalid_indent_lines = 0\n \n for line in lines:\n stripped_line = line.lstrip()\n # 忽略空行\n if not stripped_line:\n continue\n \n # 提取前导空白部分\n leading_whitespaces = line[:len(line) - len(stripped_line)]\n \n # 如果出现 Tab 字符，直接认为违规\n if '\\t' in leading_whitespaces:\n return False\n \n # 忽略类似 JsDoc 的块注释内部行（由于通常带有额外的 1 空格对齐 '*' 号）\n if stripped_line.startswith('*'):\n continue\n \n space_count = len(leading_whitespaces)\n if space_count > 0:\n # 检查缩进空格数是否为 4 的倍数\n if space_count % 4 == 0:\n valid_indent_lines += 1\n else:\n invalid_indent_lines += 1\n \n total_indented_lines = valid_indent_lines + invalid_indent_lines\n \n if total_indented_lines > 0:\n # 容错处理：允许极少量的换行参数对齐导致非 4 的倍数缩进，\n # 但如果不符合行数的比例大于 5% 并且总数超过 2 行，即认为使用了非 4 空格的缩进规范。\n if invalid_indent_lines > 2 and (invalid_indent_lines / total_indented_lines) > 0.05:\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义用于匹配各种函数定义并提取函数名称的正则表达式\n func_patterns = [\n # 匹配常规的 JS/TS 函数: function MyFunc(...) \n re.compile(r\"\\bfunction\\s+([a-zA-Z0-9_]+)\\s*\\(\"),\n # 匹配箭头函数赋值: const MyFunc = (...) => { ... } 或 const MyFunc = async e => { ... }\n re.compile(r\"\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z0-9_]+)\\s*=>\"),\n # 匹配匿名函数赋值: const MyFunc = function(...) { ... }\n re.compile(r\"\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s*)?function\\s*\\(\"),\n # 匹配 Python 函数: def MyFunc(...)\n re.compile(r\"\\bdef\\s+([a-zA-Z0-9_]+)\\s*\\(\")\n ]\n \n # 大驼峰命名正则: 必须以大写字母开头，后续仅包含字母和数字，不允许有下划线\n pascal_case_pattern = re.compile(r\"^[A-Z][a-zA-Z0-9]*$\")\n \n # 根据用户提示的 ESLint，主要针对前端相关代码文件进行扫描\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除依赖库和版本控制文件夹\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n funcs = []\n # 运用多个正则提取文件中定义的所有函数名\n for pattern in func_patterns:\n funcs.extend(pattern.findall(content))\n \n # 判断每个函数名是否为大驼峰格式\n for func_name in funcs:\n if not pascal_case_pattern.match(func_name):\n return False\n except Exception:\n # 忽略无法正常读取或非文本类型的文件\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"执行你编写的单元测试文件时，目标代码的语句覆盖率要求达到90%或以上","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":1,"instruction":"非常好，现在我们要加上核心的“热力图”视觉效果。请根据每列（即每一项行为指标）的数据分布，给表格的具体数值单元格加上渐变背景色。要求：针对每一列的数据单独计算极值，数值越高的单元格背景色偏向翠绿色，数值越低偏向珊瑚红色，居中的数值呈浅黄色或白色。确保代码是根据列的最大最小值动态计算色阶的，从而能让用户一眼看出每个段位在某项指标上的相对表现。\n请你后续所有操作始终遵循以下更新后的要求：执行你编写的单元测试文件时，目标代码的分支覆盖率要求达到95%或以上；同时每次调整项目后，务必更新progress.md文件，以便记录当前这轮完成的任务内容。","instruction_following_checklist":[{"constraint":"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于你输出的每一个代码文件，请将其中的注释全部写成中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n target_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n return False\n except Exception:\n # 忽略非编码相关的读取错误（如权限问题）\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def check_indentation(response: str, workspace_path: str) -> bool:\n import os\n\n target_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css'}\n checked_any = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in target_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n stripped = line.lstrip()\n if not stripped or stripped.startswith('*') or stripped.startswith('/*'):\n continue\n \n leading_space = line[:len(line) - len(stripped)]\n \n if '\\t' in leading_space or len(leading_space) % 4 != 0:\n return False\n checked_any = True\n except Exception:\n pass\n\n return checked_any","tags":["Style","Layout"]},{"constraint":"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能的代码文件扩展名，以 JS/TS 及前端相关为主，兼容 Python\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py'}\n # 过滤掉无关依赖及构建输出目录\n skip_dirs = {'node_modules', 'dist', 'build', 'coverage', '.git', '__pycache__', '.vscode', '.idea'}\n\n # 1. Python 中的函数定义\n py_func_pattern = re.compile(r'\\bdef\\s+([a-zA-Z_]\\w*)\\s*\\(')\n # 2. JS/TS 中的 function 函数定义\n js_func_pattern = re.compile(r'\\bfunction\\s+([a-zA-Z_]\\w*)\\s*\\(')\n # 3. JS/TS 箭头函数定义，处理跨行和类型注释，利用 [^=;]* 防止跨语句越界\n js_arrow_pattern = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_]\\w*)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z_]\\w*)[^=;]*=>')\n # 4. JS/TS 普通函数表达式\n js_expr_pattern = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_]\\w*)\\s*=\\s*(?:async\\s*)?function\\b')\n # 5. 类方法或对象内的方法简写，利用 [^{;]* 防止跨语句错误连带\n method_pattern = re.compile(r'\\b([a-zA-Z_]\\w*)\\s*\\([^)]*\\)[^{;]*\\{')\n\n # 大驼峰命名校验正则：必须全是大写字母开头，跟随字母和数字，不能有下划线\n pascal_case_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\n \n # 避免被当成函数名的方法正则干扰关键字\n keywords = {\n 'if', 'while', 'for', 'switch', 'catch', 'function', 'return', 'super', 'constructor', \n 'else', 'try', 'finally', 'with', 'typeof', 'instanceof', 'new', 'delete', 'void', 'yield', \n 'await', 'get', 'set', 'import', 'export', 'default'\n }\n\n def clean_code(code: str, ext: str) -> str:\n \"\"\"用于消除注释和字符串字面量的干扰内容\"\"\"\n if ext == '.py':\n code = re.sub(r'#.*', '', code)\n code = re.sub(r'\"\"\"(?:\\\\.|[^\\\\])*?\"\"\"', '', code, flags=re.DOTALL)\n code = re.sub(r\"'''(?:\\\\.|[^\\\\])*?'''\", '', code, flags=re.DOTALL)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', code)\n else:\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', code, flags=re.DOTALL)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', code)\n return code\n\n for root, dirs, files in os.walk(workspace_path):\n dirs[:] = [d for d in dirs if d not in skip_dirs]\n for file in files:\n ext = os.path.splitext(file)[1]\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n raw_content = f.read()\n except Exception:\n continue\n \n content = clean_code(raw_content, ext)\n \n functions = []\n if ext == '.py':\n functions.extend(py_func_pattern.findall(content))\n else:\n functions.extend(js_func_pattern.findall(content))\n functions.extend(js_arrow_pattern.findall(content))\n functions.extend(js_expr_pattern.findall(content))\n methods = method_pattern.findall(content)\n for m in methods:\n if m not in keywords:\n functions.append(m)\n\n # 逐个检查识别出的函数名\n for func in functions:\n if not func:\n continue\n if not pascal_case_pattern.match(func):\n # 发现任何一个并非大驼峰的函数即判定失败\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"执行你编写的单元测试文件时，目标代码的分支覆盖率要求达到95%或以上","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次调整项目后，务必更新progress.md文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":2,"instruction":"视觉上直观多了！接下来增加表格的交互性。第一，请实现点击表头可以按该列进行升序或降序排序的功能，排序时行数据要整体移动，并且热力图的颜色也要保持对应正确。第二，当鼠标悬停在热力图单元格上时，展示一个悬浮提示框（Tooltip），里面要清晰地拼接出该项数据的解读，例如显示：“钻石段位 - 参团率：55%（在该指标排名第 2）”。请你后续所有操作始终遵循以下更新后的要求：执行你编写的单元测试文件时，目标代码的函数覆盖率要求达到98%或以上；同时请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过2层。","instruction_following_checklist":[{"constraint":"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于你输出的每一个代码文件，请将其中的注释全部写成中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 检查工作区是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 需要忽略的文件夹，避免检查第三方依赖库或打包产物\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'coverage'}\n \n # 根据用户指令中对游戏数据表格、热力图、ESLint 和单元测试的要求，可能出现的前端代码文件类型\n code_extensions = {\n '.html', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.vue', \n '.json', '.md', '.txt', '.py'\n }\n \n found_code_file = False\n \n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试以 utf-8 编码读取文件\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件包含非 UTF-8 编码格式的内容，不满足要求\n return False\n except Exception:\n # 忽略权限不足或其他读取异常，仅关注编码错误\n pass\n \n # 如果找到了至少一个代码文件，并且没有任何一个文件因为非 UTF-8 编码而抛错，则返回 True\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见前端页面框架和相关的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.vue'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏文件夹和常见的 node_modules 依赖文件夹\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于权限或编码等问题无法读取，则跳过\n continue\n \n has_code_file = True\n space_indents = []\n \n for line in lines:\n # 去除右侧空白后如果为空，则是空行，忽略\n if not line.rstrip():\n continue\n \n # 匹配行首的前导空白字符\n m = re.match(r'^([ \\t]+)', line)\n if m:\n indent = m.group(1)\n \n # 严禁使用制表符(Tab)缩进\n if '\\t' in indent:\n return False\n \n # 忽略 JSDoc 等多行注释里的对齐空格（常以 '*' 开头）\n if line.lstrip().startswith('*'):\n continue\n \n space_indents.append(len(indent))\n \n if not space_indents:\n continue\n \n # 统计符合 4的倍数 的缩进与非 4倍数 的偶数缩进(如 2, 6, 10 等典型 2 空格缩进特征)\n mod_0 = sum(1 for x in space_indents if x % 4 == 0)\n mod_2 = sum(1 for x in space_indents if x % 2 == 0 and x % 4 != 0)\n \n # 如果典型的 2 空格类缩进出现次数超过一定阈值，且占比显著，说明没有全部遵守 4 空格缩进规则\n if mod_2 > 5 and mod_2 > mod_0 * 0.2:\n return False\n \n # 如果文件中完全不存在 4 空格类缩进，却存在明显的非 4 倍数缩进，同样判断为不合格\n if mod_0 == 0 and mod_2 > 0:\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def check_pascal_case(name: str) -> bool:\n # 大驼峰要求：首字母大写，且仅包含大小写字母和数字\n return bool(re.match(r'^[A-Z][a-zA-Z0-9]*$', name))\n\n extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\n \n # JavaScript / TypeScript 等语言的匹配规则\n # 1. 匹配 function 关键字声明的命名函数\n pattern_func_js = re.compile(r'\\bfunction\\s+([a-zA-Z_]\\w*)\\s*\\(')\n # 2. 匹配箭头函数或匿名函数的变量赋值形式 (例如 const/let/var MyFunc = ... => )\n pattern_assign_js = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_]\\w*)\\s*=\\s*(?:async\\s+)?(?:function\\s*\\(|\\([^)]*\\)\\s*=>|[a-zA-Z_]\\w*\\s*=>)')\n # 3. 匹配类或对象的方法定义形式 (例如 async MyFunc() { )\n pattern_method_js = re.compile(r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+|async\\s+)*([a-zA-Z_]\\w*)\\s*\\([^()]*\\)\\s*\\{', re.MULTILINE)\n # 4. 匹配类中的箭头函数属性赋值形式 (例如 MyFunc = () => { )\n pattern_class_arrow_js = re.compile(r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+)*([a-zA-Z_]\\w*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)\\s*=>|[a-zA-Z_]\\w*\\s*=>)', re.MULTILINE)\n \n # Python 语言匹配规则\n pattern_func_py = re.compile(r'\\bdef\\s+([a-zA-Z_]\\w*)\\s*\\(')\n \n # 要排除的代码保留字和关键控制流词汇（防止匹配为函数名）\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', \n 'constructor', 'return', 'super', 'import', 'export',\n 'default', 'typeof', 'instanceof', 'new', 'delete',\n 'try', 'finally', 'do', 'yield', 'await', 'with'\n }\n\n all_pascal = True\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤无关依赖及构建输出目录\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '__pycache__')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 简单移除代码中的注释，减少误匹配风险\n if ext != '.py':\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n else:\n content = re.sub(r'#.*', '', content)\n content = re.sub(r\"'''.*?'''\", '', content, flags=re.DOTALL)\n content = re.sub(r'\"\"\".*?\"\"\"', '', content, flags=re.DOTALL)\n \n funcs = []\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\n funcs.extend(pattern_func_js.findall(content))\n funcs.extend(pattern_assign_js.findall(content))\n funcs.extend(pattern_class_arrow_js.findall(content))\n \n # 只有方法正则有概率会匹配到控制流语法，此处加以关键字过滤\n for m in pattern_method_js.finditer(content):\n name = m.group(1)\n if name not in keywords:\n funcs.append(name)\n \n elif ext == '.py':\n funcs.extend(pattern_func_py.findall(content))\n \n # 检查所有提取的函数名称是否均符合大驼峰规范\n for func in funcs:\n if func in keywords:\n continue\n if not check_pascal_case(func):\n all_pascal = False\n break\n \n if not all_pascal:\n break\n if not all_pascal:\n break\n\n return all_pascal","tags":["Style","Naming"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"执行你编写的单元测试文件时，目标代码的函数覆盖率要求达到98%或以上","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次调整项目后，务必更新progress.md文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过2层","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":3,"instruction":"下一步，在表格上方增加一个数据筛选面板。请添加两组单选按钮或下拉框：一组是“位置过滤（全部、上单、打野、中单、下路、辅助）”，另一组是“服务器（全服、峡谷之巅、一区）”。当切换这些筛选条件时，请使用新的模拟数据重新渲染表格内容。最关键的挑战是：要求数据和颜色在切换更新时，数字变化和背景色的变化必须有平滑的过渡动画（比如渐变切换的transition效果），绝对不能是生硬的一闪而过。请你后续所有操作始终遵循以下更新后的要求：每当项目发生变更时，请务必更新process-log.md文件，以记录本轮所完成的内容，替换此前更新progress.md的要求；同时你生成的单元测试文件在运行之后，必须保障目标代码的行覆盖率指标不低于100%。","instruction_following_checklist":[{"constraint":"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于你输出的每一个代码文件，请将其中的注释全部写成中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，则代表没有违反规则的文件，返回True\n if not os.path.exists(workspace_path):\n return True\n\n # 根据项目推测的前端常见代码、配置及记录文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.md'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不属于AI直接生成或非代码文件的目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n data = f.read()\n # 尝试将读取的二进制数据以 utf-8 解码\n data.decode('utf-8')\n except UnicodeDecodeError:\n # 一旦有文件无法以 utf-8 解码，说明编码不统一，返回 False\n return False\n except Exception:\n # 忽略诸如文件无权限读取之类的其它异常\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n\n # 常见的前端核心代码扩展名（排除.json和.md等易干扰文件）\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'coverage', '.vscode'}\n\n checked_files = 0\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤不需要扫描的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_exts:\n continue\n\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n continue\n\n tab_lines = 0\n space4_lines = 0\n other_space_lines = 0\n\n for line in lines:\n # 忽略空行\n if not line.strip():\n continue\n \n # 匹配前导空白序列和后续内容\n match = re.match(r'^([ \\t]+)(.*)', line)\n if not match:\n continue\n \n leading_spaces = match.group(1)\n content = match.group(2)\n\n # 忽略多行注释块常见的起始或连接符\n if content.startswith('*') or content.startswith('/*') or content.startswith('<!--'):\n continue\n\n # 统计缩进特征\n if '\\t' in leading_spaces:\n tab_lines += 1\n else:\n if len(leading_spaces) % 4 == 0:\n space4_lines += 1\n else:\n other_space_lines += 1\n \n total_indent = tab_lines + space4_lines + other_space_lines\n if total_indent == 0:\n continue\n \n checked_files += 1\n\n # 判定 1: 检查是否使用了 Tab 控制缩进（容许极少量的意外混入）\n if tab_lines > space4_lines and tab_lines > 3:\n return False\n \n # 判定 2: 检查是否使用了 2空格 等非 4空格 的缩进体系\n # 在 4空格 工程中，非4倍数空格行数（通常是换行对齐产生的）极少。\n # 如果 other_space_lines 超过了 space4_lines，说明整体极大概率是 2空格 缩进\n if other_space_lines > space4_lines and other_space_lines > 5:\n return False\n\n # 至少需要包含 1 个有效的前端代码文件才算完成了目标代码的生成\n return checked_files > 0","tags":["Style","Layout"]},{"constraint":"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求","validation_code":"import os\nimport re\n\ndef verify_pascal_case_functions(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n # 模式1：匹配 function 声明语句，兼容 TypeScript 泛型\n pattern1 = re.compile(r'\\bfunction\\s+([a-zA-Z0-9_]+)\\s*(?:<[^>]+>)?\\s*\\(')\n \n # 模式2：匹配通过 const, let, var 声明的箭头函数或匿名函数赋值，兼容 async 和 TS 类型声明\n pattern2 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*(?::[^=]+)?\\s*=\\s*(?:async\\s+)?(?:<[^>]*>\\s*)?(?:function\\b|\\([^)]*\\)(?:\\s*:\\s*[^=>]+)?\\s*=>|[a-zA-Z0-9_]+\\s*=>)')\n \n # 模式3：匹配 Python 函数声明\n pattern3 = re.compile(r'\\bdef\\s+([a-zA-Z0-9_]+)\\s*\\(')\n \n # 模式4：匹配类方法或对象内定义的方法，避免把 if/for 捕捉到（配合后置的 exclude 词）\n pattern4 = re.compile(r'^\\s*(?:async\\s+)?(?:(?:public|private|protected|static|get|set)\\s+)?([a-zA-Z0-9_]+)\\s*(?:<[^>]+>)?\\s*\\([^)]*\\)\\s*(?::\\s*[^\\{]+)?\\s*\\{', re.MULTILINE)\n\n # 需要豁免的关键字、内置生命周期函数与测试框架内置调用\n exclude_names = {\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', 'else', 'try', 'finally',\n 'function', 'await', 'yield', 'import', 'export', 'super', 'this',\n 'render', 'componentDidMount', 'componentDidUpdate', 'componentWillUnmount',\n 'shouldComponentUpdate', 'getDerivedStateFromProps', 'getSnapshotBeforeUpdate',\n 'data', 'mounted', 'created', 'setup', 'beforeDestroy', 'destroyed', \n 'beforeMount', 'beforeUpdate', 'updated', 'errorCaptured',\n 'renderTracked', 'renderTriggered', 'activated', 'deactivated',\n 'test', 'it', 'describe', 'beforeEach', 'afterEach', 'beforeAll', 'afterAll',\n 'require', 'module', 'define', 'setTimeout', 'setInterval', 'clearTimeout', 'clearInterval',\n 'watch', 'computed', 'mixins', 'props', 'components', 'provide', 'inject', 'methods'\n }\n\n # 大驼峰验证正则：大写字母开头，后跟字母和数字，且无下划线\n pascal_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的构建与依赖目录，特别是 coverage（防止生成的覆盖率 HTML 干扰）\n if any(skip_dir in root for skip_dir in ['node_modules', '.git', '__pycache__', 'dist', 'build', 'coverage']):\n continue\n \n for file in files:\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 收集文件中所有的函数或方法名称\n all_names = []\n all_names.extend(pattern1.findall(content))\n all_names.extend(pattern2.findall(content))\n all_names.extend(pattern3.findall(content))\n all_names.extend(pattern4.findall(content))\n \n for name in all_names:\n # 如果匹配到的“函数名”为框架保留字或控制流关键字，进行安全跳过\n if name in exclude_names:\n continue\n \n # 如任意一个自定义函数不符合大驼峰，则该约束未被满足\n if not pascal_pattern.match(name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每当项目发生变更时，请务必更新process-log.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你生成的单元测试文件在运行之后，必须保障目标代码的行覆盖率指标不低于100%","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"最后，我们来完善一下高级数据可视化和实用功能。首先，在表格的最右侧新增一列“综合评分趋势”，在这一列的单元格中不要显示文字数字，而是使用 SVG 或 Canvas 为每个段位绘制一个微型的折线图（Sparkline），展示模拟的近5个版本该段位玩家的表现起伏。其次，在筛选面板旁边增加一个“导出图表”按钮，点击后能够将当前包含热力图颜色和微型折线图的整个表格保存为一张本地PNG图片。请你后续所有操作始终遵循以下更新后的要求：第一，你生成的所有代码文件里，所有循环嵌套的层级都不超过1层；第二，无论何时修改现有文件，都要输出全部代码用于完整覆盖替换，请记住绝对禁止使用 Unified Diff 或是其他补丁格式。","instruction_following_checklist":[{"constraint":"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"对于你输出的每一个代码文件，请将其中的注释全部写成中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 结合前端工程、ESLint校验以及要求中提及的.md文件，定义常见的代码及文档拓展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.md'}\n \n # 若工作区路径不存在，则无法验证\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除通常无需检验或不由助手直接创建的依赖库及环境缓存等无关目录\n ignore_dirs = {'node_modules', '.git', '.idea', '.vscode', '__pycache__', 'dist', 'build'}\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件或指定的文档文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码方式读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件未采用有效的 UTF-8 编码\n return False\n except Exception:\n # 其他不可预见的读取异常（如权限问题），可以跳过\n continue\n \n # 如果所有被检查的文件均能以 UTF-8 正常解码并读取，说明要求得到了满足\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件应全部使用 4 个空格来控制缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import collections\n\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', \n '.hpp', '.go', '.rs', '.rb', '.php', '.swift', '.kt'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n filepath = os.path.join(root, file)\n diff_counts = collections.defaultdict(int)\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 忽略无法以UTF-8读取的非文本/异常文件\n continue\n \n prev_spaces = 0\n for line in lines:\n stripped = line.strip()\n # 忽略空行和常见的注释行，防止注释里的排版干扰判断\n if not stripped or stripped.startswith(('*', '/*', '<!--', '//', '#')):\n continue\n \n # 提取行首的空白字符部分\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 如果前导空白中包含 \\t，说明违背了全空格缩进原则\n if '\\t' in leading_whitespace:\n return False\n \n spaces = len(leading_whitespace)\n \n # 计算与上一行相比的缩进增量\n if spaces > prev_spaces:\n diff_counts[spaces - prev_spaces] += 1\n \n # 更新当前行的空白数供下一行对比\n prev_spaces = spaces\n \n if diff_counts:\n # 找出文件中出现频次最高的缩进步长\n # 如果频次相同，优先将步长为 4 视为最大，防止误判\n max_step = max(diff_counts, key=lambda k: (diff_counts[k], k == 4))\n \n # 如果该文件最主要的缩进步长是 2 或 3 个空格，则明显违反 4个空格 要求\n if max_step in [2, 3]:\n return False\n \n # 如果完全没有任何一次 4个空格 的缩进，却存在 2个空格 的缩进，同样违反\n if diff_counts.get(4, 0) == 0 and diff_counts.get(2, 0) > 0:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def strip_comments_and_strings(text: str) -> str:\n # 去除多行注释\n text = re.sub(r'/\\*[\\s\\S]*?\\*/', '', text)\n # 去除单行注释 (JS/TS // 以及 Python #)\n text = re.sub(r'//.*', '', text)\n text = re.sub(r'#.*', '', text)\n # 去除字符串 (单引号，双引号，反引号模板字符串)\n text = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", text)\n text = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', text)\n text = re.sub(r'`([^`\\\\]|\\\\.)*`', '``', text)\n return text\n\n def extract_functions(code: str) -> list:\n functions = []\n \n # 1. JS/TS 具名 function 声明 (例如: function MyFunc() )\n functions.extend(re.findall(r'\\bfunction\\s+([A-Za-z0-9_]+)\\s*\\(', code))\n \n # 2. JS/TS 赋值形式的箭头函数 (例如: const MyFunc = () => )\n functions.extend(re.findall(r'\\b(?:const|let|var)\\s+([A-Za-z0-9_]+)[^=;{}]*?=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[A-Za-z0-9_]+)\\s*=>', code))\n \n # 3. JS/TS 赋值形式的常规匿名函数 (例如: const MyFunc = function() )\n functions.extend(re.findall(r'\\b(?:const|let|var)\\s+([A-Za-z0-9_]+)[^=;{}]*?=\\s*(?:async\\s+)?function\\b', code))\n \n # 4. 类或对象的方法声明 (行首或缩进后)，排除了常见的控制流关键字\n keywords = [\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', \n 'finally', 'with', 'else', 'super', 'import', 'export', \n 'await', 'yield', 'get', 'set'\n ]\n kw_pattern = '|'.join(keywords)\n method_pattern = r'^\\s*(?!(?:' + kw_pattern + r')\\b)([A-Za-z0-9_]+)\\s*\\([^)]*\\)\\s*\\{'\n functions.extend(re.findall(method_pattern, code, re.MULTILINE))\n \n # 5. JS 对象属性赋值为函数 (例如: myFunc: function() 或 myFunc: () => )\n functions.extend(re.findall(r'^\\s*([A-Za-z0-9_]+)\\s*:\\s*(?:async\\s+)?(?:function\\b|(?:\\([^)]*\\)|[A-Za-z0-9_]+)\\s*=>)', code, re.MULTILINE))\n \n # 6. Python 具名函数声明 (防备Python测试或构建脚本)\n functions.extend(re.findall(r'\\bdef\\s+([A-Za-z0-9_]+)\\s*\\(', code))\n \n return functions\n\n # 常见的不需检查的目录\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'coverage'}\n # 可能会包含函数的文件后缀\n allowed_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\n \n # 用大驼峰正则匹配: 首字母大写，且只能包含字母和数字\n pascal_case_pattern = re.compile(r'^[A-Z][A-Za-z0-9]*$')\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤排除目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in allowed_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 首先剥离代码里的注释和字符串，避免对提取造成干扰\n code_clean = strip_comments_and_strings(code)\n \n # 提取此文件里的全部函数名\n extracted_funcs = extract_functions(code_clean)\n \n # 对每一个解析到的函数名称进行大驼峰检查\n for func_name in extracted_funcs:\n if not pascal_case_pattern.match(func_name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过1层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"每当项目发生变更时，请务必更新process-log.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论何时修改现有文件，都要输出全部代码用于完整覆盖替换，请记住绝对禁止使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个基于游戏段位的行为数据表格的基础页面框架。行需要表示不同的游戏段位（比如青铜、白银、黄金、铂金、钻石、大师、王者），列表示不同的行为数据指标（比如胜率、KDA、场均输出、参团率、视野得分）。现阶段只需要把表格的静态结构搭出来，页面顶部加一个大标题即可。请用一些模拟数据填充表格内容，暂时不需要任何特殊颜色或图表，只要求表格样式干净整洁，并留好后续加功能的代码结构扩展点。请你后续所有操作始终遵循以下要求：你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过；执行你编写的单元测试文件时，目标代码的语句覆盖率要求达到90%或以上。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请将其中的注释全部写成中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据前端项目的常见特性，定义可能会创建的代码文件后缀名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json'}\\n\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的第三方包目录和版本控制目录\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'coverage'}\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明存在非 utf-8 编码（例如包含非ASCII字符的GBK编码文件等）\\n return False\\n except Exception:\\n # 其他读取异常，暂时视作不合规文件\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能的前端及配置相关代码文件扩展名\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less', '.json'}\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build'}\\n has_code_file = False\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略第三方库及构建产物等不属于“生成的代码文件”的目录\\n root_parts = set(root.replace('\\\\\\\\', '/').split('/'))\\n if root_parts & exclude_dirs:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 若文件因编码等问题无法读取，则跳过\\n continue\\n \\n has_code_file = True\\n valid_indent_lines = 0\\n invalid_indent_lines = 0\\n \\n for line in lines:\\n stripped_line = line.lstrip()\\n # 忽略空行\\n if not stripped_line:\\n continue\\n \\n # 提取前导空白部分\\n leading_whitespaces = line[:len(line) - len(stripped_line)]\\n \\n # 如果出现 Tab 字符，直接认为违规\\n if '\\\\t' in leading_whitespaces:\\n return False\\n \\n # 忽略类似 JsDoc 的块注释内部行（由于通常带有额外的 1 空格对齐 '*' 号）\\n if stripped_line.startswith('*'):\\n continue\\n \\n space_count = len(leading_whitespaces)\\n if space_count > 0:\\n # 检查缩进空格数是否为 4 的倍数\\n if space_count % 4 == 0:\\n valid_indent_lines += 1\\n else:\\n invalid_indent_lines += 1\\n \\n total_indented_lines = valid_indent_lines + invalid_indent_lines\\n \\n if total_indented_lines > 0:\\n # 容错处理：允许极少量的换行参数对齐导致非 4 的倍数缩进，\\n # 但如果不符合行数的比例大于 5% 并且总数超过 2 行，即认为使用了非 4 空格的缩进规范。\\n if invalid_indent_lines > 2 and (invalid_indent_lines / total_indented_lines) > 0.05:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义用于匹配各种函数定义并提取函数名称的正则表达式\\n func_patterns = [\\n # 匹配常规的 JS/TS 函数: function MyFunc(...) \\n re.compile(r\\\"\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(\\\"),\\n # 匹配箭头函数赋值: const MyFunc = (...) => { ... } 或 const MyFunc = async e => { ... }\\n re.compile(r\\\"\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z0-9_]+)\\\\s*=>\\\"),\\n # 匹配匿名函数赋值: const MyFunc = function(...) { ... }\\n re.compile(r\\\"\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s*)?function\\\\s*\\\\(\\\"),\\n # 匹配 Python 函数: def MyFunc(...)\\n re.compile(r\\\"\\\\bdef\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(\\\")\\n ]\\n \\n # 大驼峰命名正则: 必须以大写字母开头，后续仅包含字母和数字，不允许有下划线\\n pascal_case_pattern = re.compile(r\\\"^[A-Z][a-zA-Z0-9]*$\\\")\\n \\n # 根据用户提示的 ESLint，主要针对前端相关代码文件进行扫描\\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除依赖库和版本控制文件夹\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n funcs = []\\n # 运用多个正则提取文件中定义的所有函数名\\n for pattern in func_patterns:\\n funcs.extend(pattern.findall(content))\\n \\n # 判断每个函数名是否为大驼峰格式\\n for func_name in funcs:\\n if not pascal_case_pattern.match(func_name):\\n return False\\n except Exception:\\n # 忽略无法正常读取或非文本类型的文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"执行你编写的单元测试文件时，目标代码的语句覆盖率要求达到90%或以上\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"非常好，现在我们要加上核心的“热力图”视觉效果。请根据每列（即每一项行为指标）的数据分布，给表格的具体数值单元格加上渐变背景色。要求：针对每一列的数据单独计算极值，数值越高的单元格背景色偏向翠绿色，数值越低偏向珊瑚红色，居中的数值呈浅黄色或白色。确保代码是根据列的最大最小值动态计算色阶的，从而能让用户一眼看出每个段位在某项指标上的相对表现。\\n请你后续所有操作始终遵循以下更新后的要求：执行你编写的单元测试文件时，目标代码的分支覆盖率要求达到95%或以上；同时每次调整项目后，务必更新progress.md文件，以便记录当前这轮完成的任务内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请将其中的注释全部写成中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n return False\\n except Exception:\\n # 忽略非编码相关的读取错误（如权限问题）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def check_indentation(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css'}\\n checked_any = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in target_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n stripped = line.lstrip()\\n if not stripped or stripped.startswith('*') or stripped.startswith('/*'):\\n continue\\n \\n leading_space = line[:len(line) - len(stripped)]\\n \\n if '\\\\t' in leading_space or len(leading_space) % 4 != 0:\\n return False\\n checked_any = True\\n except Exception:\\n pass\\n\\n return checked_any\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能的代码文件扩展名，以 JS/TS 及前端相关为主，兼容 Python\\n valid_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py'}\\n # 过滤掉无关依赖及构建输出目录\\n skip_dirs = {'node_modules', 'dist', 'build', 'coverage', '.git', '__pycache__', '.vscode', '.idea'}\\n\\n # 1. Python 中的函数定义\\n py_func_pattern = re.compile(r'\\\\bdef\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n # 2. JS/TS 中的 function 函数定义\\n js_func_pattern = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n # 3. JS/TS 箭头函数定义，处理跨行和类型注释，利用 [^=;]* 防止跨语句越界\\n js_arrow_pattern = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z_]\\\\w*)[^=;]*=>')\\n # 4. JS/TS 普通函数表达式\\n js_expr_pattern = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s*)?function\\\\b')\\n # 5. 类方法或对象内的方法简写，利用 [^{;]* 防止跨语句错误连带\\n method_pattern = re.compile(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*\\\\([^)]*\\\\)[^{;]*\\\\{')\\n\\n # 大驼峰命名校验正则：必须全是大写字母开头，跟随字母和数字，不能有下划线\\n pascal_case_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\\n \\n # 避免被当成函数名的方法正则干扰关键字\\n keywords = {\\n 'if', 'while', 'for', 'switch', 'catch', 'function', 'return', 'super', 'constructor', \\n 'else', 'try', 'finally', 'with', 'typeof', 'instanceof', 'new', 'delete', 'void', 'yield', \\n 'await', 'get', 'set', 'import', 'export', 'default'\\n }\\n\\n def clean_code(code: str, ext: str) -> str:\\n \\\"\\\"\\\"用于消除注释和字符串字面量的干扰内容\\\"\\\"\\\"\\n if ext == '.py':\\n code = re.sub(r'#.*', '', code)\\n code = re.sub(r'\\\"\\\"\\\"(?:\\\\\\\\.|[^\\\\\\\\])*?\\\"\\\"\\\"', '', code, flags=re.DOTALL)\\n code = re.sub(r\\\"'''(?:\\\\\\\\.|[^\\\\\\\\])*?'''\\\", '', code, flags=re.DOTALL)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', code)\\n else:\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', code, flags=re.DOTALL)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', code)\\n return code\\n\\n for root, dirs, files in os.walk(workspace_path):\\n dirs[:] = [d for d in dirs if d not in skip_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1]\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n raw_content = f.read()\\n except Exception:\\n continue\\n \\n content = clean_code(raw_content, ext)\\n \\n functions = []\\n if ext == '.py':\\n functions.extend(py_func_pattern.findall(content))\\n else:\\n functions.extend(js_func_pattern.findall(content))\\n functions.extend(js_arrow_pattern.findall(content))\\n functions.extend(js_expr_pattern.findall(content))\\n methods = method_pattern.findall(content)\\n for m in methods:\\n if m not in keywords:\\n functions.append(m)\\n\\n # 逐个检查识别出的函数名\\n for func in functions:\\n if not func:\\n continue\\n if not pascal_case_pattern.match(func):\\n # 发现任何一个并非大驼峰的函数即判定失败\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"执行你编写的单元测试文件时，目标代码的分支覆盖率要求达到95%或以上\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次调整项目后，务必更新progress.md文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"视觉上直观多了！接下来增加表格的交互性。第一，请实现点击表头可以按该列进行升序或降序排序的功能，排序时行数据要整体移动，并且热力图的颜色也要保持对应正确。第二，当鼠标悬停在热力图单元格上时，展示一个悬浮提示框（Tooltip），里面要清晰地拼接出该项数据的解读，例如显示：“钻石段位 - 参团率：55%（在该指标排名第 2）”。请你后续所有操作始终遵循以下更新后的要求：执行你编写的单元测试文件时，目标代码的函数覆盖率要求达到98%或以上；同时请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过2层。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请将其中的注释全部写成中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 检查工作区是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 需要忽略的文件夹，避免检查第三方依赖库或打包产物\\n ignore_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'coverage'}\\n \\n # 根据用户指令中对游戏数据表格、热力图、ESLint 和单元测试的要求，可能出现的前端代码文件类型\\n code_extensions = {\\n '.html', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.vue', \\n '.json', '.md', '.txt', '.py'\\n }\\n \\n found_code_file = False\\n \\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试以 utf-8 编码读取文件\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件包含非 UTF-8 编码格式的内容，不满足要求\\n return False\\n except Exception:\\n # 忽略权限不足或其他读取异常，仅关注编码错误\\n pass\\n \\n # 如果找到了至少一个代码文件，并且没有任何一个文件因为非 UTF-8 编码而抛错，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见前端页面框架和相关的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', '.less', '.vue'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏文件夹和常见的 node_modules 依赖文件夹\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于权限或编码等问题无法读取，则跳过\\n continue\\n \\n has_code_file = True\\n space_indents = []\\n \\n for line in lines:\\n # 去除右侧空白后如果为空，则是空行，忽略\\n if not line.rstrip():\\n continue\\n \\n # 匹配行首的前导空白字符\\n m = re.match(r'^([ \\\\t]+)', line)\\n if m:\\n indent = m.group(1)\\n \\n # 严禁使用制表符(Tab)缩进\\n if '\\\\t' in indent:\\n return False\\n \\n # 忽略 JSDoc 等多行注释里的对齐空格（常以 '*' 开头）\\n if line.lstrip().startswith('*'):\\n continue\\n \\n space_indents.append(len(indent))\\n \\n if not space_indents:\\n continue\\n \\n # 统计符合 4的倍数 的缩进与非 4倍数 的偶数缩进(如 2, 6, 10 等典型 2 空格缩进特征)\\n mod_0 = sum(1 for x in space_indents if x % 4 == 0)\\n mod_2 = sum(1 for x in space_indents if x % 2 == 0 and x % 4 != 0)\\n \\n # 如果典型的 2 空格类缩进出现次数超过一定阈值，且占比显著，说明没有全部遵守 4 空格缩进规则\\n if mod_2 > 5 and mod_2 > mod_0 * 0.2:\\n return False\\n \\n # 如果文件中完全不存在 4 空格类缩进，却存在明显的非 4 倍数缩进，同样判断为不合格\\n if mod_0 == 0 and mod_2 > 0:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def check_pascal_case(name: str) -> bool:\\n # 大驼峰要求：首字母大写，且仅包含大小写字母和数字\\n return bool(re.match(r'^[A-Z][a-zA-Z0-9]*$', name))\\n\\n extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\\n \\n # JavaScript / TypeScript 等语言的匹配规则\\n # 1. 匹配 function 关键字声明的命名函数\\n pattern_func_js = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n # 2. 匹配箭头函数或匿名函数的变量赋值形式 (例如 const/let/var MyFunc = ... => )\\n pattern_assign_js = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\s*\\\\(|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_]\\\\w*\\\\s*=>)')\\n # 3. 匹配类或对象的方法定义形式 (例如 async MyFunc() { )\\n pattern_method_js = re.compile(r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+|async\\\\s+)*([a-zA-Z_]\\\\w*)\\\\s*\\\\([^()]*\\\\)\\\\s*\\\\{', re.MULTILINE)\\n # 4. 匹配类中的箭头函数属性赋值形式 (例如 MyFunc = () => { )\\n pattern_class_arrow_js = re.compile(r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+)*([a-zA-Z_]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_]\\\\w*\\\\s*=>)', re.MULTILINE)\\n \\n # Python 语言匹配规则\\n pattern_func_py = re.compile(r'\\\\bdef\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n \\n # 要排除的代码保留字和关键控制流词汇（防止匹配为函数名）\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', \\n 'constructor', 'return', 'super', 'import', 'export',\\n 'default', 'typeof', 'instanceof', 'new', 'delete',\\n 'try', 'finally', 'do', 'yield', 'await', 'with'\\n }\\n\\n all_pascal = True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤无关依赖及构建输出目录\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '__pycache__')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 简单移除代码中的注释，减少误匹配风险\\n if ext != '.py':\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n else:\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r\\\"'''.*?'''\\\", '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', content, flags=re.DOTALL)\\n \\n funcs = []\\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\\n funcs.extend(pattern_func_js.findall(content))\\n funcs.extend(pattern_assign_js.findall(content))\\n funcs.extend(pattern_class_arrow_js.findall(content))\\n \\n # 只有方法正则有概率会匹配到控制流语法，此处加以关键字过滤\\n for m in pattern_method_js.finditer(content):\\n name = m.group(1)\\n if name not in keywords:\\n funcs.append(name)\\n \\n elif ext == '.py':\\n funcs.extend(pattern_func_py.findall(content))\\n \\n # 检查所有提取的函数名称是否均符合大驼峰规范\\n for func in funcs:\\n if func in keywords:\\n continue\\n if not check_pascal_case(func):\\n all_pascal = False\\n break\\n \\n if not all_pascal:\\n break\\n if not all_pascal:\\n break\\n\\n return all_pascal\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"执行你编写的单元测试文件时，目标代码的函数覆盖率要求达到98%或以上\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次调整项目后，务必更新progress.md文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步，在表格上方增加一个数据筛选面板。请添加两组单选按钮或下拉框：一组是“位置过滤（全部、上单、打野、中单、下路、辅助）”，另一组是“服务器（全服、峡谷之巅、一区）”。当切换这些筛选条件时，请使用新的模拟数据重新渲染表格内容。最关键的挑战是：要求数据和颜色在切换更新时，数字变化和背景色的变化必须有平滑的过渡动画（比如渐变切换的transition效果），绝对不能是生硬的一闪而过。请你后续所有操作始终遵循以下更新后的要求：每当项目发生变更时，请务必更新process-log.md文件，以记录本轮所完成的内容，替换此前更新progress.md的要求；同时你生成的单元测试文件在运行之后，必须保障目标代码的行覆盖率指标不低于100%。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请将其中的注释全部写成中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，则代表没有违反规则的文件，返回True\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 根据项目推测的前端常见代码、配置及记录文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.md'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不属于AI直接生成或非代码文件的目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n data = f.read()\\n # 尝试将读取的二进制数据以 utf-8 解码\\n data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 一旦有文件无法以 utf-8 解码，说明编码不统一，返回 False\\n return False\\n except Exception:\\n # 忽略诸如文件无权限读取之类的其它异常\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 常见的前端核心代码扩展名（排除.json和.md等易干扰文件）\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', 'coverage', '.vscode'}\\n\\n checked_files = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤不需要扫描的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_exts:\\n continue\\n\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n\\n tab_lines = 0\\n space4_lines = 0\\n other_space_lines = 0\\n\\n for line in lines:\\n # 忽略空行\\n if not line.strip():\\n continue\\n \\n # 匹配前导空白序列和后续内容\\n match = re.match(r'^([ \\\\t]+)(.*)', line)\\n if not match:\\n continue\\n \\n leading_spaces = match.group(1)\\n content = match.group(2)\\n\\n # 忽略多行注释块常见的起始或连接符\\n if content.startswith('*') or content.startswith('/*') or content.startswith('<!--'):\\n continue\\n\\n # 统计缩进特征\\n if '\\\\t' in leading_spaces:\\n tab_lines += 1\\n else:\\n if len(leading_spaces) % 4 == 0:\\n space4_lines += 1\\n else:\\n other_space_lines += 1\\n \\n total_indent = tab_lines + space4_lines + other_space_lines\\n if total_indent == 0:\\n continue\\n \\n checked_files += 1\\n\\n # 判定 1: 检查是否使用了 Tab 控制缩进（容许极少量的意外混入）\\n if tab_lines > space4_lines and tab_lines > 3:\\n return False\\n \\n # 判定 2: 检查是否使用了 2空格 等非 4空格 的缩进体系\\n # 在 4空格 工程中，非4倍数空格行数（通常是换行对齐产生的）极少。\\n # 如果 other_space_lines 超过了 space4_lines，说明整体极大概率是 2空格 缩进\\n if other_space_lines > space4_lines and other_space_lines > 5:\\n return False\\n\\n # 至少需要包含 1 个有效的前端代码文件才算完成了目标代码的生成\\n return checked_files > 0\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify_pascal_case_functions(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 模式1：匹配 function 声明语句，兼容 TypeScript 泛型\\n pattern1 = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z0-9_]+)\\\\s*(?:<[^>]+>)?\\\\s*\\\\(')\\n \\n # 模式2：匹配通过 const, let, var 声明的箭头函数或匿名函数赋值，兼容 async 和 TS 类型声明\\n pattern2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*(?::[^=]+)?\\\\s*=\\\\s*(?:async\\\\s+)?(?:<[^>]*>\\\\s*)?(?:function\\\\b|\\\\([^)]*\\\\)(?:\\\\s*:\\\\s*[^=>]+)?\\\\s*=>|[a-zA-Z0-9_]+\\\\s*=>)')\\n \\n # 模式3：匹配 Python 函数声明\\n pattern3 = re.compile(r'\\\\bdef\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(')\\n \\n # 模式4：匹配类方法或对象内定义的方法，避免把 if/for 捕捉到（配合后置的 exclude 词）\\n pattern4 = re.compile(r'^\\\\s*(?:async\\\\s+)?(?:(?:public|private|protected|static|get|set)\\\\s+)?([a-zA-Z0-9_]+)\\\\s*(?:<[^>]+>)?\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^\\\\{]+)?\\\\s*\\\\{', re.MULTILINE)\\n\\n # 需要豁免的关键字、内置生命周期函数与测试框架内置调用\\n exclude_names = {\\n 'if', 'for', 'while', 'switch', 'catch', 'constructor', 'return', 'else', 'try', 'finally',\\n 'function', 'await', 'yield', 'import', 'export', 'super', 'this',\\n 'render', 'componentDidMount', 'componentDidUpdate', 'componentWillUnmount',\\n 'shouldComponentUpdate', 'getDerivedStateFromProps', 'getSnapshotBeforeUpdate',\\n 'data', 'mounted', 'created', 'setup', 'beforeDestroy', 'destroyed', \\n 'beforeMount', 'beforeUpdate', 'updated', 'errorCaptured',\\n 'renderTracked', 'renderTriggered', 'activated', 'deactivated',\\n 'test', 'it', 'describe', 'beforeEach', 'afterEach', 'beforeAll', 'afterAll',\\n 'require', 'module', 'define', 'setTimeout', 'setInterval', 'clearTimeout', 'clearInterval',\\n 'watch', 'computed', 'mixins', 'props', 'components', 'provide', 'inject', 'methods'\\n }\\n\\n # 大驼峰验证正则：大写字母开头，后跟字母和数字，且无下划线\\n pascal_pattern = re.compile(r'^[A-Z][a-zA-Z0-9]*$')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的构建与依赖目录，特别是 coverage（防止生成的覆盖率 HTML 干扰）\\n if any(skip_dir in root for skip_dir in ['node_modules', '.git', '__pycache__', 'dist', 'build', 'coverage']):\\n continue\\n \\n for file in files:\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 收集文件中所有的函数或方法名称\\n all_names = []\\n all_names.extend(pattern1.findall(content))\\n all_names.extend(pattern2.findall(content))\\n all_names.extend(pattern3.findall(content))\\n all_names.extend(pattern4.findall(content))\\n \\n for name in all_names:\\n # 如果匹配到的“函数名”为框架保留字或控制流关键字，进行安全跳过\\n if name in exclude_names:\\n continue\\n \\n # 如任意一个自定义函数不符合大驼峰，则该约束未被满足\\n if not pascal_pattern.match(name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新process-log.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你生成的单元测试文件在运行之后，必须保障目标代码的行覆盖率指标不低于100%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，我们来完善一下高级数据可视化和实用功能。首先，在表格的最右侧新增一列“综合评分趋势”，在这一列的单元格中不要显示文字数字，而是使用 SVG 或 Canvas 为每个段位绘制一个微型的折线图（Sparkline），展示模拟的近5个版本该段位玩家的表现起伏。其次，在筛选面板旁边增加一个“导出图表”按钮，点击后能够将当前包含热力图颜色和微型折线图的整个表格保存为一张本地PNG图片。请你后续所有操作始终遵循以下更新后的要求：第一，你生成的所有代码文件里，所有循环嵌套的层级都不超过1层；第二，无论何时修改现有文件，都要输出全部代码用于完整覆盖替换，请记住绝对禁止使用 Unified Diff 或是其他补丁格式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"严禁直接使用 Grep 或 Glob 工具进行代码探索，你必须尽最大可能地使用 Task 工具\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在查找或处理文件时，请尽可能通过具有 subagent_type=Explore 的 Task 工具来实现，避免直接运行搜索或文件操作命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"对于你输出的每一个代码文件，请将其中的注释全部写成中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 结合前端工程、ESLint校验以及要求中提及的.md文件，定义常见的代码及文档拓展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.json', '.md'}\\n \\n # 若工作区路径不存在，则无法验证\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除通常无需检验或不由助手直接创建的依赖库及环境缓存等无关目录\\n ignore_dirs = {'node_modules', '.git', '.idea', '.vscode', '__pycache__', 'dist', 'build'}\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件或指定的文档文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码方式读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件未采用有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 其他不可预见的读取异常（如权限问题），可以跳过\\n continue\\n \\n # 如果所有被检查的文件均能以 UTF-8 正常解码并读取，说明要求得到了满足\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用 4 个空格来控制缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import collections\\n\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \\n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', \\n '.hpp', '.go', '.rs', '.rb', '.php', '.swift', '.kt'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n diff_counts = collections.defaultdict(int)\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 忽略无法以UTF-8读取的非文本/异常文件\\n continue\\n \\n prev_spaces = 0\\n for line in lines:\\n stripped = line.strip()\\n # 忽略空行和常见的注释行，防止注释里的排版干扰判断\\n if not stripped or stripped.startswith(('*', '/*', '<!--', '//', '#')):\\n continue\\n \\n # 提取行首的空白字符部分\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 如果前导空白中包含 \\\\t，说明违背了全空格缩进原则\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n spaces = len(leading_whitespace)\\n \\n # 计算与上一行相比的缩进增量\\n if spaces > prev_spaces:\\n diff_counts[spaces - prev_spaces] += 1\\n \\n # 更新当前行的空白数供下一行对比\\n prev_spaces = spaces\\n \\n if diff_counts:\\n # 找出文件中出现频次最高的缩进步长\\n # 如果频次相同，优先将步长为 4 视为最大，防止误判\\n max_step = max(diff_counts, key=lambda k: (diff_counts[k], k == 4))\\n \\n # 如果该文件最主要的缩进步长是 2 或 3 个空格，则明显违反 4个空格 要求\\n if max_step in [2, 3]:\\n return False\\n \\n # 如果完全没有任何一次 4个空格 的缩进，却存在 2个空格 的缩进，同样违反\\n if diff_counts.get(4, 0) == 0 and diff_counts.get(2, 0) > 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请保证所生成代码文件中包含的每个函数名称均满足 大驼峰 命名要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def strip_comments_and_strings(text: str) -> str:\\n # 去除多行注释\\n text = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', text)\\n # 去除单行注释 (JS/TS // 以及 Python #)\\n text = re.sub(r'//.*', '', text)\\n text = re.sub(r'#.*', '', text)\\n # 去除字符串 (单引号，双引号，反引号模板字符串)\\n text = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", text)\\n text = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', text)\\n text = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', '``', text)\\n return text\\n\\n def extract_functions(code: str) -> list:\\n functions = []\\n \\n # 1. JS/TS 具名 function 声明 (例如: function MyFunc() )\\n functions.extend(re.findall(r'\\\\bfunction\\\\s+([A-Za-z0-9_]+)\\\\s*\\\\(', code))\\n \\n # 2. JS/TS 赋值形式的箭头函数 (例如: const MyFunc = () => )\\n functions.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([A-Za-z0-9_]+)[^=;{}]*?=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[A-Za-z0-9_]+)\\\\s*=>', code))\\n \\n # 3. JS/TS 赋值形式的常规匿名函数 (例如: const MyFunc = function() )\\n functions.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([A-Za-z0-9_]+)[^=;{}]*?=\\\\s*(?:async\\\\s+)?function\\\\b', code))\\n \\n # 4. 类或对象的方法声明 (行首或缩进后)，排除了常见的控制流关键字\\n keywords = [\\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', \\n 'finally', 'with', 'else', 'super', 'import', 'export', \\n 'await', 'yield', 'get', 'set'\\n ]\\n kw_pattern = '|'.join(keywords)\\n method_pattern = r'^\\\\s*(?!(?:' + kw_pattern + r')\\\\b)([A-Za-z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n functions.extend(re.findall(method_pattern, code, re.MULTILINE))\\n \\n # 5. JS 对象属性赋值为函数 (例如: myFunc: function() 或 myFunc: () => )\\n functions.extend(re.findall(r'^\\\\s*([A-Za-z0-9_]+)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|(?:\\\\([^)]*\\\\)|[A-Za-z0-9_]+)\\\\s*=>)', code, re.MULTILINE))\\n \\n # 6. Python 具名函数声明 (防备Python测试或构建脚本)\\n functions.extend(re.findall(r'\\\\bdef\\\\s+([A-Za-z0-9_]+)\\\\s*\\\\(', code))\\n \\n return functions\\n\\n # 常见的不需检查的目录\\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__', 'coverage'}\\n # 可能会包含函数的文件后缀\\n allowed_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\\n \\n # 用大驼峰正则匹配: 首字母大写，且只能包含字母和数字\\n pascal_case_pattern = re.compile(r'^[A-Z][A-Za-z0-9]*$')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤排除目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in allowed_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 首先剥离代码里的注释和字符串，避免对提取造成干扰\\n code_clean = strip_comments_and_strings(code)\\n \\n # 提取此文件里的全部函数名\\n extracted_funcs = extract_functions(code_clean)\\n \\n # 对每一个解析到的函数名称进行大驼峰检查\\n for func_name in extracted_funcs:\\n if not pascal_case_pattern.match(func_name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请保证你生成的所有代码文件里，所有循环嵌套的层级都不超过1层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新process-log.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论何时修改现有文件，都要输出全部代码用于完整覆盖替换，请记住绝对禁止使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面整体，顶部有一个大标题，下方包含一个完整的表格。表格最左侧列展示不同的游戏段位（如青铜、白银、黄金、铂金、钻石、大师、王者等），上方表头包含不同的行为数据指标（如胜率、KDA、场均输出、参团率、视野得分）。确认表格单元格内已用模拟数据填充，表格样式干净整洁。","观察表格的指标数值单元格，验证是否具有热力图背景色。确认在同一列的纵向对比中，数值最高的单元格背景色偏向翠绿色，数值最低的偏向珊瑚红色，中间数值偏向浅黄色或白色。","点击任意一个行为指标的表头，验证表格行数据是否按该列数据进行升序或降序重新排列。重点检查段位名称、其他列的数值及对应的热力图背景色是否随整行一起正确移动。","将鼠标悬停在表格中间任意一个具体数值的单元格上，验证页面是否弹出一个悬浮提示框（Tooltip）。确认提示框内清晰显示了正确的段位名、指标名、该单元格的具体数值以及该数值在本列的具体排名（如'钻石段位 - 参团率：55%（在该指标排名第 2）'）。","查看表格上方的筛选面板，确认存在'位置过滤'（包含全部、上单、打野、中单、下路、辅助）和'服务器'（包含全服、峡谷之巅、一区）两组选项控件（下拉框或单选按钮）。切换任意控件的值，验证表格内的指标数值和热力图颜色是否发生改变（模拟数据更新）。","在筛选面板切换选项触发数据更新时，观察表格单元格，验证其内部的数字改变和背景颜色的切换是否具有平滑的渐变过渡动画（transition），确信不存在生硬瞬间突变的情况。","查看表格的最右侧列，确认表头是否名为'综合评分趋势'。验证该列下的每一个单元格中均没有显示文字或数字，而是通过 SVG 或 Canvas 渲染了一个微型折线图（Sparkline），展示模拟的近5个版本该段位玩家表现的起伏波动。","点击页面上的'导出图表'按钮，验证浏览器是否触发了文件下载行为。打开下载的文件，确认其为一张完整清晰的 PNG 图片，且图片内容完整包含了当前的表格界面、正确的热力图背景色以及微型折线图的视觉效果。"],"string":"[\n \"查看页面整体，顶部有一个大标题，下方包含一个完整的表格。表格最左侧列展示不同的游戏段位（如青铜、白银、黄金、铂金、钻石、大师、王者等），上方表头包含不同的行为数据指标（如胜率、KDA、场均输出、参团率、视野得分）。确认表格单元格内已用模拟数据填充，表格样式干净整洁。\",\n \"观察表格的指标数值单元格，验证是否具有热力图背景色。确认在同一列的纵向对比中，数值最高的单元格背景色偏向翠绿色，数值最低的偏向珊瑚红色，中间数值偏向浅黄色或白色。\",\n \"点击任意一个行为指标的表头，验证表格行数据是否按该列数据进行升序或降序重新排列。重点检查段位名称、其他列的数值及对应的热力图背景色是否随整行一起正确移动。\",\n \"将鼠标悬停在表格中间任意一个具体数值的单元格上，验证页面是否弹出一个悬浮提示框（Tooltip）。确认提示框内清晰显示了正确的段位名、指标名、该单元格的具体数值以及该数值在本列的具体排名（如'钻石段位 - 参团率：55%（在该指标排名第 2）'）。\",\n \"查看表格上方的筛选面板，确认存在'位置过滤'（包含全部、上单、打野、中单、下路、辅助）和'服务器'（包含全服、峡谷之巅、一区）两组选项控件（下拉框或单选按钮）。切换任意控件的值，验证表格内的指标数值和热力图颜色是否发生改变（模拟数据更新）。\",\n \"在筛选面板切换选项触发数据更新时，观察表格单元格，验证其内部的数字改变和背景颜色的切换是否具有平滑的渐变过渡动画（transition），确信不存在生硬瞬间突变的情况。\",\n \"查看表格的最右侧列，确认表头是否名为'综合评分趋势'。验证该列下的每一个单元格中均没有显示文字或数字，而是通过 SVG 或 Canvas 渲染了一个微型折线图（Sparkline），展示模拟的近5个版本该段位玩家表现的起伏波动。\",\n \"点击页面上的'导出图表'按钮，验证浏览器是否触发了文件下载行为。打开下载的文件，确认其为一张完整清晰的 PNG 图片，且图片内容完整包含了当前的表格界面、正确的热力图背景色以及微型折线图的视觉效果。\"\n]","isConversational":false}}},{"rowIdx":74,"cells":{"task_id":{"kind":"number","value":75,"string":"75"},"repository_policy":{"kind":"string","value":"你在后续所有交互轮次的所有操作都必须严格遵循以下约束，不得违反：\n1. 进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改。\n2. 执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录。\n3. 在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑。\n4. 要求在生成的代码文件中必须出现递归调用。\n5. 务必在生成的代码文件中体现出三元运算符的使用。\n6. 生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布。\n7. 对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-。\n8. 针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成的代码文件中必须出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在生成的代码文件中体现出三元运算符的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区路径无效，则无法验证代码文件，返回False\n if not workspace_path or not os.path.exists(workspace_path):\n return False\n\n # 常见代码文件的扩展名列表\n code_exts = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', \n '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.php', '.swift', '.kt', '.go', '.rb', '.dart', '.rs', '.m', '.mm'\n }\n\n # 遍历当前工作区的所有目录与文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n # 根据不同语言的基础单行注释符，简单剔除注释干扰\n if ext in ['.py', '.rb']:\n code_line = line.split('#')[0]\n else:\n code_line = line.split('//')[0]\n \n # 对 Python 特有的三元运算符结构进行正则匹配\n if ext == '.py':\n # 匹配模式：前面有非空字符 + if + 中间字符 + else \n # (有效剔除传统 if-else 语句块的首行)\n if re.search(r'\\S.*?\\bif\\s+.*?\\belse\\b', code_line):\n return True\n \n # 对绝大多数 C-style / JS-style 语言的三元运算符（? :）进行匹配\n else:\n # 匹配模式：包含 ? 且其后跟随 :\n if re.search(r'\\?.*:', code_line):\n return True\n except Exception:\n # 遇到无法读取的文件（如二进制混合、权限问题等）则跳过\n continue\n\n return False","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据“import/require”等上下文推断可能的代码文件扩展名\n code_extensions = {'.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php', '.sh'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录及常见的非生成代码的依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'venv', 'env', '__pycache__')]\n \n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法按 UTF-8 读取或发生其他异常，视为未通过验证\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.c', '.cpp', '.cc', '.cxx', \n '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.html', '.css', '.sh', \n '.swift', '.kt', '.m', '.scala', '.vue', '.svelte', '.dart'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 提取文件中所有的非空行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n # 检查文件的最后5个非空行，以兼容可能存在的块注释结束符号(如 */, --> 等)\n tail_lines = non_empty_lines[-5:] if non_empty_lines else []\n \n found = False\n for line in tail_lines:\n if target_text in line:\n found = True\n break\n \n # 如果该代码文件结尾处没有找到指定的注释内容，直接返回 False\n if not found:\n return False\n except Exception:\n # 如果存在不可读文件或编码问题，忽略即可\n pass\n \n # 如果项目至少包含一个代码文件，并且都通过了检测，则返回 True\n return has_code_file","tags":["Content","Format"]}],"string":"[\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成的代码文件中必须出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在生成的代码文件中体现出三元运算符的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区路径无效，则无法验证代码文件，返回False\\n if not workspace_path or not os.path.exists(workspace_path):\\n return False\\n\\n # 常见代码文件的扩展名列表\\n code_exts = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', \\n '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.php', '.swift', '.kt', '.go', '.rb', '.dart', '.rs', '.m', '.mm'\\n }\\n\\n # 遍历当前工作区的所有目录与文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n # 根据不同语言的基础单行注释符，简单剔除注释干扰\\n if ext in ['.py', '.rb']:\\n code_line = line.split('#')[0]\\n else:\\n code_line = line.split('//')[0]\\n \\n # 对 Python 特有的三元运算符结构进行正则匹配\\n if ext == '.py':\\n # 匹配模式：前面有非空字符 + if + 中间字符 + else \\n # (有效剔除传统 if-else 语句块的首行)\\n if re.search(r'\\\\S.*?\\\\bif\\\\s+.*?\\\\belse\\\\b', code_line):\\n return True\\n \\n # 对绝大多数 C-style / JS-style 语言的三元运算符（? :）进行匹配\\n else:\\n # 匹配模式：包含 ? 且其后跟随 :\\n if re.search(r'\\\\?.*:', code_line):\\n return True\\n except Exception:\\n # 遇到无法读取的文件（如二进制混合、权限问题等）则跳过\\n continue\\n\\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据“import/require”等上下文推断可能的代码文件扩展名\\n code_extensions = {'.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php', '.sh'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录及常见的非生成代码的依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'venv', 'env', '__pycache__')]\\n \\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法按 UTF-8 读取或发生其他异常，视为未通过验证\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.c', '.cpp', '.cc', '.cxx', \\n '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.html', '.css', '.sh', \\n '.swift', '.kt', '.m', '.scala', '.vue', '.svelte', '.dart'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 提取文件中所有的非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n # 检查文件的最后5个非空行，以兼容可能存在的块注释结束符号(如 */, --> 等)\\n tail_lines = non_empty_lines[-5:] if non_empty_lines else []\\n \\n found = False\\n for line in tail_lines:\\n if target_text in line:\\n found = True\\n break\\n \\n # 如果该代码文件结尾处没有找到指定的注释内容，直接返回 False\\n if not found:\\n return False\\n except Exception:\\n # 如果存在不可读文件或编码问题，忽略即可\\n pass\\n \\n # 如果项目至少包含一个代码文件，并且都通过了检测，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我搭建一个律师案件管理仪表盘的前端基础骨架。需要包含一个左侧边栏，菜单项有：概览、案件进度、文件中心、管理员面板。顶部需要一个状态栏，展示系统Logo和当前登录账号（比如：李律师）。主内容区域默认展示“概览”页面，请在概览页放置四个简单的占位卡片（如：进行中案件数、本周开庭数、待审文件、新增客户），提供最基础的页面导航结构即可，暂时不需要复杂的业务逻辑交互。另外需要你注意，后续所有交互轮次都必须严格遵守以下要求：第一，每一个由你生成的代码文件必须精确包含3个箭头运算符(=>)；第二，你的任何用户回复都必须全程使用中文进行生成。","instruction_following_checklist":[{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成的代码文件中必须出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在生成的代码文件中体现出三元运算符的使用","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 覆盖推断可能的前端与后端相关代码文件类型\n code_extensions = {\n \".js\", \".jsx\", \".ts\", \".tsx\", \".vue\", \".html\", \".svelte\",\n \".py\", \".java\", \".cpp\", \".c\", \".cs\", \".go\", \".php\",\n \".rb\", \".swift\", \".kt\", \".rs\", \".dart\"\n }\n \n # 匹配 C-like 语言的三元运算符\n # (?![.:]) 前瞻否定，确保 ? 后面不紧跟 . 或 : （排除了可选链和紧凑的可选属性）\n # [^:?]*\\S[^:?]* 确保 ? 和 : 之间没有其他的 ? 和 :，并且至少包含一个非空字符（排除了中间带空格的可选属性形式）\n c_like_pattern = re.compile(r'\\?(?![.:])[^:?]*\\S[^:?]*:')\n \n # 匹配 Python 的三元运算符\n # 普通多行或单行的 if 语句必定包含冒号，而 Python 三元表达式没有冒号\n py_pattern = re.compile(r'\\bif\\b[^:]+\\belse\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n content = f.read()\n \n # 1. 清除多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 2. 清除 Python 等的多行字符串字面量，防止误判\n content = re.sub(r'\"\"\"(.*?)\"\"\"', '\"\"', content, flags=re.DOTALL)\n content = re.sub(r\"'''(.*?)'''\", \"''\", content, flags=re.DOTALL)\n \n # 3. 清除单行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n \n # 4. 清除各类单行字符串字面量的干扰\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n # 5. 分别应用相应的正则表达式来检查是否存在三元运算符\n if ext == \".py\":\n if py_pattern.search(content):\n return True\n else:\n if c_like_pattern.search(content):\n return True\n except Exception:\n continue\n \n return False","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义前端骨架常见代码文件的扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的目录\n for d in ['.git', 'node_modules']:\n if d in dirs:\n dirs.remove(d)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline().strip()\n # 判断首行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果无法正常读取（如编码错误或没有权限等），判定为不满足要求\n return False\n \n # 如果没有生成任何对应的代码文件，也认为不符合预期；反之则证明全部符合要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 指定要求包含的注释文本\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 前端基础骨架常用的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\n \n file_count = 0\n \n # 遍历项目地址中的文件\n for root, dirs, files in os.walk(workspace_path):\n # 排除通常无需检查的依赖目录及版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n # 过滤出所有非空行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n # 如果文件为空但被创建了，显然没有包含要求中的注释\n if not non_empty_lines:\n return False\n \n # 检查文件结尾（最后5个非空行），寻找是否存在该注释文本\n # 这样可以兼容多行注释标签（如 `-->` 或 `*/`）单独占据最后一行的情形\n found = False\n for line in non_empty_lines[-5:]:\n if target_text in line:\n found = True\n break\n \n # 一旦有生成的代码文件结尾缺失此注释，则判定为未满足要求\n if not found:\n return False\n \n file_count += 1\n \n # 必须保证有检查到代码文件且全都验证通过\n return file_count > 0","tags":["Content","Format"]},{"constraint":"每一个由你生成的代码文件必须精确包含3个箭头运算符(=>)","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉一些常见的非业务代码或忽略目录\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用utf-8编码读取文件，忽略不可解析的字符\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 统计箭头运算符的数量\n if content.count('=>') != 3:\n return False\n except Exception:\n return False\n \n # 如果工作区内存在至少一个代码文件，且它们都符合\"包含3个=>\"的要求，则返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]}]},{"round_id":1,"instruction":"框架看起来不错。接下来请实现“案件进度”页面的功能。我希望这个页面采用看板（Kanban）的形式。请创建四个垂直的列，分别代表案件阶段：前期准备、调查取证、庭审阶段、执行/结案。在这些列中生成几个样例案件卡片，卡片需显示案件编号、案件名称、负责律师和截止日期（标红显示如果快到期）。最重要的是，请实现卡片的状态流转交互：每个卡片上放一个“推进到下一阶段”的按钮，点击后该卡片会从当前列移动到右侧的下一个阶段列中；如果是处于最后一列的卡片，背景色请置灰并打上“已完结”的标记。另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“每一个由你生成的代码文件必须精确包含3个箭头运算符(=>)”替换为“你所输出的任意一个代码文件，都必须刚好存在1个\"TODO(human)\"标记”；第二，只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中。","instruction_following_checklist":[{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成的代码文件中必须出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在生成的代码文件中体现出三元运算符的使用","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查生成的代码项目中是否体现了三元运算符的使用。\n \"\"\"\n # 可能存在三元运算符的前端代码文件扩展名\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.vue', '.svelte'}\n \n # 核心正则表达式：\n # \\s\\?\\s : 匹配前后均为空白符的问号，精准过滤掉 ??, ?., ?: 等干扰项。\n # [\\s\\S]{1,2000}? : 匹配两者中间的内容，非贪婪模式，最高允许跨越2000个字符（适应JSX等复杂换行情况）。\n # :\\s : 匹配后接空白符的冒号，兼容标准的三元运算符格式。\n ternary_pattern = re.compile(r'\\s\\?\\s[\\s\\S]{1,2000}?:\\s')\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n\n # 预处理：安全地移除注释内容以防匹配到被注释掉的无用代码\n # 1. 移除多行块级注释与 HTML 注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 2. 移除单行注释\n # 优先移除整行单行注释\n content = re.sub(r'(?m)^\\s*//.*$', '', content)\n # 移除行内单行注释（要求 // 前面必须有空白符，防止误删 http:// 中的 //）\n content = re.sub(r'\\s//.*', '', content)\n\n # 执行匹配\n if ternary_pattern.search(content):\n return True\n \n except Exception:\n # 忽略无法读取或解析的文件\n continue\n\n return False","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_encoding_declaration(response: str, workspace_path: str) -> bool:\n # 推断的前端常见代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\n # 需要忽略的常见项目构建及缓存目录\n ignore_dirs = {'node_modules', '.git', 'dist', 'build', '.vscode', '.idea'}\n \n has_code_file = False\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 判断首行是否包含指定的编码声明字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果读取遇到异常（如编码错误或权限问题等），也视为验证失败\n return False\n \n # 如果遍历了相关代码文件且均符合要求，则返回True；否则如果在目录中没找到任何代码文件，则返回False。\n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 结合用户指令，目标项目为前端骨架项目，在此推断常见的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 剔除首尾的空白字符和空行，保证取到真实可见的最后一行\n content = f.read().strip()\n except Exception:\n # 忽略读取异常的文件（如非UTF-8格式的二进制文件等）\n continue\n \n # 如果是空文件，则认为不满足要求\n if not content:\n return False\n \n # 获取文件的最后一行\n lines = content.splitlines()\n last_line = lines[-1] if lines else \"\"\n \n # 兼容不同的注释符号，只需要检查最后一行是否包含目标子串即可\n if target_comment not in last_line:\n return False\n \n found_code_file = True\n \n # 只要成功匹配且工作区内确实存在代码文件即为成功\n return found_code_file","tags":["Content","Format"]},{"constraint":"你所输出的任意一个代码文件，都必须刚好存在1个\"TODO(human)\"标记","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的前端及其他代码文件扩展名，排除.txt等非代码文件\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.html', '.css', '.scss', '.less',\n '.py', '.java', '.c', '.cpp', '.h', \n '.go', '.rs', '.php', '.rb', '.swift'\n }\n \n # 定义需要跳过的无关目录\n exclude_dirs = {'node_modules', '.git', '__pycache__', '.vscode', '.idea', 'dist', 'build'}\n \n file_found = False\n for root, dirs, files in os.walk(workspace_path):\n # 过滤需要忽略的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查是否刚好存在1个\"TODO(human)\"标记\n if content.count('TODO(human)') != 1:\n return False\n except Exception:\n # 如果读取文件失败（例如编码问题），则忽略该文件\n continue\n \n # 如果没有找到任何代码文件，返回False；若全都刚好包含1个标记，则返回True\n return file_found","tags":["Content","Keyword"]},{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":2,"instruction":"很直观！现在我们来细化案件参与人的详情查看功能。请在当前的看板视图基础上增加交互：当我点击任意一个案件卡片的主体部分时，从屏幕右侧滑出一个详细信息抽屉（Drawer）。抽屉内需要使用标签页（Tabs）分为两块：“客户详情”和“律师详情”。“律师详情”里展示律师头像、执业证号以及一个简单的案件胜率进度条；“客户详情”里展示客户基础信息，并且在手机号旁边做一个“修改”小图标，点击后手机号文本变成一个输入框和保存按钮，点击保存后恢复文本显示并更新号码内容。另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“你所输出的任意一个代码文件，都必须刚好存在1个\"TODO(human)\"标记”替换为“你所输出的任意一个代码文件，都必须刚好存在2个\"TODO\"标记”；第二，无论哪一个生成的代码文件，都不可含有console.log调试语句。","instruction_following_checklist":[{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成的代码文件中必须出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在生成的代码文件中体现出三元运算符的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 支持验证的常见代码文件后缀列表\n allowed_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py', '.java', '.c', '.cpp', '.cs', '.php'}\n \n # 移除各类代码中的注释内容以防干扰检测\n def remove_comments(code: str) -> str:\n # 移除 HTML 风格注释\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n # 移除 JS/C/Java 风格多行注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除 JS/C/Java 风格单行注释\n code = re.sub(r'//.*', '', code)\n # 移除 Python 风格单行注释\n code = re.sub(r'#.*', '', code)\n return code\n\n # 清理只包含单纯文本的字符串，避免误判纯自然语言中的标点，并保留可能是Vue等模板内属性绑定的复杂情况\n def clean_strings_global(text: str) -> str:\n # 如果单双引号/反引号字符串中不同时包含 ? 和 : ，则它不大可能是三元运算符，将其替换为空\n text = re.sub(r'`([^`]*)`', lambda m: m.group(0) if '?' in m.group(1) and ':' in m.group(1) else '``', text, flags=re.DOTALL)\n text = re.sub(r'\"([^\"]*)\"', lambda m: m.group(0) if '?' in m.group(1) and ':' in m.group(1) else '\"\"', text)\n text = re.sub(r\"'([^']*)'\", lambda m: m.group(0) if '?' in m.group(1) and ':' in m.group(1) else \"''\", text)\n return text\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in allowed_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 对读取的代码内容进行注释和基础字符串剔除\n content = remove_comments(content)\n content = clean_strings_global(content)\n \n if ext == '.py':\n # 检测 Python 的三元运算符: a if condition else b\n # 要求是行内 if 和 else 且 if 后不存在普通控制语句块包含的冒号 :\n if re.search(r'\\S+\\s+if\\s+[^:]+\\s+else\\s+\\S+', content):\n return True\n else:\n # 检测 JS/TS/Vue 等前端的三元运算符: cond ? exprIfTrue : exprIfFalse\n # ? 和 : 之间应当包含非空白的其他字符以防止误匹配 TS 可选属性声明（如 a?: string）\n if re.search(r'\\?\\s*[^:;\\s][^:;]*:', content):\n return True\n except Exception:\n # 若遇到文件读取或编码异常则忽略，继续排查后续文件\n continue\n \n return False","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义前端骨架及常见通用项目可能出现的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css'}\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的无关目录，避免误判\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'dist', 'build'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查特定要求的编码声明字符串是否包含在代码文件的第一行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于某些原因文件读取失败，无法验证是否满足要求，保守返回 False\n return False\n \n # 如果项目中没有任何生成的代码文件，也视作未实现要求的验证逻辑\n return has_code_files","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 推断前端项目常见的代码文件类型\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', \n '.html', '.css', '.scss', '.less', '.vue'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n code_files = []\n # 遍历工作空间目录\n for root, dirs, files in os.walk(workspace_path):\n # 排除通常的依赖库或隐藏目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 若没有找到任何代码文件，则无法证明满足了要求\n if not code_files:\n return False\n \n # 检查每一个代码文件的结尾\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n # 过滤获取所有非空行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n if not non_empty_lines:\n return False\n \n has_target_comment = False\n # 检查最后 5 个非空行以兼容可能的多行注释包裹符号导致偏移的情况\n for line in non_empty_lines[-5:]:\n if target_text in line:\n has_target_comment = True\n break\n \n # 只要存在某一个生成的代码文件不具备该结尾注释即验证失败\n if not has_target_comment:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"你所输出的任意一个代码文件，都必须刚好存在2个\"TODO\"标记","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对前端项目定义可能的代码文件后缀名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\n code_file_found = False\n \n # 如果工作目录不存在，返回 False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的依赖和版本控制文件夹\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n code_file_found = True\n \n # 统计 \"TODO\" 标记出现的次数\n if content.count(\"TODO\") != 2:\n return False\n \n # 必须有至少一个代码文件被验证，否则视为不合规\n if not code_file_found:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论哪一个生成的代码文件，都不可含有console.log调试语句","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 若工作空间不存在，说明可能没有生成文件，默认不违反不可含有的约束\n if not os.path.isdir(workspace_path):\n return True\n\n # 基于前端项目可能生成的文件推测出的代码文件类型扩展名集合\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', '.mjs', '.cjs', '.svelte'}\n # 排除常见的第三方库或构建产物目录，避免误检\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', '.next', '.nuxt'}\n\n # 编译正则表达式，允许 console 和 log 以及点操作符之间存在空白符\n pattern = re.compile(r'console\\s*\\.\\s*log')\n\n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 列表，以便 os.walk 自动跳过被排除的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if pattern.search(content):\n # 只要发现存在 console.log 语句，则验证失败，不满足要求\n return False\n except Exception:\n # 如果由于读取权限或非文本内容等原因无法正常读取文件，则跳过\n continue\n\n # 所有目标文件遍历完毕均未发现 console.log 语句，满足要求\n return True","tags":["Content","Keyword"]}]},{"round_id":3,"instruction":"抽屉交互做得很好。现在请切换到“文件中心”页面的开发。左侧实现一个案件文件夹的树状导航（例如展开某个案件可以看到：诉讼材料、证据清单、身份证明）。右侧是当前选中文件夹的文件列表。请在列表顶部实现一个“上传材料”按钮，点击后弹出一个居中的模态框（Modal）。模态框内需要包含一个虚线框的拖拽上传区域，以及一个单选框用来标记该文件属于“公开材料”还是“保密材料”。另外，文件列表的每一行加一个“标记为核心证据”的按钮，点击后该行前面会出现一个黄色的高亮星标。另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“你所输出的任意一个代码文件，都必须刚好存在2个\"TODO\"标记”替换为“你所输出的任意一个代码文件，都必须刚好存在3个\"待优化\"标记”；第二，每次内容更新完毕后，需要更新Review.md文件以审阅现有代码，必须遵照此格式输出：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```，以上两项要求后续所有轮次都必须严格遵守。","instruction_following_checklist":[{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成的代码文件中必须出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在生成的代码文件中体现出三元运算符的使用","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令提及的“前端基础骨架”、“模态框”等推断主要为前端项目\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.py'}\n \n # Python 三元运算符模式： value if condition else value\n py_ternary_pattern = re.compile(r'\\b\\S+\\s+if\\s+.+?\\s+else\\s+')\n \n # JS/TS 等 C-Style 语言的三元运算符模式\n # 模式1: 规范的带空格写法，排除可选链 `?.`\n js_ternary_pattern_strict = re.compile(r'\\s\\?(?!\\.)\\s+[^;]+?\\s:\\s')\n \n # 模式2: 宽松的无空格写法兼容匹配。排除可选属性 `?:`，空值合并 `??` 以及可选链 `?.`\n js_ternary_pattern_loose = re.compile(r'[\\w\\])]\\s*\\?(?![.\\?:])[^;]+?:\\s*[\\w\\[{(<]')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 在匹配前清理常见的注释内容以免造成文本或字符串误匹配\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n \n if ext == '.py':\n content = re.sub(r'#.*', '', content)\n if py_ternary_pattern.search(content):\n return True\n else:\n if js_ternary_pattern_strict.search(content) or js_ternary_pattern_loose.search(content):\n return True\n except Exception:\n # 忽略无法通过 utf-8 读取等异常的文件\n continue\n \n return False","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 常见的前端及常用后端代码文件后缀\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', \n '.html', '.css', '.scss', '.less',\n '.py', '.java', '.c', '.cpp', '.h', '.go', '.rs'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的第三方依赖和打包输出目录\n dirs[:] = [d for d in dirs if d not in (\n 'node_modules', '.git', 'dist', 'build', \n '__pycache__', '.next', '.nuxt', 'coverage'\n )]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 尝试读取所有行\n lines = f.readlines()\n except Exception:\n # 如果由于特殊编码等原因无法读取文本，可选择跳过\n continue\n \n # 如果代码文件为空，则不满足在首行写入要求的情况\n if not lines:\n return False\n \n # 检查文件第一行内容\n first_line = lines[0]\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n \n has_code_file = True\n\n # 只要存在至少一个代码文件且都满足条件，即返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据任务上下文推断可能的前端及常见代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.py'}\n \n found_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉无关的目录，防止扫描依赖及构建缓存等\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__', 'dist', 'build')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果文件无法以 utf-8 读取，通常说明文件存在异常，算作不满足条件\n return False\n \n # 过滤掉文件中的所有纯空白行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n if not non_empty_lines:\n return False\n \n # 检查文件尾部（取最后 5 个非空行）是否存在目标文本\n # 此做法是为了防止因为遇到多行注释结束符号占位而导致的判断偏差\n found_target = False\n for line in non_empty_lines[-5:]:\n if target_text in line:\n found_target = True\n break\n \n if not found_target:\n # 如果发现哪怕一个代码文件末尾没有标注，即判断为未满足该要求\n return False\n \n # 确保检查了至少一个有效的代码文件\n return found_code_file","tags":["Content","Format"]},{"constraint":"你所输出的任意一个代码文件，都必须刚好存在3个\"待优化\"标记","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的前端及后端代码文件扩展名，排除.md, .txt等文档及纯文本文件\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \n '.css', '.scss', '.less', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift'\n }\n \n code_files_found = False\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的第三方依赖或版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 筛选代码文件\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计 \"待优化\" 标记在代码文件内容中出现的次数\n if content.count(\"待优化\") != 3:\n return False\n except Exception:\n # 如果文件无法正常读取（例如编码不支持），则视为不满足条件\n return False\n \n # 如果全部代码文件均刚好包含3个标记，并且确实存在至少一个代码文件，则判定通过\n return code_files_found","tags":["Content","Keyword"]},{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论哪一个生成的代码文件，都不可含有console.log调试语句","validation_code":"def check_no_console_log(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义前端项目中常见的代码文件扩展名白名单\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', '.svelte', '.mjs', '.cjs'}\n \n # 定义正则表达式，匹配 console.log 语句调用（包含可能的空格，如 console.log (...)）\n pattern = re.compile(r'console\\.log\\s*\\(')\n \n # 如果工作区路径不存在，视作未生成违规文件，返回 True\n if not os.path.isdir(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名并统一转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 只要在任意代码文件中匹配到 console.log 调试语句，立刻判定为不满足要求\n if pattern.search(content):\n return False\n except Exception:\n # 忽略特殊编码或因权限问题无法读取的文件\n continue\n \n # 如果遍历检查完所有代码文件均未发现 console.log 语句，则代表满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"每次内容更新完毕后，需要更新Review.md文件以审阅现有代码，必须遵照此格式输出：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":4,"instruction":"最后，我们来完成“管理员面板”页面，这主要供律所合伙人使用。页面上半部分请用纯CSS或者简单的图表库模拟一个“律师案件负荷统计”，用水平柱状图展示三四名律师当前手里负责的案件数量。页面的下半部分是一个权限管理表格，列出律师姓名、职务。核心交互是：表格中有一列“案件查看权限”，使用下拉选择框（Select）展示，选项包括“查看全所”、“查看本组”、“仅看自己”。当我修改任意一名律师的下拉框选项时，屏幕顶部要滑出一个绿色的全局提示（Toast）“权限设置已更新”，这个提示要在显示 3 秒后自动消失。\n另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“你所输出的任意一个代码文件，都必须刚好存在3个\"待优化\"标记”替换为“你所输出的任意一个代码文件，都必须刚好存在2个\"// FIXME\"标记”；第二，请保证每一个生成的代码文件，连同空行在内的总行数均符合质数的要求，以上要求后续所有轮次都必须严格遵守。","instruction_following_checklist":[{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"要求在生成的代码文件中必须出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在生成的代码文件中体现出三元运算符的使用","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端和后端代码文件后缀\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.svelte', '.py', '.php', '.java', '.c', '.cpp', '.cs'}\n \n # 针对 JS/TS/Java/C/C++ 等的三元运算符： condition ? expr1 : expr2\n # 特征点：\n # 1. 问号前不能是 ? 或 . （排除空值合并 ?? 和可选链 ?.)\n # 2. 问号后不能是 ? 或 . 或 : (排除 ??、?. 和 TS的可选属性连写 ?:)\n # 3. 问号和冒号之间应存在至少一个非空字符（保证是有效表达式），最多匹配 500 个字符防止无意义的跨段误判。\n js_ternary_pattern = re.compile(r'(?<!\\?|\\.)\\?(?!\\?|\\.|:)\\s*(?:\\S(?:[\\s\\S]{0,500}?))\\s*:')\n \n # 针对 Python 的三元运算符： expr1 if condition else expr2\n # 特征点：要求单行内包含非空字符 + 空白 + if + 空白 + 条件 + 空白 + else + 空白 + 非空字符\n py_ternary_pattern = re.compile(r'\\S+[^\\n]*?\\s+if\\s+[^\\n]+?\\s+else\\s+[^\\n]*?\\S+')\n \n def remove_strings_and_comments(code: str) -> str:\n # 移除单行注释\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'#.*', '', code)\n # 移除多行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code, flags=re.DOTALL)\n # 简单移除成对的字符串，防止字符串里包含虚假的 \" ? : \" 影响判断\n code = re.sub(r'\".*?\"', '\"\"', code)\n code = re.sub(r\"'.*?'\", \"''\", code)\n code = re.sub(r\"`[\\s\\S]*?`\", \"``\", code)\n return code\n\n # 遍历项目工作区\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 去除注释和字符串常量后再进行判断\n clean_content = remove_strings_and_comments(content)\n \n if js_ternary_pattern.search(clean_content) or py_ternary_pattern.search(clean_content):\n return True\n except Exception:\n continue\n \n return False","tags":["Style","Paradigm"]},{"constraint":"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，仪表盘前端项目可能的代码文件后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\n # 忽略常见的依赖和产物目录\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__'}\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n \n # 检查首行是否包含指定的编码声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件读取出现异常，则视为未满足要求\n return False\n \n # 要求生成项目中必须包含至少一个代码文件，且所有代码文件均满足首行声明要求\n return found_code_file","tags":["Content","Format"]},{"constraint":"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 针对前端项目的典型代码文件类型推断\n valid_extensions = {\".html\", \".css\", \".js\", \".jsx\", \".ts\", \".tsx\", \".vue\"}\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉无关的目录，避免扫描第三方依赖或版本控制文件夹\n dirs[:] = [d for d in dirs if d not in (\"node_modules\", \".git\", \"dist\", \"build\")]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n lines = f.readlines()\n except Exception:\n # 如果由于任何原因无法读取代码文件，认为验证不通过\n return False\n \n # 过滤掉完全为空白内容的行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n if not non_empty_lines:\n return False\n \n # 检查倒数 3 个非空行中是否包含目标注释内容\n # 此逻辑可以很好地兼容 \"*/\"、\"-->\" 等结束符号单独作为最后一行的场景\n has_target = False\n for line in non_empty_lines[-3:]:\n if target_str in line:\n has_target = True\n break\n \n if not has_target:\n return False\n\n return found_code_file","tags":["Content","Format"]},{"constraint":"你所输出的任意一个代码文件，都必须刚好存在2个\"// FIXME\"标记","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令（前端仪表盘、看板、抽屉、拖拽等），推断可能的代码文件类型\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的常规依赖或版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 检查 \"// FIXME\" 标记是否刚好存在 2 个\n if content.count('// FIXME') != 2:\n return False\n except Exception:\n # 如果文件存在但无法读取，视为验证失败\n return False\n\n # 必须至少有一个代码文件生成且均符合要求才算满足\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"无论哪一个生成的代码文件，都不可含有console.log调试语句","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令的前端项目背景，推断可能的代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.svelte'}\n \n if not os.path.exists(workspace_path):\n # 如果工作区目录不存在，可以认为尚未违规，或者视具体测试框架逻辑而定，此处默认为 True\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的第三方依赖库及编译产物目录，避免误判\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.next', '.nuxt')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若发现 console.log，则说明不满足要求\n if 'console.log' in content:\n return False\n except Exception:\n # 遇到读取异常时跳过（防止因编码问题导致的崩溃）\n continue\n \n # 如果所有代码文件均不包含 console.log，则满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"每次内容更新完毕后，需要更新Review.md文件以审阅现有代码，必须遵照此格式输出：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请保证每一个生成的代码文件，连同空行在内的总行数均符合质数的要求","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 针对前端骨架项目的常见代码后缀\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css'}\n file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除前端依赖和版本控制目录，避免干扰\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用换行符分割能够精确计入末尾空行或连续空行\n lines = content.split('\\n')\n line_count = len(lines)\n \n # 检查行数是否为质数\n if not is_prime(line_count):\n return False\n except Exception:\n # 遇到非UTF-8等异常无法读取的情况，安全跳过\n continue\n \n # 如果找到了相关的代码文件并全都符合质数要求，则返回 True\n return file_found","tags":["Quantity","Exact Value"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我搭建一个律师案件管理仪表盘的前端基础骨架。需要包含一个左侧边栏，菜单项有：概览、案件进度、文件中心、管理员面板。顶部需要一个状态栏，展示系统Logo和当前登录账号（比如：李律师）。主内容区域默认展示“概览”页面，请在概览页放置四个简单的占位卡片（如：进行中案件数、本周开庭数、待审文件、新增客户），提供最基础的页面导航结构即可，暂时不需要复杂的业务逻辑交互。另外需要你注意，后续所有交互轮次都必须严格遵守以下要求：第一，每一个由你生成的代码文件必须精确包含3个箭头运算符(=>)；第二，你的任何用户回复都必须全程使用中文进行生成。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成的代码文件中必须出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在生成的代码文件中体现出三元运算符的使用\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 覆盖推断可能的前端与后端相关代码文件类型\\n code_extensions = {\\n \\\".js\\\", \\\".jsx\\\", \\\".ts\\\", \\\".tsx\\\", \\\".vue\\\", \\\".html\\\", \\\".svelte\\\",\\n \\\".py\\\", \\\".java\\\", \\\".cpp\\\", \\\".c\\\", \\\".cs\\\", \\\".go\\\", \\\".php\\\",\\n \\\".rb\\\", \\\".swift\\\", \\\".kt\\\", \\\".rs\\\", \\\".dart\\\"\\n }\\n \\n # 匹配 C-like 语言的三元运算符\\n # (?![.:]) 前瞻否定，确保 ? 后面不紧跟 . 或 : （排除了可选链和紧凑的可选属性）\\n # [^:?]*\\\\S[^:?]* 确保 ? 和 : 之间没有其他的 ? 和 :，并且至少包含一个非空字符（排除了中间带空格的可选属性形式）\\n c_like_pattern = re.compile(r'\\\\?(?![.:])[^:?]*\\\\S[^:?]*:')\\n \\n # 匹配 Python 的三元运算符\\n # 普通多行或单行的 if 语句必定包含冒号，而 Python 三元表达式没有冒号\\n py_pattern = re.compile(r'\\\\bif\\\\b[^:]+\\\\belse\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n content = f.read()\\n \\n # 1. 清除多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 2. 清除 Python 等的多行字符串字面量，防止误判\\n content = re.sub(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', '\\\"\\\"', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'''(.*?)'''\\\", \\\"''\\\", content, flags=re.DOTALL)\\n \\n # 3. 清除单行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n \\n # 4. 清除各类单行字符串字面量的干扰\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n # 5. 分别应用相应的正则表达式来检查是否存在三元运算符\\n if ext == \\\".py\\\":\\n if py_pattern.search(content):\\n return True\\n else:\\n if c_like_pattern.search(content):\\n return True\\n except Exception:\\n continue\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义前端骨架常见代码文件的扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的目录\\n for d in ['.git', 'node_modules']:\\n if d in dirs:\\n dirs.remove(d)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline().strip()\\n # 判断首行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果无法正常读取（如编码错误或没有权限等），判定为不满足要求\\n return False\\n \\n # 如果没有生成任何对应的代码文件，也认为不符合预期；反之则证明全部符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 指定要求包含的注释文本\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 前端基础骨架常用的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\\n \\n file_count = 0\\n \\n # 遍历项目地址中的文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除通常无需检查的依赖目录及版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n # 过滤出所有非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n # 如果文件为空但被创建了，显然没有包含要求中的注释\\n if not non_empty_lines:\\n return False\\n \\n # 检查文件结尾（最后5个非空行），寻找是否存在该注释文本\\n # 这样可以兼容多行注释标签（如 `-->` 或 `*/`）单独占据最后一行的情形\\n found = False\\n for line in non_empty_lines[-5:]:\\n if target_text in line:\\n found = True\\n break\\n \\n # 一旦有生成的代码文件结尾缺失此注释，则判定为未满足要求\\n if not found:\\n return False\\n \\n file_count += 1\\n \\n # 必须保证有检查到代码文件且全都验证通过\\n return file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件必须精确包含3个箭头运算符(=>)\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉一些常见的非业务代码或忽略目录\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用utf-8编码读取文件，忽略不可解析的字符\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 统计箭头运算符的数量\\n if content.count('=>') != 3:\\n return False\\n except Exception:\\n return False\\n \\n # 如果工作区内存在至少一个代码文件，且它们都符合\\\"包含3个=>\\\"的要求，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"框架看起来不错。接下来请实现“案件进度”页面的功能。我希望这个页面采用看板（Kanban）的形式。请创建四个垂直的列，分别代表案件阶段：前期准备、调查取证、庭审阶段、执行/结案。在这些列中生成几个样例案件卡片，卡片需显示案件编号、案件名称、负责律师和截止日期（标红显示如果快到期）。最重要的是，请实现卡片的状态流转交互：每个卡片上放一个“推进到下一阶段”的按钮，点击后该卡片会从当前列移动到右侧的下一个阶段列中；如果是处于最后一列的卡片，背景色请置灰并打上“已完结”的标记。另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“每一个由你生成的代码文件必须精确包含3个箭头运算符(=>)”替换为“你所输出的任意一个代码文件，都必须刚好存在1个\\\"TODO(human)\\\"标记”；第二，只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成的代码文件中必须出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在生成的代码文件中体现出三元运算符的使用\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查生成的代码项目中是否体现了三元运算符的使用。\\n \\\"\\\"\\\"\\n # 可能存在三元运算符的前端代码文件扩展名\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.html', '.vue', '.svelte'}\\n \\n # 核心正则表达式：\\n # \\\\s\\\\?\\\\s : 匹配前后均为空白符的问号，精准过滤掉 ??, ?., ?: 等干扰项。\\n # [\\\\s\\\\S]{1,2000}? : 匹配两者中间的内容，非贪婪模式，最高允许跨越2000个字符（适应JSX等复杂换行情况）。\\n # :\\\\s : 匹配后接空白符的冒号，兼容标准的三元运算符格式。\\n ternary_pattern = re.compile(r'\\\\s\\\\?\\\\s[\\\\s\\\\S]{1,2000}?:\\\\s')\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n # 预处理：安全地移除注释内容以防匹配到被注释掉的无用代码\\n # 1. 移除多行块级注释与 HTML 注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 2. 移除单行注释\\n # 优先移除整行单行注释\\n content = re.sub(r'(?m)^\\\\s*//.*$', '', content)\\n # 移除行内单行注释（要求 // 前面必须有空白符，防止误删 http:// 中的 //）\\n content = re.sub(r'\\\\s//.*', '', content)\\n\\n # 执行匹配\\n if ternary_pattern.search(content):\\n return True\\n \\n except Exception:\\n # 忽略无法读取或解析的文件\\n continue\\n\\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n # 推断的前端常见代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\\n # 需要忽略的常见项目构建及缓存目录\\n ignore_dirs = {'node_modules', '.git', 'dist', 'build', '.vscode', '.idea'}\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 判断首行是否包含指定的编码声明字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果读取遇到异常（如编码错误或权限问题等），也视为验证失败\\n return False\\n \\n # 如果遍历了相关代码文件且均符合要求，则返回True；否则如果在目录中没找到任何代码文件，则返回False。\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 结合用户指令，目标项目为前端骨架项目，在此推断常见的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 剔除首尾的空白字符和空行，保证取到真实可见的最后一行\\n content = f.read().strip()\\n except Exception:\\n # 忽略读取异常的文件（如非UTF-8格式的二进制文件等）\\n continue\\n \\n # 如果是空文件，则认为不满足要求\\n if not content:\\n return False\\n \\n # 获取文件的最后一行\\n lines = content.splitlines()\\n last_line = lines[-1] if lines else \\\"\\\"\\n \\n # 兼容不同的注释符号，只需要检查最后一行是否包含目标子串即可\\n if target_comment not in last_line:\\n return False\\n \\n found_code_file = True\\n \\n # 只要成功匹配且工作区内确实存在代码文件即为成功\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你所输出的任意一个代码文件，都必须刚好存在1个\\\"TODO(human)\\\"标记\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的前端及其他代码文件扩展名，排除.txt等非代码文件\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.html', '.css', '.scss', '.less',\\n '.py', '.java', '.c', '.cpp', '.h', \\n '.go', '.rs', '.php', '.rb', '.swift'\\n }\\n \\n # 定义需要跳过的无关目录\\n exclude_dirs = {'node_modules', '.git', '__pycache__', '.vscode', '.idea', 'dist', 'build'}\\n \\n file_found = False\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤需要忽略的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查是否刚好存在1个\\\"TODO(human)\\\"标记\\n if content.count('TODO(human)') != 1:\\n return False\\n except Exception:\\n # 如果读取文件失败（例如编码问题），则忽略该文件\\n continue\\n \\n # 如果没有找到任何代码文件，返回False；若全都刚好包含1个标记，则返回True\\n return file_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"很直观！现在我们来细化案件参与人的详情查看功能。请在当前的看板视图基础上增加交互：当我点击任意一个案件卡片的主体部分时，从屏幕右侧滑出一个详细信息抽屉（Drawer）。抽屉内需要使用标签页（Tabs）分为两块：“客户详情”和“律师详情”。“律师详情”里展示律师头像、执业证号以及一个简单的案件胜率进度条；“客户详情”里展示客户基础信息，并且在手机号旁边做一个“修改”小图标，点击后手机号文本变成一个输入框和保存按钮，点击保存后恢复文本显示并更新号码内容。另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“你所输出的任意一个代码文件，都必须刚好存在1个\\\"TODO(human)\\\"标记”替换为“你所输出的任意一个代码文件，都必须刚好存在2个\\\"TODO\\\"标记”；第二，无论哪一个生成的代码文件，都不可含有console.log调试语句。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成的代码文件中必须出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在生成的代码文件中体现出三元运算符的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 支持验证的常见代码文件后缀列表\\n allowed_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py', '.java', '.c', '.cpp', '.cs', '.php'}\\n \\n # 移除各类代码中的注释内容以防干扰检测\\n def remove_comments(code: str) -> str:\\n # 移除 HTML 风格注释\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n # 移除 JS/C/Java 风格多行注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除 JS/C/Java 风格单行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除 Python 风格单行注释\\n code = re.sub(r'#.*', '', code)\\n return code\\n\\n # 清理只包含单纯文本的字符串，避免误判纯自然语言中的标点，并保留可能是Vue等模板内属性绑定的复杂情况\\n def clean_strings_global(text: str) -> str:\\n # 如果单双引号/反引号字符串中不同时包含 ? 和 : ，则它不大可能是三元运算符，将其替换为空\\n text = re.sub(r'`([^`]*)`', lambda m: m.group(0) if '?' in m.group(1) and ':' in m.group(1) else '``', text, flags=re.DOTALL)\\n text = re.sub(r'\\\"([^\\\"]*)\\\"', lambda m: m.group(0) if '?' in m.group(1) and ':' in m.group(1) else '\\\"\\\"', text)\\n text = re.sub(r\\\"'([^']*)'\\\", lambda m: m.group(0) if '?' in m.group(1) and ':' in m.group(1) else \\\"''\\\", text)\\n return text\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in allowed_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 对读取的代码内容进行注释和基础字符串剔除\\n content = remove_comments(content)\\n content = clean_strings_global(content)\\n \\n if ext == '.py':\\n # 检测 Python 的三元运算符: a if condition else b\\n # 要求是行内 if 和 else 且 if 后不存在普通控制语句块包含的冒号 :\\n if re.search(r'\\\\S+\\\\s+if\\\\s+[^:]+\\\\s+else\\\\s+\\\\S+', content):\\n return True\\n else:\\n # 检测 JS/TS/Vue 等前端的三元运算符: cond ? exprIfTrue : exprIfFalse\\n # ? 和 : 之间应当包含非空白的其他字符以防止误匹配 TS 可选属性声明（如 a?: string）\\n if re.search(r'\\\\?\\\\s*[^:;\\\\s][^:;]*:', content):\\n return True\\n except Exception:\\n # 若遇到文件读取或编码异常则忽略，继续排查后续文件\\n continue\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义前端骨架及常见通用项目可能出现的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css'}\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的无关目录，避免误判\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'dist', 'build'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查特定要求的编码声明字符串是否包含在代码文件的第一行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于某些原因文件读取失败，无法验证是否满足要求，保守返回 False\\n return False\\n \\n # 如果项目中没有任何生成的代码文件，也视作未实现要求的验证逻辑\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 推断前端项目常见的代码文件类型\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', \\n '.html', '.css', '.scss', '.less', '.vue'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files = []\\n # 遍历工作空间目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除通常的依赖库或隐藏目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 若没有找到任何代码文件，则无法证明满足了要求\\n if not code_files:\\n return False\\n \\n # 检查每一个代码文件的结尾\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n # 过滤获取所有非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n if not non_empty_lines:\\n return False\\n \\n has_target_comment = False\\n # 检查最后 5 个非空行以兼容可能的多行注释包裹符号导致偏移的情况\\n for line in non_empty_lines[-5:]:\\n if target_text in line:\\n has_target_comment = True\\n break\\n \\n # 只要存在某一个生成的代码文件不具备该结尾注释即验证失败\\n if not has_target_comment:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你所输出的任意一个代码文件，都必须刚好存在2个\\\"TODO\\\"标记\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对前端项目定义可能的代码文件后缀名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.less'}\\n code_file_found = False\\n \\n # 如果工作目录不存在，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的依赖和版本控制文件夹\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n code_file_found = True\\n \\n # 统计 \\\"TODO\\\" 标记出现的次数\\n if content.count(\\\"TODO\\\") != 2:\\n return False\\n \\n # 必须有至少一个代码文件被验证，否则视为不合规\\n if not code_file_found:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论哪一个生成的代码文件，都不可含有console.log调试语句\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 若工作空间不存在，说明可能没有生成文件，默认不违反不可含有的约束\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 基于前端项目可能生成的文件推测出的代码文件类型扩展名集合\\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', '.mjs', '.cjs', '.svelte'}\\n # 排除常见的第三方库或构建产物目录，避免误检\\n exclude_dirs = {'node_modules', '.git', 'dist', 'build', '.next', '.nuxt'}\\n\\n # 编译正则表达式，允许 console 和 log 以及点操作符之间存在空白符\\n pattern = re.compile(r'console\\\\s*\\\\.\\\\s*log')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 列表，以便 os.walk 自动跳过被排除的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if pattern.search(content):\\n # 只要发现存在 console.log 语句，则验证失败，不满足要求\\n return False\\n except Exception:\\n # 如果由于读取权限或非文本内容等原因无法正常读取文件，则跳过\\n continue\\n\\n # 所有目标文件遍历完毕均未发现 console.log 语句，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"抽屉交互做得很好。现在请切换到“文件中心”页面的开发。左侧实现一个案件文件夹的树状导航（例如展开某个案件可以看到：诉讼材料、证据清单、身份证明）。右侧是当前选中文件夹的文件列表。请在列表顶部实现一个“上传材料”按钮，点击后弹出一个居中的模态框（Modal）。模态框内需要包含一个虚线框的拖拽上传区域，以及一个单选框用来标记该文件属于“公开材料”还是“保密材料”。另外，文件列表的每一行加一个“标记为核心证据”的按钮，点击后该行前面会出现一个黄色的高亮星标。另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“你所输出的任意一个代码文件，都必须刚好存在2个\\\"TODO\\\"标记”替换为“你所输出的任意一个代码文件，都必须刚好存在3个\\\"待优化\\\"标记”；第二，每次内容更新完毕后，需要更新Review.md文件以审阅现有代码，必须遵照此格式输出：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```，以上两项要求后续所有轮次都必须严格遵守。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成的代码文件中必须出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在生成的代码文件中体现出三元运算符的使用\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令提及的“前端基础骨架”、“模态框”等推断主要为前端项目\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.scss', '.py'}\\n \\n # Python 三元运算符模式： value if condition else value\\n py_ternary_pattern = re.compile(r'\\\\b\\\\S+\\\\s+if\\\\s+.+?\\\\s+else\\\\s+')\\n \\n # JS/TS 等 C-Style 语言的三元运算符模式\\n # 模式1: 规范的带空格写法，排除可选链 `?.`\\n js_ternary_pattern_strict = re.compile(r'\\\\s\\\\?(?!\\\\.)\\\\s+[^;]+?\\\\s:\\\\s')\\n \\n # 模式2: 宽松的无空格写法兼容匹配。排除可选属性 `?:`，空值合并 `??` 以及可选链 `?.`\\n js_ternary_pattern_loose = re.compile(r'[\\\\w\\\\])]\\\\s*\\\\?(?![.\\\\?:])[^;]+?:\\\\s*[\\\\w\\\\[{(<]')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 在匹配前清理常见的注释内容以免造成文本或字符串误匹配\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n \\n if ext == '.py':\\n content = re.sub(r'#.*', '', content)\\n if py_ternary_pattern.search(content):\\n return True\\n else:\\n if js_ternary_pattern_strict.search(content) or js_ternary_pattern_loose.search(content):\\n return True\\n except Exception:\\n # 忽略无法通过 utf-8 读取等异常的文件\\n continue\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 常见的前端及常用后端代码文件后缀\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', \\n '.html', '.css', '.scss', '.less',\\n '.py', '.java', '.c', '.cpp', '.h', '.go', '.rs'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的第三方依赖和打包输出目录\\n dirs[:] = [d for d in dirs if d not in (\\n 'node_modules', '.git', 'dist', 'build', \\n '__pycache__', '.next', '.nuxt', 'coverage'\\n )]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 尝试读取所有行\\n lines = f.readlines()\\n except Exception:\\n # 如果由于特殊编码等原因无法读取文本，可选择跳过\\n continue\\n \\n # 如果代码文件为空，则不满足在首行写入要求的情况\\n if not lines:\\n return False\\n \\n # 检查文件第一行内容\\n first_line = lines[0]\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n \\n has_code_file = True\\n\\n # 只要存在至少一个代码文件且都满足条件，即返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据任务上下文推断可能的前端及常见代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css', '.py'}\\n \\n found_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉无关的目录，防止扫描依赖及构建缓存等\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', '__pycache__', 'dist', 'build')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果文件无法以 utf-8 读取，通常说明文件存在异常，算作不满足条件\\n return False\\n \\n # 过滤掉文件中的所有纯空白行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n if not non_empty_lines:\\n return False\\n \\n # 检查文件尾部（取最后 5 个非空行）是否存在目标文本\\n # 此做法是为了防止因为遇到多行注释结束符号占位而导致的判断偏差\\n found_target = False\\n for line in non_empty_lines[-5:]:\\n if target_text in line:\\n found_target = True\\n break\\n \\n if not found_target:\\n # 如果发现哪怕一个代码文件末尾没有标注，即判断为未满足该要求\\n return False\\n \\n # 确保检查了至少一个有效的代码文件\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你所输出的任意一个代码文件，都必须刚好存在3个\\\"待优化\\\"标记\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的前端及后端代码文件扩展名，排除.md, .txt等文档及纯文本文件\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \\n '.css', '.scss', '.less', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift'\\n }\\n \\n code_files_found = False\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的第三方依赖或版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 筛选代码文件\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计 \\\"待优化\\\" 标记在代码文件内容中出现的次数\\n if content.count(\\\"待优化\\\") != 3:\\n return False\\n except Exception:\\n # 如果文件无法正常读取（例如编码不支持），则视为不满足条件\\n return False\\n \\n # 如果全部代码文件均刚好包含3个标记，并且确实存在至少一个代码文件，则判定通过\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论哪一个生成的代码文件，都不可含有console.log调试语句\",\n \"validation_code\": \"def check_no_console_log(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义前端项目中常见的代码文件扩展名白名单\\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', '.svelte', '.mjs', '.cjs'}\\n \\n # 定义正则表达式，匹配 console.log 语句调用（包含可能的空格，如 console.log (...)）\\n pattern = re.compile(r'console\\\\.log\\\\s*\\\\(')\\n \\n # 如果工作区路径不存在，视作未生成违规文件，返回 True\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名并统一转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 只要在任意代码文件中匹配到 console.log 调试语句，立刻判定为不满足要求\\n if pattern.search(content):\\n return False\\n except Exception:\\n # 忽略特殊编码或因权限问题无法读取的文件\\n continue\\n \\n # 如果遍历检查完所有代码文件均未发现 console.log 语句，则代表满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每次内容更新完毕后，需要更新Review.md文件以审阅现有代码，必须遵照此格式输出：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后，我们来完成“管理员面板”页面，这主要供律所合伙人使用。页面上半部分请用纯CSS或者简单的图表库模拟一个“律师案件负荷统计”，用水平柱状图展示三四名律师当前手里负责的案件数量。页面的下半部分是一个权限管理表格，列出律师姓名、职务。核心交互是：表格中有一列“案件查看权限”，使用下拉选择框（Select）展示，选项包括“查看全所”、“查看本组”、“仅看自己”。当我修改任意一名律师的下拉框选项时，屏幕顶部要滑出一个绿色的全局提示（Toast）“权限设置已更新”，这个提示要在显示 3 秒后自动消失。\\n另外需要注意，从本轮起后续所有交互轮次需遵守两项更新要求：第一，原要求“你所输出的任意一个代码文件，都必须刚好存在3个\\\"待优化\\\"标记”替换为“你所输出的任意一个代码文件，都必须刚好存在2个\\\"// FIXME\\\"标记”；第二，请保证每一个生成的代码文件，连同空行在内的总行数均符合质数的要求，以上要求后续所有轮次都必须严格遵守。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤。当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在编写代码时，必须保证最短路径实现且无过度设计，严禁私自附加例如兜底、降级方案等不在需求范围内的处理逻辑\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"要求在生成的代码文件中必须出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在生成的代码文件中体现出三元运算符的使用\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端和后端代码文件后缀\\n code_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.svelte', '.py', '.php', '.java', '.c', '.cpp', '.cs'}\\n \\n # 针对 JS/TS/Java/C/C++ 等的三元运算符： condition ? expr1 : expr2\\n # 特征点：\\n # 1. 问号前不能是 ? 或 . （排除空值合并 ?? 和可选链 ?.)\\n # 2. 问号后不能是 ? 或 . 或 : (排除 ??、?. 和 TS的可选属性连写 ?:)\\n # 3. 问号和冒号之间应存在至少一个非空字符（保证是有效表达式），最多匹配 500 个字符防止无意义的跨段误判。\\n js_ternary_pattern = re.compile(r'(?<!\\\\?|\\\\.)\\\\?(?!\\\\?|\\\\.|:)\\\\s*(?:\\\\S(?:[\\\\s\\\\S]{0,500}?))\\\\s*:')\\n \\n # 针对 Python 的三元运算符： expr1 if condition else expr2\\n # 特征点：要求单行内包含非空字符 + 空白 + if + 空白 + 条件 + 空白 + else + 空白 + 非空字符\\n py_ternary_pattern = re.compile(r'\\\\S+[^\\\\n]*?\\\\s+if\\\\s+[^\\\\n]+?\\\\s+else\\\\s+[^\\\\n]*?\\\\S+')\\n \\n def remove_strings_and_comments(code: str) -> str:\\n # 移除单行注释\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'#.*', '', code)\\n # 移除多行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code, flags=re.DOTALL)\\n # 简单移除成对的字符串，防止字符串里包含虚假的 \\\" ? : \\\" 影响判断\\n code = re.sub(r'\\\".*?\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'.*?'\\\", \\\"''\\\", code)\\n code = re.sub(r\\\"`[\\\\s\\\\S]*?`\\\", \\\"``\\\", code)\\n return code\\n\\n # 遍历项目工作区\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 去除注释和字符串常量后再进行判断\\n clean_content = remove_strings_and_comments(content)\\n \\n if js_ternary_pattern.search(clean_content) or py_ternary_pattern.search(clean_content):\\n return True\\n except Exception:\\n continue\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，请严格约束其中所有的 import/require 语句，使其引用的包名保持字母降序排布\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于生成的所有代码文件，务必在其首行写入该编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，仪表盘前端项目可能的代码文件后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\\n # 忽略常见的依赖和产物目录\\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__'}\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n \\n # 检查首行是否包含指定的编码声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取出现异常，则视为未满足要求\\n return False\\n \\n # 要求生成项目中必须包含至少一个代码文件，且所有代码文件均满足首行声明要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"针对所有由你生成的代码文件，必须在结尾另起一行作清晰的注释标注：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 针对前端项目的典型代码文件类型推断\\n valid_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".jsx\\\", \\\".ts\\\", \\\".tsx\\\", \\\".vue\\\"}\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉无关的目录，避免扫描第三方依赖或版本控制文件夹\\n dirs[:] = [d for d in dirs if d not in (\\\"node_modules\\\", \\\".git\\\", \\\"dist\\\", \\\"build\\\")]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于任何原因无法读取代码文件，认为验证不通过\\n return False\\n \\n # 过滤掉完全为空白内容的行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n if not non_empty_lines:\\n return False\\n \\n # 检查倒数 3 个非空行中是否包含目标注释内容\\n # 此逻辑可以很好地兼容 \\\"*/\\\"、\\\"-->\\\" 等结束符号单独作为最后一行的场景\\n has_target = False\\n for line in non_empty_lines[-3:]:\\n if target_str in line:\\n has_target = True\\n break\\n \\n if not has_target:\\n return False\\n\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你所输出的任意一个代码文件，都必须刚好存在2个\\\"// FIXME\\\"标记\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令（前端仪表盘、看板、抽屉、拖拽等），推断可能的代码文件类型\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的常规依赖或版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 检查 \\\"// FIXME\\\" 标记是否刚好存在 2 个\\n if content.count('// FIXME') != 2:\\n return False\\n except Exception:\\n # 如果文件存在但无法读取，视为验证失败\\n return False\\n\\n # 必须至少有一个代码文件生成且均符合要求才算满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"无论哪一个生成的代码文件，都不可含有console.log调试语句\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令的前端项目背景，推断可能的代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.svelte'}\\n \\n if not os.path.exists(workspace_path):\\n # 如果工作区目录不存在，可以认为尚未违规，或者视具体测试框架逻辑而定，此处默认为 True\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的第三方依赖库及编译产物目录，避免误判\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', '.next', '.nuxt')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若发现 console.log，则说明不满足要求\\n if 'console.log' in content:\\n return False\\n except Exception:\\n # 遇到读取异常时跳过（防止因编码问题导致的崩溃）\\n continue\\n \\n # 如果所有代码文件均不包含 console.log，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每次内容更新完毕后，需要更新Review.md文件以审阅现有代码，必须遵照此格式输出：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请保证每一个生成的代码文件，连同空行在内的总行数均符合质数的要求\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 针对前端骨架项目的常见代码后缀\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.css'}\\n file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除前端依赖和版本控制目录，避免干扰\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用换行符分割能够精确计入末尾空行或连续空行\\n lines = content.split('\\\\n')\\n line_count = len(lines)\\n \\n # 检查行数是否为质数\\n if not is_prime(line_count):\\n return False\\n except Exception:\\n # 遇到非UTF-8等异常无法读取的情况，安全跳过\\n continue\\n \\n # 如果找到了相关的代码文件并全都符合质数要求，则返回 True\\n return file_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看页面左侧的导航菜单，确认是否清晰展示了‘概览’、‘案件进度’、‘文件中心’和‘管理员面板’四个菜单项。","查看页面顶部状态栏，确认是否展示了系统的Logo图片（或文字占位符）以及当前登录账号名称（如‘李律师’）。","刷新页面回到默认视图，检查主内容区域是否展示‘概览’页面，并且包含四个数据卡片：‘进行中案件数’、‘本周开庭数’、‘待审文件’和‘新增客户’。","分别点击左侧导航栏的‘案件进度’、‘文件中心’和‘管理员面板’菜单项，观察右侧主内容区域是否能正确切换到对应的页面内容。","进入‘案件进度’页面，检查页面是否以看板形式展示，并且包含四个垂直的列，列名分别为‘前期准备’、‘调查取证’、‘庭审阶段’、‘执行/结案’。","在‘案件进度’页面查看案件卡片，检查卡片上是否显示了案件编号、案件名称、负责律师和截止日期，且存在截止日期的文本标红显示的卡片（模拟快到期效果）。","在‘案件进度’页面，选择位于非最右阶段列（如‘前期准备’‘调查取证’或‘庭审阶段’）中的任意一张案件卡片，点击其上的‘推进到下一阶段’按钮，观察该卡片是否从当前列消失，并成功移动到右侧紧邻的下一阶段列中。","在‘案件进度’页面观察最右侧‘执行/结案’列中的案件卡片，检查其背景色是否置灰，且卡片上是否具有‘已完结’字样的标记。","在‘案件进度’页面，点击任意一个案件卡片的主体区域（避开按钮），观察屏幕右侧是否成功滑出一个展示详细信息的抽屉（Drawer）。","查看滑出的详情抽屉，确认抽屉顶部是否包含可点击切换的‘客户详情’和‘律师详情’两个标签页（Tabs）。","在抽屉内点击切换到‘律师详情’标签页，检查页面内容是否展示了律师头像、执业证号以及一个可视化的案件胜率进度条。","在抽屉内切换到‘客户详情’标签页，点击手机号码旁边的‘修改’小图标，检查原本展示手机号的文本是否被替换成了一个输入框以及一个‘保存’按钮。","在客户详情的手机号输入框中输入新的号码数字，点击‘保存’按钮，验证输入框和保存按钮是否消失，且页面重新以纯文本形式展示刚刚输入的新手机号码。","进入‘文件中心’页面，检查左侧是否展示了树状结构的案件文件夹导航，且包含诸如‘诉讼材料’、‘证据清单’或‘身份证明’等节点。","在‘文件中心’页面右侧的文件列表上方，点击‘上传材料’按钮，观察屏幕中间是否弹出一个居中显示的模态框（Modal）。","查看弹出的上传模态框，确认内部是否包含一个带虚线边框的拖拽上传区域，以及用于选择‘公开材料’或‘保密材料’的单选框（Radio）。","在‘文件中心’的文件列表中，点击任意一行自带的‘标记为核心证据’按钮，验证该行数据前方是否随之出现一个黄色的高亮星标。","进入‘管理员面板’页面，检查页面上半部分是否展示了一个水平排列的柱状图（或纯CSS模拟的水平条），用于统计展示数名律师各自的案件负荷数量。","查看‘管理员面板’页面下半部分，检查是否展示了权限管理表格，且表格的列头至少包含律师姓名、职务以及‘案件查看权限’。","点击权限管理表格中任意一行的‘案件查看权限’下拉选择框，验证弹出的下拉菜单中是否包含‘查看全所’、‘查看本组’和‘仅看自己’三个选项。","在表格的‘案件查看权限’下拉菜单中选择一个不同于当前显示的选项，观察屏幕顶部是否立即滑出一条背景为绿色的全局提示信息（Toast），文字内容为‘权限设置已更新’。","触发‘权限设置已更新’的绿色提示后，不进行任何鼠标键盘操作，静默观察约3秒钟，验证该提示信息是否会自动从屏幕上消失。"],"string":"[\n \"查看页面左侧的导航菜单，确认是否清晰展示了‘概览’、‘案件进度’、‘文件中心’和‘管理员面板’四个菜单项。\",\n \"查看页面顶部状态栏，确认是否展示了系统的Logo图片（或文字占位符）以及当前登录账号名称（如‘李律师’）。\",\n \"刷新页面回到默认视图，检查主内容区域是否展示‘概览’页面，并且包含四个数据卡片：‘进行中案件数’、‘本周开庭数’、‘待审文件’和‘新增客户’。\",\n \"分别点击左侧导航栏的‘案件进度’、‘文件中心’和‘管理员面板’菜单项，观察右侧主内容区域是否能正确切换到对应的页面内容。\",\n \"进入‘案件进度’页面，检查页面是否以看板形式展示，并且包含四个垂直的列，列名分别为‘前期准备’、‘调查取证’、‘庭审阶段’、‘执行/结案’。\",\n \"在‘案件进度’页面查看案件卡片，检查卡片上是否显示了案件编号、案件名称、负责律师和截止日期，且存在截止日期的文本标红显示的卡片（模拟快到期效果）。\",\n \"在‘案件进度’页面，选择位于非最右阶段列（如‘前期准备’‘调查取证’或‘庭审阶段’）中的任意一张案件卡片，点击其上的‘推进到下一阶段’按钮，观察该卡片是否从当前列消失，并成功移动到右侧紧邻的下一阶段列中。\",\n \"在‘案件进度’页面观察最右侧‘执行/结案’列中的案件卡片，检查其背景色是否置灰，且卡片上是否具有‘已完结’字样的标记。\",\n \"在‘案件进度’页面，点击任意一个案件卡片的主体区域（避开按钮），观察屏幕右侧是否成功滑出一个展示详细信息的抽屉（Drawer）。\",\n \"查看滑出的详情抽屉，确认抽屉顶部是否包含可点击切换的‘客户详情’和‘律师详情’两个标签页（Tabs）。\",\n \"在抽屉内点击切换到‘律师详情’标签页，检查页面内容是否展示了律师头像、执业证号以及一个可视化的案件胜率进度条。\",\n \"在抽屉内切换到‘客户详情’标签页，点击手机号码旁边的‘修改’小图标，检查原本展示手机号的文本是否被替换成了一个输入框以及一个‘保存’按钮。\",\n \"在客户详情的手机号输入框中输入新的号码数字，点击‘保存’按钮，验证输入框和保存按钮是否消失，且页面重新以纯文本形式展示刚刚输入的新手机号码。\",\n \"进入‘文件中心’页面，检查左侧是否展示了树状结构的案件文件夹导航，且包含诸如‘诉讼材料’、‘证据清单’或‘身份证明’等节点。\",\n \"在‘文件中心’页面右侧的文件列表上方，点击‘上传材料’按钮，观察屏幕中间是否弹出一个居中显示的模态框（Modal）。\",\n \"查看弹出的上传模态框，确认内部是否包含一个带虚线边框的拖拽上传区域，以及用于选择‘公开材料’或‘保密材料’的单选框（Radio）。\",\n \"在‘文件中心’的文件列表中，点击任意一行自带的‘标记为核心证据’按钮，验证该行数据前方是否随之出现一个黄色的高亮星标。\",\n \"进入‘管理员面板’页面，检查页面上半部分是否展示了一个水平排列的柱状图（或纯CSS模拟的水平条），用于统计展示数名律师各自的案件负荷数量。\",\n \"查看‘管理员面板’页面下半部分，检查是否展示了权限管理表格，且表格的列头至少包含律师姓名、职务以及‘案件查看权限’。\",\n \"点击权限管理表格中任意一行的‘案件查看权限’下拉选择框，验证弹出的下拉菜单中是否包含‘查看全所’、‘查看本组’和‘仅看自己’三个选项。\",\n \"在表格的‘案件查看权限’下拉菜单中选择一个不同于当前显示的选项，观察屏幕顶部是否立即滑出一条背景为绿色的全局提示信息（Toast），文字内容为‘权限设置已更新’。\",\n \"触发‘权限设置已更新’的绿色提示后，不进行任何鼠标键盘操作，静默观察约3秒钟，验证该提示信息是否会自动从屏幕上消失。\"\n]","isConversational":false}}},{"rowIdx":75,"cells":{"task_id":{"kind":"number","value":76,"string":"76"},"repository_policy":{"kind":"string","value":"以下是你后续所有轮次必须始终遵循的约束条件：\n1. 任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改。\n2. 无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置。\n3. 使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误。\n4. 请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%。\n5. 不要在你生成的任意一个代码文件中写入注释。\n6. 必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数。\n7. 请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n\n # 定义各类代码文件的扩展名\n js_ts_exts = {'.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs', '.vue'}\n py_exts = {'.py', '.sh'}\n html_exts = {'.html', '.xml'}\n css_exts = {'.css', '.scss', '.less'}\n\n # 定义正则表达式，用于剔除字符串字面量以避免符号误判（如 \"http://\"）\n # 匹配双引号、单引号、反引号内的内容\n js_str_re = re.compile(r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`')\n # 匹配Python风格的多行和单行字符串\n py_str_re = re.compile(r'\\\"\\\"\\\"(?:.|\\n)*?\\\"\\\"\\\"|\\'\\'\\'(?:.|\\n)*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'')\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的构建与依赖目录\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'dist', 'build', 'out', '__pycache__'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略无法以 UTF-8 读取的文件（例如二进制文件等）\n continue\n \n # 去除文件开头的 shebang（如 #!/usr/bin/env node），避免被误判为注释\n if content.startswith('#!'):\n parts = content.split('\\n', 1)\n content = parts[1] if len(parts) > 1 else ''\n\n # 按文件后缀类别检查是否含有注释\n if ext in js_ts_exts:\n clean_content = js_str_re.sub('', content)\n # 检查 JS/TS 的常见单行/多行注释\n if '/*' in clean_content or '//' in clean_content:\n return False\n # 兼容 vue 文件中可能存在的 HTML 注释\n if ext == '.vue' and '<!--' in clean_content:\n return False\n \n elif ext in py_exts:\n clean_content = py_str_re.sub('', content)\n if '#' in clean_content:\n return False\n \n elif ext in html_exts:\n if '<!--' in content:\n return False\n \n elif ext in css_exts:\n if '/*' in content:\n return False\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue',\n '.html', '.css', '.scss', '.less',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \n '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\n }\n \n # 定义需要忽略的目录，避免把第三方依赖或编译输出包含在内\n ignore_dirs = {\n 'node_modules', '.git', 'dist', 'build', \n 'coverage', '__pycache__', 'out'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过换行符拆分来获取真实包含空行的总行数\n # 例如：完全为空的文件，split后为 ['']，长度为1（奇数）；\n # 包含一段代码并以一个换行符结尾的文件，split后最后元素为空字符串，长度亦真实代表行数。\n lines = content.split('\\n')\n \n # 校验总行数是否为偶数，若为偶数则判定为不满足条件\n if len(lines) % 2 == 0:\n return False\n \n has_code_file = True\n except UnicodeDecodeError:\n # 如果文件无法用 UTF-8 解码（可能混入了二进制文件等），忽略该文件\n continue\n except Exception:\n # 忽略其他读取异常（例如权限问题等）\n continue\n \n # 确保项目中至少存在一个有效代码文件并全部满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]}],"string":"[\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义各类代码文件的扩展名\\n js_ts_exts = {'.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs', '.vue'}\\n py_exts = {'.py', '.sh'}\\n html_exts = {'.html', '.xml'}\\n css_exts = {'.css', '.scss', '.less'}\\n\\n # 定义正则表达式，用于剔除字符串字面量以避免符号误判（如 \\\"http://\\\"）\\n # 匹配双引号、单引号、反引号内的内容\\n js_str_re = re.compile(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`')\\n # 匹配Python风格的多行和单行字符串\\n py_str_re = re.compile(r'\\\\\\\"\\\\\\\"\\\\\\\"(?:.|\\\\n)*?\\\\\\\"\\\\\\\"\\\\\\\"|\\\\'\\\\'\\\\'(?:.|\\\\n)*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的构建与依赖目录\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', 'dist', 'build', 'out', '__pycache__'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略无法以 UTF-8 读取的文件（例如二进制文件等）\\n continue\\n \\n # 去除文件开头的 shebang（如 #!/usr/bin/env node），避免被误判为注释\\n if content.startswith('#!'):\\n parts = content.split('\\\\n', 1)\\n content = parts[1] if len(parts) > 1 else ''\\n\\n # 按文件后缀类别检查是否含有注释\\n if ext in js_ts_exts:\\n clean_content = js_str_re.sub('', content)\\n # 检查 JS/TS 的常见单行/多行注释\\n if '/*' in clean_content or '//' in clean_content:\\n return False\\n # 兼容 vue 文件中可能存在的 HTML 注释\\n if ext == '.vue' and '<!--' in clean_content:\\n return False\\n \\n elif ext in py_exts:\\n clean_content = py_str_re.sub('', content)\\n if '#' in clean_content:\\n return False\\n \\n elif ext in html_exts:\\n if '<!--' in content:\\n return False\\n \\n elif ext in css_exts:\\n if '/*' in content:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue',\\n '.html', '.css', '.scss', '.less',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt'\\n }\\n \\n # 定义需要忽略的目录，避免把第三方依赖或编译输出包含在内\\n ignore_dirs = {\\n 'node_modules', '.git', 'dist', 'build', \\n 'coverage', '__pycache__', 'out'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过换行符拆分来获取真实包含空行的总行数\\n # 例如：完全为空的文件，split后为 ['']，长度为1（奇数）；\\n # 包含一段代码并以一个换行符结尾的文件，split后最后元素为空字符串，长度亦真实代表行数。\\n lines = content.split('\\\\n')\\n \\n # 校验总行数是否为偶数，若为偶数则判定为不满足条件\\n if len(lines) % 2 == 0:\\n return False\\n \\n has_code_file = True\\n except UnicodeDecodeError:\\n # 如果文件无法用 UTF-8 解码（可能混入了二进制文件等），忽略该文件\\n continue\\n except Exception:\\n # 忽略其他读取异常（例如权限问题等）\\n continue\\n \\n # 确保项目中至少存在一个有效代码文件并全部满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用纯HTML、CSS和JavaScript写一个本地表格助手的基础框架。要求：\n1. 包含两个主要的视图容器：主页面（表格列表页）和表格编辑页面。\n2. 主页面需要有一个顶部的标题导航栏，中间是表格列表区域（目前先写死两三个占位卡片即可），以及一个明显的“新建表格”按钮。\n3. 表格编辑页面也需要有一个顶部栏，包含“返回”按钮和一个假定的表格标题；下方是一个简单的静态HTML表格（比如3行3列），不需要能编辑。\n4. 实现基础的页面切换逻辑：点击“新建表格”或者列表中的表格卡片，能隐藏主页面显示表格编辑页面；点击“返回”按钮能回到主页面。\n请注重整体的UI排版，让它看起来像一个现代的Web应用，暂时不需要实现本地存储和表格内部的具体操作。\n另外，以下两项要求请在后续所有操作中始终遵循：\n1. 请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于95%\n2. 确保在输出的任何文字中，自我称呼一律使用\"小助手\"而不是\"我\"","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义需要检查的常见前端和测试相关代码文件后缀\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n \n # 编译单行注释正则表达式\n # 匹配行首可选空白后，或者是空白符、分号、花括号之后跟随的 //\n # 此规则能有效避开诸如 http:// 或者 href=\"//cdn...\" 的误伤情况\n single_line_comment_pattern = re.compile(r'(?:^\\s*|[\\s;{}])//', re.MULTILINE)\n \n # 遍历当前工作区路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果有不可读的文件，则跳过\n continue\n \n # 检查是否包含 HTML 注释\n if '<!--' in content:\n return False\n \n # 检查是否包含多行或块级注释 (CSS/JS)\n if '/*' in content:\n return False\n \n # 检查是否包含单行注释 (JS)\n if single_line_comment_pattern.search(content):\n return False\n \n # 所有文件验证完毕，无任何注释则满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的前端与配置相关的代码文件扩展名\n code_extensions = {'.js', '.css', '.html', '.json', '.cjs', '.mjs', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件（基于扩展名或常见的ESLint配置文件）\n is_code_file = (ext in code_extensions) or file.startswith('.eslint')\n \n if is_code_file:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照换行符分割，能够精确包括空行来计算总行数\n lines_count = len(content.split('\\n'))\n \n # 若总行数不是奇数（即为偶数），说明不满足要求，直接返回 False\n if lines_count % 2 == 0:\n return False\n \n has_code_file = True\n except Exception:\n # 忽略不是纯文本或无法以UTF-8解码的非标准文件\n continue\n \n # 只要存在至少一个代码文件，且全部验证通过（没有触发偶数返回），则视为满足要求\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于95%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"小助手\"而不是\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查助手的回复中是否包含字符“我”\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]}]},{"round_id":1,"instruction":"很好，框架出来了。现在我们引入LocalStorage来实现数据的本地持久化以及主页面的完整逻辑：\n1. 把主页面的列表改为动态渲染，读取LocalStorage中保存的所有表格记录。每条记录展示表格标题、最后修改时间，并且卡片右上角要有一个“删除”按钮。\n2. 点击“新建表格”时，在LocalStorage中生成一条新记录（包含唯一ID、默认标题“未命名表格”、以及一个默认的5行5列的空数据结构），然后跳转到表格编辑页面打开它。\n3. 点击主页列表中的“删除”按钮，要有二次确认弹窗，确认后从LocalStorage删除对应记录并刷新列表。\n4. 点击列表卡片本身，可以带着当前表格的ID跳转到表格编辑页面。\n注意，目前表格编辑页面只要能接收到ID并把表格的标题显示出来就行，表格本身依然用之前的静态占位。\n\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\n1. 原单元测试行覆盖率≥95%的要求调整为：请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%。\n2. 确保你生成的任何代码文件中都不包含alert语句，删除功能的二次确认弹窗请使用自定义DOM弹窗实现。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"def check(response, workspace_path):\n import os\n import re\n\n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n\n # 构造正则表达式，按从左至右的顺序优先级进行匹配：\n # 1. string: 匹配被双引号、单引号或反引号包裹的字符串（正确处理转义字符，不匹配断行错误）\n # 2. html_comment: 匹配 HTML 注释 <!-- ... -->\n # 3. multi_line: 匹配 CSS/JS 多行注释 /* ... */\n # 4. single_line: 匹配 JS 单行注释 // ... （排除前面是冒号的情况，避免将 http:// 识别为注释）\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<html_comment><!--.*?-->)'\n r'|(?P<multi_line>/\\*.*?\\*/)'\n r'|(?P<single_line>(?<!:)//[^\\r\\n]*)',\n re.DOTALL\n )\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除前端项目中常见的无需检查注释的依赖或系统目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n if '.vscode' in dirs:\n dirs.remove('.vscode')\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果遇到不可读取的文件或解码异常则跳过\n continue\n\n # 遍历所有匹配项，查找是否存在注释\n for m in pattern.finditer(content):\n group_name = m.lastgroup\n # 如果成功匹配到了注释分组（而没有被前面的字符串分组消耗掉），说明文件中存在真实的注释\n if group_name in ['html_comment', 'multi_line', 'single_line']:\n return False\n\n # 若遍历了所有相关代码文件均未发现任何注释，判定满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码文件后缀类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的无关目录\n path_parts = root.split(os.sep)\n if 'node_modules' in path_parts or '.git' in path_parts:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n has_code_file = True\n # 空行已被 readlines 作为单独的列表元素计算在内，可直接检查长度是否为奇数\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果遇到文件编码等读取异常，则跳过\n pass\n \n # 如果没有找到任何代码文件，则也视为不满足\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"小助手\"而不是\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 1. 检查直接回复内容中是否包含“我”\n if \"我\" in response:\n return False\n\n # 2. 检查生成的代码项目文件中是否包含“我”\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if \"我\" in content:\n return False\n except Exception:\n # 忽略二进制文件或无法读取的文件\n continue\n\n return True","tags":["Content","Persona"]},{"constraint":"确保你生成的任何代码文件中都不包含alert语句","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 编译正则表达式，匹配代码中出现的 alert( 或 window.alert( 语句\n alert_pattern = re.compile(r'\\balert\\s*\\(')\n \n # 假定常见的前端代码文件类型，因为 CSS 中不会包含 JS 的 alert 语句，主要检查 HTML 和 JS 相关文件\n code_extensions = ('.html', '.js', '.jsx', '.ts', '.tsx')\n \n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果在文件内容中匹配到了 alert 语句，说明没有满足要求\n if alert_pattern.search(content):\n return False\n except Exception:\n # 如果遇到文件编码问题无法读取，则跳过继续检查下一个文件\n continue\n \n # 所有代码文件都遍历结束且未发现 alert 语句，则满足要求\n return True","tags":["Content","Keyword"]}]},{"round_id":2,"instruction":"接下来完善表格编辑页面的基础编辑和渲染功能：\n1. 进入表格编辑页面时，根据传入的ID从LocalStorage读取对应的表格数据，并动态渲染出这个表格。\n2. 支持编辑表格标题：点击顶部栏的标题可以直接修改，失去焦点时自动把新标题保存到LocalStorage。\n3. 支持编辑单元格内容：双击任意一个单元格，该单元格变为输入框（input），输入文本后失去焦点或按下回车，输入框恢复为普通文本，并且最新的内容必须立即同步更新到LocalStorage中，保证刷新页面数据不丢失。\n\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\n1. 原“确保你生成的任何代码文件中都不包含alert语句”的要求调整为：确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句，所有交互弹窗均使用自定义DOM弹窗实现。\n2. 每次编辑已有文件之前，必须先借助对应工具审阅该文件。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 匹配 JS 及其类似语法中的注释 (利用正则中的捕获组优先匹配跳过字符串和模板字符串)\n def has_js_comment(code):\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<comment>//[^\\n]*|/\\*[\\s\\S]*?\\*/)'\n )\n for match in pattern.finditer(code):\n if match.group('comment'):\n return True\n return False\n\n # 匹配 CSS 中的注释 (利用正则中的捕获组优先匹配跳过字符串)\n def has_css_comment(code):\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/)'\n )\n for match in pattern.finditer(code):\n if match.group('comment'):\n return True\n return False\n\n # 匹配 HTML 中的整体注释 (利用正则中的捕获组优先匹配跳过字符串)\n def has_html_comment(code):\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<comment><!--[\\s\\S]*?-->)'\n )\n for match in pattern.finditer(code):\n if match.group('comment'):\n return True\n return False\n\n def check_html_file(code):\n # 1. 检查HTML自带的 <!-- --> 注释\n if has_html_comment(code):\n return True\n \n # 2. 检查内嵌 script 中的 JS 注释\n script_pattern = re.compile(r'<script\\b[^>]*>([\\s\\S]*?)</script>', re.IGNORECASE)\n for match in script_pattern.finditer(code):\n if has_js_comment(match.group(1)):\n return True\n \n # 3. 检查内嵌 style 中的 CSS 注释\n style_pattern = re.compile(r'<style\\b[^>]*>([\\s\\S]*?)</style>', re.IGNORECASE)\n for match in style_pattern.finditer(code):\n if has_css_comment(match.group(1)):\n return True\n \n return False\n\n js_exts = {'.js', '.ts', '.jsx', '.tsx', '.json'}\n css_exts = {'.css', '.scss', '.less'}\n html_exts = {'.html', '.htm', '.vue'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 过滤不需要检查的非生成代码目录（如第三方依赖）\n parts = file_path.replace('\\\\', '/').split('/')\n if 'node_modules' in parts or '.git' in parts:\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in js_exts or ext in css_exts or ext in html_exts:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 按文件类型执行检查逻辑\n if ext in js_exts:\n if has_js_comment(code):\n return False\n elif ext in css_exts:\n if has_css_comment(code):\n return False\n elif ext in html_exts:\n if check_html_file(code):\n return False\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对前端要求（HTML, CSS, JS），包含常规代码文件的扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue', '.mjs', '.cjs'}\n has_code_file = False\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除不需要检查的隐藏目录和第三方包目录\n dirs[:] = [d for d in dirs if d not in ['node_modules', '__pycache__'] and not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取并忽略解析错误以防止异常中断\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 按换行符 \\n 分割计算行数，以此方式能够精准包含所有空行（包括文件末尾的空行）\n lines = content.split('\\n')\n \n # 如果行数为偶数，则说明违反了要求\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果发生意外异常导致无法读取，保守处理，认为该文件暂不满足检测条件\n return False\n \n # 如果找到了相关的代码文件，并且全都是奇数行，则返回 True\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"小助手\"而不是\"我\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复中是否不包含“我”，\n 以验证其自我称呼是否严格使用了“小助手”而不是“我”。\n \"\"\"\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 检查工作区路径是否存在\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n \n # 定义正则表达式：匹配 alert、confirm、prompt 作为函数调用的情况\n # \\b 确保是独立单词，\\s* 匹配可能存在的空格，\\( 匹配函数调用的左括号\n pattern = re.compile(r'\\b(alert|confirm|prompt)\\s*\\(')\n \n # 定义可能包含 JavaScript 代码的文件扩展名\n code_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue'}\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查可能的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查文件内容是否包含这三类弹窗语句\n if pattern.search(content):\n return False\n except Exception:\n # 如果由于编码或其他问题无法读取文件，跳过该文件\n pass\n \n # 未检测到任何原生弹窗语句，说明满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":3,"instruction":"现在给表格编辑页面增加行列的动态增删功能。我们需要在表格的上方或侧边增加一排操作按钮：\n1. 实现“选中”机制：单击任意单元格时，让该单元格高亮显示（加个边框或背景色），表示它是当前激活的单元格。\n2. 增加四个按钮：“在上方插入行”、“在下方插入行”、“在左侧插入列”、“在右侧插入列”。点击时，参照当前选中的单元格位置，在表格数据中插入对应的空行或空列，重新渲染表格并更新LocalStorage。\n3. 增加两个按钮：“删除当前行”和“删除当前列”。点击后删除选中单元格所在的行或列，如果表格只剩最后一行/列则提示不许删除。操作后立刻保存并重新渲染。\n\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\n1. 原自我称呼的要求调整为：确保在输出的任何文字中，自我称呼一律使用\"前端开发小助手\"而不是\"我\"。\n2. 必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 如果工作区目录不存在，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n \n # 用于预处理：匹配代码中的各类字符串字面量（双引号、单引号、反引号），避免将字符串内部的内容误认为注释\n string_pattern = re.compile(r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`[\\s\\S]*?`')\n \n # 匹配各类注释的正则表达式\n html_comment_pattern = re.compile(r'<!--[\\s\\S]*?-->')\n block_comment_pattern = re.compile(r'/\\*[\\s\\S]*?\\*/')\n # 匹配 // 注释，但排除前面是冒号的情况（防止错误匹配到 http:// 等 URL 结构）\n line_comment_pattern = re.compile(r'(?<!:)//')\n \n # 需要检查的常见代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的目录，例如第三方依赖和隐藏目录\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', '.idea', '.vscode']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果遇到文件无法读取的异常情况，跳过该文件继续检查其他文件\n continue\n \n # 移除源代码中的字符串字面量内容，留下纯粹的代码结构\n content_no_strings = string_pattern.sub('', content)\n \n # 检查预处理后的代码中是否仍然存在任何类型的注释\n if html_comment_pattern.search(content_no_strings):\n return False\n if block_comment_pattern.search(content_no_strings):\n return False\n if line_comment_pattern.search(content_no_strings):\n return False\n \n # 如果遍历检查完所有符合条件的代码文件均未发现注释，则满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义前端项目常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除无需检查的无关目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按照 '\\n' 分割可以准确计算总行数并包含所有空行\n # 例如内容为 \"a\\nb\" 划分为 ['a', 'b']，共2行\n # 内容为 \"a\\nb\\n\" 划分为 ['a', 'b', '']，共3行\n lines = content.split('\\n')\n line_count = len(lines)\n \n # 如果存在任何一个代码文件总行数为偶数，则不满足要求\n if line_count % 2 == 0:\n return False\n \n has_code_file = True\n except Exception:\n # 忽略不可读或编码有问题的文件\n continue\n \n # 如果项目中没有任何代码文件，严格意义上不算满足生成代码行数为奇数的条件\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"前端开发小助手\"而不是\"我\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 检查助手的回复中是否出现了“我”\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 匹配独立的弹窗函数调用，或者是 window. 前缀的调用\n # (?:^|[^\\w\\.]) : 确保前面不是字母数字下划线或其他的点（避免匹配对象的自定义属性调用如 xxx.alert）\n # (?:window\\.)? : 允许匹配 window.alert 等原生形式\n # (alert|confirm|prompt) : 匹配三种弹窗关键字\n # \\s*\\( : 匹配紧跟的括号，确认其为函数调用\n pattern = re.compile(r'(?:^|[^\\w\\.])(?:window\\.)?(alert|confirm|prompt)\\s*\\(')\n \n # 结合项目类型，定义可能包含 JavaScript 代码的文件扩展名\n valid_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略掉可能产生干扰的无关目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若文件中存在 alert, confirm 或 prompt 的原生调用，则不满足要求\n if pattern.search(content):\n return False\n except Exception:\n # 忽略不可读取的文件\n pass\n \n # 检查完所有相关代码文件均未发现原生弹窗语句，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":4,"instruction":"现在的挑战来了，我们需要实现复杂的“合并单元格”功能：\n1. 实现多选机制：用户可以通过在表格内按下鼠标并拖拽，选中一个矩形区域内的多个单元格（给被选中的区域加上半透明背景色）。\n2. 增加一个“合并单元格”按钮。当用户选中了大于1个单元格并点击此按钮时，将选区合并为一个大的单元格（使用HTML的rowspan和colspan属性）。\n3. 合并后的单元格文本保留选区内左上角第一个有内容的单元格文本，选区内其他单元格在数据结构中需标记为“被隐藏/已合并”，不要在页面中渲染出来。\n4. 每次合并操作完成后，必须将带有合并状态的数据结构保存到LocalStorage，并确保页面刷新后依然能正确读取并渲染合并后的表格外观。\n\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\n1. 原“必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置”的要求调整为：请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来。\n2. 新增要求：在对任何已有文件执行修改操作前，请务必在相同目录层级下为其生成一份修改前的备份文件。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def check_js(text: str) -> bool:\n # 正则表达式按优先级捕获：\n # 1. string: 双引号、单引号或反引号括起来的字符串字面量（包含转义字符支持）\n # 2. block_comment: /* ... */ 格式的多行注释\n # 3. line_comment: // 格式的单行注释（排除前面是冒号的情况如 http://）\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<block_comment>/\\*.*?\\*/)'\n r'|(?P<line_comment>(?<!:)//[^\\r\\n]*)',\n re.DOTALL | re.MULTILINE\n )\n for match in pattern.finditer(text):\n # 如果匹配到的内容属于注释分组，则表明包含注释\n if match.group('block_comment') or match.group('line_comment'):\n return True\n return False\n \n def check_css(text: str) -> bool:\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<block_comment>/\\*.*?\\*/)',\n re.DOTALL | re.MULTILINE\n )\n for match in pattern.finditer(text):\n if match.group('block_comment'):\n return True\n return False\n\n def check_no_comments(file_path: str) -> bool:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果文件无法读取，暂且跳过不判定为含注释\n return True\n \n ext = file_path.split('.')[-1].lower()\n \n if ext in ['js', 'jsx', 'ts', 'tsx']:\n if check_js(content):\n return False\n elif ext == 'css':\n if check_css(content):\n return False\n elif ext in ['html', 'htm']:\n # 检查直接编写的HTML注释\n if re.search(r'<!--.*?-->', content, re.DOTALL):\n return False\n # 提取所有 <script> 标签中的内容进行 JS 注释检查\n for script_match in re.finditer(r'<script[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL):\n if check_js(script_match.group(1)):\n return False\n # 提取所有 <style> 标签中的内容进行 CSS 注释检查\n for style_match in re.finditer(r'<style[^>]*>(.*?)</style>', content, re.IGNORECASE | re.DOTALL):\n if check_css(style_match.group(1)):\n return False\n \n return True\n\n valid_exts = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见无关目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n # 只要发现任意一个有效代码文件内含有注释，即不满足要求\n if not check_no_comments(file_path):\n return False\n \n # 未在任何有效代码文件里发现注释，要求得到满足\n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户的指令，推断纯HTML/CSS/JS应用及单元测试文件可能具有的代码后缀\n code_exts = {'.html', '.css', '.js'}\n \n # 需要排除的由第三方工具自动生成的或非助手直接生成的常见目录\n exclude_dirs = {'node_modules', 'coverage', '.nyc_output', '.git', 'dist', 'build'}\n \n file_found = False\n\n # 若工作区路径不存在，直接返回未满足\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 列表，使得 os.walk 跳过被排除的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n # 判断是否为推断出的代码文件类型\n if any(file.endswith(ext) for ext in code_exts):\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 依据换行符进行分割可以确保所有空行（包括文件末尾由于换行符多产生的空行）都能计算在内\n lines = content.split('\\n')\n \n # 检查总行数是否为偶数\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果某个文件因为编码或权限等原因无法读取，直接跳过，不应引起验证程序崩溃\n continue\n\n # 如果没有找到任何代码文件，说明没有生成所需代码，返回False；否则所有检查通过，返回True\n return file_found","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"前端开发小助手\"而不是\"我\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复中是否没有使用\"我\"作为自我称呼。\n 严格模式下，直接检查字符串中是否包含\"我\"字。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句","validation_code":"def check_popup_statements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return True\n \n # 定义需要检查的代码文件类型（基于纯前端项目推断）\n code_extensions = ('.html', '.js')\n \n # 遍历当前工作空间内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 去除 HTML 注释，防止误判\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 去除 JavaScript 多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 去除 JavaScript 单行注释\n content = re.sub(r'//.*', '', content)\n \n # 匹配原生弹窗函数调用，涵盖 alert(), confirm(), prompt() 及其带有空格的情况\n if re.search(r'\\b(alert|confirm|prompt)\\s*\\(', content):\n return False\n except Exception:\n # 忽略读取异常的文件\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在对任何已有文件执行修改操作前，请务必在相同目录层级下为其生成一份修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":5,"instruction":"最后一步，补充“取消合并”功能，并完善交互细节：\n1. 在操作栏增加一个“取消合并”按钮。\n2. 如果用户单击选中的是一个之前被合并过的巨大单元格（rowspan或colspan > 1），点击“取消合并”按钮后，将其恢复成原来数量的独立单格（rowspan和colspan恢复为1），原隐藏的单元格重新显示为空白单元格。\n3. 取消合并后，不仅要重新渲染视图，还要更新LocalStorage，把对应的合并标记清除。\n4. 优化整个拖拽多选的体验和界面的CSS样式，确保在包含合并单元格的表格中进行行列插入、删除时，数据结构不会乱掉（如果太复杂，可以限制为：当前选中区域如果有合并单元格时禁用增删行列按钮）。\n\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\n1. 原单元测试函数覆盖率≥98%的要求调整为：请确保生成的单元测试文件运行后，能够让目标代码的语句覆盖率大于等于99%。\n2. 新增要求：凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"不要在你生成的任意一个代码文件中写入注释","validation_code":"def check_no_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见前端代码文件扩展名\n target_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能包含依赖或版本控制文件的常见无关目录\n if 'node_modules' in root.split(os.sep) or '.git' in root.split(os.sep):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in target_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件不可读，则跳过\n continue\n \n # 为了防止误判，将字符串字面量内容剔除（替换为成对的空引号）\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', content)\n \n # 将 url() 中的内容剔除，防止无引号的链接双斜杠被识别为单行注释\n content = re.sub(r'url\\([^)]*\\)', 'url()', content)\n \n # 检测 HTML 注释\n if re.search(r'<!--[\\s\\S]*?-->', content):\n return False\n \n # 检测 块级注释 (CSS/JS 等通用)\n if re.search(r'/\\*[\\s\\S]*?\\*/', content):\n return False\n \n # 检测 JS/HTML 中的单行注释，排除形如 \"http://\" 中冒号后的斜杠\n if ext in {'.html', '.vue', '.js', '.jsx', '.ts', '.tsx'}:\n if re.search(r'(?<!:)//', content):\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_code_lines_odd(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令内容推断，可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n code_file_found = False\n \n # 遍历项目目录下的所有文件（包括子目录中的测试文件等）\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统计总行数（将空行计算在内）\n # 以 '\\n' 为分隔符进行 split 操作，返回的列表长度即为包含空行在内的确切文件行数\n line_count = len(content.split('\\n'))\n \n # 如果有任意一个代码文件的行数为偶数，则说明不满足要求\n if line_count % 2 == 0:\n return False\n \n code_file_found = True\n except Exception:\n # 忽略读取异常的文件\n pass\n \n # 如果遍历完毕未出现偶数行，且至少找到了一份有效代码文件，则要求被满足\n return code_file_found","tags":["Quantity","Exact Value"]},{"constraint":"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的语句覆盖率大于等于99%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保在输出的任何文字中，自我称呼一律使用\"前端开发小助手\"而不是\"我\"","validation_code":"def check_self_reference_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复内容中的自我称呼是否符合要求。\n 要求：确保在输出的任何文字中，自我称呼一律使用\"前端开发小助手\"而不是\"我\"。\n \"\"\"\n \n # 1. 确保回复的任何文字中没有出现“我”字\n if \"我\" in response:\n return False\n \n # 2. 确保之前轮次要求的“小助手”被准确更新为“前端开发小助手”\n # 即回复中出现所有的“小助手”时，其前缀必须是“前端开发”\n idx = response.find(\"小助手\")\n while idx != -1:\n # 提取“小助手”前面的4个字符，如果不足4个或者不是\"前端开发\"，则不符合要求\n if idx < 4 or response[idx-4:idx] != \"前端开发\":\n return False\n # 继续寻找下一个“小助手”\n idx = response.find(\"小助手\", idx + 1)\n \n return True","tags":["Content","Persona"]},{"constraint":"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断生成的代码项目中是否完全不包含 alert、confirm、prompt 三类原生JavaScript弹窗语句。\n \"\"\"\n # 可能包含 JavaScript 弹窗语句的目标代码文件后缀名\n target_extensions = ('.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.mjs', '.cjs')\n \n # 正则表达式匹配原生的 alert, confirm, prompt 调用\n # 涵盖 alert(...) 或 window.alert(...) 等形式\n # [^a-zA-Z0-9_$.] 用于确保前面没有其他变量名拼接或其他的对象调用（例如 custom_alert 或 dialog.alert）\n pattern = re.compile(r'(?:^|[^a-zA-Z0-9_$.])(?:window\\.)?(alert|confirm|prompt)\\s*\\(')\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(target_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果在代码文件中找到了任一原生弹窗语句，则代表未满足要求，直接返回 False\n if pattern.search(content):\n return False\n except Exception:\n # 忽略无法正常读取的文件（例如由于权限、编码异常等导致）\n continue\n \n # 所有的代码文件检查均未发现原生弹窗语句，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"每次编辑已有文件之前，必须先借助对应工具审阅该文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在对任何已有文件执行修改操作前，请务必在相同目录层级下为其生成一份修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md","validation_code":"","tags":["Environment","Logging"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用纯HTML、CSS和JavaScript写一个本地表格助手的基础框架。要求：\\n1. 包含两个主要的视图容器：主页面（表格列表页）和表格编辑页面。\\n2. 主页面需要有一个顶部的标题导航栏，中间是表格列表区域（目前先写死两三个占位卡片即可），以及一个明显的“新建表格”按钮。\\n3. 表格编辑页面也需要有一个顶部栏，包含“返回”按钮和一个假定的表格标题；下方是一个简单的静态HTML表格（比如3行3列），不需要能编辑。\\n4. 实现基础的页面切换逻辑：点击“新建表格”或者列表中的表格卡片，能隐藏主页面显示表格编辑页面；点击“返回”按钮能回到主页面。\\n请注重整体的UI排版，让它看起来像一个现代的Web应用，暂时不需要实现本地存储和表格内部的具体操作。\\n另外，以下两项要求请在后续所有操作中始终遵循：\\n1. 请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于95%\\n2. 确保在输出的任何文字中，自我称呼一律使用\\\"小助手\\\"而不是\\\"我\\\"\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义需要检查的常见前端和测试相关代码文件后缀\\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n \\n # 编译单行注释正则表达式\\n # 匹配行首可选空白后，或者是空白符、分号、花括号之后跟随的 //\\n # 此规则能有效避开诸如 http:// 或者 href=\\\"//cdn...\\\" 的误伤情况\\n single_line_comment_pattern = re.compile(r'(?:^\\\\s*|[\\\\s;{}])//', re.MULTILINE)\\n \\n # 遍历当前工作区路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果有不可读的文件，则跳过\\n continue\\n \\n # 检查是否包含 HTML 注释\\n if '<!--' in content:\\n return False\\n \\n # 检查是否包含多行或块级注释 (CSS/JS)\\n if '/*' in content:\\n return False\\n \\n # 检查是否包含单行注释 (JS)\\n if single_line_comment_pattern.search(content):\\n return False\\n \\n # 所有文件验证完毕，无任何注释则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的前端与配置相关的代码文件扩展名\\n code_extensions = {'.js', '.css', '.html', '.json', '.cjs', '.mjs', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件（基于扩展名或常见的ESLint配置文件）\\n is_code_file = (ext in code_extensions) or file.startswith('.eslint')\\n \\n if is_code_file:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照换行符分割，能够精确包括空行来计算总行数\\n lines_count = len(content.split('\\\\n'))\\n \\n # 若总行数不是奇数（即为偶数），说明不满足要求，直接返回 False\\n if lines_count % 2 == 0:\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 忽略不是纯文本或无法以UTF-8解码的非标准文件\\n continue\\n \\n # 只要存在至少一个代码文件，且全部验证通过（没有触发偶数返回），则视为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于95%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"小助手\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查助手的回复中是否包含字符“我”\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"很好，框架出来了。现在我们引入LocalStorage来实现数据的本地持久化以及主页面的完整逻辑：\\n1. 把主页面的列表改为动态渲染，读取LocalStorage中保存的所有表格记录。每条记录展示表格标题、最后修改时间，并且卡片右上角要有一个“删除”按钮。\\n2. 点击“新建表格”时，在LocalStorage中生成一条新记录（包含唯一ID、默认标题“未命名表格”、以及一个默认的5行5列的空数据结构），然后跳转到表格编辑页面打开它。\\n3. 点击主页列表中的“删除”按钮，要有二次确认弹窗，确认后从LocalStorage删除对应记录并刷新列表。\\n4. 点击列表卡片本身，可以带着当前表格的ID跳转到表格编辑页面。\\n注意，目前表格编辑页面只要能接收到ID并把表格的标题显示出来就行，表格本身依然用之前的静态占位。\\n\\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\\n1. 原单元测试行覆盖率≥95%的要求调整为：请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%。\\n2. 确保你生成的任何代码文件中都不包含alert语句，删除功能的二次确认弹窗请使用自定义DOM弹窗实现。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"def check(response, workspace_path):\\n import os\\n import re\\n\\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n\\n # 构造正则表达式，按从左至右的顺序优先级进行匹配：\\n # 1. string: 匹配被双引号、单引号或反引号包裹的字符串（正确处理转义字符，不匹配断行错误）\\n # 2. html_comment: 匹配 HTML 注释 <!-- ... -->\\n # 3. multi_line: 匹配 CSS/JS 多行注释 /* ... */\\n # 4. single_line: 匹配 JS 单行注释 // ... （排除前面是冒号的情况，避免将 http:// 识别为注释）\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<html_comment><!--.*?-->)'\\n r'|(?P<multi_line>/\\\\*.*?\\\\*/)'\\n r'|(?P<single_line>(?<!:)//[^\\\\r\\\\n]*)',\\n re.DOTALL\\n )\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除前端项目中常见的无需检查注释的依赖或系统目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if '.vscode' in dirs:\\n dirs.remove('.vscode')\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果遇到不可读取的文件或解码异常则跳过\\n continue\\n\\n # 遍历所有匹配项，查找是否存在注释\\n for m in pattern.finditer(content):\\n group_name = m.lastgroup\\n # 如果成功匹配到了注释分组（而没有被前面的字符串分组消耗掉），说明文件中存在真实的注释\\n if group_name in ['html_comment', 'multi_line', 'single_line']:\\n return False\\n\\n # 若遍历了所有相关代码文件均未发现任何注释，判定满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码文件后缀类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的无关目录\\n path_parts = root.split(os.sep)\\n if 'node_modules' in path_parts or '.git' in path_parts:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n has_code_file = True\\n # 空行已被 readlines 作为单独的列表元素计算在内，可直接检查长度是否为奇数\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果遇到文件编码等读取异常，则跳过\\n pass\\n \\n # 如果没有找到任何代码文件，则也视为不满足\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"小助手\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 1. 检查直接回复内容中是否包含“我”\\n if \\\"我\\\" in response:\\n return False\\n\\n # 2. 检查生成的代码项目文件中是否包含“我”\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if \\\"我\\\" in content:\\n return False\\n except Exception:\\n # 忽略二进制文件或无法读取的文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含alert语句\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 编译正则表达式，匹配代码中出现的 alert( 或 window.alert( 语句\\n alert_pattern = re.compile(r'\\\\balert\\\\s*\\\\(')\\n \\n # 假定常见的前端代码文件类型，因为 CSS 中不会包含 JS 的 alert 语句，主要检查 HTML 和 JS 相关文件\\n code_extensions = ('.html', '.js', '.jsx', '.ts', '.tsx')\\n \\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果在文件内容中匹配到了 alert 语句，说明没有满足要求\\n if alert_pattern.search(content):\\n return False\\n except Exception:\\n # 如果遇到文件编码问题无法读取，则跳过继续检查下一个文件\\n continue\\n \\n # 所有代码文件都遍历结束且未发现 alert 语句，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来完善表格编辑页面的基础编辑和渲染功能：\\n1. 进入表格编辑页面时，根据传入的ID从LocalStorage读取对应的表格数据，并动态渲染出这个表格。\\n2. 支持编辑表格标题：点击顶部栏的标题可以直接修改，失去焦点时自动把新标题保存到LocalStorage。\\n3. 支持编辑单元格内容：双击任意一个单元格，该单元格变为输入框（input），输入文本后失去焦点或按下回车，输入框恢复为普通文本，并且最新的内容必须立即同步更新到LocalStorage中，保证刷新页面数据不丢失。\\n\\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\\n1. 原“确保你生成的任何代码文件中都不包含alert语句”的要求调整为：确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句，所有交互弹窗均使用自定义DOM弹窗实现。\\n2. 每次编辑已有文件之前，必须先借助对应工具审阅该文件。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 匹配 JS 及其类似语法中的注释 (利用正则中的捕获组优先匹配跳过字符串和模板字符串)\\n def has_js_comment(code):\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<comment>//[^\\\\n]*|/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n for match in pattern.finditer(code):\\n if match.group('comment'):\\n return True\\n return False\\n\\n # 匹配 CSS 中的注释 (利用正则中的捕获组优先匹配跳过字符串)\\n def has_css_comment(code):\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n for match in pattern.finditer(code):\\n if match.group('comment'):\\n return True\\n return False\\n\\n # 匹配 HTML 中的整体注释 (利用正则中的捕获组优先匹配跳过字符串)\\n def has_html_comment(code):\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<comment><!--[\\\\s\\\\S]*?-->)'\\n )\\n for match in pattern.finditer(code):\\n if match.group('comment'):\\n return True\\n return False\\n\\n def check_html_file(code):\\n # 1. 检查HTML自带的 <!-- --> 注释\\n if has_html_comment(code):\\n return True\\n \\n # 2. 检查内嵌 script 中的 JS 注释\\n script_pattern = re.compile(r'<script\\\\b[^>]*>([\\\\s\\\\S]*?)</script>', re.IGNORECASE)\\n for match in script_pattern.finditer(code):\\n if has_js_comment(match.group(1)):\\n return True\\n \\n # 3. 检查内嵌 style 中的 CSS 注释\\n style_pattern = re.compile(r'<style\\\\b[^>]*>([\\\\s\\\\S]*?)</style>', re.IGNORECASE)\\n for match in style_pattern.finditer(code):\\n if has_css_comment(match.group(1)):\\n return True\\n \\n return False\\n\\n js_exts = {'.js', '.ts', '.jsx', '.tsx', '.json'}\\n css_exts = {'.css', '.scss', '.less'}\\n html_exts = {'.html', '.htm', '.vue'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 过滤不需要检查的非生成代码目录（如第三方依赖）\\n parts = file_path.replace('\\\\\\\\', '/').split('/')\\n if 'node_modules' in parts or '.git' in parts:\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in js_exts or ext in css_exts or ext in html_exts:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 按文件类型执行检查逻辑\\n if ext in js_exts:\\n if has_js_comment(code):\\n return False\\n elif ext in css_exts:\\n if has_css_comment(code):\\n return False\\n elif ext in html_exts:\\n if check_html_file(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对前端要求（HTML, CSS, JS），包含常规代码文件的扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue', '.mjs', '.cjs'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除不需要检查的隐藏目录和第三方包目录\\n dirs[:] = [d for d in dirs if d not in ['node_modules', '__pycache__'] and not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取并忽略解析错误以防止异常中断\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 按换行符 \\\\n 分割计算行数，以此方式能够精准包含所有空行（包括文件末尾的空行）\\n lines = content.split('\\\\n')\\n \\n # 如果行数为偶数，则说明违反了要求\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果发生意外异常导致无法读取，保守处理，认为该文件暂不满足检测条件\\n return False\\n \\n # 如果找到了相关的代码文件，并且全都是奇数行，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"小助手\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的回复中是否不包含“我”，\\n 以验证其自我称呼是否严格使用了“小助手”而不是“我”。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 检查工作区路径是否存在\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n \\n # 定义正则表达式：匹配 alert、confirm、prompt 作为函数调用的情况\\n # \\\\b 确保是独立单词，\\\\s* 匹配可能存在的空格，\\\\( 匹配函数调用的左括号\\n pattern = re.compile(r'\\\\b(alert|confirm|prompt)\\\\s*\\\\(')\\n \\n # 定义可能包含 JavaScript 代码的文件扩展名\\n code_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查可能的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查文件内容是否包含这三类弹窗语句\\n if pattern.search(content):\\n return False\\n except Exception:\\n # 如果由于编码或其他问题无法读取文件，跳过该文件\\n pass\\n \\n # 未检测到任何原生弹窗语句，说明满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在给表格编辑页面增加行列的动态增删功能。我们需要在表格的上方或侧边增加一排操作按钮：\\n1. 实现“选中”机制：单击任意单元格时，让该单元格高亮显示（加个边框或背景色），表示它是当前激活的单元格。\\n2. 增加四个按钮：“在上方插入行”、“在下方插入行”、“在左侧插入列”、“在右侧插入列”。点击时，参照当前选中的单元格位置，在表格数据中插入对应的空行或空列，重新渲染表格并更新LocalStorage。\\n3. 增加两个按钮：“删除当前行”和“删除当前列”。点击后删除选中单元格所在的行或列，如果表格只剩最后一行/列则提示不许删除。操作后立刻保存并重新渲染。\\n\\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\\n1. 原自我称呼的要求调整为：确保在输出的任何文字中，自我称呼一律使用\\\"前端开发小助手\\\"而不是\\\"我\\\"。\\n2. 必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 如果工作区目录不存在，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 用于预处理：匹配代码中的各类字符串字面量（双引号、单引号、反引号），避免将字符串内部的内容误认为注释\\n string_pattern = re.compile(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`[\\\\s\\\\S]*?`')\\n \\n # 匹配各类注释的正则表达式\\n html_comment_pattern = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n block_comment_pattern = re.compile(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n # 匹配 // 注释，但排除前面是冒号的情况（防止错误匹配到 http:// 等 URL 结构）\\n line_comment_pattern = re.compile(r'(?<!:)//')\\n \\n # 需要检查的常见代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的目录，例如第三方依赖和隐藏目录\\n dirs[:] = [d for d in dirs if d not in ['node_modules', '.git', '.idea', '.vscode']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果遇到文件无法读取的异常情况，跳过该文件继续检查其他文件\\n continue\\n \\n # 移除源代码中的字符串字面量内容，留下纯粹的代码结构\\n content_no_strings = string_pattern.sub('', content)\\n \\n # 检查预处理后的代码中是否仍然存在任何类型的注释\\n if html_comment_pattern.search(content_no_strings):\\n return False\\n if block_comment_pattern.search(content_no_strings):\\n return False\\n if line_comment_pattern.search(content_no_strings):\\n return False\\n \\n # 如果遍历检查完所有符合条件的代码文件均未发现注释，则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义前端项目常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无需检查的无关目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按照 '\\\\n' 分割可以准确计算总行数并包含所有空行\\n # 例如内容为 \\\"a\\\\nb\\\" 划分为 ['a', 'b']，共2行\\n # 内容为 \\\"a\\\\nb\\\\n\\\" 划分为 ['a', 'b', '']，共3行\\n lines = content.split('\\\\n')\\n line_count = len(lines)\\n \\n # 如果存在任何一个代码文件总行数为偶数，则不满足要求\\n if line_count % 2 == 0:\\n return False\\n \\n has_code_file = True\\n except Exception:\\n # 忽略不可读或编码有问题的文件\\n continue\\n \\n # 如果项目中没有任何代码文件，严格意义上不算满足生成代码行数为奇数的条件\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"前端开发小助手\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 检查助手的回复中是否出现了“我”\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 匹配独立的弹窗函数调用，或者是 window. 前缀的调用\\n # (?:^|[^\\\\w\\\\.]) : 确保前面不是字母数字下划线或其他的点（避免匹配对象的自定义属性调用如 xxx.alert）\\n # (?:window\\\\.)? : 允许匹配 window.alert 等原生形式\\n # (alert|confirm|prompt) : 匹配三种弹窗关键字\\n # \\\\s*\\\\( : 匹配紧跟的括号，确认其为函数调用\\n pattern = re.compile(r'(?:^|[^\\\\w\\\\.])(?:window\\\\.)?(alert|confirm|prompt)\\\\s*\\\\(')\\n \\n # 结合项目类型，定义可能包含 JavaScript 代码的文件扩展名\\n valid_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略掉可能产生干扰的无关目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若文件中存在 alert, confirm 或 prompt 的原生调用，则不满足要求\\n if pattern.search(content):\\n return False\\n except Exception:\\n # 忽略不可读取的文件\\n pass\\n \\n # 检查完所有相关代码文件均未发现原生弹窗语句，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在的挑战来了，我们需要实现复杂的“合并单元格”功能：\\n1. 实现多选机制：用户可以通过在表格内按下鼠标并拖拽，选中一个矩形区域内的多个单元格（给被选中的区域加上半透明背景色）。\\n2. 增加一个“合并单元格”按钮。当用户选中了大于1个单元格并点击此按钮时，将选区合并为一个大的单元格（使用HTML的rowspan和colspan属性）。\\n3. 合并后的单元格文本保留选区内左上角第一个有内容的单元格文本，选区内其他单元格在数据结构中需标记为“被隐藏/已合并”，不要在页面中渲染出来。\\n4. 每次合并操作完成后，必须将带有合并状态的数据结构保存到LocalStorage，并确保页面刷新后依然能正确读取并渲染合并后的表格外观。\\n\\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\\n1. 原“必须始终且仅用双引号，来括住你回复内容中出现的文件路径或代码位置”的要求调整为：请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来。\\n2. 新增要求：在对任何已有文件执行修改操作前，请务必在相同目录层级下为其生成一份修改前的备份文件。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def check_js(text: str) -> bool:\\n # 正则表达式按优先级捕获：\\n # 1. string: 双引号、单引号或反引号括起来的字符串字面量（包含转义字符支持）\\n # 2. block_comment: /* ... */ 格式的多行注释\\n # 3. line_comment: // 格式的单行注释（排除前面是冒号的情况如 http://）\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<block_comment>/\\\\*.*?\\\\*/)'\\n r'|(?P<line_comment>(?<!:)//[^\\\\r\\\\n]*)',\\n re.DOTALL | re.MULTILINE\\n )\\n for match in pattern.finditer(text):\\n # 如果匹配到的内容属于注释分组，则表明包含注释\\n if match.group('block_comment') or match.group('line_comment'):\\n return True\\n return False\\n \\n def check_css(text: str) -> bool:\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<block_comment>/\\\\*.*?\\\\*/)',\\n re.DOTALL | re.MULTILINE\\n )\\n for match in pattern.finditer(text):\\n if match.group('block_comment'):\\n return True\\n return False\\n\\n def check_no_comments(file_path: str) -> bool:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果文件无法读取，暂且跳过不判定为含注释\\n return True\\n \\n ext = file_path.split('.')[-1].lower()\\n \\n if ext in ['js', 'jsx', 'ts', 'tsx']:\\n if check_js(content):\\n return False\\n elif ext == 'css':\\n if check_css(content):\\n return False\\n elif ext in ['html', 'htm']:\\n # 检查直接编写的HTML注释\\n if re.search(r'<!--.*?-->', content, re.DOTALL):\\n return False\\n # 提取所有 <script> 标签中的内容进行 JS 注释检查\\n for script_match in re.finditer(r'<script[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL):\\n if check_js(script_match.group(1)):\\n return False\\n # 提取所有 <style> 标签中的内容进行 CSS 注释检查\\n for style_match in re.finditer(r'<style[^>]*>(.*?)</style>', content, re.IGNORECASE | re.DOTALL):\\n if check_css(style_match.group(1)):\\n return False\\n \\n return True\\n\\n valid_exts = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见无关目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n # 只要发现任意一个有效代码文件内含有注释，即不满足要求\\n if not check_no_comments(file_path):\\n return False\\n \\n # 未在任何有效代码文件里发现注释，要求得到满足\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_odd_lines_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户的指令，推断纯HTML/CSS/JS应用及单元测试文件可能具有的代码后缀\\n code_exts = {'.html', '.css', '.js'}\\n \\n # 需要排除的由第三方工具自动生成的或非助手直接生成的常见目录\\n exclude_dirs = {'node_modules', 'coverage', '.nyc_output', '.git', 'dist', 'build'}\\n \\n file_found = False\\n\\n # 若工作区路径不存在，直接返回未满足\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 列表，使得 os.walk 跳过被排除的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n # 判断是否为推断出的代码文件类型\\n if any(file.endswith(ext) for ext in code_exts):\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 依据换行符进行分割可以确保所有空行（包括文件末尾由于换行符多产生的空行）都能计算在内\\n lines = content.split('\\\\n')\\n \\n # 检查总行数是否为偶数\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果某个文件因为编码或权限等原因无法读取，直接跳过，不应引起验证程序崩溃\\n continue\\n\\n # 如果没有找到任何代码文件，说明没有生成所需代码，返回False；否则所有检查通过，返回True\\n return file_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的函数覆盖率大于等于98%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"前端开发小助手\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复中是否没有使用\\\"我\\\"作为自我称呼。\\n 严格模式下，直接检查字符串中是否包含\\\"我\\\"字。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句\",\n \"validation_code\": \"def check_popup_statements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义需要检查的代码文件类型（基于纯前端项目推断）\\n code_extensions = ('.html', '.js')\\n \\n # 遍历当前工作空间内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 去除 HTML 注释，防止误判\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 去除 JavaScript 多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 去除 JavaScript 单行注释\\n content = re.sub(r'//.*', '', content)\\n \\n # 匹配原生弹窗函数调用，涵盖 alert(), confirm(), prompt() 及其带有空格的情况\\n if re.search(r'\\\\b(alert|confirm|prompt)\\\\s*\\\\(', content):\\n return False\\n except Exception:\\n # 忽略读取异常的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在对任何已有文件执行修改操作前，请务必在相同目录层级下为其生成一份修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后一步，补充“取消合并”功能，并完善交互细节：\\n1. 在操作栏增加一个“取消合并”按钮。\\n2. 如果用户单击选中的是一个之前被合并过的巨大单元格（rowspan或colspan > 1），点击“取消合并”按钮后，将其恢复成原来数量的独立单格（rowspan和colspan恢复为1），原隐藏的单元格重新显示为空白单元格。\\n3. 取消合并后，不仅要重新渲染视图，还要更新LocalStorage，把对应的合并标记清除。\\n4. 优化整个拖拽多选的体验和界面的CSS样式，确保在包含合并单元格的表格中进行行列插入、删除时，数据结构不会乱掉（如果太复杂，可以限制为：当前选中区域如果有合并单元格时禁用增删行列按钮）。\\n\\n另外，有两项要求更新及新增，请在后续所有操作中始终遵循：\\n1. 原单元测试函数覆盖率≥98%的要求调整为：请确保生成的单元测试文件运行后，能够让目标代码的语句覆盖率大于等于99%。\\n2. 新增要求：凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"无论进行何种文件操作，前提条件都是先运行`pwd`来获取当前目录位置\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"使用 ESLint 对你编写的代码进行校验是必须的，同时你要排查并修复一切报错，保证最终 ESLint 的检查结果没有任何错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的分支覆盖率大于等于90%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"不要在你生成的任意一个代码文件中写入注释\",\n \"validation_code\": \"def check_no_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见前端代码文件扩展名\\n target_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能包含依赖或版本控制文件的常见无关目录\\n if 'node_modules' in root.split(os.sep) or '.git' in root.split(os.sep):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in target_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件不可读，则跳过\\n continue\\n \\n # 为了防止误判，将字符串字面量内容剔除（替换为成对的空引号）\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', content)\\n \\n # 将 url() 中的内容剔除，防止无引号的链接双斜杠被识别为单行注释\\n content = re.sub(r'url\\\\([^)]*\\\\)', 'url()', content)\\n \\n # 检测 HTML 注释\\n if re.search(r'<!--[\\\\s\\\\S]*?-->', content):\\n return False\\n \\n # 检测 块级注释 (CSS/JS 等通用)\\n if re.search(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content):\\n return False\\n \\n # 检测 JS/HTML 中的单行注释，排除形如 \\\"http://\\\" 中冒号后的斜杠\\n if ext in {'.html', '.vue', '.js', '.jsx', '.ts', '.tsx'}:\\n if re.search(r'(?<!:)//', content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_code_lines_odd(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令内容推断，可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n code_file_found = False\\n \\n # 遍历项目目录下的所有文件（包括子目录中的测试文件等）\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统计总行数（将空行计算在内）\\n # 以 '\\\\n' 为分隔符进行 split 操作，返回的列表长度即为包含空行在内的确切文件行数\\n line_count = len(content.split('\\\\n'))\\n \\n # 如果有任意一个代码文件的行数为偶数，则说明不满足要求\\n if line_count % 2 == 0:\\n return False\\n \\n code_file_found = True\\n except Exception:\\n # 忽略读取异常的文件\\n pass\\n \\n # 如果遍历完毕未出现偶数行，且至少找到了一份有效代码文件，则要求被满足\\n return code_file_found\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请确保对生成代码产生的所有中间测试代码得以保留，且被安置在项目目录的__tests__子目录内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的语句覆盖率大于等于99%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保在输出的任何文字中，自我称呼一律使用\\\"前端开发小助手\\\"而不是\\\"我\\\"\",\n \"validation_code\": \"def check_self_reference_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复内容中的自我称呼是否符合要求。\\n 要求：确保在输出的任何文字中，自我称呼一律使用\\\"前端开发小助手\\\"而不是\\\"我\\\"。\\n \\\"\\\"\\\"\\n \\n # 1. 确保回复的任何文字中没有出现“我”字\\n if \\\"我\\\" in response:\\n return False\\n \\n # 2. 确保之前轮次要求的“小助手”被准确更新为“前端开发小助手”\\n # 即回复中出现所有的“小助手”时，其前缀必须是“前端开发”\\n idx = response.find(\\\"小助手\\\")\\n while idx != -1:\\n # 提取“小助手”前面的4个字符，如果不足4个或者不是\\\"前端开发\\\"，则不符合要求\\n if idx < 4 or response[idx-4:idx] != \\\"前端开发\\\":\\n return False\\n # 继续寻找下一个“小助手”\\n idx = response.find(\\\"小助手\\\", idx + 1)\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"确保你生成的任何代码文件中都不包含alert、confirm、prompt三类原生JavaScript弹窗语句\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断生成的代码项目中是否完全不包含 alert、confirm、prompt 三类原生JavaScript弹窗语句。\\n \\\"\\\"\\\"\\n # 可能包含 JavaScript 弹窗语句的目标代码文件后缀名\\n target_extensions = ('.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.mjs', '.cjs')\\n \\n # 正则表达式匹配原生的 alert, confirm, prompt 调用\\n # 涵盖 alert(...) 或 window.alert(...) 等形式\\n # [^a-zA-Z0-9_$.] 用于确保前面没有其他变量名拼接或其他的对象调用（例如 custom_alert 或 dialog.alert）\\n pattern = re.compile(r'(?:^|[^a-zA-Z0-9_$.])(?:window\\\\.)?(alert|confirm|prompt)\\\\s*\\\\(')\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(target_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果在代码文件中找到了任一原生弹窗语句，则代表未满足要求，直接返回 False\\n if pattern.search(content):\\n return False\\n except Exception:\\n # 忽略无法正常读取的文件（例如由于权限、编码异常等导致）\\n continue\\n \\n # 所有的代码文件检查均未发现原生弹窗语句，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每次编辑已有文件之前，必须先借助对应工具审阅该文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在你的回复里，只要涉及到文件路径或代码位置，就只能使用单引号将其括起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在对任何已有文件执行修改操作前，请务必在相同目录层级下为其生成一份修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进task_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["验证页面初始化。打开应用主页，截图验证页面包含顶部标题导航栏、中间有表格列表区域、且有明显的'新建表格'按钮。","验证新建表格与跳转功能。在主页点击'新建表格'按钮，验证页面是否隐藏主页并跳转到表格编辑页面，编辑页顶部出现'返回'按钮，且页面渲染出一个默认的5行5列的空白表格。","验证表格默认标题与返回功能。在新建的表格编辑页中，验证顶部显示的标题默认是否为'未命名表格'；点击'返回'按钮，验证页面是否切回主页面，且主页列表中新增了一张包含修改时间的表格卡片。","验证编辑标题功能与持久化。在表格编辑页中点击顶部的表格标题将其修改为'测试表格'并点击页面空白处失焦；点击返回主页再重新点击该卡片进入，截图验证标题是否依然保持为'测试表格'。","验证单元格单击高亮。在表格编辑页中，用鼠标左键单击任意一个普通单元格，截图验证该单元格是否出现高亮边框或背景色，以表示被选中激活。","验证单元格编辑逻辑。双击任意单元格，验证其是否变为输入框（input）；输入文本'测试数据'后按回车键或点击其他区域失焦，验证输入框是否恢复为普通文本并展示刚输入的内容。","验证单元格数据本地持久化。完成单元格内容编辑并展示文本后，刷新浏览器当前网页，截图验证刚刚输入的文本是否依然保留在原单元格中，未发生丢失。","验证插入行功能。单击选中任意单元格，分别点击'在上方插入行'和'在下方插入行'按钮各一次，截图验证当前表格的总行数是否正确增加了2行，原单元格内容位置是否正确平移。","验证插入列功能。单击选中任意单元格，分别点击'在左侧插入列'和'在右侧插入列'按钮各一次，截图验证当前表格的总列数是否正确增加了2列。","验证删除行列功能与持久化。在表格中单击选中某一个空白单元格，依次点击'删除当前行'和'删除当前列'，验证表格的行列数是否各自减少了1；刷新网页，验证删除后的表格结构尺寸是否被保存。","验证行列删除极值保护。在一个表格中连续点击'删除当前行'直到只剩最后1行，再次点击'删除当前行'，截图并验证此时表格是否至少保留1行未被删除（或者有弹窗提示不许删除）；列的极值删除验证同理。","验证鼠标拖拽多选交互。在表格内按住鼠标左键并在相邻的多个单元格上拖拽（比如跨越2x2区域），松开鼠标，截图验证被拖拽选中的矩形区域单元格是否都带有半透明背景色高亮。","验证合并单元格功能及文本保留。在左上角单元格双击输入'合并测试'，接着鼠标拖拽选中包含该带有文字单元格的2x2矩形选区，点击'合并单元格'按钮，截图验证这4个单元格是否在界面上合并为1个大的单元格，且大单元格内保留文字'合并测试'。","验证合并状态本地持久化。完成上一条合并操作后，刷新当前网页，截图验证合并后的巨大单元格外观以及保留的文本是否能完整且结构正确地恢复渲染。","验证取消合并功能。单击刚刚合并得到的大单元格使其高亮，然后点击'取消合并'按钮，截图验证该大单元格是否重新恢复为4个独立的基础单元格，且除了原本带字的单元格外，其余拆分出的单元格均显示为空白。","验证取消合并后的持久化。完成取消合并操作后，刷新网页，截图验证表格是否仍呈现为拆分后的独立单元格状态，没有错误恢复成此前合并的大单元格。","验证主页表格删除与二次确认逻辑。在编辑页点击返回主页，点击某个表格卡片右上角的'删除'按钮，验证是否弹出确认弹窗；点击弹窗确认后，截图验证该卡片是否从列表中消失；刷新主页面验证被删卡片未重新出现。","验证存在合并单元格时的行列增删安全机制。在一个已经存在合并单元格的表格中，当选中包含了合并单元格的区域时，尝试操作插入或删除行列，验证是否实现了对应的安全保护逻辑（如成功限制并禁用了插入/删除行列的按钮，使按钮置灰不可点击，或进行了不致数据错乱的安全渲染处理）。"],"string":"[\n \"验证页面初始化。打开应用主页，截图验证页面包含顶部标题导航栏、中间有表格列表区域、且有明显的'新建表格'按钮。\",\n \"验证新建表格与跳转功能。在主页点击'新建表格'按钮，验证页面是否隐藏主页并跳转到表格编辑页面，编辑页顶部出现'返回'按钮，且页面渲染出一个默认的5行5列的空白表格。\",\n \"验证表格默认标题与返回功能。在新建的表格编辑页中，验证顶部显示的标题默认是否为'未命名表格'；点击'返回'按钮，验证页面是否切回主页面，且主页列表中新增了一张包含修改时间的表格卡片。\",\n \"验证编辑标题功能与持久化。在表格编辑页中点击顶部的表格标题将其修改为'测试表格'并点击页面空白处失焦；点击返回主页再重新点击该卡片进入，截图验证标题是否依然保持为'测试表格'。\",\n \"验证单元格单击高亮。在表格编辑页中，用鼠标左键单击任意一个普通单元格，截图验证该单元格是否出现高亮边框或背景色，以表示被选中激活。\",\n \"验证单元格编辑逻辑。双击任意单元格，验证其是否变为输入框（input）；输入文本'测试数据'后按回车键或点击其他区域失焦，验证输入框是否恢复为普通文本并展示刚输入的内容。\",\n \"验证单元格数据本地持久化。完成单元格内容编辑并展示文本后，刷新浏览器当前网页，截图验证刚刚输入的文本是否依然保留在原单元格中，未发生丢失。\",\n \"验证插入行功能。单击选中任意单元格，分别点击'在上方插入行'和'在下方插入行'按钮各一次，截图验证当前表格的总行数是否正确增加了2行，原单元格内容位置是否正确平移。\",\n \"验证插入列功能。单击选中任意单元格，分别点击'在左侧插入列'和'在右侧插入列'按钮各一次，截图验证当前表格的总列数是否正确增加了2列。\",\n \"验证删除行列功能与持久化。在表格中单击选中某一个空白单元格，依次点击'删除当前行'和'删除当前列'，验证表格的行列数是否各自减少了1；刷新网页，验证删除后的表格结构尺寸是否被保存。\",\n \"验证行列删除极值保护。在一个表格中连续点击'删除当前行'直到只剩最后1行，再次点击'删除当前行'，截图并验证此时表格是否至少保留1行未被删除（或者有弹窗提示不许删除）；列的极值删除验证同理。\",\n \"验证鼠标拖拽多选交互。在表格内按住鼠标左键并在相邻的多个单元格上拖拽（比如跨越2x2区域），松开鼠标，截图验证被拖拽选中的矩形区域单元格是否都带有半透明背景色高亮。\",\n \"验证合并单元格功能及文本保留。在左上角单元格双击输入'合并测试'，接着鼠标拖拽选中包含该带有文字单元格的2x2矩形选区，点击'合并单元格'按钮，截图验证这4个单元格是否在界面上合并为1个大的单元格，且大单元格内保留文字'合并测试'。\",\n \"验证合并状态本地持久化。完成上一条合并操作后，刷新当前网页，截图验证合并后的巨大单元格外观以及保留的文本是否能完整且结构正确地恢复渲染。\",\n \"验证取消合并功能。单击刚刚合并得到的大单元格使其高亮，然后点击'取消合并'按钮，截图验证该大单元格是否重新恢复为4个独立的基础单元格，且除了原本带字的单元格外，其余拆分出的单元格均显示为空白。\",\n \"验证取消合并后的持久化。完成取消合并操作后，刷新网页，截图验证表格是否仍呈现为拆分后的独立单元格状态，没有错误恢复成此前合并的大单元格。\",\n \"验证主页表格删除与二次确认逻辑。在编辑页点击返回主页，点击某个表格卡片右上角的'删除'按钮，验证是否弹出确认弹窗；点击弹窗确认后，截图验证该卡片是否从列表中消失；刷新主页面验证被删卡片未重新出现。\",\n \"验证存在合并单元格时的行列增删安全机制。在一个已经存在合并单元格的表格中，当选中包含了合并单元格的区域时，尝试操作插入或删除行列，验证是否实现了对应的安全保护逻辑（如成功限制并禁用了插入/删除行列的按钮，使按钮置灰不可点击，或进行了不致数据错乱的安全渲染处理）。\"\n]","isConversational":false}}},{"rowIdx":76,"cells":{"task_id":{"kind":"number","value":77,"string":"77"},"repository_policy":{"kind":"string","value":"后续所有交互轮次请你始终遵循以下约束要求：\n1. 任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改。\n2. 在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录。\n3. 请将生成的代码项目目录深度控制在2层及以内。\n4. 请将你生成的代码文件中所有函数的圈复杂度控制在3或以下。\n5. 请确保你生成的所有代码文件中都包含 \"TODO\"。\n6. 无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词。\n7. 请在对所有用户进行回复的最后，统一加上\"呜～\"语气词。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"import os\nfrom pathlib import Path\n\ndef check_project_directory_depth(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证生成的代码项目目录深度是否控制在2层及以内。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: bool，代表是否满足深度限制要求\n \"\"\"\n # 如果工作空间路径不存在，说明没有生成代码项目，无法满足“控制深度”的前提，返回False\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏目录以及常见的编译/环境缓存等非代码逻辑目录\n # 修改 dirs 列表本身，可以阻止 os.walk 继续向下遍历这些目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\n \n # 遍历当前层级下的所有文件和合格的子目录\n for name in dirs + files:\n # 过滤隐藏文件\n if name.startswith('.'):\n continue\n \n # 拼接绝对路径\n full_path = os.path.join(root, name)\n # 计算相对于工作空间的相对路径\n rel_path = os.path.relpath(full_path, workspace_path)\n \n # 获取相对路径的层级分段\n parts = Path(rel_path).parts\n \n # 若分段数大于2，说明存在超过2层结构（即第3层及以上的文件/目录），不满足要求\n if len(parts) > 2:\n return False\n \n # 遍历结束且没有发现超过2层的结构，说明符合要求\n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件后缀名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', \n '.go', '.rb', '.php', '.cs', '.sh', '.swift', '.kt', \n '.rs', '.m', '.scala', '.dart', '.sql', '.vue'\n }\n \n found_code_file = False\n \n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录（如 .git, .vscode, .idea 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n # 读取代码文件内容\n content = \"\"\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n # 如果UTF-8解码失败，尝试使用GBK编码读取\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n content = f.read()\n except Exception:\n return False\n except Exception:\n return False\n \n # 检查文件内容中是否包含大写的 \"TODO\"\n if \"TODO\" not in content:\n return False\n \n # 确保生成了至少一个代码文件\n if not found_code_file:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 移除回复可能存在的前导空白字符，并判断最开头是否为\"喵～\"\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 使用正则表达式检查回复是否以\"呜～\"结尾\n # \\W* 允许\"呜～\"后面跟有零个或多个标点符号或空白字符（如换行、空格、句号等）\n # $ 匹配字符串的末尾\n pattern = r'呜～\\W*$'\n \n return bool(re.search(pattern, response))","tags":["Content","Persona"]}],"string":"[\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"import os\\nfrom pathlib import Path\\n\\ndef check_project_directory_depth(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证生成的代码项目目录深度是否控制在2层及以内。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: bool，代表是否满足深度限制要求\\n \\\"\\\"\\\"\\n # 如果工作空间路径不存在，说明没有生成代码项目，无法满足“控制深度”的前提，返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏目录以及常见的编译/环境缓存等非代码逻辑目录\\n # 修改 dirs 列表本身，可以阻止 os.walk 继续向下遍历这些目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('__pycache__', 'node_modules', 'venv', 'env')]\\n \\n # 遍历当前层级下的所有文件和合格的子目录\\n for name in dirs + files:\\n # 过滤隐藏文件\\n if name.startswith('.'):\\n continue\\n \\n # 拼接绝对路径\\n full_path = os.path.join(root, name)\\n # 计算相对于工作空间的相对路径\\n rel_path = os.path.relpath(full_path, workspace_path)\\n \\n # 获取相对路径的层级分段\\n parts = Path(rel_path).parts\\n \\n # 若分段数大于2，说明存在超过2层结构（即第3层及以上的文件/目录），不满足要求\\n if len(parts) > 2:\\n return False\\n \\n # 遍历结束且没有发现超过2层的结构，说明符合要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件后缀名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \\n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.scss', \\n '.go', '.rb', '.php', '.cs', '.sh', '.swift', '.kt', \\n '.rs', '.m', '.scala', '.dart', '.sql', '.vue'\\n }\\n \\n found_code_file = False\\n \\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录（如 .git, .vscode, .idea 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 读取代码文件内容\\n content = \\\"\\\"\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n # 如果UTF-8解码失败，尝试使用GBK编码读取\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n content = f.read()\\n except Exception:\\n return False\\n except Exception:\\n return False\\n \\n # 检查文件内容中是否包含大写的 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n \\n # 确保生成了至少一个代码文件\\n if not found_code_file:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 移除回复可能存在的前导空白字符，并判断最开头是否为\\\"喵～\\\"\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 使用正则表达式检查回复是否以\\\"呜～\\\"结尾\\n # \\\\W* 允许\\\"呜～\\\"后面跟有零个或多个标点符号或空白字符（如换行、空格、句号等）\\n # $ 匹配字符串的末尾\\n pattern = r'呜～\\\\W*$'\\n \\n return bool(re.search(pattern, response))\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"我们需要用HTML、CSS和JavaScript实现一个经典的网页小游戏“黄金矿工”。这一步先搭建游戏的基础骨架。请创建一个页面，包含一个“首页”视图和一个“游戏”视图。首页中心有两个按钮：“单人模式”和“双人模式”。点击“单人模式”后能隐藏首页并显示游戏视图。游戏视图使用Canvas绘制，划分为上下两部分：上部占屏幕大约1/4的蓝天和地表，下部占3/4的棕色地下矿区。在游戏视图顶部的地表正中央，画一个简单的正方形代表矿工的位置，并在其下方画一个半圆代表“钩子”的初始占位。为了后续扩展，请规划好整体的JavaScript类或对象结构。请你在后续全流程开发中始终遵循以下两个额外要求：第一，生成的各个代码文件内，注释部分需完全使用英文撰写；第二，强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n # 检查工作区路径是否存在且为一个合法的目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 遍历整个项目的目录树\n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏目录，不将 .git, .idea 等非代码结构性的隐藏文件夹计入深度考量中\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 获取当前目录相对项目根目录的路径\n rel_root = os.path.relpath(root, workspace_path)\n \n # 如果是项目根目录（即 '.'），其深度定义为1层，暂不处理直接跳过\n if rel_root == '.':\n continue\n \n # 使用 pathlib 的 parts 属性将相对路径跨平台地分割为元组\n # 例如 'js/utils' 会被分割为 ('js', 'utils')\n parts = Path(rel_root).parts\n \n # 当前目录总深度 = 根目录的1层 + 相对路径的层级数\n depth = 1 + len(parts)\n \n # 要求目录深度控制在2层及以内，一旦大于2层则直接判错\n if depth > 2:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n code_extensions = ['.html', '.css', '.js']\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 检查文件内容是否包含 \"TODO\"\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果由于某种原因文件无法读取，出于严谨性考虑，直接返回 False\n return False\n \n # 如果找到至少一个代码文件且都通过了检查，返回 True；如果没有找到任何代码文件，返回 False\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除回复两端可能存在的空白字符（如空格、换行符等）\n cleaned_response = response.strip()\n \n # 判断清理后的回复是否以\"喵～\"开头\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n import re\n \n # 移除回复末尾可能存在的空白字符（空格、换行等）以及常见的标点符号\n cleaned_response = re.sub(r'[\\s。，！!？\\.]+$', '', response)\n \n # 判断清理后的字符串是否以\"呜～\"结尾，同时为了以防万一，也检查原生清理空白后的字符串\n if cleaned_response.endswith('呜～') or response.strip().endswith('呜～'):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"生成的各个代码文件内，注释部分需完全使用英文撰写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对前端游戏项目的常规代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n code_files_found = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n # 使用utf-8编码读取文件，忽略可能的解码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n \n # 判断顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 文件读取发生异常（如无权限等）也视为未满足\n return False\n\n # 必须至少找到一个代码文件且全部满足要求才返回 True\n return code_files_found","tags":["Content","Format"]}]},{"round_id":1,"instruction":"好，基础页面有了，接下来让单人模式下的钩子动起来。请实现钩子的自动摆动逻辑：钩子应当以顶部正方形为锚点，在左侧-75度到右侧75度的范围内像钟摆一样来回匀速摆动。接着加入键盘事件监听，当玩家按下键盘的“S”键时，钩子停止摆动，并沿着按下瞬间所指向的角度直线往下伸出。如果钩子触碰到了Canvas的左右边缘或底部边缘，则停止伸出，并沿着原路倒退收回。完全收回到起点后，重新开始左右摆动。\n另外从本轮开发开始，后续全流程需始终遵循以下更新要求：第一，原注释全英文要求调整为无论生成多少个代码文件，其内部的注释内容均要求全是中文；第二，限制要求：你生成的全部代码文件中一律不得出现 \"debugger\"。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，判定为不满足\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for name in files + dirs:\n # 拼接当前文件或文件夹的绝对路径\n full_path = os.path.join(root, name)\n \n # 计算相对于工作区根目录的相对路径\n rel_path = os.path.relpath(full_path, workspace_path)\n \n # 统一路径分隔符以便后续拆分\n rel_path = rel_path.replace('\\\\', '/')\n \n # 拆分出路径层级组件，并过滤掉空字符串和当前目录占位符\n components = [c for c in rel_path.split('/') if c and c != '.']\n \n # 如果组件长度大于 2，代表其目录深度达到了3层或以上，不满足要求\n if len(components) > 2:\n return False\n\n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"import os\n\ndef check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令推测相关的代码文件后缀名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果文件读取失败，无法验证是否包含 \"TODO\"，保守返回 False\n return False\n \n # 如果找到了至少一个代码文件，并且上面的检查全通过，则满足条件\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 检查回复的最开头是否是\"喵～\"，兼容前面可能有空白字符的情况\n return response.startswith(\"喵～\") or response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复最后是否统一加上了\"呜～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除末尾的空白字符以及常见的结束标点符号，避免标点符号影响末尾判断\n cleaned_response = response.rstrip(' \\t\\n\\r。！!.()[]{}（）【】\\'\\\"”’')\n \n # 检查清理后的回复是否以\"呜～\"结尾\n return cleaned_response.endswith('呜～')","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令，推断可能的代码文件扩展名\n code_extensions = ['.html', '.css', '.js']\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 只读取文件的第一行\n first_line = f.readline()\n \n # 检查第一行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法正常读取，视作不满足要求\n return False\n \n # 如果没有找到任何相关的代码文件，则也视为不满足\n return found_code_file","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"限制要求：你生成的全部代码文件中一律不得出现 \"debugger\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推测可能的代码文件后缀类型\n code_extensions = {'.html', '.css', '.js'}\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是对应的代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查内容中是否包含 \"debugger\"\n if 'debugger' in content:\n return False\n except Exception:\n # 遇到无法使用utf-8解码的文件直接跳过\n continue\n \n return True","tags":["Content","Keyword"]}]},{"round_id":2,"instruction":"非常棒，钩子能正常伸缩了。现在来丰富地下矿区。请在游戏开始时，在地下区域随机生成以下静态物品：金矿（黄色圆形，分为大、中、小三种尺寸）、石头（灰色矩形，分为大、中、小三种尺寸）、钻石（亮蓝色菱形，要求每次游戏至少生成3个）、炸弹（黑色圆形，顶部带一根红线）。这些物品的位置应该随机分布在地下区域，且尽量避免严重的重叠现象。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原禁止出现\"debugger\"的要求调整为你生成的全部代码文件中一律不得出现\"console.log\"；第二，生成的代码文件必须满足：单个函数的规模不得大于40行。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录（以 . 开头的目录）以及 __pycache__ 等无关目录，避免干扰层级判断\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n # 计算相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n if rel_path == '.':\n depth = 1 # 根目录自身计为第1层\n else:\n # 统一路径分隔符并拆分，计算相对于根目录嵌套了多少层\n parts = rel_path.replace('\\\\', '/').split('/')\n parts = [p for p in parts if p] # 去除可能出现的空字符串\n depth = len(parts) + 1\n \n # 如果当前遍历到的目录深度大于 2 层，则违反了“控制在2层及以内”的要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 根据用户指令背景，推断代码文件类型主要是前端相关代码以及可能存在的脚本文件\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n # 遍历当前项目工作目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选出代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 以 utf-8 编码读取代码文件内容，忽略少量无法解码的异常字符\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 判断 \"TODO\" 字符串是否在代码文件内容中\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果由于权限等问题无法读取，为了严格验证可以视为不满足条件\n return False\n \n # 如果项目中存在代码文件并且都包含\"TODO\"，则满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复的最开头是否带有\"喵～\"语气词。\n \"\"\"\n # 去除回复内容最前面的所有空白字符（包括空格、换行等），判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 去除前后空白字符\n response = response.strip()\n if not response:\n return False\n \n if response.endswith(\"呜～\") or response.endswith(\"呜~\"):\n return True\n \n # 定义允许出现在“呜～”之后的常见标点符号和空白字符（不包含“～”和“~”）\n end_chars = \"!?.！。？ \\n\\r\\t\\\"'”’\"\n \n # 从尾部逐个去除这些允许的结尾符号\n temp_resp = response\n while temp_resp and temp_resp[-1] in end_chars:\n temp_resp = temp_resp[:-1]\n \n # 判断去除尾部标点后是否以指定的语气词结尾\n return temp_resp.endswith(\"呜～\") or temp_resp.endswith(\"呜~\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件无法正常读取，视作未满足要求\n return False\n\n # 必须保证项目中包含至少一个代码文件\n return has_code_file","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见前端项目代码文件的后缀名\n code_extensions = ['.html', '.js', '.css', '.ts', '.jsx', '.tsx']\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 判断是否包含 \"console.log\"\n if 'console.log' in content:\n return False\n except Exception:\n # 如果遇到无法用 utf-8 读取的文件则跳过\n continue\n\n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于40行","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n def check_js_functions(js_code: str) -> bool:\n # 使用状态机将注释和字符串字面量覆盖为空格，避免其内部的大括号和关键字干扰正则和计数，同时保留换行符保持行号一致\n state = 'NORMAL'\n masked = []\n i = 0\n n = len(js_code)\n while i < n:\n char = js_code[i]\n if state == 'NORMAL':\n if char in \"'\\\"`\":\n state = 'IN_STRING_' + char\n masked.append(' ')\n elif char == '/' and i + 1 < n and js_code[i+1] == '/':\n state = 'IN_COMMENT_SL'\n masked.append(' ')\n masked.append(' ')\n i += 1\n elif char == '/' and i + 1 < n and js_code[i+1] == '*':\n state = 'IN_COMMENT_ML'\n masked.append(' ')\n masked.append(' ')\n i += 1\n else:\n masked.append(char)\n elif state.startswith('IN_STRING_'):\n quote = state[-1]\n if char == '\\\\':\n masked.append(' ')\n if i + 1 < n:\n masked.append('\\n' if js_code[i+1] == '\\n' else ' ')\n i += 1\n elif char == quote:\n state = 'NORMAL'\n masked.append(' ')\n else:\n masked.append('\\n' if char == '\\n' else ' ')\n elif state == 'IN_COMMENT_SL':\n if char == '\\n':\n state = 'NORMAL'\n masked.append('\\n')\n else:\n masked.append(' ')\n elif state == 'IN_COMMENT_ML':\n if char == '*' and i + 1 < n and js_code[i+1] == '/':\n state = 'NORMAL'\n masked.append(' ')\n masked.append(' ')\n i += 1\n else:\n masked.append('\\n' if char == '\\n' else ' ')\n i += 1\n \n masked_code = \"\".join(masked)\n \n # 寻找疑似函数定义的开头（包括常规 function 声明、箭头函数、类方法及对象方法的首行定义）\n pattern = re.compile(\n r'\\bfunction\\s*(?:[a-zA-Z_$][\\w$]*\\s*)?\\([^)]*\\)\\s*\\{'\n r'|(?:\\([^)]*\\)|\\b[a-zA-Z_$][\\w$]*)\\s*=>\\s*\\{'\n r'|^[ \\t]*(?!(?:if|for|while|catch|switch|with|else|return|typeof|instanceof)\\b)[a-zA-Z_$][\\w$]*\\s*\\([^)]*\\)\\s*\\{',\n re.MULTILINE\n )\n \n for match in pattern.finditer(masked_code):\n start_idx = match.end() - 1\n # 找到这段匹配中的大括号确切位置\n for j in range(match.start(), match.end()):\n if masked_code[j] == '{':\n start_idx = j\n break\n \n brace_count = 0\n end_idx = -1\n # 寻找函数体对应的右大括号\n for j in range(start_idx, n):\n if masked_code[j] == '{':\n brace_count += 1\n elif masked_code[j] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_idx = j\n break\n \n # 如果成功找到闭合大括号，则统计这区间的行数\n if end_idx != -1:\n func_lines = masked_code[match.start():end_idx+1].count('\\n') + 1\n if func_lines > 40:\n return False\n return True\n\n def check_html_functions(html_code: str) -> bool:\n # 匹配提取 HTML 文件中的所有 <script> 块内容\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', html_code, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_js_functions(script):\n return False\n return True\n\n def check_py_functions(py_code: str) -> bool:\n try:\n tree = ast.parse(py_code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n lines = node.end_lineno - node.lineno + 1\n if lines > 40:\n return False\n return True\n except Exception:\n # 语法错误的代码跳过不计\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n\n # 针对不同扩展名调用检测函数\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n if not check_js_functions(content):\n return False\n elif file.endswith(('.html', '.htm')):\n if not check_html_functions(content):\n return False\n elif file.endswith('.py'):\n if not check_py_functions(content):\n return False\n \n return True","tags":["Quantity","Range"]}]},{"round_id":3,"instruction":"下一步实现核心的抓取逻辑。在钩子伸出阶段，请加入碰撞检测功能。当钩子与地下的任意物品（金矿、石头、钻石或炸弹）发生碰撞重叠时，钩子立刻停止向下伸出，转为收回状态。在收回的过程中，被抓到的物品需要吸附在钩子末端，随着钩子一起移动回到起点。一旦完全拉回到起点，该物品从画面中消失，钩子恢复摆动准备下一次抓取。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于40行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于30行；第二，务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n if not os.path.isdir(workspace_path):\n return False\n \n for dirpath, dirnames, filenames in os.walk(workspace_path):\n rel_path = os.path.relpath(dirpath, workspace_path)\n \n # 根目录的相对路径为 '.'\n if rel_path == '.':\n depth = 0\n else:\n # 统一路径格式，利用系统路径分隔符(os.sep)计算当前目录的相对深度\n rel_path = os.path.normpath(rel_path)\n depth = len(rel_path.split(os.sep))\n \n # 深度 >= 2 意味着当前所在目录（如 workspace/dir1/dir2）已经是项目的第2层\n # 如果此目录下还有子文件夹(dirnames)或文件(filenames)，则代表存在第3层甚至更深的项目结构\n if depth >= 2 and (len(dirnames) > 0 or len(filenames) > 0):\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令，推断目标代码文件类型为HTML、CSS和JavaScript\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 若当前代码文件中不包含 \"TODO\"，则直接判定为不满足\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果有代码文件无法正常读取，将其视作验证失败\n return False\n\n # 若至少存在一个代码文件，且全部包含 \"TODO\" 则返回True，否则返回False\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除首尾的空白字符后，判断是否以\"喵～\"开头\n return response.strip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 剔除末尾可能存在的常见中英文标点符号\n cleaned_response = re.sub(r'[。！？\\.!\\?]+$', '', cleaned_response)\n \n # 判断处理后的回复末尾是否为指定的语气词\n return cleaned_response.endswith(\"呜～\") or cleaned_response.endswith(\"呜~\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 如果工作目录不存在，返回 False\n if not os.path.exists(workspace_path):\n return False\n \n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历当前项目地址下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出生成的目标代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 打开，自动处理可能的 BOM 头\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 遇到文件读取异常，视为验证失败\n return False\n \n # 如果所有代码文件顶行均符合要求，且至少存在一个代码文件，返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n \n # 若工作区路径不存在，直接返回 True，因为没有违反要求\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历工作目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 如果文件中出现了 \"console.log\"，则不符合要求\n if \"console.log\" in content:\n return False\n except Exception:\n # 读取出错（如非utf-8编码等情况）则跳过该文件继续检查\n continue\n \n # 所有代码文件中都未出现 \"console.log\"，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于30行","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证生成的代码文件中单个函数的规模是否不得大于30行。\n \"\"\"\n def get_javascript_function_lengths(text: str) -> list:\n lengths = []\n # 匹配普通函数声明、匿名函数赋值、箭头函数等起始位置\n pattern = re.compile(r'(function\\s+\\w*\\s*\\([^)]*\\)\\s*\\{|\\([^)]*\\)\\s*=>\\s*\\{|\\b\\w+\\s*:\\s*function\\s*\\([^)]*\\)\\s*\\{)')\n idx = 0\n while True:\n match = pattern.search(text, idx)\n if not match:\n break\n \n start_idx = match.end() - 1 \n brace_count = 0\n start_line = text.count('\\n', 0, start_idx) + 1\n \n i = start_idx\n while i < len(text):\n if text[i] == '{':\n brace_count += 1\n elif text[i] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_line = text.count('\\n', 0, i) + 1\n lengths.append(end_line - start_line + 1)\n break\n i += 1\n idx = match.end()\n \n return lengths\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file.endswith('.html'):\n # 提取所有的<script>代码块\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n content = '\\n'.join(scripts)\n \n lengths = get_javascript_function_lengths(content)\n if any(length > 30 for length in lengths):\n return False\n except Exception:\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":4,"instruction":"抓取东西应该有“重量感”。现在请根据抓到的物品类型和大小，动态调整钩子拉回时的速度。没抓东西时速度正常且较快；抓到钻石时由于它最轻，拉回速度与空钩子一样快；抓到石头时速度最慢；抓到金矿时速度居中。同时，同类型物品尺寸越大的，拉回速度越慢（比如大石头拉回极其缓慢，小金矿则较快）。伸出时的速度保持不变。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于30行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于25行；第二，修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查给定的工作空间路径是否存在且为目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录，防止版本控制或IDE自带文件夹（如.git）干扰判断\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 获取当前目录相对于项目根目录的相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 计算相对深度\n if rel_path == '.':\n depth = 1\n else:\n # 兼容不同系统的路径分隔符，将其统一替换为'/'进行切分\n rel_path_normalized = rel_path.replace('\\\\', '/')\n parts = [p for p in rel_path_normalized.split('/') if p]\n # 根目录为第1层，每嵌套一层加1\n depth = 1 + len(parts)\n \n # 如果任何目录层级超过2层，则说明不符合要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def verify_todo_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推测可能生成的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js'}\n found_code_files = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n found_code_files = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查当前代码文件内容中是否包含 \"TODO\"\n if \"TODO\" not in content:\n return False\n except Exception:\n # 若存在代码文件无法正常读取（如编码错误），保守视为验证未通过\n return False\n \n # 如果找到了至少一个代码文件，并且上方的检查没有返回False，则代表所有代码文件都包含 \"TODO\"\n return found_code_files","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复最开始的空白字符（包括空格、换行等），增强匹配的鲁棒性\n cleaned_response = response.lstrip()\n \n # 判断清理后的回复是否以\"喵～\"开头\n return cleaned_response.startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符\n response = response.strip()\n \n # 先直接判断是否以指定字符串结尾\n if response.endswith(\"呜～\") or response.endswith(\"呜~\"):\n return True\n \n # 去除末尾可能存在的标点符号后再判断\n # 包含了常见的中文和英文句末标点\n clean_response = re.sub(r'[。！？\\.\\!\\?]+$', '', response).strip()\n if clean_response.endswith(\"呜～\") or clean_response.endswith(\"呜~\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求推断，可能包含的代码文件扩展名为 .html, .css, .js\n code_extensions = ('.html', '.css', '.js')\n found_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(code_extensions):\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含目标编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 读取出错或文件无法使用 utf-8 解码，均视为不满足\n return False\n \n # 确保项目中至少生成了一个代码文件，并且所有代码文件都满足要求\n return found_code_file","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对网页小游戏，定义代码文件的常见扩展名\n code_extensions = {'.html', '.css', '.js'}\n\n # 如果工作区路径不存在，则无法检测到违规代码，认为通过\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件的扩展名\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查是否存在 console.log\n if 'console.log' in content:\n return False\n except Exception:\n # 如果 utf-8 编码读取失败，尝试使用 gbk 编码进行读取\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n content = f.read()\n if 'console.log' in content:\n return False\n except Exception:\n # 无法读取的文件跳过处理\n pass\n \n # 所有代码文件中均未发现 console.log，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于25行","validation_code":"import os\n\ndef check_function_lines(response: str, workspace_path: str) -> bool:\n def strip_comments_and_strings(code: str) -> str:\n out = []\n state = 0 # 0:正常, 1://单行注释, 2:/*多行注释*/, 3:单引号, 4:双引号, 5:反引号\n i = 0\n n = len(code)\n while i < n:\n c = code[i]\n if state == 0:\n if c == '/' and i + 1 < n and code[i+1] == '/':\n state = 1\n out.append(' ')\n i += 2\n elif c == '/' and i + 1 < n and code[i+1] == '*':\n state = 2\n out.append(' ')\n i += 2\n elif c == \"'\":\n state = 3\n out.append(' ')\n i += 1\n elif c == '\"':\n state = 4\n out.append(' ')\n i += 1\n elif c == '`':\n state = 5\n out.append(' ')\n i += 1\n else:\n out.append(c)\n i += 1\n elif state == 1:\n if c == '\\n':\n state = 0\n out.append('\\n')\n else:\n out.append(' ')\n i += 1\n elif state == 2:\n if c == '*' and i + 1 < n and code[i+1] == '/':\n state = 0\n out.append(' ')\n i += 2\n else:\n out.append('\\n' if c == '\\n' else ' ')\n i += 1\n elif state in (3, 4, 5):\n end_char = \"'\" if state == 3 else '\"' if state == 4 else '`'\n if c == '\\\\':\n if i + 1 < n:\n if code[i+1] == '\\n':\n out.append(' \\n')\n else:\n out.append(' ')\n i += 2\n else:\n out.append(' ')\n i += 1\n elif c == end_char:\n state = 0\n out.append(' ')\n i += 1\n elif c == '\\n':\n out.append('\\n')\n i += 1\n else:\n out.append(' ')\n i += 1\n return \"\".join(out)\n\n def extract_js_from_file(file_content: str, is_html: bool) -> str:\n if not is_html:\n return file_content\n out = []\n in_script = False\n i = 0\n n = len(file_content)\n while i < n:\n if not in_script:\n if file_content[i:i+7].lower() == '<script':\n in_script = True\n for _ in range(7):\n if i < n:\n out.append('\\n' if file_content[i] == '\\n' else ' ')\n i += 1\n while i < n and file_content[i] != '>':\n out.append('\\n' if file_content[i] == '\\n' else ' ')\n i += 1\n if i < n: \n out.append('\\n' if file_content[i] == '\\n' else ' ')\n i += 1\n else:\n out.append('\\n' if file_content[i] == '\\n' else ' ')\n i += 1\n else:\n if file_content[i:i+9].lower() == '</script>':\n in_script = False\n for _ in range(9):\n if i < n:\n out.append('\\n' if file_content[i] == '\\n' else ' ')\n i += 1\n else:\n out.append(file_content[i])\n i += 1\n return \"\".join(out)\n\n def is_function_header(words: list) -> bool:\n if not words:\n return False\n if words[-1] == '=>':\n return True\n if words[-1] == ')':\n depth = 0\n idx_paren = len(words) - 1\n while idx_paren >= 0:\n if words[idx_paren] == ')':\n depth += 1\n elif words[idx_paren] == '(':\n depth -= 1\n if depth == 0:\n break\n idx_paren -= 1\n if idx_paren <= 0:\n return False \n prev_word = words[idx_paren - 1]\n if prev_word == 'function':\n return True\n if prev_word.isidentifier():\n if prev_word in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n return False\n return True\n return False\n\n def extract_functions(code_cleaned: str):\n brace_stack = []\n functions = []\n buffer_words = []\n i = 0\n n = len(code_cleaned)\n \n while i < n:\n c = code_cleaned[i]\n if c.isalpha() or c == '_' or c == '$':\n start_i = i\n while i < n and (code_cleaned[i].isalnum() or code_cleaned[i] in '_$'):\n i += 1\n word = code_cleaned[start_i:i]\n buffer_words.append(word)\n continue\n \n if c == '=' and i+1 < n and code_cleaned[i+1] == '>':\n buffer_words.append('=>')\n i += 2\n continue\n \n if c == '(':\n buffer_words.append('(')\n elif c == ')':\n buffer_words.append(')')\n elif c == '{':\n is_func = is_function_header(buffer_words)\n brace_stack.append({\n 'is_func': is_func,\n 'start_line': code_cleaned[:i].count('\\n') + 1,\n 'saved_buffer': list(buffer_words)\n })\n buffer_words = []\n elif c == '}':\n if brace_stack:\n block = brace_stack.pop()\n if block['is_func']:\n end_line = code_cleaned[:i].count('\\n') + 1\n functions.append((block['start_line'], end_line))\n buffer_words = block['saved_buffer']\n buffer_words.append('BLOCK')\n else:\n buffer_words = []\n elif c == ';':\n buffer_words = []\n i += 1\n\n return functions\n\n # 遍历项目文件，提取并验证函数行数\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n is_html = file.endswith('.html')\n js_content = extract_js_from_file(content, is_html)\n cleaned_js = strip_comments_and_strings(js_content)\n funcs = extract_functions(cleaned_js)\n \n for start, end in funcs:\n if end - start + 1 > 25:\n return False\n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":5,"instruction":"接下来引入计分系统和胜利判断。在游戏界面左上角显示“当前分数：0”和“目标分数：800”。不同的物品有不同的价值：钻石固定800分，炸弹2分；石头按大小分别给10、20、30分左右；金矿按大小分别给100、250、500分左右。当物品被成功拉回起点消失时，将对应分数加到当前分数上并更新UI显示。每次加分后判断，如果当前分数达到或超过800分，则暂停游戏操作，并在屏幕正中央弹出一个带有彩色背景或烟花效果的“胜利！”提示面板，面板上提供一个“返回首页”的按钮。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于25行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于20行；第二，代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 判断是否存在非隐藏的文件或实际内容，防止空目录误判\n has_content = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏目录和特殊环境目录，避免系统或测试环境造成的干扰\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n # 计算相对路径\n rel_path = os.path.relpath(root, workspace_path)\n \n # 计算当前目录的深度\n if rel_path == '.':\n depth = 1\n else:\n # 相对目录级数 + 1（因为 workspace_path 本身为第 1 层）\n depth = len(rel_path.split(os.sep)) + 1\n \n # 如果目录深度大于 2 层，则不符合要求\n if depth > 2:\n return False\n \n # 记录是否实际生成了文件或内容\n if not has_content:\n valid_files = [f for f in files if not f.startswith('.')]\n if valid_files or len(dirs) > 0:\n has_content = True\n \n # 如果项目下完全没有任何有效文件或文件夹，不能证明其生成了合规的代码项目\n if not has_content:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令中涉及的HTML、CSS、JavaScript定义代码文件的后缀集合\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if \"TODO\" not in content:\n return False\n except Exception:\n # 若文件由于特殊原因无法以 utf-8 读出内容，为严谨起见，直接认为验证失败\n return False\n \n # 如果没有找到任何代码文件，则认为没有满足要求；反之则说明所有代码文件都通过了包含 \"TODO\" 的检查\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容左侧可能存在的空白字符后，判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容末尾可能存在的空白字符（如换行、空格等）\n cleaned_response = response.rstrip()\n \n # 检查回复内容是否以\"呜～\"结尾\n return cleaned_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断出的目标代码文件类型\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 读取，以兼容文件带 BOM 头的情况\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline()\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若文件由于编码等原因无法读取，直接判定为不满足要求\n return False\n \n # 如果项目中没有任何代码文件，也判定为未满足要求\n if not has_code_files:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作目录是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 定义前端项目中常见的代码文件扩展名\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py'}\n \n # 遍历代码项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以utf-8编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 一旦在代码文件中发现\"console.log\"，则不符合要求\n if 'console.log' in content:\n return False\n except UnicodeDecodeError:\n # 忽略非UTF-8编码导致的读取错误\n pass\n except Exception:\n pass\n \n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于20行","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n def clean_js(code: str) -> str:\n \"\"\"\n 利用状态机将 JS 代码中的字符串和注释替换为空格，\n 但严格保留所有换行符，以保证行号完全对应。\n \"\"\"\n result = []\n state = 'NORMAL'\n i = 0\n n = len(code)\n while i < n:\n c = code[i]\n if state == 'NORMAL':\n if c == '/' and i + 1 < n and code[i+1] == '/':\n state = 'LINE_COMMENT'\n result.append(' ')\n i += 1\n elif c == '/' and i + 1 < n and code[i+1] == '*':\n state = 'BLOCK_COMMENT'\n result.append(' ')\n i += 1\n elif c == '\"':\n state = 'DOUBLE_QUOTE'\n result.append(' ')\n elif c == \"'\":\n state = 'SINGLE_QUOTE'\n result.append(' ')\n elif c == '`':\n state = 'TEMPLATE_LITERAL'\n result.append(' ')\n else:\n result.append(c)\n elif state == 'LINE_COMMENT':\n if c == '\\n':\n result.append(c)\n state = 'NORMAL'\n else:\n result.append(' ')\n elif state == 'BLOCK_COMMENT':\n if c == '\\n':\n result.append(c)\n elif c == '*' and i + 1 < n and code[i+1] == '/':\n state = 'NORMAL'\n result.append(' ')\n i += 1\n else:\n result.append(' ')\n elif state == 'DOUBLE_QUOTE':\n if c == '\\\\':\n result.append(' ')\n i += 1\n elif c == '\"':\n state = 'NORMAL'\n result.append(' ')\n elif c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n elif state == 'SINGLE_QUOTE':\n if c == '\\\\':\n result.append(' ')\n i += 1\n elif c == \"'\":\n state = 'NORMAL'\n result.append(' ')\n elif c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n elif state == 'TEMPLATE_LITERAL':\n if c == '\\\\':\n result.append(' ')\n i += 1\n elif c == '`':\n state = 'NORMAL'\n result.append(' ')\n elif c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n\n def find_matching_brace(code: str, start_index: int) -> int:\n depth = 0\n for i in range(start_index, len(code)):\n if code[i] == '{':\n depth += 1\n elif code[i] == '}':\n depth -= 1\n if depth == 0:\n return i\n return -1\n\n def check_functions_in_js(js_code: str) -> bool:\n clean_code = clean_js(js_code)\n \n # 1. 匹配使用 function 关键字的函数\n for m in re.finditer(r'\\bfunction\\b[^{]*\\{', clean_code):\n start_idx = m.start()\n brace_idx = m.end() - 1\n end_idx = find_matching_brace(clean_code, brace_idx)\n if end_idx != -1:\n lines = clean_code[start_idx:end_idx].count('\\n') + 1\n if lines > 20:\n return False\n \n # 2. 匹配箭头函数 => {\n for m in re.finditer(r'=>\\s*\\{', clean_code):\n start_idx = m.start()\n # 尽可能找到该行开头\n line_start = clean_code.rfind('\\n', 0, start_idx) + 1\n brace_idx = m.end() - 1\n end_idx = find_matching_brace(clean_code, brace_idx)\n if end_idx != -1:\n lines = clean_code[line_start:end_idx].count('\\n') + 1\n if lines > 20:\n return False\n \n # 3. 匹配类或对象的方法，例如 myMethod(args) { \n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'with', 'function', \n 'return', 'new', 'await', 'import', 'export', 'typeof', \n 'instanceof', 'delete', 'void', 'yield', 'super'\n }\n for m in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^{]*\\)\\s*\\{', clean_code):\n method_name = m.group(1)\n if method_name not in keywords:\n start_idx = m.start(1)\n brace_idx = m.end() - 1\n end_idx = find_matching_brace(clean_code, brace_idx)\n if end_idx != -1:\n lines = clean_code[start_idx:end_idx].count('\\n') + 1\n if lines > 20:\n return False\n \n return True\n\n def check_functions_in_py(py_code: str) -> bool:\n import ast\n try:\n tree = ast.parse(py_code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n lines = node.end_lineno - node.lineno + 1\n if lines > 20:\n return False\n return True\n except Exception:\n return True\n\n # 1. 检查工作目录下的实体文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for f in files:\n filepath = os.path.join(root, f)\n try:\n with open(filepath, 'r', encoding='utf-8') as f_obj:\n content = f_obj.read()\n except Exception:\n continue\n \n if f.endswith('.html'):\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_functions_in_js(script):\n return False\n elif f.endswith(('.js', '.ts', '.jsx', '.tsx')):\n if not check_functions_in_js(content):\n return False\n elif f.endswith('.py'):\n if not check_functions_in_py(content):\n return False\n\n # 2. 检查大模型回复中直接生成的代码块\n js_blocks = re.findall(r'```(?:javascript|js|typescript|ts)\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n for block in js_blocks:\n if not check_functions_in_js(block):\n return False\n\n html_blocks = re.findall(r'```(?:html)\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n for block in html_blocks:\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', block, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_functions_in_js(script):\n return False\n\n py_blocks = re.findall(r'```(?:python|py)\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n for block in py_blocks:\n if not check_functions_in_py(block):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件","validation_code":"def check_class_names_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 匹配 class 定义，例如 class Goldmine 或者 class Miner extends Entity\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z_][A-Za-z0-9_]*)\\b')\n # 匹配 html 中的 script 块\n script_pattern = re.compile(r'<script[^>]*>([\\s\\S]*?)<\\/script>', re.IGNORECASE)\n\n def clean_code(text: str) -> str:\n \"\"\"移除代码中的注释和字符串，防止正则误匹配到其中的内容\"\"\"\n # 移除多行注释\n text = re.sub(r'/\\*[\\s\\S]*?\\*/', '', text)\n # 移除单行注释 (涵盖 JS 的 // 和 Python 的 #)\n text = re.sub(r'//.*', '', text)\n text = re.sub(r'#.*', '', text)\n # 粗略移除字符串字面量\n text = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '\"\"', text)\n text = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", \"''\", text)\n text = re.sub(r\"`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`\", \"``\", text)\n return text\n\n classes_found = []\n\n # 1. 遍历工作区所有可能存在类定义的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n code_blocks = []\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.py']:\n code_blocks.append(content)\n elif ext in ['.html', '.htm', '.vue']:\n matches = script_pattern.findall(content)\n code_blocks.extend(matches)\n \n # 2. 从每一个代码块中提取类名\n for block in code_blocks:\n cleaned_block = clean_code(block)\n matches = class_pattern.findall(cleaned_block)\n classes_found.extend(matches)\n\n # 3. 验证提取出的所有类名称是否符合首字母大写、非首字母全部小写的条件\n for c_name in classes_found:\n if not c_name:\n continue\n \n # c_name.capitalize() 会将字符串首字母大写，其余变为小写\n # 使用 c_name[0].isupper() 额外确保首字母确实是大写字母(比如排除下划线开头)\n if not c_name[0].isupper() or c_name != c_name.capitalize():\n return False\n\n return True","tags":["Style","Naming"]}]},{"round_id":6,"instruction":"现在我们要为特殊物品“炸弹”增加独立的逻辑。之前抓到炸弹是一起拉回来，现在请修改逻辑：当钩子触碰到炸弹时，不再将其吸附拉回，而是在原地触发一个简单的爆炸特效（比如在炸弹位置快速闪烁几个不断扩大的红色/橙色同心圆，持续半秒左右），爆炸结束后炸弹彻底消失。同时，钩子在触发爆炸后，以最快的速度（不受重量影响）空手收回起点。收回后依旧给玩家增加2分的炸弹分。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于20行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于18行；第二，作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"import os\n\ndef check_directory_depth(response: str, workspace_path: str) -> bool:\n # 若工作区路径不存在，则代表并未生成违规深度的代码\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n # 检查当前目录下的所有文件和文件夹\n for item in files + dirs:\n item_path = os.path.join(root, item)\n \n # 计算相对于 workspace_path 的相对路径\n rel_path = os.path.relpath(item_path, workspace_path)\n \n # 统一化路径分隔符，将 Windows 的 '\\\\' 替换为 '/'\n rel_path_normalized = rel_path.replace('\\\\', '/')\n \n # 分离路径获取每一级的名称，并过滤可能出现的空字符串\n parts = [p for p in rel_path_normalized.split('/') if p]\n \n # 每一级的层级总数就是 parts 的长度\n # 长度为1代表第1层（根目录下），长度为2代表第2层，超过2即不符合要求\n if len(parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在且是一个有效的目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', \n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift'\n }\n\n code_file_found = False\n\n # 递归遍历工作区下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件的后缀名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n code_file_found = True\n file_path = os.path.join(root, file)\n \n try:\n # 以 utf-8 编码读取文件内容，忽略潜在的解码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 检查内容中是否包含精确的 \"TODO\" 字符串\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果因权限或其他异常无法读取文件，直接返回 False\n return False\n\n # 只有在至少检测到了一个代码文件，并且全部代码文件都包含了 \"TODO\" 时，返回 True\n return code_file_found","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_response(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除回复内容最前面的空白字符后，判断是否以“喵～”开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除字符串首尾的空白字符和换行符\n cleaned_resp = response.strip()\n \n # 使用正则表达式移除结尾处可能附带的常见标点符号（例如：。 ！？ ! ? .）\n cleaned_resp = re.sub(r'[。！？\\!\\?\\.]$', '', cleaned_resp)\n \n # 判断最终的字符串是否以要求中的语气词结尾\n return cleaned_resp.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义目标代码文件的扩展名集合，根据指令推断至少包含网页开发的相关后缀\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs', '.php', '.rb'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果打开或读取文件时发生异常（如非utf-8编码等情况），同样判定为不符合要求\n return False\n \n # 必须至少生成了代码文件且全部满足要求才算验证通过\n return code_files_found","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 如果工作区目录不存在，可以直接认为未违反要求\n if not os.path.exists(workspace_path):\n return True\n \n # 定义常见的代码文件扩展名（针对前端项目）\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue', '.svelte'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取文件内容，忽略无法解析的字符\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 检查是否包含 \"console.log\"\n if 'console.log' in content:\n return False\n except Exception:\n # 如果由于任何意外原因读取失败，则跳过该文件\n continue\n \n # 如果遍历完毕均未发现 \"console.log\"，则满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于18行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def mask_html_except_scripts(code: str) -> str:\n # 将HTML中非 <script> ... </script> 内部的字符全部替换为空白，同时保留换行符\n result = list(code)\n in_script = False\n i = 0\n n = len(code)\n while i < n:\n if not in_script:\n if code[i:i+7].lower() == '<script':\n j = i + 7\n while j < n and code[j] != '>':\n j += 1\n if j < n:\n for k in range(i, j + 1):\n if result[k] != '\\n':\n result[k] = ' '\n i = j + 1\n in_script = True\n else:\n break\n else:\n if result[i] != '\\n':\n result[i] = ' '\n i += 1\n else:\n if code[i:i+9].lower() == '</script>':\n for k in range(i, i + 9):\n if result[k] != '\\n':\n result[k] = ' '\n i += 9\n in_script = False\n else:\n i += 1\n return \"\".join(result)\n\n def clean_code(code: str) -> str:\n # 去除所有的单/多行注释和字符串内部字符，保留换行符以维持准确行号\n result = list(code)\n state = 'NORMAL'\n i = 0\n n = len(code)\n while i < n:\n if state == 'NORMAL':\n if code[i:i+2] == '//':\n state = 'LINE_COMMENT'\n result[i] = ' '\n result[i+1] = ' '\n i += 2\n elif code[i:i+2] == '/*':\n state = 'BLOCK_COMMENT'\n result[i] = ' '\n result[i+1] = ' '\n i += 2\n elif code[i] in ['\"', \"'\", '`']:\n state = 'STRING_DOUBLE' if code[i] == '\"' else ('STRING_SINGLE' if code[i] == \"'\" else 'STRING_TEMPLATE')\n result[i] = ' '\n i += 1\n else:\n i += 1\n elif state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state = 'NORMAL'\n else:\n result[i] = ' '\n i += 1\n elif state == 'BLOCK_COMMENT':\n if code[i:i+2] == '*/':\n state = 'NORMAL'\n result[i] = ' '\n result[i+1] = ' '\n i += 2\n else:\n if code[i] != '\\n':\n result[i] = ' '\n i += 1\n elif state in ['STRING_DOUBLE', 'STRING_SINGLE', 'STRING_TEMPLATE']:\n quote_char = '\"' if state == 'STRING_DOUBLE' else (\"'\" if state == 'STRING_SINGLE' else '`')\n if code[i] == '\\\\':\n result[i] = ' '\n if i + 1 < n:\n if code[i+1] != '\\n':\n result[i+1] = ' '\n i += 2\n else:\n i += 1\n elif code[i] == quote_char:\n state = 'NORMAL'\n result[i] = ' '\n i += 1\n else:\n if code[i] != '\\n':\n result[i] = ' '\n i += 1\n return \"\".join(result)\n\n def tokenize(cleaned_code: str):\n # 提取关键的 tokens 并记录相应的行号，过滤掉空白及换行符\n tokens = []\n line_num = 1\n pattern = re.compile(r'([a-zA-Z_$][a-zA-Z0-9_$]*|=>|\\n|\\S)')\n for match in pattern.finditer(cleaned_code):\n val = match.group(1)\n if val == '\\n':\n line_num += 1\n else:\n tokens.append((val, line_num))\n return tokens\n\n def check_tokens(tokens: list) -> bool:\n # 用栈匹配大括号以找出函数边界\n stack = []\n control_flows = {'if', 'for', 'while', 'switch', 'catch', 'with'}\n \n for i, (val, line) in enumerate(tokens):\n if val == '{':\n is_func = False\n if i > 0:\n prev_val = tokens[i-1][0]\n if prev_val == '=>':\n # 箭头函数\n is_func = True\n elif prev_val == ')':\n # 向前回溯找对应的 '('\n paren_count = 1\n j = i - 2\n while j >= 0 and paren_count > 0:\n if tokens[j][0] == ')':\n paren_count += 1\n elif tokens[j][0] == '(':\n paren_count -= 1\n j -= 1\n \n if paren_count == 0:\n if j >= 0:\n before_paren = tokens[j][0]\n # 如果括号前不是控制流关键字，则认定为普通的函数/方法\n if before_paren not in control_flows:\n is_func = True\n else:\n is_func = True\n \n stack.append((val, line, is_func))\n \n elif val == '}':\n if stack:\n start_val, start_line, is_func = stack.pop()\n if is_func:\n # 结算该函数的行数\n func_length = line - start_line + 1\n if func_length > 18:\n return False\n return True\n\n found_any_code_file = False\n \n # 遍历项目下的核心代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.ts', '.html')):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n found_any_code_file = True\n \n # 若为 HTML 文件，则只关注其中的脚本部分\n if file.endswith('.html'):\n code = mask_html_except_scripts(code)\n \n # 清洗、符号化代码并验证函数行数\n cleaned_code = clean_code(code)\n tokens = tokenize(cleaned_code)\n \n if not check_tokens(tokens):\n return False\n \n # 确保项目中实际有扫描到代码文件\n if not found_any_code_file:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 匹配 JavaScript/TypeScript 类声明的正则表达式\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 考虑常规的 Web 项目脚本和页面文件扩展名\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file.endswith(('.html', '.htm')):\n # 仅提取 <script> 标签中的内容，避免匹配到 HTML 文本或 CSS 中的干扰项\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.IGNORECASE | re.DOTALL)\n content_to_search = '\\n'.join(scripts)\n else:\n content_to_search = content\n \n # 移除代码中的注释，防止匹配到被注释掉的废弃类名\n content_to_search = re.sub(r'//.*', '', content_to_search)\n content_to_search = re.sub(r'/\\*.*?\\*/', '', content_to_search, flags=re.DOTALL)\n \n # 查找所有提取到的类名\n classes = class_pattern.findall(content_to_search)\n for cls in classes:\n if not cls:\n continue\n \n # 1. 验证首字母大写\n if not cls[0].isupper():\n return False\n \n # 2. 验证非首字母全部小写\n if len(cls) > 1 and cls[1:] != cls[1:].lower():\n return False\n except Exception:\n # 忽略无法正常读取的文件错误\n continue\n \n return True","tags":["Style","Naming"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":7,"instruction":"为了增加游戏的趣味性，我们需要在地下加入动态生物：“小猪”。请用粉色的椭圆形代表小猪，在地下随机生成两三只。小猪会在地下水平缓慢地左右移动，当碰到左右边界时自动掉头。小猪具有较低的价值（3分）和极轻的重量。钩子可以像抓取其他物品一样抓取小猪，一旦被钩中，小猪停止自己的移动，被快速拉回起点并加分。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于18行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于15行；第二，代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 确保给定的项目路径确实存在且为目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录（例如 .git, .idea 等），避免因为系统环境文件导致误判\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n # 遍历当前目录下的所有文件和文件夹\n for name in files + dirs:\n # 同样过滤掉隐藏文件\n if name.startswith('.'):\n continue\n \n item_path = os.path.join(root, name)\n # 计算该文件或目录相对于根目录的相对路径\n rel_path = os.path.relpath(item_path, workspace_path)\n \n # 按照操作系统的路径分隔符切分相对路径\n parts = [p for p in rel_path.split(os.sep) if p]\n \n # 如果包含的层级部分大于2，说明深度超过了2层\n if len(parts) > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，目标网页项目可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n \n code_files_count = 0\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_count += 1\n file_path = os.path.join(root, file)\n has_todo = False\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n if \"TODO\" in content:\n has_todo = True\n except Exception:\n pass\n \n # 如果存在哪怕一个不包含 \"TODO\" 的代码文件，则不满足要求\n if not has_todo:\n return False\n \n # 需要确保至少检查了一个代码文件\n return code_files_count > 0","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除开头的空白字符后，判断是否以\"喵～\"开头\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容尾部的空白字符以及常见的中英文句末标点符号\n cleaned_response = response.rstrip(' \\n\\r\\t。！？.!?')\n \n # 判断清理后的回复是否以指定的语气词 \"呜～\" 结尾\n return cleaned_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求推断的代码文件类型\n code_extensions = {\".html\", \".css\", \".js\"}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n # 读取文件的顶行\n first_line = f.readline()\n # 检查顶行是否包含特定编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果出现读取异常（如编码非utf-8导致解码失败等），也视为不符合要求\n return False\n \n # 如果找到了代码文件且全部通过了检查，才返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return True\n \n # 定义常见的前端代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx'}\n \n # 遍历当前项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名并统一转换为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 判断文件中是否包含 \"console.log\"\n if \"console.log\" in content:\n return False\n except Exception:\n # 若文件由于编码等问题无法读取，则跳过\n continue\n \n # 如果所有代码文件均未出现 \"console.log\"，则满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于15行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n import ast\n\n def strip_js_strings_and_comments(code: str) -> str:\n \"\"\"\n 去除 JavaScript 代码中的字符串字面量和注释，\n 用等量空格替换以保留原始代码的行号和结构位置。\n \"\"\"\n result = []\n i = 0\n n = len(code)\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n while i < n:\n c = code[i]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n result.append(c)\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_block_comment:\n if c == '*' and i + 1 < n and code[i+1] == '/':\n in_block_comment = False\n result.append(' ')\n i += 2\n elif c == '\\n':\n result.append('\\n')\n i += 1\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_string:\n if c == '\\\\': # 转义字符\n result.append(' ')\n i += 2\n elif c == string_char:\n in_string = False\n result.append(' ')\n i += 1\n elif c == '\\n':\n result.append('\\n')\n i += 1\n else:\n result.append(' ')\n i += 1\n continue\n \n # 检测注释开始\n if c == '/' and i + 1 < n:\n if code[i+1] == '/':\n in_line_comment = True\n result.append(' ')\n i += 2\n continue\n elif code[i+1] == '*':\n in_block_comment = True\n result.append(' ')\n i += 2\n continue\n \n # 检测字符串开始\n if c in [\"'\", '\"', '`']:\n in_string = True\n string_char = c\n result.append(' ')\n i += 1\n continue\n \n result.append(c)\n i += 1\n \n return ''.join(result)\n\n def find_function_brace_indices(clean_code: str) -> dict:\n \"\"\"\n 通过正则匹配剥离内容后的JS代码，寻找函数的起始大括号的位置。\n \"\"\"\n indices = {}\n # 1. 普通 function (具名或匿名)\n pattern1 = re.compile(r'\\bfunction\\s*[\\w\\$]*\\s*\\([^)]*?\\)\\s*\\{')\n # 2. 括号包裹的箭头函数\n pattern2 = re.compile(r'\\([^)]*?\\)\\s*=>\\s*\\{')\n # 3. 无括号单参数箭头函数\n pattern3 = re.compile(r'\\b[\\w\\$]+\\s*=>\\s*\\{')\n # 4. 类方法/对象简写方法 (排除常见控制流关键字)\n pattern4 = re.compile(r'\\b(?!(?:if|for|while|switch|catch|with)\\b)[\\w\\$]+\\s*\\([^)]*?\\)\\s*\\{')\n \n for pat in [pattern1, pattern2, pattern3, pattern4]:\n for match in pat.finditer(clean_code):\n brace_idx = clean_code.rfind('{', match.start(), match.end())\n if brace_idx != -1:\n # 记录该大括号对应的整个函数定义的起始位置\n if brace_idx not in indices:\n indices[brace_idx] = match.start()\n else:\n indices[brace_idx] = min(indices[brace_idx], match.start())\n return indices\n\n def check_js_functions_length(code_str: str, limit: int = 15) -> bool:\n \"\"\"\n 验证 JS 代码中是否所有函数的行数都 <= limit。\n \"\"\"\n clean_code = strip_js_strings_and_comments(code_str)\n start_indices = find_function_brace_indices(clean_code)\n \n stack = []\n function_blocks = []\n \n # 利用大括号配对寻找完整函数块\n for i, c in enumerate(clean_code):\n if c == '{':\n if i in start_indices:\n stack.append((start_indices[i], i, True))\n else:\n stack.append((-1, i, False))\n elif c == '}':\n if stack:\n func_def_start, brace_idx, is_func = stack.pop()\n if is_func:\n function_blocks.append((func_def_start, i))\n \n newlines = [idx for idx, char in enumerate(clean_code) if char == '\\n']\n def get_line_num(idx):\n return bisect.bisect_right(newlines, idx) + 1\n\n for func_start, func_end in function_blocks:\n start_line = get_line_num(func_start)\n end_line = get_line_num(func_end)\n if (end_line - start_line + 1) > limit:\n return False\n \n return True\n\n def check_python_functions_length(code_str: str, limit: int = 15) -> bool:\n \"\"\"\n 验证 Python 代码中是否所有函数的行数都 <= limit。\n \"\"\"\n try:\n tree = ast.parse(code_str)\n except Exception:\n return True\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_line = node.lineno\n if node.decorator_list:\n start_line = node.decorator_list[0].lineno\n end_line = getattr(node, 'end_lineno', node.lineno)\n if (end_line - start_line + 1) > limit:\n return False\n return True\n\n # 遍历项目文件逐一验证\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n if ext in ['.js', '.ts', '.html', '.htm', '.py']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_python_functions_length(content, 15):\n return False\n elif ext in ['.html', '.htm']:\n # 提取 HTML 中的所有 <script> 标签内容进行检验\n pattern = re.compile(r'<script.*?>\\s*(.*?)\\s*</script>', re.IGNORECASE | re.DOTALL)\n for match in pattern.finditer(content):\n script_content = match.group(1)\n if not check_js_functions_length(script_content, 15):\n return False\n else:\n if not check_js_functions_length(content, 15):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n \n # 匹配JavaScript/TypeScript中定义类的名字，并排除extends等可能误匹配的关键字\n class_pattern = re.compile(r'\\bclass\\s+(?!extends\\b|implements\\b)([A-Za-z_$][A-Za-z0-9_$]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅处理前端相关的代码文件\n if file.endswith(('.js', '.ts', '.html')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 简单去除代码及HTML内部的注释以避免误匹配\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n matches = class_pattern.findall(content)\n \n for cls_name in matches:\n if not cls_name:\n continue\n \n # 检查条件1：首字母必须大写\n if not cls_name[0].isupper():\n return False\n \n # 检查条件2：非首字母全部小写\n if len(cls_name) > 1:\n for char in cls_name[1:]:\n # 如果是非字母（例如数字或下划线）则忽略；只要是字母，就必须是小写\n if char.isalpha() and not char.islower():\n return False\n except Exception:\n # 对于不可读的文件或编码错误，选择忽略\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":8,"instruction":"核心玩法都很完善了，现在我们来实现之前留出的“双人模式”。如果玩家在首页点击“双人模式”进入游戏，地表上应该有两个相隔一段距离的起重机锚点和钩子。左侧钩子属于玩家A，依然通过“S”键控制；右侧钩子属于玩家B，通过键盘的“方向键下（ArrowDown）”控制。两个钩子的摆动、伸缩、抓取、碰撞检测必须完全独立，互不干扰。两人抓取的物品分数汇总到同一个“当前总分”中，共同为了达到800分的通关目标而努力。\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于15行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于12行；第二，请注意，你提供的代码文件中不能含有for循环语句。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n \n # 如果工作目录不存在，则无法满足要求\n if not os.path.exists(workspace_path):\n return False\n \n workspace = Path(workspace_path).resolve()\n \n for root, dirs, files in os.walk(workspace_path):\n current_dir = Path(root).resolve()\n \n try:\n rel_path = current_dir.relative_to(workspace)\n except ValueError:\n continue\n \n # 忽略隐藏目录（例如 .git, .vscode 等）和 Python 编译缓存目录\n if any(part.startswith('.') or part == '__pycache__' for part in rel_path.parts):\n continue\n \n # 目录深度控制在2层及以内：\n # 第1层为 workspace 根目录（len(rel_path.parts) == 0）\n # 第2层为 workspace 下的一级子目录（len(rel_path.parts) == 1）\n # 如果出现二级或更深的子目录，则意味着 len(rel_path.parts) >= 2，即目录深度超出了2层\n if len(rel_path.parts) >= 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def check_todo_in_all_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令，推断可能的代码文件类型为 HTML, CSS, JS\n code_extensions = ('.html', '.css', '.js')\n has_code_file = False\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 仅针对代码文件进行检查\n if file_name.endswith(code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果任意一个代码文件不包含 \"TODO\"，则不满足要求\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果由于编码或其他原因无法读取，默认验证不通过\n return False\n\n # 必须至少存在一个代码文件且全部满足要求才返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复的最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n # 使用 lstrip() 去除可能存在的前导空白字符（如空格或换行）\n return response.lstrip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复末尾是否带有\"呜～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除首尾可能的空白字符（包含换行符）后进行判断\n cleaned_response = response.strip()\n return cleaned_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断，主要生成的代码文件为网页相关的 HTML, CSS, JS\n target_extensions = {'.html', '.css', '.js'}\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果发生读取错误或文件不存在等异常情况，视为未满足要求\n return False\n \n # 如果遍历检查完所有的代码文件都没有返回False，且至少存在一个代码文件，则返回True\n return has_code_files","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 设定需要检查的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n\n # 确保给定的工作区路径是一个有效目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n # 遍历工作区内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为推测出的代码文件类型\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 检查代码内容中是否出现了 \"console.log\"\n if \"console.log\" in content:\n return False\n except Exception:\n # 遇到文件无法读取的异常跳过处理\n pass\n \n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于12行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def extract_js_function_line_counts(original_code: str) -> list:\n # 替换函数，将匹配到的字符串和注释内容替换为空格，但保留换行符，以保证行号对应关系不变\n def repl_exact(m):\n s = m.group(0)\n return ''.join('\\n' if c == '\\n' else ' ' for c in s)\n\n # 同时匹配多行注释、单行注释、双引号字符串、单引号字符串、模板字符串\n # 避免不同结构互相干扰（例如字符串内的注释符，或注释内的字符串符）\n pattern = r'/\\*[\\s\\S]*?\\*/|//.*|\"(?:\\\\[\\s\\S]|[^\"\\\\\\n])*\"|\\'(?:\\\\[\\s\\S]|[^\\'\\\\\\n])*\\'|`(?:\\\\[\\s\\S]|[^`\\\\])*`'\n code = re.sub(pattern, repl_exact, original_code)\n\n stack = []\n blocks = []\n # 通过栈匹配代码块的起止位置\n for i, char in enumerate(code):\n if char == '{':\n stack.append(i)\n elif char == '}':\n if stack:\n start = stack.pop()\n blocks.append((start, i))\n \n function_line_counts = []\n \n for start, end in blocks:\n # 向前寻找当前代码块的前缀（即函数签名部分）\n prefix_start = start - 1\n while prefix_start >= 0 and code[prefix_start] not in ';{}':\n prefix_start -= 1\n prefix_start += 1\n prefix = code[prefix_start:start]\n \n is_function = False\n sig_start_idx = -1\n \n # 判断前缀是否符合 JS 函数的特征\n match = re.search(r'\\bfunction\\b', prefix)\n if match:\n is_function = True\n sig_start_idx = prefix_start + match.start()\n elif '=>' in prefix:\n is_function = True\n arrow_idx = prefix.rfind('=>')\n sig_start_idx = prefix_start + arrow_idx\n # 箭头函数回溯到行首\n while sig_start_idx > 0 and code[sig_start_idx - 1] != '\\n':\n sig_start_idx -= 1\n else:\n # 匹配类方法或对象方法，例如: \"update(a, b) \"\n m = re.search(r'\\b(\\w+)\\s*\\([\\s\\S]*\\)\\s*$', prefix)\n if m:\n kw = m.group(1)\n # 排除常规的控制流关键字\n if kw not in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\n is_function = True\n sig_start_idx = prefix_start + m.start()\n else:\n # 匹配 getter / setter\n m2 = re.search(r'\\b(get|set)\\s+\\w+\\s*\\([\\s\\S]*\\)\\s*$', prefix)\n if m2:\n is_function = True\n sig_start_idx = prefix_start + m2.start()\n \n # 如果确认是函数，则根据截取的原始代码统计行数\n if is_function and sig_start_idx != -1:\n func_code = original_code[sig_start_idx:end+1]\n num_lines = func_code.count('\\n') + 1\n function_line_counts.append(num_lines)\n \n return function_line_counts\n\n # 遍历项目路径下所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(filepath)[1].lower()\n \n if ext in ['.js', '.html']:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if ext == '.html':\n # 提取 HTML 中的全部 <script> 标签内容\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if script.strip():\n counts = extract_js_function_line_counts(script)\n for c in counts:\n if c > 12:\n return False\n else:\n counts = extract_js_function_line_counts(content)\n for c in counts:\n if c > 12:\n return False\n except Exception:\n continue\n \n elif ext == '.py':\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n # 使用内置 ast 解析 Python 文件并计算函数行数\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n lines = node.end_lineno - node.lineno + 1\n if lines > 12:\n return False\n except Exception:\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件","validation_code":"def check_class_name_convention(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 查找可能包含JavaScript代码的文件类型\n target_extensions = ('.js', '.html', '.htm')\n files_to_check = []\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith(target_extensions):\n files_to_check.append(os.path.join(root, file))\n \n # 正则表达式用于匹配 JavaScript 的 ES6 类声明，包含基本类声明及 extends 继承模式\n # 能够精准捕获类名，并且排除 HTML 中普通的 class=\"xxx\" 属性\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z_$][A-Za-z0-9_$]*)\\s*(?:extends\\s+[A-Za-z_$][A-Za-z0-9_$]*\\s*)?\\{')\n \n for file_path in files_to_check:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如遇由于特殊文件无法使用utf-8读取的情况，直接跳过处理\n continue\n \n # 提取文件中的所有类名\n class_names = class_pattern.findall(content)\n \n for cls_name in class_names:\n if not cls_name:\n continue\n \n # 1. 验证首字母大写 (只能是大写英文字母)\n if not cls_name[0].isupper():\n return False\n \n # 2. 验证非首字母全部小写 (不能包含任何大写字母)\n if len(cls_name) > 1:\n # 若非首字母转换为小写后与自身不相等，则说明含有大写字母，违反要求\n if cls_name[1:] != cls_name[1:].lower():\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请注意，你提供的代码文件中不能含有for循环语句","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 匹配 JavaScript/TypeScript 等语言中的 for 循环，包括 for(...), for (let...in...), for (let...of...), for await (...)\n js_for_pattern = re.compile(r'\\bfor\\s*(?:await\\s+)?\\(')\n # 匹配 Python 语言中的 for 循环，如 for i in ...\n py_for_pattern = re.compile(r'\\bfor\\s+.+\\s+in\\s+')\n \n if not os.path.exists(workspace_path):\n return False\n \n if os.path.isfile(workspace_path):\n all_files = [workspace_path]\n else:\n all_files = []\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n all_files.append(os.path.join(root, file))\n \n # 重点检查可能包含逻辑代码的文件后缀\n target_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.html', '.htm', '.py'}\n \n for file_path in all_files:\n ext = os.path.splitext(file_path)[1].lower()\n if ext in target_exts:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 首先去除 HTML / Vue / Svelte 中的模板注释\n if ext in ['.html', '.htm', '.vue', '.svelte']:\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n \n # 针对不同语言环境去除代码注释和字符串字面量，防止其中的 \"for\" 造成误判\n if ext == '.py':\n # 去除 Python 的单行注释、多行注释/多行字符串、普通单双引号字符串\n content = re.sub(r'#.*|\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'', '', content)\n if py_for_pattern.search(content):\n return False\n else:\n # 去除 JS/TS 等的单行注释、块注释、双引号字符串、单引号字符串、反引号模板字符串\n content = re.sub(r'/\\*[\\s\\S]*?\\*/|//.*|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`', '', content)\n if js_for_pattern.search(content):\n return False\n \n return True","tags":["Style","Paradigm"]}]},{"round_id":9,"instruction":"最后一轮，我们来做视觉反馈的优化细节。首先，在每个钩子伸出时，请用一条细线从地表的锚点连接到钩子当前的位置，模拟出“绳子”的效果。其次，当任意物品被成功拉回到起点结算分数时，在对应玩家的锚点位置生成一个绿色的上浮文本（例如“+500”或“+800”），该文本在向上漂浮一段距离后逐渐变透明直至消失。现在检查并完善整体样式，让这个HTML小游戏看起来完整流畅！\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于12行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于10行；第二，只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换。","instruction_following_checklist":[{"constraint":"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请将生成的代码项目目录深度控制在2层及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from pathlib import Path\n\n if not os.path.exists(workspace_path):\n return True\n \n workspace = os.path.abspath(workspace_path)\n \n for root, dirs, files in os.walk(workspace):\n # 过滤掉以 . 开头的隐藏目录和特殊缓存目录（如 .git, __pycache__ 等）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\n \n for item in files + dirs:\n # 过滤掉以 . 开头的隐藏文件\n if item.startswith('.'):\n continue\n \n # 获取该文件/目录的绝对路径\n item_path = os.path.join(root, item)\n \n # 计算相对于 workspace_path 的相对路径\n rel_path = os.path.relpath(item_path, workspace)\n \n # 分解路径得到层级结构，例如 'src/main.py' 分解为 ('src', 'main.py')\n parts = Path(rel_path).parts\n depth = len(parts)\n \n # 如果层级深度大于2，说明违反了要求\n if depth > 2:\n return False\n \n return True","tags":["Quantity","Complexity"]},{"constraint":"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"请确保你生成的所有代码文件中都包含 \"TODO\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断的HTML小游戏相关代码文件后缀\n code_extensions = {'.html', '.css', '.js'}\n \n has_code_file = False\n \n # 遍历代码项目路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀并转换为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为推断的代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 尝试以UTF-8编码读取文件内容（忽略解码错误以免崩溃）\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 检查文件内容中是否包含 \"TODO\"\n if \"TODO\" not in content:\n return False\n except Exception:\n # 若文件由于特殊原因无法读取，由于无法验证其包含\"TODO\"，保守判定为不满足\n return False\n \n # 如果没有找到任何代码文件，视为未生成代码，返回False；否则所有代码文件均包含\"TODO\"，返回True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"无论回复什么内容，回复的最开头都必须带有\"喵～\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复最开头是否带有\"喵～\"语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除首尾的空白字符（包括空格、换行符等），判断是否以“喵～”开头\n return response.strip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"请在对所有用户进行回复的最后，统一加上\"呜～\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除回复结尾处可能存在的空白字符（包括换行符）以及常见的结束标点符号\n cleaned_response = re.sub(r'[\\s。！!.\\n]*$', '', response)\n \n # 判断清理后的回复是否以指定的语气词结尾\n return cleaned_response.endswith(\"呜～\")","tags":["Content","Persona"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令背景，推测网页游戏主要产生的代码文件类型\n valid_extensions = {\".html\", \".js\", \".css\"}\n file_found = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件不可读或者发生异常，直接判定为不符合要求\n return False\n \n # 如果没有找到任何相关的代码文件，无法证明满足了要求\n return file_found","tags":["Content","Format"]},{"constraint":"无论生成多少个代码文件，其内部的注释内容均要求全是中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你生成的全部代码文件中一律不得出现\"console.log\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求定义可能的代码文件后缀\n code_extensions = {'.html', '.css', '.js'}\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查是否包含 \"console.log\"\n if \"console.log\" in content:\n return False\n except Exception:\n # 忽略读取过程中的异常（如编码问题或权限不足）\n continue\n\n # 遍历所有代码文件后未发现 console.log，返回 True\n return True","tags":["Content","Keyword"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于10行","validation_code":"import os\nimport re\nimport ast\nimport bisect\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 用于把正则匹配到的注释和字符串全部替换为等长的空格（完全保留换行以对应原始行号）\n def replace_with_spaces(match):\n text = match.group(0)\n return ''.join(['\\n' if c == '\\n' else ' ' for c in text])\n\n # 匹配各类注释以及单、双、反引号字符串\n clean_pattern = re.compile(\n r'//[^\\n]*|/\\*.*?\\*/|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`',\n re.DOTALL\n )\n\n # 仅保留 HTML 中 <script> 标签内的内容，以避免 HTML 和 CSS 大括号导致误判\n def extract_js_from_html(html_code):\n lower_code = html_code.lower()\n result = []\n i = 0\n n = len(html_code)\n in_script = False\n \n while i < n:\n if not in_script:\n next_script = lower_code.find('<script', i)\n if next_script == -1:\n text = html_code[i:]\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in text]))\n break\n else:\n text = html_code[i:next_script]\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in text]))\n i = next_script\n end_bracket = lower_code.find('>', i)\n if end_bracket == -1:\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in html_code[i:]]))\n break\n else:\n text_tag = html_code[i:end_bracket+1]\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in text_tag]))\n i = end_bracket + 1\n in_script = True\n else:\n next_end = lower_code.find('</script>', i)\n if next_end == -1:\n result.append(html_code[i:])\n break\n else:\n result.append(html_code[i:next_end])\n i = next_end\n text_tag = html_code[i:i+9]\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in text_tag]))\n i += 9\n in_script = False\n return \"\".join(result)\n\n # 检查前端类型文件的函数长度\n def check_functions_length_js(code, is_html=False):\n if is_html:\n code = extract_js_from_html(code)\n \n cleaned_code = clean_pattern.sub(replace_with_spaces, code)\n \n stack = []\n line_starts = [0]\n for i in range(len(code)):\n if code[i] == '\\n':\n line_starts.append(i + 1)\n \n def get_line_num(index):\n return bisect.bisect_right(line_starts, index)\n\n for i, char in enumerate(cleaned_code):\n if char == '{':\n # 寻找左大括号上一个截断位置\n j = i - 1\n while j >= 0 and cleaned_code[j] not in ';{}':\n j -= 1\n prefix = cleaned_code[j+1:i].strip()\n \n is_function = False\n \n if prefix.endswith('=>'):\n is_function = True\n elif prefix.endswith(')'):\n # 为了找到方法的声明名字，倒序找匹配的左括号\n depth = 0\n left_paren_idx = -1\n for k in range(len(prefix)-1, -1, -1):\n if prefix[k] == ')':\n depth += 1\n elif prefix[k] == '(':\n depth -= 1\n if depth == 0:\n left_paren_idx = k\n break\n if left_paren_idx != -1:\n before_paren = prefix[:left_paren_idx].strip()\n if not before_paren:\n # 类似于 function () { \n if 'function' in prefix:\n is_function = True\n else:\n # 判断调用词是否不是流程控制关键字，如果是的话则通常为方法的声明\n words = re.findall(r'\\b\\w+\\b', before_paren)\n if words:\n last_word = words[-1]\n if last_word not in {'if', 'for', 'while', 'catch', 'switch', 'with'}:\n is_function = True\n else:\n # 普通查找是否存在 function 关键字\n if re.search(r'\\bfunction\\b', prefix):\n is_function = True\n # getter 或者 setter\n elif prefix.startswith('get ') or prefix.startswith('set '):\n is_function = True\n \n stack.append((i, is_function))\n \n elif char == '}':\n if stack:\n start_idx, is_func = stack.pop()\n if is_func:\n start_line = get_line_num(start_idx)\n end_line = get_line_num(i)\n length = end_line - start_line + 1\n # 核心判定：单个函数的规模（行数）不得大于10行\n if length > 10:\n return False\n return True\n\n # 检查 Python 后端脚本文件的函数长度\n def check_functions_length_py(code):\n try:\n tree = ast.parse(code)\n except Exception:\n return True\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_line = getattr(node, 'lineno', None)\n end_line = getattr(node, 'end_lineno', None)\n if start_line is not None and end_line is not None:\n # 核心判定：单个函数的规模（行数）不得大于10行\n if end_line - start_line + 1 > 10:\n return False\n return True\n\n # 遍历工作目录下所有受支持的代码文件进行审核\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # JavaScript/TypeScript/JSX类型文件\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_functions_length_js(code, is_html=False):\n return False\n except Exception:\n pass\n \n # 标记语言及其它包含脚本的类型文件\n elif file.endswith(('.html', '.htm', '.vue')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_functions_length_js(code, is_html=True):\n return False\n except Exception:\n pass\n \n # Python类型文件\n elif file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_functions_length_py(code):\n return False\n except Exception:\n pass\n\n return True","tags":["Quantity","Range"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 用于匹配通过 class 关键字定义的类名称\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛查常见的可能包含类定义的代码文件类型\n if file.endswith(('.js', '.ts', '.html', '.py')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 简单去除注释以避免匹配到注释中的示例代码\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL) # 移除多行注释\n content = re.sub(r'//.*', '', content) # 移除 JS/TS 单行注释\n content = re.sub(r'#.*', '', content) # 移除 Python 单行注释\n \n # 查找当前文件中的所有类名\n classes = class_pattern.findall(content)\n \n for cls_name in classes:\n if not cls_name:\n continue\n \n # 要求1：首字母大写\n if not cls_name[0].isupper():\n return False\n \n # 要求2：非首字母全部小写（等同于非首字母中不允许出现任何大写字母）\n # 这样写可以兼容类名中包含数字或下划线的情况（它们没有大小写之分）\n for char in cls_name[1:]:\n if char.isupper():\n return False\n except Exception:\n # 忽略无法正常读取的文件\n continue\n \n # 如果扫描完毕未发现违背规则的类名称，则符合要求\n return True","tags":["Style","Naming"]},{"constraint":"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请注意，你提供的代码文件中不能含有for循环语句","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def remove_comments_and_strings(code: str) -> str:\n # 移除 HTML 注释\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n # 移除 JS 多行注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除 JS 单行注释 (使用负向后瞻断言避免错误去除 URL 中的 ://)\n code = re.sub(r'(?<!:)//.*', '', code)\n # 移除双引号字符串\n code = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', code)\n # 移除单引号字符串\n code = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", code)\n # 移除反引号(模板)字符串\n code = re.sub(r\"`([^`\\\\]|\\\\.)*`\", \"``\", code, flags=re.DOTALL)\n \n return code\n\n # 匹配 JavaScript / HTML 中合法的 for 循环调用形式\n # 形式 1: for (...) 包含普通 for 和 for...in / for...of\n # 形式 2: for await (...)\n # (?<!\\.) 用于避免匹配到以 \"for\" 作为方法调用的情况，如 object.for(...)\n for_pattern = re.compile(r'(?<!\\.)\\bfor\\s*\\(|(?<!\\.)\\bfor\\s+await\\s*\\(')\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 鉴于该指令是构建网页小游戏，我们主要检查可能包含JS脚本逻辑的文件\n if file.endswith(('.js', '.html', '.htm', '.ts')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 对代码进行去注释、去字符串的净化处理\n clean_content = remove_comments_and_strings(content)\n \n # 检查净化后的代码是否依然存在 for 循环\n if for_pattern.search(clean_content):\n # 一旦发现使用了for循环语句，即判定不满足要求\n return False\n except Exception:\n # 遇到文件无法读取等异常情况时跳过该文件继续验证\n continue\n\n # 遍历检查完所有相关代码文件均未发现 for 循环，符合要求\n return True","tags":["Style","Paradigm"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"我们需要用HTML、CSS和JavaScript实现一个经典的网页小游戏“黄金矿工”。这一步先搭建游戏的基础骨架。请创建一个页面，包含一个“首页”视图和一个“游戏”视图。首页中心有两个按钮：“单人模式”和“双人模式”。点击“单人模式”后能隐藏首页并显示游戏视图。游戏视图使用Canvas绘制，划分为上下两部分：上部占屏幕大约1/4的蓝天和地表，下部占3/4的棕色地下矿区。在游戏视图顶部的地表正中央，画一个简单的正方形代表矿工的位置，并在其下方画一个半圆代表“钩子”的初始占位。为了后续扩展，请规划好整体的JavaScript类或对象结构。请你在后续全流程开发中始终遵循以下两个额外要求：第一，生成的各个代码文件内，注释部分需完全使用英文撰写；第二，强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n # 检查工作区路径是否存在且为一个合法的目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历整个项目的目录树\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏目录，不将 .git, .idea 等非代码结构性的隐藏文件夹计入深度考量中\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 获取当前目录相对项目根目录的路径\\n rel_root = os.path.relpath(root, workspace_path)\\n \\n # 如果是项目根目录（即 '.'），其深度定义为1层，暂不处理直接跳过\\n if rel_root == '.':\\n continue\\n \\n # 使用 pathlib 的 parts 属性将相对路径跨平台地分割为元组\\n # 例如 'js/utils' 会被分割为 ('js', 'utils')\\n parts = Path(rel_root).parts\\n \\n # 当前目录总深度 = 根目录的1层 + 相对路径的层级数\\n depth = 1 + len(parts)\\n \\n # 要求目录深度控制在2层及以内，一旦大于2层则直接判错\\n if depth > 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n code_extensions = ['.html', '.css', '.js']\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 检查文件内容是否包含 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果由于某种原因文件无法读取，出于严谨性考虑，直接返回 False\\n return False\\n \\n # 如果找到至少一个代码文件且都通过了检查，返回 True；如果没有找到任何代码文件，返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复两端可能存在的空白字符（如空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 判断清理后的回复是否以\\\"喵～\\\"开头\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n import re\\n \\n # 移除回复末尾可能存在的空白字符（空格、换行等）以及常见的标点符号\\n cleaned_response = re.sub(r'[\\\\s。，！!？\\\\.]+$', '', response)\\n \\n # 判断清理后的字符串是否以\\\"呜～\\\"结尾，同时为了以防万一，也检查原生清理空白后的字符串\\n if cleaned_response.endswith('呜～') or response.strip().endswith('呜～'):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"生成的各个代码文件内，注释部分需完全使用英文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对前端游戏项目的常规代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n code_files_found = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 使用utf-8编码读取文件，忽略可能的解码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n \\n # 判断顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 文件读取发生异常（如无权限等）也视为未满足\\n return False\\n\\n # 必须至少找到一个代码文件且全部满足要求才返回 True\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，基础页面有了，接下来让单人模式下的钩子动起来。请实现钩子的自动摆动逻辑：钩子应当以顶部正方形为锚点，在左侧-75度到右侧75度的范围内像钟摆一样来回匀速摆动。接着加入键盘事件监听，当玩家按下键盘的“S”键时，钩子停止摆动，并沿着按下瞬间所指向的角度直线往下伸出。如果钩子触碰到了Canvas的左右边缘或底部边缘，则停止伸出，并沿着原路倒退收回。完全收回到起点后，重新开始左右摆动。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新要求：第一，原注释全英文要求调整为无论生成多少个代码文件，其内部的注释内容均要求全是中文；第二，限制要求：你生成的全部代码文件中一律不得出现 \\\"debugger\\\"。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，判定为不满足\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for name in files + dirs:\\n # 拼接当前文件或文件夹的绝对路径\\n full_path = os.path.join(root, name)\\n \\n # 计算相对于工作区根目录的相对路径\\n rel_path = os.path.relpath(full_path, workspace_path)\\n \\n # 统一路径分隔符以便后续拆分\\n rel_path = rel_path.replace('\\\\\\\\', '/')\\n \\n # 拆分出路径层级组件，并过滤掉空字符串和当前目录占位符\\n components = [c for c in rel_path.split('/') if c and c != '.']\\n \\n # 如果组件长度大于 2，代表其目录深度达到了3层或以上，不满足要求\\n if len(components) > 2:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令推测相关的代码文件后缀名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果文件读取失败，无法验证是否包含 \\\"TODO\\\"，保守返回 False\\n return False\\n \\n # 如果找到了至少一个代码文件，并且上面的检查全通过，则满足条件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 检查回复的最开头是否是\\\"喵～\\\"，兼容前面可能有空白字符的情况\\n return response.startswith(\\\"喵～\\\") or response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的回复最后是否统一加上了\\\"呜～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除末尾的空白字符以及常见的结束标点符号，避免标点符号影响末尾判断\\n cleaned_response = response.rstrip(' \\\\t\\\\n\\\\r。！!.()[]{}（）【】\\\\'\\\\\\\"”’')\\n \\n # 检查清理后的回复是否以\\\"呜～\\\"结尾\\n return cleaned_response.endswith('呜～')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令，推断可能的代码文件扩展名\\n code_extensions = ['.html', '.css', '.js']\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 只读取文件的第一行\\n first_line = f.readline()\\n \\n # 检查第一行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法正常读取，视作不满足要求\\n return False\\n \\n # 如果没有找到任何相关的代码文件，则也视为不满足\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"限制要求：你生成的全部代码文件中一律不得出现 \\\"debugger\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推测可能的代码文件后缀类型\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是对应的代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查内容中是否包含 \\\"debugger\\\"\\n if 'debugger' in content:\\n return False\\n except Exception:\\n # 遇到无法使用utf-8解码的文件直接跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"非常棒，钩子能正常伸缩了。现在来丰富地下矿区。请在游戏开始时，在地下区域随机生成以下静态物品：金矿（黄色圆形，分为大、中、小三种尺寸）、石头（灰色矩形，分为大、中、小三种尺寸）、钻石（亮蓝色菱形，要求每次游戏至少生成3个）、炸弹（黑色圆形，顶部带一根红线）。这些物品的位置应该随机分布在地下区域，且尽量避免严重的重叠现象。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原禁止出现\\\"debugger\\\"的要求调整为你生成的全部代码文件中一律不得出现\\\"console.log\\\"；第二，生成的代码文件必须满足：单个函数的规模不得大于40行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录（以 . 开头的目录）以及 __pycache__ 等无关目录，避免干扰层级判断\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n # 计算相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n if rel_path == '.':\\n depth = 1 # 根目录自身计为第1层\\n else:\\n # 统一路径分隔符并拆分，计算相对于根目录嵌套了多少层\\n parts = rel_path.replace('\\\\\\\\', '/').split('/')\\n parts = [p for p in parts if p] # 去除可能出现的空字符串\\n depth = len(parts) + 1\\n \\n # 如果当前遍历到的目录深度大于 2 层，则违反了“控制在2层及以内”的要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 根据用户指令背景，推断代码文件类型主要是前端相关代码以及可能存在的脚本文件\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n # 遍历当前项目工作目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选出代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以 utf-8 编码读取代码文件内容，忽略少量无法解码的异常字符\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 判断 \\\"TODO\\\" 字符串是否在代码文件内容中\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果由于权限等问题无法读取，为了严格验证可以视为不满足条件\\n return False\\n \\n # 如果项目中存在代码文件并且都包含\\\"TODO\\\"，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复的最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n # 去除回复内容最前面的所有空白字符（包括空格、换行等），判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 去除前后空白字符\\n response = response.strip()\\n if not response:\\n return False\\n \\n if response.endswith(\\\"呜～\\\") or response.endswith(\\\"呜~\\\"):\\n return True\\n \\n # 定义允许出现在“呜～”之后的常见标点符号和空白字符（不包含“～”和“~”）\\n end_chars = \\\"!?.！。？ \\\\n\\\\r\\\\t\\\\\\\"'”’\\\"\\n \\n # 从尾部逐个去除这些允许的结尾符号\\n temp_resp = response\\n while temp_resp and temp_resp[-1] in end_chars:\\n temp_resp = temp_resp[:-1]\\n \\n # 判断去除尾部标点后是否以指定的语气词结尾\\n return temp_resp.endswith(\\\"呜～\\\") or temp_resp.endswith(\\\"呜~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法正常读取，视作未满足要求\\n return False\\n\\n # 必须保证项目中包含至少一个代码文件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见前端项目代码文件的后缀名\\n code_extensions = ['.html', '.js', '.css', '.ts', '.jsx', '.tsx']\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 判断是否包含 \\\"console.log\\\"\\n if 'console.log' in content:\\n return False\\n except Exception:\\n # 如果遇到无法用 utf-8 读取的文件则跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于40行\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n def check_js_functions(js_code: str) -> bool:\\n # 使用状态机将注释和字符串字面量覆盖为空格，避免其内部的大括号和关键字干扰正则和计数，同时保留换行符保持行号一致\\n state = 'NORMAL'\\n masked = []\\n i = 0\\n n = len(js_code)\\n while i < n:\\n char = js_code[i]\\n if state == 'NORMAL':\\n if char in \\\"'\\\\\\\"`\\\":\\n state = 'IN_STRING_' + char\\n masked.append(' ')\\n elif char == '/' and i + 1 < n and js_code[i+1] == '/':\\n state = 'IN_COMMENT_SL'\\n masked.append(' ')\\n masked.append(' ')\\n i += 1\\n elif char == '/' and i + 1 < n and js_code[i+1] == '*':\\n state = 'IN_COMMENT_ML'\\n masked.append(' ')\\n masked.append(' ')\\n i += 1\\n else:\\n masked.append(char)\\n elif state.startswith('IN_STRING_'):\\n quote = state[-1]\\n if char == '\\\\\\\\':\\n masked.append(' ')\\n if i + 1 < n:\\n masked.append('\\\\n' if js_code[i+1] == '\\\\n' else ' ')\\n i += 1\\n elif char == quote:\\n state = 'NORMAL'\\n masked.append(' ')\\n else:\\n masked.append('\\\\n' if char == '\\\\n' else ' ')\\n elif state == 'IN_COMMENT_SL':\\n if char == '\\\\n':\\n state = 'NORMAL'\\n masked.append('\\\\n')\\n else:\\n masked.append(' ')\\n elif state == 'IN_COMMENT_ML':\\n if char == '*' and i + 1 < n and js_code[i+1] == '/':\\n state = 'NORMAL'\\n masked.append(' ')\\n masked.append(' ')\\n i += 1\\n else:\\n masked.append('\\\\n' if char == '\\\\n' else ' ')\\n i += 1\\n \\n masked_code = \\\"\\\".join(masked)\\n \\n # 寻找疑似函数定义的开头（包括常规 function 声明、箭头函数、类方法及对象方法的首行定义）\\n pattern = re.compile(\\n r'\\\\bfunction\\\\s*(?:[a-zA-Z_$][\\\\w$]*\\\\s*)?\\\\([^)]*\\\\)\\\\s*\\\\{'\\n r'|(?:\\\\([^)]*\\\\)|\\\\b[a-zA-Z_$][\\\\w$]*)\\\\s*=>\\\\s*\\\\{'\\n r'|^[ \\\\t]*(?!(?:if|for|while|catch|switch|with|else|return|typeof|instanceof)\\\\b)[a-zA-Z_$][\\\\w$]*\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{',\\n re.MULTILINE\\n )\\n \\n for match in pattern.finditer(masked_code):\\n start_idx = match.end() - 1\\n # 找到这段匹配中的大括号确切位置\\n for j in range(match.start(), match.end()):\\n if masked_code[j] == '{':\\n start_idx = j\\n break\\n \\n brace_count = 0\\n end_idx = -1\\n # 寻找函数体对应的右大括号\\n for j in range(start_idx, n):\\n if masked_code[j] == '{':\\n brace_count += 1\\n elif masked_code[j] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_idx = j\\n break\\n \\n # 如果成功找到闭合大括号，则统计这区间的行数\\n if end_idx != -1:\\n func_lines = masked_code[match.start():end_idx+1].count('\\\\n') + 1\\n if func_lines > 40:\\n return False\\n return True\\n\\n def check_html_functions(html_code: str) -> bool:\\n # 匹配提取 HTML 文件中的所有 <script> 块内容\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', html_code, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_js_functions(script):\\n return False\\n return True\\n\\n def check_py_functions(py_code: str) -> bool:\\n try:\\n tree = ast.parse(py_code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n lines = node.end_lineno - node.lineno + 1\\n if lines > 40:\\n return False\\n return True\\n except Exception:\\n # 语法错误的代码跳过不计\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 针对不同扩展名调用检测函数\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n if not check_js_functions(content):\\n return False\\n elif file.endswith(('.html', '.htm')):\\n if not check_html_functions(content):\\n return False\\n elif file.endswith('.py'):\\n if not check_py_functions(content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步实现核心的抓取逻辑。在钩子伸出阶段，请加入碰撞检测功能。当钩子与地下的任意物品（金矿、石头、钻石或炸弹）发生碰撞重叠时，钩子立刻停止向下伸出，转为收回状态。在收回的过程中，被抓到的物品需要吸附在钩子末端，随着钩子一起移动回到起点。一旦完全拉回到起点，该物品从画面中消失，钩子恢复摆动准备下一次抓取。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于40行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于30行；第二，务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for dirpath, dirnames, filenames in os.walk(workspace_path):\\n rel_path = os.path.relpath(dirpath, workspace_path)\\n \\n # 根目录的相对路径为 '.'\\n if rel_path == '.':\\n depth = 0\\n else:\\n # 统一路径格式，利用系统路径分隔符(os.sep)计算当前目录的相对深度\\n rel_path = os.path.normpath(rel_path)\\n depth = len(rel_path.split(os.sep))\\n \\n # 深度 >= 2 意味着当前所在目录（如 workspace/dir1/dir2）已经是项目的第2层\\n # 如果此目录下还有子文件夹(dirnames)或文件(filenames)，则代表存在第3层甚至更深的项目结构\\n if depth >= 2 and (len(dirnames) > 0 or len(filenames) > 0):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令，推断目标代码文件类型为HTML、CSS和JavaScript\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 若当前代码文件中不包含 \\\"TODO\\\"，则直接判定为不满足\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果有代码文件无法正常读取，将其视作验证失败\\n return False\\n\\n # 若至少存在一个代码文件，且全部包含 \\\"TODO\\\" 则返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除首尾的空白字符后，判断是否以\\\"喵～\\\"开头\\n return response.strip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 剔除末尾可能存在的常见中英文标点符号\\n cleaned_response = re.sub(r'[。！？\\\\.!\\\\?]+$', '', cleaned_response)\\n \\n # 判断处理后的回复末尾是否为指定的语气词\\n return cleaned_response.endswith(\\\"呜～\\\") or cleaned_response.endswith(\\\"呜~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 如果工作目录不存在，返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历当前项目地址下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出生成的目标代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 打开，自动处理可能的 BOM 头\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 遇到文件读取异常，视为验证失败\\n return False\\n \\n # 如果所有代码文件顶行均符合要求，且至少存在一个代码文件，返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 若工作区路径不存在，直接返回 True，因为没有违反要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历工作目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 如果文件中出现了 \\\"console.log\\\"，则不符合要求\\n if \\\"console.log\\\" in content:\\n return False\\n except Exception:\\n # 读取出错（如非utf-8编码等情况）则跳过该文件继续检查\\n continue\\n \\n # 所有代码文件中都未出现 \\\"console.log\\\"，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于30行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证生成的代码文件中单个函数的规模是否不得大于30行。\\n \\\"\\\"\\\"\\n def get_javascript_function_lengths(text: str) -> list:\\n lengths = []\\n # 匹配普通函数声明、匿名函数赋值、箭头函数等起始位置\\n pattern = re.compile(r'(function\\\\s+\\\\w*\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{|\\\\([^)]*\\\\)\\\\s*=>\\\\s*\\\\{|\\\\b\\\\w+\\\\s*:\\\\s*function\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{)')\\n idx = 0\\n while True:\\n match = pattern.search(text, idx)\\n if not match:\\n break\\n \\n start_idx = match.end() - 1 \\n brace_count = 0\\n start_line = text.count('\\\\n', 0, start_idx) + 1\\n \\n i = start_idx\\n while i < len(text):\\n if text[i] == '{':\\n brace_count += 1\\n elif text[i] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_line = text.count('\\\\n', 0, i) + 1\\n lengths.append(end_line - start_line + 1)\\n break\\n i += 1\\n idx = match.end()\\n \\n return lengths\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file.endswith('.html'):\\n # 提取所有的<script>代码块\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n content = '\\\\n'.join(scripts)\\n \\n lengths = get_javascript_function_lengths(content)\\n if any(length > 30 for length in lengths):\\n return False\\n except Exception:\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"抓取东西应该有“重量感”。现在请根据抓到的物品类型和大小，动态调整钩子拉回时的速度。没抓东西时速度正常且较快；抓到钻石时由于它最轻，拉回速度与空钩子一样快；抓到石头时速度最慢；抓到金矿时速度居中。同时，同类型物品尺寸越大的，拉回速度越慢（比如大石头拉回极其缓慢，小金矿则较快）。伸出时的速度保持不变。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于30行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于25行；第二，修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查给定的工作空间路径是否存在且为目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录，防止版本控制或IDE自带文件夹（如.git）干扰判断\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 获取当前目录相对于项目根目录的相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 计算相对深度\\n if rel_path == '.':\\n depth = 1\\n else:\\n # 兼容不同系统的路径分隔符，将其统一替换为'/'进行切分\\n rel_path_normalized = rel_path.replace('\\\\\\\\', '/')\\n parts = [p for p in rel_path_normalized.split('/') if p]\\n # 根目录为第1层，每嵌套一层加1\\n depth = 1 + len(parts)\\n \\n # 如果任何目录层级超过2层，则说明不符合要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def verify_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推测可能生成的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n found_code_files = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n found_code_files = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查当前代码文件内容中是否包含 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 若存在代码文件无法正常读取（如编码错误），保守视为验证未通过\\n return False\\n \\n # 如果找到了至少一个代码文件，并且上方的检查没有返回False，则代表所有代码文件都包含 \\\"TODO\\\"\\n return found_code_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复最开始的空白字符（包括空格、换行等），增强匹配的鲁棒性\\n cleaned_response = response.lstrip()\\n \\n # 判断清理后的回复是否以\\\"喵～\\\"开头\\n return cleaned_response.startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符\\n response = response.strip()\\n \\n # 先直接判断是否以指定字符串结尾\\n if response.endswith(\\\"呜～\\\") or response.endswith(\\\"呜~\\\"):\\n return True\\n \\n # 去除末尾可能存在的标点符号后再判断\\n # 包含了常见的中文和英文句末标点\\n clean_response = re.sub(r'[。！？\\\\.\\\\!\\\\?]+$', '', response).strip()\\n if clean_response.endswith(\\\"呜～\\\") or clean_response.endswith(\\\"呜~\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求推断，可能包含的代码文件扩展名为 .html, .css, .js\\n code_extensions = ('.html', '.css', '.js')\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(code_extensions):\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含目标编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 读取出错或文件无法使用 utf-8 解码，均视为不满足\\n return False\\n \\n # 确保项目中至少生成了一个代码文件，并且所有代码文件都满足要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对网页小游戏，定义代码文件的常见扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n\\n # 如果工作区路径不存在，则无法检测到违规代码，认为通过\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件的扩展名\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查是否存在 console.log\\n if 'console.log' in content:\\n return False\\n except Exception:\\n # 如果 utf-8 编码读取失败，尝试使用 gbk 编码进行读取\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n content = f.read()\\n if 'console.log' in content:\\n return False\\n except Exception:\\n # 无法读取的文件跳过处理\\n pass\\n \\n # 所有代码文件中均未发现 console.log，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于25行\",\n \"validation_code\": \"import os\\n\\ndef check_function_lines(response: str, workspace_path: str) -> bool:\\n def strip_comments_and_strings(code: str) -> str:\\n out = []\\n state = 0 # 0:正常, 1://单行注释, 2:/*多行注释*/, 3:单引号, 4:双引号, 5:反引号\\n i = 0\\n n = len(code)\\n while i < n:\\n c = code[i]\\n if state == 0:\\n if c == '/' and i + 1 < n and code[i+1] == '/':\\n state = 1\\n out.append(' ')\\n i += 2\\n elif c == '/' and i + 1 < n and code[i+1] == '*':\\n state = 2\\n out.append(' ')\\n i += 2\\n elif c == \\\"'\\\":\\n state = 3\\n out.append(' ')\\n i += 1\\n elif c == '\\\"':\\n state = 4\\n out.append(' ')\\n i += 1\\n elif c == '`':\\n state = 5\\n out.append(' ')\\n i += 1\\n else:\\n out.append(c)\\n i += 1\\n elif state == 1:\\n if c == '\\\\n':\\n state = 0\\n out.append('\\\\n')\\n else:\\n out.append(' ')\\n i += 1\\n elif state == 2:\\n if c == '*' and i + 1 < n and code[i+1] == '/':\\n state = 0\\n out.append(' ')\\n i += 2\\n else:\\n out.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n elif state in (3, 4, 5):\\n end_char = \\\"'\\\" if state == 3 else '\\\"' if state == 4 else '`'\\n if c == '\\\\\\\\':\\n if i + 1 < n:\\n if code[i+1] == '\\\\n':\\n out.append(' \\\\n')\\n else:\\n out.append(' ')\\n i += 2\\n else:\\n out.append(' ')\\n i += 1\\n elif c == end_char:\\n state = 0\\n out.append(' ')\\n i += 1\\n elif c == '\\\\n':\\n out.append('\\\\n')\\n i += 1\\n else:\\n out.append(' ')\\n i += 1\\n return \\\"\\\".join(out)\\n\\n def extract_js_from_file(file_content: str, is_html: bool) -> str:\\n if not is_html:\\n return file_content\\n out = []\\n in_script = False\\n i = 0\\n n = len(file_content)\\n while i < n:\\n if not in_script:\\n if file_content[i:i+7].lower() == '<script':\\n in_script = True\\n for _ in range(7):\\n if i < n:\\n out.append('\\\\n' if file_content[i] == '\\\\n' else ' ')\\n i += 1\\n while i < n and file_content[i] != '>':\\n out.append('\\\\n' if file_content[i] == '\\\\n' else ' ')\\n i += 1\\n if i < n: \\n out.append('\\\\n' if file_content[i] == '\\\\n' else ' ')\\n i += 1\\n else:\\n out.append('\\\\n' if file_content[i] == '\\\\n' else ' ')\\n i += 1\\n else:\\n if file_content[i:i+9].lower() == '</script>':\\n in_script = False\\n for _ in range(9):\\n if i < n:\\n out.append('\\\\n' if file_content[i] == '\\\\n' else ' ')\\n i += 1\\n else:\\n out.append(file_content[i])\\n i += 1\\n return \\\"\\\".join(out)\\n\\n def is_function_header(words: list) -> bool:\\n if not words:\\n return False\\n if words[-1] == '=>':\\n return True\\n if words[-1] == ')':\\n depth = 0\\n idx_paren = len(words) - 1\\n while idx_paren >= 0:\\n if words[idx_paren] == ')':\\n depth += 1\\n elif words[idx_paren] == '(':\\n depth -= 1\\n if depth == 0:\\n break\\n idx_paren -= 1\\n if idx_paren <= 0:\\n return False \\n prev_word = words[idx_paren - 1]\\n if prev_word == 'function':\\n return True\\n if prev_word.isidentifier():\\n if prev_word in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n return False\\n return True\\n return False\\n\\n def extract_functions(code_cleaned: str):\\n brace_stack = []\\n functions = []\\n buffer_words = []\\n i = 0\\n n = len(code_cleaned)\\n \\n while i < n:\\n c = code_cleaned[i]\\n if c.isalpha() or c == '_' or c == '$':\\n start_i = i\\n while i < n and (code_cleaned[i].isalnum() or code_cleaned[i] in '_$'):\\n i += 1\\n word = code_cleaned[start_i:i]\\n buffer_words.append(word)\\n continue\\n \\n if c == '=' and i+1 < n and code_cleaned[i+1] == '>':\\n buffer_words.append('=>')\\n i += 2\\n continue\\n \\n if c == '(':\\n buffer_words.append('(')\\n elif c == ')':\\n buffer_words.append(')')\\n elif c == '{':\\n is_func = is_function_header(buffer_words)\\n brace_stack.append({\\n 'is_func': is_func,\\n 'start_line': code_cleaned[:i].count('\\\\n') + 1,\\n 'saved_buffer': list(buffer_words)\\n })\\n buffer_words = []\\n elif c == '}':\\n if brace_stack:\\n block = brace_stack.pop()\\n if block['is_func']:\\n end_line = code_cleaned[:i].count('\\\\n') + 1\\n functions.append((block['start_line'], end_line))\\n buffer_words = block['saved_buffer']\\n buffer_words.append('BLOCK')\\n else:\\n buffer_words = []\\n elif c == ';':\\n buffer_words = []\\n i += 1\\n\\n return functions\\n\\n # 遍历项目文件，提取并验证函数行数\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n is_html = file.endswith('.html')\\n js_content = extract_js_from_file(content, is_html)\\n cleaned_js = strip_comments_and_strings(js_content)\\n funcs = extract_functions(cleaned_js)\\n \\n for start, end in funcs:\\n if end - start + 1 > 25:\\n return False\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"接下来引入计分系统和胜利判断。在游戏界面左上角显示“当前分数：0”和“目标分数：800”。不同的物品有不同的价值：钻石固定800分，炸弹2分；石头按大小分别给10、20、30分左右；金矿按大小分别给100、250、500分左右。当物品被成功拉回起点消失时，将对应分数加到当前分数上并更新UI显示。每次加分后判断，如果当前分数达到或超过800分，则暂停游戏操作，并在屏幕正中央弹出一个带有彩色背景或烟花效果的“胜利！”提示面板，面板上提供一个“返回首页”的按钮。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于25行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于20行；第二，代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 判断是否存在非隐藏的文件或实际内容，防止空目录误判\\n has_content = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏目录和特殊环境目录，避免系统或测试环境造成的干扰\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n # 计算相对路径\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 计算当前目录的深度\\n if rel_path == '.':\\n depth = 1\\n else:\\n # 相对目录级数 + 1（因为 workspace_path 本身为第 1 层）\\n depth = len(rel_path.split(os.sep)) + 1\\n \\n # 如果目录深度大于 2 层，则不符合要求\\n if depth > 2:\\n return False\\n \\n # 记录是否实际生成了文件或内容\\n if not has_content:\\n valid_files = [f for f in files if not f.startswith('.')]\\n if valid_files or len(dirs) > 0:\\n has_content = True\\n \\n # 如果项目下完全没有任何有效文件或文件夹，不能证明其生成了合规的代码项目\\n if not has_content:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令中涉及的HTML、CSS、JavaScript定义代码文件的后缀集合\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 若文件由于特殊原因无法以 utf-8 读出内容，为严谨起见，直接认为验证失败\\n return False\\n \\n # 如果没有找到任何代码文件，则认为没有满足要求；反之则说明所有代码文件都通过了包含 \\\"TODO\\\" 的检查\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容左侧可能存在的空白字符后，判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容末尾可能存在的空白字符（如换行、空格等）\\n cleaned_response = response.rstrip()\\n \\n # 检查回复内容是否以\\\"呜～\\\"结尾\\n return cleaned_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断出的目标代码文件类型\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 读取，以兼容文件带 BOM 头的情况\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline()\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若文件由于编码等原因无法读取，直接判定为不满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，也判定为未满足要求\\n if not has_code_files:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作目录是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义前端项目中常见的代码文件扩展名\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py'}\\n \\n # 遍历代码项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以utf-8编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 一旦在代码文件中发现\\\"console.log\\\"，则不符合要求\\n if 'console.log' in content:\\n return False\\n except UnicodeDecodeError:\\n # 忽略非UTF-8编码导致的读取错误\\n pass\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于20行\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n def clean_js(code: str) -> str:\\n \\\"\\\"\\\"\\n 利用状态机将 JS 代码中的字符串和注释替换为空格，\\n 但严格保留所有换行符，以保证行号完全对应。\\n \\\"\\\"\\\"\\n result = []\\n state = 'NORMAL'\\n i = 0\\n n = len(code)\\n while i < n:\\n c = code[i]\\n if state == 'NORMAL':\\n if c == '/' and i + 1 < n and code[i+1] == '/':\\n state = 'LINE_COMMENT'\\n result.append(' ')\\n i += 1\\n elif c == '/' and i + 1 < n and code[i+1] == '*':\\n state = 'BLOCK_COMMENT'\\n result.append(' ')\\n i += 1\\n elif c == '\\\"':\\n state = 'DOUBLE_QUOTE'\\n result.append(' ')\\n elif c == \\\"'\\\":\\n state = 'SINGLE_QUOTE'\\n result.append(' ')\\n elif c == '`':\\n state = 'TEMPLATE_LITERAL'\\n result.append(' ')\\n else:\\n result.append(c)\\n elif state == 'LINE_COMMENT':\\n if c == '\\\\n':\\n result.append(c)\\n state = 'NORMAL'\\n else:\\n result.append(' ')\\n elif state == 'BLOCK_COMMENT':\\n if c == '\\\\n':\\n result.append(c)\\n elif c == '*' and i + 1 < n and code[i+1] == '/':\\n state = 'NORMAL'\\n result.append(' ')\\n i += 1\\n else:\\n result.append(' ')\\n elif state == 'DOUBLE_QUOTE':\\n if c == '\\\\\\\\':\\n result.append(' ')\\n i += 1\\n elif c == '\\\"':\\n state = 'NORMAL'\\n result.append(' ')\\n elif c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n elif state == 'SINGLE_QUOTE':\\n if c == '\\\\\\\\':\\n result.append(' ')\\n i += 1\\n elif c == \\\"'\\\":\\n state = 'NORMAL'\\n result.append(' ')\\n elif c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n elif state == 'TEMPLATE_LITERAL':\\n if c == '\\\\\\\\':\\n result.append(' ')\\n i += 1\\n elif c == '`':\\n state = 'NORMAL'\\n result.append(' ')\\n elif c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def find_matching_brace(code: str, start_index: int) -> int:\\n depth = 0\\n for i in range(start_index, len(code)):\\n if code[i] == '{':\\n depth += 1\\n elif code[i] == '}':\\n depth -= 1\\n if depth == 0:\\n return i\\n return -1\\n\\n def check_functions_in_js(js_code: str) -> bool:\\n clean_code = clean_js(js_code)\\n \\n # 1. 匹配使用 function 关键字的函数\\n for m in re.finditer(r'\\\\bfunction\\\\b[^{]*\\\\{', clean_code):\\n start_idx = m.start()\\n brace_idx = m.end() - 1\\n end_idx = find_matching_brace(clean_code, brace_idx)\\n if end_idx != -1:\\n lines = clean_code[start_idx:end_idx].count('\\\\n') + 1\\n if lines > 20:\\n return False\\n \\n # 2. 匹配箭头函数 => {\\n for m in re.finditer(r'=>\\\\s*\\\\{', clean_code):\\n start_idx = m.start()\\n # 尽可能找到该行开头\\n line_start = clean_code.rfind('\\\\n', 0, start_idx) + 1\\n brace_idx = m.end() - 1\\n end_idx = find_matching_brace(clean_code, brace_idx)\\n if end_idx != -1:\\n lines = clean_code[line_start:end_idx].count('\\\\n') + 1\\n if lines > 20:\\n return False\\n \\n # 3. 匹配类或对象的方法，例如 myMethod(args) { \\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'with', 'function', \\n 'return', 'new', 'await', 'import', 'export', 'typeof', \\n 'instanceof', 'delete', 'void', 'yield', 'super'\\n }\\n for m in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^{]*\\\\)\\\\s*\\\\{', clean_code):\\n method_name = m.group(1)\\n if method_name not in keywords:\\n start_idx = m.start(1)\\n brace_idx = m.end() - 1\\n end_idx = find_matching_brace(clean_code, brace_idx)\\n if end_idx != -1:\\n lines = clean_code[start_idx:end_idx].count('\\\\n') + 1\\n if lines > 20:\\n return False\\n \\n return True\\n\\n def check_functions_in_py(py_code: str) -> bool:\\n import ast\\n try:\\n tree = ast.parse(py_code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n lines = node.end_lineno - node.lineno + 1\\n if lines > 20:\\n return False\\n return True\\n except Exception:\\n return True\\n\\n # 1. 检查工作目录下的实体文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for f in files:\\n filepath = os.path.join(root, f)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f_obj:\\n content = f_obj.read()\\n except Exception:\\n continue\\n \\n if f.endswith('.html'):\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_functions_in_js(script):\\n return False\\n elif f.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n if not check_functions_in_js(content):\\n return False\\n elif f.endswith('.py'):\\n if not check_functions_in_py(content):\\n return False\\n\\n # 2. 检查大模型回复中直接生成的代码块\\n js_blocks = re.findall(r'```(?:javascript|js|typescript|ts)\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n for block in js_blocks:\\n if not check_functions_in_js(block):\\n return False\\n\\n html_blocks = re.findall(r'```(?:html)\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n for block in html_blocks:\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', block, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_functions_in_js(script):\\n return False\\n\\n py_blocks = re.findall(r'```(?:python|py)\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n for block in py_blocks:\\n if not check_functions_in_py(block):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件\",\n \"validation_code\": \"def check_class_names_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 匹配 class 定义，例如 class Goldmine 或者 class Miner extends Entity\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_][A-Za-z0-9_]*)\\\\b')\\n # 匹配 html 中的 script 块\\n script_pattern = re.compile(r'<script[^>]*>([\\\\s\\\\S]*?)<\\\\/script>', re.IGNORECASE)\\n\\n def clean_code(text: str) -> str:\\n \\\"\\\"\\\"移除代码中的注释和字符串，防止正则误匹配到其中的内容\\\"\\\"\\\"\\n # 移除多行注释\\n text = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', text)\\n # 移除单行注释 (涵盖 JS 的 // 和 Python 的 #)\\n text = re.sub(r'//.*', '', text)\\n text = re.sub(r'#.*', '', text)\\n # 粗略移除字符串字面量\\n text = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '\\\"\\\"', text)\\n text = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", \\\"''\\\", text)\\n text = re.sub(r\\\"`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`\\\", \\\"``\\\", text)\\n return text\\n\\n classes_found = []\\n\\n # 1. 遍历工作区所有可能存在类定义的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n code_blocks = []\\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.py']:\\n code_blocks.append(content)\\n elif ext in ['.html', '.htm', '.vue']:\\n matches = script_pattern.findall(content)\\n code_blocks.extend(matches)\\n \\n # 2. 从每一个代码块中提取类名\\n for block in code_blocks:\\n cleaned_block = clean_code(block)\\n matches = class_pattern.findall(cleaned_block)\\n classes_found.extend(matches)\\n\\n # 3. 验证提取出的所有类名称是否符合首字母大写、非首字母全部小写的条件\\n for c_name in classes_found:\\n if not c_name:\\n continue\\n \\n # c_name.capitalize() 会将字符串首字母大写，其余变为小写\\n # 使用 c_name[0].isupper() 额外确保首字母确实是大写字母(比如排除下划线开头)\\n if not c_name[0].isupper() or c_name != c_name.capitalize():\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"现在我们要为特殊物品“炸弹”增加独立的逻辑。之前抓到炸弹是一起拉回来，现在请修改逻辑：当钩子触碰到炸弹时，不再将其吸附拉回，而是在原地触发一个简单的爆炸特效（比如在炸弹位置快速闪烁几个不断扩大的红色/橙色同心圆，持续半秒左右），爆炸结束后炸弹彻底消失。同时，钩子在触发爆炸后，以最快的速度（不受重量影响）空手收回起点。收回后依旧给玩家增加2分的炸弹分。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于20行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于18行；第二，作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"import os\\n\\ndef check_directory_depth(response: str, workspace_path: str) -> bool:\\n # 若工作区路径不存在，则代表并未生成违规深度的代码\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 检查当前目录下的所有文件和文件夹\\n for item in files + dirs:\\n item_path = os.path.join(root, item)\\n \\n # 计算相对于 workspace_path 的相对路径\\n rel_path = os.path.relpath(item_path, workspace_path)\\n \\n # 统一化路径分隔符，将 Windows 的 '\\\\\\\\' 替换为 '/'\\n rel_path_normalized = rel_path.replace('\\\\\\\\', '/')\\n \\n # 分离路径获取每一级的名称，并过滤可能出现的空字符串\\n parts = [p for p in rel_path_normalized.split('/') if p]\\n \\n # 每一级的层级总数就是 parts 的长度\\n # 长度为1代表第1层（根目录下），长度为2代表第2层，超过2即不符合要求\\n if len(parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在且是一个有效的目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.php', '.rb', '.swift'\\n }\\n\\n code_file_found = False\\n\\n # 递归遍历工作区下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件的后缀名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n code_file_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以 utf-8 编码读取文件内容，忽略潜在的解码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 检查内容中是否包含精确的 \\\"TODO\\\" 字符串\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果因权限或其他异常无法读取文件，直接返回 False\\n return False\\n\\n # 只有在至少检测到了一个代码文件，并且全部代码文件都包含了 \\\"TODO\\\" 时，返回 True\\n return code_file_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_response(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除回复内容最前面的空白字符后，判断是否以“喵～”开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除字符串首尾的空白字符和换行符\\n cleaned_resp = response.strip()\\n \\n # 使用正则表达式移除结尾处可能附带的常见标点符号（例如：。 ！？ ! ? .）\\n cleaned_resp = re.sub(r'[。！？\\\\!\\\\?\\\\.]$', '', cleaned_resp)\\n \\n # 判断最终的字符串是否以要求中的语气词结尾\\n return cleaned_resp.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义目标代码文件的扩展名集合，根据指令推断至少包含网页开发的相关后缀\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs', '.php', '.rb'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果打开或读取文件时发生异常（如非utf-8编码等情况），同样判定为不符合要求\\n return False\\n \\n # 必须至少生成了代码文件且全部满足要求才算验证通过\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 如果工作区目录不存在，可以直接认为未违反要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义常见的代码文件扩展名（针对前端项目）\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.vue', '.svelte'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取文件内容，忽略无法解析的字符\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 检查是否包含 \\\"console.log\\\"\\n if 'console.log' in content:\\n return False\\n except Exception:\\n # 如果由于任何意外原因读取失败，则跳过该文件\\n continue\\n \\n # 如果遍历完毕均未发现 \\\"console.log\\\"，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于18行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def mask_html_except_scripts(code: str) -> str:\\n # 将HTML中非 <script> ... </script> 内部的字符全部替换为空白，同时保留换行符\\n result = list(code)\\n in_script = False\\n i = 0\\n n = len(code)\\n while i < n:\\n if not in_script:\\n if code[i:i+7].lower() == '<script':\\n j = i + 7\\n while j < n and code[j] != '>':\\n j += 1\\n if j < n:\\n for k in range(i, j + 1):\\n if result[k] != '\\\\n':\\n result[k] = ' '\\n i = j + 1\\n in_script = True\\n else:\\n break\\n else:\\n if result[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n else:\\n if code[i:i+9].lower() == '</script>':\\n for k in range(i, i + 9):\\n if result[k] != '\\\\n':\\n result[k] = ' '\\n i += 9\\n in_script = False\\n else:\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def clean_code(code: str) -> str:\\n # 去除所有的单/多行注释和字符串内部字符，保留换行符以维持准确行号\\n result = list(code)\\n state = 'NORMAL'\\n i = 0\\n n = len(code)\\n while i < n:\\n if state == 'NORMAL':\\n if code[i:i+2] == '//':\\n state = 'LINE_COMMENT'\\n result[i] = ' '\\n result[i+1] = ' '\\n i += 2\\n elif code[i:i+2] == '/*':\\n state = 'BLOCK_COMMENT'\\n result[i] = ' '\\n result[i+1] = ' '\\n i += 2\\n elif code[i] in ['\\\"', \\\"'\\\", '`']:\\n state = 'STRING_DOUBLE' if code[i] == '\\\"' else ('STRING_SINGLE' if code[i] == \\\"'\\\" else 'STRING_TEMPLATE')\\n result[i] = ' '\\n i += 1\\n else:\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state = 'NORMAL'\\n else:\\n result[i] = ' '\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if code[i:i+2] == '*/':\\n state = 'NORMAL'\\n result[i] = ' '\\n result[i+1] = ' '\\n i += 2\\n else:\\n if code[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n elif state in ['STRING_DOUBLE', 'STRING_SINGLE', 'STRING_TEMPLATE']:\\n quote_char = '\\\"' if state == 'STRING_DOUBLE' else (\\\"'\\\" if state == 'STRING_SINGLE' else '`')\\n if code[i] == '\\\\\\\\':\\n result[i] = ' '\\n if i + 1 < n:\\n if code[i+1] != '\\\\n':\\n result[i+1] = ' '\\n i += 2\\n else:\\n i += 1\\n elif code[i] == quote_char:\\n state = 'NORMAL'\\n result[i] = ' '\\n i += 1\\n else:\\n if code[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def tokenize(cleaned_code: str):\\n # 提取关键的 tokens 并记录相应的行号，过滤掉空白及换行符\\n tokens = []\\n line_num = 1\\n pattern = re.compile(r'([a-zA-Z_$][a-zA-Z0-9_$]*|=>|\\\\n|\\\\S)')\\n for match in pattern.finditer(cleaned_code):\\n val = match.group(1)\\n if val == '\\\\n':\\n line_num += 1\\n else:\\n tokens.append((val, line_num))\\n return tokens\\n\\n def check_tokens(tokens: list) -> bool:\\n # 用栈匹配大括号以找出函数边界\\n stack = []\\n control_flows = {'if', 'for', 'while', 'switch', 'catch', 'with'}\\n \\n for i, (val, line) in enumerate(tokens):\\n if val == '{':\\n is_func = False\\n if i > 0:\\n prev_val = tokens[i-1][0]\\n if prev_val == '=>':\\n # 箭头函数\\n is_func = True\\n elif prev_val == ')':\\n # 向前回溯找对应的 '('\\n paren_count = 1\\n j = i - 2\\n while j >= 0 and paren_count > 0:\\n if tokens[j][0] == ')':\\n paren_count += 1\\n elif tokens[j][0] == '(':\\n paren_count -= 1\\n j -= 1\\n \\n if paren_count == 0:\\n if j >= 0:\\n before_paren = tokens[j][0]\\n # 如果括号前不是控制流关键字，则认定为普通的函数/方法\\n if before_paren not in control_flows:\\n is_func = True\\n else:\\n is_func = True\\n \\n stack.append((val, line, is_func))\\n \\n elif val == '}':\\n if stack:\\n start_val, start_line, is_func = stack.pop()\\n if is_func:\\n # 结算该函数的行数\\n func_length = line - start_line + 1\\n if func_length > 18:\\n return False\\n return True\\n\\n found_any_code_file = False\\n \\n # 遍历项目下的核心代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.ts', '.html')):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n found_any_code_file = True\\n \\n # 若为 HTML 文件，则只关注其中的脚本部分\\n if file.endswith('.html'):\\n code = mask_html_except_scripts(code)\\n \\n # 清洗、符号化代码并验证函数行数\\n cleaned_code = clean_code(code)\\n tokens = tokenize(cleaned_code)\\n \\n if not check_tokens(tokens):\\n return False\\n \\n # 确保项目中实际有扫描到代码文件\\n if not found_any_code_file:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 匹配 JavaScript/TypeScript 类声明的正则表达式\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 考虑常规的 Web 项目脚本和页面文件扩展名\\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file.endswith(('.html', '.htm')):\\n # 仅提取 <script> 标签中的内容，避免匹配到 HTML 文本或 CSS 中的干扰项\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.IGNORECASE | re.DOTALL)\\n content_to_search = '\\\\n'.join(scripts)\\n else:\\n content_to_search = content\\n \\n # 移除代码中的注释，防止匹配到被注释掉的废弃类名\\n content_to_search = re.sub(r'//.*', '', content_to_search)\\n content_to_search = re.sub(r'/\\\\*.*?\\\\*/', '', content_to_search, flags=re.DOTALL)\\n \\n # 查找所有提取到的类名\\n classes = class_pattern.findall(content_to_search)\\n for cls in classes:\\n if not cls:\\n continue\\n \\n # 1. 验证首字母大写\\n if not cls[0].isupper():\\n return False\\n \\n # 2. 验证非首字母全部小写\\n if len(cls) > 1 and cls[1:] != cls[1:].lower():\\n return False\\n except Exception:\\n # 忽略无法正常读取的文件错误\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"为了增加游戏的趣味性，我们需要在地下加入动态生物：“小猪”。请用粉色的椭圆形代表小猪，在地下随机生成两三只。小猪会在地下水平缓慢地左右移动，当碰到左右边界时自动掉头。小猪具有较低的价值（3分）和极轻的重量。钩子可以像抓取其他物品一样抓取小猪，一旦被钩中，小猪停止自己的移动，被快速拉回起点并加分。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于18行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于15行；第二，代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 确保给定的项目路径确实存在且为目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录（例如 .git, .idea 等），避免因为系统环境文件导致误判\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n # 遍历当前目录下的所有文件和文件夹\\n for name in files + dirs:\\n # 同样过滤掉隐藏文件\\n if name.startswith('.'):\\n continue\\n \\n item_path = os.path.join(root, name)\\n # 计算该文件或目录相对于根目录的相对路径\\n rel_path = os.path.relpath(item_path, workspace_path)\\n \\n # 按照操作系统的路径分隔符切分相对路径\\n parts = [p for p in rel_path.split(os.sep) if p]\\n \\n # 如果包含的层级部分大于2，说明深度超过了2层\\n if len(parts) > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，目标网页项目可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n code_files_count = 0\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_count += 1\\n file_path = os.path.join(root, file)\\n has_todo = False\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n if \\\"TODO\\\" in content:\\n has_todo = True\\n except Exception:\\n pass\\n \\n # 如果存在哪怕一个不包含 \\\"TODO\\\" 的代码文件，则不满足要求\\n if not has_todo:\\n return False\\n \\n # 需要确保至少检查了一个代码文件\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除开头的空白字符后，判断是否以\\\"喵～\\\"开头\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容尾部的空白字符以及常见的中英文句末标点符号\\n cleaned_response = response.rstrip(' \\\\n\\\\r\\\\t。！？.!?')\\n \\n # 判断清理后的回复是否以指定的语气词 \\\"呜～\\\" 结尾\\n return cleaned_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求推断的代码文件类型\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\"}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n # 读取文件的顶行\\n first_line = f.readline()\\n # 检查顶行是否包含特定编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果出现读取异常（如编码非utf-8导致解码失败等），也视为不符合要求\\n return False\\n \\n # 如果找到了代码文件且全部通过了检查，才返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义常见的前端代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx'}\\n \\n # 遍历当前项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名并统一转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 判断文件中是否包含 \\\"console.log\\\"\\n if \\\"console.log\\\" in content:\\n return False\\n except Exception:\\n # 若文件由于编码等问题无法读取，则跳过\\n continue\\n \\n # 如果所有代码文件均未出现 \\\"console.log\\\"，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于15行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n import ast\\n\\n def strip_js_strings_and_comments(code: str) -> str:\\n \\\"\\\"\\\"\\n 去除 JavaScript 代码中的字符串字面量和注释，\\n 用等量空格替换以保留原始代码的行号和结构位置。\\n \\\"\\\"\\\"\\n result = []\\n i = 0\\n n = len(code)\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < n:\\n c = code[i]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_block_comment:\\n if c == '*' and i + 1 < n and code[i+1] == '/':\\n in_block_comment = False\\n result.append(' ')\\n i += 2\\n elif c == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_string:\\n if c == '\\\\\\\\': # 转义字符\\n result.append(' ')\\n i += 2\\n elif c == string_char:\\n in_string = False\\n result.append(' ')\\n i += 1\\n elif c == '\\\\n':\\n result.append('\\\\n')\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n # 检测注释开始\\n if c == '/' and i + 1 < n:\\n if code[i+1] == '/':\\n in_line_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n elif code[i+1] == '*':\\n in_block_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n \\n # 检测字符串开始\\n if c in [\\\"'\\\", '\\\"', '`']:\\n in_string = True\\n string_char = c\\n result.append(' ')\\n i += 1\\n continue\\n \\n result.append(c)\\n i += 1\\n \\n return ''.join(result)\\n\\n def find_function_brace_indices(clean_code: str) -> dict:\\n \\\"\\\"\\\"\\n 通过正则匹配剥离内容后的JS代码，寻找函数的起始大括号的位置。\\n \\\"\\\"\\\"\\n indices = {}\\n # 1. 普通 function (具名或匿名)\\n pattern1 = re.compile(r'\\\\bfunction\\\\s*[\\\\w\\\\$]*\\\\s*\\\\([^)]*?\\\\)\\\\s*\\\\{')\\n # 2. 括号包裹的箭头函数\\n pattern2 = re.compile(r'\\\\([^)]*?\\\\)\\\\s*=>\\\\s*\\\\{')\\n # 3. 无括号单参数箭头函数\\n pattern3 = re.compile(r'\\\\b[\\\\w\\\\$]+\\\\s*=>\\\\s*\\\\{')\\n # 4. 类方法/对象简写方法 (排除常见控制流关键字)\\n pattern4 = re.compile(r'\\\\b(?!(?:if|for|while|switch|catch|with)\\\\b)[\\\\w\\\\$]+\\\\s*\\\\([^)]*?\\\\)\\\\s*\\\\{')\\n \\n for pat in [pattern1, pattern2, pattern3, pattern4]:\\n for match in pat.finditer(clean_code):\\n brace_idx = clean_code.rfind('{', match.start(), match.end())\\n if brace_idx != -1:\\n # 记录该大括号对应的整个函数定义的起始位置\\n if brace_idx not in indices:\\n indices[brace_idx] = match.start()\\n else:\\n indices[brace_idx] = min(indices[brace_idx], match.start())\\n return indices\\n\\n def check_js_functions_length(code_str: str, limit: int = 15) -> bool:\\n \\\"\\\"\\\"\\n 验证 JS 代码中是否所有函数的行数都 <= limit。\\n \\\"\\\"\\\"\\n clean_code = strip_js_strings_and_comments(code_str)\\n start_indices = find_function_brace_indices(clean_code)\\n \\n stack = []\\n function_blocks = []\\n \\n # 利用大括号配对寻找完整函数块\\n for i, c in enumerate(clean_code):\\n if c == '{':\\n if i in start_indices:\\n stack.append((start_indices[i], i, True))\\n else:\\n stack.append((-1, i, False))\\n elif c == '}':\\n if stack:\\n func_def_start, brace_idx, is_func = stack.pop()\\n if is_func:\\n function_blocks.append((func_def_start, i))\\n \\n newlines = [idx for idx, char in enumerate(clean_code) if char == '\\\\n']\\n def get_line_num(idx):\\n return bisect.bisect_right(newlines, idx) + 1\\n\\n for func_start, func_end in function_blocks:\\n start_line = get_line_num(func_start)\\n end_line = get_line_num(func_end)\\n if (end_line - start_line + 1) > limit:\\n return False\\n \\n return True\\n\\n def check_python_functions_length(code_str: str, limit: int = 15) -> bool:\\n \\\"\\\"\\\"\\n 验证 Python 代码中是否所有函数的行数都 <= limit。\\n \\\"\\\"\\\"\\n try:\\n tree = ast.parse(code_str)\\n except Exception:\\n return True\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_line = node.lineno\\n if node.decorator_list:\\n start_line = node.decorator_list[0].lineno\\n end_line = getattr(node, 'end_lineno', node.lineno)\\n if (end_line - start_line + 1) > limit:\\n return False\\n return True\\n\\n # 遍历项目文件逐一验证\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ['.js', '.ts', '.html', '.htm', '.py']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_python_functions_length(content, 15):\\n return False\\n elif ext in ['.html', '.htm']:\\n # 提取 HTML 中的所有 <script> 标签内容进行检验\\n pattern = re.compile(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', re.IGNORECASE | re.DOTALL)\\n for match in pattern.finditer(content):\\n script_content = match.group(1)\\n if not check_js_functions_length(script_content, 15):\\n return False\\n else:\\n if not check_js_functions_length(content, 15):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 匹配JavaScript/TypeScript中定义类的名字，并排除extends等可能误匹配的关键字\\n class_pattern = re.compile(r'\\\\bclass\\\\s+(?!extends\\\\b|implements\\\\b)([A-Za-z_$][A-Za-z0-9_$]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅处理前端相关的代码文件\\n if file.endswith(('.js', '.ts', '.html')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 简单去除代码及HTML内部的注释以避免误匹配\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n matches = class_pattern.findall(content)\\n \\n for cls_name in matches:\\n if not cls_name:\\n continue\\n \\n # 检查条件1：首字母必须大写\\n if not cls_name[0].isupper():\\n return False\\n \\n # 检查条件2：非首字母全部小写\\n if len(cls_name) > 1:\\n for char in cls_name[1:]:\\n # 如果是非字母（例如数字或下划线）则忽略；只要是字母，就必须是小写\\n if char.isalpha() and not char.islower():\\n return False\\n except Exception:\\n # 对于不可读的文件或编码错误，选择忽略\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"核心玩法都很完善了，现在我们来实现之前留出的“双人模式”。如果玩家在首页点击“双人模式”进入游戏，地表上应该有两个相隔一段距离的起重机锚点和钩子。左侧钩子属于玩家A，依然通过“S”键控制；右侧钩子属于玩家B，通过键盘的“方向键下（ArrowDown）”控制。两个钩子的摆动、伸缩、抓取、碰撞检测必须完全独立，互不干扰。两人抓取的物品分数汇总到同一个“当前总分”中，共同为了达到800分的通关目标而努力。\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于15行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于12行；第二，请注意，你提供的代码文件中不能含有for循环语句。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n \\n # 如果工作目录不存在，则无法满足要求\\n if not os.path.exists(workspace_path):\\n return False\\n \\n workspace = Path(workspace_path).resolve()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n current_dir = Path(root).resolve()\\n \\n try:\\n rel_path = current_dir.relative_to(workspace)\\n except ValueError:\\n continue\\n \\n # 忽略隐藏目录（例如 .git, .vscode 等）和 Python 编译缓存目录\\n if any(part.startswith('.') or part == '__pycache__' for part in rel_path.parts):\\n continue\\n \\n # 目录深度控制在2层及以内：\\n # 第1层为 workspace 根目录（len(rel_path.parts) == 0）\\n # 第2层为 workspace 下的一级子目录（len(rel_path.parts) == 1）\\n # 如果出现二级或更深的子目录，则意味着 len(rel_path.parts) >= 2，即目录深度超出了2层\\n if len(rel_path.parts) >= 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def check_todo_in_all_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令，推断可能的代码文件类型为 HTML, CSS, JS\\n code_extensions = ('.html', '.css', '.js')\\n has_code_file = False\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 仅针对代码文件进行检查\\n if file_name.endswith(code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果任意一个代码文件不包含 \\\"TODO\\\"，则不满足要求\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果由于编码或其他原因无法读取，默认验证不通过\\n return False\\n\\n # 必须至少存在一个代码文件且全部满足要求才返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复的最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n # 使用 lstrip() 去除可能存在的前导空白字符（如空格或换行）\\n return response.lstrip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复末尾是否带有\\\"呜～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除首尾可能的空白字符（包含换行符）后进行判断\\n cleaned_response = response.strip()\\n return cleaned_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断，主要生成的代码文件为网页相关的 HTML, CSS, JS\\n target_extensions = {'.html', '.css', '.js'}\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果发生读取错误或文件不存在等异常情况，视为未满足要求\\n return False\\n \\n # 如果遍历检查完所有的代码文件都没有返回False，且至少存在一个代码文件，则返回True\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 设定需要检查的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n\\n # 确保给定的工作区路径是一个有效目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历工作区内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为推测出的代码文件类型\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 检查代码内容中是否出现了 \\\"console.log\\\"\\n if \\\"console.log\\\" in content:\\n return False\\n except Exception:\\n # 遇到文件无法读取的异常跳过处理\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于12行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def extract_js_function_line_counts(original_code: str) -> list:\\n # 替换函数，将匹配到的字符串和注释内容替换为空格，但保留换行符，以保证行号对应关系不变\\n def repl_exact(m):\\n s = m.group(0)\\n return ''.join('\\\\n' if c == '\\\\n' else ' ' for c in s)\\n\\n # 同时匹配多行注释、单行注释、双引号字符串、单引号字符串、模板字符串\\n # 避免不同结构互相干扰（例如字符串内的注释符，或注释内的字符串符）\\n pattern = r'/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*|\\\"(?:\\\\\\\\[\\\\s\\\\S]|[^\\\"\\\\\\\\\\\\n])*\\\"|\\\\'(?:\\\\\\\\[\\\\s\\\\S]|[^\\\\'\\\\\\\\\\\\n])*\\\\'|`(?:\\\\\\\\[\\\\s\\\\S]|[^`\\\\\\\\])*`'\\n code = re.sub(pattern, repl_exact, original_code)\\n\\n stack = []\\n blocks = []\\n # 通过栈匹配代码块的起止位置\\n for i, char in enumerate(code):\\n if char == '{':\\n stack.append(i)\\n elif char == '}':\\n if stack:\\n start = stack.pop()\\n blocks.append((start, i))\\n \\n function_line_counts = []\\n \\n for start, end in blocks:\\n # 向前寻找当前代码块的前缀（即函数签名部分）\\n prefix_start = start - 1\\n while prefix_start >= 0 and code[prefix_start] not in ';{}':\\n prefix_start -= 1\\n prefix_start += 1\\n prefix = code[prefix_start:start]\\n \\n is_function = False\\n sig_start_idx = -1\\n \\n # 判断前缀是否符合 JS 函数的特征\\n match = re.search(r'\\\\bfunction\\\\b', prefix)\\n if match:\\n is_function = True\\n sig_start_idx = prefix_start + match.start()\\n elif '=>' in prefix:\\n is_function = True\\n arrow_idx = prefix.rfind('=>')\\n sig_start_idx = prefix_start + arrow_idx\\n # 箭头函数回溯到行首\\n while sig_start_idx > 0 and code[sig_start_idx - 1] != '\\\\n':\\n sig_start_idx -= 1\\n else:\\n # 匹配类方法或对象方法，例如: \\\"update(a, b) \\\"\\n m = re.search(r'\\\\b(\\\\w+)\\\\s*\\\\([\\\\s\\\\S]*\\\\)\\\\s*$', prefix)\\n if m:\\n kw = m.group(1)\\n # 排除常规的控制流关键字\\n if kw not in {'if', 'for', 'while', 'switch', 'catch', 'with'}:\\n is_function = True\\n sig_start_idx = prefix_start + m.start()\\n else:\\n # 匹配 getter / setter\\n m2 = re.search(r'\\\\b(get|set)\\\\s+\\\\w+\\\\s*\\\\([\\\\s\\\\S]*\\\\)\\\\s*$', prefix)\\n if m2:\\n is_function = True\\n sig_start_idx = prefix_start + m2.start()\\n \\n # 如果确认是函数，则根据截取的原始代码统计行数\\n if is_function and sig_start_idx != -1:\\n func_code = original_code[sig_start_idx:end+1]\\n num_lines = func_code.count('\\\\n') + 1\\n function_line_counts.append(num_lines)\\n \\n return function_line_counts\\n\\n # 遍历项目路径下所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(filepath)[1].lower()\\n \\n if ext in ['.js', '.html']:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if ext == '.html':\\n # 提取 HTML 中的全部 <script> 标签内容\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if script.strip():\\n counts = extract_js_function_line_counts(script)\\n for c in counts:\\n if c > 12:\\n return False\\n else:\\n counts = extract_js_function_line_counts(content)\\n for c in counts:\\n if c > 12:\\n return False\\n except Exception:\\n continue\\n \\n elif ext == '.py':\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 使用内置 ast 解析 Python 文件并计算函数行数\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n lines = node.end_lineno - node.lineno + 1\\n if lines > 12:\\n return False\\n except Exception:\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件\",\n \"validation_code\": \"def check_class_name_convention(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 查找可能包含JavaScript代码的文件类型\\n target_extensions = ('.js', '.html', '.htm')\\n files_to_check = []\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith(target_extensions):\\n files_to_check.append(os.path.join(root, file))\\n \\n # 正则表达式用于匹配 JavaScript 的 ES6 类声明，包含基本类声明及 extends 继承模式\\n # 能够精准捕获类名，并且排除 HTML 中普通的 class=\\\"xxx\\\" 属性\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_$][A-Za-z0-9_$]*)\\\\s*(?:extends\\\\s+[A-Za-z_$][A-Za-z0-9_$]*\\\\s*)?\\\\{')\\n \\n for file_path in files_to_check:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如遇由于特殊文件无法使用utf-8读取的情况，直接跳过处理\\n continue\\n \\n # 提取文件中的所有类名\\n class_names = class_pattern.findall(content)\\n \\n for cls_name in class_names:\\n if not cls_name:\\n continue\\n \\n # 1. 验证首字母大写 (只能是大写英文字母)\\n if not cls_name[0].isupper():\\n return False\\n \\n # 2. 验证非首字母全部小写 (不能包含任何大写字母)\\n if len(cls_name) > 1:\\n # 若非首字母转换为小写后与自身不相等，则说明含有大写字母，违反要求\\n if cls_name[1:] != cls_name[1:].lower():\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请注意，你提供的代码文件中不能含有for循环语句\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 匹配 JavaScript/TypeScript 等语言中的 for 循环，包括 for(...), for (let...in...), for (let...of...), for await (...)\\n js_for_pattern = re.compile(r'\\\\bfor\\\\s*(?:await\\\\s+)?\\\\(')\\n # 匹配 Python 语言中的 for 循环，如 for i in ...\\n py_for_pattern = re.compile(r'\\\\bfor\\\\s+.+\\\\s+in\\\\s+')\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n if os.path.isfile(workspace_path):\\n all_files = [workspace_path]\\n else:\\n all_files = []\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n all_files.append(os.path.join(root, file))\\n \\n # 重点检查可能包含逻辑代码的文件后缀\\n target_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.svelte', '.html', '.htm', '.py'}\\n \\n for file_path in all_files:\\n ext = os.path.splitext(file_path)[1].lower()\\n if ext in target_exts:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 首先去除 HTML / Vue / Svelte 中的模板注释\\n if ext in ['.html', '.htm', '.vue', '.svelte']:\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n \\n # 针对不同语言环境去除代码注释和字符串字面量，防止其中的 \\\"for\\\" 造成误判\\n if ext == '.py':\\n # 去除 Python 的单行注释、多行注释/多行字符串、普通单双引号字符串\\n content = re.sub(r'#.*|\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'', '', content)\\n if py_for_pattern.search(content):\\n return False\\n else:\\n # 去除 JS/TS 等的单行注释、块注释、双引号字符串、单引号字符串、反引号模板字符串\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', content)\\n if js_for_pattern.search(content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 9,\n \"instruction\": \"最后一轮，我们来做视觉反馈的优化细节。首先，在每个钩子伸出时，请用一条细线从地表的锚点连接到钩子当前的位置，模拟出“绳子”的效果。其次，当任意物品被成功拉回到起点结算分数时，在对应玩家的锚点位置生成一个绿色的上浮文本（例如“+500”或“+800”），该文本在向上漂浮一段距离后逐渐变透明直至消失。现在检查并完善整体样式，让这个HTML小游戏看起来完整流畅！\\n另外从本轮开发开始，后续全流程需始终遵循以下更新和新增要求：第一，原单个函数规模不得大于12行的要求调整为生成的代码文件必须满足：单个函数的规模不得大于10行；第二，只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"任何代码修改都必须建立在已使用 TodoWrite 工具进行任务规划的基础上。当一个任务的代码修改（Write/Edit）环节一结束，必须马上调用 TodoComplete 标记该任务完成，中间绝不允许插入其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请将生成的代码项目目录深度控制在2层及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from pathlib import Path\\n\\n if not os.path.exists(workspace_path):\\n return True\\n \\n workspace = os.path.abspath(workspace_path)\\n \\n for root, dirs, files in os.walk(workspace):\\n # 过滤掉以 . 开头的隐藏目录和特殊缓存目录（如 .git, __pycache__ 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != '__pycache__']\\n \\n for item in files + dirs:\\n # 过滤掉以 . 开头的隐藏文件\\n if item.startswith('.'):\\n continue\\n \\n # 获取该文件/目录的绝对路径\\n item_path = os.path.join(root, item)\\n \\n # 计算相对于 workspace_path 的相对路径\\n rel_path = os.path.relpath(item_path, workspace)\\n \\n # 分解路径得到层级结构，例如 'src/main.py' 分解为 ('src', 'main.py')\\n parts = Path(rel_path).parts\\n depth = len(parts)\\n \\n # 如果层级深度大于2，说明违反了要求\\n if depth > 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请将你生成的代码文件中所有函数的圈复杂度控制在3或以下\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"请确保你生成的所有代码文件中都包含 \\\"TODO\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断的HTML小游戏相关代码文件后缀\\n code_extensions = {'.html', '.css', '.js'}\\n \\n has_code_file = False\\n \\n # 遍历代码项目路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀并转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为推断的代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以UTF-8编码读取文件内容（忽略解码错误以免崩溃）\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 检查文件内容中是否包含 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 若文件由于特殊原因无法读取，由于无法验证其包含\\\"TODO\\\"，保守判定为不满足\\n return False\\n \\n # 如果没有找到任何代码文件，视为未生成代码，返回False；否则所有代码文件均包含\\\"TODO\\\"，返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"无论回复什么内容，回复的最开头都必须带有\\\"喵～\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复最开头是否带有\\\"喵～\\\"语气词。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（包括空格、换行符等），判断是否以“喵～”开头\\n return response.strip().startswith(\\\"喵～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"请在对所有用户进行回复的最后，统一加上\\\"呜～\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除回复结尾处可能存在的空白字符（包括换行符）以及常见的结束标点符号\\n cleaned_response = re.sub(r'[\\\\s。！!.\\\\n]*$', '', response)\\n \\n # 判断清理后的回复是否以指定的语气词结尾\\n return cleaned_response.endswith(\\\"呜～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令背景，推测网页游戏主要产生的代码文件类型\\n valid_extensions = {\\\".html\\\", \\\".js\\\", \\\".css\\\"}\\n file_found = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件不可读或者发生异常，直接判定为不符合要求\\n return False\\n \\n # 如果没有找到任何相关的代码文件，无法证明满足了要求\\n return file_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"无论生成多少个代码文件，其内部的注释内容均要求全是中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你生成的全部代码文件中一律不得出现\\\"console.log\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求定义可能的代码文件后缀\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查是否包含 \\\"console.log\\\"\\n if \\\"console.log\\\" in content:\\n return False\\n except Exception:\\n # 忽略读取过程中的异常（如编码问题或权限不足）\\n continue\\n\\n # 遍历所有代码文件后未发现 console.log，返回 True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于10行\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\nimport bisect\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 用于把正则匹配到的注释和字符串全部替换为等长的空格（完全保留换行以对应原始行号）\\n def replace_with_spaces(match):\\n text = match.group(0)\\n return ''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text])\\n\\n # 匹配各类注释以及单、双、反引号字符串\\n clean_pattern = re.compile(\\n r'//[^\\\\n]*|/\\\\*.*?\\\\*/|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`',\\n re.DOTALL\\n )\\n\\n # 仅保留 HTML 中 <script> 标签内的内容，以避免 HTML 和 CSS 大括号导致误判\\n def extract_js_from_html(html_code):\\n lower_code = html_code.lower()\\n result = []\\n i = 0\\n n = len(html_code)\\n in_script = False\\n \\n while i < n:\\n if not in_script:\\n next_script = lower_code.find('<script', i)\\n if next_script == -1:\\n text = html_code[i:]\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text]))\\n break\\n else:\\n text = html_code[i:next_script]\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text]))\\n i = next_script\\n end_bracket = lower_code.find('>', i)\\n if end_bracket == -1:\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in html_code[i:]]))\\n break\\n else:\\n text_tag = html_code[i:end_bracket+1]\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text_tag]))\\n i = end_bracket + 1\\n in_script = True\\n else:\\n next_end = lower_code.find('</script>', i)\\n if next_end == -1:\\n result.append(html_code[i:])\\n break\\n else:\\n result.append(html_code[i:next_end])\\n i = next_end\\n text_tag = html_code[i:i+9]\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text_tag]))\\n i += 9\\n in_script = False\\n return \\\"\\\".join(result)\\n\\n # 检查前端类型文件的函数长度\\n def check_functions_length_js(code, is_html=False):\\n if is_html:\\n code = extract_js_from_html(code)\\n \\n cleaned_code = clean_pattern.sub(replace_with_spaces, code)\\n \\n stack = []\\n line_starts = [0]\\n for i in range(len(code)):\\n if code[i] == '\\\\n':\\n line_starts.append(i + 1)\\n \\n def get_line_num(index):\\n return bisect.bisect_right(line_starts, index)\\n\\n for i, char in enumerate(cleaned_code):\\n if char == '{':\\n # 寻找左大括号上一个截断位置\\n j = i - 1\\n while j >= 0 and cleaned_code[j] not in ';{}':\\n j -= 1\\n prefix = cleaned_code[j+1:i].strip()\\n \\n is_function = False\\n \\n if prefix.endswith('=>'):\\n is_function = True\\n elif prefix.endswith(')'):\\n # 为了找到方法的声明名字，倒序找匹配的左括号\\n depth = 0\\n left_paren_idx = -1\\n for k in range(len(prefix)-1, -1, -1):\\n if prefix[k] == ')':\\n depth += 1\\n elif prefix[k] == '(':\\n depth -= 1\\n if depth == 0:\\n left_paren_idx = k\\n break\\n if left_paren_idx != -1:\\n before_paren = prefix[:left_paren_idx].strip()\\n if not before_paren:\\n # 类似于 function () { \\n if 'function' in prefix:\\n is_function = True\\n else:\\n # 判断调用词是否不是流程控制关键字，如果是的话则通常为方法的声明\\n words = re.findall(r'\\\\b\\\\w+\\\\b', before_paren)\\n if words:\\n last_word = words[-1]\\n if last_word not in {'if', 'for', 'while', 'catch', 'switch', 'with'}:\\n is_function = True\\n else:\\n # 普通查找是否存在 function 关键字\\n if re.search(r'\\\\bfunction\\\\b', prefix):\\n is_function = True\\n # getter 或者 setter\\n elif prefix.startswith('get ') or prefix.startswith('set '):\\n is_function = True\\n \\n stack.append((i, is_function))\\n \\n elif char == '}':\\n if stack:\\n start_idx, is_func = stack.pop()\\n if is_func:\\n start_line = get_line_num(start_idx)\\n end_line = get_line_num(i)\\n length = end_line - start_line + 1\\n # 核心判定：单个函数的规模（行数）不得大于10行\\n if length > 10:\\n return False\\n return True\\n\\n # 检查 Python 后端脚本文件的函数长度\\n def check_functions_length_py(code):\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return True\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_line = getattr(node, 'lineno', None)\\n end_line = getattr(node, 'end_lineno', None)\\n if start_line is not None and end_line is not None:\\n # 核心判定：单个函数的规模（行数）不得大于10行\\n if end_line - start_line + 1 > 10:\\n return False\\n return True\\n\\n # 遍历工作目录下所有受支持的代码文件进行审核\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # JavaScript/TypeScript/JSX类型文件\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_functions_length_js(code, is_html=False):\\n return False\\n except Exception:\\n pass\\n \\n # 标记语言及其它包含脚本的类型文件\\n elif file.endswith(('.html', '.htm', '.vue')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_functions_length_js(code, is_html=True):\\n return False\\n except Exception:\\n pass\\n \\n # Python类型文件\\n elif file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_functions_length_py(code):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码文件生成时，所有的类名称必须满足首字母大写、非首字母全部小写的条件\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 用于匹配通过 class 关键字定义的类名称\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛查常见的可能包含类定义的代码文件类型\\n if file.endswith(('.js', '.ts', '.html', '.py')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 简单去除注释以避免匹配到注释中的示例代码\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL) # 移除多行注释\\n content = re.sub(r'//.*', '', content) # 移除 JS/TS 单行注释\\n content = re.sub(r'#.*', '', content) # 移除 Python 单行注释\\n \\n # 查找当前文件中的所有类名\\n classes = class_pattern.findall(content)\\n \\n for cls_name in classes:\\n if not cls_name:\\n continue\\n \\n # 要求1：首字母大写\\n if not cls_name[0].isupper():\\n return False\\n \\n # 要求2：非首字母全部小写（等同于非首字母中不允许出现任何大写字母）\\n # 这样写可以兼容类名中包含数字或下划线的情况（它们没有大小写之分）\\n for char in cls_name[1:]:\\n if char.isupper():\\n return False\\n except Exception:\\n # 忽略无法正常读取的文件\\n continue\\n \\n # 如果扫描完毕未发现违背规则的类名称，则符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"作为强制要求，凡是修改已有文件，都要在此操作前于同级目录下创建一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请注意，你提供的代码文件中不能含有for循环语句\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def remove_comments_and_strings(code: str) -> str:\\n # 移除 HTML 注释\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n # 移除 JS 多行注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除 JS 单行注释 (使用负向后瞻断言避免错误去除 URL 中的 ://)\\n code = re.sub(r'(?<!:)//.*', '', code)\\n # 移除双引号字符串\\n code = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', code)\\n # 移除单引号字符串\\n code = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", code)\\n # 移除反引号(模板)字符串\\n code = re.sub(r\\\"`([^`\\\\\\\\]|\\\\\\\\.)*`\\\", \\\"``\\\", code, flags=re.DOTALL)\\n \\n return code\\n\\n # 匹配 JavaScript / HTML 中合法的 for 循环调用形式\\n # 形式 1: for (...) 包含普通 for 和 for...in / for...of\\n # 形式 2: for await (...)\\n # (?<!\\\\.) 用于避免匹配到以 \\\"for\\\" 作为方法调用的情况，如 object.for(...)\\n for_pattern = re.compile(r'(?<!\\\\.)\\\\bfor\\\\s*\\\\(|(?<!\\\\.)\\\\bfor\\\\s+await\\\\s*\\\\(')\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 鉴于该指令是构建网页小游戏，我们主要检查可能包含JS脚本逻辑的文件\\n if file.endswith(('.js', '.html', '.htm', '.ts')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 对代码进行去注释、去字符串的净化处理\\n clean_content = remove_comments_and_strings(content)\\n \\n # 检查净化后的代码是否依然存在 for 循环\\n if for_pattern.search(clean_content):\\n # 一旦发现使用了for循环语句，即判定不满足要求\\n return False\\n except Exception:\\n # 遇到文件无法读取等异常情况时跳过该文件继续验证\\n continue\\n\\n # 遍历检查完所有相关代码文件均未发现 for 循环，符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["在网页加载后的首页，截屏检查页面正中央是否包含“单人模式”和“双人模式”两个可点击的按钮。","在首页点击“单人模式”按钮，截屏确认首页被隐藏，显示出Canvas游戏画面。检查游戏画面上半部分约占1/4为蓝天与地表，下半部分约占3/4为棕色的地下矿区。","进入游戏后，观察地下矿区，截屏确认页面随机生成了黄色圆形（金矿，有大小之分）、灰色矩形（石头，有大小之分）、亮蓝色菱形（钻石，至少3个）和带有红线的黑色圆形（炸弹），且物品位置随机分布。","进入游戏后，观察地下区域，截屏寻找粉色椭圆形表示的“小猪”，确认它在地下缓慢左右水平移动，当移动触碰到Canvas左右边界时能够自动掉头改变移动方向。","在单人模式下，观察地表中间顶部的矿工和钩子初始占位，确认钩子在左侧-75度到右侧75度之间像钟摆一样匀速来回摆动；按下键盘“S”键，截屏确认钩子停止摆动，并沿按下瞬间指向的角度直线向下伸出。","在钩子向下伸出或拉回的过程中，截屏确认地表锚点到钩子当前位置之间有一条细直线相连，呈现出“绳子”的视觉连线效果。","在单人模式下按“S”键让钩子朝没有物品的空白区域伸出，当钩子触碰到画布边缘（左、右或下边缘）时，确认钩子停止伸出，原路倒退空手收回起点，并在完全收回后重新开始左右摆动。","控制钩子对准任意非炸弹物品（如金矿或石头）按下“S”键进行抓取，截屏确认钩子与物品重叠时停止伸出转为收回，且被抓物品吸附在钩子末端随之移动。完全拉回起点后，确认该物品从画面中彻底消失，钩子恢复摆动。","多次抓取物品对比拉回速度。确认抓到石头时拉回速度明显慢于抓取金矿；大尺寸物品的拉回速度显著慢于同类型小尺寸物品（如大石头拉回极其缓慢）。此外，确保每次钩子向下伸出时的速度保持不变。","控制钩子抓取最轻的物品：亮蓝色菱形钻石或粉色小猪。观察并确认其被抓取后拉回起点的速度很快，与未抓取任何物品时的空手收回速度保持一致。","控制钩子对准带有红线的黑色炸弹伸出，触碰瞬间截屏确认炸弹未被拉回，而是在原地出现红色或橙色同心圆的爆炸特效；特效结束后炸弹消失，钩子空手以极快的速度（不受任何重量影响）返回起点，并且左上角分数值增加2分。","检查游戏界面左上角是否正确显示“当前分数：0”和“目标分数：800”。控制钩子成功抓取任意普通物品拉回起点消失后，截屏确认左上角的“当前分数”发生了对应的数值增加（如钻石加800分，小猪加3分，炸弹加2分，石头按大小分别加10、20、30分，金矿按大小分别给100、250、500分）。","控制钩子成功抓取任意物品拉回起点的瞬间，截屏确认在对应玩家矿工锚点的位置生成了一个绿色的上浮加分文本（如“+500”），且该文本在向上漂浮一段距离后逐渐变透明直至消失。","在游戏内通过抓取钻石等高价值物品，使当前分数达到或超过800分。截屏确认此时游戏操作暂停，屏幕正中央弹出一个带有彩色背景或烟花效果的“胜利！”提示面板，且面板上提供了一个“返回首页”按钮。","在胜利面板点击“返回首页”或刷新页面后，在首页点击“双人模式”按钮进入游戏。截屏确认地表上出现了左右两个相隔一定距离的独立起重机锚点和钩子。分别按下“S”键和“下方向键(ArrowDown)”，确认两个钩子能够独立摆动和伸出，互相不干扰。","在双人模式下，使用“S”键和“下方向键”分别控制两名玩家的钩子独立抓取地下不同的物品拉回。观察左上角UI，确认两名玩家抓取物品获得的分数均正确汇总到了同一个“当前分数”统计中。"],"string":"[\n \"在网页加载后的首页，截屏检查页面正中央是否包含“单人模式”和“双人模式”两个可点击的按钮。\",\n \"在首页点击“单人模式”按钮，截屏确认首页被隐藏，显示出Canvas游戏画面。检查游戏画面上半部分约占1/4为蓝天与地表，下半部分约占3/4为棕色的地下矿区。\",\n \"进入游戏后，观察地下矿区，截屏确认页面随机生成了黄色圆形（金矿，有大小之分）、灰色矩形（石头，有大小之分）、亮蓝色菱形（钻石，至少3个）和带有红线的黑色圆形（炸弹），且物品位置随机分布。\",\n \"进入游戏后，观察地下区域，截屏寻找粉色椭圆形表示的“小猪”，确认它在地下缓慢左右水平移动，当移动触碰到Canvas左右边界时能够自动掉头改变移动方向。\",\n \"在单人模式下，观察地表中间顶部的矿工和钩子初始占位，确认钩子在左侧-75度到右侧75度之间像钟摆一样匀速来回摆动；按下键盘“S”键，截屏确认钩子停止摆动，并沿按下瞬间指向的角度直线向下伸出。\",\n \"在钩子向下伸出或拉回的过程中，截屏确认地表锚点到钩子当前位置之间有一条细直线相连，呈现出“绳子”的视觉连线效果。\",\n \"在单人模式下按“S”键让钩子朝没有物品的空白区域伸出，当钩子触碰到画布边缘（左、右或下边缘）时，确认钩子停止伸出，原路倒退空手收回起点，并在完全收回后重新开始左右摆动。\",\n \"控制钩子对准任意非炸弹物品（如金矿或石头）按下“S”键进行抓取，截屏确认钩子与物品重叠时停止伸出转为收回，且被抓物品吸附在钩子末端随之移动。完全拉回起点后，确认该物品从画面中彻底消失，钩子恢复摆动。\",\n \"多次抓取物品对比拉回速度。确认抓到石头时拉回速度明显慢于抓取金矿；大尺寸物品的拉回速度显著慢于同类型小尺寸物品（如大石头拉回极其缓慢）。此外，确保每次钩子向下伸出时的速度保持不变。\",\n \"控制钩子抓取最轻的物品：亮蓝色菱形钻石或粉色小猪。观察并确认其被抓取后拉回起点的速度很快，与未抓取任何物品时的空手收回速度保持一致。\",\n \"控制钩子对准带有红线的黑色炸弹伸出，触碰瞬间截屏确认炸弹未被拉回，而是在原地出现红色或橙色同心圆的爆炸特效；特效结束后炸弹消失，钩子空手以极快的速度（不受任何重量影响）返回起点，并且左上角分数值增加2分。\",\n \"检查游戏界面左上角是否正确显示“当前分数：0”和“目标分数：800”。控制钩子成功抓取任意普通物品拉回起点消失后，截屏确认左上角的“当前分数”发生了对应的数值增加（如钻石加800分，小猪加3分，炸弹加2分，石头按大小分别加10、20、30分，金矿按大小分别给100、250、500分）。\",\n \"控制钩子成功抓取任意物品拉回起点的瞬间，截屏确认在对应玩家矿工锚点的位置生成了一个绿色的上浮加分文本（如“+500”），且该文本在向上漂浮一段距离后逐渐变透明直至消失。\",\n \"在游戏内通过抓取钻石等高价值物品，使当前分数达到或超过800分。截屏确认此时游戏操作暂停，屏幕正中央弹出一个带有彩色背景或烟花效果的“胜利！”提示面板，且面板上提供了一个“返回首页”按钮。\",\n \"在胜利面板点击“返回首页”或刷新页面后，在首页点击“双人模式”按钮进入游戏。截屏确认地表上出现了左右两个相隔一定距离的独立起重机锚点和钩子。分别按下“S”键和“下方向键(ArrowDown)”，确认两个钩子能够独立摆动和伸出，互相不干扰。\",\n \"在双人模式下，使用“S”键和“下方向键”分别控制两名玩家的钩子独立抓取地下不同的物品拉回。观察左上角UI，确认两名玩家抓取物品获得的分数均正确汇总到了同一个“当前分数”统计中。\"\n]","isConversational":false}}},{"rowIdx":77,"cells":{"task_id":{"kind":"number","value":78,"string":"78"},"repository_policy":{"kind":"string","value":"请你在后续所有交互轮次中始终遵循以下约束条件：\n1. 你的任何用户回复都必须全程使用中文进行生成。\n2. 你所生成的所有代码文件都必须使用UTF-8进行编码。\n3. 你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内。\n4. 在生成的代码文件中，必须保证所有函数的圈复杂度上限为4。\n5. 在生成的代码文件中，严禁使用多层if。\n6. 所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层。\n7. 所生成的各个文件，行数均不得大于200行。\n8. 务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码、脚本、配置文件及文本文档扩展名集合\n code_extensions = {\n '.py', '.pyx', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.hxx', \n '.java', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', \n '.scss', '.sass', '.less', '.go', '.rs', '.php', '.rb', '.cs', \n '.swift', '.kt', '.kts', '.sh', '.bash', '.zsh', '.bat', '.cmd', \n '.ps1', '.sql', '.xml', '.json', '.yaml', '.yml', '.md', '.txt', \n '.ini', '.cfg', '.conf', '.vue', '.svelte', '.dart', '.m', '.pl', \n '.pm', '.lua', '.r', '.jl'\n }\n\n # 如果工作区路径不存在，视作没有违反规则\n if not os.path.exists(workspace_path):\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取整个文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明包含非 utf-8 编码字符\n return False\n except Exception:\n # 如果是因为其他原因（如文件不可读或无权限）导致报错，则直接跳过\n continue\n\n # 所有被验证的代码文件均成功使用 utf-8 读取，返回 True\n return True","tags":["Language","File Encoding"]},{"constraint":"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import ast\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历项目目录结构\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅解析 Python 代码文件\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n source_code = f.read()\n \n # 生成抽象语法树\n tree = ast.parse(source_code)\n \n # 遍历语法树所有的节点\n for node in ast.walk(tree):\n # 判断是否为普通函数定义或异步函数定义（包含类中的方法定义）\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_line = node.lineno\n end_line = node.end_lineno\n \n # 计算函数的总行数 (包括可能存在的装饰器)\n func_length = end_line - start_line + 1\n \n # 校验该函数行数是否在 20-40 范围内\n if func_length < 20 or func_length > 40:\n return False\n except Exception:\n # 如果遇到语法错误导致解析失败等异常，忽略并跳过当前文件\n continue\n\n # 如果所有函数检查完毕均未违反规则，且无其他问题，则返回 True\n return True","tags":["Quantity","Range"]},{"constraint":"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中，严禁使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的各个文件，行数均不得大于200行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作空间目录不存在，则没有违反文件行数的条件\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n file_path = os.path.join(root, file_name)\n \n # 确保是文件\n if not os.path.isfile(file_path):\n continue\n \n try:\n # 尝试以 UTF-8 编码读取文件，统计行数\n with open(file_path, 'r', encoding='utf-8') as f:\n line_count = 0\n for _ in f:\n line_count += 1\n \n # 如果某个文件的行数大于200行，则不满足要求\n if line_count > 200:\n return False\n \n except UnicodeDecodeError:\n # 可能是二进制文件或是非UTF-8编码文件，跳过统计\n continue\n except Exception:\n # 忽略其他读取异常\n continue\n\n # 所有文件检查完毕，均未超过200行\n return True","tags":["Quantity","Range"]},{"constraint":"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的业务逻辑源代码文件后缀\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.js', '.jsx', \n '.ts', '.tsx', '.go', '.rs', '.cs', '.php', '.swift', '.kt', \n '.rb', '.sh'\n }\n\n # 辅助函数：清洗单行注释文本中多余的常见注释边框标记\n def clean_comment_line(line):\n line = line.strip()\n while True:\n old_line = line\n if line.startswith('/*'): line = line[2:].strip()\n if line.endswith('*/'): line = line[:-2].strip()\n if line.startswith('<!--'): line = line[4:].strip()\n if line.endswith('-->'): line = line[:-3].strip()\n if line.startswith('//'): line = line[2:].strip()\n if line.startswith('#'): line = line[1:].strip()\n if line.startswith('*'): line = line[1:].strip()\n # 如果没有变化，说明边界清理完成\n if line == old_line:\n break\n return line\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_exts:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 如果是空文件，则跳过不计较注释\n if not content.strip():\n continue\n\n has_code_file = True\n comments = []\n\n # 根据扩展名采取不同的正则，利用正则的“捕获组顺序”跳过各种语言的字符串字面量\n if ext in {'.py', '.rb', '.sh'}:\n # 匹配 Python/Ruby 等形式的三引号多行字符串，或单双引号字符串，或者是 '#' 开头的注释\n pattern = r'(\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*?\"|\\'(?:\\\\.|[^\\'\\\\])*?\\')|(#[^\\n]*)'\n for match in re.finditer(pattern, content):\n if match.group(2):\n comments.append(match.group(2))\n elif ext in {'.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.js', '.jsx', '.ts', '.tsx', '.go', '.rs', '.cs', '.php', '.swift', '.kt'}:\n # 匹配 C-Family 等的反引号(多行)，或单双引号，或者是 // ... 或 /* ... */ 的注释\n pattern = r'(`[\\s\\S]*?`|\"(?:\\\\.|[^\"\\\\])*?\"|\\'(?:\\\\.|[^\\'\\\\])*?\\')|(//[^\\n]*|/\\*[\\s\\S]*?\\*/)'\n for match in re.finditer(pattern, content):\n if match.group(2):\n comments.append(match.group(2))\n\n valid_lines = []\n for comment_text in comments:\n # 拆分多行注释（例如 /* ... */ 往往可能包含很多行）\n lines = comment_text.split('\\n')\n for line in lines:\n cleaned = clean_comment_line(line)\n if cleaned:\n # 忽略仅包含纯标点或分隔符（如 \"###########\" 或 \"--------\"）的非实质性注释行\n # 一行有效的注释理应包含字母、数字或中日韩文字\n if not re.search(r'[A-Za-z0-9\\u4e00-\\u9fa5]', cleaned):\n continue\n valid_lines.append(cleaned)\n\n # 该代码文件必须保证有注释\n if not valid_lines:\n return False\n \n # 校验每一行有效注释文字的字数是否满足 >= 20 的条件\n for line in valid_lines:\n if count_word(line) < 20:\n return False\n\n return True","tags":["Quantity","Range"]}],"string":"[\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码、脚本、配置文件及文本文档扩展名集合\\n code_extensions = {\\n '.py', '.pyx', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.hxx', \\n '.java', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.css', \\n '.scss', '.sass', '.less', '.go', '.rs', '.php', '.rb', '.cs', \\n '.swift', '.kt', '.kts', '.sh', '.bash', '.zsh', '.bat', '.cmd', \\n '.ps1', '.sql', '.xml', '.json', '.yaml', '.yml', '.md', '.txt', \\n '.ini', '.cfg', '.conf', '.vue', '.svelte', '.dart', '.m', '.pl', \\n '.pm', '.lua', '.r', '.jl'\\n }\\n\\n # 如果工作区路径不存在，视作没有违反规则\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取整个文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明包含非 utf-8 编码字符\\n return False\\n except Exception:\\n # 如果是因为其他原因（如文件不可读或无权限）导致报错，则直接跳过\\n continue\\n\\n # 所有被验证的代码文件均成功使用 utf-8 读取，返回 True\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历项目目录结构\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅解析 Python 代码文件\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n source_code = f.read()\\n \\n # 生成抽象语法树\\n tree = ast.parse(source_code)\\n \\n # 遍历语法树所有的节点\\n for node in ast.walk(tree):\\n # 判断是否为普通函数定义或异步函数定义（包含类中的方法定义）\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_line = node.lineno\\n end_line = node.end_lineno\\n \\n # 计算函数的总行数 (包括可能存在的装饰器)\\n func_length = end_line - start_line + 1\\n \\n # 校验该函数行数是否在 20-40 范围内\\n if func_length < 20 or func_length > 40:\\n return False\\n except Exception:\\n # 如果遇到语法错误导致解析失败等异常，忽略并跳过当前文件\\n continue\\n\\n # 如果所有函数检查完毕均未违反规则，且无其他问题，则返回 True\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，严禁使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的各个文件，行数均不得大于200行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作空间目录不存在，则没有违反文件行数的条件\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n file_path = os.path.join(root, file_name)\\n \\n # 确保是文件\\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n # 尝试以 UTF-8 编码读取文件，统计行数\\n with open(file_path, 'r', encoding='utf-8') as f:\\n line_count = 0\\n for _ in f:\\n line_count += 1\\n \\n # 如果某个文件的行数大于200行，则不满足要求\\n if line_count > 200:\\n return False\\n \\n except UnicodeDecodeError:\\n # 可能是二进制文件或是非UTF-8编码文件，跳过统计\\n continue\\n except Exception:\\n # 忽略其他读取异常\\n continue\\n\\n # 所有文件检查完毕，均未超过200行\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的业务逻辑源代码文件后缀\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.js', '.jsx', \\n '.ts', '.tsx', '.go', '.rs', '.cs', '.php', '.swift', '.kt', \\n '.rb', '.sh'\\n }\\n\\n # 辅助函数：清洗单行注释文本中多余的常见注释边框标记\\n def clean_comment_line(line):\\n line = line.strip()\\n while True:\\n old_line = line\\n if line.startswith('/*'): line = line[2:].strip()\\n if line.endswith('*/'): line = line[:-2].strip()\\n if line.startswith('<!--'): line = line[4:].strip()\\n if line.endswith('-->'): line = line[:-3].strip()\\n if line.startswith('//'): line = line[2:].strip()\\n if line.startswith('#'): line = line[1:].strip()\\n if line.startswith('*'): line = line[1:].strip()\\n # 如果没有变化，说明边界清理完成\\n if line == old_line:\\n break\\n return line\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_exts:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 如果是空文件，则跳过不计较注释\\n if not content.strip():\\n continue\\n\\n has_code_file = True\\n comments = []\\n\\n # 根据扩展名采取不同的正则，利用正则的“捕获组顺序”跳过各种语言的字符串字面量\\n if ext in {'.py', '.rb', '.sh'}:\\n # 匹配 Python/Ruby 等形式的三引号多行字符串，或单双引号字符串，或者是 '#' 开头的注释\\n pattern = r'(\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*?\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*?\\\\')|(#[^\\\\n]*)'\\n for match in re.finditer(pattern, content):\\n if match.group(2):\\n comments.append(match.group(2))\\n elif ext in {'.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.js', '.jsx', '.ts', '.tsx', '.go', '.rs', '.cs', '.php', '.swift', '.kt'}:\\n # 匹配 C-Family 等的反引号(多行)，或单双引号，或者是 // ... 或 /* ... */ 的注释\\n pattern = r'(`[\\\\s\\\\S]*?`|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*?\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*?\\\\')|(//[^\\\\n]*|/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n for match in re.finditer(pattern, content):\\n if match.group(2):\\n comments.append(match.group(2))\\n\\n valid_lines = []\\n for comment_text in comments:\\n # 拆分多行注释（例如 /* ... */ 往往可能包含很多行）\\n lines = comment_text.split('\\\\n')\\n for line in lines:\\n cleaned = clean_comment_line(line)\\n if cleaned:\\n # 忽略仅包含纯标点或分隔符（如 \\\"###########\\\" 或 \\\"--------\\\"）的非实质性注释行\\n # 一行有效的注释理应包含字母、数字或中日韩文字\\n if not re.search(r'[A-Za-z0-9\\\\u4e00-\\\\u9fa5]', cleaned):\\n continue\\n valid_lines.append(cleaned)\\n\\n # 该代码文件必须保证有注释\\n if not valid_lines:\\n return False\\n \\n # 校验每一行有效注释文字的字数是否满足 >= 20 的条件\\n for line in valid_lines:\\n if count_word(line) < 20:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"我需要用HTML、CSS和JS实现一个科普实验网站，演示'两个物体同时下落'。第一步，先帮我搭一个页面的基础骨架。页面采用左右布局：左侧是参数控制面板，现在只要放几个带边框的空div作为占位符即可；右侧是实验显示区，画一个带有背景色的较高矩形区域代表下落空间。在右侧区域的顶部，放置两个简单的纯色圆形代表准备下落的物体。在页面底部放'开始'和'重置'两个按钮。这一轮暂时不需要实现任何下落动画逻辑，只要页面布局合理、自适应屏幕且稍微美观一点就行。另外，后续所有轮次你都需要始终遵循以下要求：第一，修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它；第二，你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能生成的代码文件扩展名\n target_extensions = {'.html', '.css', '.js'}\n has_checked_files = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_checked_files = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果出现解码错误，则说明不是合法的 UTF-8 编码\n return False\n except Exception:\n # 如果由于其他异常无法读取，视为不满足\n return False\n\n # 若至少存在一个代码文件，且均未引发解码错误，则视为满足要求\n return has_checked_files","tags":["Language","File Encoding"]},{"constraint":"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def clean_code(code: str) -> str:\n \"\"\"\n 简单的词法清理：将字符串内内容及注释替换为空格，避免大括号等特殊符号干扰解析，\n 同时保持原字符串的长度和换行符完全一致。\n \"\"\"\n result = []\n in_single_quote = False\n in_double_quote = False\n in_template_quote = False\n in_line_comment = False\n in_block_comment = False\n i = 0\n n = len(code)\n while i < n:\n c = code[i]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n result.append(c)\n else:\n result.append(' ')\n i += 1\n elif in_block_comment:\n if c == '*' and i + 1 < n and code[i+1] == '/':\n in_block_comment = False\n result.append(' ')\n i += 2\n else:\n if c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n elif in_single_quote:\n if c == '\\\\' and i + 1 < n:\n if code[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n elif c == \"'\":\n in_single_quote = False\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n elif in_double_quote:\n if c == '\\\\' and i + 1 < n:\n if code[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n elif c == '\"':\n in_double_quote = False\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n elif in_template_quote:\n if c == '\\\\' and i + 1 < n:\n if code[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n elif c == '`':\n in_template_quote = False\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n else:\n if c == '/' and i + 1 < n and code[i+1] == '/':\n in_line_comment = True\n result.append(' ')\n i += 2\n elif c == '/' and i + 1 < n and code[i+1] == '*':\n in_block_comment = True\n result.append(' ')\n i += 2\n elif c == \"'\":\n in_single_quote = True\n result.append(' ')\n i += 1\n elif c == '\"':\n in_double_quote = True\n result.append(' ')\n i += 1\n elif c == '`':\n in_template_quote = True\n result.append(' ')\n i += 1\n else:\n result.append(c)\n i += 1\n return \"\".join(result)\n\n function_lengths = []\n \n if not os.path.exists(workspace_path):\n return False\n\n # 遍历代码项目\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n js_contents = []\n if file.endswith('.js'):\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n js_contents.append(f.read())\n elif file.endswith('.html'):\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n content = f.read()\n # 提取全部 <script> 内的内容\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n js_contents.extend(scripts)\n \n for js_code in js_contents:\n if not js_code.strip():\n continue\n # 获取剔除干扰后的“干净代码面具”\n js_code_clean = clean_code(js_code)\n \n # 寻找各类函数的起始特征\n matches = list(re.finditer(r'\\bfunction\\b[^{]*\\{', js_code_clean))\n matches += list(re.finditer(r'=>\\s*\\{', js_code_clean))\n method_pattern = r'\\b([a-zA-Z_$][\\w_$]*)\\s*\\([^)]*\\)\\s*\\{'\n for m in re.finditer(method_pattern, js_code_clean):\n keyword = m.group(1)\n # 排除控制流关键字的误判\n if keyword not in ['if', 'for', 'while', 'catch', 'switch', 'with', 'function']:\n matches.append(m)\n \n # 去重重叠的匹配（共享同一个左大括号的匹配项视为同一个函数起点）\n starts = set()\n valid_matches = []\n for m in matches:\n brace_idx = m.end() - 1\n if brace_idx not in starts:\n starts.add(brace_idx)\n valid_matches.append(m)\n \n # 从匹配处往后解析匹配对应大括号提取函数体范围\n for m in valid_matches:\n start_index = m.start()\n brace_index = m.end() - 1\n \n stack_count = 1\n j = brace_index + 1\n while j < len(js_code_clean) and stack_count > 0:\n if js_code_clean[j] == '{':\n stack_count += 1\n elif js_code_clean[j] == '}':\n stack_count -= 1\n j += 1\n \n if stack_count == 0:\n end_index = j\n # 使用原代码切片以获取精准包含的代码段\n func_body = js_code[start_index:end_index]\n line_count = func_body.count('\\n') + 1\n function_lengths.append(line_count)\n\n # 验证是否任意单个函数的行数均在20-40（闭区间）内\n for length in function_lengths:\n if length < 20 or length > 40:\n return False\n \n # 若全部符合要求，或尚无任何函数被创建，都视为符合规则\n return True","tags":["Quantity","Range"]},{"constraint":"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中，严禁使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的各个文件，行数均不得大于200行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 根据用户指令，推断可能的代码文件类型\n target_extensions = {'.html', '.css', '.js'}\n\n # 遍历工作区内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 获取文件后缀名\n ext = os.path.splitext(file_name)[1].lower()\n \n # 仅检查目标类型的文件\n if ext in target_extensions:\n file_path = os.path.join(root, file_name)\n try:\n # 读取文件行数\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 判断行数是否大于200行\n if len(lines) > 200:\n return False\n except Exception:\n # 如果由于非预期原因（如编码问题等）无法读取，为了鲁棒性暂且忽略该不可读文件\n pass\n\n # 如果所有被检查的目标文件行数均未超过200行，则满足要求\n return True","tags":["Quantity","Range"]},{"constraint":"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 网站相关的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return False\n \n comments_lines = []\n \n # HTML 注释提取 (<!-- ... -->)\n if ext == '.html':\n for m in re.finditer(r'<!--(.*?)-->', content, re.DOTALL):\n comments_lines.extend(m.group(1).split('\\n'))\n \n # 块级注释提取 (/* ... */) 适用于 HTML, CSS, JS\n if ext in {'.html', '.css', '.js'}:\n for m in re.finditer(r'/\\*(.*?)\\*/', content, re.DOTALL):\n comments_lines.extend(m.group(1).split('\\n'))\n \n # 单行注释提取 (// ...) 适用于 HTML, JS\n if ext in {'.html', '.js'}:\n # (?<!:) 用于防止匹配 http:// 等 URL 结构\n for m in re.finditer(r'(?<!:)\\/\\/(.*)', content):\n comments_lines.append(m.group(1))\n \n file_has_valid_comment = False\n \n for line in comments_lines:\n text = line.strip()\n if not text:\n continue\n \n # 多行注释的每行常以 * 开头，需将其剥离以保留纯文本\n if text.startswith('*'):\n text = text.lstrip('*').strip()\n \n if not text:\n continue\n \n file_has_valid_comment = True\n \n # 验证每行注释的字数是否 >= 20\n if count_word(text) < 20:\n return False\n \n # 必须保证每个代码文件都包含注释\n if not file_has_valid_comment:\n return False\n \n # 如果遍历完成且至少检查过一个代码文件，则要求满足\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":1,"instruction":"好，接下来我们把左侧的控制面板完善一下，并加上基础的触发机制。请用真实的表单元素替换掉之前的占位符，表单需要包含：物体1类型（下拉选择：小球/羽毛）、物体1重量（滑块：1-100）、物体2类型、物体2重量、是否为真空环境（开关或复选框）、空气阻力系数（滑块）、重力加速度（滑块，默认9.8）。然后，给底部的按钮加上功能：点击'开始'后，右侧的两个圆圈以一个固定的像素值向下匀速移动，碰到容器底部停止；点击'重置'让它们回到顶部。暂时不用管真实的物理公式，只要表单能用、物体能动就行。此外，后续所有轮次你都需要始终遵循以下新增要求：第一，在更新已有文件内容时，必须输出严格的Unified Diff格式，不允许直接输出全量代码替换；第二，务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目描述推断可能的代码文件和文本文件扩展名\n code_extensions = {'.html', '.css', '.js', '.log'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试将文件内容按照 utf-8 格式解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明不是合法的 utf-8 编码\n return False\n except Exception:\n # 其他读取异常等保守起见视为验证不通过\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内","validation_code":"def verify_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 遍历工作区目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html'):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于某种原因遇到读取错误则跳过该文件\n continue\n \n codes_to_parse = []\n if file.endswith('.js'):\n codes_to_parse.append(content)\n else:\n # 对于HTML文件，只提取 <script> 标签内的内容\n pattern = re.compile(r'<script.*?>([\\s\\S]*?)</script>', re.IGNORECASE)\n for match in pattern.finditer(content):\n codes_to_parse.append(match.group(1))\n \n for code in codes_to_parse:\n n = len(code)\n sanitized_code = list(code)\n i = 0\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n # 第一阶段：剥离字符串与注释，将其替换为空格，保留换行符\n while i < n:\n c = code[i]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n else:\n sanitized_code[i] = ' '\n i += 1\n continue\n \n if in_block_comment:\n if c == '*' and i + 1 < n and code[i+1] == '/':\n sanitized_code[i] = ' '\n sanitized_code[i+1] = ' '\n in_block_comment = False\n i += 2\n else:\n if c != '\\n':\n sanitized_code[i] = ' '\n i += 1\n continue\n \n if in_string:\n if c == '\\\\':\n sanitized_code[i] = ' '\n if i + 1 < n:\n if sanitized_code[i+1] != '\\n':\n sanitized_code[i+1] = ' '\n i += 2\n elif c == string_char:\n in_string = False\n sanitized_code[i] = ' '\n i += 1\n else:\n if c != '\\n':\n sanitized_code[i] = ' '\n i += 1\n continue\n \n if c in (\"'\", '\"', '`'):\n in_string = True\n string_char = c\n sanitized_code[i] = ' '\n i += 1\n continue\n \n if c == '/':\n if i + 1 < n and code[i+1] == '/':\n in_line_comment = True\n sanitized_code[i] = ' '\n sanitized_code[i+1] = ' '\n i += 2\n continue\n elif i + 1 < n and code[i+1] == '*':\n in_block_comment = True\n sanitized_code[i] = ' '\n sanitized_code[i+1] = ' '\n i += 2\n continue\n \n i += 1\n\n sanitized = \"\".join(sanitized_code)\n \n function_starts = []\n \n # 第二阶段：寻找所有可能的函数起始标识\n # 1. 查找 function 关键字\n idx = 0\n while True:\n idx = sanitized.find('function', idx)\n if idx == -1:\n break\n # 确保是一个独立的单词而不是变量名的后缀/前缀\n if (idx == 0 or not (sanitized[idx-1].isalnum() or sanitized[idx-1] == '_')) and \\\n (idx + 8 >= len(sanitized) or not (sanitized[idx+8].isalnum() or sanitized[idx+8] == '_')):\n function_starts.append(idx)\n idx += 8\n \n # 2. 查找箭头函数 =>\n idx = 0\n while True:\n idx = sanitized.find('=>', idx)\n if idx == -1:\n break\n function_starts.append(idx)\n idx += 2\n \n # 3. 查找类方法/对象简写方法 (如 myMethod(arg) { )\n method_pattern = re.compile(r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\([^)]*\\)\\s*\\{')\n for match in method_pattern.finditer(sanitized):\n name = match.group(1)\n if name not in ('if', 'for', 'while', 'catch', 'switch', 'function', 'return', 'with'):\n function_starts.append(match.start())\n \n function_starts = sorted(list(set(function_starts)))\n \n # 过滤重叠定义，建立匹配大括号的关联\n brace_to_start = {}\n for start_idx in function_starts:\n brace_idx = sanitized.find('{', start_idx)\n if brace_idx == -1:\n continue\n \n # 如果标记到最近大括号间有严重中断符(;)，可能属误识别，忽略\n between = sanitized[start_idx:brace_idx]\n if ';' in between:\n continue\n \n # 确保 => 紧连着大括号（允许换行/空格），防止捕获隐式返回的箭头函数后的无关大括号\n if sanitized[start_idx:start_idx+2] == '=>':\n between_arrow = sanitized[start_idx+2:brace_idx]\n if not between_arrow.isspace() and between_arrow != \"\":\n continue\n \n if brace_idx not in brace_to_start:\n brace_to_start[brace_idx] = start_idx\n else:\n if start_idx < brace_to_start[brace_idx]:\n brace_to_start[brace_idx] = start_idx\n \n # 第三阶段：解析范围及计算行数\n for brace_idx, start_idx in brace_to_start.items():\n true_start = start_idx\n # 对箭头函数，尝试向左寻找包裹参数的 '(' 以计算完整行数\n if sanitized[start_idx:start_idx+2] == '=>':\n left_ptr = start_idx - 1\n while left_ptr >= 0 and sanitized[left_ptr].isspace():\n left_ptr -= 1\n if left_ptr >= 0 and sanitized[left_ptr] == ')':\n count = 1\n curr = left_ptr - 1\n while curr >= 0:\n if sanitized[curr] == ')':\n count += 1\n elif sanitized[curr] == '(':\n count -= 1\n if count == 0:\n true_start = curr\n break\n curr -= 1\n \n # 寻找匹配的结束大括号\n count = 1\n end_idx = -1\n for j in range(brace_idx + 1, len(sanitized)):\n if sanitized[j] == '{':\n count += 1\n elif sanitized[j] == '}':\n count -= 1\n if count == 0:\n end_idx = j\n break\n \n if end_idx != -1:\n # 统计真实起点到终止大括号之间换行符的数量，获取函数代码块所占行数\n start_line = sanitized.count('\\n', 0, true_start) + 1\n end_line = sanitized.count('\\n', 0, end_idx) + 1\n line_count = end_line - start_line + 1\n \n # 如果函数的行数不满足限定在20-40行的标准，返回 False\n if line_count < 20 or line_count > 40:\n return False\n return True","tags":["Quantity","Range"]},{"constraint":"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中，严禁使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的各个文件，行数均不得大于200行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在且为目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 根据用户指令推断可能生成的文件类型\n valid_extensions = {'.html', '.css', '.js', '.log'}\n files_checked = False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n files_checked = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 统计文件的行数\n line_count = sum(1 for _ in f)\n \n # 检查行数是否大于200行\n if line_count > 200:\n return False\n except Exception:\n # 如果遇到文件读取等异常，跳过处理\n continue\n\n # 至少检查了一个相关文件且没有出现大于200行的情况才视为符合\n return files_checked","tags":["Quantity","Range"]},{"constraint":"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字","validation_code":"def check_comment_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.isdir(workspace_path):\n return False\n \n valid_exts = {'.html', '.css', '.js'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 使用正则表达式提取文件中的各类注释\n # 1. 匹配 HTML 注释：<!-- ... -->\n html_comments = re.findall(r'<!--([\\s\\S]*?)-->', content)\n # 2. 匹配 CSS/JS 多行注释：/* ... */\n multiline_comments = re.findall(r'/\\*([\\s\\S]*?)\\*/', content)\n # 3. 匹配 JS 单行注释：// ...（仅匹配行首或在空白字符之后的//，避开网址中的//）\n singleline_comments = re.findall(r'(?:^|\\s)//(.*)', content)\n \n valid_comment_lines = 0\n \n # 检查 HTML 注释的内容\n for comment in html_comments:\n for line in comment.split('\\n'):\n clean_line = line.strip()\n if clean_line:\n valid_comment_lines += 1\n if count_word(clean_line) < 20:\n return False\n \n # 检查 CSS/JS 多行注释的内容\n for comment in multiline_comments:\n for line in comment.split('\\n'):\n # 去除开头的 '*' 符号（多行注释换行经常带有装饰性星号），再去除两端空白\n clean_line = line.strip().lstrip('*').strip()\n if clean_line:\n valid_comment_lines += 1\n if count_word(clean_line) < 20:\n return False\n \n # 检查 JS 单行注释的内容\n for comment in singleline_comments:\n clean_line = comment.strip()\n if clean_line:\n valid_comment_lines += 1\n if count_word(clean_line) < 20:\n return False\n \n # 每个代码文件必须至少包含一行有效注释\n if valid_comment_lines == 0:\n return False\n \n # 如果项目中没有任何代码文件，也判定为不满足\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在更新已有文件内容时，必须输出严格的Unified Diff格式，不允许直接输出全量代码替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":2,"instruction":"基础交互做得很好了。现在我们要把核心的物理引擎加进去，让这两个物体的下落真正遵循物理规律。请修改JS里的动画计算逻辑：使用 RequestAnimationFrame 重构循环。在每一帧中，利用牛顿第二定律计算加速度(a = (mg - f)/m)。其中重力G=mg，空气阻力设定为f=k*v（k为表单中的风阻系数，v为当前下落速度）。注意：如果勾选了'真空环境'，则阻力f恒为0；另外，下拉框选择'羽毛'时，请在代码里给它加上一个阻力放大系数，模拟其迎风面积大的特点。修改完成后，点击开始，非真空下应该能明显看到羽毛下落比小球慢，而真空中两者应该完全同时落地。另外，后续所有轮次你都需要始终遵循以下调整后的要求：第一，只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式；第二，进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令推断，主要生成的代码文件包括 HTML, CSS, JS\n code_extensions = {'.html', '.css', '.js'}\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，则文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 如果发生其他读取错误，保守判定为不满足\n return False\n\n # 必须至少找到一个代码文件且全部验证通过，才算真正满足了要求\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 屏蔽代码中的字符串和注释，替换为空格但保留换行符，以防止误匹配符号并保持原行数不变\n def mask_code(c: str) -> str:\n result = list(c)\n i = 0\n n = len(c)\n while i < n:\n if c[i:i+2] == '//':\n j = i\n while j < n and c[j] != '\\n':\n result[j] = ' '\n j += 1\n i = j\n elif c[i:i+2] == '/*':\n j = i\n while j < n and c[j:j+2] != '*/':\n if result[j] != '\\n':\n result[j] = ' '\n j += 1\n if j < n:\n result[j] = ' '\n result[j+1] = ' '\n j += 2\n i = j\n elif c[i] in [\"'\", '\"', \"`\"]:\n quote = c[i]\n result[i] = ' '\n j = i + 1\n while j < n and c[j] != quote:\n if c[j] == '\\\\':\n result[j] = ' '\n j += 1\n if j < n and result[j] != '\\n':\n result[j] = ' '\n else:\n if result[j] != '\\n':\n result[j] = ' '\n j += 1\n if j < n:\n result[j] = ' '\n j += 1\n i = j\n else:\n i += 1\n return \"\".join(result)\n\n def extract_and_check(code: str) -> bool:\n masked = mask_code(code)\n n = len(masked)\n \n for i in range(n):\n if masked[i] == '{':\n is_func = False\n start_pos = i\n \n j = i - 1\n while j >= 0 and masked[j].isspace():\n j -= 1\n if j < 0:\n continue\n \n # 箭头函数的 =>\n if masked[j] == '>':\n if j >= 1 and masked[j-1] == '=':\n is_func = True\n start_pos = j - 1\n k = start_pos - 1\n while k >= 0 and masked[k].isspace(): k -= 1\n if k >= 0 and masked[k] == ')':\n paren_c = 1\n k -= 1\n while k >= 0 and paren_c > 0:\n if masked[k] == ')': paren_c += 1\n elif masked[k] == '(': paren_c -= 1\n k -= 1\n start_pos = k + 1\n else:\n while k >= 0 and (masked[k].isalnum() or masked[k] in '_$'): k -= 1\n start_pos = k + 1\n \n # 普通函数或方法的参数 )\n elif masked[j] == ')':\n paren_c = 1\n k = j - 1\n while k >= 0 and paren_c > 0:\n if masked[k] == ')': paren_c += 1\n elif masked[k] == '(': paren_c -= 1\n k -= 1\n \n if k >= 0:\n start_pos = k + 1\n p = k - 1\n while p >= 0 and masked[p].isspace(): p -= 1\n \n if p >= 0:\n if masked[p].isalnum() or masked[p] in '_$':\n q = p\n while q >= 0 and (masked[q].isalnum() or masked[q] in '_$'): q -= 1\n word = masked[q+1:p+1]\n \n if word == 'function':\n is_func = True\n start_pos = q + 1\n # 排查非函数控制流语句块\n elif word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n w2_p = q\n while w2_p >= 0 and masked[w2_p].isspace(): w2_p -= 1\n w2_q = w2_p\n while w2_q >= 0 and (masked[w2_q].isalnum() or masked[w2_q] in '_$'): w2_q -= 1\n word2 = masked[w2_q+1:w2_p+1] if w2_q < w2_p else \"\"\n \n if word2 == 'function':\n is_func = True\n start_pos = w2_q + 1\n else:\n is_func = True\n if word2 in ['get', 'set']:\n start_pos = w2_q + 1\n else:\n start_pos = q + 1\n else:\n # 对于动态命名或未定名的函数形式，只要以)结束，极大可能为合法的函数语法树\n is_func = True\n \n # 如果定位到是函数，向后匹配完整的 {} 区域并计算行数\n if is_func:\n brace_c = 1\n end_pos = -1\n for end_i in range(i + 1, n):\n if masked[end_i] == '{': brace_c += 1\n elif masked[end_i] == '}':\n brace_c -= 1\n if brace_c == 0:\n end_pos = end_i\n break\n \n if end_pos != -1:\n start_line = code.count('\\n', 0, start_pos) + 1\n end_line = code.count('\\n', 0, end_pos) + 1\n lines = end_line - start_line + 1\n \n # 检查行数是否在20-40之间\n if lines < 20 or lines > 40:\n return False\n \n return True\n\n # 遍历 workspace_path 中生成的 js 与 html 文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html'):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if file.endswith('.js'):\n if not extract_and_check(content):\n return False\n elif file.endswith('.html'):\n # 匹配内联的 JavaScript 脚本块\n blocks = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for block in blocks:\n if not extract_and_check(block):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中，严禁使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的各个文件，行数均不得大于200行","validation_code":"def check_file_line_count(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return True\n\n # 根据用户指令推断可能的代码文件类型（HTML、CSS、JS、LOG）\n target_extensions = ('.html', '.css', '.js', '.log')\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if file_name.endswith(target_extensions):\n file_path = os.path.join(root, file_name)\n try:\n # 系统提示明确要求使用 UTF-8 进行编码\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n # 判断行数是否大于200\n if len(lines) > 200:\n return False\n except Exception:\n # 忽略读取或编码异常的文件\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字","validation_code":"def evaluate(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def clean_strings(text: str) -> str:\n # 移除代码中双引号、单引号和反引号包裹的字符串内容，避免内部的 // 或 /* 被误认为注释\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', text)\n text = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', text)\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', text)\n return text\n\n # 推断可能的代码文件类型\n code_extensions = ['.html', '.htm', '.css', '.js']\n code_files = []\n \n # 查找工作区下的所有目标代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.lower().endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n if not code_files:\n return False\n \n # 逐一检查每个代码文件\n for filepath in code_files:\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return False\n \n ext = os.path.splitext(filepath)[1].lower()\n comments = []\n \n if ext in ['.html', '.htm']:\n # 提取 HTML 注释\n comments.extend(re.findall(r'<!--([\\s\\S]*?)-->', content))\n \n # 提取内部的 script 和 style 内容处理内部 JS 和 CSS 注释\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n styles = re.findall(r'<style[^>]*>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n js_css_content = '\\n'.join(scripts + styles)\n \n if js_css_content:\n js_css_content = clean_strings(js_css_content)\n comments.extend(re.findall(r'/\\*([\\s\\S]*?)\\*/', js_css_content))\n # 使用后向否定断言避免匹配 http:// 和 https:// 中的双斜杠\n comments.extend(re.findall(r'(?<!:)//(.*)', js_css_content))\n \n elif ext in ['.js', '.css']:\n content_cleaned = clean_strings(content)\n comments.extend(re.findall(r'/\\*([\\s\\S]*?)\\*/', content_cleaned))\n comments.extend(re.findall(r'(?<!:)//(.*)', content_cleaned))\n \n # 若文件中没有任何注释提取出来，说明不满足“每个生成的代码文件里注释详细”的要求\n if not comments:\n return False\n \n has_valid_comment = False\n \n for c in comments:\n lines = c.split('\\n')\n for line in lines:\n line = line.strip()\n # 过滤掉多行注释里常规的前缀符号\n if line.startswith('*'):\n line = line[1:].strip()\n \n # 忽略完全为空的行\n if not line:\n continue\n \n has_valid_comment = True\n \n # 验证每行有效注释的字数是否满足下限要求\n if count_word(line) < 20:\n return False\n \n # 若注释全部是空行等无效内容，也不达标\n if not has_valid_comment:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"物理引擎看起来生效了，非常棒！下一步，我们需要给实验加上精确的数据统计和结果展示功能。请在右侧实验区的顶端中间加上一个'高精度计时器'，点击开始后从 0.00 秒开始实时滚动更新。当任意一个物体碰到地面（下边界）时，瞬间记录下它的落地时间并在其底部显示出来。当两个物体都触底后，在实验区中央弹出一个半透明的实验报告面板，清晰地对比：物体1用时多少，物体2用时多少，并根据时间差自动生成一句简单的物理结论（例如'真空中物体下落加速度与质量无关'或'受空气阻力影响，物体落地时间产生差异'）。此外，后续所有轮次你都需要始终遵循以下调整及新增要求：第一，只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换；第二，请务必尽量通过专用工具（如Read, Write, Edit）来完成文件操作，而不是借助 bash 命令。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断，科普实验网站项目主要包含以下类型的前端代码文件\n code_extensions = {'.html', '.htm', '.css', '.js'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试使用 utf-8 进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 抛出 UnicodeDecodeError 说明该文件含有非 UTF-8 编码的字符\n return False\n except Exception:\n # 发生其他读取异常也视为校验失败\n return False\n\n # 若项目中至少存在一个代码文件且全部通过了 UTF-8 解码测试，则返回 True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内","validation_code":"def check_function_lines(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def mask_code(code: str) -> str:\n \"\"\"屏蔽字符串和注释，将它们替换为等长的空格，以保证后续匹配的安全性\"\"\"\n masked = list(code)\n code_len = len(code)\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n i = 0\n while i < code_len:\n char = code[i]\n if not in_string and not in_line_comment and not in_block_comment:\n if char == '/' and i + 1 < code_len:\n if code[i+1] == '/':\n in_line_comment = True\n masked[i] = ' '\n masked[i+1] = ' '\n i += 1\n elif code[i+1] == '*':\n in_block_comment = True\n masked[i] = ' '\n masked[i+1] = ' '\n i += 1\n elif char in [\"'\", '\"', \"`\"]:\n in_string = True\n string_char = char\n masked[i] = ' '\n elif in_string:\n masked[i] = ' '\n # 处理转义字符\n if char == '\\\\':\n if i + 1 < code_len:\n masked[i+1] = ' '\n i += 1\n elif char == string_char:\n in_string = False\n elif in_line_comment:\n if char == '\\n':\n in_line_comment = False\n else:\n masked[i] = ' '\n elif in_block_comment:\n masked[i] = ' '\n if char == '*' and i + 1 < code_len and code[i+1] == '/':\n masked[i+1] = ' '\n in_block_comment = False\n i += 1\n i += 1\n return \"\".join(masked)\n\n def get_functions_from_js(code: str):\n \"\"\"利用正则和括号匹配从JS代码中提取完整的函数体代码块\"\"\"\n masked = mask_code(code)\n funcs = []\n \n # 特判：如果不带大括号的箭头函数存在，其行数必然极少（一定低于20行），此处加入一个违规的占位伪代码即可使后续检查返回False\n if re.search(r'=>\\s*(?:[^{]|$)', masked):\n funcs.append(\"() => 1;\")\n \n code_len = len(masked)\n idx = 0\n \n # 匹配三种常见的JS函数起始声明形式\n pattern = re.compile(\n r'\\bfunction\\b[^{]*\\{|'\n r'(?:\\b[a-zA-Z_]\\w*\\s*=>\\s*\\{|\\([^)]*\\)\\s*=>\\s*\\{)|'\n r'\\b(?!(?:if|for|while|switch|catch|with|return)\\b)[a-zA-Z_]\\w*\\s*\\([^)]*\\)\\s*\\{'\n )\n \n while idx < code_len:\n match = pattern.search(masked, idx)\n if not match:\n break\n \n absolute_match_start = match.start()\n start_idx = match.end() - 1 # 找到的大括号 '{' 的下标\n \n brace_count = 0\n end_idx = -1\n \n i = start_idx\n while i < code_len:\n char = masked[i]\n if char == '{':\n brace_count += 1\n elif char == '}':\n brace_count -= 1\n if brace_count == 0:\n end_idx = i\n break\n i += 1\n \n if end_idx != -1:\n # 从原始原代码片段截取真实代码\n func_code = code[absolute_match_start : end_idx + 1]\n funcs.append(func_code)\n # 为了能够捕捉到嵌套在内部的子函数，游标只向前推进到大括号内侧\n idx = start_idx + 1\n else:\n idx = match.end()\n \n return funcs\n\n if not os.path.isdir(workspace_path):\n return False\n \n js_codes = []\n py_files = []\n \n # 遍历并收集代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n js_codes.append(f.read())\n except:\n pass\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n html_code = f.read()\n # 从HTML中提取出所有的 <script> 标签内联代码\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', html_code, re.IGNORECASE | re.DOTALL)\n js_codes.extend(scripts)\n except:\n pass\n elif file.endswith('.py'):\n py_files.append(file_path)\n\n all_funcs = []\n for code in js_codes:\n all_funcs.extend(get_functions_from_js(code))\n \n # 验证提取出的所有JS函数行数\n for func in all_funcs:\n lines = func.strip().split('\\n')\n if not (20 <= len(lines) <= 40):\n return False\n \n # 验证提取出的所有Python函数行数\n for py_file in py_files:\n try:\n with open(py_file, 'r', encoding='utf-8') as f:\n code = f.read()\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n end = getattr(node, 'end_lineno', start)\n num_lines = end - start + 1\n if not (20 <= num_lines <= 40):\n return False\n except:\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中，严禁使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的各个文件，行数均不得大于200行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，视作没有违反行数条件\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉可能的版本控制相关隐藏文件夹，如.git\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 逐行统计文件行数，这样也能节省大文件的内存开销\n line_count = sum(1 for _ in f)\n \n # 只要有任何一个文件的行数大于200，就不满足要求\n if line_count > 200:\n return False\n except Exception:\n # 忽略因非UTF-8编码或其他不可读原因导致的异常文件（例如二进制文件）\n pass\n\n # 遍历完全部文件后，没有发现超过200行的文件\n return True","tags":["Quantity","Range"]},{"constraint":"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 确保给定的项目路径存在\n if not os.path.isdir(workspace_path):\n return False\n \n code_extensions = ('.html', '.htm', '.css', '.js')\n \n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if not file.endswith(code_extensions):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n \n # 根据文件类型用对应的正则表达式提取注释\n if file.endswith(('.html', '.htm')):\n # 匹配 HTML 注释\n for match in re.finditer(r'<!--([\\s\\S]*?)-->', content):\n comments.extend(match.group(1).split('\\n'))\n \n # 提取 <script> 内的 js 注释\n for match in re.finditer(r'<script[^>]*>([\\s\\S]*?)</script>', content, flags=re.IGNORECASE):\n script_content = match.group(1)\n # js 单行注释 (使用(?<!:)避免匹配到 URL 的协议部分如 http://)\n for m in re.finditer(r'(?<!:)\\/\\/(.*)', script_content):\n comments.append(m.group(1))\n # js 多行注释\n for m in re.finditer(r'/\\*([\\s\\S]*?)\\*/', script_content):\n comments.extend(m.group(1).split('\\n'))\n \n # 提取 <style> 内的 css 注释\n for match in re.finditer(r'<style[^>]*>([\\s\\S]*?)</style>', content, flags=re.IGNORECASE):\n style_content = match.group(1)\n for m in re.finditer(r'/\\*([\\s\\S]*?)\\*/', style_content):\n comments.extend(m.group(1).split('\\n'))\n \n elif file.endswith('.js'):\n # JS 的单行和多行注释\n for m in re.finditer(r'(?<!:)\\/\\/(.*)', content):\n comments.append(m.group(1))\n for m in re.finditer(r'/\\*([\\s\\S]*?)\\*/', content):\n comments.extend(m.group(1).split('\\n'))\n \n elif file.endswith('.css'):\n # CSS 的多行注释\n for m in re.finditer(r'/\\*([\\s\\S]*?)\\*/', content):\n comments.extend(m.group(1).split('\\n'))\n \n valid_comment_count = 0\n \n # 校验每一行提取出来的注释\n for line in comments:\n clean_line = line.strip()\n # 移除多行注释或者装饰性注释常带的前导无意义字符（如 * 或 -）\n clean_line = re.sub(r'^[\\*\\-\\/]+', '', clean_line).strip()\n clean_line = clean_line.strip('*').strip()\n \n # 跳过空行\n if not clean_line:\n continue\n \n valid_comment_count += 1\n \n # 检查单行字数下限\n if count_word(clean_line) < 20:\n return False\n \n # 必须保证每个生成的代码文件里都有至少一条有效注释以满足\"注释详细\"的要求\n if valid_comment_count == 0:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请务必尽量通过专用工具（如Read, Write, Edit）来完成文件操作，而不是借助 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":4,"instruction":"最后一步，我们来进一步提升这个科普网站的视觉表现力和交互体验。首先，给左侧的所有滑块加上实时数值反馈，拖动滑块时要在旁边精准显示当前的数值。其次，优化右侧物体的视觉：如果选的是'小球'，请用CSS径向渐变把它画成具有3D立体光泽的球体；如果选的是'羽毛'，请用简单的SVG或特征符号替换原来的圆圈。最重要的是，在它们下落的过程中，请用Canvas或动态DOM在它们背后生成带有透明度渐隐的'运动轨迹拖尾'（残影效果），这样用户就能直观地通过拖尾的疏密程度感受到速度的实时变化。完善这些后，实验就大功告成了！此外，后续所有轮次你都需要始终遵循以下调整及新增要求：第一，修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式；第二，你生成的所有代码文件中，注释内容必须统一使用中文编写。","instruction_following_checklist":[{"constraint":"你的任何用户回复都必须全程使用中文进行生成","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 根据用户指令推断可能生成的代码及日志文件类型\n target_extensions = {'.html', '.css', '.js', '.log'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 utf-8 编码进行解码\n # 如果文件使用了其他编码(如 GBK 等)，这里可能会抛出 UnicodeDecodeError\n content.decode('utf-8')\n except UnicodeDecodeError:\n return False\n except Exception:\n # 捕获其他任何读取或解码异常，直接视为不满足要求\n return False\n \n # 如果找到了至少一个代码文件且全部验证通过，则返回 True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 移除JS代码中的注释和字符串字面量以避免干扰匹配，用等长空格替换以保持原始代码的行号和位置\n def clean_js_code(code: str) -> str:\n result = []\n i = 0\n n = len(code)\n state = 'NORMAL'\n while i < n:\n if state == 'NORMAL':\n if i + 1 < n and code[i:i+2] == '//':\n state = 'LINE_COMMENT'\n result.append(' ')\n i += 2\n elif i + 1 < n and code[i:i+2] == '/*':\n state = 'BLOCK_COMMENT'\n result.append(' ')\n i += 2\n elif code[i] == \"'\":\n state = 'STRING_SINGLE'\n result.append(\"'\")\n i += 1\n elif code[i] == '\"':\n state = 'STRING_DOUBLE'\n result.append('\"')\n i += 1\n elif code[i] == '`':\n state = 'STRING_BACKTICK'\n result.append('`')\n i += 1\n else:\n result.append(code[i])\n i += 1\n elif state == 'LINE_COMMENT':\n if code[i] == '\\n':\n state = 'NORMAL'\n result.append('\\n')\n i += 1\n else:\n result.append(' ')\n i += 1\n elif state == 'BLOCK_COMMENT':\n if i + 1 < n and code[i:i+2] == '*/':\n state = 'NORMAL'\n result.append(' ')\n i += 2\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'STRING_SINGLE':\n if code[i] == '\\\\' and i + 1 < n:\n result.append(' ')\n i += 2\n elif code[i] == \"'\":\n state = 'NORMAL'\n result.append(\"'\")\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'STRING_DOUBLE':\n if code[i] == '\\\\' and i + 1 < n:\n result.append(' ')\n i += 2\n elif code[i] == '\"':\n state = 'NORMAL'\n result.append('\"')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 'STRING_BACKTICK':\n if code[i] == '\\\\' and i + 1 < n:\n result.append(' ')\n i += 2\n elif code[i] == '`':\n state = 'NORMAL'\n result.append('`')\n i += 1\n else:\n if code[i] == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n\n def check_js_code(code: str) -> bool:\n clean_code = clean_js_code(code)\n blocks = []\n \n # 1. 匹配传统的 function 关键字\n for match in re.finditer(r'\\bfunction\\b', clean_code):\n start_idx = match.start()\n brace_idx = clean_code.find('{', start_idx)\n if brace_idx != -1:\n blocks.append((start_idx, brace_idx))\n \n # 2. 匹配箭头函数 => 且带有块级 { } 的形式\n for match in re.finditer(r'=>', clean_code):\n start_idx = match.start()\n i = start_idx + 2\n while i < len(clean_code) and clean_code[i] in ' \\t\\n\\r':\n i += 1\n if i < len(clean_code) and clean_code[i] == '{':\n blocks.append((start_idx, i))\n \n # 3. 匹配对象内声明的方法或者类方法，形如 myFunc(args) { ... }\n exclude_keywords = ('if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else', 'elif', 'with')\n for match in re.finditer(r'\\b([A-Za-z0-9_]+)\\s*\\([^)]*\\)\\s*\\{', clean_code):\n if match.group(1) not in exclude_keywords:\n brace_idx = clean_code.find('{', match.start())\n if brace_idx != -1:\n if not any(b == brace_idx for _, b in blocks):\n blocks.append((match.start(), brace_idx))\n\n # 映射索引到行号\n line_starts = [0]\n for i, c in enumerate(clean_code):\n if c == '\\n':\n line_starts.append(i + 1)\n \n def get_line_num(idx: int) -> int:\n low, high = 0, len(line_starts) - 1\n while low <= high:\n mid = (low + high) // 2\n if line_starts[mid] <= idx:\n if mid == len(line_starts) - 1 or line_starts[mid+1] > idx:\n return mid + 1\n else:\n low = mid + 1\n else:\n high = mid - 1\n return 1\n\n # 计算大括号嵌套匹配，得出该函数的结束索引与行数\n for start_idx, brace_idx in blocks:\n count = 0\n end_idx = -1\n for i in range(brace_idx, len(clean_code)):\n if clean_code[i] == '{':\n count += 1\n elif clean_code[i] == '}':\n count -= 1\n if count == 0:\n end_idx = i\n break\n \n if end_idx != -1:\n start_line = get_line_num(start_idx)\n end_line = get_line_num(end_idx)\n lines = end_line - start_line + 1\n if not (20 <= lines <= 40):\n return False\n \n return True\n\n def check_py_file(content: str) -> bool:\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_line = node.lineno\n end_line = node.end_lineno\n if end_line is not None:\n lines = end_line - start_line + 1\n if not (20 <= lines <= 40):\n return False\n except Exception:\n pass\n return True\n\n def extract_js_from_html(html_code: str) -> list:\n scripts = []\n pattern = re.compile(r'<script.*?>([\\s\\S]*?)</script>', re.IGNORECASE)\n for match in pattern.finditer(html_code):\n scripts.append(match.group(1))\n return scripts\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略常规的依赖与虚拟环境目录\n for d in ['.git', 'node_modules', '__pycache__', 'venv', 'env']:\n if d in dirs:\n dirs.remove(d)\n \n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n if ext == '.py':\n if not check_py_file(content):\n return False\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n if not check_js_code(content):\n return False\n elif ext in ['.html', '.htm']:\n scripts = extract_js_from_html(content)\n for script in scripts:\n if not check_js_code(script):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在生成的代码文件中，严禁使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所生成的各个文件，行数均不得大于200行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n file_checked = False\n \n # 遍历当前工作目录\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，防止遍历如 .git 等不需要的文件\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n try:\n # 尝试以 utf-8 编码读取内容并判断行数\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 如果行数大于 200，则该文件不符合要求\n if len(lines) > 200:\n return False\n \n # 标记成功读取并检查了至少一个有效文本文件\n file_checked = True\n \n except UnicodeDecodeError:\n # 忽略非 utf-8 编码的二进制文件 (例如图像等)\n pass\n except Exception:\n # 忽略因权限等其他原因造成的不可读文件\n pass\n \n # 必须至少存在一个生成的文件，且所有文件行数均不超过200行时才算满足要求\n return file_checked","tags":["Quantity","Range"]},{"constraint":"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 常见的有效代码文件后缀集合\n exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go'}\n \n # 编译匹配模式：优先跳过字符串，进而提取 HTML、块级、单行以及 Python 单行注释\n # 对单双引号限制了不跨行，以避免在HTML或异常语法下将大规模跨行内容当成单一字符串导致漏抓注释\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\\\n])*\"|\\'(?:\\\\.|[^\\'\\\\\\n])*\\'|`(?:\\\\.|[^`\\\\])*`)|'\n r'(?P<html_comment><!--[\\s\\S]*?-->)|'\n r'(?P<block_comment>/\\*[\\s\\S]*?\\*/)|'\n r'(?P<line_comment>(?<!http:)(?<!https:)//[^\\n]*)|'\n r'(?P<py_comment>#[^\\n]*)'\n )\n \n file_count = 0\n for root, dirs, files in os.walk(workspace_path):\n for f in files:\n _, ext = os.path.splitext(f)\n if ext.lower() in exts:\n file_count += 1\n filepath = os.path.join(root, f)\n try:\n with open(filepath, 'r', encoding='utf-8') as file:\n content = file.read()\n except Exception:\n # 如果不能用 UTF-8 读取或发生其他异常，视为未能按照要求生成规范代码文件\n return False\n \n has_valid_comment = False\n # 遍历所有正则匹配结果\n for match in pattern.finditer(content):\n comment_text = None\n is_block = False\n \n # 识别具体匹配的注释类型，并剥离首尾的特殊修饰字符\n if match.group('html_comment'):\n comment_text = match.group('html_comment')[4:-3]\n elif match.group('block_comment'):\n comment_text = match.group('block_comment')[2:-2]\n is_block = True\n elif match.group('line_comment'):\n comment_text = match.group('line_comment')[2:]\n elif match.group('py_comment'):\n comment_text = match.group('py_comment')[1:]\n \n if comment_text is not None:\n # 针对多行文本，将其进行换行切分后对每一行进行校验\n for line in comment_text.split('\\n'):\n clean_line = line.strip()\n if is_block:\n # 对于块级注释中间部分的行，需要去除前导的星号修饰（如 JSDoc 风格）\n clean_line = re.sub(r'^\\*+\\s*', '', clean_line).strip()\n \n # 当此行为具有实际意义的非空行时才进行判定\n if clean_line:\n has_valid_comment = True\n # 使用已实现好的人类专家函数，检验该行是否满足20字的下限要求\n if count_word(clean_line) < 20:\n return False\n \n # 若某个代码文件内未能找到任何有内容的注释，违背了“每个文件里注释详细”的要求\n if not has_valid_comment:\n return False\n \n # 如果项目中没有生成符合目标扩展名的任何代码文件，则不符合要求\n if file_count == 0:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"请务必尽量通过专用工具（如Read, Write, Edit）来完成文件操作，而不是借助 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"我需要用HTML、CSS和JS实现一个科普实验网站，演示'两个物体同时下落'。第一步，先帮我搭一个页面的基础骨架。页面采用左右布局：左侧是参数控制面板，现在只要放几个带边框的空div作为占位符即可；右侧是实验显示区，画一个带有背景色的较高矩形区域代表下落空间。在右侧区域的顶部，放置两个简单的纯色圆形代表准备下落的物体。在页面底部放'开始'和'重置'两个按钮。这一轮暂时不需要实现任何下落动画逻辑，只要页面布局合理、自适应屏幕且稍微美观一点就行。另外，后续所有轮次你都需要始终遵循以下要求：第一，修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它；第二，你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能生成的代码文件扩展名\\n target_extensions = {'.html', '.css', '.js'}\\n has_checked_files = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_checked_files = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果出现解码错误，则说明不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 如果由于其他异常无法读取，视为不满足\\n return False\\n\\n # 若至少存在一个代码文件，且均未引发解码错误，则视为满足要求\\n return has_checked_files\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def clean_code(code: str) -> str:\\n \\\"\\\"\\\"\\n 简单的词法清理：将字符串内内容及注释替换为空格，避免大括号等特殊符号干扰解析，\\n 同时保持原字符串的长度和换行符完全一致。\\n \\\"\\\"\\\"\\n result = []\\n in_single_quote = False\\n in_double_quote = False\\n in_template_quote = False\\n in_line_comment = False\\n in_block_comment = False\\n i = 0\\n n = len(code)\\n while i < n:\\n c = code[i]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n elif in_block_comment:\\n if c == '*' and i + 1 < n and code[i+1] == '/':\\n in_block_comment = False\\n result.append(' ')\\n i += 2\\n else:\\n if c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n elif in_single_quote:\\n if c == '\\\\\\\\' and i + 1 < n:\\n if code[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif c == \\\"'\\\":\\n in_single_quote = False\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n elif in_double_quote:\\n if c == '\\\\\\\\' and i + 1 < n:\\n if code[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif c == '\\\"':\\n in_double_quote = False\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n elif in_template_quote:\\n if c == '\\\\\\\\' and i + 1 < n:\\n if code[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n elif c == '`':\\n in_template_quote = False\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n else:\\n if c == '/' and i + 1 < n and code[i+1] == '/':\\n in_line_comment = True\\n result.append(' ')\\n i += 2\\n elif c == '/' and i + 1 < n and code[i+1] == '*':\\n in_block_comment = True\\n result.append(' ')\\n i += 2\\n elif c == \\\"'\\\":\\n in_single_quote = True\\n result.append(' ')\\n i += 1\\n elif c == '\\\"':\\n in_double_quote = True\\n result.append(' ')\\n i += 1\\n elif c == '`':\\n in_template_quote = True\\n result.append(' ')\\n i += 1\\n else:\\n result.append(c)\\n i += 1\\n return \\\"\\\".join(result)\\n\\n function_lengths = []\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历代码项目\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n js_contents = []\\n if file.endswith('.js'):\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n js_contents.append(f.read())\\n elif file.endswith('.html'):\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 提取全部 <script> 内的内容\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n js_contents.extend(scripts)\\n \\n for js_code in js_contents:\\n if not js_code.strip():\\n continue\\n # 获取剔除干扰后的“干净代码面具”\\n js_code_clean = clean_code(js_code)\\n \\n # 寻找各类函数的起始特征\\n matches = list(re.finditer(r'\\\\bfunction\\\\b[^{]*\\\\{', js_code_clean))\\n matches += list(re.finditer(r'=>\\\\s*\\\\{', js_code_clean))\\n method_pattern = r'\\\\b([a-zA-Z_$][\\\\w_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n for m in re.finditer(method_pattern, js_code_clean):\\n keyword = m.group(1)\\n # 排除控制流关键字的误判\\n if keyword not in ['if', 'for', 'while', 'catch', 'switch', 'with', 'function']:\\n matches.append(m)\\n \\n # 去重重叠的匹配（共享同一个左大括号的匹配项视为同一个函数起点）\\n starts = set()\\n valid_matches = []\\n for m in matches:\\n brace_idx = m.end() - 1\\n if brace_idx not in starts:\\n starts.add(brace_idx)\\n valid_matches.append(m)\\n \\n # 从匹配处往后解析匹配对应大括号提取函数体范围\\n for m in valid_matches:\\n start_index = m.start()\\n brace_index = m.end() - 1\\n \\n stack_count = 1\\n j = brace_index + 1\\n while j < len(js_code_clean) and stack_count > 0:\\n if js_code_clean[j] == '{':\\n stack_count += 1\\n elif js_code_clean[j] == '}':\\n stack_count -= 1\\n j += 1\\n \\n if stack_count == 0:\\n end_index = j\\n # 使用原代码切片以获取精准包含的代码段\\n func_body = js_code[start_index:end_index]\\n line_count = func_body.count('\\\\n') + 1\\n function_lengths.append(line_count)\\n\\n # 验证是否任意单个函数的行数均在20-40（闭区间）内\\n for length in function_lengths:\\n if length < 20 or length > 40:\\n return False\\n \\n # 若全部符合要求，或尚无任何函数被创建，都视为符合规则\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，严禁使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的各个文件，行数均不得大于200行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据用户指令，推断可能的代码文件类型\\n target_extensions = {'.html', '.css', '.js'}\\n\\n # 遍历工作区内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 获取文件后缀名\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 仅检查目标类型的文件\\n if ext in target_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n # 读取文件行数\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 判断行数是否大于200行\\n if len(lines) > 200:\\n return False\\n except Exception:\\n # 如果由于非预期原因（如编码问题等）无法读取，为了鲁棒性暂且忽略该不可读文件\\n pass\\n\\n # 如果所有被检查的目标文件行数均未超过200行，则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 网站相关的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return False\\n \\n comments_lines = []\\n \\n # HTML 注释提取 (<!-- ... -->)\\n if ext == '.html':\\n for m in re.finditer(r'<!--(.*?)-->', content, re.DOTALL):\\n comments_lines.extend(m.group(1).split('\\\\n'))\\n \\n # 块级注释提取 (/* ... */) 适用于 HTML, CSS, JS\\n if ext in {'.html', '.css', '.js'}:\\n for m in re.finditer(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL):\\n comments_lines.extend(m.group(1).split('\\\\n'))\\n \\n # 单行注释提取 (// ...) 适用于 HTML, JS\\n if ext in {'.html', '.js'}:\\n # (?<!:) 用于防止匹配 http:// 等 URL 结构\\n for m in re.finditer(r'(?<!:)\\\\/\\\\/(.*)', content):\\n comments_lines.append(m.group(1))\\n \\n file_has_valid_comment = False\\n \\n for line in comments_lines:\\n text = line.strip()\\n if not text:\\n continue\\n \\n # 多行注释的每行常以 * 开头，需将其剥离以保留纯文本\\n if text.startswith('*'):\\n text = text.lstrip('*').strip()\\n \\n if not text:\\n continue\\n \\n file_has_valid_comment = True\\n \\n # 验证每行注释的字数是否 >= 20\\n if count_word(text) < 20:\\n return False\\n \\n # 必须保证每个代码文件都包含注释\\n if not file_has_valid_comment:\\n return False\\n \\n # 如果遍历完成且至少检查过一个代码文件，则要求满足\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来我们把左侧的控制面板完善一下，并加上基础的触发机制。请用真实的表单元素替换掉之前的占位符，表单需要包含：物体1类型（下拉选择：小球/羽毛）、物体1重量（滑块：1-100）、物体2类型、物体2重量、是否为真空环境（开关或复选框）、空气阻力系数（滑块）、重力加速度（滑块，默认9.8）。然后，给底部的按钮加上功能：点击'开始'后，右侧的两个圆圈以一个固定的像素值向下匀速移动，碰到容器底部停止；点击'重置'让它们回到顶部。暂时不用管真实的物理公式，只要表单能用、物体能动就行。此外，后续所有轮次你都需要始终遵循以下新增要求：第一，在更新已有文件内容时，必须输出严格的Unified Diff格式，不允许直接输出全量代码替换；第二，务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目描述推断可能的代码文件和文本文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.log'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试将文件内容按照 utf-8 格式解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明不是合法的 utf-8 编码\\n return False\\n except Exception:\\n # 其他读取异常等保守起见视为验证不通过\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内\",\n \"validation_code\": \"def verify_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 遍历工作区目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html'):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于某种原因遇到读取错误则跳过该文件\\n continue\\n \\n codes_to_parse = []\\n if file.endswith('.js'):\\n codes_to_parse.append(content)\\n else:\\n # 对于HTML文件，只提取 <script> 标签内的内容\\n pattern = re.compile(r'<script.*?>([\\\\s\\\\S]*?)</script>', re.IGNORECASE)\\n for match in pattern.finditer(content):\\n codes_to_parse.append(match.group(1))\\n \\n for code in codes_to_parse:\\n n = len(code)\\n sanitized_code = list(code)\\n i = 0\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n # 第一阶段：剥离字符串与注释，将其替换为空格，保留换行符\\n while i < n:\\n c = code[i]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n else:\\n sanitized_code[i] = ' '\\n i += 1\\n continue\\n \\n if in_block_comment:\\n if c == '*' and i + 1 < n and code[i+1] == '/':\\n sanitized_code[i] = ' '\\n sanitized_code[i+1] = ' '\\n in_block_comment = False\\n i += 2\\n else:\\n if c != '\\\\n':\\n sanitized_code[i] = ' '\\n i += 1\\n continue\\n \\n if in_string:\\n if c == '\\\\\\\\':\\n sanitized_code[i] = ' '\\n if i + 1 < n:\\n if sanitized_code[i+1] != '\\\\n':\\n sanitized_code[i+1] = ' '\\n i += 2\\n elif c == string_char:\\n in_string = False\\n sanitized_code[i] = ' '\\n i += 1\\n else:\\n if c != '\\\\n':\\n sanitized_code[i] = ' '\\n i += 1\\n continue\\n \\n if c in (\\\"'\\\", '\\\"', '`'):\\n in_string = True\\n string_char = c\\n sanitized_code[i] = ' '\\n i += 1\\n continue\\n \\n if c == '/':\\n if i + 1 < n and code[i+1] == '/':\\n in_line_comment = True\\n sanitized_code[i] = ' '\\n sanitized_code[i+1] = ' '\\n i += 2\\n continue\\n elif i + 1 < n and code[i+1] == '*':\\n in_block_comment = True\\n sanitized_code[i] = ' '\\n sanitized_code[i+1] = ' '\\n i += 2\\n continue\\n \\n i += 1\\n\\n sanitized = \\\"\\\".join(sanitized_code)\\n \\n function_starts = []\\n \\n # 第二阶段：寻找所有可能的函数起始标识\\n # 1. 查找 function 关键字\\n idx = 0\\n while True:\\n idx = sanitized.find('function', idx)\\n if idx == -1:\\n break\\n # 确保是一个独立的单词而不是变量名的后缀/前缀\\n if (idx == 0 or not (sanitized[idx-1].isalnum() or sanitized[idx-1] == '_')) and \\\\\\n (idx + 8 >= len(sanitized) or not (sanitized[idx+8].isalnum() or sanitized[idx+8] == '_')):\\n function_starts.append(idx)\\n idx += 8\\n \\n # 2. 查找箭头函数 =>\\n idx = 0\\n while True:\\n idx = sanitized.find('=>', idx)\\n if idx == -1:\\n break\\n function_starts.append(idx)\\n idx += 2\\n \\n # 3. 查找类方法/对象简写方法 (如 myMethod(arg) { )\\n method_pattern = re.compile(r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n for match in method_pattern.finditer(sanitized):\\n name = match.group(1)\\n if name not in ('if', 'for', 'while', 'catch', 'switch', 'function', 'return', 'with'):\\n function_starts.append(match.start())\\n \\n function_starts = sorted(list(set(function_starts)))\\n \\n # 过滤重叠定义，建立匹配大括号的关联\\n brace_to_start = {}\\n for start_idx in function_starts:\\n brace_idx = sanitized.find('{', start_idx)\\n if brace_idx == -1:\\n continue\\n \\n # 如果标记到最近大括号间有严重中断符(;)，可能属误识别，忽略\\n between = sanitized[start_idx:brace_idx]\\n if ';' in between:\\n continue\\n \\n # 确保 => 紧连着大括号（允许换行/空格），防止捕获隐式返回的箭头函数后的无关大括号\\n if sanitized[start_idx:start_idx+2] == '=>':\\n between_arrow = sanitized[start_idx+2:brace_idx]\\n if not between_arrow.isspace() and between_arrow != \\\"\\\":\\n continue\\n \\n if brace_idx not in brace_to_start:\\n brace_to_start[brace_idx] = start_idx\\n else:\\n if start_idx < brace_to_start[brace_idx]:\\n brace_to_start[brace_idx] = start_idx\\n \\n # 第三阶段：解析范围及计算行数\\n for brace_idx, start_idx in brace_to_start.items():\\n true_start = start_idx\\n # 对箭头函数，尝试向左寻找包裹参数的 '(' 以计算完整行数\\n if sanitized[start_idx:start_idx+2] == '=>':\\n left_ptr = start_idx - 1\\n while left_ptr >= 0 and sanitized[left_ptr].isspace():\\n left_ptr -= 1\\n if left_ptr >= 0 and sanitized[left_ptr] == ')':\\n count = 1\\n curr = left_ptr - 1\\n while curr >= 0:\\n if sanitized[curr] == ')':\\n count += 1\\n elif sanitized[curr] == '(':\\n count -= 1\\n if count == 0:\\n true_start = curr\\n break\\n curr -= 1\\n \\n # 寻找匹配的结束大括号\\n count = 1\\n end_idx = -1\\n for j in range(brace_idx + 1, len(sanitized)):\\n if sanitized[j] == '{':\\n count += 1\\n elif sanitized[j] == '}':\\n count -= 1\\n if count == 0:\\n end_idx = j\\n break\\n \\n if end_idx != -1:\\n # 统计真实起点到终止大括号之间换行符的数量，获取函数代码块所占行数\\n start_line = sanitized.count('\\\\n', 0, true_start) + 1\\n end_line = sanitized.count('\\\\n', 0, end_idx) + 1\\n line_count = end_line - start_line + 1\\n \\n # 如果函数的行数不满足限定在20-40行的标准，返回 False\\n if line_count < 20 or line_count > 40:\\n return False\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，严禁使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的各个文件，行数均不得大于200行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在且为目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据用户指令推断可能生成的文件类型\\n valid_extensions = {'.html', '.css', '.js', '.log'}\\n files_checked = False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n files_checked = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 统计文件的行数\\n line_count = sum(1 for _ in f)\\n \\n # 检查行数是否大于200行\\n if line_count > 200:\\n return False\\n except Exception:\\n # 如果遇到文件读取等异常，跳过处理\\n continue\\n\\n # 至少检查了一个相关文件且没有出现大于200行的情况才视为符合\\n return files_checked\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字\",\n \"validation_code\": \"def check_comment_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n valid_exts = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 使用正则表达式提取文件中的各类注释\\n # 1. 匹配 HTML 注释：<!-- ... -->\\n html_comments = re.findall(r'<!--([\\\\s\\\\S]*?)-->', content)\\n # 2. 匹配 CSS/JS 多行注释：/* ... */\\n multiline_comments = re.findall(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', content)\\n # 3. 匹配 JS 单行注释：// ...（仅匹配行首或在空白字符之后的//，避开网址中的//）\\n singleline_comments = re.findall(r'(?:^|\\\\s)//(.*)', content)\\n \\n valid_comment_lines = 0\\n \\n # 检查 HTML 注释的内容\\n for comment in html_comments:\\n for line in comment.split('\\\\n'):\\n clean_line = line.strip()\\n if clean_line:\\n valid_comment_lines += 1\\n if count_word(clean_line) < 20:\\n return False\\n \\n # 检查 CSS/JS 多行注释的内容\\n for comment in multiline_comments:\\n for line in comment.split('\\\\n'):\\n # 去除开头的 '*' 符号（多行注释换行经常带有装饰性星号），再去除两端空白\\n clean_line = line.strip().lstrip('*').strip()\\n if clean_line:\\n valid_comment_lines += 1\\n if count_word(clean_line) < 20:\\n return False\\n \\n # 检查 JS 单行注释的内容\\n for comment in singleline_comments:\\n clean_line = comment.strip()\\n if clean_line:\\n valid_comment_lines += 1\\n if count_word(clean_line) < 20:\\n return False\\n \\n # 每个代码文件必须至少包含一行有效注释\\n if valid_comment_lines == 0:\\n return False\\n \\n # 如果项目中没有任何代码文件，也判定为不满足\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在更新已有文件内容时，必须输出严格的Unified Diff格式，不允许直接输出全量代码替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"基础交互做得很好了。现在我们要把核心的物理引擎加进去，让这两个物体的下落真正遵循物理规律。请修改JS里的动画计算逻辑：使用 RequestAnimationFrame 重构循环。在每一帧中，利用牛顿第二定律计算加速度(a = (mg - f)/m)。其中重力G=mg，空气阻力设定为f=k*v（k为表单中的风阻系数，v为当前下落速度）。注意：如果勾选了'真空环境'，则阻力f恒为0；另外，下拉框选择'羽毛'时，请在代码里给它加上一个阻力放大系数，模拟其迎风面积大的特点。修改完成后，点击开始，非真空下应该能明显看到羽毛下落比小球慢，而真空中两者应该完全同时落地。另外，后续所有轮次你都需要始终遵循以下调整后的要求：第一，只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式；第二，进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令推断，主要生成的代码文件包括 HTML, CSS, JS\\n code_extensions = {'.html', '.css', '.js'}\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，则文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 如果发生其他读取错误，保守判定为不满足\\n return False\\n\\n # 必须至少找到一个代码文件且全部验证通过，才算真正满足了要求\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 屏蔽代码中的字符串和注释，替换为空格但保留换行符，以防止误匹配符号并保持原行数不变\\n def mask_code(c: str) -> str:\\n result = list(c)\\n i = 0\\n n = len(c)\\n while i < n:\\n if c[i:i+2] == '//':\\n j = i\\n while j < n and c[j] != '\\\\n':\\n result[j] = ' '\\n j += 1\\n i = j\\n elif c[i:i+2] == '/*':\\n j = i\\n while j < n and c[j:j+2] != '*/':\\n if result[j] != '\\\\n':\\n result[j] = ' '\\n j += 1\\n if j < n:\\n result[j] = ' '\\n result[j+1] = ' '\\n j += 2\\n i = j\\n elif c[i] in [\\\"'\\\", '\\\"', \\\"`\\\"]:\\n quote = c[i]\\n result[i] = ' '\\n j = i + 1\\n while j < n and c[j] != quote:\\n if c[j] == '\\\\\\\\':\\n result[j] = ' '\\n j += 1\\n if j < n and result[j] != '\\\\n':\\n result[j] = ' '\\n else:\\n if result[j] != '\\\\n':\\n result[j] = ' '\\n j += 1\\n if j < n:\\n result[j] = ' '\\n j += 1\\n i = j\\n else:\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def extract_and_check(code: str) -> bool:\\n masked = mask_code(code)\\n n = len(masked)\\n \\n for i in range(n):\\n if masked[i] == '{':\\n is_func = False\\n start_pos = i\\n \\n j = i - 1\\n while j >= 0 and masked[j].isspace():\\n j -= 1\\n if j < 0:\\n continue\\n \\n # 箭头函数的 =>\\n if masked[j] == '>':\\n if j >= 1 and masked[j-1] == '=':\\n is_func = True\\n start_pos = j - 1\\n k = start_pos - 1\\n while k >= 0 and masked[k].isspace(): k -= 1\\n if k >= 0 and masked[k] == ')':\\n paren_c = 1\\n k -= 1\\n while k >= 0 and paren_c > 0:\\n if masked[k] == ')': paren_c += 1\\n elif masked[k] == '(': paren_c -= 1\\n k -= 1\\n start_pos = k + 1\\n else:\\n while k >= 0 and (masked[k].isalnum() or masked[k] in '_$'): k -= 1\\n start_pos = k + 1\\n \\n # 普通函数或方法的参数 )\\n elif masked[j] == ')':\\n paren_c = 1\\n k = j - 1\\n while k >= 0 and paren_c > 0:\\n if masked[k] == ')': paren_c += 1\\n elif masked[k] == '(': paren_c -= 1\\n k -= 1\\n \\n if k >= 0:\\n start_pos = k + 1\\n p = k - 1\\n while p >= 0 and masked[p].isspace(): p -= 1\\n \\n if p >= 0:\\n if masked[p].isalnum() or masked[p] in '_$':\\n q = p\\n while q >= 0 and (masked[q].isalnum() or masked[q] in '_$'): q -= 1\\n word = masked[q+1:p+1]\\n \\n if word == 'function':\\n is_func = True\\n start_pos = q + 1\\n # 排查非函数控制流语句块\\n elif word not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n w2_p = q\\n while w2_p >= 0 and masked[w2_p].isspace(): w2_p -= 1\\n w2_q = w2_p\\n while w2_q >= 0 and (masked[w2_q].isalnum() or masked[w2_q] in '_$'): w2_q -= 1\\n word2 = masked[w2_q+1:w2_p+1] if w2_q < w2_p else \\\"\\\"\\n \\n if word2 == 'function':\\n is_func = True\\n start_pos = w2_q + 1\\n else:\\n is_func = True\\n if word2 in ['get', 'set']:\\n start_pos = w2_q + 1\\n else:\\n start_pos = q + 1\\n else:\\n # 对于动态命名或未定名的函数形式，只要以)结束，极大可能为合法的函数语法树\\n is_func = True\\n \\n # 如果定位到是函数，向后匹配完整的 {} 区域并计算行数\\n if is_func:\\n brace_c = 1\\n end_pos = -1\\n for end_i in range(i + 1, n):\\n if masked[end_i] == '{': brace_c += 1\\n elif masked[end_i] == '}':\\n brace_c -= 1\\n if brace_c == 0:\\n end_pos = end_i\\n break\\n \\n if end_pos != -1:\\n start_line = code.count('\\\\n', 0, start_pos) + 1\\n end_line = code.count('\\\\n', 0, end_pos) + 1\\n lines = end_line - start_line + 1\\n \\n # 检查行数是否在20-40之间\\n if lines < 20 or lines > 40:\\n return False\\n \\n return True\\n\\n # 遍历 workspace_path 中生成的 js 与 html 文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html'):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if file.endswith('.js'):\\n if not extract_and_check(content):\\n return False\\n elif file.endswith('.html'):\\n # 匹配内联的 JavaScript 脚本块\\n blocks = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for block in blocks:\\n if not extract_and_check(block):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，严禁使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的各个文件，行数均不得大于200行\",\n \"validation_code\": \"def check_file_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 根据用户指令推断可能的代码文件类型（HTML、CSS、JS、LOG）\\n target_extensions = ('.html', '.css', '.js', '.log')\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if file_name.endswith(target_extensions):\\n file_path = os.path.join(root, file_name)\\n try:\\n # 系统提示明确要求使用 UTF-8 进行编码\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n # 判断行数是否大于200\\n if len(lines) > 200:\\n return False\\n except Exception:\\n # 忽略读取或编码异常的文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字\",\n \"validation_code\": \"def evaluate(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def clean_strings(text: str) -> str:\\n # 移除代码中双引号、单引号和反引号包裹的字符串内容，避免内部的 // 或 /* 被误认为注释\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', text)\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', text)\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', text)\\n return text\\n\\n # 推断可能的代码文件类型\\n code_extensions = ['.html', '.htm', '.css', '.js']\\n code_files = []\\n \\n # 查找工作区下的所有目标代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.lower().endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n if not code_files:\\n return False\\n \\n # 逐一检查每个代码文件\\n for filepath in code_files:\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return False\\n \\n ext = os.path.splitext(filepath)[1].lower()\\n comments = []\\n \\n if ext in ['.html', '.htm']:\\n # 提取 HTML 注释\\n comments.extend(re.findall(r'<!--([\\\\s\\\\S]*?)-->', content))\\n \\n # 提取内部的 script 和 style 内容处理内部 JS 和 CSS 注释\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n styles = re.findall(r'<style[^>]*>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n js_css_content = '\\\\n'.join(scripts + styles)\\n \\n if js_css_content:\\n js_css_content = clean_strings(js_css_content)\\n comments.extend(re.findall(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', js_css_content))\\n # 使用后向否定断言避免匹配 http:// 和 https:// 中的双斜杠\\n comments.extend(re.findall(r'(?<!:)//(.*)', js_css_content))\\n \\n elif ext in ['.js', '.css']:\\n content_cleaned = clean_strings(content)\\n comments.extend(re.findall(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', content_cleaned))\\n comments.extend(re.findall(r'(?<!:)//(.*)', content_cleaned))\\n \\n # 若文件中没有任何注释提取出来，说明不满足“每个生成的代码文件里注释详细”的要求\\n if not comments:\\n return False\\n \\n has_valid_comment = False\\n \\n for c in comments:\\n lines = c.split('\\\\n')\\n for line in lines:\\n line = line.strip()\\n # 过滤掉多行注释里常规的前缀符号\\n if line.startswith('*'):\\n line = line[1:].strip()\\n \\n # 忽略完全为空的行\\n if not line:\\n continue\\n \\n has_valid_comment = True\\n \\n # 验证每行有效注释的字数是否满足下限要求\\n if count_word(line) < 20:\\n return False\\n \\n # 若注释全部是空行等无效内容，也不达标\\n if not has_valid_comment:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"只要是对已有文件进行更改，就要求输出该文件的全量代码以供覆盖替换，切勿使用 Unified Diff 或是其他补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"物理引擎看起来生效了，非常棒！下一步，我们需要给实验加上精确的数据统计和结果展示功能。请在右侧实验区的顶端中间加上一个'高精度计时器'，点击开始后从 0.00 秒开始实时滚动更新。当任意一个物体碰到地面（下边界）时，瞬间记录下它的落地时间并在其底部显示出来。当两个物体都触底后，在实验区中央弹出一个半透明的实验报告面板，清晰地对比：物体1用时多少，物体2用时多少，并根据时间差自动生成一句简单的物理结论（例如'真空中物体下落加速度与质量无关'或'受空气阻力影响，物体落地时间产生差异'）。此外，后续所有轮次你都需要始终遵循以下调整及新增要求：第一，只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换；第二，请务必尽量通过专用工具（如Read, Write, Edit）来完成文件操作，而不是借助 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断，科普实验网站项目主要包含以下类型的前端代码文件\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试使用 utf-8 进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 抛出 UnicodeDecodeError 说明该文件含有非 UTF-8 编码的字符\\n return False\\n except Exception:\\n # 发生其他读取异常也视为校验失败\\n return False\\n\\n # 若项目中至少存在一个代码文件且全部通过了 UTF-8 解码测试，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内\",\n \"validation_code\": \"def check_function_lines(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def mask_code(code: str) -> str:\\n \\\"\\\"\\\"屏蔽字符串和注释，将它们替换为等长的空格，以保证后续匹配的安全性\\\"\\\"\\\"\\n masked = list(code)\\n code_len = len(code)\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n i = 0\\n while i < code_len:\\n char = code[i]\\n if not in_string and not in_line_comment and not in_block_comment:\\n if char == '/' and i + 1 < code_len:\\n if code[i+1] == '/':\\n in_line_comment = True\\n masked[i] = ' '\\n masked[i+1] = ' '\\n i += 1\\n elif code[i+1] == '*':\\n in_block_comment = True\\n masked[i] = ' '\\n masked[i+1] = ' '\\n i += 1\\n elif char in [\\\"'\\\", '\\\"', \\\"`\\\"]:\\n in_string = True\\n string_char = char\\n masked[i] = ' '\\n elif in_string:\\n masked[i] = ' '\\n # 处理转义字符\\n if char == '\\\\\\\\':\\n if i + 1 < code_len:\\n masked[i+1] = ' '\\n i += 1\\n elif char == string_char:\\n in_string = False\\n elif in_line_comment:\\n if char == '\\\\n':\\n in_line_comment = False\\n else:\\n masked[i] = ' '\\n elif in_block_comment:\\n masked[i] = ' '\\n if char == '*' and i + 1 < code_len and code[i+1] == '/':\\n masked[i+1] = ' '\\n in_block_comment = False\\n i += 1\\n i += 1\\n return \\\"\\\".join(masked)\\n\\n def get_functions_from_js(code: str):\\n \\\"\\\"\\\"利用正则和括号匹配从JS代码中提取完整的函数体代码块\\\"\\\"\\\"\\n masked = mask_code(code)\\n funcs = []\\n \\n # 特判：如果不带大括号的箭头函数存在，其行数必然极少（一定低于20行），此处加入一个违规的占位伪代码即可使后续检查返回False\\n if re.search(r'=>\\\\s*(?:[^{]|$)', masked):\\n funcs.append(\\\"() => 1;\\\")\\n \\n code_len = len(masked)\\n idx = 0\\n \\n # 匹配三种常见的JS函数起始声明形式\\n pattern = re.compile(\\n r'\\\\bfunction\\\\b[^{]*\\\\{|'\\n r'(?:\\\\b[a-zA-Z_]\\\\w*\\\\s*=>\\\\s*\\\\{|\\\\([^)]*\\\\)\\\\s*=>\\\\s*\\\\{)|'\\n r'\\\\b(?!(?:if|for|while|switch|catch|with|return)\\\\b)[a-zA-Z_]\\\\w*\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n )\\n \\n while idx < code_len:\\n match = pattern.search(masked, idx)\\n if not match:\\n break\\n \\n absolute_match_start = match.start()\\n start_idx = match.end() - 1 # 找到的大括号 '{' 的下标\\n \\n brace_count = 0\\n end_idx = -1\\n \\n i = start_idx\\n while i < code_len:\\n char = masked[i]\\n if char == '{':\\n brace_count += 1\\n elif char == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_idx = i\\n break\\n i += 1\\n \\n if end_idx != -1:\\n # 从原始原代码片段截取真实代码\\n func_code = code[absolute_match_start : end_idx + 1]\\n funcs.append(func_code)\\n # 为了能够捕捉到嵌套在内部的子函数，游标只向前推进到大括号内侧\\n idx = start_idx + 1\\n else:\\n idx = match.end()\\n \\n return funcs\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n js_codes = []\\n py_files = []\\n \\n # 遍历并收集代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n js_codes.append(f.read())\\n except:\\n pass\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n html_code = f.read()\\n # 从HTML中提取出所有的 <script> 标签内联代码\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', html_code, re.IGNORECASE | re.DOTALL)\\n js_codes.extend(scripts)\\n except:\\n pass\\n elif file.endswith('.py'):\\n py_files.append(file_path)\\n\\n all_funcs = []\\n for code in js_codes:\\n all_funcs.extend(get_functions_from_js(code))\\n \\n # 验证提取出的所有JS函数行数\\n for func in all_funcs:\\n lines = func.strip().split('\\\\n')\\n if not (20 <= len(lines) <= 40):\\n return False\\n \\n # 验证提取出的所有Python函数行数\\n for py_file in py_files:\\n try:\\n with open(py_file, 'r', encoding='utf-8') as f:\\n code = f.read()\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n end = getattr(node, 'end_lineno', start)\\n num_lines = end - start + 1\\n if not (20 <= num_lines <= 40):\\n return False\\n except:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，严禁使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的各个文件，行数均不得大于200行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，视作没有违反行数条件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉可能的版本控制相关隐藏文件夹，如.git\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 逐行统计文件行数，这样也能节省大文件的内存开销\\n line_count = sum(1 for _ in f)\\n \\n # 只要有任何一个文件的行数大于200，就不满足要求\\n if line_count > 200:\\n return False\\n except Exception:\\n # 忽略因非UTF-8编码或其他不可读原因导致的异常文件（例如二进制文件）\\n pass\\n\\n # 遍历完全部文件后，没有发现超过200行的文件\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 确保给定的项目路径存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n code_extensions = ('.html', '.htm', '.css', '.js')\\n \\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if not file.endswith(code_extensions):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n \\n # 根据文件类型用对应的正则表达式提取注释\\n if file.endswith(('.html', '.htm')):\\n # 匹配 HTML 注释\\n for match in re.finditer(r'<!--([\\\\s\\\\S]*?)-->', content):\\n comments.extend(match.group(1).split('\\\\n'))\\n \\n # 提取 <script> 内的 js 注释\\n for match in re.finditer(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, flags=re.IGNORECASE):\\n script_content = match.group(1)\\n # js 单行注释 (使用(?<!:)避免匹配到 URL 的协议部分如 http://)\\n for m in re.finditer(r'(?<!:)\\\\/\\\\/(.*)', script_content):\\n comments.append(m.group(1))\\n # js 多行注释\\n for m in re.finditer(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', script_content):\\n comments.extend(m.group(1).split('\\\\n'))\\n \\n # 提取 <style> 内的 css 注释\\n for match in re.finditer(r'<style[^>]*>([\\\\s\\\\S]*?)</style>', content, flags=re.IGNORECASE):\\n style_content = match.group(1)\\n for m in re.finditer(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', style_content):\\n comments.extend(m.group(1).split('\\\\n'))\\n \\n elif file.endswith('.js'):\\n # JS 的单行和多行注释\\n for m in re.finditer(r'(?<!:)\\\\/\\\\/(.*)', content):\\n comments.append(m.group(1))\\n for m in re.finditer(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', content):\\n comments.extend(m.group(1).split('\\\\n'))\\n \\n elif file.endswith('.css'):\\n # CSS 的多行注释\\n for m in re.finditer(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', content):\\n comments.extend(m.group(1).split('\\\\n'))\\n \\n valid_comment_count = 0\\n \\n # 校验每一行提取出来的注释\\n for line in comments:\\n clean_line = line.strip()\\n # 移除多行注释或者装饰性注释常带的前导无意义字符（如 * 或 -）\\n clean_line = re.sub(r'^[\\\\*\\\\-\\\\/]+', '', clean_line).strip()\\n clean_line = clean_line.strip('*').strip()\\n \\n # 跳过空行\\n if not clean_line:\\n continue\\n \\n valid_comment_count += 1\\n \\n # 检查单行字数下限\\n if count_word(clean_line) < 20:\\n return False\\n \\n # 必须保证每个生成的代码文件里都有至少一条有效注释以满足\\\"注释详细\\\"的要求\\n if valid_comment_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请务必尽量通过专用工具（如Read, Write, Edit）来完成文件操作，而不是借助 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后一步，我们来进一步提升这个科普网站的视觉表现力和交互体验。首先，给左侧的所有滑块加上实时数值反馈，拖动滑块时要在旁边精准显示当前的数值。其次，优化右侧物体的视觉：如果选的是'小球'，请用CSS径向渐变把它画成具有3D立体光泽的球体；如果选的是'羽毛'，请用简单的SVG或特征符号替换原来的圆圈。最重要的是，在它们下落的过程中，请用Canvas或动态DOM在它们背后生成带有透明度渐隐的'运动轨迹拖尾'（残影效果），这样用户就能直观地通过拖尾的疏密程度感受到速度的实时变化。完善这些后，实验就大功告成了！此外，后续所有轮次你都需要始终遵循以下调整及新增要求：第一，修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式；第二，你生成的所有代码文件中，注释内容必须统一使用中文编写。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你的任何用户回复都必须全程使用中文进行生成\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 根据用户指令推断可能生成的代码及日志文件类型\\n target_extensions = {'.html', '.css', '.js', '.log'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 utf-8 编码进行解码\\n # 如果文件使用了其他编码(如 GBK 等)，这里可能会抛出 UnicodeDecodeError\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n return False\\n except Exception:\\n # 捕获其他任何读取或解码异常，直接视为不满足要求\\n return False\\n \\n # 如果找到了至少一个代码文件且全部验证通过，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，任何单个函数的行数都需限定在20-40的标准内\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 移除JS代码中的注释和字符串字面量以避免干扰匹配，用等长空格替换以保持原始代码的行号和位置\\n def clean_js_code(code: str) -> str:\\n result = []\\n i = 0\\n n = len(code)\\n state = 'NORMAL'\\n while i < n:\\n if state == 'NORMAL':\\n if i + 1 < n and code[i:i+2] == '//':\\n state = 'LINE_COMMENT'\\n result.append(' ')\\n i += 2\\n elif i + 1 < n and code[i:i+2] == '/*':\\n state = 'BLOCK_COMMENT'\\n result.append(' ')\\n i += 2\\n elif code[i] == \\\"'\\\":\\n state = 'STRING_SINGLE'\\n result.append(\\\"'\\\")\\n i += 1\\n elif code[i] == '\\\"':\\n state = 'STRING_DOUBLE'\\n result.append('\\\"')\\n i += 1\\n elif code[i] == '`':\\n state = 'STRING_BACKTICK'\\n result.append('`')\\n i += 1\\n else:\\n result.append(code[i])\\n i += 1\\n elif state == 'LINE_COMMENT':\\n if code[i] == '\\\\n':\\n state = 'NORMAL'\\n result.append('\\\\n')\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'BLOCK_COMMENT':\\n if i + 1 < n and code[i:i+2] == '*/':\\n state = 'NORMAL'\\n result.append(' ')\\n i += 2\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'STRING_SINGLE':\\n if code[i] == '\\\\\\\\' and i + 1 < n:\\n result.append(' ')\\n i += 2\\n elif code[i] == \\\"'\\\":\\n state = 'NORMAL'\\n result.append(\\\"'\\\")\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'STRING_DOUBLE':\\n if code[i] == '\\\\\\\\' and i + 1 < n:\\n result.append(' ')\\n i += 2\\n elif code[i] == '\\\"':\\n state = 'NORMAL'\\n result.append('\\\"')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 'STRING_BACKTICK':\\n if code[i] == '\\\\\\\\' and i + 1 < n:\\n result.append(' ')\\n i += 2\\n elif code[i] == '`':\\n state = 'NORMAL'\\n result.append('`')\\n i += 1\\n else:\\n if code[i] == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def check_js_code(code: str) -> bool:\\n clean_code = clean_js_code(code)\\n blocks = []\\n \\n # 1. 匹配传统的 function 关键字\\n for match in re.finditer(r'\\\\bfunction\\\\b', clean_code):\\n start_idx = match.start()\\n brace_idx = clean_code.find('{', start_idx)\\n if brace_idx != -1:\\n blocks.append((start_idx, brace_idx))\\n \\n # 2. 匹配箭头函数 => 且带有块级 { } 的形式\\n for match in re.finditer(r'=>', clean_code):\\n start_idx = match.start()\\n i = start_idx + 2\\n while i < len(clean_code) and clean_code[i] in ' \\\\t\\\\n\\\\r':\\n i += 1\\n if i < len(clean_code) and clean_code[i] == '{':\\n blocks.append((start_idx, i))\\n \\n # 3. 匹配对象内声明的方法或者类方法，形如 myFunc(args) { ... }\\n exclude_keywords = ('if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'else', 'elif', 'with')\\n for match in re.finditer(r'\\\\b([A-Za-z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', clean_code):\\n if match.group(1) not in exclude_keywords:\\n brace_idx = clean_code.find('{', match.start())\\n if brace_idx != -1:\\n if not any(b == brace_idx for _, b in blocks):\\n blocks.append((match.start(), brace_idx))\\n\\n # 映射索引到行号\\n line_starts = [0]\\n for i, c in enumerate(clean_code):\\n if c == '\\\\n':\\n line_starts.append(i + 1)\\n \\n def get_line_num(idx: int) -> int:\\n low, high = 0, len(line_starts) - 1\\n while low <= high:\\n mid = (low + high) // 2\\n if line_starts[mid] <= idx:\\n if mid == len(line_starts) - 1 or line_starts[mid+1] > idx:\\n return mid + 1\\n else:\\n low = mid + 1\\n else:\\n high = mid - 1\\n return 1\\n\\n # 计算大括号嵌套匹配，得出该函数的结束索引与行数\\n for start_idx, brace_idx in blocks:\\n count = 0\\n end_idx = -1\\n for i in range(brace_idx, len(clean_code)):\\n if clean_code[i] == '{':\\n count += 1\\n elif clean_code[i] == '}':\\n count -= 1\\n if count == 0:\\n end_idx = i\\n break\\n \\n if end_idx != -1:\\n start_line = get_line_num(start_idx)\\n end_line = get_line_num(end_idx)\\n lines = end_line - start_line + 1\\n if not (20 <= lines <= 40):\\n return False\\n \\n return True\\n\\n def check_py_file(content: str) -> bool:\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_line = node.lineno\\n end_line = node.end_lineno\\n if end_line is not None:\\n lines = end_line - start_line + 1\\n if not (20 <= lines <= 40):\\n return False\\n except Exception:\\n pass\\n return True\\n\\n def extract_js_from_html(html_code: str) -> list:\\n scripts = []\\n pattern = re.compile(r'<script.*?>([\\\\s\\\\S]*?)</script>', re.IGNORECASE)\\n for match in pattern.finditer(html_code):\\n scripts.append(match.group(1))\\n return scripts\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常规的依赖与虚拟环境目录\\n for d in ['.git', 'node_modules', '__pycache__', 'venv', 'env']:\\n if d in dirs:\\n dirs.remove(d)\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n if ext == '.py':\\n if not check_py_file(content):\\n return False\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n if not check_js_code(content):\\n return False\\n elif ext in ['.html', '.htm']:\\n scripts = extract_js_from_html(content)\\n for script in scripts:\\n if not check_js_code(script):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，必须保证所有函数的圈复杂度上限为4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在生成的代码文件中，严禁使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的代码文件必须满足一个条件：循环结构的嵌套不能超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所生成的各个文件，行数均不得大于200行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n file_checked = False\\n \\n # 遍历当前工作目录\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，防止遍历如 .git 等不需要的文件\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n # 尝试以 utf-8 编码读取内容并判断行数\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 如果行数大于 200，则该文件不符合要求\\n if len(lines) > 200:\\n return False\\n \\n # 标记成功读取并检查了至少一个有效文本文件\\n file_checked = True\\n \\n except UnicodeDecodeError:\\n # 忽略非 utf-8 编码的二进制文件 (例如图像等)\\n pass\\n except Exception:\\n # 忽略因权限等其他原因造成的不可读文件\\n pass\\n \\n # 必须至少存在一个生成的文件，且所有文件行数均不超过200行时才算满足要求\\n return file_checked\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"务必保证每个生成的代码文件里注释详细，其中每行注释的字数下限为20个字\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 常见的有效代码文件后缀集合\\n exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.h', '.cs', '.go'}\\n \\n # 编译匹配模式：优先跳过字符串，进而提取 HTML、块级、单行以及 Python 单行注释\\n # 对单双引号限制了不跨行，以避免在HTML或异常语法下将大规模跨行内容当成单一字符串导致漏抓注释\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\\\\\n])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(?P<html_comment><!--[\\\\s\\\\S]*?-->)|'\\n r'(?P<block_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)|'\\n r'(?P<line_comment>(?<!http:)(?<!https:)//[^\\\\n]*)|'\\n r'(?P<py_comment>#[^\\\\n]*)'\\n )\\n \\n file_count = 0\\n for root, dirs, files in os.walk(workspace_path):\\n for f in files:\\n _, ext = os.path.splitext(f)\\n if ext.lower() in exts:\\n file_count += 1\\n filepath = os.path.join(root, f)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as file:\\n content = file.read()\\n except Exception:\\n # 如果不能用 UTF-8 读取或发生其他异常，视为未能按照要求生成规范代码文件\\n return False\\n \\n has_valid_comment = False\\n # 遍历所有正则匹配结果\\n for match in pattern.finditer(content):\\n comment_text = None\\n is_block = False\\n \\n # 识别具体匹配的注释类型，并剥离首尾的特殊修饰字符\\n if match.group('html_comment'):\\n comment_text = match.group('html_comment')[4:-3]\\n elif match.group('block_comment'):\\n comment_text = match.group('block_comment')[2:-2]\\n is_block = True\\n elif match.group('line_comment'):\\n comment_text = match.group('line_comment')[2:]\\n elif match.group('py_comment'):\\n comment_text = match.group('py_comment')[1:]\\n \\n if comment_text is not None:\\n # 针对多行文本，将其进行换行切分后对每一行进行校验\\n for line in comment_text.split('\\\\n'):\\n clean_line = line.strip()\\n if is_block:\\n # 对于块级注释中间部分的行，需要去除前导的星号修饰（如 JSDoc 风格）\\n clean_line = re.sub(r'^\\\\*+\\\\s*', '', clean_line).strip()\\n \\n # 当此行为具有实际意义的非空行时才进行判定\\n if clean_line:\\n has_valid_comment = True\\n # 使用已实现好的人类专家函数，检验该行是否满足20字的下限要求\\n if count_word(clean_line) < 20:\\n return False\\n \\n # 若某个代码文件内未能找到任何有内容的注释，违背了“每个文件里注释详细”的要求\\n if not has_valid_comment:\\n return False\\n \\n # 如果项目中没有生成符合目标扩展名的任何代码文件，则不符合要求\\n if file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"修改任何已有文件的首要步骤，必须是使用相应的工具来阅读它\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码应专注实现当前的明确要求，不可过度设计或预留后续功能，且需将任何未曾使用的代码与依赖统统删去\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在每次修改项目之后新建文件 change_log_X.log 保存本次的修改记录，X 为以 0 为起点的改动计数值\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"进行代码修改前，不可缺少使用 TodoWrite 工具规划任务这一步骤，当你完成了一个任务对应的代码修改（Write/Edit）后，紧跟的第一步操作必须是调用 TodoComplete 标记该任务完成，严禁在此阶段内穿插其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"请务必尽量通过专用工具（如Read, Write, Edit）来完成文件操作，而不是借助 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用Unified Diff等任何形式的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查页面布局与基础元素：页面分为左右两栏布局。左侧控制面板包含'物体1类型'、'物体1重量'滑块、'物体2类型'、'物体2重量'滑块、'真空环境'开关或复选框、'空气阻力系数'滑块、'重力加速度'滑块且默认值为9.8；右侧为较高的矩形实验展示区，顶部有两个待下落物体，底部有'开始'和'重置'按钮。","验证滑块的实时数值反馈：在左侧控制面板用鼠标拖动'重力加速度'或'空气阻力系数'等滑块，观察滑块旁边或下方是否能实时、精准地显示当前拖动位置对应的具体数值。","验证物体类型的UI渲染表现：在左侧面板将物体1选为'小球'，物体2选为'羽毛'。观察右侧实验区顶部的两个物体，'小球'必须展示为带有CSS径向渐变、具有3D立体光泽的球体，'羽毛'必须展示为SVG图形或具备羽毛特征的符号。","验证运动轨迹拖尾（残影）效果：点击'开始'按钮，在物体下落过程中观察或截屏。物体背后的移动路径上应生成带有透明度渐隐的残影或运动轨迹拖尾，并且能通过拖尾疏密程度直观体现速度的实时变化。","验证非真空环境下的下落物理逻辑：确保取消勾选'真空环境'，将物体1选为'小球'，物体2选为'羽毛'，点击'开始'按钮。在下落过程中截屏观察，必须能明显看到小球下落速度快于羽毛（小球的纵向位置低于羽毛）。","验证真空环境下的下落物理逻辑：勾选左侧的'真空环境'，将物体1选为'小球'，物体2选为'羽毛'，点击'开始'按钮。在下落过程中多次截屏观察，无论类型和重量如何，两个物体必须始终保持在完全相同的水平高度同步下落，且同时触底。","验证高精度计时器与单体落地时间显示：点击'开始'按钮，观察右侧顶端中间的计时器是否从0.00秒开始实时滚动增加。当某一个物体碰到下边界（地面）停止时，观察该物体的正下方是否瞬间显示其具体的落地用时（例如'x.xx秒'）。","验证实验报告面板弹窗：等待两个物体都触碰到下边界停止后，观察右侧实验区中央是否自动弹出一个半透明的实验报告面板。该面板上必须清晰展示物体1的用时、物体2的用时，并包含一句生成的物理结论文本（如'真空中物体下落加速度与质量无关'等）。","验证重置功能：在物体下落途中或落地后，点击底部的'重置'按钮。观察两个物体是否立刻回到右侧展示区的顶部初始位置，顶部全局计时器是否归零，两个物体下方的落地时间是否清空，运动轨迹拖尾是否清除，且实验报告弹窗面板是否消失。","验证物理动画计算实现：检查JS动画逻辑是否使用RequestAnimationFrame驱动每帧更新，并按照a=(mg-f)/m计算加速度；其中重力G=mg，空气阻力f=k*v，k来自表单中的空气阻力系数，v为当前下落速度；勾选'真空环境'时阻力f应恒为0，选择'羽毛'时应存在阻力放大系数以模拟迎风面积更大。","验证开始按钮的基础触发与停止机制：点击'开始'后两个物体应从顶部开始下落，动画过程中位置持续更新；当任意物体碰到容器下边界时，该物体应停止在底部而不是穿出容器或继续移动。"],"string":"[\n \"检查页面布局与基础元素：页面分为左右两栏布局。左侧控制面板包含'物体1类型'、'物体1重量'滑块、'物体2类型'、'物体2重量'滑块、'真空环境'开关或复选框、'空气阻力系数'滑块、'重力加速度'滑块且默认值为9.8；右侧为较高的矩形实验展示区，顶部有两个待下落物体，底部有'开始'和'重置'按钮。\",\n \"验证滑块的实时数值反馈：在左侧控制面板用鼠标拖动'重力加速度'或'空气阻力系数'等滑块，观察滑块旁边或下方是否能实时、精准地显示当前拖动位置对应的具体数值。\",\n \"验证物体类型的UI渲染表现：在左侧面板将物体1选为'小球'，物体2选为'羽毛'。观察右侧实验区顶部的两个物体，'小球'必须展示为带有CSS径向渐变、具有3D立体光泽的球体，'羽毛'必须展示为SVG图形或具备羽毛特征的符号。\",\n \"验证运动轨迹拖尾（残影）效果：点击'开始'按钮，在物体下落过程中观察或截屏。物体背后的移动路径上应生成带有透明度渐隐的残影或运动轨迹拖尾，并且能通过拖尾疏密程度直观体现速度的实时变化。\",\n \"验证非真空环境下的下落物理逻辑：确保取消勾选'真空环境'，将物体1选为'小球'，物体2选为'羽毛'，点击'开始'按钮。在下落过程中截屏观察，必须能明显看到小球下落速度快于羽毛（小球的纵向位置低于羽毛）。\",\n \"验证真空环境下的下落物理逻辑：勾选左侧的'真空环境'，将物体1选为'小球'，物体2选为'羽毛'，点击'开始'按钮。在下落过程中多次截屏观察，无论类型和重量如何，两个物体必须始终保持在完全相同的水平高度同步下落，且同时触底。\",\n \"验证高精度计时器与单体落地时间显示：点击'开始'按钮，观察右侧顶端中间的计时器是否从0.00秒开始实时滚动增加。当某一个物体碰到下边界（地面）停止时，观察该物体的正下方是否瞬间显示其具体的落地用时（例如'x.xx秒'）。\",\n \"验证实验报告面板弹窗：等待两个物体都触碰到下边界停止后，观察右侧实验区中央是否自动弹出一个半透明的实验报告面板。该面板上必须清晰展示物体1的用时、物体2的用时，并包含一句生成的物理结论文本（如'真空中物体下落加速度与质量无关'等）。\",\n \"验证重置功能：在物体下落途中或落地后，点击底部的'重置'按钮。观察两个物体是否立刻回到右侧展示区的顶部初始位置，顶部全局计时器是否归零，两个物体下方的落地时间是否清空，运动轨迹拖尾是否清除，且实验报告弹窗面板是否消失。\",\n \"验证物理动画计算实现：检查JS动画逻辑是否使用RequestAnimationFrame驱动每帧更新，并按照a=(mg-f)/m计算加速度；其中重力G=mg，空气阻力f=k*v，k来自表单中的空气阻力系数，v为当前下落速度；勾选'真空环境'时阻力f应恒为0，选择'羽毛'时应存在阻力放大系数以模拟迎风面积更大。\",\n \"验证开始按钮的基础触发与停止机制：点击'开始'后两个物体应从顶部开始下落，动画过程中位置持续更新；当任意物体碰到容器下边界时，该物体应停止在底部而不是穿出容器或继续移动。\"\n]","isConversational":false}}},{"rowIdx":78,"cells":{"task_id":{"kind":"number","value":79,"string":"79"},"repository_policy":{"kind":"string","value":"请你在后续所有轮次的回复中始终严格遵循以下约束条件：\n1. 必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改。\n2. 想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径。\n3. 确保在代码开发完成之后，测试所有的代码并保证它们均通过测试。\n4. 必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中。\n5. 在你提供的代码文件内，严禁包含任何递归操作。\n6. 确保每一个生成的代码文件里，注释部分完全使用中文。\n7. 你所生成的所有代码文件都必须使用UTF-8进行编码。\n8. 运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', '.scss', '.sass', '.less',\n '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs', '.go', '.rs', '.swift', '.kt',\n '.rb', '.php', '.sh', '.bat', '.ps1', '.sql', '.xml', '.json', '.yaml', '.yml', '.md'\n }\n\n # 定义需要忽略的常见环境目录或构建输出目录\n ignore_dirs = {\n '.git', '__pycache__', 'node_modules', 'venv', 'env', \n '.env', '.idea', '.vscode', 'build', 'dist', 'out', 'target'\n }\n\n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的子目录\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用utf-8严格解码，若文件非UTF-8编码且含有非ASCII字符，将抛出异常\n content.decode('utf-8')\n except UnicodeDecodeError:\n return False\n except Exception:\n # 若存在其他不可读错误等，出于容错也可以视作不满足或略过\n # 在此处为了严格要求，如果代码文件读取失败直接返回False\n return False\n\n return True","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]}],"string":"[\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', '.scss', '.sass', '.less',\\n '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs', '.go', '.rs', '.swift', '.kt',\\n '.rb', '.php', '.sh', '.bat', '.ps1', '.sql', '.xml', '.json', '.yaml', '.yml', '.md'\\n }\\n\\n # 定义需要忽略的常见环境目录或构建输出目录\\n ignore_dirs = {\\n '.git', '__pycache__', 'node_modules', 'venv', 'env', \\n '.env', '.idea', '.vscode', 'build', 'dist', 'out', 'target'\\n }\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的子目录\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用utf-8严格解码，若文件非UTF-8编码且含有非ASCII字符，将抛出异常\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n return False\\n except Exception:\\n # 若存在其他不可读错误等，出于容错也可以视作不满足或略过\\n # 在此处为了严格要求，如果代码文件读取失败直接返回False\\n return False\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我做一个网页版的SVG电子鱼缸。作为第一步，先帮我把基础的静态环境搭好：在页面中央创建一个足够大的鱼缸，利用SVG的渐变或半透明图层做出水体的透明质感以及微微的玻璃反光效果。然后在鱼缸底部，用SVG绘制几棵绿色的水草和几块错落有致的棕色石头。暂时不需要放鱼，先把这个带底景的空鱼缸框架搭建出来。请注意后续开发需始终遵循以下要求：第一，确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项；第二，必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作区路径不存在，说明未生成文件，不违背此单一条件\n if not os.path.exists(workspace_path):\n return True\n \n # 定义常见的代码文件后缀名（包括网页、样式、脚本、配置文件等）\n code_extensions = {\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.svg', '.py', '.json'\n }\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件，尝试用 UTF-8 编码进行读取\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦捕获到解码错误，说明存在非 UTF-8 编码的文件\n return False\n except Exception:\n # 对于非编码导致的异常（如权限问题等），安全跳过即可\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 将目标字符串的所有空白字符剔除，方便后续宽容匹配\n target_str_clean = re.sub(r'\\s+', '', target_str)\n \n # 定义可能生成的代码文件后缀集合\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.svg'}\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰的目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if not content.strip():\n # 若代码文件内容为空，必然没有指定注释\n return False\n\n # 取文件最后 1000 个字符（应对结尾可能的 HTML/SVG 闭合标签和多行注释符）\n tail_content = content[-1000:]\n # 剔除空白字符进行匹配，兼容各种注释折行格式\n tail_clean = re.sub(r'\\s+', '', tail_content)\n \n if target_str_clean not in tail_clean:\n return False\n except Exception:\n # 如果无法读取、解码或者产生其他异常，按不满足要求处理\n return False\n\n return has_code_file","tags":["Content","Format"]}]},{"round_id":1,"instruction":"场景看起来很不错。现在我们需要往鱼缸里放入主角：请用纯SVG在鱼缸里绘制出两条金色的鲤鱼，体型一大一小。每条鱼必须有具体的细节，包括眼睛、身体、尾巴和尾鳍等元素。此外，请为它们的尾鳍添加左右摇动的独立动画（可以使用CSS动画），让它们看起来像是在水中呼吸，目前让它们先静止停留在水里即可。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件结尾添加的指定注释，调整为在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，生成的代码文件中不得使用任何形式的三元运算符。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目描述推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.py', '.svg', '.json'}\n found_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 以二进制读取文件内容\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用utf-8进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出UnicodeDecodeError，说明并非UTF-8编码\n return False\n \n # 确保项目中至少生成了相关的代码文件\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 结合网页版SVG项目和系统提示中对测试脚本的要求，推断常见代码文件扩展名\n code_extensions = {\".html\", \".js\", \".css\", \".svg\", \".ts\", \".jsx\", \".tsx\", \".py\"}\n \n code_files_count = 0\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件前15行内容，足以覆盖文件\"开头处\"的范围\n lines = [f.readline() for _ in range(15)]\n except Exception:\n continue\n \n code_files_count += 1\n \n # 判断目标注释字符串是否出现在文件开头的任意一行中\n has_comment = any(target_comment in line for line in lines if line)\n \n # 若某个代码文件在开头处没有包含目标注释，则不满足要求\n if not has_comment:\n return False\n \n # 如果至少检查到了一个代码文件，且均包含了目标注释，则要求达成\n return code_files_count > 0","tags":["Content","Format"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_js_ts_code(code: str) -> bool:\n \"\"\"\n 判断 JS/TS 代码字符串中是否包含三元运算符\n \"\"\"\n # 去除所有类型的字符串字面量（单引号、双引号、反引号），避免匹配字符串中合法的问号及冒号\n c = re.sub(r'([\"\\'`])(?:(?=(\\\\?))\\2.)*?\\1', '\"\"', code, flags=re.DOTALL)\n \n # 去除多行注释\n c = re.sub(r'/\\*.*?\\*/', '', c, flags=re.DOTALL)\n \n # 去除单行注释\n c = re.sub(r'//.*', '', c)\n \n # 移除 JS/TS 中其他涉及问号和冒号的正常语法干扰，如可选链、空值合并以及可选属性\n c = c.replace('?.', '').replace('??', '').replace('?:', '')\n \n # 检查是否还存在 `?` 后面跟着某个表达式然后有 `:` 的模式\n # 表达式中间通常不会包含分号或大括号\n if re.search(r'\\?[^;\\{\\}]*?:', c):\n return True\n \n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略包和版本管理控制等无关文件夹\n if any(ignored in root for ignored in ['node_modules', '.git', '__pycache__', 'dist', 'build']):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext in ['.js', '.jsx', '.ts', '.tsx']:\n if check_js_ts_code(content):\n # 若发现三元运算符，立刻返回 False 表示不满足要求\n return False\n \n elif ext in ['.html', '.htm', '.svg', '.vue']:\n # 提取 script 标签内容进行检测\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n for script_content in scripts:\n if check_js_ts_code(script_content):\n return False\n \n # 提取内联事件脚本（例如 onclick=\"...\"）进行检测\n content_no_script = re.sub(r'<script\\b[^>]*>.*?</script>', '', content, flags=re.DOTALL | re.IGNORECASE)\n inline_scripts = re.findall(r'\\bon[a-z]+\\s*=\\s*([\"\\'])(.*?)\\1', content_no_script, flags=re.IGNORECASE)\n for _, inline_code in inline_scripts:\n if check_js_ts_code(inline_code):\n return False\n \n elif ext == '.py':\n # 对于 Python 文件，直接使用 ast 解析以获取 `x if cond else y` 的 IfExp 节点\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, ast.IfExp):\n return False\n except Exception:\n pass\n\n # 如果遍历完毕均未发现三元运算符，则认为满足给定要求\n return True","tags":["Style","Paradigm"]}]},{"round_id":2,"instruction":"接下来让这两条鱼活过来。请加入JavaScript逻辑，让两条鱼能在鱼缸内部的二维空间里持续且平滑地随机游动（上下左右均可）。这里有两个重要要求：一是鱼在改变游动方向时，鱼头的朝向必须平滑旋转对准前进的方向；二是做好边界碰撞检测，鱼不能游出鱼缸的玻璃范围，靠近边缘时要自然转身。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件开头添加的指定注释，调整为在每个被生成的代码文件中，要在结尾的独立一行添加注释，以此清晰标明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，后续所有开发需要保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件后缀\n code_extensions = {\n '.html', '.js', '.css', '.svg', '.ts', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.cs', '.go', '.rb', '.php', '.jsx', '.tsx', '.json', \n '.sh', '.vue', '.svelte', '.sql', '.xml', '.yml', '.yaml'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉无需检查的隐藏目录和依赖目录\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取文件字节内容并尝试以 utf-8 进行解码\n with open(file_path, 'rb') as f:\n content = f.read()\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明存在非 UTF-8 编码的文件\n return False\n except Exception:\n # 忽略诸如权限不足等其他读取异常情况\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在每个被生成的代码文件中，要在结尾的独立一行添加注释，以此清晰标明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 推断可能的生成的代码及测试文件扩展名\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.svg', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 统一路径分隔符，排除常见的非AI生成的缓存、版本控制、依赖环境等文件夹\n parts = root.replace('\\\\', '/').split('/')\n if any(exclude in parts for exclude in ['.git', 'node_modules', '__pycache__', '.pytest_cache', 'env', 'venv']):\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 依据系统提示要求，统一按 UTF-8 进行读取\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 寻找文件的最后一个非空行\n last_valid_line = \"\"\n for line in reversed(lines):\n if line.strip():\n last_valid_line = line.strip()\n break\n \n # 验证最后一条非空行内是否包含了目标注释说明\n if target_str not in last_valid_line:\n return False\n except Exception:\n # 若文件读取失败（例如非UTF-8编码异常）视作不满足要求\n return False\n \n # 如果没有找到任何有效的代码文件，同样视为未满足\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的代码文件中不得使用任何形式的三元运算符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def remove_strings_and_comments(text: str) -> str:\n # 移除单行/多行/HTML注释，以及三种常见的字符串字面量\n pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|<!--.*?-->|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`',\n re.DOTALL | re.MULTILINE\n )\n return re.sub(pattern, '', text)\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据项目上下文，推测可能涉及的前端代码文件格式\n if file.endswith(('.html', '.js', '.css', '.svg', '.ts')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 过滤掉注释与字符串内容，避免误判\n clean_content = remove_strings_and_comments(content)\n \n # 查找是否出现三元运算符\n # 规则：匹配 '?' 且后面不能紧跟 '.' (如可选链) 或 '?' (如空值合并) 或 ':' (如类型注解)\n # 中间内容不包含 ';' 和 '{}' 以避免跨语句导致误匹配，最后匹配到冒号 ':'\n if re.search(r'\\?(?!\\.|\\?|:)[^;{}]*:', clean_content):\n return False\n \n return True","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"游动的动作很自然了！下一步我们增加点击互动的喂食功能。请实现：当鼠标在鱼缸内的任意位置点击时，就在该点击位置的上方（模拟水面处）生成3-5颗小小的棕色饲料颗粒。这些饲料出现后，要在重力的作用下缓缓向下沉，最终落到鱼缸底部并停留在水草或石头旁边。请注意后续开发需始终遵循以下更新要求：第一，原要求生成的代码文件中不得使用任何形式的三元运算符，调整为生成的代码文件中请务必包含三元运算符；第二，后续所有开发请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回未满足要求\n if not os.path.exists(workspace_path):\n return False\n\n # 根据上下文推断可能的代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.svg', '.py', '.json'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试用 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明该文件不是使用UTF-8编码的\n return False\n except Exception:\n # 忽略其他如权限问题等非编码相关的读取错误\n pass\n \n # 只要找到了代码文件且均未发生解码错误，即可判定满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在每个被生成的代码文件中，要在结尾的独立一行添加注释，以此清晰标明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据上下文推断，项目涉及网页和SVG，可能的代码文件后缀如下\n code_extensions = ['.html', '.css', '.js', '.svg']\n file_count = 0\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 检查是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n file_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 寻找文件的最后一个非空行\n last_non_empty_line = \"\"\n for line in reversed(lines):\n if line.strip():\n last_non_empty_line = line.strip()\n break\n \n # 如果最后一个非空行中不包含指定的注释文本，则不满足要求\n if target_comment not in last_non_empty_line:\n return False\n except Exception:\n # 如果文件无法按UTF-8读取或其他异常，视为未满足要求\n return False\n\n # 必须有代码文件被检查过且全部满足要求才返回True\n if file_count == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"生成的代码文件中请务必包含三元运算符","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 如果工作区不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n \n # 定义函数以利用正则表达式安全剔除代码中的注释和字符串文本\n def strip_comments_and_strings(text: str) -> str:\n pattern = re.compile(\n r'<!--.*?-->|' # HTML 注释\n r'//.*?$|' # 单行注释\n r'/\\*.*?\\*/|' # 多行注释\n r\"'(?:\\\\.|[^'\\\\])*'|\" # 单引号字符串\n r'\"(?:\\\\.|[^\"\\\\])*\"|' # 双引号字符串\n r'`(?:\\\\.|[^`\\\\])*`', # 模板字符串 (反引号)\n re.MULTILINE | re.DOTALL\n )\n return pattern.sub('', text)\n\n # 遍历工作区内目标前端代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.html', '.ts', '.jsx', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n # 获取剔除掉字符串和注释后的纯净代码结构\n clean_code = strip_comments_and_strings(code)\n \n # 匹配逻辑：\n # \\? -> 匹配问号\n # (?![.?]) -> 排除可选链 (?.) 和空值合并 (??) 的情况\n # [^;]*? -> 中间可以有任意字符，但不跨越分号语句边界\n # : -> 匹配冒号\n if re.search(r'\\?(?![.?])[^;]*?:', clean_code, re.DOTALL):\n return True\n except Exception:\n # 忽略无法读取或解析的文件\n continue\n \n # 如果全都没有找到\n return False","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def mask_strings_and_comments(text: str) -> str:\n # 正则匹配各类注释和不同类型的字符串\n pattern = re.compile(\n r'(?P<html_comment><!--.*?-->)|'\n r'(?P<multi_comment>/\\*.*?\\*/)|'\n r'(?P<single_comment>//.*)|'\n r'(?P<double_quote>\"(?:\\\\.|[^\"\\\\])*\")|'\n r'(?P<single_quote>\\'(?:\\\\.|[^\\'\\\\])*\\')|'\n r'(?P<back_tick>`(?:\\\\.|[^`\\\\])*`)',\n re.DOTALL\n )\n \n def replacer(match):\n # 将匹配到的字符串或注释替换为等长的 'X'，保持原文件文本长度和字符索引不偏移\n return 'X' * len(match.group(0))\n \n return pattern.sub(replacer, text)\n\n valid_extensions = ('.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.svelte')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 清除字符串和注释内容对等号验证的干扰\n masked_content = mask_strings_and_comments(content)\n \n # 寻找所有非复合型的单独等号 \"=\"\n # 排除带有如下前置符的情况：=, +, -, *, /, %, &, |, ^, !, <, >, ?, ~, $\n # 排除带有如下后置符的情况：=, >\n matches = re.finditer(r\"(?<![=+\\-*/%&|^!<>?~$])=(?![=>])\", masked_content)\n \n for match in matches:\n idx = match.start()\n \n # 1. 检查等号前方是否有且仅有两个空格\n if idx < 2 or masked_content[idx-2:idx] != ' ':\n return False\n # 前方第三个字符存在且为空格，说明空格数量超过两个，也视为违规\n if idx >= 3 and masked_content[idx-3] == ' ':\n return False\n \n # 2. 检查等号后方是否有且仅有两个空格\n if idx + 3 > len(masked_content) or masked_content[idx+1:idx+3] != ' ':\n return False\n # 后方第三个字符存在且为空格，说明空格数量超过两个，也视为违规\n if idx + 3 < len(masked_content) and masked_content[idx+3] == ' ':\n return False\n\n # 若遍历了所有的代码且未发现不合规的等号，则说明要求已满足\n return True","tags":["Style","Layout"]}]},{"round_id":4,"instruction":"食物掉下来了，但鱼还没有反应。现在请给鱼加上“视觉追踪”逻辑：每当鱼缸中出现饲料时，两条鲤鱼必须立刻中断当前的随机游动状态。它们需要计算距离自己最近的饲料，转动身体朝向该饲料，并加速向目标饲料游去。由于饲料在下沉，鱼的游动轨迹需要能够实时追踪移动中的饲料。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件结尾添加的指定注释，调整为在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，包含网页基础、SVG、JS逻辑、CSS及测试代码，故指定常见前端代码文件类型\n code_extensions = {'.html', '.htm', '.js', '.css', '.svg', '.ts', '.jsx', '.tsx'}\n code_files_found = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验，过滤掉日志文件(.log)等非代码文件\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 utf-8 编码进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果引发 UnicodeDecodeError，证明其包含非 UTF-8 编码字符\n return False\n except Exception:\n # 如果有其它无法读取的异常（如权限问题等），也视为验证不通过\n return False\n \n # 确保项目中存在代码文件并且所有代码文件均顺利通过了 UTF-8 解码测试\n return code_files_found","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义预期可能被认为是代码文件的扩展名后缀\n code_extensions = {'.html', '.htm', '.js', '.css', '.svg', '.py', '.ts'}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）和无关目录（如 node_modules）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n file_meets_req = False\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 检查文件开头处（取前5行兼容文档声明头部）是否包含该注释\n for _ in range(5):\n line = f.readline()\n if not line: # 已到文件末尾\n break\n if target_comment in line:\n file_meets_req = True\n break\n except Exception:\n # 如果遇到文件无法以utf-8读取或其他异常情况，保守判定为不符合要求\n pass\n \n # 若任意一个代码文件不满足条件，则立即判定不通过\n if not file_meets_req:\n return False\n \n # 如果项目中没有任何代码文件，也视作未满足要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的代码文件中请务必包含三元运算符","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def check_ternary(code: str) -> bool:\n # 移除块注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除 HTML 注释\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n # 移除行注释\n code = re.sub(r'//.*', '', code)\n # 移除字符串常量（包括双引号、单引号、模板字符串）\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', code)\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', code)\n \n # 移除可选链 ?. 和空值合并 ?? 以防将其中的 ? 误判为三元运算符\n code = code.replace('?.', '').replace('??', '')\n \n # 逐行检查是否存在三元运算符模式 (避免跨越不同语句的误判)\n # 三元运算符 ? 和 : 之间必定包含表达式 (即非空的内容)\n for line in code.split('\\n'):\n if re.search(r'\\?[^:]+:', line):\n return True\n \n # 检查可能跨行的三元运算符 (使用 [^;:] 限制，确保不会跨越语句分号匹配)\n if re.search(r'\\?[^;:]+:', code):\n return True\n \n return False\n\n # 遍历项目路径下所有可能包含 JS 逻辑的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx')):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n if check_ternary(code):\n return True\n except Exception:\n continue\n \n return False","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证代码项目中的代码文件中，独立的 \"=\" 前后是否分别恰好有两个空格。\n \"\"\"\n if not os.path.exists(workspace_path):\n return False\n \n # 定义该前端项目可能包含的代码文件扩展名\n valid_extensions = {'.js', '.html', '.svg', '.css', '.ts', '.jsx', '.tsx'}\n code_files = []\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n code_files.append(os.path.join(root, file))\n \n if not code_files:\n return False\n \n # 定义匹配单个等号的正则表达式\n # 1. 后发断言 (?<!...) 排除等号前面是: = < > ! + * / % & | ^ ~ - \n # （注：连字符 - 放在字符集的最后以避免被解析为范围）\n # 2. 先行断言 (?!=...) 排除等号后面是: = >\n pattern = re.compile(r'(?<![=<>!+*/%&|^~-])=(?![=>])')\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n for match in pattern.finditer(content):\n idx = match.start()\n \n # 1. 检查等号前方是否有且恰好有两个空格\n # 前方两个字符必须是 ' '\n if idx < 2 or content[idx-2:idx] != ' ':\n return False\n # 前方第三个字符（如果存在）不能是 ' '\n if idx >= 3 and content[idx-3] == ' ':\n return False\n \n # 2. 检查等号后方是否有且恰好有两个空格\n # 后方两个字符必须是 ' '\n if idx + 3 > len(content) or content[idx+1:idx+3] != ' ':\n return False\n # 后方第三个字符（如果存在）不能是 ' '\n if idx + 3 < len(content) and content[idx+3] == ' ':\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":5,"instruction":"最后一步，完善“进食”的物理判定效果。请在JS中加入碰撞检测，当鱼头（嘴部）碰到目标饲料时，将这颗饲料从DOM中销毁，模拟出被鱼吃掉的效果。当鱼吃完一颗饲料后会去追下一颗，直到鱼缸内所有的饲料都被吃完，两条鱼再平滑地恢复到之前那种漫无目的的随机游动状态。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件开头添加的指定注释，调整为在你生成的每一个代码文件的结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径。","instruction_following_checklist":[{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你提供的代码文件内，严禁包含任何递归操作","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"确保每一个生成的代码文件里，注释部分完全使用中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\n # 常见的网页/前端和Python项目代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.svg', '.py', '.json', '.md', '.txt'}\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录或者特殊目录\n if '.git' in root or '__pycache__' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明不是合法的 UTF-8 编码\n return False\n except Exception:\n # 其他读取异常，这里可以忽略或视作失败，稳妥起见视作失败\n return False\n \n # 如果找到了至少一个代码文件，且全部都能用UTF-8读取成功，则返回True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的每一个代码文件的结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 目标注释文本\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常见的代码文件后缀名\n code_exts = {\n '.py', '.js', '.ts', '.html', '.css', '.scss', \n '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs',\n '.php', '.rb', '.swift', '.kt', '.sql', '.svg'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见无关依赖及隐藏目录，防止误查不需要添加注释的文件\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', '.venv', 'venv', 'env', '.idea', '.vscode'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果发生编码错误无法读取则跳过\n continue\n \n # 过滤出所有非空行\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n if not non_empty_lines:\n return False\n \n # 考虑到块状注释可能会有闭合标签（例如 `-->` 或 `*/`）独立成行\n # 因此我们检查文件的最后3个非空行，看看是否包含目标注释字符串\n found = any(target_text in line for line in non_empty_lines[-3:])\n \n if not found:\n # 只要有一个代码文件末尾未包含该注释，则不满足要求\n return False\n \n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的代码文件中请务必包含三元运算符","validation_code":"import os\nimport re\n\ndef check_ternary_operator(response: str, workspace_path: str) -> bool:\n def strip_comments_and_strings(code: str, is_py: bool) -> str:\n \"\"\"\n 去除代码中的字符串字面量以及注释，防止符号误判\n \"\"\"\n if is_py:\n # 移除 Python 多行字符串/注释\n code = re.sub(r\"'''[\\s\\S]*?'''\", \"\", code)\n code = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', \"\", code)\n # 移除 Python 单行字符串\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"\", code)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', \"\", code)\n # 移除 Python 单行注释\n code = re.sub(r'#.*', \"\", code)\n else:\n # 移除 JS/CSS 多行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', \"\", code)\n # 移除 HTML 注释\n code = re.sub(r'<!--[\\s\\S]*?-->', \"\", code)\n # 移除单行/多行字符串\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"\", code)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', \"\", code)\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', \"\", code)\n # 移除 JS 单行注释\n code = re.sub(r'//.*', \"\", code)\n return code\n\n # 定义可能包含逻辑代码的文件后缀\n valid_exts = {'.js', '.ts', '.html', '.htm', '.py', '.jsx', '.tsx', '.vue', '.svelte'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n is_py = (ext == '.py')\n cleaned_code = strip_comments_and_strings(content, is_py)\n\n if is_py:\n # Python三元运算符: A if condition else B\n # 要求一行内含有 if 和 else，且 if 前方有其他非空字符\n for line in cleaned_code.splitlines():\n if re.search(r'\\S+\\s+\\bif\\b\\s+.*\\belse\\b', line):\n return True\n else:\n # 前端语言三元运算符: condition ? exprIfTrue : exprIfFalse\n # 使用正则 \\?(?![.\\?:])[^:;]+: \n # 作用：匹配 ? 但排除可选链(?.)、双问号(??)、和TS属性(?:) \n # 且 ? 和 : 之间不允许出现分号(;)，避免跨越不同语句导致误判\n if re.search(r'\\?(?![.\\?:])[^:;]+:', cleaned_code):\n return True\n\n return False","tags":["Style","Paradigm"]},{"constraint":"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格","validation_code":"def check_equals_spacing(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义前端代码文件常见的后缀集合\n code_exts = {'.js', '.html', '.css', '.svg', '.ts', '.jsx', '.tsx'}\n \n # 匹配单独的等号，排除各种常见的带等号的组合运算符\n # 前面排除了：<, >, =, !, +, -, *, /, %, &, |, ^, ~, $\n # 后面排除了：=, > (防止匹配 ==, ===, => 等)\n pattern = re.compile(r'(?<![<>=!\\+\\-\\*\\/%&|^~$])=(?![=>])')\n \n found_any_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能的版本控制与依赖目录\n if '.git' in root or 'node_modules' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_any_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n # 如果读取非文本或读取失败则跳过\n continue\n \n # 寻找代码文件里所有符合条件的独立等号\n matches = list(pattern.finditer(text))\n for match in matches:\n start = match.start()\n end = match.end()\n \n # 向前统计连续空格的数量\n left_spaces = 0\n i = start - 1\n while i >= 0 and text[i] == ' ':\n left_spaces += 1\n i -= 1\n \n # 向后统计连续空格的数量\n right_spaces = 0\n j = end\n while j < len(text) and text[j] == ' ':\n right_spaces += 1\n j += 1\n \n # 只要有一个等号的前方或后方不正好是两个空格，则违反约束条件\n if left_spaces != 2 or right_spaces != 2:\n return False\n \n # 如果找到了代码文件且所有独立等号均符合前后两个空格的要求，则返回 True\n return found_any_code_file","tags":["Style","Layout"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我做一个网页版的SVG电子鱼缸。作为第一步，先帮我把基础的静态环境搭好：在页面中央创建一个足够大的鱼缸，利用SVG的渐变或半透明图层做出水体的透明质感以及微微的玻璃反光效果。然后在鱼缸底部，用SVG绘制几棵绿色的水草和几块错落有致的棕色石头。暂时不需要放鱼，先把这个带底景的空鱼缸框架搭建出来。请注意后续开发需始终遵循以下要求：第一，确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项；第二，必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作区路径不存在，说明未生成文件，不违背此单一条件\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义常见的代码文件后缀名（包括网页、样式、脚本、配置文件等）\\n code_extensions = {\\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.svg', '.py', '.json'\\n }\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件，尝试用 UTF-8 编码进行读取\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦捕获到解码错误，说明存在非 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 对于非编码导致的异常（如权限问题等），安全跳过即可\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 将目标字符串的所有空白字符剔除，方便后续宽容匹配\\n target_str_clean = re.sub(r'\\\\s+', '', target_str)\\n \\n # 定义可能生成的代码文件后缀集合\\n code_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.svg'}\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰的目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if not content.strip():\\n # 若代码文件内容为空，必然没有指定注释\\n return False\\n\\n # 取文件最后 1000 个字符（应对结尾可能的 HTML/SVG 闭合标签和多行注释符）\\n tail_content = content[-1000:]\\n # 剔除空白字符进行匹配，兼容各种注释折行格式\\n tail_clean = re.sub(r'\\\\s+', '', tail_content)\\n \\n if target_str_clean not in tail_clean:\\n return False\\n except Exception:\\n # 如果无法读取、解码或者产生其他异常，按不满足要求处理\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"场景看起来很不错。现在我们需要往鱼缸里放入主角：请用纯SVG在鱼缸里绘制出两条金色的鲤鱼，体型一大一小。每条鱼必须有具体的细节，包括眼睛、身体、尾巴和尾鳍等元素。此外，请为它们的尾鳍添加左右摇动的独立动画（可以使用CSS动画），让它们看起来像是在水中呼吸，目前让它们先静止停留在水里即可。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件结尾添加的指定注释，调整为在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，生成的代码文件中不得使用任何形式的三元运算符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目描述推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.py', '.svg', '.json'}\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制读取文件内容\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用utf-8进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出UnicodeDecodeError，说明并非UTF-8编码\\n return False\\n \\n # 确保项目中至少生成了相关的代码文件\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 结合网页版SVG项目和系统提示中对测试脚本的要求，推断常见代码文件扩展名\\n code_extensions = {\\\".html\\\", \\\".js\\\", \\\".css\\\", \\\".svg\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".py\\\"}\\n \\n code_files_count = 0\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件前15行内容，足以覆盖文件\\\"开头处\\\"的范围\\n lines = [f.readline() for _ in range(15)]\\n except Exception:\\n continue\\n \\n code_files_count += 1\\n \\n # 判断目标注释字符串是否出现在文件开头的任意一行中\\n has_comment = any(target_comment in line for line in lines if line)\\n \\n # 若某个代码文件在开头处没有包含目标注释，则不满足要求\\n if not has_comment:\\n return False\\n \\n # 如果至少检查到了一个代码文件，且均包含了目标注释，则要求达成\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_js_ts_code(code: str) -> bool:\\n \\\"\\\"\\\"\\n 判断 JS/TS 代码字符串中是否包含三元运算符\\n \\\"\\\"\\\"\\n # 去除所有类型的字符串字面量（单引号、双引号、反引号），避免匹配字符串中合法的问号及冒号\\n c = re.sub(r'([\\\"\\\\'`])(?:(?=(\\\\\\\\?))\\\\2.)*?\\\\1', '\\\"\\\"', code, flags=re.DOTALL)\\n \\n # 去除多行注释\\n c = re.sub(r'/\\\\*.*?\\\\*/', '', c, flags=re.DOTALL)\\n \\n # 去除单行注释\\n c = re.sub(r'//.*', '', c)\\n \\n # 移除 JS/TS 中其他涉及问号和冒号的正常语法干扰，如可选链、空值合并以及可选属性\\n c = c.replace('?.', '').replace('??', '').replace('?:', '')\\n \\n # 检查是否还存在 `?` 后面跟着某个表达式然后有 `:` 的模式\\n # 表达式中间通常不会包含分号或大括号\\n if re.search(r'\\\\?[^;\\\\{\\\\}]*?:', c):\\n return True\\n \\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略包和版本管理控制等无关文件夹\\n if any(ignored in root for ignored in ['node_modules', '.git', '__pycache__', 'dist', 'build']):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext in ['.js', '.jsx', '.ts', '.tsx']:\\n if check_js_ts_code(content):\\n # 若发现三元运算符，立刻返回 False 表示不满足要求\\n return False\\n \\n elif ext in ['.html', '.htm', '.svg', '.vue']:\\n # 提取 script 标签内容进行检测\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n for script_content in scripts:\\n if check_js_ts_code(script_content):\\n return False\\n \\n # 提取内联事件脚本（例如 onclick=\\\"...\\\"）进行检测\\n content_no_script = re.sub(r'<script\\\\b[^>]*>.*?</script>', '', content, flags=re.DOTALL | re.IGNORECASE)\\n inline_scripts = re.findall(r'\\\\bon[a-z]+\\\\s*=\\\\s*([\\\"\\\\'])(.*?)\\\\1', content_no_script, flags=re.IGNORECASE)\\n for _, inline_code in inline_scripts:\\n if check_js_ts_code(inline_code):\\n return False\\n \\n elif ext == '.py':\\n # 对于 Python 文件，直接使用 ast 解析以获取 `x if cond else y` 的 IfExp 节点\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.IfExp):\\n return False\\n except Exception:\\n pass\\n\\n # 如果遍历完毕均未发现三元运算符，则认为满足给定要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来让这两条鱼活过来。请加入JavaScript逻辑，让两条鱼能在鱼缸内部的二维空间里持续且平滑地随机游动（上下左右均可）。这里有两个重要要求：一是鱼在改变游动方向时，鱼头的朝向必须平滑旋转对准前进的方向；二是做好边界碰撞检测，鱼不能游出鱼缸的玻璃范围，靠近边缘时要自然转身。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件开头添加的指定注释，调整为在每个被生成的代码文件中，要在结尾的独立一行添加注释，以此清晰标明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，后续所有开发需要保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件后缀\\n code_extensions = {\\n '.html', '.js', '.css', '.svg', '.ts', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.cs', '.go', '.rb', '.php', '.jsx', '.tsx', '.json', \\n '.sh', '.vue', '.svelte', '.sql', '.xml', '.yml', '.yaml'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉无需检查的隐藏目录和依赖目录\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取文件字节内容并尝试以 utf-8 进行解码\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明存在非 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 忽略诸如权限不足等其他读取异常情况\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在每个被生成的代码文件中，要在结尾的独立一行添加注释，以此清晰标明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 推断可能的生成的代码及测试文件扩展名\\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.svg', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 统一路径分隔符，排除常见的非AI生成的缓存、版本控制、依赖环境等文件夹\\n parts = root.replace('\\\\\\\\', '/').split('/')\\n if any(exclude in parts for exclude in ['.git', 'node_modules', '__pycache__', '.pytest_cache', 'env', 'venv']):\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 依据系统提示要求，统一按 UTF-8 进行读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 寻找文件的最后一个非空行\\n last_valid_line = \\\"\\\"\\n for line in reversed(lines):\\n if line.strip():\\n last_valid_line = line.strip()\\n break\\n \\n # 验证最后一条非空行内是否包含了目标注释说明\\n if target_str not in last_valid_line:\\n return False\\n except Exception:\\n # 若文件读取失败（例如非UTF-8编码异常）视作不满足要求\\n return False\\n \\n # 如果没有找到任何有效的代码文件，同样视为未满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中不得使用任何形式的三元运算符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def remove_strings_and_comments(text: str) -> str:\\n # 移除单行/多行/HTML注释，以及三种常见的字符串字面量\\n pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|<!--.*?-->|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`',\\n re.DOTALL | re.MULTILINE\\n )\\n return re.sub(pattern, '', text)\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据项目上下文，推测可能涉及的前端代码文件格式\\n if file.endswith(('.html', '.js', '.css', '.svg', '.ts')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 过滤掉注释与字符串内容，避免误判\\n clean_content = remove_strings_and_comments(content)\\n \\n # 查找是否出现三元运算符\\n # 规则：匹配 '?' 且后面不能紧跟 '.' (如可选链) 或 '?' (如空值合并) 或 ':' (如类型注解)\\n # 中间内容不包含 ';' 和 '{}' 以避免跨语句导致误匹配，最后匹配到冒号 ':'\\n if re.search(r'\\\\?(?!\\\\.|\\\\?|:)[^;{}]*:', clean_content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"游动的动作很自然了！下一步我们增加点击互动的喂食功能。请实现：当鼠标在鱼缸内的任意位置点击时，就在该点击位置的上方（模拟水面处）生成3-5颗小小的棕色饲料颗粒。这些饲料出现后，要在重力的作用下缓缓向下沉，最终落到鱼缸底部并停留在水草或石头旁边。请注意后续开发需始终遵循以下更新要求：第一，原要求生成的代码文件中不得使用任何形式的三元运算符，调整为生成的代码文件中请务必包含三元运算符；第二，后续所有开发请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回未满足要求\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据上下文推断可能的代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.svg', '.py', '.json'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试用 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明该文件不是使用UTF-8编码的\\n return False\\n except Exception:\\n # 忽略其他如权限问题等非编码相关的读取错误\\n pass\\n \\n # 只要找到了代码文件且均未发生解码错误，即可判定满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在每个被生成的代码文件中，要在结尾的独立一行添加注释，以此清晰标明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据上下文推断，项目涉及网页和SVG，可能的代码文件后缀如下\\n code_extensions = ['.html', '.css', '.js', '.svg']\\n file_count = 0\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 检查是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 寻找文件的最后一个非空行\\n last_non_empty_line = \\\"\\\"\\n for line in reversed(lines):\\n if line.strip():\\n last_non_empty_line = line.strip()\\n break\\n \\n # 如果最后一个非空行中不包含指定的注释文本，则不满足要求\\n if target_comment not in last_non_empty_line:\\n return False\\n except Exception:\\n # 如果文件无法按UTF-8读取或其他异常，视为未满足要求\\n return False\\n\\n # 必须有代码文件被检查过且全部满足要求才返回True\\n if file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中请务必包含三元运算符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 如果工作区不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义函数以利用正则表达式安全剔除代码中的注释和字符串文本\\n def strip_comments_and_strings(text: str) -> str:\\n pattern = re.compile(\\n r'<!--.*?-->|' # HTML 注释\\n r'//.*?$|' # 单行注释\\n r'/\\\\*.*?\\\\*/|' # 多行注释\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\" # 单引号字符串\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|' # 双引号字符串\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', # 模板字符串 (反引号)\\n re.MULTILINE | re.DOTALL\\n )\\n return pattern.sub('', text)\\n\\n # 遍历工作区内目标前端代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.html', '.ts', '.jsx', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n # 获取剔除掉字符串和注释后的纯净代码结构\\n clean_code = strip_comments_and_strings(code)\\n \\n # 匹配逻辑：\\n # \\\\? -> 匹配问号\\n # (?![.?]) -> 排除可选链 (?.) 和空值合并 (??) 的情况\\n # [^;]*? -> 中间可以有任意字符，但不跨越分号语句边界\\n # : -> 匹配冒号\\n if re.search(r'\\\\?(?![.?])[^;]*?:', clean_code, re.DOTALL):\\n return True\\n except Exception:\\n # 忽略无法读取或解析的文件\\n continue\\n \\n # 如果全都没有找到\\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def mask_strings_and_comments(text: str) -> str:\\n # 正则匹配各类注释和不同类型的字符串\\n pattern = re.compile(\\n r'(?P<html_comment><!--.*?-->)|'\\n r'(?P<multi_comment>/\\\\*.*?\\\\*/)|'\\n r'(?P<single_comment>//.*)|'\\n r'(?P<double_quote>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\")|'\\n r'(?P<single_quote>\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|'\\n r'(?P<back_tick>`(?:\\\\\\\\.|[^`\\\\\\\\])*`)',\\n re.DOTALL\\n )\\n \\n def replacer(match):\\n # 将匹配到的字符串或注释替换为等长的 'X'，保持原文件文本长度和字符索引不偏移\\n return 'X' * len(match.group(0))\\n \\n return pattern.sub(replacer, text)\\n\\n valid_extensions = ('.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.svelte')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 清除字符串和注释内容对等号验证的干扰\\n masked_content = mask_strings_and_comments(content)\\n \\n # 寻找所有非复合型的单独等号 \\\"=\\\"\\n # 排除带有如下前置符的情况：=, +, -, *, /, %, &, |, ^, !, <, >, ?, ~, $\\n # 排除带有如下后置符的情况：=, >\\n matches = re.finditer(r\\\"(?<![=+\\\\-*/%&|^!<>?~$])=(?![=>])\\\", masked_content)\\n \\n for match in matches:\\n idx = match.start()\\n \\n # 1. 检查等号前方是否有且仅有两个空格\\n if idx < 2 or masked_content[idx-2:idx] != ' ':\\n return False\\n # 前方第三个字符存在且为空格，说明空格数量超过两个，也视为违规\\n if idx >= 3 and masked_content[idx-3] == ' ':\\n return False\\n \\n # 2. 检查等号后方是否有且仅有两个空格\\n if idx + 3 > len(masked_content) or masked_content[idx+1:idx+3] != ' ':\\n return False\\n # 后方第三个字符存在且为空格，说明空格数量超过两个，也视为违规\\n if idx + 3 < len(masked_content) and masked_content[idx+3] == ' ':\\n return False\\n\\n # 若遍历了所有的代码且未发现不合规的等号，则说明要求已满足\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"食物掉下来了，但鱼还没有反应。现在请给鱼加上“视觉追踪”逻辑：每当鱼缸中出现饲料时，两条鲤鱼必须立刻中断当前的随机游动状态。它们需要计算距离自己最近的饲料，转动身体朝向该饲料，并加速向目标饲料游去。由于饲料在下沉，鱼的游动轨迹需要能够实时追踪移动中的饲料。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件结尾添加的指定注释，调整为在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，包含网页基础、SVG、JS逻辑、CSS及测试代码，故指定常见前端代码文件类型\\n code_extensions = {'.html', '.htm', '.js', '.css', '.svg', '.ts', '.jsx', '.tsx'}\\n code_files_found = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验，过滤掉日志文件(.log)等非代码文件\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 utf-8 编码进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果引发 UnicodeDecodeError，证明其包含非 UTF-8 编码字符\\n return False\\n except Exception:\\n # 如果有其它无法读取的异常（如权限问题等），也视为验证不通过\\n return False\\n \\n # 确保项目中存在代码文件并且所有代码文件均顺利通过了 UTF-8 解码测试\\n return code_files_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件的开头处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 定义预期可能被认为是代码文件的扩展名后缀\\n code_extensions = {'.html', '.htm', '.js', '.css', '.svg', '.py', '.ts'}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）和无关目录（如 node_modules）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n file_meets_req = False\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 检查文件开头处（取前5行兼容文档声明头部）是否包含该注释\\n for _ in range(5):\\n line = f.readline()\\n if not line: # 已到文件末尾\\n break\\n if target_comment in line:\\n file_meets_req = True\\n break\\n except Exception:\\n # 如果遇到文件无法以utf-8读取或其他异常情况，保守判定为不符合要求\\n pass\\n \\n # 若任意一个代码文件不满足条件，则立即判定不通过\\n if not file_meets_req:\\n return False\\n \\n # 如果项目中没有任何代码文件，也视作未满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中请务必包含三元运算符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def check_ternary(code: str) -> bool:\\n # 移除块注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除 HTML 注释\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n # 移除行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除字符串常量（包括双引号、单引号、模板字符串）\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', code)\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', code)\\n \\n # 移除可选链 ?. 和空值合并 ?? 以防将其中的 ? 误判为三元运算符\\n code = code.replace('?.', '').replace('??', '')\\n \\n # 逐行检查是否存在三元运算符模式 (避免跨越不同语句的误判)\\n # 三元运算符 ? 和 : 之间必定包含表达式 (即非空的内容)\\n for line in code.split('\\\\n'):\\n if re.search(r'\\\\?[^:]+:', line):\\n return True\\n \\n # 检查可能跨行的三元运算符 (使用 [^;:] 限制，确保不会跨越语句分号匹配)\\n if re.search(r'\\\\?[^;:]+:', code):\\n return True\\n \\n return False\\n\\n # 遍历项目路径下所有可能包含 JS 逻辑的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx')):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if check_ternary(code):\\n return True\\n except Exception:\\n continue\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证代码项目中的代码文件中，独立的 \\\"=\\\" 前后是否分别恰好有两个空格。\\n \\\"\\\"\\\"\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义该前端项目可能包含的代码文件扩展名\\n valid_extensions = {'.js', '.html', '.svg', '.css', '.ts', '.jsx', '.tsx'}\\n code_files = []\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n if not code_files:\\n return False\\n \\n # 定义匹配单个等号的正则表达式\\n # 1. 后发断言 (?<!...) 排除等号前面是: = < > ! + * / % & | ^ ~ - \\n # （注：连字符 - 放在字符集的最后以避免被解析为范围）\\n # 2. 先行断言 (?!=...) 排除等号后面是: = >\\n pattern = re.compile(r'(?<![=<>!+*/%&|^~-])=(?![=>])')\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n for match in pattern.finditer(content):\\n idx = match.start()\\n \\n # 1. 检查等号前方是否有且恰好有两个空格\\n # 前方两个字符必须是 ' '\\n if idx < 2 or content[idx-2:idx] != ' ':\\n return False\\n # 前方第三个字符（如果存在）不能是 ' '\\n if idx >= 3 and content[idx-3] == ' ':\\n return False\\n \\n # 2. 检查等号后方是否有且恰好有两个空格\\n # 后方两个字符必须是 ' '\\n if idx + 3 > len(content) or content[idx+1:idx+3] != ' ':\\n return False\\n # 后方第三个字符（如果存在）不能是 ' '\\n if idx + 3 < len(content) and content[idx+3] == ' ':\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后一步，完善“进食”的物理判定效果。请在JS中加入碰撞检测，当鱼头（嘴部）碰到目标饲料时，将这颗饲料从DOM中销毁，模拟出被鱼吃掉的效果。当鱼吃完一颗饲料后会去追下一颗，直到鱼缸内所有的饲料都被吃完，两条鱼再平滑地恢复到之前那种漫无目的的随机游动状态。请注意后续开发需始终遵循以下更新要求：第一，原要求在每个代码文件开头添加的指定注释，调整为在你生成的每一个代码文件的结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”；第二，你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"想要操作文件，请先保证已经运行了`pwd`以确认当前的目录路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"必须妥善保留对生成代码进行的所有中间测试代码，并将它们统一放入项目目录的__tests__子目录中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件内，严禁包含任何递归操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"确保每一个生成的代码文件里，注释部分完全使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n # 常见的网页/前端和Python项目代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.svg', '.py', '.json', '.md', '.txt'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录或者特殊目录\\n if '.git' in root or '__pycache__' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 其他读取异常，这里可以忽略或视作失败，稳妥起见视作失败\\n return False\\n \\n # 如果找到了至少一个代码文件，且全部都能用UTF-8读取成功，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在90%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"确保代码只包含当前指定的功能，严禁任何过度设计或预留未来特性，并务必清理掉所有未使用的代码与依赖项\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的每一个代码文件的结尾处单独换行添加注释，并清楚地说明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 目标注释文本\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常见的代码文件后缀名\\n code_exts = {\\n '.py', '.js', '.ts', '.html', '.css', '.scss', \\n '.java', '.c', '.cpp', '.h', '.cs', '.go', '.rs',\\n '.php', '.rb', '.swift', '.kt', '.sql', '.svg'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见无关依赖及隐藏目录，防止误查不需要添加注释的文件\\n dirs[:] = [d for d in dirs if d not in {'.git', 'node_modules', '__pycache__', '.venv', 'venv', 'env', '.idea', '.vscode'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果发生编码错误无法读取则跳过\\n continue\\n \\n # 过滤出所有非空行\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n if not non_empty_lines:\\n return False\\n \\n # 考虑到块状注释可能会有闭合标签（例如 `-->` 或 `*/`）独立成行\\n # 因此我们检查文件的最后3个非空行，看看是否包含目标注释字符串\\n found = any(target_text in line for line in non_empty_lines[-3:])\\n \\n if not found:\\n # 只要有一个代码文件末尾未包含该注释，则不满足要求\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中请务必包含三元运算符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_ternary_operator(response: str, workspace_path: str) -> bool:\\n def strip_comments_and_strings(code: str, is_py: bool) -> str:\\n \\\"\\\"\\\"\\n 去除代码中的字符串字面量以及注释，防止符号误判\\n \\\"\\\"\\\"\\n if is_py:\\n # 移除 Python 多行字符串/注释\\n code = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", \\\"\\\", code)\\n code = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', \\\"\\\", code)\\n # 移除 Python 单行字符串\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"\\\", code)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', \\\"\\\", code)\\n # 移除 Python 单行注释\\n code = re.sub(r'#.*', \\\"\\\", code)\\n else:\\n # 移除 JS/CSS 多行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', \\\"\\\", code)\\n # 移除 HTML 注释\\n code = re.sub(r'<!--[\\\\s\\\\S]*?-->', \\\"\\\", code)\\n # 移除单行/多行字符串\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"\\\", code)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', \\\"\\\", code)\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', \\\"\\\", code)\\n # 移除 JS 单行注释\\n code = re.sub(r'//.*', \\\"\\\", code)\\n return code\\n\\n # 定义可能包含逻辑代码的文件后缀\\n valid_exts = {'.js', '.ts', '.html', '.htm', '.py', '.jsx', '.tsx', '.vue', '.svelte'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n is_py = (ext == '.py')\\n cleaned_code = strip_comments_and_strings(content, is_py)\\n\\n if is_py:\\n # Python三元运算符: A if condition else B\\n # 要求一行内含有 if 和 else，且 if 前方有其他非空字符\\n for line in cleaned_code.splitlines():\\n if re.search(r'\\\\S+\\\\s+\\\\bif\\\\b\\\\s+.*\\\\belse\\\\b', line):\\n return True\\n else:\\n # 前端语言三元运算符: condition ? exprIfTrue : exprIfFalse\\n # 使用正则 \\\\?(?![.\\\\?:])[^:;]+: \\n # 作用：匹配 ? 但排除可选链(?.)、双问号(??)、和TS属性(?:) \\n # 且 ? 和 : 之间不允许出现分号(;)，避免跨越不同语句导致误判\\n if re.search(r'\\\\?(?![.\\\\?:])[^:;]+:', cleaned_code):\\n return True\\n\\n return False\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"保证目标代码的分支覆盖率在执行你生成的单元测试文件后，至少能达到95%的标准\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，“=”的前方和后方都分别有两个空格\",\n \"validation_code\": \"def check_equals_spacing(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义前端代码文件常见的后缀集合\\n code_exts = {'.js', '.html', '.css', '.svg', '.ts', '.jsx', '.tsx'}\\n \\n # 匹配单独的等号，排除各种常见的带等号的组合运算符\\n # 前面排除了：<, >, =, !, +, -, *, /, %, &, |, ^, ~, $\\n # 后面排除了：=, > (防止匹配 ==, ===, => 等)\\n pattern = re.compile(r'(?<![<>=!\\\\+\\\\-\\\\*\\\\/%&|^~$])=(?![=>])')\\n \\n found_any_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能的版本控制与依赖目录\\n if '.git' in root or 'node_modules' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_any_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n # 如果读取非文本或读取失败则跳过\\n continue\\n \\n # 寻找代码文件里所有符合条件的独立等号\\n matches = list(pattern.finditer(text))\\n for match in matches:\\n start = match.start()\\n end = match.end()\\n \\n # 向前统计连续空格的数量\\n left_spaces = 0\\n i = start - 1\\n while i >= 0 and text[i] == ' ':\\n left_spaces += 1\\n i -= 1\\n \\n # 向后统计连续空格的数量\\n right_spaces = 0\\n j = end\\n while j < len(text) and text[j] == ' ':\\n right_spaces += 1\\n j += 1\\n \\n # 只要有一个等号的前方或后方不正好是两个空格，则违反约束条件\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n \\n # 如果找到了代码文件且所有独立等号均符合前后两个空格的要求，则返回 True\\n return found_any_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["观察页面中央，确认显示了一个足够大的鱼缸，并且带有半透明水体效果和微微的玻璃反光效果。","观察鱼缸底部区域，确认绘制有绿色的水草和棕色的石头，且错落有致没有超出鱼缸范围。","观察鱼缸内部，确认存在两条金色的鲤鱼，体型为一大一小，并且鱼的视觉细节包含了眼睛、身体、尾巴和尾鳍等部分。","持续观察两条金鲤鱼，确认它们的尾鳍部分带有左右摇动的独立持续动画效果，看起来像在呼吸或划水。","在不进行任何点击操作的情况下观察一段时间，确认两条鱼在鱼缸范围内能自主、持续、平滑地进行上下左右随机移动。","仔细观察鱼在改变方向时的表现，确认鱼在游动时鱼头始终对准前进方向，且转向时发生了平滑的旋转过渡，而不是生硬突变。","观察鱼游向鱼缸边缘的情况，确认具有边界限制逻辑，鱼不会游出鱼缸的玻璃范围，并在靠近边缘时能够自然转身游回鱼缸内。","使用鼠标在鱼缸内部任意位置点击一次，确认会在点击位置的上方（模拟水面位置处）立刻生成3至5颗小巧的棕色饲料颗粒。","出现饲料后，观察饲料的运动状态，确认饲料颗粒会在重力作用下缓缓向下沉，最终落在鱼缸底部并停留在水草或石头区域，不再下沉掉出屏幕外。","点击屏幕生成饲料，立即观察鱼的反应，确认两条鱼立刻中断了原有的随机游动状态，迅速转动身体朝向距离自己最近的饲料，并加速向该下沉饲料游去实现动态追踪。","观察鱼触碰饲料的瞬间，确认当鱼头（嘴部位置）接触到目标饲料时，该颗饲料立即从页面中消失，呈现出被吃掉的视觉效果。","观察进食后的行为，确认鱼在吃掉一颗饲料后会继续寻找下一颗饲料，直到鱼缸内所有饲料被吃完，随后两条鱼平滑地恢复到漫无目的的随机游动状态。","检查页面源代码或DOM结构，确认鱼缸主体、水体/玻璃反光、水草、石头、两条金色鲤鱼以及饲料颗粒等核心视觉元素均使用SVG元素绘制或承载，而不是用Canvas或纯HTML普通元素替代核心图形。"],"string":"[\n \"观察页面中央，确认显示了一个足够大的鱼缸，并且带有半透明水体效果和微微的玻璃反光效果。\",\n \"观察鱼缸底部区域，确认绘制有绿色的水草和棕色的石头，且错落有致没有超出鱼缸范围。\",\n \"观察鱼缸内部，确认存在两条金色的鲤鱼，体型为一大一小，并且鱼的视觉细节包含了眼睛、身体、尾巴和尾鳍等部分。\",\n \"持续观察两条金鲤鱼，确认它们的尾鳍部分带有左右摇动的独立持续动画效果，看起来像在呼吸或划水。\",\n \"在不进行任何点击操作的情况下观察一段时间，确认两条鱼在鱼缸范围内能自主、持续、平滑地进行上下左右随机移动。\",\n \"仔细观察鱼在改变方向时的表现，确认鱼在游动时鱼头始终对准前进方向，且转向时发生了平滑的旋转过渡，而不是生硬突变。\",\n \"观察鱼游向鱼缸边缘的情况，确认具有边界限制逻辑，鱼不会游出鱼缸的玻璃范围，并在靠近边缘时能够自然转身游回鱼缸内。\",\n \"使用鼠标在鱼缸内部任意位置点击一次，确认会在点击位置的上方（模拟水面位置处）立刻生成3至5颗小巧的棕色饲料颗粒。\",\n \"出现饲料后，观察饲料的运动状态，确认饲料颗粒会在重力作用下缓缓向下沉，最终落在鱼缸底部并停留在水草或石头区域，不再下沉掉出屏幕外。\",\n \"点击屏幕生成饲料，立即观察鱼的反应，确认两条鱼立刻中断了原有的随机游动状态，迅速转动身体朝向距离自己最近的饲料，并加速向该下沉饲料游去实现动态追踪。\",\n \"观察鱼触碰饲料的瞬间，确认当鱼头（嘴部位置）接触到目标饲料时，该颗饲料立即从页面中消失，呈现出被吃掉的视觉效果。\",\n \"观察进食后的行为，确认鱼在吃掉一颗饲料后会继续寻找下一颗饲料，直到鱼缸内所有饲料被吃完，随后两条鱼平滑地恢复到漫无目的的随机游动状态。\",\n \"检查页面源代码或DOM结构，确认鱼缸主体、水体/玻璃反光、水草、石头、两条金色鲤鱼以及饲料颗粒等核心视觉元素均使用SVG元素绘制或承载，而不是用Canvas或纯HTML普通元素替代核心图形。\"\n]","isConversational":false}}},{"rowIdx":79,"cells":{"task_id":{"kind":"number","value":80,"string":"80"},"repository_policy":{"kind":"string","value":"以下是你在后续所有轮次交互中必须严格遵循的约束条件，所有要求均需永久生效：\n1. 在生成针对用户的回复时，必须且只能一直采用中文。\n2. 强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-。\n3. 必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n4. 在你生成的所有代码文件中，注释内容必须统一使用中文编写。\n5. 你输出的代码文件中严格禁止出现递归调用。\n6. 在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列。\n7. 每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。\n8. 写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令中的“import/require”等特征推断，涵盖主要的开发语言代码扩展名\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.cjs', '.mjs', \n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \n '.rb', '.php', '.swift', '.kt', '.rs', '.html', '.css'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 读取第一行（即顶行）\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于某种原因无法读取文件，根据严格验证原则可能视为无法判断或不满足\n # 这里保守视为未满足该条件\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 常见的代码文件扩展名白名单\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \n '.cs', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', \n '.rb', '.php', '.sh', '.swift', '.kt', '.rs', '.m', '.scala', \n '.vue', '.svelte', '.dart', '.lua', '.sql'\n }\n \n if not os.path.isdir(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read().strip()\n \n # 空文件跳过\n if not content:\n continue\n \n # 检查是否包含指定的文本\n if target_str not in content:\n return False\n \n # 将内容按行分割，并只保留非空行\n lines = [line.strip() for line in content.split('\\n') if line.strip()]\n \n # 查找目标字符串所在的最后位置\n last_index = -1\n for i in range(len(lines) - 1, -1, -1):\n if target_str in lines[i]:\n last_index = i\n break\n \n # 判断目标字符串是否在结尾位置\n # 考虑到结尾可能有 `*/`、`-->` 等注释闭合符号，允许其距离有效行的末尾最大偏移量为 2 行\n if last_index == -1 or (len(lines) - 1 - last_index > 2):\n return False\n \n except Exception:\n # 如果读取文件出现异常（如文件无法被 utf-8 解析等），跳过该文件\n continue\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]}],"string":"[\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令中的“import/require”等特征推断，涵盖主要的开发语言代码扩展名\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.cjs', '.mjs', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', \\n '.rb', '.php', '.swift', '.kt', '.rs', '.html', '.css'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 读取第一行（即顶行）\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于某种原因无法读取文件，根据严格验证原则可能视为无法判断或不满足\\n # 这里保守视为未满足该条件\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 常见的代码文件扩展名白名单\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \\n '.cs', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', \\n '.rb', '.php', '.sh', '.swift', '.kt', '.rs', '.m', '.scala', \\n '.vue', '.svelte', '.dart', '.lua', '.sql'\\n }\\n \\n if not os.path.isdir(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read().strip()\\n \\n # 空文件跳过\\n if not content:\\n continue\\n \\n # 检查是否包含指定的文本\\n if target_str not in content:\\n return False\\n \\n # 将内容按行分割，并只保留非空行\\n lines = [line.strip() for line in content.split('\\\\n') if line.strip()]\\n \\n # 查找目标字符串所在的最后位置\\n last_index = -1\\n for i in range(len(lines) - 1, -1, -1):\\n if target_str in lines[i]:\\n last_index = i\\n break\\n \\n # 判断目标字符串是否在结尾位置\\n # 考虑到结尾可能有 `*/`、`-->` 等注释闭合符号，允许其距离有效行的末尾最大偏移量为 2 行\\n if last_index == -1 or (len(lines) - 1 - last_index > 2):\\n return False\\n \\n except Exception:\\n # 如果读取文件出现异常（如文件无法被 utf-8 解析等），跳过该文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"UI/UX Optimization"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用SVG画一个展示“医疗急诊处理流程”的泳道图草图。这第一步只需要画出图表的骨架，不需要任何流程节点和连线。具体要求：1. 画布要足够大（建议至少1200x800），顶部写上大标题“医疗急诊处理流程图”。2. 左侧划分出5个横向泳道，并写上角色/部门名称：分诊护士、急诊医生、检查/检验科、专科/手术团队、行政/出入院管理。3. 顶部划分出6列纵向的阶段表头：患者到达/分诊判断、急诊医生初诊、检查与诊断、急救/重症处理、医生诊断与决策、出院/住院/转院流程。4. 给不同的横向泳道加上区分度较好的浅色交替背景，用实线或虚线画出网格线以便后续放置内容。另外请你后续轮次始终遵循以下要求：第一，在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符；第二，代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 根据用户指令推测，可能生成的代码文件为 SVG 文件、嵌套 SVG 的网页文件或相关生成脚本\n code_extensions = {'.svg', '.xml', '.html', '.htm', '.py', '.js', '.ts', '.sh'}\n \n files_checked = 0\n # 1. 首先检查工作区中的实际文件\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 验证顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n files_checked += 1\n except Exception:\n # 忽略可能由于某些原因无法以 UTF-8 读取的文件\n pass\n \n # 2. 如果工作区内未查找到符合条件的代码文件，降级检查回复文本中的代码块\n if files_checked == 0:\n # 匹配 markdown 格式的代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n # 过滤掉空代码块\n if not block.strip():\n continue\n \n # 获取代码块的顶行\n first_line = block.split('\\n')[0]\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名集合，其中必须包含 .svg\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cs', '.h', '.hpp', '.js', '.jsx', \n '.ts', '.tsx', '.html', '.htm', '.css', '.scss', '.less', '.svg', \n '.xml', '.sh', '.bash', '.go', '.rs', '.rb', '.php', '.swift', \n '.kt', '.dart', '.m', '.mm', '.json', '.yaml', '.yml', '.sql'\n }\n \n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n # 如果遇到无法读取的文件，保险起见按未满足要求处理\n return False\n \n # 移除文件末尾的空白行\n while lines and not lines[-1].strip():\n lines.pop()\n \n # 如果文件为空，必然不满足包含注释的要求\n if not lines:\n return False\n \n # 检查文件的结尾位置（例如最后 5 行之内）是否包含目标字符串\n # 倒数5行足以覆盖注释单独占据一行或是用块注释包裹的场景\n found = False\n for line in lines[-5:]:\n if target_text in line:\n found = True\n break\n \n # 若任意一个代码文件缺失该注释，即为不满足要求\n if not found:\n return False\n \n # 如果没有找到任何代码文件，说明助手没有输出需要的文件，不符合要求\n return code_files_found","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":1,"instruction":"好的，骨架搭得很清晰。现在我们在第一列“患者到达/分诊判断”区域加入内容。在分诊护士的泳道里，画一个绿色的开始节点“患者到达”，连线到一个矩形节点“接待/登记/测体征”，然后再连到一个菱形判断框“初步分级判断”。从这个判断框引出三条关键连线：1. “危重”（红线），直接向下穿透，连到第四行专科/手术团队对应的第二或第四列位置，画一个红色的目标节点“急救处理”；2. “严重”（黄线），连到第二行急诊医生对应的第一列边缘，画个“转初诊”占位；3. “轻症”（绿线），连到第二行急诊医生的一个叫“排队候诊”的节点。请你后续轮次始终遵循以下新增及调整的要求：第一，你提供的SVG代码文件中，任何出现“=”的地方，其两边都必须各有两个空格；第二，原有的代码编写完成后需对所有代码进行测试并保证测试通过的要求作废，改为写完代码之后，绝不能对其进行任何测试。请用对应颜色的线条，并在连线上写好文字标识。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在或者不是一个目录，无法验证\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 根据用户指令推断，主要生成的代码类型应该是 svg 格式，\n # 为了鲁棒性，同时囊括其他常见的代码和标记语言的扩展名\n target_extensions = {'.svg', '.xml', '.html', '.htm', '.js', '.css', '.py'}\n \n code_files_count = 0\n\n # 遍历当前工作区的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件的后缀名\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n code_files_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 仅读取文件的顶行（第一行）\n first_line = f.readline()\n \n # 判断顶行是否包含了指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果无法按utf-8格式成功读取，也算作失败\n return False\n \n # 如果找到了至少一个代码文件，且全部通过了检查，则返回 True\n return code_files_count > 0","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_sentence = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n checked_items = 0\n \n # 1. 检查 response 中的 markdown 代码块\n code_blocks = re.findall(r'```.*?\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n block_stripped = block.strip()\n if block_stripped:\n checked_items += 1\n # 截取末尾 500 个字符，验证是否在结尾位置包含目标注释\n tail_content = block_stripped[-500:]\n if target_sentence not in tail_content:\n return False\n \n # 2. 检查 workspace_path 中的生成的代码文件\n if os.path.exists(workspace_path):\n # 常见可能生成的文件后缀，依据指令推断SVG及相关常见语言\n code_exts = {'.svg', '.py', '.js', '.html', '.css', '.xml'}\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read().strip()\n if content:\n checked_items += 1\n # 截取末尾 500 个字符，验证是否在结尾位置包含目标注释\n tail_content = content[-500:]\n if target_sentence not in tail_content:\n return False\n except Exception:\n # 忽略非文本或无法用utf-8读取的文件\n continue\n \n # 3. 如果没找到任何代码块或代码文件，视为未生成所需代码或未满足要求\n if checked_items == 0:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"写完代码之后，绝不能对其进行任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的SVG代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n svg_contents = []\n \n # 1. 尝试从工作区路径中提取 .svg 文件内容\n if workspace_path and os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n svg_contents.append(f.read())\n except Exception:\n pass\n \n # 2. 尝试从 response 的 Markdown 代码块中提取 SVG 内容\n # 匹配所有的代码块，寻找带有 <svg 标签的内容\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n if '<svg' in block.lower():\n svg_contents.append(block)\n \n # 如果没有找到任何 SVG 代码文件或代码块，则无法满足要求\n if not svg_contents:\n return False\n \n # 3. 验证所有 SVG 内容中 `=` 的空格情况\n for content in svg_contents:\n if '=' not in content:\n continue\n \n for i, char in enumerate(content):\n if char == '=':\n # 检查左侧是否有且仅有两个空格\n if i < 2:\n return False\n if content[i-1] != ' ' or content[i-2] != ' ':\n return False\n if i >= 3 and content[i-3] == ' ':\n return False\n \n # 检查右侧是否有且仅有两个空格\n if i + 2 >= len(content):\n return False\n if content[i+1] != ' ' or content[i+2] != ' ':\n return False\n if i + 3 < len(content) and content[i+3] == ' ':\n return False\n \n return True","tags":["Style","Layout"]}]},{"round_id":2,"instruction":"接下来完善第二列“急诊医生初诊”。在急诊医生的泳道里，接收刚刚排队或转初诊的患者流，依次画出“问诊”、“体格检查”、“开具检查单”这三个矩形节点用箭头相连。在这个流程后加入一个“病情评估”的菱形判断框，并分出三条路：1. 如果判断“需急救”，画一条红线连接到刚才我们在专科团队那里画的“急救处理”节点；2. 如果判断“需进一步检查”，画一条线向下连到第三行检查科的对应列，先放一个占位节点“流向检查科”；3. 如果是“轻症可处理”，直接向右拉一条线到第五行行政管理的最后一列，放一个“开处方/出院”节点。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的写完代码之后绝不能对其进行任何测试的要求作废，改为写完代码后必须对所有的代码进行测试以确保测试通过；第二，你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_coding_declaration(response: str, workspace_path: str) -> bool:\n import os\n \n # 结合用户指令中的“SVG”等特征，定义需要检查的代码文件扩展名集合\n code_extensions = {\n '.svg', '.xml', '.py', '.js', '.ts', '.html', '.css', \n '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh'\n }\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n code_files_found = False\n \n # 遍历当前代码项目地址下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对推断为代码类型的文件进行验证\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的顶行（第一行）\n first_line = f.readline()\n \n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法使用 utf-8 解码或存在其他读取错误，则判定为不满足\n return False\n \n # 如果找到了代码文件且均满足要求，则返回 True；如果未生成任何代码文件，则视为未达成条件\n return code_files_found","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令背景推断，此处的代码文件主要是指 SVG 文件\n if file.lower().endswith(\".svg\"):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n lines = f.readlines()\n \n # 移除末尾的空白行，以便准确获取实际代码或注释的尾部内容\n while lines and not lines[-1].strip():\n lines.pop()\n \n if not lines:\n return False\n \n # 检查文件的结尾位置（最后10个有效非空行）内是否包含特定的注释字符串\n found_in_tail = False\n for line in lines[-10:]:\n if target_comment in line:\n found_in_tail = True\n break\n \n if not found_in_tail:\n return False\n\n except Exception:\n # 出现无法读取等异常情况时，判定为失败\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你提供的SVG代码文件中，任何出现“=”的地方，其两边都必须各有两个空格","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n svg_contents = []\n \n # 1. 遍历工作区寻找 .svg 文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.svg'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n svg_contents.append(f.read())\n except Exception:\n pass\n \n # 2. 若工作区中没有 SVG 文件，尝试从回复中提取 svg 代码块作为兜底\n if not svg_contents:\n svg_blocks = re.findall(r'```svg\\s*(.*?)\\s*```', response, re.DOTALL)\n svg_contents.extend(svg_blocks)\n \n # 如果依然没有任何 SVG 代码内容，则认定未满足输出 SVG 代码的前提\n if not svg_contents:\n return False\n \n # 3. 遍历所有 SVG 内容并检查 `=` 周围的空格\n for content in svg_contents:\n for i, char in enumerate(content):\n if char == '=':\n # 统计左边的连续空格数量\n left_spaces = 0\n j = i - 1\n while j >= 0 and content[j] == ' ':\n left_spaces += 1\n j -= 1\n \n # 统计右边的连续空格数量\n right_spaces = 0\n k = i + 1\n while k < len(content) and content[k] == ' ':\n right_spaces += 1\n k += 1\n \n # 检查是否都严格恰好各有两个空格\n if left_spaces != 2 or right_spaces != 2:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"写完代码后必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n svg_contents = []\n \n # 1. 尝试在工作区查找并读取 .svg 文件的内容\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n svg_contents.append(f.read())\n except Exception:\n pass\n \n # 2. 如果未找到任何文件，退而从 response 提取 SVG 代码块进行验证\n if not svg_contents:\n pattern = re.compile(r'```[^\\n]*\\n(.*?)```', re.DOTALL)\n matches = pattern.findall(response)\n for match in matches:\n if '<svg' in match.lower() or target_str in match:\n svg_contents.append(match)\n \n # 3. 如果既没有 .svg 文件也没有在回复中找到任何疑似的 SVG 代码内容，判定为失败\n if not svg_contents:\n return False\n \n # 4. 对收集到的每一个 SVG 内容进行校验\n for content in svg_contents:\n # 按换行切割并去掉空行\n lines = [line.strip() for line in content.split('\\n') if line.strip()]\n found = False\n \n # 仅检查开头的 5 个非空行\n for line in lines[:5]:\n if target_str in line:\n # 确保是“另起一行”：不能与 <?xml 或 <svg 等主体标签混杂在同一行\n if '<svg' not in line.lower() and '<?xml' not in line.lower():\n found = True\n break\n \n if not found:\n return False\n \n return True","tags":["Content","Format"]}]},{"round_id":3,"instruction":"很棒！现在填补第三列“检查与诊断”的内容。在检查/检验科的泳道中，从刚才的“流向检查科”节点出发，平行画三个检查节点：“化验血液/尿液”、“影像学(X光/CT/MRI)”、“心电图监测”。将这三个节点汇总到一个“检查结果”的菱形判断框中。接着处理分支：如果“检查结果正常”，向上引一条线回到第二行急诊医生泳道的第五列位置，建一个“结合检查做诊断”的节点；如果“检查异常”，向下引一条线到专科/手术团队泳道，建一个“通知介入”的预警节点。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的你提供的SVG代码文件中任何出现“=”的地方其两边都必须各有两个空格的要求作废，改为生成的SVG代码文件中任何“=”周围都不允许带有空白字符；第二，你生成的所有SVG代码文件都必须使用UTF-8进行编码。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，目标代码文件主要为 SVG 文件\n valid_extensions = {'.svg', '.xml', '.html', '.htm', '.py', '.js', '.css'}\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行（顶行）\n first_line = f.readline()\n \n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若文件无法被正常读取（如编码错误或没有读取权限），视作不符合要求\n return False\n\n # 必须至少找到一个代码文件才可能视为成功，否则未生成目标代码判定为失败\n return found_code_file","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户指令推断，主要的代码文件类型为SVG文件以及可能的测试代码文件\n valid_exts = {'.svg', '.py', '.html', '.js', '.sh', '.xml'}\n \n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n code_files_found = True\n \n # 获取文件的所有非空行\n lines = [line.strip() for line in content.splitlines() if line.strip()]\n \n if not lines:\n return False\n \n # 检查目标文本是否出现在文件的末尾部分（最后 10 个非空行内）\n found_in_end = False\n for line in lines[-10:]:\n if target_text in line:\n found_in_end = True\n break\n \n # 如果有任何一个代码文件的末尾没有包含该注释，则判为不满足\n if not found_in_end:\n return False\n \n # 只有当至少检查了一个代码文件，并且全部通过时，才返回 True\n return code_files_found","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的SVG代码文件中任何“=”周围都不允许带有空白字符","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n svg_contents = []\n \n # 1. 从工作区路径中扫描并读取生成的 .svg 文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n svg_contents.append(f.read())\n except Exception:\n pass\n \n # 2. 如果在工作区中没有找到 SVG 文件，降级从助手的回复内容中提取 SVG 代码块\n if not svg_contents:\n blocks = re.findall(r'```([a-zA-Z0-9]*)\\n(.*?)```', response, re.DOTALL)\n for lang, code in blocks:\n # 匹配语言类型为 svg、xml、html 或空，并确保内部含有 <svg 标签\n if lang.lower() in ['svg', 'xml', 'html', ''] and '<svg' in code.lower():\n svg_contents.append(code)\n \n # 3. 如果根本没有找到任何 SVG 代码内容，视为未满足生成要求\n if not svg_contents:\n return False\n \n # 4. 验证内容中是否存在“=”周围带有空白字符的情况\n for content in svg_contents:\n # \\s 匹配任何空白字符（空格、换行、制表符等）\n # \\s= 匹配 '=' 前面有空白字符的情况\n # =\\s 匹配 '=' 后面有空白字符的情况\n if re.search(r'\\s=|=\\s', content):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"写完代码后必须对所有的代码进行测试以确保测试通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 收集所有的 .svg 文件路径\n svg_files = []\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n svg_files.append(os.path.join(root, file))\n\n # 如果未找到任何 SVG 文件，视为未满足生成文件的要求\n if not svg_files:\n return False\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n\n # 遍历所有 SVG 文件，检查是否在开头处包含指定注释\n for file_path in svg_files:\n has_target_comment = False\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取前 10 行以涵盖“开头处另起一行”的可能情况\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n \n if target_comment in line:\n has_target_comment = True\n break\n except Exception:\n # 任何文件读取错误直接判定为不满足要求\n return False\n\n # 如果某一个文件缺少该声明，则整体判定为失败\n if not has_target_comment:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"你生成的所有SVG代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n svg_files = []\n # 获取所有的 .svg 文件\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n svg_files.append(os.path.join(root, file))\n \n # 如果没有生成任何 SVG 文件，则无法证明满足了要求\n if not svg_files:\n return False\n \n for svg_file in svg_files:\n try:\n with open(svg_file, 'rb') as f:\n content = f.read()\n \n # 尝试使用 UTF-8 进行解码以验证文件编码\n text = content.decode('utf-8')\n \n # 检查 XML 声明中的 encoding 属性（如果存在）\n # 例如: <?xml version=\"1.0\" encoding=\"UTF-8\"?>\n xml_decl_match = re.search(r'<\\?xml[^>]+encoding\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']', text, re.IGNORECASE)\n if xml_decl_match:\n encoding = xml_decl_match.group(1).lower()\n if encoding not in ['utf-8', 'utf8']:\n return False\n \n except UnicodeDecodeError:\n # 如果解码抛出异常，说明文件的实际字节流不是合法的 UTF-8 编码\n return False\n except Exception:\n return False\n \n return True","tags":["Language","File Encoding"]}]},{"round_id":4,"instruction":"这步我们来细化第四列“急救/重症处理”。在专科/手术团队的泳道里，扩充原有的“急救处理”节点，在旁边加上“心肺复苏、气道管理、用药抢救”等具体动作的文字或小节点。然后紧接着加一个“急救结果判断”的菱形框：1. 若“病情稳定”，画线向上汇入急诊医生的“结合检查做诊断”节点；2. 若“病情恶化”，画一条带有“再次急救”字样的折线，循环指回前方的“急救处理”节点；3. 若“无效”，直接画一条线引向一个黑色的终点节点，写上“宣布死亡（流程结束）”。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的写完代码后必须对所有的代码进行测试以确保测试通过的要求作废，改为代码一旦编写完毕，严禁进行任何形式的测试；第二，对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef verify_encoding_declaration(response: str, workspace_path: str) -> bool:\n code_files = []\n \n # 遍历当前项目工作目录，查找所有的 .svg 代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.svg'):\n code_files.append(os.path.join(root, file))\n \n # 如果没有生成预期的SVG代码文件，视为未满足要求\n if not code_files:\n return False\n \n # 遍历所有查找到的代码文件，检查顶行内容\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行（顶行）\n first_line = f.readline()\n # 检查指定的编码声明是否包含在顶行中\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法读取或发生其他异常，保守判定为不满足\n return False\n \n # 所有代码文件的顶行都包含了特定声明，则满足要求\n return True","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 依据用户指令推断代码文件类型，以 .svg, .html, .xml 为主\n valid_extensions = ('.svg', '.html', '.xml')\n code_files = []\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(valid_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果没有任何代码文件，无法证明要求被满足\n if not code_files:\n return False\n \n expected_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 去除文件内容整体为空的情况\n if not content.strip():\n return False\n \n # 获取所有非空行\n lines = [line.strip() for line in content.split('\\n') if line.strip()]\n \n # 检查指定的注释文本是否出现在结尾位置（考察末尾 5 个非空行）\n found_at_end = False\n for line in lines[-5:]:\n if expected_text in line:\n found_at_end = True\n break\n \n if not found_at_end:\n return False\n \n except Exception:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的SVG代码文件中任何“=”周围都不允许带有空白字符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n svg_contents = []\n\n # 1. 从工作区中寻找生成的 .svg 文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n svg_contents.append(f.read())\n except Exception:\n pass\n\n # 2. 从 response 中提取可能存在的 SVG 代码块\n code_blocks = re.findall(r'```(?:[a-zA-Z0-9_-]+)?\\s*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n # 只保留包含 svg 标签或者 xmlns 声明的代码块\n if '<svg' in block.lower() or 'xmlns=' in block.lower():\n svg_contents.append(block)\n\n # 3. 如果仍未找到，尝试直接提取裸 <svg> 标签对内容\n if not svg_contents:\n raw_svgs = re.findall(r'(<svg[\\s>].*?</svg>)', response, re.DOTALL | re.IGNORECASE)\n svg_contents.extend(raw_svgs)\n\n # 如果完全没有找到任何 SVG 代码，视为无法满足生成目标要求\n if not svg_contents:\n return False\n\n # 4. 验证核心要求：任何 \"=\" 周围不允许存在空白字符 (\\s=|=\\s)\n # \\s 会匹配空格、制表符 \\t、换行符 \\n 等\n for text in svg_contents:\n if re.search(r'\\s=|=\\s', text):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n target_str = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n svg_contents = []\n \n # 1. 尝试从当前代码工作区加载所有生成的 .svg 后缀的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n svg_contents.append(f.read())\n except Exception:\n pass\n \n # 2. 如果工作区未发现已保存的 .svg 文件，则尝试在最近一轮的 response 寻找包含 SVG 源码的 Markdown 代码块\n if not svg_contents:\n blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in blocks:\n # 判断该代码块是否具有 SVG 的基本标识\n if '<svg' in block.lower():\n svg_contents.append(block)\n \n # 若都没有，则代表其根本未生成或修改任何包含 SVG 的图表文件，验证失败\n if not svg_contents:\n return False\n \n # 3. 对每一个识别到的 SVG 文件内容进行检验\n for content in svg_contents:\n # 去除全空行，防止空行干扰开头行数的判断\n lines = [line.strip() for line in content.split('\\n') if line.strip()]\n \n found_in_beginning = False\n # 取文件的前20行作为“开头处”的标准范围\n for line in lines[:20]:\n if target_str in line:\n # 去除可能的各种注释符号，以便判断它是否是另起一行的干净注释\n clean_line = re.sub(r'<!--|-->|/\\*|\\*/|//|#', '', line)\n clean_line = clean_line.replace(target_str, '').strip()\n \n # 若行内不再包含其他标签(如'<')，并且残留文本较短，则确认是“另起一行进行注释声明”\n if len(clean_line) < 50 and '<' not in clean_line:\n found_in_beginning = True\n break\n \n # 只要有任何一个 SVG 文件不满足要求，则判定为失败\n if not found_in_beginning:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"你生成的所有SVG代码文件都必须使用UTF-8进行编码","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n svg_files = []\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.svg'):\n svg_files.append(os.path.join(root, file))\n \n # 如果没有找到任何SVG文件，则认为未完成生成代码文件的基础任务，返回 False\n if not svg_files:\n return False\n \n for file_path in svg_files:\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n \n # 1. 检查物理文件是否能够以 UTF-8 成功解码\n text = content.decode('utf-8')\n \n # 2. 如果文件包含 XML 声明，检查其 encoding 属性是否明确指定为 utf-8\n # 例如匹配：<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n match = re.search(r'<\\?xml[^>]*encoding\\s*=\\s*[\\'\"]([^\\'\"]+)[\\'\"]', text, re.IGNORECASE)\n if match:\n encoding_value = match.group(1).lower()\n if encoding_value not in ['utf-8', 'utf8']:\n return False\n \n except UnicodeDecodeError:\n # 如果解码抛出异常，说明文件物理上不是纯净的 UTF-8 编码\n return False\n \n return True","tags":["Language","File Encoding"]},{"constraint":"代码一旦编写完毕，严禁进行任何形式的测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":5,"instruction":"逻辑越来越严密了。现在来完善第五列“医生诊断与决策”。在急诊医生泳道里，从“结合检查做诊断”这个节点出来，加上一个关键的“决策”菱形分支，引出四条去向的线：1. “门诊治疗”，向下连到第五行行政管理泳道的“办理出院结算”；2. “需住院”，连到行政泳道的“办理住院”节点；3. “需手术”，向下连到专科团队泳道的“准备手术”节点；4. “需转院”，连到行政泳道的“协调转院”节点。请把这些连线排布整齐，不要过度交叉打结。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的代码一旦编写完毕严禁进行任何形式的测试的要求作废，改为编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过；第二，在对现有SVG文件进行修改之前，请务必先调用相关工具读取文件内容。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n target_statement = \"-*- coding: utf-8 -*-\"\n \n # 定义常见的代码文件扩展名，特别需包含对话中重点生成的 .svg 文件\n # 同时过滤掉 .log 等非代码类文件\n code_extensions = {\n '.svg', '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \n '.css', '.xml', '.c', '.cpp', '.h', '.hpp', '.java', '.go', \n '.rs', '.php', '.rb', '.sh', '.bash'\n }\n \n found_code_file = False\n \n # 优先检查工作区目录下的实际生成的代码文件\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline() # 读取第一行（顶行）\n if target_statement not in first_line:\n return False\n except Exception:\n return False\n \n # 若在工作区中找到了至少一个代码文件并且全经验证通过，则判定符合要求\n if found_code_file:\n return True\n \n # 如果工作区中不存在代码文件（可能尚未真实写盘），则退化为在AI回复内容中提取代码块检查\n pattern = re.compile(r'```[^\\n]*\\n(.*?)```', re.DOTALL)\n blocks = pattern.findall(response)\n if not blocks:\n # 如果既没有生成文件，也没有输出代码块，无法判定满足了生成代码的要求\n return False\n \n for block in blocks:\n lines = block.split('\\n')\n if not lines:\n return False\n first_line = lines[0] # 获取代码块的顶行\n if target_statement not in first_line:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n svg_found = False\n\n # 1. 检查 workspace_path 下的 .svg 代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n svg_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 过滤掉纯空行，方便提取“结尾位置”的有效内容\n lines = [line.strip() for line in content.splitlines() if line.strip()]\n if not lines:\n return False\n \n # 检查目标文本是否出现在最后5个非空行中（兼容尾部含有多行注释标记如 --> 的情况）\n if not any(target_text in line for line in lines[-5:]):\n return False\n except Exception:\n return False\n\n # 2. 检查 response 中的 SVG 代码块\n # 匹配 Markdown 格式的代码块\n code_blocks = re.findall(r'```(?:[a-zA-Z0-9_-]+)?[ \\t]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n # 如果代码块中包含 svg 标签，则判定为生成的 svg 代码\n if '<svg' in block or '</svg>' in block:\n svg_found = True\n lines = [line.strip() for line in block.splitlines() if line.strip()]\n if not lines:\n return False\n \n # 同样检查目标文本是否在代码块结尾位置\n if not any(target_text in line for line in lines[-5:]):\n return False\n\n # 如果找到了至少一个 SVG 文件/代码块且都通过了测试，则要求满足\n return svg_found","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的SVG代码文件中任何“=”周围都不允许带有空白字符","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查生成的SVG代码文件中任何“=”周围都不允许带有空白字符。\n \"\"\"\n has_svg_file = False\n \n # 匹配“=”前面或后面带有空白字符的情况（\\s 匹配任何空白字符，包括空格、制表符、换行符等）\n # \\s= 匹配 \"[空白字符]=\"\n # =\\s 匹配 \"=[空白字符]\"\n pattern = re.compile(r'\\s=|=\\s')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.svg'):\n has_svg_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果匹配到等号周围存在空白字符，则直接返回 False\n if pattern.search(content):\n return False\n except Exception as e:\n # 若文件读取失败，无法验证其是否满足要求，保守起见返回 False\n return False\n \n # 如果找到了至少一个 SVG 文件，且所有 SVG 文件的等号周围均没有空白字符，则返回 True\n return has_svg_file","tags":["Style","Layout"]},{"constraint":"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 收集所有的 SVG 代码文件\n svg_files = []\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n svg_files.append(os.path.join(root, file))\n\n if not svg_files:\n return False\n\n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n\n # 检查每一个 SVG 文件的开头是否包含要求的注释声明\n for file_path in svg_files:\n found_in_head = False\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 只需检查文件开头处，取前 10 行以涵盖各类声明\n for line in lines[:10]:\n if target_comment in line:\n # 确保是“另起一行”，即行内不能掺杂实质性的 SVG 标签\n if \"<svg\" not in line and \"<path\" not in line and \"<g\" not in line and \"<rect\" not in line:\n found_in_head = True\n break\n except Exception:\n return False\n\n if not found_in_head:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"你生成的所有SVG代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n svg_files_found = False\n \n # 遍历 workspace_path 找到所有 .svg 文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n svg_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制读取文件内容\n with open(file_path, 'rb') as f:\n content = f.read()\n \n # 尝试以 utf-8 解码，如果抛出异常则说明并非使用 UTF-8 编码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 发现存在非 UTF-8 编码的 SVG 文件\n return False\n except Exception:\n # 针对其他异常情况，例如文件无法读取，出于严谨也视作验证失败\n return False\n \n # 至少要生成了 SVG 文件并且全都符合 UTF-8 编码要求，才返回 True\n return svg_files_found","tags":["Language","File Encoding"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在对现有SVG文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":6,"instruction":"最后一轮，把第六列“出院/住院/转院流程”的收尾工作做完，并全面美化UI。在行政管理泳道，将出院结算引向一个结束节点“流程结束”；将办理住院引向“病房分配”，再连向专科团队泳道的“住院治疗及护理交接”节点；把协调转院引向“转院完成（流程结束）”节点。同时在专科团队补充“实施手术”并连向“住院治疗”。美化方面：给所有矩形节点加上圆角和阴影，给判断框使用醒目的橙色或边框，确保所有文字排版居中不越界，连线统一改为平滑的折角线条(Orthogonal 风格)。在画布右上角加一个简单的图例说明（节点种类和红黄绿通道），让这张业务架构图看起来非常专业！\n另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的生成的SVG代码文件中任何“=”周围都不允许带有空白字符的要求作废，改为在编写SVG代码文件时，请保证“=”的两端分别用两个空格隔开；第二，每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容。","instruction_following_checklist":[{"constraint":"在生成针对用户的回复时，必须且只能一直采用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断，主要的生成代码文件是SVG\n code_file_extensions = ['.svg']\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if any(file_name.endswith(ext) for ext in code_file_extensions):\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行（顶行）\n first_line = f.readline()\n # 判断是否包含特定编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法读取或发生异常，视为不满足要求\n return False\n\n # 若所有匹配的代码文件都满足要求，且至少找到了一个代码文件，则判定为True\n return has_code_file","tags":["Content","Format"]},{"constraint":"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件后缀，根据对话上下文主要考察 .svg\n code_extensions = ['.svg', '.py', '.js', '.html', '.css', '.java', '.c', '.cpp', '.h']\n \n code_files = []\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果没有生成任何代码文件，则无法验证满足要求\n if not code_files:\n return False\n \n expected_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read().strip()\n \n if not content:\n return False\n \n # 将代码按行分割，过滤掉纯空行\n lines = content.split('\\n')\n non_empty_lines = [line.strip() for line in lines if line.strip()]\n \n if not non_empty_lines:\n return False\n \n # 提取最后几行进行检查，容忍多行注释（如XML/HTML的注释标签占用了不同行）\n found_at_end = False\n # 检查最后 5 行是否包含预期的注释文本\n for line in reversed(non_empty_lines[-5:]):\n if expected_comment in line:\n found_at_end = True\n break\n \n if not found_at_end:\n return False\n \n except Exception:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件中严格禁止出现递归调用","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写SVG代码文件时，请保证“=”的两端分别用两个空格隔开","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n svg_files = []\n # 查找工作区下所有的 .svg 文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.svg'):\n svg_files.append(os.path.join(root, file))\n\n # 如果未找到任何 SVG 文件，则认为要求未满足\n if not svg_files:\n return False\n \n for file_path in svg_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 遍历文件内容，寻找所有 '=' 字符\n for i, char in enumerate(content):\n if char == '=':\n # 提取 '=' 左侧的2个和3个字符\n left_2 = content[i-2:i] if i >= 2 else content[:i]\n left_3 = content[i-3:i] if i >= 3 else content[:i]\n \n # 提取 '=' 右侧的2个和3个字符\n right_2 = content[i+1:i+3] if i + 3 <= len(content) else content[i+1:]\n right_3 = content[i+1:i+4] if i + 4 <= len(content) else content[i+1:]\n \n # 验证左侧是否正好是两个空格\n if left_2 != \" \":\n return False\n if len(left_3) == 3 and left_3[0] == \" \":\n return False\n \n # 验证右侧是否正好是两个空格\n if right_2 != \" \":\n return False\n if len(right_3) == 3 and right_3[-1] == \" \":\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n target_comment = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n svg_found = False\n \n # 1. 优先检查 workspace_path 下是否存在 .svg 文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.lower().endswith('.svg'):\n svg_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 剔除开头空白字符后按行分割\n lines = content.lstrip().split('\\n')\n \n # 检查开头处（这里取前10行作为判断开头的冗余范围）\n comment_found = False\n for line in lines[:10]:\n if target_comment in line:\n comment_found = True\n break\n \n if not comment_found:\n return False\n except Exception:\n return False\n\n # 如果在本地工作区内找到了至少一个 .svg 文件，并且全部通过了上述测试，则符合要求\n if svg_found:\n return True\n \n # 2. 如果工作区中没有提取到 .svg 文件，检查 response 文本中是否包含 SVG 代码块\n svg_blocks = []\n \n # 匹配 markdown 格式的代码块\n code_blocks = re.findall(r'```([a-zA-Z0-9]*)\\n(.*?)```', response, re.DOTALL)\n for lang, code in code_blocks:\n if lang.lower() == 'svg' or '<svg' in code.lower():\n svg_blocks.append(code)\n \n # 如果没有找到 markdown 代码块，尝试匹配原生的 <svg> ... </svg> 标签\n if not svg_blocks:\n raw_svgs = re.findall(r'(<svg.*?</svg>)', response, re.DOTALL | re.IGNORECASE)\n svg_blocks.extend(raw_svgs)\n \n # 如果既没有对应的生成文件，回复里也没有 SVG 代码，则无从验证，返回 False\n if not svg_blocks:\n return False\n \n # 检查从回复文本中提取到的所有 SVG 代码段\n for block in svg_blocks:\n lines = block.lstrip().split('\\n')\n comment_found = False\n for line in lines[:10]:\n if target_comment in line:\n comment_found = True\n break\n \n # 只要存在任何一个 SVG 未按照要求声明，均视作失败\n if not comment_found:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"你生成的所有SVG代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n has_svg = False\n \n # 遍历项目目录结构查找所有的 SVG 文件\n for root, dirs, files in os.walk(workspace_path):\n for filename in files:\n if filename.lower().endswith('.svg'):\n has_svg = True\n file_path = os.path.join(root, filename)\n try:\n # 尝试以 UTF-8 编码读取该 SVG 文件的全部内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果触发了解码异常，证明文件并不是使用 UTF-8 编码的\n return False\n except Exception:\n # 如果有其他的读取异常，出于严格验证的目的视为未满足要求\n return False\n \n # 如果找到了至少一个 SVG 文件并且全部都通过了解码测试，则返回 True\n return has_svg","tags":["Language","File Encoding"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在对现有SVG文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用SVG画一个展示“医疗急诊处理流程”的泳道图草图。这第一步只需要画出图表的骨架，不需要任何流程节点和连线。具体要求：1. 画布要足够大（建议至少1200x800），顶部写上大标题“医疗急诊处理流程图”。2. 左侧划分出5个横向泳道，并写上角色/部门名称：分诊护士、急诊医生、检查/检验科、专科/手术团队、行政/出入院管理。3. 顶部划分出6列纵向的阶段表头：患者到达/分诊判断、急诊医生初诊、检查与诊断、急救/重症处理、医生诊断与决策、出院/住院/转院流程。4. 给不同的横向泳道加上区分度较好的浅色交替背景，用实线或虚线画出网格线以便后续放置内容。另外请你后续轮次始终遵循以下要求：第一，在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符；第二，代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推测，可能生成的代码文件为 SVG 文件、嵌套 SVG 的网页文件或相关生成脚本\\n code_extensions = {'.svg', '.xml', '.html', '.htm', '.py', '.js', '.ts', '.sh'}\\n \\n files_checked = 0\\n # 1. 首先检查工作区中的实际文件\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 验证顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n files_checked += 1\\n except Exception:\\n # 忽略可能由于某些原因无法以 UTF-8 读取的文件\\n pass\\n \\n # 2. 如果工作区内未查找到符合条件的代码文件，降级检查回复文本中的代码块\\n if files_checked == 0:\\n # 匹配 markdown 格式的代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n # 过滤掉空代码块\\n if not block.strip():\\n continue\\n \\n # 获取代码块的顶行\\n first_line = block.split('\\\\n')[0]\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名集合，其中必须包含 .svg\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cs', '.h', '.hpp', '.js', '.jsx', \\n '.ts', '.tsx', '.html', '.htm', '.css', '.scss', '.less', '.svg', \\n '.xml', '.sh', '.bash', '.go', '.rs', '.rb', '.php', '.swift', \\n '.kt', '.dart', '.m', '.mm', '.json', '.yaml', '.yml', '.sql'\\n }\\n \\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果遇到无法读取的文件，保险起见按未满足要求处理\\n return False\\n \\n # 移除文件末尾的空白行\\n while lines and not lines[-1].strip():\\n lines.pop()\\n \\n # 如果文件为空，必然不满足包含注释的要求\\n if not lines:\\n return False\\n \\n # 检查文件的结尾位置（例如最后 5 行之内）是否包含目标字符串\\n # 倒数5行足以覆盖注释单独占据一行或是用块注释包裹的场景\\n found = False\\n for line in lines[-5:]:\\n if target_text in line:\\n found = True\\n break\\n \\n # 若任意一个代码文件缺失该注释，即为不满足要求\\n if not found:\\n return False\\n \\n # 如果没有找到任何代码文件，说明助手没有输出需要的文件，不符合要求\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好的，骨架搭得很清晰。现在我们在第一列“患者到达/分诊判断”区域加入内容。在分诊护士的泳道里，画一个绿色的开始节点“患者到达”，连线到一个矩形节点“接待/登记/测体征”，然后再连到一个菱形判断框“初步分级判断”。从这个判断框引出三条关键连线：1. “危重”（红线），直接向下穿透，连到第四行专科/手术团队对应的第二或第四列位置，画一个红色的目标节点“急救处理”；2. “严重”（黄线），连到第二行急诊医生对应的第一列边缘，画个“转初诊”占位；3. “轻症”（绿线），连到第二行急诊医生的一个叫“排队候诊”的节点。请你后续轮次始终遵循以下新增及调整的要求：第一，你提供的SVG代码文件中，任何出现“=”的地方，其两边都必须各有两个空格；第二，原有的代码编写完成后需对所有代码进行测试并保证测试通过的要求作废，改为写完代码之后，绝不能对其进行任何测试。请用对应颜色的线条，并在连线上写好文字标识。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在或者不是一个目录，无法验证\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 根据用户指令推断，主要生成的代码类型应该是 svg 格式，\\n # 为了鲁棒性，同时囊括其他常见的代码和标记语言的扩展名\\n target_extensions = {'.svg', '.xml', '.html', '.htm', '.js', '.css', '.py'}\\n \\n code_files_count = 0\\n\\n # 遍历当前工作区的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件的后缀名\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n code_files_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 仅读取文件的顶行（第一行）\\n first_line = f.readline()\\n \\n # 判断顶行是否包含了指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果无法按utf-8格式成功读取，也算作失败\\n return False\\n \\n # 如果找到了至少一个代码文件，且全部通过了检查，则返回 True\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_sentence = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n checked_items = 0\\n \\n # 1. 检查 response 中的 markdown 代码块\\n code_blocks = re.findall(r'```.*?\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n block_stripped = block.strip()\\n if block_stripped:\\n checked_items += 1\\n # 截取末尾 500 个字符，验证是否在结尾位置包含目标注释\\n tail_content = block_stripped[-500:]\\n if target_sentence not in tail_content:\\n return False\\n \\n # 2. 检查 workspace_path 中的生成的代码文件\\n if os.path.exists(workspace_path):\\n # 常见可能生成的文件后缀，依据指令推断SVG及相关常见语言\\n code_exts = {'.svg', '.py', '.js', '.html', '.css', '.xml'}\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read().strip()\\n if content:\\n checked_items += 1\\n # 截取末尾 500 个字符，验证是否在结尾位置包含目标注释\\n tail_content = content[-500:]\\n if target_sentence not in tail_content:\\n return False\\n except Exception:\\n # 忽略非文本或无法用utf-8读取的文件\\n continue\\n \\n # 3. 如果没找到任何代码块或代码文件，视为未生成所需代码或未满足要求\\n if checked_items == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"写完代码之后，绝不能对其进行任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的SVG代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n svg_contents = []\\n \\n # 1. 尝试从工作区路径中提取 .svg 文件内容\\n if workspace_path and os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n svg_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 尝试从 response 的 Markdown 代码块中提取 SVG 内容\\n # 匹配所有的代码块，寻找带有 <svg 标签的内容\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n if '<svg' in block.lower():\\n svg_contents.append(block)\\n \\n # 如果没有找到任何 SVG 代码文件或代码块，则无法满足要求\\n if not svg_contents:\\n return False\\n \\n # 3. 验证所有 SVG 内容中 `=` 的空格情况\\n for content in svg_contents:\\n if '=' not in content:\\n continue\\n \\n for i, char in enumerate(content):\\n if char == '=':\\n # 检查左侧是否有且仅有两个空格\\n if i < 2:\\n return False\\n if content[i-1] != ' ' or content[i-2] != ' ':\\n return False\\n if i >= 3 and content[i-3] == ' ':\\n return False\\n \\n # 检查右侧是否有且仅有两个空格\\n if i + 2 >= len(content):\\n return False\\n if content[i+1] != ' ' or content[i+2] != ' ':\\n return False\\n if i + 3 < len(content) and content[i+3] == ' ':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"接下来完善第二列“急诊医生初诊”。在急诊医生的泳道里，接收刚刚排队或转初诊的患者流，依次画出“问诊”、“体格检查”、“开具检查单”这三个矩形节点用箭头相连。在这个流程后加入一个“病情评估”的菱形判断框，并分出三条路：1. 如果判断“需急救”，画一条红线连接到刚才我们在专科团队那里画的“急救处理”节点；2. 如果判断“需进一步检查”，画一条线向下连到第三行检查科的对应列，先放一个占位节点“流向检查科”；3. 如果是“轻症可处理”，直接向右拉一条线到第五行行政管理的最后一列，放一个“开处方/出院”节点。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的写完代码之后绝不能对其进行任何测试的要求作废，改为写完代码后必须对所有的代码进行测试以确保测试通过；第二，你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_coding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 结合用户指令中的“SVG”等特征，定义需要检查的代码文件扩展名集合\\n code_extensions = {\\n '.svg', '.xml', '.py', '.js', '.ts', '.html', '.css', \\n '.java', '.c', '.cpp', '.cs', '.go', '.rs', '.php', '.rb', '.sh'\\n }\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files_found = False\\n \\n # 遍历当前代码项目地址下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对推断为代码类型的文件进行验证\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的顶行（第一行）\\n first_line = f.readline()\\n \\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法使用 utf-8 解码或存在其他读取错误，则判定为不满足\\n return False\\n \\n # 如果找到了代码文件且均满足要求，则返回 True；如果未生成任何代码文件，则视为未达成条件\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令背景推断，此处的代码文件主要是指 SVG 文件\\n if file.lower().endswith(\\\".svg\\\"):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n lines = f.readlines()\\n \\n # 移除末尾的空白行，以便准确获取实际代码或注释的尾部内容\\n while lines and not lines[-1].strip():\\n lines.pop()\\n \\n if not lines:\\n return False\\n \\n # 检查文件的结尾位置（最后10个有效非空行）内是否包含特定的注释字符串\\n found_in_tail = False\\n for line in lines[-10:]:\\n if target_comment in line:\\n found_in_tail = True\\n break\\n \\n if not found_in_tail:\\n return False\\n\\n except Exception:\\n # 出现无法读取等异常情况时，判定为失败\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你提供的SVG代码文件中，任何出现“=”的地方，其两边都必须各有两个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n svg_contents = []\\n \\n # 1. 遍历工作区寻找 .svg 文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.svg'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n svg_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 若工作区中没有 SVG 文件，尝试从回复中提取 svg 代码块作为兜底\\n if not svg_contents:\\n svg_blocks = re.findall(r'```svg\\\\s*(.*?)\\\\s*```', response, re.DOTALL)\\n svg_contents.extend(svg_blocks)\\n \\n # 如果依然没有任何 SVG 代码内容，则认定未满足输出 SVG 代码的前提\\n if not svg_contents:\\n return False\\n \\n # 3. 遍历所有 SVG 内容并检查 `=` 周围的空格\\n for content in svg_contents:\\n for i, char in enumerate(content):\\n if char == '=':\\n # 统计左边的连续空格数量\\n left_spaces = 0\\n j = i - 1\\n while j >= 0 and content[j] == ' ':\\n left_spaces += 1\\n j -= 1\\n \\n # 统计右边的连续空格数量\\n right_spaces = 0\\n k = i + 1\\n while k < len(content) and content[k] == ' ':\\n right_spaces += 1\\n k += 1\\n \\n # 检查是否都严格恰好各有两个空格\\n if left_spaces != 2 or right_spaces != 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"写完代码后必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n svg_contents = []\\n \\n # 1. 尝试在工作区查找并读取 .svg 文件的内容\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n svg_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 如果未找到任何文件，退而从 response 提取 SVG 代码块进行验证\\n if not svg_contents:\\n pattern = re.compile(r'```[^\\\\n]*\\\\n(.*?)```', re.DOTALL)\\n matches = pattern.findall(response)\\n for match in matches:\\n if '<svg' in match.lower() or target_str in match:\\n svg_contents.append(match)\\n \\n # 3. 如果既没有 .svg 文件也没有在回复中找到任何疑似的 SVG 代码内容，判定为失败\\n if not svg_contents:\\n return False\\n \\n # 4. 对收集到的每一个 SVG 内容进行校验\\n for content in svg_contents:\\n # 按换行切割并去掉空行\\n lines = [line.strip() for line in content.split('\\\\n') if line.strip()]\\n found = False\\n \\n # 仅检查开头的 5 个非空行\\n for line in lines[:5]:\\n if target_str in line:\\n # 确保是“另起一行”：不能与 <?xml 或 <svg 等主体标签混杂在同一行\\n if '<svg' not in line.lower() and '<?xml' not in line.lower():\\n found = True\\n break\\n \\n if not found:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"很棒！现在填补第三列“检查与诊断”的内容。在检查/检验科的泳道中，从刚才的“流向检查科”节点出发，平行画三个检查节点：“化验血液/尿液”、“影像学(X光/CT/MRI)”、“心电图监测”。将这三个节点汇总到一个“检查结果”的菱形判断框中。接着处理分支：如果“检查结果正常”，向上引一条线回到第二行急诊医生泳道的第五列位置，建一个“结合检查做诊断”的节点；如果“检查异常”，向下引一条线到专科/手术团队泳道，建一个“通知介入”的预警节点。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的你提供的SVG代码文件中任何出现“=”的地方其两边都必须各有两个空格的要求作废，改为生成的SVG代码文件中任何“=”周围都不允许带有空白字符；第二，你生成的所有SVG代码文件都必须使用UTF-8进行编码。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，目标代码文件主要为 SVG 文件\\n valid_extensions = {'.svg', '.xml', '.html', '.htm', '.py', '.js', '.css'}\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行（顶行）\\n first_line = f.readline()\\n \\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若文件无法被正常读取（如编码错误或没有读取权限），视作不符合要求\\n return False\\n\\n # 必须至少找到一个代码文件才可能视为成功，否则未生成目标代码判定为失败\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户指令推断，主要的代码文件类型为SVG文件以及可能的测试代码文件\\n valid_exts = {'.svg', '.py', '.html', '.js', '.sh', '.xml'}\\n \\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n code_files_found = True\\n \\n # 获取文件的所有非空行\\n lines = [line.strip() for line in content.splitlines() if line.strip()]\\n \\n if not lines:\\n return False\\n \\n # 检查目标文本是否出现在文件的末尾部分（最后 10 个非空行内）\\n found_in_end = False\\n for line in lines[-10:]:\\n if target_text in line:\\n found_in_end = True\\n break\\n \\n # 如果有任何一个代码文件的末尾没有包含该注释，则判为不满足\\n if not found_in_end:\\n return False\\n \\n # 只有当至少检查了一个代码文件，并且全部通过时，才返回 True\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的SVG代码文件中任何“=”周围都不允许带有空白字符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n svg_contents = []\\n \\n # 1. 从工作区路径中扫描并读取生成的 .svg 文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n svg_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 如果在工作区中没有找到 SVG 文件，降级从助手的回复内容中提取 SVG 代码块\\n if not svg_contents:\\n blocks = re.findall(r'```([a-zA-Z0-9]*)\\\\n(.*?)```', response, re.DOTALL)\\n for lang, code in blocks:\\n # 匹配语言类型为 svg、xml、html 或空，并确保内部含有 <svg 标签\\n if lang.lower() in ['svg', 'xml', 'html', ''] and '<svg' in code.lower():\\n svg_contents.append(code)\\n \\n # 3. 如果根本没有找到任何 SVG 代码内容，视为未满足生成要求\\n if not svg_contents:\\n return False\\n \\n # 4. 验证内容中是否存在“=”周围带有空白字符的情况\\n for content in svg_contents:\\n # \\\\s 匹配任何空白字符（空格、换行、制表符等）\\n # \\\\s= 匹配 '=' 前面有空白字符的情况\\n # =\\\\s 匹配 '=' 后面有空白字符的情况\\n if re.search(r'\\\\s=|=\\\\s', content):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"写完代码后必须对所有的代码进行测试以确保测试通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 收集所有的 .svg 文件路径\\n svg_files = []\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n svg_files.append(os.path.join(root, file))\\n\\n # 如果未找到任何 SVG 文件，视为未满足生成文件的要求\\n if not svg_files:\\n return False\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n\\n # 遍历所有 SVG 文件，检查是否在开头处包含指定注释\\n for file_path in svg_files:\\n has_target_comment = False\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取前 10 行以涵盖“开头处另起一行”的可能情况\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n \\n if target_comment in line:\\n has_target_comment = True\\n break\\n except Exception:\\n # 任何文件读取错误直接判定为不满足要求\\n return False\\n\\n # 如果某一个文件缺少该声明，则整体判定为失败\\n if not has_target_comment:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的所有SVG代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n svg_files = []\\n # 获取所有的 .svg 文件\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n svg_files.append(os.path.join(root, file))\\n \\n # 如果没有生成任何 SVG 文件，则无法证明满足了要求\\n if not svg_files:\\n return False\\n \\n for svg_file in svg_files:\\n try:\\n with open(svg_file, 'rb') as f:\\n content = f.read()\\n \\n # 尝试使用 UTF-8 进行解码以验证文件编码\\n text = content.decode('utf-8')\\n \\n # 检查 XML 声明中的 encoding 属性（如果存在）\\n # 例如: <?xml version=\\\"1.0\\\" encoding=\\\"UTF-8\\\"?>\\n xml_decl_match = re.search(r'<\\\\?xml[^>]+encoding\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']', text, re.IGNORECASE)\\n if xml_decl_match:\\n encoding = xml_decl_match.group(1).lower()\\n if encoding not in ['utf-8', 'utf8']:\\n return False\\n \\n except UnicodeDecodeError:\\n # 如果解码抛出异常，说明文件的实际字节流不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"这步我们来细化第四列“急救/重症处理”。在专科/手术团队的泳道里，扩充原有的“急救处理”节点，在旁边加上“心肺复苏、气道管理、用药抢救”等具体动作的文字或小节点。然后紧接着加一个“急救结果判断”的菱形框：1. 若“病情稳定”，画线向上汇入急诊医生的“结合检查做诊断”节点；2. 若“病情恶化”，画一条带有“再次急救”字样的折线，循环指回前方的“急救处理”节点；3. 若“无效”，直接画一条线引向一个黑色的终点节点，写上“宣布死亡（流程结束）”。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的写完代码后必须对所有的代码进行测试以确保测试通过的要求作废，改为代码一旦编写完毕，严禁进行任何形式的测试；第二，对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef verify_encoding_declaration(response: str, workspace_path: str) -> bool:\\n code_files = []\\n \\n # 遍历当前项目工作目录，查找所有的 .svg 代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.svg'):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有生成预期的SVG代码文件，视为未满足要求\\n if not code_files:\\n return False\\n \\n # 遍历所有查找到的代码文件，检查顶行内容\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行（顶行）\\n first_line = f.readline()\\n # 检查指定的编码声明是否包含在顶行中\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法读取或发生其他异常，保守判定为不满足\\n return False\\n \\n # 所有代码文件的顶行都包含了特定声明，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 依据用户指令推断代码文件类型，以 .svg, .html, .xml 为主\\n valid_extensions = ('.svg', '.html', '.xml')\\n code_files = []\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(valid_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有任何代码文件，无法证明要求被满足\\n if not code_files:\\n return False\\n \\n expected_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 去除文件内容整体为空的情况\\n if not content.strip():\\n return False\\n \\n # 获取所有非空行\\n lines = [line.strip() for line in content.split('\\\\n') if line.strip()]\\n \\n # 检查指定的注释文本是否出现在结尾位置（考察末尾 5 个非空行）\\n found_at_end = False\\n for line in lines[-5:]:\\n if expected_text in line:\\n found_at_end = True\\n break\\n \\n if not found_at_end:\\n return False\\n \\n except Exception:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的SVG代码文件中任何“=”周围都不允许带有空白字符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n svg_contents = []\\n\\n # 1. 从工作区中寻找生成的 .svg 文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n svg_contents.append(f.read())\\n except Exception:\\n pass\\n\\n # 2. 从 response 中提取可能存在的 SVG 代码块\\n code_blocks = re.findall(r'```(?:[a-zA-Z0-9_-]+)?\\\\s*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n # 只保留包含 svg 标签或者 xmlns 声明的代码块\\n if '<svg' in block.lower() or 'xmlns=' in block.lower():\\n svg_contents.append(block)\\n\\n # 3. 如果仍未找到，尝试直接提取裸 <svg> 标签对内容\\n if not svg_contents:\\n raw_svgs = re.findall(r'(<svg[\\\\s>].*?</svg>)', response, re.DOTALL | re.IGNORECASE)\\n svg_contents.extend(raw_svgs)\\n\\n # 如果完全没有找到任何 SVG 代码，视为无法满足生成目标要求\\n if not svg_contents:\\n return False\\n\\n # 4. 验证核心要求：任何 \\\"=\\\" 周围不允许存在空白字符 (\\\\s=|=\\\\s)\\n # \\\\s 会匹配空格、制表符 \\\\t、换行符 \\\\n 等\\n for text in svg_contents:\\n if re.search(r'\\\\s=|=\\\\s', text):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n target_str = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n svg_contents = []\\n \\n # 1. 尝试从当前代码工作区加载所有生成的 .svg 后缀的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n svg_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 如果工作区未发现已保存的 .svg 文件，则尝试在最近一轮的 response 寻找包含 SVG 源码的 Markdown 代码块\\n if not svg_contents:\\n blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in blocks:\\n # 判断该代码块是否具有 SVG 的基本标识\\n if '<svg' in block.lower():\\n svg_contents.append(block)\\n \\n # 若都没有，则代表其根本未生成或修改任何包含 SVG 的图表文件，验证失败\\n if not svg_contents:\\n return False\\n \\n # 3. 对每一个识别到的 SVG 文件内容进行检验\\n for content in svg_contents:\\n # 去除全空行，防止空行干扰开头行数的判断\\n lines = [line.strip() for line in content.split('\\\\n') if line.strip()]\\n \\n found_in_beginning = False\\n # 取文件的前20行作为“开头处”的标准范围\\n for line in lines[:20]:\\n if target_str in line:\\n # 去除可能的各种注释符号，以便判断它是否是另起一行的干净注释\\n clean_line = re.sub(r'<!--|-->|/\\\\*|\\\\*/|//|#', '', line)\\n clean_line = clean_line.replace(target_str, '').strip()\\n \\n # 若行内不再包含其他标签(如'<')，并且残留文本较短，则确认是“另起一行进行注释声明”\\n if len(clean_line) < 50 and '<' not in clean_line:\\n found_in_beginning = True\\n break\\n \\n # 只要有任何一个 SVG 文件不满足要求，则判定为失败\\n if not found_in_beginning:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的所有SVG代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n svg_files = []\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.svg'):\\n svg_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何SVG文件，则认为未完成生成代码文件的基础任务，返回 False\\n if not svg_files:\\n return False\\n \\n for file_path in svg_files:\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n \\n # 1. 检查物理文件是否能够以 UTF-8 成功解码\\n text = content.decode('utf-8')\\n \\n # 2. 如果文件包含 XML 声明，检查其 encoding 属性是否明确指定为 utf-8\\n # 例如匹配：<?xml version=\\\"1.0\\\" encoding=\\\"UTF-8\\\"?>\\n match = re.search(r'<\\\\?xml[^>]*encoding\\\\s*=\\\\s*[\\\\'\\\"]([^\\\\'\\\"]+)[\\\\'\\\"]', text, re.IGNORECASE)\\n if match:\\n encoding_value = match.group(1).lower()\\n if encoding_value not in ['utf-8', 'utf8']:\\n return False\\n \\n except UnicodeDecodeError:\\n # 如果解码抛出异常，说明文件物理上不是纯净的 UTF-8 编码\\n return False\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"代码一旦编写完毕，严禁进行任何形式的测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"逻辑越来越严密了。现在来完善第五列“医生诊断与决策”。在急诊医生泳道里，从“结合检查做诊断”这个节点出来，加上一个关键的“决策”菱形分支，引出四条去向的线：1. “门诊治疗”，向下连到第五行行政管理泳道的“办理出院结算”；2. “需住院”，连到行政泳道的“办理住院”节点；3. “需手术”，向下连到专科团队泳道的“准备手术”节点；4. “需转院”，连到行政泳道的“协调转院”节点。请把这些连线排布整齐，不要过度交叉打结。另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的代码一旦编写完毕严禁进行任何形式的测试的要求作废，改为编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过；第二，在对现有SVG文件进行修改之前，请务必先调用相关工具读取文件内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n target_statement = \\\"-*- coding: utf-8 -*-\\\"\\n \\n # 定义常见的代码文件扩展名，特别需包含对话中重点生成的 .svg 文件\\n # 同时过滤掉 .log 等非代码类文件\\n code_extensions = {\\n '.svg', '.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', \\n '.css', '.xml', '.c', '.cpp', '.h', '.hpp', '.java', '.go', \\n '.rs', '.php', '.rb', '.sh', '.bash'\\n }\\n \\n found_code_file = False\\n \\n # 优先检查工作区目录下的实际生成的代码文件\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline() # 读取第一行（顶行）\\n if target_statement not in first_line:\\n return False\\n except Exception:\\n return False\\n \\n # 若在工作区中找到了至少一个代码文件并且全经验证通过，则判定符合要求\\n if found_code_file:\\n return True\\n \\n # 如果工作区中不存在代码文件（可能尚未真实写盘），则退化为在AI回复内容中提取代码块检查\\n pattern = re.compile(r'```[^\\\\n]*\\\\n(.*?)```', re.DOTALL)\\n blocks = pattern.findall(response)\\n if not blocks:\\n # 如果既没有生成文件，也没有输出代码块，无法判定满足了生成代码的要求\\n return False\\n \\n for block in blocks:\\n lines = block.split('\\\\n')\\n if not lines:\\n return False\\n first_line = lines[0] # 获取代码块的顶行\\n if target_statement not in first_line:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n svg_found = False\\n\\n # 1. 检查 workspace_path 下的 .svg 代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n svg_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 过滤掉纯空行，方便提取“结尾位置”的有效内容\\n lines = [line.strip() for line in content.splitlines() if line.strip()]\\n if not lines:\\n return False\\n \\n # 检查目标文本是否出现在最后5个非空行中（兼容尾部含有多行注释标记如 --> 的情况）\\n if not any(target_text in line for line in lines[-5:]):\\n return False\\n except Exception:\\n return False\\n\\n # 2. 检查 response 中的 SVG 代码块\\n # 匹配 Markdown 格式的代码块\\n code_blocks = re.findall(r'```(?:[a-zA-Z0-9_-]+)?[ \\\\t]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n # 如果代码块中包含 svg 标签，则判定为生成的 svg 代码\\n if '<svg' in block or '</svg>' in block:\\n svg_found = True\\n lines = [line.strip() for line in block.splitlines() if line.strip()]\\n if not lines:\\n return False\\n \\n # 同样检查目标文本是否在代码块结尾位置\\n if not any(target_text in line for line in lines[-5:]):\\n return False\\n\\n # 如果找到了至少一个 SVG 文件/代码块且都通过了测试，则要求满足\\n return svg_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的SVG代码文件中任何“=”周围都不允许带有空白字符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查生成的SVG代码文件中任何“=”周围都不允许带有空白字符。\\n \\\"\\\"\\\"\\n has_svg_file = False\\n \\n # 匹配“=”前面或后面带有空白字符的情况（\\\\s 匹配任何空白字符，包括空格、制表符、换行符等）\\n # \\\\s= 匹配 \\\"[空白字符]=\\\"\\n # =\\\\s 匹配 \\\"=[空白字符]\\\"\\n pattern = re.compile(r'\\\\s=|=\\\\s')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.svg'):\\n has_svg_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果匹配到等号周围存在空白字符，则直接返回 False\\n if pattern.search(content):\\n return False\\n except Exception as e:\\n # 若文件读取失败，无法验证其是否满足要求，保守起见返回 False\\n return False\\n \\n # 如果找到了至少一个 SVG 文件，且所有 SVG 文件的等号周围均没有空白字符，则返回 True\\n return has_svg_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 收集所有的 SVG 代码文件\\n svg_files = []\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n svg_files.append(os.path.join(root, file))\\n\\n if not svg_files:\\n return False\\n\\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n\\n # 检查每一个 SVG 文件的开头是否包含要求的注释声明\\n for file_path in svg_files:\\n found_in_head = False\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 只需检查文件开头处，取前 10 行以涵盖各类声明\\n for line in lines[:10]:\\n if target_comment in line:\\n # 确保是“另起一行”，即行内不能掺杂实质性的 SVG 标签\\n if \\\"<svg\\\" not in line and \\\"<path\\\" not in line and \\\"<g\\\" not in line and \\\"<rect\\\" not in line:\\n found_in_head = True\\n break\\n except Exception:\\n return False\\n\\n if not found_in_head:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的所有SVG代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n svg_files_found = False\\n \\n # 遍历 workspace_path 找到所有 .svg 文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n svg_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制读取文件内容\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n \\n # 尝试以 utf-8 解码，如果抛出异常则说明并非使用 UTF-8 编码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 发现存在非 UTF-8 编码的 SVG 文件\\n return False\\n except Exception:\\n # 针对其他异常情况，例如文件无法读取，出于严谨也视作验证失败\\n return False\\n \\n # 至少要生成了 SVG 文件并且全都符合 UTF-8 编码要求，才返回 True\\n return svg_files_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在对现有SVG文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"最后一轮，把第六列“出院/住院/转院流程”的收尾工作做完，并全面美化UI。在行政管理泳道，将出院结算引向一个结束节点“流程结束”；将办理住院引向“病房分配”，再连向专科团队泳道的“住院治疗及护理交接”节点；把协调转院引向“转院完成（流程结束）”节点。同时在专科团队补充“实施手术”并连向“住院治疗”。美化方面：给所有矩形节点加上圆角和阴影，给判断框使用醒目的橙色或边框，确保所有文字排版居中不越界，连线统一改为平滑的折角线条(Orthogonal 风格)。在画布右上角加一个简单的图例说明（节点种类和红黄绿通道），让这张业务架构图看起来非常专业！\\n另外请你后续轮次始终遵循以下新增及调整的要求：第一，原有的生成的SVG代码文件中任何“=”周围都不允许带有空白字符的要求作废，改为在编写SVG代码文件时，请保证“=”的两端分别用两个空格隔开；第二，每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在生成针对用户的回复时，必须且只能一直采用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断，主要的生成代码文件是SVG\\n code_file_extensions = ['.svg']\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if any(file_name.endswith(ext) for ext in code_file_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行（顶行）\\n first_line = f.readline()\\n # 判断是否包含特定编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法读取或发生异常，视为不满足要求\\n return False\\n\\n # 若所有匹配的代码文件都满足要求，且至少找到了一个代码文件，则判定为True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在生成的每个代码文件的结尾位置换行写下注释，清晰地注明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件后缀，根据对话上下文主要考察 .svg\\n code_extensions = ['.svg', '.py', '.js', '.html', '.css', '.java', '.c', '.cpp', '.h']\\n \\n code_files = []\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有生成任何代码文件，则无法验证满足要求\\n if not code_files:\\n return False\\n \\n expected_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read().strip()\\n \\n if not content:\\n return False\\n \\n # 将代码按行分割，过滤掉纯空行\\n lines = content.split('\\\\n')\\n non_empty_lines = [line.strip() for line in lines if line.strip()]\\n \\n if not non_empty_lines:\\n return False\\n \\n # 提取最后几行进行检查，容忍多行注释（如XML/HTML的注释标签占用了不同行）\\n found_at_end = False\\n # 检查最后 5 行是否包含预期的注释文本\\n for line in reversed(non_empty_lines[-5:]):\\n if expected_comment in line:\\n found_at_end = True\\n break\\n \\n if not found_at_end:\\n return False\\n \\n except Exception:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件中严格禁止出现递归调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的代码中，务必确保所有 import/require 语句引入的包名严格遵循字母顺序升序排列\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试结束时，你应当在回复内严格遵守下面的XML规范，分别说明执行结果和描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"写代码时请杜绝过度设计，用最短路径去实现即可，千万别自作主张加入兜底或降级方案等需求外的东西\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在你生成的SVG代码文件里，请首选正斜杠 / 作为路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写SVG代码文件时，请保证“=”的两端分别用两个空格隔开\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n svg_files = []\\n # 查找工作区下所有的 .svg 文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.svg'):\\n svg_files.append(os.path.join(root, file))\\n\\n # 如果未找到任何 SVG 文件，则认为要求未满足\\n if not svg_files:\\n return False\\n \\n for file_path in svg_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 遍历文件内容，寻找所有 '=' 字符\\n for i, char in enumerate(content):\\n if char == '=':\\n # 提取 '=' 左侧的2个和3个字符\\n left_2 = content[i-2:i] if i >= 2 else content[:i]\\n left_3 = content[i-3:i] if i >= 3 else content[:i]\\n \\n # 提取 '=' 右侧的2个和3个字符\\n right_2 = content[i+1:i+3] if i + 3 <= len(content) else content[i+1:]\\n right_3 = content[i+1:i+4] if i + 4 <= len(content) else content[i+1:]\\n \\n # 验证左侧是否正好是两个空格\\n if left_2 != \\\" \\\":\\n return False\\n if len(left_3) == 3 and left_3[0] == \\\" \\\":\\n return False\\n \\n # 验证右侧是否正好是两个空格\\n if right_2 != \\\" \\\":\\n return False\\n if len(right_3) == 3 and right_3[-1] == \\\" \\\":\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个SVG代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n target_comment = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n svg_found = False\\n \\n # 1. 优先检查 workspace_path 下是否存在 .svg 文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.lower().endswith('.svg'):\\n svg_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 剔除开头空白字符后按行分割\\n lines = content.lstrip().split('\\\\n')\\n \\n # 检查开头处（这里取前10行作为判断开头的冗余范围）\\n comment_found = False\\n for line in lines[:10]:\\n if target_comment in line:\\n comment_found = True\\n break\\n \\n if not comment_found:\\n return False\\n except Exception:\\n return False\\n\\n # 如果在本地工作区内找到了至少一个 .svg 文件，并且全部通过了上述测试，则符合要求\\n if svg_found:\\n return True\\n \\n # 2. 如果工作区中没有提取到 .svg 文件，检查 response 文本中是否包含 SVG 代码块\\n svg_blocks = []\\n \\n # 匹配 markdown 格式的代码块\\n code_blocks = re.findall(r'```([a-zA-Z0-9]*)\\\\n(.*?)```', response, re.DOTALL)\\n for lang, code in code_blocks:\\n if lang.lower() == 'svg' or '<svg' in code.lower():\\n svg_blocks.append(code)\\n \\n # 如果没有找到 markdown 代码块，尝试匹配原生的 <svg> ... </svg> 标签\\n if not svg_blocks:\\n raw_svgs = re.findall(r'(<svg.*?</svg>)', response, re.DOTALL | re.IGNORECASE)\\n svg_blocks.extend(raw_svgs)\\n \\n # 如果既没有对应的生成文件，回复里也没有 SVG 代码，则无从验证，返回 False\\n if not svg_blocks:\\n return False\\n \\n # 检查从回复文本中提取到的所有 SVG 代码段\\n for block in svg_blocks:\\n lines = block.lstrip().split('\\\\n')\\n comment_found = False\\n for line in lines[:10]:\\n if target_comment in line:\\n comment_found = True\\n break\\n \\n # 只要存在任何一个 SVG 未按照要求声明，均视作失败\\n if not comment_found:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的所有SVG代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n has_svg = False\\n \\n # 遍历项目目录结构查找所有的 SVG 文件\\n for root, dirs, files in os.walk(workspace_path):\\n for filename in files:\\n if filename.lower().endswith('.svg'):\\n has_svg = True\\n file_path = os.path.join(root, filename)\\n try:\\n # 尝试以 UTF-8 编码读取该 SVG 文件的全部内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果触发了解码异常，证明文件并不是使用 UTF-8 编码的\\n return False\\n except Exception:\\n # 如果有其他的读取异常，出于严格验证的目的视为未满足要求\\n return False\\n \\n # 如果找到了至少一个 SVG 文件并且全部都通过了解码测试，则返回 True\\n return has_svg\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在对现有SVG文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看截屏，确认图表顶部居中位置是否显示有文字大标题“医疗急诊处理流程图”。","查看截屏，确认画布左侧划分为5个从上到下的横向区域（泳道），并写有角色/部门名称：分诊护士、急诊医生、检查/检验科、专科/手术团队、行政/出入院管理。","查看截屏，确认画布顶部划分了6列纵向表头阶段，依次为：患者到达/分诊判断、急诊医生初诊、检查与诊断、急救/重症处理、医生诊断与决策、出院/住院/转院流程。","查看截屏，确认各横向泳道具有交替区分的浅色背景，并且图表背景中有纵横的网格线（实线或虚线）作为排版辅助。","查看截屏分诊护士泳道（第一列区域），确认包含一个绿色的“患者到达”节点，该节点用线连接到“接待/登记/测体征”矩形节点，然后再连向一个“初步分级判断”菱形框。","查看截屏，确认“初步分级判断”菱形框向下引出了一条红色的连线，连线上标有“危重”字样，且最终向下穿透指向专科/手术团队泳道中的红色“急救处理”目标节点。","查看截屏，确认“初步分级判断”菱形框还引出了一条标有“严重”的黄色连线指向“转初诊”节点，以及一条标有“轻症”的绿色连线指向“排队候诊”节点。","查看截屏急诊医生泳道（第二列区域），确认在此区域中存在三个依次用箭头相连的矩形节点：“问诊”、“体格检查”、“开具检查单”。","查看截屏急诊医生泳道，确认第二列包含“病情评估”菱形判断框，且引出带有文字的三条连线：标有“需急救”的线（红线）、标有“需进一步检查”的线以及标有“轻症可处理”的线。","查看截屏检查/检验科泳道（第三列区域），确认存在三个上下平行的检查节点：“化验血液/尿液”、“影像学(X光/CT/MRI)”、“心电图监测”。","查看截屏，确认上述三个平行的检查节点都汇总向“检查结果”菱形框；且该菱形框引出标有“检查结果正常”的线指向上方急诊医生泳道的“结合检查做诊断”节点，引出标有“检查异常”的线指向下方专科团队的“通知介入”预警节点。","查看截屏专科/手术团队泳道（第四列区域），确认“急救处理”节点已被扩充，其旁边附带有“心肺复苏”、“气道管理”、“用药抢救”等具体急救动作的文字或小节点。","查看截屏，确认专科/手术团队泳道包含“急救结果判断”菱形框，且引出带有“病情恶化”字样的折线，折线上标有“再次急救”并向回循环指向了前方的“急救处理”节点。","查看截屏，确认“急救结果判断”菱形框引出的“病情稳定”线向上连入急诊医生的“结合检查做诊断”节点，而“无效”线连向一个独立的黑色节点并标有“宣布死亡”。","查看截屏急诊医生泳道（第五列区域），确认存在“决策”菱形分支框，并成功引出了四条去向的连线，线上的文字包含：门诊治疗、需住院、需手术、需转院。","查看截屏行政/出入院管理泳道（第六列区域），确认“门诊治疗”连向了“办理出院结算”再连向“流程结束”节点，同时“需转院”连向了“协调转院”再连向“转院完成”节点。","查看截屏，确认“办理住院”节点（连向病房分配）的后续路径，与专科团队区域的“实施手术”节点的后续路径，两者最终都连入了专科/手术团队泳道中的“住院治疗及护理交接”节点。","查看截屏节点及连线的UI样式，确认图表内所有矩形节点带有圆角和阴影效果，所有的判断框（菱形框）采用了醒目的颜色（如橙色）或醒目的边框，且所有连线均为正交风格的平滑折角线（即线段均为水平或垂直，拐角平滑）。","查看截屏右上角区域，确认存在一个图例说明面板，内容包含对节点种类的解释说明以及颜色通道（红、黄、绿）代表的含义说明。"],"string":"[\n \"查看截屏，确认图表顶部居中位置是否显示有文字大标题“医疗急诊处理流程图”。\",\n \"查看截屏，确认画布左侧划分为5个从上到下的横向区域（泳道），并写有角色/部门名称：分诊护士、急诊医生、检查/检验科、专科/手术团队、行政/出入院管理。\",\n \"查看截屏，确认画布顶部划分了6列纵向表头阶段，依次为：患者到达/分诊判断、急诊医生初诊、检查与诊断、急救/重症处理、医生诊断与决策、出院/住院/转院流程。\",\n \"查看截屏，确认各横向泳道具有交替区分的浅色背景，并且图表背景中有纵横的网格线（实线或虚线）作为排版辅助。\",\n \"查看截屏分诊护士泳道（第一列区域），确认包含一个绿色的“患者到达”节点，该节点用线连接到“接待/登记/测体征”矩形节点，然后再连向一个“初步分级判断”菱形框。\",\n \"查看截屏，确认“初步分级判断”菱形框向下引出了一条红色的连线，连线上标有“危重”字样，且最终向下穿透指向专科/手术团队泳道中的红色“急救处理”目标节点。\",\n \"查看截屏，确认“初步分级判断”菱形框还引出了一条标有“严重”的黄色连线指向“转初诊”节点，以及一条标有“轻症”的绿色连线指向“排队候诊”节点。\",\n \"查看截屏急诊医生泳道（第二列区域），确认在此区域中存在三个依次用箭头相连的矩形节点：“问诊”、“体格检查”、“开具检查单”。\",\n \"查看截屏急诊医生泳道，确认第二列包含“病情评估”菱形判断框，且引出带有文字的三条连线：标有“需急救”的线（红线）、标有“需进一步检查”的线以及标有“轻症可处理”的线。\",\n \"查看截屏检查/检验科泳道（第三列区域），确认存在三个上下平行的检查节点：“化验血液/尿液”、“影像学(X光/CT/MRI)”、“心电图监测”。\",\n \"查看截屏，确认上述三个平行的检查节点都汇总向“检查结果”菱形框；且该菱形框引出标有“检查结果正常”的线指向上方急诊医生泳道的“结合检查做诊断”节点，引出标有“检查异常”的线指向下方专科团队的“通知介入”预警节点。\",\n \"查看截屏专科/手术团队泳道（第四列区域），确认“急救处理”节点已被扩充，其旁边附带有“心肺复苏”、“气道管理”、“用药抢救”等具体急救动作的文字或小节点。\",\n \"查看截屏，确认专科/手术团队泳道包含“急救结果判断”菱形框，且引出带有“病情恶化”字样的折线，折线上标有“再次急救”并向回循环指向了前方的“急救处理”节点。\",\n \"查看截屏，确认“急救结果判断”菱形框引出的“病情稳定”线向上连入急诊医生的“结合检查做诊断”节点，而“无效”线连向一个独立的黑色节点并标有“宣布死亡”。\",\n \"查看截屏急诊医生泳道（第五列区域），确认存在“决策”菱形分支框，并成功引出了四条去向的连线，线上的文字包含：门诊治疗、需住院、需手术、需转院。\",\n \"查看截屏行政/出入院管理泳道（第六列区域），确认“门诊治疗”连向了“办理出院结算”再连向“流程结束”节点，同时“需转院”连向了“协调转院”再连向“转院完成”节点。\",\n \"查看截屏，确认“办理住院”节点（连向病房分配）的后续路径，与专科团队区域的“实施手术”节点的后续路径，两者最终都连入了专科/手术团队泳道中的“住院治疗及护理交接”节点。\",\n \"查看截屏节点及连线的UI样式，确认图表内所有矩形节点带有圆角和阴影效果，所有的判断框（菱形框）采用了醒目的颜色（如橙色）或醒目的边框，且所有连线均为正交风格的平滑折角线（即线段均为水平或垂直，拐角平滑）。\",\n \"查看截屏右上角区域，确认存在一个图例说明面板，内容包含对节点种类的解释说明以及颜色通道（红、黄、绿）代表的含义说明。\"\n]","isConversational":false}}},{"rowIdx":80,"cells":{"task_id":{"kind":"number","value":81,"string":"81"},"repository_policy":{"kind":"string","value":"以下是你后续所有轮次都必须严格遵循的约束条件：\n1. 在你生成的所有代码文件中，注释内容必须统一使用中文编写。\n2. 生成代码文件的缩进量必须固定为4个空格。\n3. 你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间。\n4. emoji表情符号 绝对不能出现在任何代码文件的注释内容中。\n5. 不要让你生成的任何代码文件里存在空行。\n6. 每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)。\n7. 你创建的每一个文件，其行数都应保持在200行或以下。\n8. 请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 包含了绝大部分可能使用 \"? :\" 三元运算符的常见编程语言后缀\n code_exts = {\n '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp',\n '.cs', '.js', '.ts', '.jsx', '.tsx', '.php', \n '.swift', '.go', '.rs', '.m', '.mm', '.rb', '.kt', '.scala', '.dart'\n }\n \n lines_to_check = []\n \n # 优先从 workspace_path 中获取生成的代码文件内容\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines_to_check.extend(f.readlines())\n except Exception:\n pass\n\n # 若工作区未提取到有效代码文件内容，则从助手的回复文本中提取 Markdown 代码块\n if not lines_to_check:\n code_blocks = re.findall(r'```.*?\\n(.*?)\\n```', response, re.DOTALL)\n for block in code_blocks:\n lines_to_check.extend(block.split('\\n'))\n\n # 若既未生成代码文件，回复中也无代码片段，则视为不满足条件\n if not lines_to_check:\n return False\n\n for line in lines_to_check:\n # 去除行尾换行符\n line_content = line.rstrip('\\n\\r')\n \n # 跳过空行（空行的有无由其他要求判断，此处专注处理\"缩进量\"）\n if not line_content.strip():\n continue\n \n # 剥离前导空白字符后的主体部分\n lstripped = line_content.lstrip()\n \n # 计算前导空白字符的长度和完整字符串\n leading_len = len(line_content) - len(lstripped)\n leading_whitespace = line_content[:leading_len]\n \n # 条件1：前导空白字符必须全部是由空格组成（不能含有 \\t 制表符等）\n if leading_whitespace != ' ' * leading_len:\n return False\n \n # 条件2：前导空格数量必须是 4 的倍数（0, 4, 8, ...）\n if leading_len % 4 != 0:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_python_funcs(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n source = f.read()\n tree = ast.parse(source)\n except Exception:\n return []\n \n counts = []\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n counts.append(node.end_lineno - node.lineno + 1)\n elif node.body:\n counts.append(node.body[-1].lineno - node.lineno + 1)\n return counts\n\n def check_c_like_functions(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n return []\n\n function_line_counts = []\n depth = 0\n in_function = False\n func_start_line = 0\n func_start_depth = 0\n has_entered_body = False\n \n in_str = False\n str_char = ''\n in_multiline_comment = False\n \n for i, line in enumerate(lines):\n line_num = i + 1\n \n # 使用简单的状态机清理字符串与注释，避免大括号统计受到干扰\n clean_chars = []\n j = 0\n while j < len(line):\n c = line[j]\n \n if in_multiline_comment:\n if c == '*' and j + 1 < len(line) and line[j+1] == '/':\n in_multiline_comment = False\n j += 1\n elif in_str:\n if c == '\\\\':\n j += 1 \n elif c == str_char:\n in_str = False\n else:\n if c in ('\"', \"'\", \"`\"):\n in_str = True\n str_char = c\n elif c == '/' and j + 1 < len(line) and line[j+1] == '*':\n in_multiline_comment = True\n j += 1\n elif c == '/' and j + 1 < len(line) and line[j+1] == '/':\n break \n else:\n clean_chars.append(c)\n \n j += 1\n \n clean_line = ''.join(clean_chars)\n \n open_braces = clean_line.count('{')\n close_braces = clean_line.count('}')\n \n if not in_function:\n # 排除普通的控制流关键字与 new 的对象创建等操作\n is_control = bool(re.search(r'\\b(if|for|while|switch|catch|class|interface|namespace|enum|struct)\\b', clean_line))\n is_new_call = bool(re.search(r'\\bnew\\b\\s+[\\w<>]+\\s*\\(', clean_line))\n has_parens = '(' in clean_line and ')' in clean_line\n \n if has_parens and not is_control and not is_new_call and (open_braces > 0 or (i + 1 < len(lines) and '{' in lines[i+1])):\n in_function = True\n func_start_line = line_num\n func_start_depth = depth\n has_entered_body = False\n \n depth += open_braces\n if in_function and depth > func_start_depth:\n has_entered_body = True\n \n depth -= close_braces\n \n if in_function and has_entered_body and depth == func_start_depth:\n function_line_counts.append(line_num - func_start_line + 1)\n in_function = False\n \n return function_line_counts\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in ['.py', '.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go', '.php', '.h', '.hpp', '.swift', '.kt', '.rs']:\n filepath = os.path.join(root, file)\n if ext == '.py':\n counts = check_python_funcs(filepath)\n else:\n counts = check_c_like_functions(filepath)\n \n for c in counts:\n if not (15 <= c <= 30):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 涵盖绝大多数常见编程语言与标记语言的代码文件后缀\n code_extensions = {\n '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.java', '.js', '.ts', '.jsx', '.tsx',\n '.cs', '.php', '.go', '.swift', '.rs', '.m', '.mm', '.py',\n '.dart', '.kt', '.scala', '.rb', '.sh', '.bash', '.pl',\n '.html', '.htm', '.css', '.xml', '.vue', '.svelte', '.yaml', '.yml'\n }\n \n # 编写正则表达式：匹配可能的注释和字符串内容\n # 通过将注释和字符串放到一起匹配，能有效避开字符串内出现的伪注释（如 \"http://...\"）\n pattern = re.compile(\n r'(?P<comment>/\\*[\\s\\S]*?\\*/|//[^\\n]*|#[^\\n]*|<!--[\\s\\S]*?-->)'\n r'|(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:[\\s\\S]*?)`)',\n re.MULTILINE\n )\n \n # 遍历工作区下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 匹配所有注释和字符串\n for match in pattern.finditer(content):\n comment_text = match.group('comment')\n \n # 如果匹配到的是注释组的内容\n if comment_text:\n # 逐字符检查是否是 emoji\n for char in comment_text:\n if is_emoji(char):\n return False\n except Exception:\n # 遇到非纯文本文件或编码错误时，安全跳过当前文件\n continue\n \n # 所有代码文件的注释中均未发现 emoji，返回满足条件\n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据包含三元运算符（? :）的要求，推断常见且支持该语法的代码文件扩展名\n code_extensions = {\n '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.hxx', \n '.java', '.js', '.ts', '.jsx', '.tsx', \n '.cs', '.php', '.swift', '.go', '.rs', '.dart'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n for line in lines:\n # 如果该行去除空白字符后为空，说明存在空行\n if not line.strip():\n return False\n except UnicodeDecodeError:\n # 如果 utf-8 解码失败，回退尝试 gbk 编码进行读取\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n lines = f.readlines()\n for line in lines:\n if not line.strip():\n return False\n except Exception:\n pass\n except Exception:\n pass\n \n return True","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义常见的代码文件扩展名\n code_exts = {\n '.py', '.java', '.c', '.cpp', '.cs', '.js', '.ts', '.go', \n '.php', '.rb', '.swift', '.kt', '.rs', '.dart', '.h', '.hpp', '.m'\n }\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return False\n \n # 预处理：移除 PHP 等标签中的问号干扰\n content = content.replace('<?php', '')\n content = content.replace('?>', '')\n content = content.replace('<?=', '')\n \n # 处理 Python 三引号多行字符串/注释的干扰\n content = re.sub(r\"'''(.*?)'''\", '', content, flags=re.DOTALL)\n content = re.sub(r'\"\"\"(.*?)\"\"\"', '', content, flags=re.DOTALL)\n \n # 移除代码中常见的单行注释、多行注释、以及各种引号包裹的字符串字面量\n pattern = re.compile(\n r'//.*?$|#.*?$|/\\*.*?\\*/|' +\n r\"'(?:\\\\.|[^\\\\'])*'|\" +\n r'\"(?:\\\\.|[^\\\\\"])*\"|' +\n r'`(?:\\\\.|[^\\\\`])*`',\n re.DOTALL | re.MULTILINE\n )\n cleaned_text = pattern.sub('', content)\n \n # 替换换行和不可见空白字符\n cleaned_text = re.sub(r'\\s+', ' ', cleaned_text)\n \n # 移除可选链和空值合并等合法的、非三元运算符的问号\n cleaned_text = cleaned_text.replace('?.', '')\n cleaned_text = cleaned_text.replace('??', '')\n \n # 验证：检查当前文件的纯代码中是否精确包含了 2 个 '?'\n if cleaned_text.count('?') != 2:\n return False\n \n # 寻找这两个 '?' 所在的位置\n idx1 = cleaned_text.find('?')\n idx2 = cleaned_text.find('?', idx1 + 1)\n \n # 三元运算符的严格结构还要求 '?' 后存在 ':'\n colon_after_idx1 = cleaned_text.find(':', idx1)\n colon_after_idx2 = cleaned_text.find(':', idx2)\n \n if colon_after_idx1 == -1 or colon_after_idx2 == -1:\n return False\n \n # 2个三元运算说明从第一个 '?' 开始算起，其后必须至少提供 2 个冒号供结构使用\n if cleaned_text.count(':', idx1) < 2:\n return False\n\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n has_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if not os.path.isfile(file_path):\n continue\n \n has_files = True\n \n # 尝试以 utf-8 编码读取文件并统计行数\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n if sum(1 for _ in f) > 200:\n return False\n except UnicodeDecodeError:\n # 如果 utf-8 失败，尝试以 gbk 编码读取\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n if sum(1 for _ in f) > 200:\n return False\n except Exception:\n # 如果仍然解码失败，可能为二进制文件，跳过\n continue\n except Exception:\n # 忽略其他读取异常（如权限问题等）\n continue\n \n # 如果存在文件且全部没有超过200行，则满足要求\n return has_files","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def check_end_of_line_comments(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件后缀及单行/行尾注释起始符\n comment_markers = {\n '.c': ['//', '/*'],\n '.cpp': ['//', '/*'],\n '.cc': ['//', '/*'],\n '.cxx': ['//', '/*'],\n '.h': ['//', '/*'],\n '.hpp': ['//', '/*'],\n '.java': ['//', '/*'],\n '.js': ['//', '/*'],\n '.ts': ['//', '/*'],\n '.jsx': ['//', '/*'],\n '.tsx': ['//', '/*'],\n '.cs': ['//', '/*'],\n '.go': ['//', '/*'],\n '.php': ['//', '/*', '#'],\n '.rs': ['//', '/*'],\n '.swift': ['//', '/*'],\n '.kt': ['//', '/*'],\n '.py': ['#'],\n '.rb': ['#'],\n '.sh': ['#'],\n '.sql': ['--'],\n '.lua': ['--']\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n def remove_strings_from_line(line: str) -> str:\n \"\"\"\n 简单剔除双引号、单引号和反引号中的内容，避免内部的注释符干扰（如 http://）。\n 将被引号包裹的字符以及转义字符替换为 'x'，以便准确匹配代码外的注释符。\n \"\"\"\n res = []\n in_double = False\n in_single = False\n in_backtick = False\n escape = False\n for char in line:\n if escape:\n res.append('x')\n escape = False\n continue\n if char == '\\\\':\n res.append('x')\n escape = True\n continue\n \n if char == '\"' and not in_single and not in_backtick:\n in_double = not in_double\n res.append(char)\n continue\n \n if char == \"'\" and not in_double and not in_backtick:\n in_single = not in_single\n res.append(char)\n continue\n \n if char == '`' and not in_double and not in_single:\n in_backtick = not in_backtick\n res.append(char)\n continue\n \n if not in_double and not in_single and not in_backtick:\n res.append(char)\n else:\n res.append('x')\n return \"\".join(res)\n\n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in comment_markers:\n has_code_files = True\n file_path = os.path.join(root, file)\n markers = comment_markers[ext]\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except (UnicodeDecodeError, FileNotFoundError):\n continue\n \n for line in lines:\n stripped = line.strip()\n if not stripped:\n continue # 跳过空行\n \n # 1. 检查是否为独立行注释\n # 如果一行去掉首尾空白后，直接以注释符开头，即判定为独立行注释\n is_independent = False\n for marker in markers:\n if stripped.startswith(marker):\n is_independent = True\n break\n \n if is_independent:\n return False\n \n # 2. 检查是否包含行尾注释\n # 必须包含至少一个 marker，且不是字符串字面量内部的假注释（如 URL 的 //）\n clean_line = remove_strings_from_line(stripped)\n has_comment = False\n for marker in markers:\n if marker in clean_line:\n has_comment = True\n break\n \n if not has_comment:\n return False\n \n return has_code_files","tags":["Style","Layout"]}],"string":"[\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 包含了绝大部分可能使用 \\\"? :\\\" 三元运算符的常见编程语言后缀\\n code_exts = {\\n '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp',\\n '.cs', '.js', '.ts', '.jsx', '.tsx', '.php', \\n '.swift', '.go', '.rs', '.m', '.mm', '.rb', '.kt', '.scala', '.dart'\\n }\\n \\n lines_to_check = []\\n \\n # 优先从 workspace_path 中获取生成的代码文件内容\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines_to_check.extend(f.readlines())\\n except Exception:\\n pass\\n\\n # 若工作区未提取到有效代码文件内容，则从助手的回复文本中提取 Markdown 代码块\\n if not lines_to_check:\\n code_blocks = re.findall(r'```.*?\\\\n(.*?)\\\\n```', response, re.DOTALL)\\n for block in code_blocks:\\n lines_to_check.extend(block.split('\\\\n'))\\n\\n # 若既未生成代码文件，回复中也无代码片段，则视为不满足条件\\n if not lines_to_check:\\n return False\\n\\n for line in lines_to_check:\\n # 去除行尾换行符\\n line_content = line.rstrip('\\\\n\\\\r')\\n \\n # 跳过空行（空行的有无由其他要求判断，此处专注处理\\\"缩进量\\\"）\\n if not line_content.strip():\\n continue\\n \\n # 剥离前导空白字符后的主体部分\\n lstripped = line_content.lstrip()\\n \\n # 计算前导空白字符的长度和完整字符串\\n leading_len = len(line_content) - len(lstripped)\\n leading_whitespace = line_content[:leading_len]\\n \\n # 条件1：前导空白字符必须全部是由空格组成（不能含有 \\\\t 制表符等）\\n if leading_whitespace != ' ' * leading_len:\\n return False\\n \\n # 条件2：前导空格数量必须是 4 的倍数（0, 4, 8, ...）\\n if leading_len % 4 != 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_python_funcs(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n source = f.read()\\n tree = ast.parse(source)\\n except Exception:\\n return []\\n \\n counts = []\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n counts.append(node.end_lineno - node.lineno + 1)\\n elif node.body:\\n counts.append(node.body[-1].lineno - node.lineno + 1)\\n return counts\\n\\n def check_c_like_functions(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n return []\\n\\n function_line_counts = []\\n depth = 0\\n in_function = False\\n func_start_line = 0\\n func_start_depth = 0\\n has_entered_body = False\\n \\n in_str = False\\n str_char = ''\\n in_multiline_comment = False\\n \\n for i, line in enumerate(lines):\\n line_num = i + 1\\n \\n # 使用简单的状态机清理字符串与注释，避免大括号统计受到干扰\\n clean_chars = []\\n j = 0\\n while j < len(line):\\n c = line[j]\\n \\n if in_multiline_comment:\\n if c == '*' and j + 1 < len(line) and line[j+1] == '/':\\n in_multiline_comment = False\\n j += 1\\n elif in_str:\\n if c == '\\\\\\\\':\\n j += 1 \\n elif c == str_char:\\n in_str = False\\n else:\\n if c in ('\\\"', \\\"'\\\", \\\"`\\\"):\\n in_str = True\\n str_char = c\\n elif c == '/' and j + 1 < len(line) and line[j+1] == '*':\\n in_multiline_comment = True\\n j += 1\\n elif c == '/' and j + 1 < len(line) and line[j+1] == '/':\\n break \\n else:\\n clean_chars.append(c)\\n \\n j += 1\\n \\n clean_line = ''.join(clean_chars)\\n \\n open_braces = clean_line.count('{')\\n close_braces = clean_line.count('}')\\n \\n if not in_function:\\n # 排除普通的控制流关键字与 new 的对象创建等操作\\n is_control = bool(re.search(r'\\\\b(if|for|while|switch|catch|class|interface|namespace|enum|struct)\\\\b', clean_line))\\n is_new_call = bool(re.search(r'\\\\bnew\\\\b\\\\s+[\\\\w<>]+\\\\s*\\\\(', clean_line))\\n has_parens = '(' in clean_line and ')' in clean_line\\n \\n if has_parens and not is_control and not is_new_call and (open_braces > 0 or (i + 1 < len(lines) and '{' in lines[i+1])):\\n in_function = True\\n func_start_line = line_num\\n func_start_depth = depth\\n has_entered_body = False\\n \\n depth += open_braces\\n if in_function and depth > func_start_depth:\\n has_entered_body = True\\n \\n depth -= close_braces\\n \\n if in_function and has_entered_body and depth == func_start_depth:\\n function_line_counts.append(line_num - func_start_line + 1)\\n in_function = False\\n \\n return function_line_counts\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ['.py', '.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go', '.php', '.h', '.hpp', '.swift', '.kt', '.rs']:\\n filepath = os.path.join(root, file)\\n if ext == '.py':\\n counts = check_python_funcs(filepath)\\n else:\\n counts = check_c_like_functions(filepath)\\n \\n for c in counts:\\n if not (15 <= c <= 30):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 涵盖绝大多数常见编程语言与标记语言的代码文件后缀\\n code_extensions = {\\n '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.java', '.js', '.ts', '.jsx', '.tsx',\\n '.cs', '.php', '.go', '.swift', '.rs', '.m', '.mm', '.py',\\n '.dart', '.kt', '.scala', '.rb', '.sh', '.bash', '.pl',\\n '.html', '.htm', '.css', '.xml', '.vue', '.svelte', '.yaml', '.yml'\\n }\\n \\n # 编写正则表达式：匹配可能的注释和字符串内容\\n # 通过将注释和字符串放到一起匹配，能有效避开字符串内出现的伪注释（如 \\\"http://...\\\"）\\n pattern = re.compile(\\n r'(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//[^\\\\n]*|#[^\\\\n]*|<!--[\\\\s\\\\S]*?-->)'\\n r'|(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:[\\\\s\\\\S]*?)`)',\\n re.MULTILINE\\n )\\n \\n # 遍历工作区下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 匹配所有注释和字符串\\n for match in pattern.finditer(content):\\n comment_text = match.group('comment')\\n \\n # 如果匹配到的是注释组的内容\\n if comment_text:\\n # 逐字符检查是否是 emoji\\n for char in comment_text:\\n if is_emoji(char):\\n return False\\n except Exception:\\n # 遇到非纯文本文件或编码错误时，安全跳过当前文件\\n continue\\n \\n # 所有代码文件的注释中均未发现 emoji，返回满足条件\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据包含三元运算符（? :）的要求，推断常见且支持该语法的代码文件扩展名\\n code_extensions = {\\n '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.hxx', \\n '.java', '.js', '.ts', '.jsx', '.tsx', \\n '.cs', '.php', '.swift', '.go', '.rs', '.dart'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n for line in lines:\\n # 如果该行去除空白字符后为空，说明存在空行\\n if not line.strip():\\n return False\\n except UnicodeDecodeError:\\n # 如果 utf-8 解码失败，回退尝试 gbk 编码进行读取\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n lines = f.readlines()\\n for line in lines:\\n if not line.strip():\\n return False\\n except Exception:\\n pass\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义常见的代码文件扩展名\\n code_exts = {\\n '.py', '.java', '.c', '.cpp', '.cs', '.js', '.ts', '.go', \\n '.php', '.rb', '.swift', '.kt', '.rs', '.dart', '.h', '.hpp', '.m'\\n }\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return False\\n \\n # 预处理：移除 PHP 等标签中的问号干扰\\n content = content.replace('<?php', '')\\n content = content.replace('?>', '')\\n content = content.replace('<?=', '')\\n \\n # 处理 Python 三引号多行字符串/注释的干扰\\n content = re.sub(r\\\"'''(.*?)'''\\\", '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', '', content, flags=re.DOTALL)\\n \\n # 移除代码中常见的单行注释、多行注释、以及各种引号包裹的字符串字面量\\n pattern = re.compile(\\n r'//.*?$|#.*?$|/\\\\*.*?\\\\*/|' +\\n r\\\"'(?:\\\\\\\\.|[^\\\\\\\\'])*'|\\\" +\\n r'\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|' +\\n r'`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.DOTALL | re.MULTILINE\\n )\\n cleaned_text = pattern.sub('', content)\\n \\n # 替换换行和不可见空白字符\\n cleaned_text = re.sub(r'\\\\s+', ' ', cleaned_text)\\n \\n # 移除可选链和空值合并等合法的、非三元运算符的问号\\n cleaned_text = cleaned_text.replace('?.', '')\\n cleaned_text = cleaned_text.replace('??', '')\\n \\n # 验证：检查当前文件的纯代码中是否精确包含了 2 个 '?'\\n if cleaned_text.count('?') != 2:\\n return False\\n \\n # 寻找这两个 '?' 所在的位置\\n idx1 = cleaned_text.find('?')\\n idx2 = cleaned_text.find('?', idx1 + 1)\\n \\n # 三元运算符的严格结构还要求 '?' 后存在 ':'\\n colon_after_idx1 = cleaned_text.find(':', idx1)\\n colon_after_idx2 = cleaned_text.find(':', idx2)\\n \\n if colon_after_idx1 == -1 or colon_after_idx2 == -1:\\n return False\\n \\n # 2个三元运算说明从第一个 '?' 开始算起，其后必须至少提供 2 个冒号供结构使用\\n if cleaned_text.count(':', idx1) < 2:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n has_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if not os.path.isfile(file_path):\\n continue\\n \\n has_files = True\\n \\n # 尝试以 utf-8 编码读取文件并统计行数\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n if sum(1 for _ in f) > 200:\\n return False\\n except UnicodeDecodeError:\\n # 如果 utf-8 失败，尝试以 gbk 编码读取\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n if sum(1 for _ in f) > 200:\\n return False\\n except Exception:\\n # 如果仍然解码失败，可能为二进制文件，跳过\\n continue\\n except Exception:\\n # 忽略其他读取异常（如权限问题等）\\n continue\\n \\n # 如果存在文件且全部没有超过200行，则满足要求\\n return has_files\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def check_end_of_line_comments(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件后缀及单行/行尾注释起始符\\n comment_markers = {\\n '.c': ['//', '/*'],\\n '.cpp': ['//', '/*'],\\n '.cc': ['//', '/*'],\\n '.cxx': ['//', '/*'],\\n '.h': ['//', '/*'],\\n '.hpp': ['//', '/*'],\\n '.java': ['//', '/*'],\\n '.js': ['//', '/*'],\\n '.ts': ['//', '/*'],\\n '.jsx': ['//', '/*'],\\n '.tsx': ['//', '/*'],\\n '.cs': ['//', '/*'],\\n '.go': ['//', '/*'],\\n '.php': ['//', '/*', '#'],\\n '.rs': ['//', '/*'],\\n '.swift': ['//', '/*'],\\n '.kt': ['//', '/*'],\\n '.py': ['#'],\\n '.rb': ['#'],\\n '.sh': ['#'],\\n '.sql': ['--'],\\n '.lua': ['--']\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n def remove_strings_from_line(line: str) -> str:\\n \\\"\\\"\\\"\\n 简单剔除双引号、单引号和反引号中的内容，避免内部的注释符干扰（如 http://）。\\n 将被引号包裹的字符以及转义字符替换为 'x'，以便准确匹配代码外的注释符。\\n \\\"\\\"\\\"\\n res = []\\n in_double = False\\n in_single = False\\n in_backtick = False\\n escape = False\\n for char in line:\\n if escape:\\n res.append('x')\\n escape = False\\n continue\\n if char == '\\\\\\\\':\\n res.append('x')\\n escape = True\\n continue\\n \\n if char == '\\\"' and not in_single and not in_backtick:\\n in_double = not in_double\\n res.append(char)\\n continue\\n \\n if char == \\\"'\\\" and not in_double and not in_backtick:\\n in_single = not in_single\\n res.append(char)\\n continue\\n \\n if char == '`' and not in_double and not in_single:\\n in_backtick = not in_backtick\\n res.append(char)\\n continue\\n \\n if not in_double and not in_single and not in_backtick:\\n res.append(char)\\n else:\\n res.append('x')\\n return \\\"\\\".join(res)\\n\\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in comment_markers:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n markers = comment_markers[ext]\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except (UnicodeDecodeError, FileNotFoundError):\\n continue\\n \\n for line in lines:\\n stripped = line.strip()\\n if not stripped:\\n continue # 跳过空行\\n \\n # 1. 检查是否为独立行注释\\n # 如果一行去掉首尾空白后，直接以注释符开头，即判定为独立行注释\\n is_independent = False\\n for marker in markers:\\n if stripped.startswith(marker):\\n is_independent = True\\n break\\n \\n if is_independent:\\n return False\\n \\n # 2. 检查是否包含行尾注释\\n # 必须包含至少一个 marker，且不是字符串字面量内部的假注释（如 URL 的 //）\\n clean_line = remove_strings_from_line(stripped)\\n has_comment = False\\n for marker in markers:\\n if marker in clean_line:\\n has_comment = True\\n break\\n \\n if not has_comment:\\n return False\\n \\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用 HTML5 Canvas 和 JavaScript 写一个网页版飞机大战的游戏骨架。首先，创建一个居中的黑色背景画布（比如 400x600）。在画布底部中央绘制一架玩家的飞机（不用找图片，用简单的几个几何图形组合一下，比如一个大三角或者矩形组合，看起来像战斗机就行）。加上键盘事件监听，使得玩家可以通过键盘的左右方向键或者 A/D 键左右平移飞机。注意要做好边界检测，飞机不能飞出画布的最左和最右侧。另外请你后续始终遵循以下要求：你必须在代码写完后测试全部的代码，并确保测试成功通过；同时请注意，你提供的代码文件中不能含有for循环语句。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对网页小游戏，定义可能生成的代码文件扩展名\n code_exts = {'.html', '.htm', '.js', '.css'}\n files_checked = 0\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n files_checked += 1\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n # 忽略不包含实际代码/注释文本的纯空白行\n if not line.strip():\n continue\n \n # 提取行首的前导空白字符（空格或制表符）\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_spaces = match.group(1)\n \n # 验证前导空白不能包含制表符\n if '\\t' in leading_spaces:\n return False\n \n # 验证前导空白的长度（空格数）必须固定为4的整数倍\n if len(leading_spaces) % 4 != 0:\n return False\n except Exception:\n # 忽略无法正常读取的非标准文件异常\n pass\n\n # 如果没有找到任何相关的代码文件，视为未生成项目代码，不满足要求\n if files_checked == 0:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def tokenize_js(js_code: str):\n tokens = []\n i = 0\n n = len(js_code)\n line_number = 1\n \n in_line_comment = False\n in_block_comment = False\n in_string = False\n string_char = ''\n \n while i < n:\n char = js_code[i]\n \n # 处理换行符，更新全局行号\n if char == '\\n':\n line_number += 1\n if in_line_comment:\n in_line_comment = False\n i += 1\n continue\n \n if in_line_comment:\n i += 1\n continue\n \n if in_block_comment:\n if char == '*' and i + 1 < n and js_code[i+1] == '/':\n in_block_comment = False\n i += 2\n else:\n if char == '\\n':\n line_number += 1\n i += 1\n continue\n \n if in_string:\n if char == '\\\\':\n if i + 1 < n:\n if js_code[i+1] == '\\n':\n line_number += 1\n i += 2\n else:\n i += 1\n elif char == string_char:\n in_string = False\n i += 1\n else:\n if char == '\\n':\n line_number += 1\n i += 1\n continue\n \n # 检测注释或正则表达式\n if char == '/' and i + 1 < n:\n if js_code[i+1] == '/':\n in_line_comment = True\n i += 2\n continue\n elif js_code[i+1] == '*':\n in_block_comment = True\n i += 2\n continue\n \n # 简单试探判断是否为正则表达式，以避免正则内的 {} 破坏括号匹配\n is_regex = False\n if tokens:\n prev_type, prev_val, _ = tokens[-1]\n if prev_type == 'PUNCT' and prev_val in '=(:,[?!&|':\n is_regex = True\n elif prev_type == 'WORD' and prev_val in ('return', 'case', 'yield'):\n is_regex = True\n \n if is_regex:\n start_i = i\n i += 1 \n in_char_class = False\n while i < n:\n if js_code[i] == '\\\\':\n if i + 1 < n:\n i += 2\n else:\n i += 1\n continue\n if js_code[i] == '[':\n in_char_class = True\n elif js_code[i] == ']':\n in_char_class = False\n elif js_code[i] == '/' and not in_char_class:\n i += 1\n break\n elif js_code[i] == '\\n':\n # 若遇换行，判断为并非正则表达式，回退\n i = start_i\n break\n i += 1\n if i > start_i:\n tokens.append(('STRING', '<regex>', line_number))\n continue\n \n # 检测字符串起始\n if char in (\"'\", '\"', '`'):\n in_string = True\n string_char = char\n tokens.append(('STRING', '<str>', line_number))\n i += 1\n continue\n \n if char.isspace():\n i += 1\n continue\n \n # 提取标识符或关键字\n if char.isalpha() or char in '_$':\n start = i\n while i < n and (js_code[i].isalnum() or js_code[i] in '_$'):\n i += 1\n word = js_code[start:i]\n tokens.append(('WORD', word, line_number))\n continue\n \n # 提取箭头函数双符组合\n if char == '=' and i + 1 < n and js_code[i+1] == '>':\n tokens.append(('PUNCT', '=>', line_number))\n i += 2\n continue\n \n tokens.append(('PUNCT', char, line_number))\n i += 1\n \n return tokens\n\n def check_functions(tokens):\n # 排除可能会有括号加大括号的控制流关键字\n control_keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\n has_func = False\n i = 0\n \n # 第一遍扫描，查找所有函数的起始大括号位置并打上 FUNC_BRACE 标记\n while i < len(tokens):\n tok_type, tok_val, line = tokens[i]\n \n # 1. 正常 function 关键字函数\n if tok_type == 'WORD' and tok_val == 'function':\n j = i + 1\n while j < len(tokens) and tokens[j][1] != '{':\n j += 1\n if j < len(tokens):\n tokens[j] = ('FUNC_BRACE', '{', tokens[j][2], line)\n i += 1\n continue\n \n # 2. 箭头函数\n if tok_type == 'PUNCT' and tok_val == '=>':\n j = i + 1\n if j < len(tokens):\n if tokens[j][1] == '{':\n tokens[j] = ('FUNC_BRACE', '{', tokens[j][2], line)\n else:\n # 这是一个直接返回的表达式的箭头函数，尝试向后寻找表达式边界来测算行数\n has_func = True\n exp_start_line = line\n k = j\n paren_depth = 0\n bracket_depth = 0\n brace_depth_local = 0\n exp_end_line = tokens[k][2]\n \n while k < len(tokens):\n t_val = tokens[k][1]\n if t_val == '(': paren_depth += 1\n elif t_val == ')': \n if paren_depth == 0: break\n paren_depth -= 1\n elif t_val == '[': bracket_depth += 1\n elif t_val == ']': \n if bracket_depth == 0: break\n bracket_depth -= 1\n elif t_val == '{': brace_depth_local += 1\n elif t_val == '}': \n if brace_depth_local == 0: break\n brace_depth_local -= 1\n elif t_val in (';', ','):\n if paren_depth == 0 and bracket_depth == 0 and brace_depth_local == 0:\n break\n exp_end_line = tokens[k][2]\n k += 1\n lines = exp_end_line - exp_start_line + 1\n \n if lines < 15 or lines > 30:\n return False, True\n i += 1\n continue\n \n # 3. 类方法、对象的方法 (例如: `update() { ... }`)\n if tok_type == 'WORD' and tok_val not in control_keywords:\n if i + 1 < len(tokens) and tokens[i+1][1] == '(':\n j = i + 1\n paren_depth = 0\n while j < len(tokens):\n if tokens[j][1] == '(': paren_depth += 1\n elif tokens[j][1] == ')': \n paren_depth -= 1\n if paren_depth == 0:\n break\n j += 1\n if j < len(tokens) and paren_depth == 0:\n if j + 1 < len(tokens) and tokens[j+1][1] == '{':\n tokens[j+1] = ('FUNC_BRACE', '{', tokens[j+1][2], line)\n i += 1\n\n # 第二遍扫描，进行基于括号层级的行数统计\n brace_depth = 0\n active_functions = [] \n \n for t in tokens:\n if t[0] == 'FUNC_BRACE':\n has_func = True\n brace_depth += 1\n # 记录 (当前层级, 函数起始行号)\n active_functions.append((brace_depth, t[3]))\n elif t[0] == 'PUNCT' and t[1] == '{':\n brace_depth += 1\n elif t[0] == 'PUNCT' and t[1] == '}':\n # 在遇到结尾 `}` 时，关闭当前层级对应的函数计算\n for func in active_functions[:]:\n if func[0] == brace_depth:\n func_start = func[1]\n func_end = t[2] \n lines = func_end - func_start + 1\n if lines < 15 or lines > 30:\n return False, True\n active_functions.remove(func)\n brace_depth -= 1\n \n return True, has_func\n\n # 主体逻辑：遍历所有的 JS 和 HTML 文件\n has_checked_any_function = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in ('.js', '.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except:\n continue\n \n blocks = []\n if ext == '.js':\n blocks.append(content)\n elif ext == '.html':\n script_pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\n for match in script_pattern.finditer(content):\n start_index = match.start(1)\n # 为了保障行号一致性，在提取出的 JS 块前部注入原本包含的相同换行数\n prefix = content[:start_index]\n newlines = '\\n' * prefix.count('\\n')\n blocks.append(newlines + match.group(1))\n \n for block in blocks:\n tokens = tokenize_js(block)\n ok, has_f = check_functions(tokens)\n if not ok:\n return False\n if has_f:\n has_checked_any_function = True\n\n # 必须至少有检查到一个函数，才可以视为成功通过\n return has_checked_any_function","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 针对代码中常见注释形式的正则表达式\n html_comment_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\n c_multi_comment_pattern = re.compile(r'/\\*(.*?)\\*/', re.DOTALL)\n c_single_comment_pattern = re.compile(r'//(.*)', re.MULTILINE)\n hash_single_comment_pattern = re.compile(r'#(.*)', re.MULTILINE)\n\n # 常见的代码文件后缀类型\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.py', '.php', '.java', '.c', '.cpp', '.h'}\n\n # 遍历代码项目地址\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取该文件内所有注释的内容\n html_comments = html_comment_pattern.findall(content)\n c_multi_comments = c_multi_comment_pattern.findall(content)\n c_single_comments = c_single_comment_pattern.findall(content)\n hash_single_comments = hash_single_comment_pattern.findall(content)\n\n # 合并所有注释\n all_comments = html_comments + c_multi_comments + c_single_comments + hash_single_comments\n\n # 遍历检查提取出的每一段注释文本\n for comment in all_comments:\n for char in comment:\n # 调用预定义的 is_emoji 专家函数判断字符\n if is_emoji(char):\n # 若发现任何一个 emoji 表情符号，则不满足要求\n return False\n \n # 如果全部遍历结束均未发现 emoji，则满足给定要求\n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check_no_empty_lines_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n\n # 验证工作空间路径是否有效\n if not workspace_path or not os.path.exists(workspace_path):\n return True\n\n # 推断可能的代码文件扩展名\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 检查是否为代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n \n # 尝试以不同编码读取文件内容\n lines = []\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except UnicodeDecodeError:\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n lines = f.readlines()\n except Exception:\n continue\n except Exception:\n continue\n \n # 检查是否存在空行\n for line in lines:\n if line.strip() == \"\":\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义可能的游戏相关代码文件类型\n code_extensions = {'.html', '.js', '.css', '.ts', '.htm'}\n files_checked = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 去除转义字符，避免转义引号破坏后续匹配\n content = re.sub(r'\\\\.', '', content)\n \n # 去除所有的字符串文字\n content = re.sub(r'\".*?\"', '', content, flags=re.DOTALL)\n content = re.sub(r\"'.*?'\", '', content, flags=re.DOTALL)\n content = re.sub(r\"`.*?`\", '', content, flags=re.DOTALL)\n \n # 去除所有的 HTML 注释、JS/CSS 的多行和单行注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n \n # 剔除可选链和空值合并语法引发的问号干扰\n content = content.replace('?.', '')\n content = content.replace('??', '')\n \n # 统计剩下的独立问号字符（即三元运算符的主标志）\n ternary_count = content.count('?')\n \n # 判断当前代码文件是否精确包含2个三元运算符\n if ternary_count != 2:\n return False\n \n files_checked += 1\n \n # 如果没查找到任何属于前端代码的文件，视为不合规\n if files_checked == 0:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def check_file_line_count(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作空间路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n has_checked_file = False\n \n # 遍历代码项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 逐行读取并统计总行数，避免大文件占用过多内存\n line_count = sum(1 for _ in f)\n \n # 若行数大于200，则不满足要求\n if line_count > 200:\n return False\n \n has_checked_file = True\n except UnicodeDecodeError:\n # 若文件不是文本文件（如图片等二进制格式），则忽略并跳过\n continue\n except Exception:\n # 忽略其他可能导致读取失败的异常（如权限问题等）\n continue\n\n # 确保至少检查了一个生成的代码文件\n return has_checked_file","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n valid_extensions = {'.html', '.js', '.css'}\n files_processed = False\n \n for root, _, files in os.walk(workspace_path):\n for file_name in files:\n if any(file_name.endswith(ext) for ext in valid_extensions):\n files_processed = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n stripped_line = line.strip()\n if not stripped_line:\n continue\n \n if (stripped_line.startswith('//') or \n stripped_line.startswith('/*') or \n stripped_line.startswith('*') or \n stripped_line.startswith('<!--')):\n return False\n \n has_inline_comment = False\n if '/*' in stripped_line or '<!--' in stripped_line:\n has_inline_comment = True\n elif '//' in stripped_line:\n idx = stripped_line.find('//')\n while idx != -1:\n if idx > 0 and stripped_line[idx-1] != ':':\n has_inline_comment = True\n break\n idx = stripped_line.find('//', idx + 2)\n \n if not has_inline_comment:\n return False\n \n return files_processed","tags":["Style","Layout"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请注意，你提供的代码文件中不能含有for循环语句","validation_code":"def check_for_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区目录不存在，可以认为是空项目，没有违反“不包含for循环”的要求\n if not os.path.exists(workspace_path):\n return True\n\n # 匹配 for 循环的正则表达式：for单词后接零或多个空白，再接可选的await及空白，最后是一个左圆括号\n for_pattern = re.compile(r'\\bfor\\s*(?:await\\s+)?\\(')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 针对本次网页游戏骨架生成，主要检查 HTML 和 JS 等代码相关文件\n if file.endswith(('.html', '.htm', '.js', '.ts', '.jsx', '.tsx')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于权限或编码等问题无法读取，直接跳过\n continue\n \n # 剔除 HTML 注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 剔除 JavaScript 多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 剔除 JavaScript 单行注释\n content = re.sub(r'//.*', '', content)\n \n # 剔除双引号字符串\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', content)\n # 剔除单引号字符串\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', content)\n # 剔除反引号（模板）字符串\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', content, flags=re.DOTALL)\n \n # 检查剥离字符串和注释后的内容是否存在 for 循环特征\n if for_pattern.search(content):\n return False\n\n return True","tags":["Style","Paradigm"]}]},{"round_id":1,"instruction":"好，接下来加入开火功能。当玩家按下空格键时，可以让飞机发射子弹。子弹用黄色的细长矩形表示，从飞机机头的位置生成，并向上方垂直飞行。为了让游戏体验更好，请增加射击的冷却时间（比如每 200 毫秒最多只能发射一发子弹，不能因为狂按空格就连成一条线）。此外，当子弹飞出画布上边缘时，必须将它从代码的数组里销毁掉，防止数组无限增长。另外请你后续始终遵循以下调整后的要求：原代码文件不能含有for循环的要求调整为你提供的代码文件中不得出现while循环；同时对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断代码文件可能的类型，排除.log等非代码文件\n code_exts = {'.html', '.js', '.css', '.ts'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_exts):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 忽略纯空行（关于空行的独立要求应由其它测试点负责）\n if not line.strip():\n continue\n \n # 匹配行首的空白字符序列\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_whitespace = match.group(1)\n \n # 缩进中不允许出现制表符(Tab)\n if '\\t' in leading_whitespace:\n return False\n \n # 缩进量必须是 4 个空格的倍数\n if len(leading_whitespace) % 4 != 0:\n return False\n\n # 若成功找到并检查过代码文件，且均未违规，则返回True\n return has_code_file","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def get_function_lengths(code, is_html=False):\n # 构建字符索引到行号的映射列表\n line_indices = []\n lines = code.split('\\n')\n for i, line in enumerate(lines):\n line_indices.extend([i] * (len(line) + 1))\n \n def replacer(m):\n return ' ' * len(m.group(0))\n \n # 替换掉所有的JS单行/多行注释以及字符串(以防止内部包含大括号导致误判)\n pattern = re.compile(\n r'//.*?$|/\\*.*?\\*/|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`',\n re.MULTILINE | re.DOTALL\n )\n \n if is_html:\n # 针对HTML文件，提取 <script>...</script> 中的内容\n script_pattern = re.compile(r'(<script[^>]*>)(.*?)(</script>)', re.IGNORECASE | re.DOTALL)\n blanked_html = list(' ' * len(code))\n for m in script_pattern.finditer(code):\n start, end = m.span(2)\n blanked_html[start:end] = list(code[start:end])\n code_to_clean = \"\".join(blanked_html)\n else:\n code_to_clean = code\n\n cleaned_code = pattern.sub(replacer, code_to_clean)\n \n # 匹配 JS 各种函数定义情况\n SIMPLE_ID = r'[a-zA-Z_$][a-zA-Z0-9_$]*'\n # 1. 传统的 function 声明\n p1 = rf'(?<![a-zA-Z0-9_$])function(?:\\s+{SIMPLE_ID})?\\s*\\(.*?\\)\\s*'\n # 2. 有括号的箭头函数 (args) => {\n p2 = r'\\(.*?\\)\\s*=>\\s*'\n # 3. 无括号的单参数箭头函数 arg => {\n p3 = rf'(?<![a-zA-Z0-9_$]){SIMPLE_ID}\\s*=>\\s*'\n # 4. 类/对象内的方法定义 method() {，需排除控制流关键字\n p4 = rf'(?<![a-zA-Z0-9_$])(?!(?:if|while|for|switch|catch|with)\\b){SIMPLE_ID}\\s*\\(.*?\\)\\s*'\n\n func_pattern = re.compile(rf'(?:{p1}|{p2}|{p3}|{p4})\\{{', re.DOTALL)\n \n # 获取所有函数的 '{' 起始索引位置\n func_starts = [m.end() - 1 for m in func_pattern.finditer(cleaned_code)]\n \n lengths = []\n for start_idx in func_starts:\n brace_level = 0\n end_idx = start_idx\n for i in range(start_idx, len(cleaned_code)):\n if cleaned_code[i] == '{':\n brace_level += 1\n elif cleaned_code[i] == '}':\n brace_level -= 1\n if brace_level == 0:\n end_idx = i\n break\n \n start_line = line_indices[start_idx]\n end_line = line_indices[end_idx]\n lengths.append(end_line - start_line + 1)\n \n return lengths\n\n all_lengths = []\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n all_lengths.extend(get_function_lengths(code, is_html=False))\n except Exception:\n pass\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n all_lengths.extend(get_function_lengths(code, is_html=True))\n except Exception:\n pass\n \n # 如果项目中没有任何函数，视为不满足游戏骨架的基本产出要求\n if not all_lengths:\n return False\n \n # 任何一个函数的行数必须介于15和30之间\n for length in all_lengths:\n if length < 15 or length > 30:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义目标代码文件的后缀名\n code_extensions = {'.html', '.htm', '.js', '.css'}\n\n # 构造正则表达式，用于匹配注释或字符串\n # (?P<comment>...) 捕获单行注释、多行块注释、HTML注释\n # (?P<string>...) 捕获双引号、单引号或反引号包裹的字符串，这可以防止诸如 http://... 被误认为注释\n pattern_str = (\n r'(?P<comment>//[^\\n]*|/\\*.*?\\*/|<!--.*?-->)'\n r'|(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n )\n # 使用 re.DOTALL 使得 .*? 能够匹配跨行内容（仅针对块注释和HTML注释，// 注释已被 [^\\n]* 限制）\n pattern = re.compile(pattern_str, re.DOTALL)\n\n # 遍历当前工作区的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于编码或其他问题无法读取，则跳过\n continue\n\n # 查找内容中所有的注释和字符串\n for match in pattern.finditer(content):\n comment = match.group('comment')\n # 如果匹配到的是注释内容\n if comment:\n # 遍历注释内容里的每一个字符，判断是否存在 emoji\n for char in comment:\n if is_emoji(char):\n return False # 一旦发现注释里有emoji，返回未满足要求\n\n # 所有代码文件的注释中均未发现emoji表情，返回符合要求\n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名白名单，涵盖前端代码及其他常见语言\n code_extensions = {\n '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.php', '.go', '.rs', '.swift', '.kt', '.sh'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历代码项目路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行检查\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 若一行代码去除两端空白字符（包含空格、制表符、换行符等）后长度为0，判定为空行\n if line.strip() == \"\":\n return False\n except Exception:\n # 忽略可能出现的文件读取异常（如编码错误等）\n continue\n \n # 只要存在至少一个代码文件，并且里面没有任何空行，就满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端相关代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx'}\n file_count = 0\n\n # 正则表达式：用于去除代码中的注释与字符串，避免将它们内部的内容误认为三元运算符\n # 包含了：多行注释、HTML注释、单行注释、双引号字符串、单引号字符串、反引号多行字符串\n clean_pattern = re.compile(\n r'/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->|//[^\\n]*|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n )\n \n # 正则表达式：匹配三元运算符的问号 ?\n # (?<!<) 排除 <? 的情况 (如XML声明等)\n # \\? 匹配问号本身\n # (?![?.]) 排除紧随其后的是 ? 或 . 的情况，进而排除可选链 `?.` 和空值合并 `??`\n ternary_pattern = re.compile(r'(?<!<)\\?(?![?.])')\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果有文件无法读取或者编码不兼容，按未通过要求处理\n return False\n \n # 第一步：清洗所有注释和字符串内容\n cleaned_content = clean_pattern.sub('', content)\n \n # 第二步：在清洗后的纯代码中匹配满足条件的三元运算符\n matches = ternary_pattern.findall(cleaned_content)\n \n # 若匹配数量不为 2，说明该文件未精确包含2个三元运算符\n if len(matches) != 2:\n return False\n \n # 如果找到了至少一个代码文件且全都符合要求，返回 True\n return file_count > 0","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n if len(lines) > 200:\n return False\n except UnicodeDecodeError:\n # 忽略可能存在的一些不可读取的二进制文件或非 utf-8 编码文件\n continue\n except Exception:\n # 忽略其他读取异常（例如权限问题等）\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对网页游戏项目，检查常规的代码文件，忽略如 .log 等非代码文件\n code_extensions = ['.html', '.js', '.css']\n has_code_files = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n stripped = line.strip()\n # 忽略空行\n if not stripped:\n continue\n \n # 1. 检查是否存在独立行注释（即以注释符号作为该行开头的内容）\n if stripped.startswith('//') or stripped.startswith('/*') or stripped.startswith('<!--'):\n return False\n \n # 2. 检查该行是否配有行尾注释（即包含至少一种注释标志）\n if '//' not in stripped and '/*' not in stripped and '<!--' not in stripped:\n return False\n\n # 必须至少包含一个代码文件且全部校验通过才算满足要求\n return has_code_files","tags":["Style","Layout"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码文件中不得出现while循环","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 如果工作空间路径不存在，视作没有代码文件，不违背不出现while的要求\n if not os.path.exists(workspace_path):\n return True\n \n # 遍历工作空间目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令，推断代码文件最可能为 HTML 和 JavaScript 文件\n if file.endswith(('.html', '.js')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找整个单词 while，防止匹配到类似于 while_loop 这样的变量名\n # 由于系统要求注释必须为中文，无需担心注释中的英文干扰\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 忽略读取异常的文件\n pass\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":2,"instruction":"现在来添加敌机系统。请编写一个定时器或利用帧循环，在画布顶部随机的横坐标位置不断生成敌机（可以用红色的方块表示）。生成后，敌机需要持续向下垂直移动。为了增加不可预测性，每架敌机的下落速度可以附带一点随机的差异，大小也可以有轻微的随机变化。同样，如果敌机飞出了画布的下边缘，也要及时从内存中清理掉它们。另外请你后续始终遵循以下调整和新增要求：原要求代码写完后必须测试全部代码并确保通过调整为编写代码结束后，绝对禁止你做任何测试；同时生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def check_indentation_rule(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的代码文件后缀，排除日志等非代码文件\n code_exts = {'.html', '.js', '.css'}\n has_code_files = False\n \n if not os.path.isdir(workspace_path):\n return False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 去除行末尾的换行符和回车符\n line_no_newline = line.rstrip('\\r\\n')\n \n # 匹配行首的前导空白字符\n match = re.match(r'^(\\s*)', line_no_newline)\n if match:\n leading_ws = match.group(1)\n # 如果没有前导空白（顶格写），属于合规的无缩进行为，不计入违反\n if not leading_ws:\n continue\n \n # 检查前导空白是否完全由普通空格组成（不能含有制表符 \\t 等）\n if not all(c == ' ' for c in leading_ws):\n return False\n \n # 检查连续空格的数量是否严格是 4 的倍数\n if len(leading_ws) % 4 != 0:\n return False\n \n return has_code_files","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"import os\nimport re\n\ndef check_requirements(response, workspace_path):\n js_files = []\n # 查找项目下所有的.js和.html文件\n for root, dirs, files in os.walk(workspace_path):\n for f in files:\n if f.endswith('.js') or f.endswith('.html'):\n js_files.append(os.path.join(root, f))\n \n if not js_files:\n return False\n \n def strip_comments_and_strings(code_str):\n # 简单状态机，替换注释和字符串内容为空格，以防正则被干扰\n # 0:normal, 1:单行注释, 2:多行注释, 3:单引号, 4:双引号, 5:模板字符串\n state = 0\n result = []\n i = 0\n n = len(code_str)\n while i < n:\n c = code_str[i]\n if state == 0:\n if c == '/' and i + 1 < n and code_str[i+1] == '/':\n state = 1\n result.append(' ')\n i += 2\n continue\n elif c == '/' and i + 1 < n and code_str[i+1] == '*':\n state = 2\n result.append(' ')\n i += 2\n continue\n elif c == \"'\":\n state = 3\n result.append(' ')\n i += 1\n continue\n elif c == '\"':\n state = 4\n result.append(' ')\n i += 1\n continue\n elif c == '`':\n state = 5\n result.append(' ')\n i += 1\n continue\n else:\n result.append(c)\n i += 1\n elif state == 1:\n if c == '\\n':\n state = 0\n result.append(c)\n else:\n result.append(' ')\n i += 1\n elif state == 2:\n if c == '*' and i + 1 < n and code_str[i+1] == '/':\n state = 0\n result.append(' ')\n i += 2\n continue\n elif c == '\\n':\n result.append(c)\n i += 1\n else:\n result.append(' ')\n i += 1\n elif state == 3:\n if c == '\\\\' and i + 1 < n:\n if code_str[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n continue\n elif c == \"'\":\n state = 0\n result.append(' ')\n i += 1\n elif c == '\\n':\n state = 0\n result.append(c)\n i += 1\n else:\n result.append(' ')\n i += 1\n elif state == 4:\n if c == '\\\\' and i + 1 < n:\n if code_str[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n continue\n elif c == '\"':\n state = 0\n result.append(' ')\n i += 1\n elif c == '\\n':\n state = 0\n result.append(c)\n i += 1\n else:\n result.append(' ')\n i += 1\n elif state == 5:\n if c == '\\\\' and i + 1 < n:\n if code_str[i+1] == '\\n':\n result.append(' \\n')\n else:\n result.append(' ')\n i += 2\n continue\n elif c == '`':\n state = 0\n result.append(' ')\n i += 1\n elif c == '\\n':\n result.append(c)\n i += 1\n else:\n result.append(' ')\n i += 1\n return \"\".join(result)\n \n has_functions = False\n \n for file_path in js_files:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n if file_path.endswith('.html'):\n script_blocks = re.findall(r'<script\\b[^>]*>(.*?)</script>', code, flags=re.DOTALL | re.IGNORECASE)\n blocks = script_blocks\n else:\n blocks = [code]\n \n for block in blocks:\n text = strip_comments_and_strings(block)\n function_braces = {}\n \n def get_line_num(t, pos):\n return t.count('\\n', 0, pos) + 1\n\n def find_matching_brace(t, start_pos):\n count = 0\n for i in range(start_pos, len(t)):\n if t[i] == '{':\n count += 1\n elif t[i] == '}':\n count -= 1\n if count == 0:\n return i\n return -1\n\n def add_function(start_pos, brace_pos):\n if brace_pos in function_braces:\n old_start = function_braces[brace_pos][0]\n new_start = get_line_num(text, start_pos)\n end_line = function_braces[brace_pos][1]\n function_braces[brace_pos] = (min(old_start, new_start), end_line)\n else:\n end_brace_pos = find_matching_brace(text, brace_pos)\n if end_brace_pos != -1:\n start_line = get_line_num(text, start_pos)\n end_line = get_line_num(text, end_brace_pos)\n function_braces[brace_pos] = (start_line, end_line)\n \n # 匹配模式1: function 关键字声明\n for match in re.finditer(r'\\bfunction\\b', text):\n brace_match = re.search(r'\\{', text[match.end():])\n if brace_match:\n sub_text = text[match.end(): match.end() + brace_match.start()]\n if '{' not in sub_text and '}' not in sub_text and ';' not in sub_text:\n brace_pos = match.end() + brace_match.start()\n add_function(match.start(), brace_pos)\n \n # 匹配模式2: 箭头函数\n for match in re.finditer(r'=>', text):\n sub_text = text[match.end():]\n m = re.match(r'\\s*\\{', sub_text)\n if m:\n brace_pos = match.end() + sub_text.find('{')\n add_function(match.start(), brace_pos)\n else:\n start_line = get_line_num(text, match.start())\n # 对单表达式箭头函数用负数偏移作为占位键\n function_braces[-match.start() - 1] = (start_line, start_line)\n\n # 匹配模式3: ES6 方法或对象方法\n for match in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{', text):\n method_name = match.group(1)\n # 排除语法结构上的关键字干扰\n if method_name not in ['if', 'for', 'while', 'switch', 'catch', 'function', 'with', 'return']:\n brace_pos = match.start() + match.group(0).find('{')\n add_function(match.start(), brace_pos)\n \n for key, (start, end) in function_braces.items():\n has_functions = True\n lines = end - start + 1\n if lines < 15 or lines > 30:\n return False\n \n return has_functions","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 预设不同类型的注释正则模式\n html_comment_re = re.compile(r'<!--(.*?)-->', re.DOTALL)\n block_comment_re = re.compile(r'/\\*(.*?)\\*/', re.DOTALL)\n line_comment_re = re.compile(r'//(.*)$', re.MULTILINE)\n hash_comment_re = re.compile(r'#(.*)$', re.MULTILINE)\n \n # 定义属于代码文件的后缀和 Markdown 代码块语言标识\n code_extensions = {\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.java', '.c', '.cpp', \n '.cs', '.go', '.py', '.rb', '.sh', '.php'\n }\n \n code_langs = {\n 'html', 'xml', 'js', 'javascript', 'jsx', 'ts', 'typescript', 'tsx',\n 'css', 'scss', 'less', 'java', 'c', 'cpp', 'cs', 'go', 'py', 'python',\n 'rb', 'ruby', 'sh', 'bash', 'php'\n }\n \n def check_comments_for_emoji(comments: list) -> bool:\n \"\"\"检查提取出来的注释列表中是否含有emoji\"\"\"\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n return True\n\n def extract_and_check(content: str, ext_or_lang: str) -> bool:\n \"\"\"根据不同的文件扩展名或语言类型提取注释并验证\"\"\"\n comments = []\n ext_or_lang = ext_or_lang.lower()\n \n # HTML 往往内嵌JS和CSS，因此将各种注释格式都提出来\n if ext_or_lang in {'.html', '.htm', 'html', 'xml'}:\n comments.extend(html_comment_re.findall(content))\n comments.extend(block_comment_re.findall(content))\n comments.extend(line_comment_re.findall(content))\n # 脚本或偏向后端的语言处理\n elif ext_or_lang in {'.py', '.rb', '.sh', 'py', 'python', 'rb', 'ruby', 'sh', 'bash'}:\n comments.extend(hash_comment_re.findall(content))\n # JS、CSS、TS等常规前端与C系语言处理\n else:\n comments.extend(block_comment_re.findall(content))\n comments.extend(line_comment_re.findall(content))\n if ext_or_lang in {'.php', 'php'}:\n comments.extend(hash_comment_re.findall(content))\n \n return check_comments_for_emoji(comments)\n\n # 1. 检查 workspace_path 下生成的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对代码文件进行注释排查（跳过.log等文件）\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n if not extract_and_check(content, ext):\n return False\n except Exception:\n pass\n \n # 2. 检查 response 回复文本中的代码块\n for match in re.finditer(r'```([a-zA-Z0-9]*)\\n(.*?)```', response, re.DOTALL):\n lang = match.group(1).lower()\n code_content = match.group(2)\n \n # 针对匹配到语言标签属于代码，或者根本没写语言标签的代码块均做检查\n if lang in code_langs or not lang:\n ext_lang = lang if lang else 'js'\n if not extract_and_check(code_content, ext_lang):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能的代码文件后缀\n code_extensions = {'.html', '.htm', '.js', '.css'}\n \n # 如果工作区路径不存在，直接返回 True（或视为未违背规则）\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件，忽略日志文件等其他格式\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行首尾的空格、换行符等空白字符，若为空则判定为空行\n if not line.strip():\n return False\n except Exception:\n # 读取出错或编码不支持时，跳过该文件继续检查\n continue\n \n # 所有代码文件遍历完毕未发现空行，满足要求\n return True","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def evaluate_response(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 根据用户指令推断可能生成的代码文件扩展名\n code_extensions = ['.html', '.js', '.css', '.ts', '.jsx', '.tsx']\n has_code_file = False\n \n def count_ternary_operators(code: str) -> int:\n # 去除 HTML 注释\n code = re.sub(r'<!--[\\s\\S]*?-->', '', code)\n # 去除 JS/CSS 多行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n # 去除 JS 单行注释\n code = re.sub(r'//.*', '', code)\n \n # 移除字符串 (双引号，单引号，反引号)\n # 采用高效且防止回溯灾难的正则表达式写法\n code = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '\"\"', code)\n code = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", \"''\", code)\n code = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '``', code)\n \n # 排除 JavaScript/TypeScript 特有的含有 ? 的非法三元运算符情况\n # 1. 可选链操作符: ?.\n code = code.replace('?.', '')\n # 2. 空值合并操作符: ??\n code = code.replace('??', '')\n \n # 统计纯逻辑语句中剩下的 '?' 符号的数量，即三元运算符的数量\n return code.count('?')\n\n # 遍历当前生成的项目路径\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理推断出的代码文件类型\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 每一个代码文件都必须精确包含 2 个三元运算符\n if count_ternary_operators(content) != 2:\n return False\n \n # 如果没有找到任何代码文件，则也视为不符合要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作空间路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 遍历项目中的每一个文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件，忽略可能出现的编码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n line_count = sum(1 for _ in f)\n # 如果任何文件行数大于 200，则不满足要求\n if line_count > 200:\n return False\n except Exception:\n # 若读取发生异常，则跳过该文件继续检查\n continue\n \n # 所有文件行数均不超过 200，满足要求\n return True","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\n files_checked = 0\n \n # 首先检查工作区中的实际代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n files_checked += 1\n for line in lines:\n s = line.strip()\n # 忽略空行\n if not s:\n continue\n \n # 检查是否为被禁止的独立行注释 (即当前行直接以注释符开头)\n if s.startswith(\"//\") or s.startswith(\"/*\") or s.startswith(\"<!--\"):\n return False\n \n has_comment = False\n \n # 检查是否包含单行注释 //，并排除网址如 http:// 的干扰\n idx = s.find(\"//\")\n while idx != -1:\n if idx > 0 and s[idx-1] == ':':\n idx = s.find(\"//\", idx + 2)\n else:\n has_comment = True\n break\n \n # 检查是否包含块注释或HTML注释\n if not has_comment and \"/*\" in s:\n has_comment = True\n if not has_comment and \"<!--\" in s:\n has_comment = True\n \n # 如果当前行没有任何注释符，则视为没有配有行尾注释，违反要求\n if not has_comment:\n return False\n \n # 如果工作区中有代码文件并且全都通过了检查，返回 True\n if files_checked > 0:\n return True\n\n # 如果工作区未找到相关代码文件，则回退(fallback)检查 response 中的代码块\n code_blocks = re.findall(r'```(.*?)\\n(.*?)```', response, re.DOTALL)\n valid_blocks = 0\n \n for lang, block in code_blocks:\n lang = lang.strip().lower()\n # 仅针对前端代码块及未标明语言的代码块进行验证\n if lang in ['html', 'js', 'javascript', 'css', 'ts', 'typescript', 'jsx', 'tsx', '']:\n valid_blocks += 1\n lines = block.split('\\n')\n for line in lines:\n s = line.strip()\n if not s:\n continue\n \n if s.startswith(\"//\") or s.startswith(\"/*\") or s.startswith(\"<!--\"):\n return False\n \n has_comment = False\n idx = s.find(\"//\")\n while idx != -1:\n if idx > 0 and s[idx-1] == ':':\n idx = s.find(\"//\", idx + 2)\n else:\n has_comment = True\n break\n \n if not has_comment and \"/*\" in s:\n has_comment = True\n if not has_comment and \"<!--\" in s:\n has_comment = True\n \n if not has_comment:\n return False\n\n return valid_blocks > 0","tags":["Style","Layout"]},{"constraint":"编写代码结束后，绝对禁止你做任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码文件中不得出现while循环","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义可能包含代码的扩展名\n code_extensions = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除 JavaScript/TypeScript 的单行注释 (// ...)\n content = re.sub(r'//.*', '', content)\n # 移除 JavaScript/TypeScript 的多行注释 (/* ... */)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 移除 HTML 注释 (<!-- ... -->)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 正则匹配 while 循环特征，例如 while( 或者是 while (\n if re.search(r'\\bwhile\\s*\\(', content):\n return False\n except Exception:\n # 如果文件无法按 UTF-8 读取或发生其他读取异常，跳过该文件继续验证\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"下一步是核心的战斗结算。请实现子弹和敌机之间的矩形碰撞检测。如果子弹打中敌机，子弹和敌机都要立刻被销毁。同时，在画布左上角增加一个白色的得分面板。为了让击杀反馈更爽快，请在敌机被击毁的位置，短暂地显示一个『+10』的文字特效，要求这个加分文字会缓缓向上飘移，并且在飘移过程中渐渐变透明直到完全消失。另外请你后续始终遵循以下调整和新增要求：原要求生成代码后务必使用ESLint进行检查并清除所有提示的报错调整为对于你产出的代码，严禁使用ESLint进行校验；同时，你所生成的所有代码文件都必须使用UTF-8进行编码。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义目标代码文件的后缀名\n code_extensions = {'.html', '.js', '.css'}\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except UnicodeDecodeError:\n # 尝试读取时忽略无法解析的特殊字符\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n continue\n\n for line in lines:\n # 忽略没有任何内容的空行（虽然系统提示禁止空行，但在此仅判断缩进要求）\n if not line.strip():\n continue\n\n leading_spaces = 0\n has_tab = False\n \n # 统计该行的前导空白字符情况\n for char in line:\n if char == ' ':\n leading_spaces += 1\n elif char == '\\t':\n has_tab = True\n break\n else:\n # 遇到非空格和非制表符（即代码内容的开始），停止统计\n break\n \n # 如果包含了制表符，或者前导空格的数量不是4的整数倍，直接判定失败\n if has_tab:\n return False\n if leading_spaces % 4 != 0:\n return False\n\n # 必须要有至少一个代码文件且全部校验通过才算真正满足了条件\n return has_code_file","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n all_functions_valid = True\n found_any_function = False\n\n def get_function_start_line(tokens, arrow_idx):\n if arrow_idx == 0:\n return tokens[arrow_idx][2]\n prev_type, prev_val, prev_line = tokens[arrow_idx - 1]\n \n if prev_type == 'identifier':\n return prev_line\n elif prev_type == ')':\n paren_count = 1\n curr_idx = arrow_idx - 2\n while curr_idx >= 0:\n ctype, cval, cline = tokens[curr_idx]\n if ctype == ')':\n paren_count += 1\n elif ctype == '(':\n paren_count -= 1\n if paren_count == 0:\n return cline\n curr_idx -= 1\n return prev_line \n \n return tokens[arrow_idx][2]\n\n def is_function_brace(tokens, brace_idx):\n if brace_idx == 0:\n return False, -1\n \n prev_idx = brace_idx - 1\n prev_type, prev_val, prev_line = tokens[prev_idx]\n \n # 匹配箭头函数\n if prev_type == '=>':\n return True, get_function_start_line(tokens, prev_idx)\n \n # 匹配 function () { 或者 function foo { (尽管不合法但若存在则算作函数) \n if prev_type == 'identifier' and prev_val == 'function':\n return True, prev_line\n \n # 匹配带有参数括号的情况: identifier(args) { ... }\n if prev_type == ')':\n paren_count = 1\n curr_idx = prev_idx - 1\n while curr_idx >= 0:\n ctype, cval, cline = tokens[curr_idx]\n if ctype == ')':\n paren_count += 1\n elif ctype == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n curr_idx -= 1\n \n if paren_count == 0:\n if curr_idx > 0:\n before_paren_type, before_paren_val, before_paren_line = tokens[curr_idx - 1]\n if before_paren_type == 'identifier':\n if before_paren_val == 'function':\n return True, before_paren_line\n # 排除常见控制流之后，剩下的通常是方法名称或普通具名函数名\n elif before_paren_val not in ('if', 'for', 'while', 'switch', 'catch', 'with'):\n if curr_idx > 1 and tokens[curr_idx - 2][1] == 'function':\n return True, tokens[curr_idx - 2][2]\n return True, before_paren_line\n elif before_paren_type == '=>':\n pass\n \n # 兼容匿名闭包 (function() {...})()\n if curr_idx + 1 < prev_idx and tokens[curr_idx + 1][1] == 'function':\n return True, tokens[curr_idx + 1][2]\n \n return False, -1\n\n def process_js(js_code):\n nonlocal all_functions_valid, found_any_function\n lines = js_code.split('\\n')\n tokens = []\n \n in_block_comment = False\n in_string = False\n string_char = ''\n \n # 将代码转化为词法 tokens 列表并过滤干扰项\n i = 0\n while i < len(lines):\n line = lines[i]\n j = 0\n while j < len(line):\n if in_block_comment:\n if j + 1 < len(line) and line[j:j+2] == '*/':\n in_block_comment = False\n j += 2\n continue\n j += 1\n continue\n \n if in_string:\n if line[j] == '\\\\':\n j += 2\n continue\n if line[j] == string_char:\n in_string = False\n j += 1\n continue\n \n if j + 1 < len(line) and line[j:j+2] == '//':\n break \n \n if j + 1 < len(line) and line[j:j+2] == '/*':\n in_block_comment = True\n j += 2\n continue\n \n if line[j] in ('\"', \"'\", '`'):\n in_string = True\n string_char = line[j]\n j += 1\n continue\n \n c = line[j]\n if c in '{}()':\n tokens.append((c, c, i))\n j += 1\n elif c.isalpha() or c == '_' or c == '$':\n start = j\n while j < len(line) and (line[j].isalnum() or line[j] in '_$'):\n j += 1\n val = line[start:j]\n tokens.append(('identifier', val, i))\n elif j + 1 < len(line) and line[j:j+2] == '=>':\n tokens.append(('=>', '=>', i))\n j += 2\n else:\n j += 1\n i += 1\n \n # 根据大括号配对解析并校验函数长度\n stack = []\n for idx, (t_type, t_val, t_line) in enumerate(tokens):\n if t_type == '{':\n is_func, start_line = is_function_brace(tokens, idx)\n stack.append((is_func, start_line))\n elif t_type == '}':\n if stack:\n is_func, start_line = stack.pop()\n if is_func:\n found_any_function = True\n length = t_line - start_line + 1\n if not (15 <= length <= 30):\n all_functions_valid = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if file.endswith('.html'):\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.IGNORECASE | re.DOTALL)\n for script in scripts:\n process_js(script)\n else:\n process_js(content)\n\n if not found_any_function:\n return False\n \n return all_functions_valid","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义常见前端代码文件的扩展名\n code_extensions = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css'}\n \n # 编译正则表达式，用于解析 字符串常量、HTML注释 和 JS/CSS的块/行注释。\n # 优先匹配字符串是为了防止正则引擎将包含在字符串中的 `//` 或 `/*` 错误识别为注释（例如URL：http://...）\n pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<html_comment><!--[\\s\\S]*?-->)'\n r'|(?P<js_comment>//.*|/\\*[\\s\\S]*?\\*/)'\n )\n \n # 如果代码项目路径不存在，无法判定，默认可以返回False或者视具体情况\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取文件内容，使用utf-8编码，并忽略可能的解码错误以防意外中断\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 遍历查找代码文件中的所有匹配项\n for match in pattern.finditer(content):\n # 提取匹配到的注释内容；如果是字符串常量，则该部分为空\n comment = match.group('html_comment') or match.group('js_comment')\n \n # 如果确实匹配到了注释内容\n if comment:\n # 遍历注释内容中的每一个字符\n for char in comment:\n # is_emoji 为已提供的专家函数，直接判断当前字符是否属于emoji\n if is_emoji(char):\n return False\n except Exception:\n # 若因其他异常导致部分文件无法读取，跳过继续处理剩余文件\n continue\n \n # 如果遍历所有代码文件均未在注释中发现 emoji，则要求满足\n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义前端网页开发中常见的代码文件后缀\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n lines = content.split('\\n')\n \n # 排除文件末尾单独的一个换行符导致的最后一个空字符串元素\n if len(lines) > 0 and lines[-1] == \"\":\n lines = lines[:-1]\n \n # 遍历判断每一行是否为空行\n for line in lines:\n if line.strip() == \"\":\n return False # 发现空行，未满足要求\n except Exception:\n # 对于非 UTF-8 编码读取失败的文件或特殊情况，予以跳过\n continue\n \n return True","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def check_ternary_operator_count(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_extensions = {'.js', '.html', '.css', '.htm'}\n code_files = []\n \n # 查找工作区下所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in code_extensions:\n code_files.append(os.path.join(root, file))\n \n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n code = f.read()\n except Exception:\n return False\n \n # 若为 HTML 文件，提取其中 <script> 标签内的内容避免受到普通文本内问号的干扰\n if file_path.lower().endswith('.html') or file_path.lower().endswith('.htm'):\n scripts = re.findall(r'<script[\\s\\S]*?>([\\s\\S]*?)</script>', code, re.IGNORECASE)\n code_to_check = '\\n'.join(scripts)\n else:\n code_to_check = code\n\n # 正则表达式模式：匹配字符串和注释\n pattern = r'''\n \"(?:\\\\.|[^\"\\\\])*\" | # 双引号字符串\n '(?:\\\\.|[^'\\\\])*' | # 单引号字符串\n `(?:\\\\.|[^`\\\\])*` | # 模板字符串\n //[^\\n]* | # 单行注释\n /\\*[\\s\\S]*?\\*/ | # 多行注释\n <!--[\\s\\S]*?--> # HTML 注释\n '''\n regex = re.compile(pattern, re.VERBOSE)\n # 移除代码中的字符串和注释内容\n cleaned_code = regex.sub('', code_to_check)\n \n # 移除可选链操作符(?:.) 和 空值合并运算符(?:??)\n cleaned_code = cleaned_code.replace('?.', '').replace('??', '')\n \n # 在去除了干扰项后，剩下的问号 '?' 即代表了三元运算符的数量\n if cleaned_code.count('?') != 2:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def check_file_line_count(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n line_count = 0\n for _ in f:\n line_count += 1\n if line_count > 200:\n return False\n except Exception:\n # 忽略由于编码问题（如二进制文件）或其他原因无法读取的文件\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def check_end_of_line_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常规的代码文件类型，排除类似.log等非代码文件\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs'}\n \n # 匹配在行尾出现的注释（允许注释后仅跟随空白字符）\n # 使用了 (?:(?!\\*/).)* 和 (?:(?!-->).)* 以保证定位非贪婪的正确闭合区域，避免吃掉同一行中可能存在的内部有效代码\n pattern = re.compile(r'(//.*|#.*|/\\*(?:(?!\\*/).)*\\*/|<!--(?:(?!-->).)*-->)\\s*$')\n has_code_file = False\n \n def remove_strings(code_line: str) -> str:\n res = []\n in_string = False\n string_char = ''\n escape = False\n for char in code_line:\n if not in_string:\n if char in (\"'\", '\"', '`'):\n in_string = True\n string_char = char\n res.append(char)\n else:\n res.append(char)\n else:\n if escape:\n res.append(' ')\n escape = False\n elif char == '\\\\':\n escape = True\n res.append(' ')\n elif char == string_char:\n in_string = False\n res.append(char)\n else:\n res.append(' ')\n return \"\".join(res)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except UnicodeDecodeError:\n continue\n \n for line in lines:\n if not line.strip():\n # 跳过空白行，空行是否允许由其他的系统提示限制判断，此项特指“代码行”\n continue\n \n # 避免字符串内容及常见未被带引号URL造成的 // 误判干扰\n processed_line = remove_strings(line)\n processed_line = processed_line.replace('http://', 'http: ').replace('https://', 'https: ')\n \n # 1. 检查当前行是否存在位于行尾的注释闭合\n if not pattern.search(processed_line):\n return False\n \n # 2. 检查是否为独立行注释（即去除一切符合规则的注释部分后，观察是否还有实质代码）\n temp = processed_line\n temp = re.sub(r'//.*', '', temp)\n temp = re.sub(r'#.*', '', temp)\n temp = re.sub(r'/\\*.*?\\*/', '', temp)\n temp = re.sub(r'<!--.*?-->', '', temp)\n \n if not temp.strip():\n # 去除所有注释后没有实质代码可见字符了，属于严格被禁的独立行注释\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"编写代码结束后，绝对禁止你做任何测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码文件中不得出现while循环","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def has_while_loop(code_content: str) -> bool:\n # 去除 HTML 注释\n code_content = re.sub(r'<!--[\\s\\S]*?-->', '', code_content)\n # 去除 多行注释 (/* ... */)\n code_content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code_content)\n # 去除 单行注释 (// ...)\n code_content = re.sub(r'//.*', '', code_content)\n \n # 查找 while 关键字后面可能伴随空白符并紧跟左括号的情况\n if re.search(r'\\bwhile\\s*\\(', code_content):\n return True\n return False\n\n code_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx'}\n\n # 1. 遍历并检查 workspace_path 下生成的实际代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if has_while_loop(content):\n return False\n except Exception:\n continue\n \n # 2. 检查 response 中的代码块 (包括 markdown 格式下的各类前端代码块)\n code_blocks = re.findall(r'```[a-zA-Z]*\\s*\\n([\\s\\S]*?)```', response)\n for block in code_blocks:\n if has_while_loop(block):\n return False\n\n return True","tags":["Style","Paradigm"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"对于你产出的代码，严禁使用ESLint进行校验","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的代码文件或日志文件后缀\n code_extensions = {'.html', '.js', '.css', '.log'}\n \n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_checked_files = False\n \n # 遍历项目工作区路径\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_checked_files = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码模式读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明该文件不是合法的 UTF-8 编码（例如使用了 GBK 并包含中文注释）\n return False\n except Exception:\n # 若遇到其他读取异常，同样视为验证失败\n return False\n \n # 只有当成功检查了至少一个代码文件且均无异常时，才返回 True\n return has_checked_files","tags":["Language","File Encoding"]}]},{"round_id":4,"instruction":"现在来完善玩家的生存和游戏失败机制。给玩家设定初始 3 点生命值，用几颗小红心或纯文字显示在画布右上角。如果有敌机直接撞击到玩家的飞机，或者有敌机成功逃脱飞出了屏幕下边界，玩家都要扣除 1 点生命值。玩家受伤扣血时，让整个画布背景短暂闪烁一下红色作为警告。当生命值降为 0 时，游戏暂停（停止所有的物理运动和生成），在画布正中央渲染大字号的『Game Over』，并在其下方绘制一个『重新开始』的按钮形状，点击该区域后可以完全重置游戏状态（清空实体、恢复生命、分数归零）。另外请你后续始终遵循以下调整和新增要求：原要求编写代码结束后绝对禁止你做任何测试调整为编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过；同时处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.isdir(workspace_path):\n return False\n\n # 可能的代码文件类型\n code_extensions = {'.html', '.js', '.css'}\n file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 忽略没有任何非空白字符的空行\n if not line.strip():\n continue\n \n # 提取前导空白字符\n leading_spaces = \"\"\n for char in line:\n if char in [' ', '\\t']:\n leading_spaces += char\n else:\n break\n \n # 规则1：不能出现制表符 (Tab) 缩进\n if '\\t' in leading_spaces:\n return False\n \n # 规则2：缩进量必须是 4 的倍数\n if len(leading_spaces) % 4 != 0:\n # 豁免多行注释的特殊情况（形如 \" * \" 开头的行通常会多出不满足倍数要求的单个空格）\n if line.lstrip().startswith('*'):\n continue\n \n return False\n\n return file_found","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def extract_js_from_html(html_content):\n # 提取 HTML 文件中所有的 <script> 标签内容\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', html_content, flags=re.DOTALL | re.IGNORECASE)\n return \"\\n\".join(scripts)\n \n def remove_strings_and_comments(code):\n # 将 JS 中的字符串内容和注释替换为空格，但保留换行符以维持原有行号准确无误\n new_code = []\n i = 0\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n while i < len(code):\n c = code[i]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n new_code.append(c)\n else:\n new_code.append(' ')\n i += 1\n elif in_block_comment:\n if c == '*' and i + 1 < len(code) and code[i+1] == '/':\n in_block_comment = False\n new_code.append(' ')\n new_code.append(' ')\n i += 2\n elif c == '\\n':\n new_code.append(c)\n i += 1\n else:\n new_code.append(' ')\n i += 1\n elif in_string:\n if c == '\\\\' and i + 1 < len(code):\n new_code.append(' ')\n new_code.append(' ')\n if code[i+1] == '\\n':\n new_code[-1] = '\\n'\n i += 2\n elif c == string_char:\n in_string = False\n new_code.append(' ')\n i += 1\n elif c == '\\n':\n new_code.append(c)\n i += 1\n else:\n new_code.append(' ')\n i += 1\n else:\n if c == '/' and i + 1 < len(code) and code[i+1] == '/':\n in_line_comment = True\n new_code.append(' ')\n new_code.append(' ')\n i += 2\n elif c == '/' and i + 1 < len(code) and code[i+1] == '*':\n in_block_comment = True\n new_code.append(' ')\n new_code.append(' ')\n i += 2\n elif c in (\"'\", '\"', '`'):\n in_string = True\n string_char = c\n new_code.append(' ')\n i += 1\n else:\n new_code.append(c)\n i += 1\n return \"\".join(new_code)\n\n code_files = []\n for root, dirs, files in os.walk(workspace_path):\n for f in files:\n if f.endswith('.js') or f.endswith('.html'):\n code_files.append(os.path.join(root, f))\n \n if not code_files:\n return True\n \n for f in code_files:\n try:\n with open(f, 'r', encoding='utf-8') as file:\n content = file.read()\n except Exception:\n continue\n \n if f.endswith('.html'):\n content = extract_js_from_html(content)\n \n if not content.strip():\n continue\n \n # 清理字符串和注释内容，以防止对 `{` 和 `}` 的误判\n clean_code = remove_strings_and_comments(content)\n \n i = 0\n stack = []\n \n while i < len(clean_code):\n # 检查是否有单行箭头函数 (例如 () => expression)\n if clean_code[i:i+2] == '=>':\n j = i + 2\n while j < len(clean_code) and clean_code[j] in ' \\t\\r\\n':\n j += 1\n if j < len(clean_code) and clean_code[j] != '{':\n # 如果不存在大括号且为单行箭头函数，其行数必然是1行，不满足 15 <= lines <= 30 的要求\n return False\n\n c = clean_code[i]\n if c == '{':\n # 寻找紧挨着 { 前的非空白字符\n j = i - 1\n while j >= 0 and clean_code[j] in ' \\t\\r\\n':\n j -= 1\n \n is_function = False\n start_line_index = clean_code.count('\\n', 0, i)\n \n if j >= 0:\n # 截取大括号前侧的文本作为判定依据\n start_look = max(0, j - 100)\n pre_text = clean_code[start_look:j+1]\n \n # 匹配普通函数 function ... ()\n if re.search(r'\\bfunction\\b\\s*(\\w+\\s*)?\\(', pre_text):\n is_function = True\n match_func = re.search(r'\\bfunction\\b', pre_text)\n if match_func:\n start_pos = start_look + match_func.start()\n start_line_index = clean_code.count('\\n', 0, start_pos)\n # 匹配多行箭头函数 => {\n elif re.search(r'=>$', pre_text):\n is_function = True\n match_arrow = re.search(r'=>', pre_text)\n if match_arrow:\n start_pos = start_look + match_arrow.start()\n start_line_index = clean_code.count('\\n', 0, start_pos)\n # 匹配对象的方法或是类方法 name() { \n elif re.search(r'\\w+\\s*\\([^)]*\\)$', pre_text):\n # 排除掉 if、for、while、switch、catch 等带括号的代码块\n if not re.search(r'\\b(if|for|while|switch|catch)\\b\\s*\\([^)]*\\)$', pre_text):\n is_function = True\n match_method = re.search(r'\\b\\w+\\s*\\(', pre_text)\n if match_method:\n start_pos = start_look + match_method.start()\n start_line_index = clean_code.count('\\n', 0, start_pos)\n \n # 入栈当前的起始行\n stack.append((start_line_index, is_function))\n \n elif c == '}':\n if stack:\n start_line_index, is_function = stack.pop()\n if is_function:\n end_line_index = clean_code.count('\\n', 0, i)\n # 计算函数的行数 (包含首尾两行)\n func_lines = end_line_index - start_line_index + 1\n if func_lines < 15 or func_lines > 30:\n return False\n i += 1\n \n # 若所有提取出的函数均在 [15, 30] 范围内，则满足要求\n return True","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 若工作区路径不存在，则无法判定存在违规代码，默认返回 True\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n code_extensions = ['.html', '.js', '.css']\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 仅处理根据用户指令推断出的网页类型代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 使用正则表达式去除字符串面值内容（支持去除单引号、双引号和反引号所包裹的字符串）\n # 这样可以避免字符串内部类似 http:// 的内容被错误判断为单行注释\n try:\n content_no_str = re.sub(r'([\"\\'`])(?:\\\\.|(?!\\1).)*?\\1', '', content)\n except Exception:\n # 如果正则处理出现异常，则降级为使用原始内容\n content_no_str = content\n \n # 提取各种格式的注释文本\n comments = []\n comments.extend(re.findall(r'//(.*)', content_no_str))\n comments.extend(re.findall(r'/\\*([\\s\\S]*?)\\*/', content_no_str))\n comments.extend(re.findall(r'<!--([\\s\\S]*?)-->', content_no_str))\n \n # 校验提取出的注释文本中是否含有 emoji 符号\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀，涵盖HTML/JS项目以及其他常用语言\n code_extensions = {\n '.js', '.html', '.css', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.php'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断文件是否为代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 判定空行：去除头尾空白字符后为空字符串\n if line.strip() == \"\":\n return False\n except Exception:\n # 若文件由于编码等问题无法读取，忽略该文件跳过\n continue\n\n return True","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的代码文件扩展名，排除.log等非代码文件\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n # 构造综合正则表达式，用于匹配并抹除：\n # 多行注释、HTML注释、单行注释、双引号字符串、单引号字符串、反引号模板字符串\n # 使用三引号的原生字符串以避免Python中的转义和单双引号解析混乱\n pattern = r\"\"\"/\\*[\\s\\S]*?\\*/|<!--[\\s\\S]*?-->|//.*|\"[^\"\\\\]*(?:\\\\[\\s\\S][^\"\\\\]*)*\"|'[^'\\\\]*(?:\\\\[\\s\\S][^'\\\\]*)*'|`[^`\\\\]*(?:\\\\[\\s\\S][^`\\\\]*)*`\"\"\"\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件存在格式异常无法用utf-8读取，跳过即可，这有可能导致漏判或假阴性\n continue\n \n # 剔除所有的代码注释及字符串字面量片段，避免它们内部的'?'符号产生干扰\n text = re.sub(pattern, '', content)\n \n # 统计剩余纯代码文本中独立的 '?'\n # (?<!\\?) 排除前面的?，(?!\\.|\\?) 排除后面的.或?，从而有效避开 '?.' 和 '??'\n matches = re.findall(r'(?<!\\?)\\?(?!\\.|\\?)', text)\n \n # 判断当前代码文件中的三元运算符数量是否精确为2\n if len(matches) != 2:\n return False\n \n # 所有代码文件都精确满足时返回True，前提是必须存在至少一个代码文件\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return True\n \n # 遍历项目地址下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n # 尝试以UTF-8格式读取文件内容\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n line_count = 0\n for line in f:\n line_count += 1\n # 提早退出，一旦行数超过200行则直接返回False\n if line_count > 200:\n return False\n except Exception:\n # 忽略文件读取过程中可能出现的异常（如无读取权限或特殊文件）\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义我们要检查的代码文件扩展名\n valid_extensions = {'.js', '.html', '.css'}\n # 定义常见的注释标识符\n comment_markers = ['//', '/*', '<!--']\n \n has_code_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于某种原因无法读取文件，可以跳过该文件处理\n continue\n \n for line in lines:\n stripped = line.strip()\n \n # 忽略空行\n if not stripped:\n continue\n \n # 1. 检查是否存在独立行注释 (即仅有注释没有代码)\n # 如果字符串以任意注释符开头，说明是独立注释，不符合要求\n is_independent = False\n for marker in comment_markers:\n if stripped.startswith(marker):\n is_independent = True\n break\n \n if is_independent:\n return False\n \n # 2. 检查该行代码是否配有行尾注释\n # 即该行不仅要有代码，还要在结尾或中间包含注释符\n has_inline = False\n for marker in comment_markers:\n if marker in stripped:\n has_inline = True\n break\n \n if not has_inline:\n return False\n\n # 必须有代码文件被解析到且全部满足条件，才返回True\n return has_code_files","tags":["Style","Layout"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码文件中不得出现while循环","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n # 如果项目目录不存在，默认未违反要求\n return True\n \n # 定义需要检查的代码文件扩展名\n valid_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 移除 HTML 注释 <!-- ... -->\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 移除 JS/TS 多行注释 /* ... */\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 移除 JS/TS 单行注释 // ...\n content = re.sub(r'//.*', '', content)\n \n # 查找是否存在 while 关键字后紧跟可能的空格与左括号的结构\n # 这可以准确识别 while (...) 和 do ... while (...)\n if re.search(r'\\bwhile\\s*\\(', content):\n return False\n except Exception:\n # 忽略读取异常跳过文件\n continue\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"对于你产出的代码，严禁使用ESLint进行校验","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推断出的潜在代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.ts'}\n found_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行编码检测\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n \n # 尝试严格使用 UTF-8 格式解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 只要有一个代码文件无法用 UTF-8 解码，说明不满足要求\n return False\n except Exception:\n # 其他读取异常忽略，继续检查其他文件\n pass\n\n # 如果存在代码文件且全部通过了 UTF-8 解码，则满足要求\n return found_code_file","tags":["Language","File Encoding"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":5,"instruction":"游戏目前比较单调，我们加入难度递增机制和特殊道具。第一，引入等级机制：分数每达到 100 分的整数倍时，游戏提升一级，并在屏幕中央闪烁提示『Level Up!』。等级越高，敌机的生成频率越快，整体下落速度也越快。第二，加入补给包：每隔约 15 秒，从顶部掉落一个蓝色的圆形『火力补给包』。如果玩家飞机接到了它，在接下来的 5 秒内，玩家飞机的射击模式临时变为『双发模式』（即每次按空格，会在机翼两侧平行发射两发子弹），5秒后恢复单发。另外请你后续始终遵循以下调整和新增要求：原要求严禁使用ESLint校验你产出的代码调整为你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过；同时，执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def is_valid_indent(line: str) -> bool:\n # 匹配前导空白字符\n match = re.match(r'^([ \\t]*)', line)\n if not match:\n return True\n leading_ws = match.group(1)\n\n # 剥离换行符后的整行内容\n content = line.rstrip('\\r\\n')\n \n # 如果整行都是空白字符（空行），只要没有制表符即认为合规\n if len(leading_ws) == len(content):\n return '\\t' not in leading_ws\n\n # 如果前导空白中含有制表符，直接判定为不合规\n if '\\t' in leading_ws:\n return False\n\n # 去除前导空白后剩下的可见内容\n stripped = content.lstrip(' \\t')\n\n # 容错：多行注释中的中间行（通常以 '*' 开头对齐）可能会有一或额外几个空格\n if stripped.startswith('*') or stripped.startswith('<!--') or stripped.startswith('-->'):\n return True\n\n # 核心判断：前导空格数量必须是 4 的整数倍\n if len(leading_ws) % 4 != 0:\n return False\n\n return True\n\n valid_extensions = ['.html', '.css', '.js', '.jsx', '.ts', '.tsx']\n has_checked_code = False\n\n # 1. 优先检查 workspace_path 中的文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file_name in files:\n if any(file_name.endswith(ext) for ext in valid_extensions):\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n if not is_valid_indent(line):\n return False\n has_checked_code = True\n except Exception:\n continue\n\n # 2. 如果工作区没有生成有效代码文件，回退到检查 response 中的代码块\n if not has_checked_code:\n # 匹配 Markdown 中的代码块\n code_blocks = re.findall(r'```(?:html|javascript|js|css|typescript|ts)?\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n if not code_blocks:\n # 既没有物理文件也没有代码块，无法确认满足条件\n return False\n \n for block in code_blocks:\n for line in block.split('\\n'):\n if not is_valid_indent(line):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n def mask_js(code: str) -> str:\n \"\"\"\n 屏蔽代码中的字符串和注释，将其替换为空格（保留换行符保证行号不变）\n \"\"\"\n masked = []\n i = 0\n n = len(code)\n \n in_single = False\n in_double = False\n in_backtick = False\n in_line_comment = False\n in_block_comment = False\n \n while i < n:\n c = code[i]\n if in_single:\n if c == '\\\\':\n masked.append(' ')\n i += 1\n if i < n:\n masked.append('\\n' if code[i] == '\\n' else ' ')\n i += 1\n elif c == \"'\":\n in_single = False\n masked.append(' ')\n i += 1\n else:\n masked.append('\\n' if c == '\\n' else ' ')\n i += 1\n elif in_double:\n if c == '\\\\':\n masked.append(' ')\n i += 1\n if i < n:\n masked.append('\\n' if code[i] == '\\n' else ' ')\n i += 1\n elif c == '\"':\n in_double = False\n masked.append(' ')\n i += 1\n else:\n masked.append('\\n' if c == '\\n' else ' ')\n i += 1\n elif in_backtick:\n if c == '\\\\':\n masked.append(' ')\n i += 1\n if i < n:\n masked.append('\\n' if code[i] == '\\n' else ' ')\n i += 1\n elif c == \"`\":\n in_backtick = False\n masked.append(' ')\n i += 1\n else:\n masked.append('\\n' if c == '\\n' else ' ')\n i += 1\n elif in_line_comment:\n if c == '\\n':\n in_line_comment = False\n masked.append('\\n')\n i += 1\n else:\n masked.append(' ')\n i += 1\n elif in_block_comment:\n if c == '*' and i + 1 < n and code[i+1] == '/':\n in_block_comment = False\n masked.append(' ')\n masked.append(' ')\n i += 2\n else:\n masked.append('\\n' if c == '\\n' else ' ')\n i += 1\n else:\n if c == \"'\" :\n in_single = True\n masked.append(' ')\n i += 1\n elif c == '\"':\n in_double = True\n masked.append(' ')\n i += 1\n elif c == '`':\n in_backtick = True\n masked.append(' ')\n i += 1\n elif c == '/' and i + 1 < n and code[i+1] == '/':\n in_line_comment = True\n masked.append(' ')\n masked.append(' ')\n i += 2\n elif c == '/' and i + 1 < n and code[i+1] == '*':\n in_block_comment = True\n masked.append(' ')\n masked.append(' ')\n i += 2\n else:\n masked.append(c)\n i += 1\n return \"\".join(masked)\n\n def get_function_line_counts(code: str) -> list:\n masked_code = mask_js(code)\n \n # 预计算行起始索引以加快行号查找\n line_starts = [0]\n for i, c in enumerate(code):\n if c == '\\n':\n line_starts.append(i + 1)\n \n def get_line_num(idx: int) -> int:\n return bisect.bisect_right(line_starts, idx)\n \n def get_prev_non_ws(start_idx: int) -> int:\n curr = start_idx\n while curr >= 0 and masked_code[curr].isspace():\n curr -= 1\n return curr\n\n def is_function_start(idx: int) -> bool:\n p1 = get_prev_non_ws(idx - 1)\n if p1 < 0:\n return False\n \n # 1. 遇到前一个有效字符是箭头函数的 '=>'\n if masked_code[p1] == '>' and p1 > 0 and masked_code[p1-1] == '=':\n return True\n \n # 2. 遇到前一个是 ')' 对应常规函数/方法结构\n if masked_code[p1] == ')':\n depth = 0\n p2 = p1\n while p2 >= 0:\n if masked_code[p2] == ')':\n depth += 1\n elif masked_code[p2] == '(':\n depth -= 1\n if depth == 0:\n break\n p2 -= 1\n \n if p2 < 0:\n return False\n \n p3 = get_prev_non_ws(p2 - 1)\n if p3 < 0:\n return True\n \n end_ident = p3\n while p3 >= 0 and (masked_code[p3].isalnum() or masked_code[p3] in ('_', '$')):\n p3 -= 1\n \n if p3 == end_ident:\n # 匿名函数没有名称，如 `function() {` 或者 `(function() {`\n return True\n \n ident = masked_code[p3+1 : end_ident+1]\n # 排除由于控制流等引起的 `) {` 结构\n reserved = {'if', 'for', 'while', 'catch', 'switch', 'with', 'await'}\n if ident in reserved:\n return False\n return True\n \n return False\n\n function_lengths = []\n stack = []\n \n i = 0\n n = len(masked_code)\n while i < n:\n c = masked_code[i]\n \n # 检测并处理无花括号的单行箭头函数 (例如 `const f = a => a + 1;`)\n if c == '=' and i + 1 < n and masked_code[i+1] == '>':\n nxt = i + 2\n while nxt < n and masked_code[nxt].isspace():\n nxt += 1\n if nxt < n and masked_code[nxt] != '{':\n # 无花括号的箭头函数通常只有1行或极短，默认长度记为1（它必将违背 15-30 的要求，成功阻截）\n function_lengths.append(1)\n i += 2\n continue\n \n if c == '{':\n is_func = is_function_start(i)\n start_line = get_line_num(i)\n stack.append((is_func, start_line))\n i += 1\n continue\n \n if c == '}':\n if stack:\n is_func, start_line = stack.pop()\n if is_func:\n end_line = get_line_num(i)\n function_lengths.append(end_line - start_line + 1)\n i += 1\n continue\n \n i += 1\n \n return function_lengths\n\n js_codes = []\n \n # 获取所有的 JS 和 HTML 文件内容\n for root, dirs, files in os.walk(workspace_path):\n # 避开不需要检测的三方库及构建缓存目录\n for exclude_dir in ['node_modules', '.git', 'dist', 'build', 'coverage', '__pycache__']:\n if exclude_dir in dirs:\n dirs.remove(exclude_dir)\n \n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n js_codes.append(f.read())\n except Exception:\n pass\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 使用正则提取出所有script块的内容\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n js_codes.extend(scripts)\n except Exception:\n pass\n \n all_lengths = []\n for code in js_codes:\n all_lengths.extend(get_function_line_counts(code))\n \n if not all_lengths:\n # 如果项目中没有任何函数代码被解析出来，视为未达标准（通常一个合格的Canvas骨架肯定包含函数）\n return False\n \n for length in all_lengths:\n if not (15 <= length <= 30):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 构造正则表达式，用于安全、准确地提取注释片段。\n # 核心原理：先匹配任何形式的字符串常量并消耗掉，避免字符串内部的 //, /* 等内容被误认为注释\n pattern = re.compile(\n r'(?P<string>\"([^\"\\\\]|\\\\.)*\"|\\'([^\\'\\\\]|\\\\.)*\\'|`([^`\\\\]|\\\\.)*`)' # 匹配双引号、单引号和反引号内的字符串\n r'|(?P<html_comment><!--[\\s\\S]*?-->)' # 匹配 HTML 注释\n r'|(?P<multi_comment>/\\*[\\s\\S]*?\\*/)' # 匹配 JS/CSS 多行注释\n r'|(?P<single_comment>//[^\\n]*)' # 匹配 JS/TS 单行注释\n )\n \n # 根据用户指令中对HTML5 Canvas、JS以及ESLint等要素的描述，推断可能存在的代码文件类型\n valid_extensions = ('.html', '.js', '.css', '.ts', '.jsx', '.tsx')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 在当前文件内容中查找所有模式\n for match in pattern.finditer(content):\n # 如果当前成功匹配的是某一种注释分组，则检查该组中的内容\n if match.lastgroup in ('html_comment', 'multi_comment', 'single_comment'):\n comment_text = match.group()\n \n # 遍历该注释内容中的每一个字符\n for char in comment_text:\n # 题目约定了可以直接使用提供的 is_emoji 函数\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据指令推测可能的代码文件后缀\n code_extensions = {'.html', '.htm', '.js', '.css'}\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 判定是否为空行（仅包含空白字符或换行符的行 strip() 后都会变为空字符串）\n if line.strip() == \"\":\n return False\n except Exception:\n # 如果遇到文件编码或读取异常，说明存在生成问题或不符合UTF-8的要求，返回False\n return False\n \n return has_code_file","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码文件后缀类型\n code_extensions = ['.js', '.html', '.css', '.ts', '.jsx', '.tsx']\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n # 1. 移除字符串字面量（避免字符串内部的 `?` 和 `:` 造成的统计干扰）\n # 处理双引号、单引号和反引号，并在正则中匹配可能的转义字符\n code = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '', code)\n code = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", '', code)\n code = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '', code, flags=re.DOTALL)\n \n # 2. 移除代码中的注释\n # 移除 /* ... */ 块注释\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除 // ... 行注释\n code = re.sub(r'//.*', '', code)\n # 移除 HTML <!-- ... --> 注释\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n \n # 3. 排除正则表达式中非捕获组的 (?:\n code = code.replace('(?:', '')\n \n # 4. 统计三元运算符数量\n # 查找 `?`，并利用负向先行断言排除后面紧跟 `.` (可选链) 或 `?` (空值合并) 的情况\n matches = re.findall(r'\\?(?!\\.|[?])', code)\n \n # 每个代码文件必须精确包含 2 个三元运算符\n if len(matches) != 2:\n return False\n except Exception:\n # 如果文件读取或解析抛出异常（如编码错误），视为未满足要求\n return False\n \n # 如果没有找到任何代码文件，也是未满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def verify_file_line_count(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 根据多轮指令推断相关可能生成的代码和文件扩展名\n allowed_extensions = {'.html', '.js', '.css', '.log', '.json'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的非AI直接生成的依赖库或代码控制相关目录\n if 'node_modules' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 匹配推测的扩展名，或者特殊配置文件（如 .eslintrc 及其变体）\n if ext in allowed_extensions or file.startswith('.eslintrc'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 迭代计算该文件内的行数\n line_count = sum(1 for _ in f)\n \n # 检查行数是否大于200，超过则违反规定\n if line_count > 200:\n return False\n except Exception:\n # 如果由于读取非纯文本文件或编码问题导致读取异常，则跳过继续检查下一个文件\n continue\n \n # 如果所有被检查的文件行数均 <= 200，则满足要求\n return True","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 推断前端项目可能的代码文件类型\n code_extensions = {'.html', '.htm', '.js', '.ts', '.css', '.jsx', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n line = line.strip()\n if not line:\n continue # 忽略完全空白的行\n \n # 剥离字符串内容，以避免将字符串内的 URL (`http://`) 或 注释符内容误判为真实注释\n s_line = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', line)\n s_line = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", s_line)\n s_line = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', s_line)\n \n s_line_stripped = s_line.strip()\n if not s_line_stripped:\n continue\n \n # 确定对应文件支持的注释符号\n syms = []\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\n syms = ['//', '/*']\n elif ext in ['.html', '.htm']:\n # HTML 文件既能包含 HTML 注释，也可能包含内置脚本片段里的注释\n syms = ['<!--', '//', '/*']\n elif ext in ['.css']:\n syms = ['/*']\n \n comment_idx = -1\n first_sym = None\n \n # 寻找本行中最先出现的有效注释符号\n for sym in syms:\n idx = s_line.find(sym)\n if idx != -1:\n if comment_idx == -1 or idx < comment_idx:\n comment_idx = idx\n first_sym = sym\n \n if comment_idx == -1:\n # 没有任何注释符，且此行非空，说明这是只有代码没有行尾注释的行\n return False\n \n left_part = s_line[:comment_idx].strip()\n if not left_part:\n # 找到了注释符，但其左侧没有任何代码，属于独立行注释\n return False\n \n # 确保多行/块级注释不会跨行（跨行的块级注释必然会在其他行构成独立注释行）\n if first_sym == '/*':\n if '*/' not in s_line[comment_idx + 2:]:\n return False\n elif first_sym == '<!--':\n if '-->' not in s_line[comment_idx + 4:]:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码文件中不得出现while循环","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def remove_comments_and_strings(code: str) -> str:\n # 正则表达式：匹配双引号、单引号、反引号字符串，以及块注释和行注释\n pattern = r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`|/\\*.*?\\*/|//[^\\n]*'\n # 替换匹配到的内容为空，实现过滤\n return re.sub(pattern, '', code, flags=re.DOTALL)\n\n def extract_script_content(html: str) -> str:\n # 正则表达式：提取 <script> 标签内的内容\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', html, flags=re.IGNORECASE | re.DOTALL)\n return \"\\n\".join(scripts)\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # 处理 JS / TS 类文件\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n cleaned_code = remove_comments_and_strings(content)\n # 匹配 while 循环结构，即单词 while 后面跟着零或多个空格和左括号 (\n if re.search(r'\\bwhile\\s*\\(', cleaned_code):\n return False\n except Exception:\n # 忽略无法读取或编码有问题的文件\n pass\n \n # 处理 HTML 类文件\n elif file.endswith(('.html', '.htm')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n scripts = extract_script_content(content)\n cleaned_code = remove_comments_and_strings(scripts)\n if re.search(r'\\bwhile\\s*\\(', cleaned_code):\n return False\n except Exception:\n pass\n\n # 若所有相关代码文件中均未发现 while 循环，则满足要求\n return True","tags":["Style","Paradigm"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推断可能存在的代码文件类型\n code_extensions = {'.html', '.js', '.css'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前工作空间的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 utf-8 对文件内容进行解码\n # 因为包含中文注释，非 utf-8 编码必定会在此处抛出 UnicodeDecodeError\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 解码失败，说明存在非 UTF-8 编码的代码文件\n return False\n except Exception:\n # 应对其他异常（如读取错误等）\n return False\n \n # 如果找到了代码文件，并且全都没有抛出异常，则满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":6,"instruction":"最后，为了让视觉效果更现代、更炫酷，我们需要增加粒子爆炸特效。请编写一个粒子系统：当敌机被子弹击毁时，不要让它直接生硬地消失，而是要在敌机的中心坐标生成 10 到 15 个颜色在橙色和红色之间随机的小圆点粒子。这些粒子生成后，需要以随机的角度和速度向四面八方炸开，并且在飞散的过程中受到一点点『阻力』减速，同时逐渐缩小并变透明，直到生命周期结束完全消失。另外请你后续始终遵循以下调整和新增要求：原要求你提供的代码文件中不得出现while循环调整为在编写生成的代码文件时，请避免使用for循环；同时，请确保生成的代码文件中不包含多层if。","instruction_following_checklist":[{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"生成代码文件的缩进量必须固定为4个空格","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return False\n \n code_extensions = {'.html', '.htm', '.js', '.css'}\n has_code_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 去除行尾的换行符和空白字符，如果该行全空，则跳过\n if not line.strip():\n continue\n \n # 匹配行首的空白字符（仅匹配空格和制表符）\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_whitespace = match.group(1)\n \n # 检查1：不能包含制表符（Tab）\n if '\\t' in leading_whitespace:\n return False\n \n # 检查2：空格的数量必须是4的整数倍\n if len(leading_whitespace) % 4 != 0:\n return False\n \n return has_code_files","tags":["Style","Layout"]},{"constraint":"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def mask_code(text: str) -> str:\n \"\"\"\n 将代码中的注释和字符串内容替换为空格，但保留换行符，\n 从而在不改变任何代码行号的前提下，避免字符串和注释中的 '{' 和 '}' 干扰代码块解析。\n \"\"\"\n result = list(text)\n i = 0\n n = len(text)\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n while i < n:\n if in_line_comment:\n if text[i] == '\\n':\n in_line_comment = False\n else:\n result[i] = ' '\n i += 1\n elif in_block_comment:\n if text[i:i+2] == '*/':\n result[i] = ' '\n if i + 1 < n:\n result[i+1] = ' '\n in_block_comment = False\n i += 2\n else:\n if text[i] != '\\n':\n result[i] = ' '\n i += 1\n elif in_string:\n if text[i] == '\\\\':\n result[i] = ' '\n if i + 1 < n:\n result[i+1] = ' '\n i += 2\n elif text[i] == string_char:\n result[i] = ' '\n in_string = False\n i += 1\n else:\n if text[i] != '\\n':\n result[i] = ' '\n i += 1\n else:\n if text[i:i+2] == '//':\n in_line_comment = True\n result[i] = ' '\n if i + 1 < n:\n result[i+1] = ' '\n i += 2\n elif text[i:i+2] == '/*':\n in_block_comment = True\n result[i] = ' '\n if i + 1 < n:\n result[i+1] = ' '\n i += 2\n elif text[i] in (\"'\", '\"', \"`\"):\n in_string = True\n string_char = text[i]\n result[i] = ' '\n i += 1\n else:\n i += 1\n return \"\".join(result)\n\n def get_functions_line_counts(code: str, is_html: bool) -> list:\n if is_html:\n scripts = []\n # 提取所有 <script> 标签中的内容\n for match in re.finditer(r'<script\\b[^>]*>(.*?)</script>', code, re.IGNORECASE | re.DOTALL):\n scripts.append(match.group(1))\n codes_to_check = scripts\n else:\n codes_to_check = [code]\n \n counts = []\n for c in codes_to_check:\n masked = mask_code(c)\n \n # 定义三种常见的JS函数入口特征正则表达式\n # p1: function keyword (e.g. function name() { or function() { )\n p1 = r'\\bfunction\\b[^{]*\\{'\n # p2: arrow functions with braces (e.g. () => { or args => { )\n p2 = r'(?:[a-zA-Z0-9_$]+|\\([^)]*\\))\\s*=>\\s*\\{'\n # p3: class methods or object literal methods (e.g. methodName() { )\n p3 = r'\\b(?!(?:if|while|for|switch|catch|return|with|function|new)\\b)[a-zA-Z0-9_$]+\\s*\\([^)]*\\)\\s*\\{'\n \n brace_indices = [i for i, char in enumerate(masked) if char == '{']\n \n for brace_idx in brace_indices:\n # 截取大括号前的一段文本，判断是否以函数的特征结尾\n prefix = masked[max(0, brace_idx - 300):brace_idx + 1]\n \n is_func = False\n match_start_idx = -1\n \n m1 = re.search(p1 + r'$', prefix)\n if m1:\n is_func = True\n match_start_idx = brace_idx - len(prefix) + m1.start()\n else:\n m2 = re.search(p2 + r'$', prefix)\n if m2:\n is_func = True\n match_start_idx = brace_idx - len(prefix) + m2.start()\n else:\n m3 = re.search(p3 + r'$', prefix)\n if m3:\n is_func = True\n match_start_idx = brace_idx - len(prefix) + m3.start()\n \n if is_func:\n brace_count = 1\n end_idx = brace_idx + 1\n while end_idx < len(masked) and brace_count > 0:\n if masked[end_idx] == '{':\n brace_count += 1\n elif masked[end_idx] == '}':\n brace_count -= 1\n end_idx += 1\n \n if brace_count == 0:\n # 计算行数差值\n start_line = masked[:match_start_idx].count('\\n')\n end_line = masked[:end_idx].count('\\n')\n counts.append(end_line - start_line + 1)\n \n # 单独匹配没有任何大括号 `{` 的单行箭头函数：例如 x => x * 2;\n # 这种函数即使由于换行被拆成了两三行，也极难达到 15 行的下限，因此计入后将直接使验证失败\n for m in re.finditer(r'(?<!=)=>', masked):\n idx = m.end()\n while idx < len(masked) and masked[idx].isspace():\n idx += 1\n if idx < len(masked) and masked[idx] != '{':\n counts.append(1) # 简化处理，统一记录为1行，后续均会不达标\n \n return counts\n\n all_counts = []\n files_checked = 0\n \n # 优先解析项目文件夹下生成的代码文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for f in files:\n if f.endswith('.js'):\n with open(os.path.join(root, f), 'r', encoding='utf-8') as file:\n code = file.read()\n all_counts.extend(get_functions_line_counts(code, is_html=False))\n files_checked += 1\n elif f.endswith('.html'):\n with open(os.path.join(root, f), 'r', encoding='utf-8') as file:\n code = file.read()\n all_counts.extend(get_functions_line_counts(code, is_html=True))\n files_checked += 1\n \n # 如果项目文件夹内未找到，则尝试回退到解析 response 里提供的 Markdown 代码块\n if files_checked == 0:\n for match in re.finditer(r'```(?:javascript|js)\\s*(.*?)```', response, re.DOTALL | re.IGNORECASE):\n all_counts.extend(get_functions_line_counts(match.group(1), is_html=False))\n for match in re.finditer(r'```html\\s*(.*?)```', response, re.DOTALL | re.IGNORECASE):\n all_counts.extend(get_functions_line_counts(match.group(1), is_html=True))\n \n # 如果没有找到任何函数，那么不能认为满足了要求（对于游戏项目来说这是不符合逻辑的）\n if not all_counts:\n return False\n \n # 检查所有的函数行数是否都符合 [15, 30] 闭区间的要求\n for c in all_counts:\n if c < 15 or c > 30:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"emoji表情符号 绝对不能出现在任何代码文件的注释内容中","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 构造正则表达式以匹配 字符串字面量 和 各种注释风格\n # Group 1: 各种引号包裹的字符串（过滤掉字符串内的 // 等字符，避免误判）\n # Group 2: 单行注释 (JS, TS, CSS 等)\n # Group 3: 多行注释 (JS, TS, CSS 等)\n # Group 4: HTML 注释\n # Group 5: 单行注释 (Python, Shell 等)\n pattern = re.compile(\n r'(\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|'\n r'(//.*?$)|'\n r'(/\\*[\\s\\S]*?\\*/)|'\n r'(<!--[\\s\\S]*?-->)|'\n r'(#.*?$)',\n re.MULTILINE\n )\n\n def has_emoji_in_comments(content: str) -> bool:\n \"\"\"检查给定内容中的注释是否含有 emoji\"\"\"\n for match in pattern.finditer(content):\n # 获取注释的内容（如果在字符串内，则 group 2, 3, 4, 5 均为 None）\n comment = match.group(2) or match.group(3) or match.group(4) or match.group(5)\n if comment:\n # 遍历注释中的每一个字符\n for char in comment:\n if is_emoji(char):\n return True\n return False\n\n valid_extensions = ('.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.sh', '.json')\n\n # 1. 检查 workspace_path 下生成的代码文件中的注释\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若文件中注释含有 emoji，要求不被满足\n if has_emoji_in_comments(content):\n return False\n except Exception:\n continue\n\n # 2. 检查 response 文本中自带的代码块内的注释\n code_blocks = re.findall(r'```(?:\\w*)\\s*\\n([\\s\\S]*?)```', response)\n for block in code_blocks:\n if has_emoji_in_comments(block):\n return False\n\n # 没有发现任何违规情况，要求被满足\n return True","tags":["Content","Keyword"]},{"constraint":"不要让你生成的任何代码文件里存在空行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据上下文推断可能生成的代码文件扩展名\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx'}\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n def check_no_empty_lines_in_file(file_path: str) -> bool:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去除行首尾的空白字符（包含空格、制表符、回车换行等）\n # 如果结果是空字符串，意味着当前行是空行\n if line.strip() == '':\n return False\n except Exception:\n # 若文件读取失败（例如编码不支持），则跳过\n pass\n return True\n\n # 如果传入的路径本身是文件\n if os.path.isfile(workspace_path):\n ext = os.path.splitext(workspace_path)[1].lower()\n if ext in code_extensions:\n return check_no_empty_lines_in_file(workspace_path)\n return True\n\n # 如果传入的是目录，则递归遍历其下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n if not check_no_empty_lines_in_file(file_path):\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)","validation_code":"def check_ternary_operator_in_files(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的代码文件类型\n valid_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\n files_checked = 0\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 1. 剔除字符串内容 (双引号、单引号、反引号)，避免匹配到字符串里的符号\n # 处理转义字符，避免匹配提前结束\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', content)\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', content)\n content = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '', content)\n \n # 2. 剔除注释内容\n # 多行注释\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n # HTML 注释\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n # 单行注释 (在去除了字符串之后进行，避免误删 URL 中的 //)\n content = re.sub(r'//.*', '', content)\n \n # 3. 统计三元运算符中的问号数量\n # 排除可选链 `?.` 和空值合并 `??`\n ternary_count = len(re.findall(r'\\?(?![?.])', content))\n \n # 如果任意一个代码文件不满足精确包含2个三元运算符的要求，直接返回 False\n if ternary_count != 2:\n return False\n \n files_checked += 1\n \n # 如果没有扫描到任何有效的代码文件，返回 False\n if files_checked == 0:\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"你创建的每一个文件，其行数都应保持在200行或以下","validation_code":"def check_file_length_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区不存在，无法判断是否有超行数的文件，直接返回 False\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令推断的可能代码及日志文件类型\n valid_extensions = {'.html', '.js', '.css', '.log'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉非生成环境所产生的目录，防止干扰\n for exclude_dir in ['node_modules', '.git']:\n if exclude_dir in dirs:\n dirs.remove(exclude_dir)\n\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n # 若文件的行数大于200，则不满足要求\n if len(lines) > 200:\n return False\n except Exception:\n # 对于不可读或编码不符的文件跳过检查\n pass\n\n # 遍历结束无超过200行的文件则满足要求\n return True","tags":["Quantity","Range"]},{"constraint":"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n valid_exts = ['.js', '.html', '.css', '.ts']\n code_file_count = 0\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n\n filepath = os.path.join(root, file)\n code_file_count += 1\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n\n for line in lines:\n s_line = line.strip()\n if not s_line:\n continue\n\n # 检查是否为独立行注释（不允许存在位于代码上方或下方的独立行注释）\n if s_line.startswith('//') or s_line.startswith('/*') or s_line.startswith('<!--'):\n return False\n\n # 检查该行代码是否包含行尾注释\n has_comment = False\n if ext in ['.js', '.ts', '.html']:\n if '/*' in s_line:\n has_comment = True\n elif '<!--' in s_line and ext == '.html':\n has_comment = True\n else:\n # 查找 // 注释，避开类似 http:// 的非注释情况\n idx = s_line.find('//')\n while idx != -1:\n if idx > 0 and s_line[idx-1] == ':':\n idx = s_line.find('//', idx + 2)\n else:\n has_comment = True\n break\n elif ext == '.css':\n if '/*' in s_line:\n has_comment = True\n\n # 如果这行代码没有找到任何注释标识符，代表缺失了行尾注释\n if not has_comment:\n return False\n\n # 至少需要检查到一个代码文件且均满足要求时才认为任务成功\n return code_file_count > 0","tags":["Style","Layout"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在编写生成的代码文件时，请避免使用for循环","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def remove_comments_and_strings(code: str) -> str:\n # 正则表达式用于匹配并提取可能的干扰项：\n # 1. /* ... */ 多行注释\n # 2. // ... 单行注释\n # 3. \"...\" 双引号字符串\n # 4. '...' 单引号字符串\n # 5. `...` 反引号字符串\n pattern = r'(/\\*.*?\\*/)|(//[^\\n]*)|(\"(?:\\\\.|[^\\\\\"])*\")|(\\'(?:\\\\.|[^\\\\\\'])*\\')|(`(?:\\\\.|[^\\\\`])*`)'\n \n # 将上述匹配到的注释或字符串内容替换为空格，以防止其中的内容被错误地识别为代码关键字\n return re.sub(pattern, ' ', code, flags=re.DOTALL)\n\n # 如果工作区路径不存在，直接返回 True（可视为未违反要求）\n if not os.path.isdir(workspace_path):\n return True\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 针对飞机大战游戏，主要检查 html 与 js 后缀的源码文件\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n js_content = \"\"\n if file.endswith('.html'):\n # 仅提取 HTML 文件中 <script> 标签内的 JS 代码\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\n js_content = '\\n'.join(scripts)\n else:\n js_content = content\n \n # 去除注释和字符串后的代码片段\n clean_js = remove_comments_and_strings(js_content)\n \n # 检查清理后的代码中是否出现 for 循环，包括 for(...) 和 for await(...)\n # \\bfor\\b : 确保匹配的是完整关键字 for\n # \\s* : 匹配可能存在的空格\n # (?:await\\s*)? : 兼容 JavaScript 中的 for await 语法\n # \\( : 匹配语法左括号\n if re.search(r'\\bfor\\s*(?:await\\s*)?\\(', clean_js):\n return False\n \n return True","tags":["Style","Paradigm"]},{"constraint":"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 针对前端和常见JS项目，定义相关的代码文件扩展名\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts', '.json'}\n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n \n try:\n # 以二进制模式读取整个文件\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试将其按照 utf-8 进行解码\n # 如果不是 UTF-8 编码或 ASCII 编码（比如 GBK 等含中文的内容），此处将会抛出异常\n raw_data.decode('utf-8')\n \n except UnicodeDecodeError:\n # 只要有一个代码文件不是 utf-8 编码，就不满足要求\n return False\n except Exception:\n # 其他读取相关的异常安全起见也视为验证失败\n return False\n \n # 必须至少包含一个代码文件，并且所有代码文件均符合 UTF-8 编码，才算达成目标\n return code_files_found","tags":["Language","File Encoding"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"请确保生成的代码文件中不包含多层if","validation_code":"","tags":["Quantity","Complexity"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用 HTML5 Canvas 和 JavaScript 写一个网页版飞机大战的游戏骨架。首先，创建一个居中的黑色背景画布（比如 400x600）。在画布底部中央绘制一架玩家的飞机（不用找图片，用简单的几个几何图形组合一下，比如一个大三角或者矩形组合，看起来像战斗机就行）。加上键盘事件监听，使得玩家可以通过键盘的左右方向键或者 A/D 键左右平移飞机。注意要做好边界检测，飞机不能飞出画布的最左和最右侧。另外请你后续始终遵循以下要求：你必须在代码写完后测试全部的代码，并确保测试成功通过；同时请注意，你提供的代码文件中不能含有for循环语句。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对网页小游戏，定义可能生成的代码文件扩展名\\n code_exts = {'.html', '.htm', '.js', '.css'}\\n files_checked = 0\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n files_checked += 1\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n # 忽略不包含实际代码/注释文本的纯空白行\\n if not line.strip():\\n continue\\n \\n # 提取行首的前导空白字符（空格或制表符）\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_spaces = match.group(1)\\n \\n # 验证前导空白不能包含制表符\\n if '\\\\t' in leading_spaces:\\n return False\\n \\n # 验证前导空白的长度（空格数）必须固定为4的整数倍\\n if len(leading_spaces) % 4 != 0:\\n return False\\n except Exception:\\n # 忽略无法正常读取的非标准文件异常\\n pass\\n\\n # 如果没有找到任何相关的代码文件，视为未生成项目代码，不满足要求\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def tokenize_js(js_code: str):\\n tokens = []\\n i = 0\\n n = len(js_code)\\n line_number = 1\\n \\n in_line_comment = False\\n in_block_comment = False\\n in_string = False\\n string_char = ''\\n \\n while i < n:\\n char = js_code[i]\\n \\n # 处理换行符，更新全局行号\\n if char == '\\\\n':\\n line_number += 1\\n if in_line_comment:\\n in_line_comment = False\\n i += 1\\n continue\\n \\n if in_line_comment:\\n i += 1\\n continue\\n \\n if in_block_comment:\\n if char == '*' and i + 1 < n and js_code[i+1] == '/':\\n in_block_comment = False\\n i += 2\\n else:\\n if char == '\\\\n':\\n line_number += 1\\n i += 1\\n continue\\n \\n if in_string:\\n if char == '\\\\\\\\':\\n if i + 1 < n:\\n if js_code[i+1] == '\\\\n':\\n line_number += 1\\n i += 2\\n else:\\n i += 1\\n elif char == string_char:\\n in_string = False\\n i += 1\\n else:\\n if char == '\\\\n':\\n line_number += 1\\n i += 1\\n continue\\n \\n # 检测注释或正则表达式\\n if char == '/' and i + 1 < n:\\n if js_code[i+1] == '/':\\n in_line_comment = True\\n i += 2\\n continue\\n elif js_code[i+1] == '*':\\n in_block_comment = True\\n i += 2\\n continue\\n \\n # 简单试探判断是否为正则表达式，以避免正则内的 {} 破坏括号匹配\\n is_regex = False\\n if tokens:\\n prev_type, prev_val, _ = tokens[-1]\\n if prev_type == 'PUNCT' and prev_val in '=(:,[?!&|':\\n is_regex = True\\n elif prev_type == 'WORD' and prev_val in ('return', 'case', 'yield'):\\n is_regex = True\\n \\n if is_regex:\\n start_i = i\\n i += 1 \\n in_char_class = False\\n while i < n:\\n if js_code[i] == '\\\\\\\\':\\n if i + 1 < n:\\n i += 2\\n else:\\n i += 1\\n continue\\n if js_code[i] == '[':\\n in_char_class = True\\n elif js_code[i] == ']':\\n in_char_class = False\\n elif js_code[i] == '/' and not in_char_class:\\n i += 1\\n break\\n elif js_code[i] == '\\\\n':\\n # 若遇换行，判断为并非正则表达式，回退\\n i = start_i\\n break\\n i += 1\\n if i > start_i:\\n tokens.append(('STRING', '<regex>', line_number))\\n continue\\n \\n # 检测字符串起始\\n if char in (\\\"'\\\", '\\\"', '`'):\\n in_string = True\\n string_char = char\\n tokens.append(('STRING', '<str>', line_number))\\n i += 1\\n continue\\n \\n if char.isspace():\\n i += 1\\n continue\\n \\n # 提取标识符或关键字\\n if char.isalpha() or char in '_$':\\n start = i\\n while i < n and (js_code[i].isalnum() or js_code[i] in '_$'):\\n i += 1\\n word = js_code[start:i]\\n tokens.append(('WORD', word, line_number))\\n continue\\n \\n # 提取箭头函数双符组合\\n if char == '=' and i + 1 < n and js_code[i+1] == '>':\\n tokens.append(('PUNCT', '=>', line_number))\\n i += 2\\n continue\\n \\n tokens.append(('PUNCT', char, line_number))\\n i += 1\\n \\n return tokens\\n\\n def check_functions(tokens):\\n # 排除可能会有括号加大括号的控制流关键字\\n control_keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\\n has_func = False\\n i = 0\\n \\n # 第一遍扫描，查找所有函数的起始大括号位置并打上 FUNC_BRACE 标记\\n while i < len(tokens):\\n tok_type, tok_val, line = tokens[i]\\n \\n # 1. 正常 function 关键字函数\\n if tok_type == 'WORD' and tok_val == 'function':\\n j = i + 1\\n while j < len(tokens) and tokens[j][1] != '{':\\n j += 1\\n if j < len(tokens):\\n tokens[j] = ('FUNC_BRACE', '{', tokens[j][2], line)\\n i += 1\\n continue\\n \\n # 2. 箭头函数\\n if tok_type == 'PUNCT' and tok_val == '=>':\\n j = i + 1\\n if j < len(tokens):\\n if tokens[j][1] == '{':\\n tokens[j] = ('FUNC_BRACE', '{', tokens[j][2], line)\\n else:\\n # 这是一个直接返回的表达式的箭头函数，尝试向后寻找表达式边界来测算行数\\n has_func = True\\n exp_start_line = line\\n k = j\\n paren_depth = 0\\n bracket_depth = 0\\n brace_depth_local = 0\\n exp_end_line = tokens[k][2]\\n \\n while k < len(tokens):\\n t_val = tokens[k][1]\\n if t_val == '(': paren_depth += 1\\n elif t_val == ')': \\n if paren_depth == 0: break\\n paren_depth -= 1\\n elif t_val == '[': bracket_depth += 1\\n elif t_val == ']': \\n if bracket_depth == 0: break\\n bracket_depth -= 1\\n elif t_val == '{': brace_depth_local += 1\\n elif t_val == '}': \\n if brace_depth_local == 0: break\\n brace_depth_local -= 1\\n elif t_val in (';', ','):\\n if paren_depth == 0 and bracket_depth == 0 and brace_depth_local == 0:\\n break\\n exp_end_line = tokens[k][2]\\n k += 1\\n lines = exp_end_line - exp_start_line + 1\\n \\n if lines < 15 or lines > 30:\\n return False, True\\n i += 1\\n continue\\n \\n # 3. 类方法、对象的方法 (例如: `update() { ... }`)\\n if tok_type == 'WORD' and tok_val not in control_keywords:\\n if i + 1 < len(tokens) and tokens[i+1][1] == '(':\\n j = i + 1\\n paren_depth = 0\\n while j < len(tokens):\\n if tokens[j][1] == '(': paren_depth += 1\\n elif tokens[j][1] == ')': \\n paren_depth -= 1\\n if paren_depth == 0:\\n break\\n j += 1\\n if j < len(tokens) and paren_depth == 0:\\n if j + 1 < len(tokens) and tokens[j+1][1] == '{':\\n tokens[j+1] = ('FUNC_BRACE', '{', tokens[j+1][2], line)\\n i += 1\\n\\n # 第二遍扫描，进行基于括号层级的行数统计\\n brace_depth = 0\\n active_functions = [] \\n \\n for t in tokens:\\n if t[0] == 'FUNC_BRACE':\\n has_func = True\\n brace_depth += 1\\n # 记录 (当前层级, 函数起始行号)\\n active_functions.append((brace_depth, t[3]))\\n elif t[0] == 'PUNCT' and t[1] == '{':\\n brace_depth += 1\\n elif t[0] == 'PUNCT' and t[1] == '}':\\n # 在遇到结尾 `}` 时，关闭当前层级对应的函数计算\\n for func in active_functions[:]:\\n if func[0] == brace_depth:\\n func_start = func[1]\\n func_end = t[2] \\n lines = func_end - func_start + 1\\n if lines < 15 or lines > 30:\\n return False, True\\n active_functions.remove(func)\\n brace_depth -= 1\\n \\n return True, has_func\\n\\n # 主体逻辑：遍历所有的 JS 和 HTML 文件\\n has_checked_any_function = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ('.js', '.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except:\\n continue\\n \\n blocks = []\\n if ext == '.js':\\n blocks.append(content)\\n elif ext == '.html':\\n script_pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\\n for match in script_pattern.finditer(content):\\n start_index = match.start(1)\\n # 为了保障行号一致性，在提取出的 JS 块前部注入原本包含的相同换行数\\n prefix = content[:start_index]\\n newlines = '\\\\n' * prefix.count('\\\\n')\\n blocks.append(newlines + match.group(1))\\n \\n for block in blocks:\\n tokens = tokenize_js(block)\\n ok, has_f = check_functions(tokens)\\n if not ok:\\n return False\\n if has_f:\\n has_checked_any_function = True\\n\\n # 必须至少有检查到一个函数，才可以视为成功通过\\n return has_checked_any_function\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 针对代码中常见注释形式的正则表达式\\n html_comment_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\\n c_multi_comment_pattern = re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL)\\n c_single_comment_pattern = re.compile(r'//(.*)', re.MULTILINE)\\n hash_single_comment_pattern = re.compile(r'#(.*)', re.MULTILINE)\\n\\n # 常见的代码文件后缀类型\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.py', '.php', '.java', '.c', '.cpp', '.h'}\\n\\n # 遍历代码项目地址\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取该文件内所有注释的内容\\n html_comments = html_comment_pattern.findall(content)\\n c_multi_comments = c_multi_comment_pattern.findall(content)\\n c_single_comments = c_single_comment_pattern.findall(content)\\n hash_single_comments = hash_single_comment_pattern.findall(content)\\n\\n # 合并所有注释\\n all_comments = html_comments + c_multi_comments + c_single_comments + hash_single_comments\\n\\n # 遍历检查提取出的每一段注释文本\\n for comment in all_comments:\\n for char in comment:\\n # 调用预定义的 is_emoji 专家函数判断字符\\n if is_emoji(char):\\n # 若发现任何一个 emoji 表情符号，则不满足要求\\n return False\\n \\n # 如果全部遍历结束均未发现 emoji，则满足给定要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check_no_empty_lines_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 验证工作空间路径是否有效\\n if not workspace_path or not os.path.exists(workspace_path):\\n return True\\n\\n # 推断可能的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 检查是否为代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n \\n # 尝试以不同编码读取文件内容\\n lines = []\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except UnicodeDecodeError:\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n except Exception:\\n continue\\n \\n # 检查是否存在空行\\n for line in lines:\\n if line.strip() == \\\"\\\":\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义可能的游戏相关代码文件类型\\n code_extensions = {'.html', '.js', '.css', '.ts', '.htm'}\\n files_checked = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 去除转义字符，避免转义引号破坏后续匹配\\n content = re.sub(r'\\\\\\\\.', '', content)\\n \\n # 去除所有的字符串文字\\n content = re.sub(r'\\\".*?\\\"', '', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'.*?'\\\", '', content, flags=re.DOTALL)\\n content = re.sub(r\\\"`.*?`\\\", '', content, flags=re.DOTALL)\\n \\n # 去除所有的 HTML 注释、JS/CSS 的多行和单行注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n \\n # 剔除可选链和空值合并语法引发的问号干扰\\n content = content.replace('?.', '')\\n content = content.replace('??', '')\\n \\n # 统计剩下的独立问号字符（即三元运算符的主标志）\\n ternary_count = content.count('?')\\n \\n # 判断当前代码文件是否精确包含2个三元运算符\\n if ternary_count != 2:\\n return False\\n \\n files_checked += 1\\n \\n # 如果没查找到任何属于前端代码的文件，视为不合规\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def check_file_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作空间路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_checked_file = False\\n \\n # 遍历代码项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 逐行读取并统计总行数，避免大文件占用过多内存\\n line_count = sum(1 for _ in f)\\n \\n # 若行数大于200，则不满足要求\\n if line_count > 200:\\n return False\\n \\n has_checked_file = True\\n except UnicodeDecodeError:\\n # 若文件不是文本文件（如图片等二进制格式），则忽略并跳过\\n continue\\n except Exception:\\n # 忽略其他可能导致读取失败的异常（如权限问题等）\\n continue\\n\\n # 确保至少检查了一个生成的代码文件\\n return has_checked_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n valid_extensions = {'.html', '.js', '.css'}\\n files_processed = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file_name in files:\\n if any(file_name.endswith(ext) for ext in valid_extensions):\\n files_processed = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n stripped_line = line.strip()\\n if not stripped_line:\\n continue\\n \\n if (stripped_line.startswith('//') or \\n stripped_line.startswith('/*') or \\n stripped_line.startswith('*') or \\n stripped_line.startswith('<!--')):\\n return False\\n \\n has_inline_comment = False\\n if '/*' in stripped_line or '<!--' in stripped_line:\\n has_inline_comment = True\\n elif '//' in stripped_line:\\n idx = stripped_line.find('//')\\n while idx != -1:\\n if idx > 0 and stripped_line[idx-1] != ':':\\n has_inline_comment = True\\n break\\n idx = stripped_line.find('//', idx + 2)\\n \\n if not has_inline_comment:\\n return False\\n \\n return files_processed\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请注意，你提供的代码文件中不能含有for循环语句\",\n \"validation_code\": \"def check_for_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区目录不存在，可以认为是空项目，没有违反“不包含for循环”的要求\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 匹配 for 循环的正则表达式：for单词后接零或多个空白，再接可选的await及空白，最后是一个左圆括号\\n for_pattern = re.compile(r'\\\\bfor\\\\s*(?:await\\\\s+)?\\\\(')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 针对本次网页游戏骨架生成，主要检查 HTML 和 JS 等代码相关文件\\n if file.endswith(('.html', '.htm', '.js', '.ts', '.jsx', '.tsx')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于权限或编码等问题无法读取，直接跳过\\n continue\\n \\n # 剔除 HTML 注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 剔除 JavaScript 多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 剔除 JavaScript 单行注释\\n content = re.sub(r'//.*', '', content)\\n \\n # 剔除双引号字符串\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', content)\\n # 剔除单引号字符串\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', content)\\n # 剔除反引号（模板）字符串\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', content, flags=re.DOTALL)\\n \\n # 检查剥离字符串和注释后的内容是否存在 for 循环特征\\n if for_pattern.search(content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来加入开火功能。当玩家按下空格键时，可以让飞机发射子弹。子弹用黄色的细长矩形表示，从飞机机头的位置生成，并向上方垂直飞行。为了让游戏体验更好，请增加射击的冷却时间（比如每 200 毫秒最多只能发射一发子弹，不能因为狂按空格就连成一条线）。此外，当子弹飞出画布上边缘时，必须将它从代码的数组里销毁掉，防止数组无限增长。另外请你后续始终遵循以下调整后的要求：原代码文件不能含有for循环的要求调整为你提供的代码文件中不得出现while循环；同时对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断代码文件可能的类型，排除.log等非代码文件\\n code_exts = {'.html', '.js', '.css', '.ts'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_exts):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 忽略纯空行（关于空行的独立要求应由其它测试点负责）\\n if not line.strip():\\n continue\\n \\n # 匹配行首的空白字符序列\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_whitespace = match.group(1)\\n \\n # 缩进中不允许出现制表符(Tab)\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 缩进量必须是 4 个空格的倍数\\n if len(leading_whitespace) % 4 != 0:\\n return False\\n\\n # 若成功找到并检查过代码文件，且均未违规，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def get_function_lengths(code, is_html=False):\\n # 构建字符索引到行号的映射列表\\n line_indices = []\\n lines = code.split('\\\\n')\\n for i, line in enumerate(lines):\\n line_indices.extend([i] * (len(line) + 1))\\n \\n def replacer(m):\\n return ' ' * len(m.group(0))\\n \\n # 替换掉所有的JS单行/多行注释以及字符串(以防止内部包含大括号导致误判)\\n pattern = re.compile(\\n r'//.*?$|/\\\\*.*?\\\\*/|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`',\\n re.MULTILINE | re.DOTALL\\n )\\n \\n if is_html:\\n # 针对HTML文件，提取 <script>...</script> 中的内容\\n script_pattern = re.compile(r'(<script[^>]*>)(.*?)(</script>)', re.IGNORECASE | re.DOTALL)\\n blanked_html = list(' ' * len(code))\\n for m in script_pattern.finditer(code):\\n start, end = m.span(2)\\n blanked_html[start:end] = list(code[start:end])\\n code_to_clean = \\\"\\\".join(blanked_html)\\n else:\\n code_to_clean = code\\n\\n cleaned_code = pattern.sub(replacer, code_to_clean)\\n \\n # 匹配 JS 各种函数定义情况\\n SIMPLE_ID = r'[a-zA-Z_$][a-zA-Z0-9_$]*'\\n # 1. 传统的 function 声明\\n p1 = rf'(?<![a-zA-Z0-9_$])function(?:\\\\s+{SIMPLE_ID})?\\\\s*\\\\(.*?\\\\)\\\\s*'\\n # 2. 有括号的箭头函数 (args) => {\\n p2 = r'\\\\(.*?\\\\)\\\\s*=>\\\\s*'\\n # 3. 无括号的单参数箭头函数 arg => {\\n p3 = rf'(?<![a-zA-Z0-9_$]){SIMPLE_ID}\\\\s*=>\\\\s*'\\n # 4. 类/对象内的方法定义 method() {，需排除控制流关键字\\n p4 = rf'(?<![a-zA-Z0-9_$])(?!(?:if|while|for|switch|catch|with)\\\\b){SIMPLE_ID}\\\\s*\\\\(.*?\\\\)\\\\s*'\\n\\n func_pattern = re.compile(rf'(?:{p1}|{p2}|{p3}|{p4})\\\\{{', re.DOTALL)\\n \\n # 获取所有函数的 '{' 起始索引位置\\n func_starts = [m.end() - 1 for m in func_pattern.finditer(cleaned_code)]\\n \\n lengths = []\\n for start_idx in func_starts:\\n brace_level = 0\\n end_idx = start_idx\\n for i in range(start_idx, len(cleaned_code)):\\n if cleaned_code[i] == '{':\\n brace_level += 1\\n elif cleaned_code[i] == '}':\\n brace_level -= 1\\n if brace_level == 0:\\n end_idx = i\\n break\\n \\n start_line = line_indices[start_idx]\\n end_line = line_indices[end_idx]\\n lengths.append(end_line - start_line + 1)\\n \\n return lengths\\n\\n all_lengths = []\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n all_lengths.extend(get_function_lengths(code, is_html=False))\\n except Exception:\\n pass\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n all_lengths.extend(get_function_lengths(code, is_html=True))\\n except Exception:\\n pass\\n \\n # 如果项目中没有任何函数，视为不满足游戏骨架的基本产出要求\\n if not all_lengths:\\n return False\\n \\n # 任何一个函数的行数必须介于15和30之间\\n for length in all_lengths:\\n if length < 15 or length > 30:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义目标代码文件的后缀名\\n code_extensions = {'.html', '.htm', '.js', '.css'}\\n\\n # 构造正则表达式，用于匹配注释或字符串\\n # (?P<comment>...) 捕获单行注释、多行块注释、HTML注释\\n # (?P<string>...) 捕获双引号、单引号或反引号包裹的字符串，这可以防止诸如 http://... 被误认为注释\\n pattern_str = (\\n r'(?P<comment>//[^\\\\n]*|/\\\\*.*?\\\\*/|<!--.*?-->)'\\n r'|(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n )\\n # 使用 re.DOTALL 使得 .*? 能够匹配跨行内容（仅针对块注释和HTML注释，// 注释已被 [^\\\\n]* 限制）\\n pattern = re.compile(pattern_str, re.DOTALL)\\n\\n # 遍历当前工作区的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于编码或其他问题无法读取，则跳过\\n continue\\n\\n # 查找内容中所有的注释和字符串\\n for match in pattern.finditer(content):\\n comment = match.group('comment')\\n # 如果匹配到的是注释内容\\n if comment:\\n # 遍历注释内容里的每一个字符，判断是否存在 emoji\\n for char in comment:\\n if is_emoji(char):\\n return False # 一旦发现注释里有emoji，返回未满足要求\\n\\n # 所有代码文件的注释中均未发现emoji表情，返回符合要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名白名单，涵盖前端代码及其他常见语言\\n code_extensions = {\\n '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.php', '.go', '.rs', '.swift', '.kt', '.sh'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历代码项目路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行检查\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 若一行代码去除两端空白字符（包含空格、制表符、换行符等）后长度为0，判定为空行\\n if line.strip() == \\\"\\\":\\n return False\\n except Exception:\\n # 忽略可能出现的文件读取异常（如编码错误等）\\n continue\\n \\n # 只要存在至少一个代码文件，并且里面没有任何空行，就满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端相关代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx'}\\n file_count = 0\\n\\n # 正则表达式：用于去除代码中的注释与字符串，避免将它们内部的内容误认为三元运算符\\n # 包含了：多行注释、HTML注释、单行注释、双引号字符串、单引号字符串、反引号多行字符串\\n clean_pattern = re.compile(\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->|//[^\\\\n]*|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n )\\n \\n # 正则表达式：匹配三元运算符的问号 ?\\n # (?<!<) 排除 <? 的情况 (如XML声明等)\\n # \\\\? 匹配问号本身\\n # (?![?.]) 排除紧随其后的是 ? 或 . 的情况，进而排除可选链 `?.` 和空值合并 `??`\\n ternary_pattern = re.compile(r'(?<!<)\\\\?(?![?.])')\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果有文件无法读取或者编码不兼容，按未通过要求处理\\n return False\\n \\n # 第一步：清洗所有注释和字符串内容\\n cleaned_content = clean_pattern.sub('', content)\\n \\n # 第二步：在清洗后的纯代码中匹配满足条件的三元运算符\\n matches = ternary_pattern.findall(cleaned_content)\\n \\n # 若匹配数量不为 2，说明该文件未精确包含2个三元运算符\\n if len(matches) != 2:\\n return False\\n \\n # 如果找到了至少一个代码文件且全都符合要求，返回 True\\n return file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n if len(lines) > 200:\\n return False\\n except UnicodeDecodeError:\\n # 忽略可能存在的一些不可读取的二进制文件或非 utf-8 编码文件\\n continue\\n except Exception:\\n # 忽略其他读取异常（例如权限问题等）\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对网页游戏项目，检查常规的代码文件，忽略如 .log 等非代码文件\\n code_extensions = ['.html', '.js', '.css']\\n has_code_files = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n stripped = line.strip()\\n # 忽略空行\\n if not stripped:\\n continue\\n \\n # 1. 检查是否存在独立行注释（即以注释符号作为该行开头的内容）\\n if stripped.startswith('//') or stripped.startswith('/*') or stripped.startswith('<!--'):\\n return False\\n \\n # 2. 检查该行是否配有行尾注释（即包含至少一种注释标志）\\n if '//' not in stripped and '/*' not in stripped and '<!--' not in stripped:\\n return False\\n\\n # 必须至少包含一个代码文件且全部校验通过才算满足要求\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中不得出现while循环\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 如果工作空间路径不存在，视作没有代码文件，不违背不出现while的要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 遍历工作空间目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令，推断代码文件最可能为 HTML 和 JavaScript 文件\\n if file.endswith(('.html', '.js')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找整个单词 while，防止匹配到类似于 while_loop 这样的变量名\\n # 由于系统要求注释必须为中文，无需担心注释中的英文干扰\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 忽略读取异常的文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来添加敌机系统。请编写一个定时器或利用帧循环，在画布顶部随机的横坐标位置不断生成敌机（可以用红色的方块表示）。生成后，敌机需要持续向下垂直移动。为了增加不可预测性，每架敌机的下落速度可以附带一点随机的差异，大小也可以有轻微的随机变化。同样，如果敌机飞出了画布的下边缘，也要及时从内存中清理掉它们。另外请你后续始终遵循以下调整和新增要求：原要求代码写完后必须测试全部代码并确保通过调整为编写代码结束后，绝对禁止你做任何测试；同时生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def check_indentation_rule(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的代码文件后缀，排除日志等非代码文件\\n code_exts = {'.html', '.js', '.css'}\\n has_code_files = False\\n \\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 去除行末尾的换行符和回车符\\n line_no_newline = line.rstrip('\\\\r\\\\n')\\n \\n # 匹配行首的前导空白字符\\n match = re.match(r'^(\\\\s*)', line_no_newline)\\n if match:\\n leading_ws = match.group(1)\\n # 如果没有前导空白（顶格写），属于合规的无缩进行为，不计入违反\\n if not leading_ws:\\n continue\\n \\n # 检查前导空白是否完全由普通空格组成（不能含有制表符 \\\\t 等）\\n if not all(c == ' ' for c in leading_ws):\\n return False\\n \\n # 检查连续空格的数量是否严格是 4 的倍数\\n if len(leading_ws) % 4 != 0:\\n return False\\n \\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response, workspace_path):\\n js_files = []\\n # 查找项目下所有的.js和.html文件\\n for root, dirs, files in os.walk(workspace_path):\\n for f in files:\\n if f.endswith('.js') or f.endswith('.html'):\\n js_files.append(os.path.join(root, f))\\n \\n if not js_files:\\n return False\\n \\n def strip_comments_and_strings(code_str):\\n # 简单状态机，替换注释和字符串内容为空格，以防正则被干扰\\n # 0:normal, 1:单行注释, 2:多行注释, 3:单引号, 4:双引号, 5:模板字符串\\n state = 0\\n result = []\\n i = 0\\n n = len(code_str)\\n while i < n:\\n c = code_str[i]\\n if state == 0:\\n if c == '/' and i + 1 < n and code_str[i+1] == '/':\\n state = 1\\n result.append(' ')\\n i += 2\\n continue\\n elif c == '/' and i + 1 < n and code_str[i+1] == '*':\\n state = 2\\n result.append(' ')\\n i += 2\\n continue\\n elif c == \\\"'\\\":\\n state = 3\\n result.append(' ')\\n i += 1\\n continue\\n elif c == '\\\"':\\n state = 4\\n result.append(' ')\\n i += 1\\n continue\\n elif c == '`':\\n state = 5\\n result.append(' ')\\n i += 1\\n continue\\n else:\\n result.append(c)\\n i += 1\\n elif state == 1:\\n if c == '\\\\n':\\n state = 0\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 2:\\n if c == '*' and i + 1 < n and code_str[i+1] == '/':\\n state = 0\\n result.append(' ')\\n i += 2\\n continue\\n elif c == '\\\\n':\\n result.append(c)\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 3:\\n if c == '\\\\\\\\' and i + 1 < n:\\n if code_str[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n continue\\n elif c == \\\"'\\\":\\n state = 0\\n result.append(' ')\\n i += 1\\n elif c == '\\\\n':\\n state = 0\\n result.append(c)\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 4:\\n if c == '\\\\\\\\' and i + 1 < n:\\n if code_str[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n continue\\n elif c == '\\\"':\\n state = 0\\n result.append(' ')\\n i += 1\\n elif c == '\\\\n':\\n state = 0\\n result.append(c)\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 5:\\n if c == '\\\\\\\\' and i + 1 < n:\\n if code_str[i+1] == '\\\\n':\\n result.append(' \\\\n')\\n else:\\n result.append(' ')\\n i += 2\\n continue\\n elif c == '`':\\n state = 0\\n result.append(' ')\\n i += 1\\n elif c == '\\\\n':\\n result.append(c)\\n i += 1\\n else:\\n result.append(' ')\\n i += 1\\n return \\\"\\\".join(result)\\n \\n has_functions = False\\n \\n for file_path in js_files:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n if file_path.endswith('.html'):\\n script_blocks = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', code, flags=re.DOTALL | re.IGNORECASE)\\n blocks = script_blocks\\n else:\\n blocks = [code]\\n \\n for block in blocks:\\n text = strip_comments_and_strings(block)\\n function_braces = {}\\n \\n def get_line_num(t, pos):\\n return t.count('\\\\n', 0, pos) + 1\\n\\n def find_matching_brace(t, start_pos):\\n count = 0\\n for i in range(start_pos, len(t)):\\n if t[i] == '{':\\n count += 1\\n elif t[i] == '}':\\n count -= 1\\n if count == 0:\\n return i\\n return -1\\n\\n def add_function(start_pos, brace_pos):\\n if brace_pos in function_braces:\\n old_start = function_braces[brace_pos][0]\\n new_start = get_line_num(text, start_pos)\\n end_line = function_braces[brace_pos][1]\\n function_braces[brace_pos] = (min(old_start, new_start), end_line)\\n else:\\n end_brace_pos = find_matching_brace(text, brace_pos)\\n if end_brace_pos != -1:\\n start_line = get_line_num(text, start_pos)\\n end_line = get_line_num(text, end_brace_pos)\\n function_braces[brace_pos] = (start_line, end_line)\\n \\n # 匹配模式1: function 关键字声明\\n for match in re.finditer(r'\\\\bfunction\\\\b', text):\\n brace_match = re.search(r'\\\\{', text[match.end():])\\n if brace_match:\\n sub_text = text[match.end(): match.end() + brace_match.start()]\\n if '{' not in sub_text and '}' not in sub_text and ';' not in sub_text:\\n brace_pos = match.end() + brace_match.start()\\n add_function(match.start(), brace_pos)\\n \\n # 匹配模式2: 箭头函数\\n for match in re.finditer(r'=>', text):\\n sub_text = text[match.end():]\\n m = re.match(r'\\\\s*\\\\{', sub_text)\\n if m:\\n brace_pos = match.end() + sub_text.find('{')\\n add_function(match.start(), brace_pos)\\n else:\\n start_line = get_line_num(text, match.start())\\n # 对单表达式箭头函数用负数偏移作为占位键\\n function_braces[-match.start() - 1] = (start_line, start_line)\\n\\n # 匹配模式3: ES6 方法或对象方法\\n for match in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', text):\\n method_name = match.group(1)\\n # 排除语法结构上的关键字干扰\\n if method_name not in ['if', 'for', 'while', 'switch', 'catch', 'function', 'with', 'return']:\\n brace_pos = match.start() + match.group(0).find('{')\\n add_function(match.start(), brace_pos)\\n \\n for key, (start, end) in function_braces.items():\\n has_functions = True\\n lines = end - start + 1\\n if lines < 15 or lines > 30:\\n return False\\n \\n return has_functions\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 预设不同类型的注释正则模式\\n html_comment_re = re.compile(r'<!--(.*?)-->', re.DOTALL)\\n block_comment_re = re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL)\\n line_comment_re = re.compile(r'//(.*)$', re.MULTILINE)\\n hash_comment_re = re.compile(r'#(.*)$', re.MULTILINE)\\n \\n # 定义属于代码文件的后缀和 Markdown 代码块语言标识\\n code_extensions = {\\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.java', '.c', '.cpp', \\n '.cs', '.go', '.py', '.rb', '.sh', '.php'\\n }\\n \\n code_langs = {\\n 'html', 'xml', 'js', 'javascript', 'jsx', 'ts', 'typescript', 'tsx',\\n 'css', 'scss', 'less', 'java', 'c', 'cpp', 'cs', 'go', 'py', 'python',\\n 'rb', 'ruby', 'sh', 'bash', 'php'\\n }\\n \\n def check_comments_for_emoji(comments: list) -> bool:\\n \\\"\\\"\\\"检查提取出来的注释列表中是否含有emoji\\\"\\\"\\\"\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n return True\\n\\n def extract_and_check(content: str, ext_or_lang: str) -> bool:\\n \\\"\\\"\\\"根据不同的文件扩展名或语言类型提取注释并验证\\\"\\\"\\\"\\n comments = []\\n ext_or_lang = ext_or_lang.lower()\\n \\n # HTML 往往内嵌JS和CSS，因此将各种注释格式都提出来\\n if ext_or_lang in {'.html', '.htm', 'html', 'xml'}:\\n comments.extend(html_comment_re.findall(content))\\n comments.extend(block_comment_re.findall(content))\\n comments.extend(line_comment_re.findall(content))\\n # 脚本或偏向后端的语言处理\\n elif ext_or_lang in {'.py', '.rb', '.sh', 'py', 'python', 'rb', 'ruby', 'sh', 'bash'}:\\n comments.extend(hash_comment_re.findall(content))\\n # JS、CSS、TS等常规前端与C系语言处理\\n else:\\n comments.extend(block_comment_re.findall(content))\\n comments.extend(line_comment_re.findall(content))\\n if ext_or_lang in {'.php', 'php'}:\\n comments.extend(hash_comment_re.findall(content))\\n \\n return check_comments_for_emoji(comments)\\n\\n # 1. 检查 workspace_path 下生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对代码文件进行注释排查（跳过.log等文件）\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n if not extract_and_check(content, ext):\\n return False\\n except Exception:\\n pass\\n \\n # 2. 检查 response 回复文本中的代码块\\n for match in re.finditer(r'```([a-zA-Z0-9]*)\\\\n(.*?)```', response, re.DOTALL):\\n lang = match.group(1).lower()\\n code_content = match.group(2)\\n \\n # 针对匹配到语言标签属于代码，或者根本没写语言标签的代码块均做检查\\n if lang in code_langs or not lang:\\n ext_lang = lang if lang else 'js'\\n if not extract_and_check(code_content, ext_lang):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能的代码文件后缀\\n code_extensions = {'.html', '.htm', '.js', '.css'}\\n \\n # 如果工作区路径不存在，直接返回 True（或视为未违背规则）\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件，忽略日志文件等其他格式\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行首尾的空格、换行符等空白字符，若为空则判定为空行\\n if not line.strip():\\n return False\\n except Exception:\\n # 读取出错或编码不支持时，跳过该文件继续检查\\n continue\\n \\n # 所有代码文件遍历完毕未发现空行，满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def evaluate_response(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 根据用户指令推断可能生成的代码文件扩展名\\n code_extensions = ['.html', '.js', '.css', '.ts', '.jsx', '.tsx']\\n has_code_file = False\\n \\n def count_ternary_operators(code: str) -> int:\\n # 去除 HTML 注释\\n code = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', code)\\n # 去除 JS/CSS 多行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n # 去除 JS 单行注释\\n code = re.sub(r'//.*', '', code)\\n \\n # 移除字符串 (双引号，单引号，反引号)\\n # 采用高效且防止回溯灾难的正则表达式写法\\n code = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", \\\"''\\\", code)\\n code = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '``', code)\\n \\n # 排除 JavaScript/TypeScript 特有的含有 ? 的非法三元运算符情况\\n # 1. 可选链操作符: ?.\\n code = code.replace('?.', '')\\n # 2. 空值合并操作符: ??\\n code = code.replace('??', '')\\n \\n # 统计纯逻辑语句中剩下的 '?' 符号的数量，即三元运算符的数量\\n return code.count('?')\\n\\n # 遍历当前生成的项目路径\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理推断出的代码文件类型\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 每一个代码文件都必须精确包含 2 个三元运算符\\n if count_ternary_operators(content) != 2:\\n return False\\n \\n # 如果没有找到任何代码文件，则也视为不符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作空间路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 遍历项目中的每一个文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件，忽略可能出现的编码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n line_count = sum(1 for _ in f)\\n # 如果任何文件行数大于 200，则不满足要求\\n if line_count > 200:\\n return False\\n except Exception:\\n # 若读取发生异常，则跳过该文件继续检查\\n continue\\n \\n # 所有文件行数均不超过 200，满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\\n files_checked = 0\\n \\n # 首先检查工作区中的实际代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n files_checked += 1\\n for line in lines:\\n s = line.strip()\\n # 忽略空行\\n if not s:\\n continue\\n \\n # 检查是否为被禁止的独立行注释 (即当前行直接以注释符开头)\\n if s.startswith(\\\"//\\\") or s.startswith(\\\"/*\\\") or s.startswith(\\\"<!--\\\"):\\n return False\\n \\n has_comment = False\\n \\n # 检查是否包含单行注释 //，并排除网址如 http:// 的干扰\\n idx = s.find(\\\"//\\\")\\n while idx != -1:\\n if idx > 0 and s[idx-1] == ':':\\n idx = s.find(\\\"//\\\", idx + 2)\\n else:\\n has_comment = True\\n break\\n \\n # 检查是否包含块注释或HTML注释\\n if not has_comment and \\\"/*\\\" in s:\\n has_comment = True\\n if not has_comment and \\\"<!--\\\" in s:\\n has_comment = True\\n \\n # 如果当前行没有任何注释符，则视为没有配有行尾注释，违反要求\\n if not has_comment:\\n return False\\n \\n # 如果工作区中有代码文件并且全都通过了检查，返回 True\\n if files_checked > 0:\\n return True\\n\\n # 如果工作区未找到相关代码文件，则回退(fallback)检查 response 中的代码块\\n code_blocks = re.findall(r'```(.*?)\\\\n(.*?)```', response, re.DOTALL)\\n valid_blocks = 0\\n \\n for lang, block in code_blocks:\\n lang = lang.strip().lower()\\n # 仅针对前端代码块及未标明语言的代码块进行验证\\n if lang in ['html', 'js', 'javascript', 'css', 'ts', 'typescript', 'jsx', 'tsx', '']:\\n valid_blocks += 1\\n lines = block.split('\\\\n')\\n for line in lines:\\n s = line.strip()\\n if not s:\\n continue\\n \\n if s.startswith(\\\"//\\\") or s.startswith(\\\"/*\\\") or s.startswith(\\\"<!--\\\"):\\n return False\\n \\n has_comment = False\\n idx = s.find(\\\"//\\\")\\n while idx != -1:\\n if idx > 0 and s[idx-1] == ':':\\n idx = s.find(\\\"//\\\", idx + 2)\\n else:\\n has_comment = True\\n break\\n \\n if not has_comment and \\\"/*\\\" in s:\\n has_comment = True\\n if not has_comment and \\\"<!--\\\" in s:\\n has_comment = True\\n \\n if not has_comment:\\n return False\\n\\n return valid_blocks > 0\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"编写代码结束后，绝对禁止你做任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中不得出现while循环\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义可能包含代码的扩展名\\n code_extensions = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除 JavaScript/TypeScript 的单行注释 (// ...)\\n content = re.sub(r'//.*', '', content)\\n # 移除 JavaScript/TypeScript 的多行注释 (/* ... */)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 移除 HTML 注释 (<!-- ... -->)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 正则匹配 while 循环特征，例如 while( 或者是 while (\\n if re.search(r'\\\\bwhile\\\\s*\\\\(', content):\\n return False\\n except Exception:\\n # 如果文件无法按 UTF-8 读取或发生其他读取异常，跳过该文件继续验证\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码后，务必使用 ESLint 进行检查并清除所有提示的报错，从而确保最终的 ESLint 运行无错误返回\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"下一步是核心的战斗结算。请实现子弹和敌机之间的矩形碰撞检测。如果子弹打中敌机，子弹和敌机都要立刻被销毁。同时，在画布左上角增加一个白色的得分面板。为了让击杀反馈更爽快，请在敌机被击毁的位置，短暂地显示一个『+10』的文字特效，要求这个加分文字会缓缓向上飘移，并且在飘移过程中渐渐变透明直到完全消失。另外请你后续始终遵循以下调整和新增要求：原要求生成代码后务必使用ESLint进行检查并清除所有提示的报错调整为对于你产出的代码，严禁使用ESLint进行校验；同时，你所生成的所有代码文件都必须使用UTF-8进行编码。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义目标代码文件的后缀名\\n code_extensions = {'.html', '.js', '.css'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except UnicodeDecodeError:\\n # 尝试读取时忽略无法解析的特殊字符\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n\\n for line in lines:\\n # 忽略没有任何内容的空行（虽然系统提示禁止空行，但在此仅判断缩进要求）\\n if not line.strip():\\n continue\\n\\n leading_spaces = 0\\n has_tab = False\\n \\n # 统计该行的前导空白字符情况\\n for char in line:\\n if char == ' ':\\n leading_spaces += 1\\n elif char == '\\\\t':\\n has_tab = True\\n break\\n else:\\n # 遇到非空格和非制表符（即代码内容的开始），停止统计\\n break\\n \\n # 如果包含了制表符，或者前导空格的数量不是4的整数倍，直接判定失败\\n if has_tab:\\n return False\\n if leading_spaces % 4 != 0:\\n return False\\n\\n # 必须要有至少一个代码文件且全部校验通过才算真正满足了条件\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n all_functions_valid = True\\n found_any_function = False\\n\\n def get_function_start_line(tokens, arrow_idx):\\n if arrow_idx == 0:\\n return tokens[arrow_idx][2]\\n prev_type, prev_val, prev_line = tokens[arrow_idx - 1]\\n \\n if prev_type == 'identifier':\\n return prev_line\\n elif prev_type == ')':\\n paren_count = 1\\n curr_idx = arrow_idx - 2\\n while curr_idx >= 0:\\n ctype, cval, cline = tokens[curr_idx]\\n if ctype == ')':\\n paren_count += 1\\n elif ctype == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n return cline\\n curr_idx -= 1\\n return prev_line \\n \\n return tokens[arrow_idx][2]\\n\\n def is_function_brace(tokens, brace_idx):\\n if brace_idx == 0:\\n return False, -1\\n \\n prev_idx = brace_idx - 1\\n prev_type, prev_val, prev_line = tokens[prev_idx]\\n \\n # 匹配箭头函数\\n if prev_type == '=>':\\n return True, get_function_start_line(tokens, prev_idx)\\n \\n # 匹配 function () { 或者 function foo { (尽管不合法但若存在则算作函数) \\n if prev_type == 'identifier' and prev_val == 'function':\\n return True, prev_line\\n \\n # 匹配带有参数括号的情况: identifier(args) { ... }\\n if prev_type == ')':\\n paren_count = 1\\n curr_idx = prev_idx - 1\\n while curr_idx >= 0:\\n ctype, cval, cline = tokens[curr_idx]\\n if ctype == ')':\\n paren_count += 1\\n elif ctype == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n curr_idx -= 1\\n \\n if paren_count == 0:\\n if curr_idx > 0:\\n before_paren_type, before_paren_val, before_paren_line = tokens[curr_idx - 1]\\n if before_paren_type == 'identifier':\\n if before_paren_val == 'function':\\n return True, before_paren_line\\n # 排除常见控制流之后，剩下的通常是方法名称或普通具名函数名\\n elif before_paren_val not in ('if', 'for', 'while', 'switch', 'catch', 'with'):\\n if curr_idx > 1 and tokens[curr_idx - 2][1] == 'function':\\n return True, tokens[curr_idx - 2][2]\\n return True, before_paren_line\\n elif before_paren_type == '=>':\\n pass\\n \\n # 兼容匿名闭包 (function() {...})()\\n if curr_idx + 1 < prev_idx and tokens[curr_idx + 1][1] == 'function':\\n return True, tokens[curr_idx + 1][2]\\n \\n return False, -1\\n\\n def process_js(js_code):\\n nonlocal all_functions_valid, found_any_function\\n lines = js_code.split('\\\\n')\\n tokens = []\\n \\n in_block_comment = False\\n in_string = False\\n string_char = ''\\n \\n # 将代码转化为词法 tokens 列表并过滤干扰项\\n i = 0\\n while i < len(lines):\\n line = lines[i]\\n j = 0\\n while j < len(line):\\n if in_block_comment:\\n if j + 1 < len(line) and line[j:j+2] == '*/':\\n in_block_comment = False\\n j += 2\\n continue\\n j += 1\\n continue\\n \\n if in_string:\\n if line[j] == '\\\\\\\\':\\n j += 2\\n continue\\n if line[j] == string_char:\\n in_string = False\\n j += 1\\n continue\\n \\n if j + 1 < len(line) and line[j:j+2] == '//':\\n break \\n \\n if j + 1 < len(line) and line[j:j+2] == '/*':\\n in_block_comment = True\\n j += 2\\n continue\\n \\n if line[j] in ('\\\"', \\\"'\\\", '`'):\\n in_string = True\\n string_char = line[j]\\n j += 1\\n continue\\n \\n c = line[j]\\n if c in '{}()':\\n tokens.append((c, c, i))\\n j += 1\\n elif c.isalpha() or c == '_' or c == '$':\\n start = j\\n while j < len(line) and (line[j].isalnum() or line[j] in '_$'):\\n j += 1\\n val = line[start:j]\\n tokens.append(('identifier', val, i))\\n elif j + 1 < len(line) and line[j:j+2] == '=>':\\n tokens.append(('=>', '=>', i))\\n j += 2\\n else:\\n j += 1\\n i += 1\\n \\n # 根据大括号配对解析并校验函数长度\\n stack = []\\n for idx, (t_type, t_val, t_line) in enumerate(tokens):\\n if t_type == '{':\\n is_func, start_line = is_function_brace(tokens, idx)\\n stack.append((is_func, start_line))\\n elif t_type == '}':\\n if stack:\\n is_func, start_line = stack.pop()\\n if is_func:\\n found_any_function = True\\n length = t_line - start_line + 1\\n if not (15 <= length <= 30):\\n all_functions_valid = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if file.endswith('.html'):\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n process_js(script)\\n else:\\n process_js(content)\\n\\n if not found_any_function:\\n return False\\n \\n return all_functions_valid\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义常见前端代码文件的扩展名\\n code_extensions = {'.html', '.htm', '.js', '.jsx', '.ts', '.tsx', '.css'}\\n \\n # 编译正则表达式，用于解析 字符串常量、HTML注释 和 JS/CSS的块/行注释。\\n # 优先匹配字符串是为了防止正则引擎将包含在字符串中的 `//` 或 `/*` 错误识别为注释（例如URL：http://...）\\n pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<html_comment><!--[\\\\s\\\\S]*?-->)'\\n r'|(?P<js_comment>//.*|/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n \\n # 如果代码项目路径不存在，无法判定，默认可以返回False或者视具体情况\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取文件内容，使用utf-8编码，并忽略可能的解码错误以防意外中断\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 遍历查找代码文件中的所有匹配项\\n for match in pattern.finditer(content):\\n # 提取匹配到的注释内容；如果是字符串常量，则该部分为空\\n comment = match.group('html_comment') or match.group('js_comment')\\n \\n # 如果确实匹配到了注释内容\\n if comment:\\n # 遍历注释内容中的每一个字符\\n for char in comment:\\n # is_emoji 为已提供的专家函数，直接判断当前字符是否属于emoji\\n if is_emoji(char):\\n return False\\n except Exception:\\n # 若因其他异常导致部分文件无法读取，跳过继续处理剩余文件\\n continue\\n \\n # 如果遍历所有代码文件均未在注释中发现 emoji，则要求满足\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义前端网页开发中常见的代码文件后缀\\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n lines = content.split('\\\\n')\\n \\n # 排除文件末尾单独的一个换行符导致的最后一个空字符串元素\\n if len(lines) > 0 and lines[-1] == \\\"\\\":\\n lines = lines[:-1]\\n \\n # 遍历判断每一行是否为空行\\n for line in lines:\\n if line.strip() == \\\"\\\":\\n return False # 发现空行，未满足要求\\n except Exception:\\n # 对于非 UTF-8 编码读取失败的文件或特殊情况，予以跳过\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def check_ternary_operator_count(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_extensions = {'.js', '.html', '.css', '.htm'}\\n code_files = []\\n \\n # 查找工作区下所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in code_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n code = f.read()\\n except Exception:\\n return False\\n \\n # 若为 HTML 文件，提取其中 <script> 标签内的内容避免受到普通文本内问号的干扰\\n if file_path.lower().endswith('.html') or file_path.lower().endswith('.htm'):\\n scripts = re.findall(r'<script[\\\\s\\\\S]*?>([\\\\s\\\\S]*?)</script>', code, re.IGNORECASE)\\n code_to_check = '\\\\n'.join(scripts)\\n else:\\n code_to_check = code\\n\\n # 正则表达式模式：匹配字符串和注释\\n pattern = r'''\\n \\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\" | # 双引号字符串\\n '(?:\\\\\\\\.|[^'\\\\\\\\])*' | # 单引号字符串\\n `(?:\\\\\\\\.|[^`\\\\\\\\])*` | # 模板字符串\\n //[^\\\\n]* | # 单行注释\\n /\\\\*[\\\\s\\\\S]*?\\\\*/ | # 多行注释\\n <!--[\\\\s\\\\S]*?--> # HTML 注释\\n '''\\n regex = re.compile(pattern, re.VERBOSE)\\n # 移除代码中的字符串和注释内容\\n cleaned_code = regex.sub('', code_to_check)\\n \\n # 移除可选链操作符(?:.) 和 空值合并运算符(?:??)\\n cleaned_code = cleaned_code.replace('?.', '').replace('??', '')\\n \\n # 在去除了干扰项后，剩下的问号 '?' 即代表了三元运算符的数量\\n if cleaned_code.count('?') != 2:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def check_file_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n line_count = 0\\n for _ in f:\\n line_count += 1\\n if line_count > 200:\\n return False\\n except Exception:\\n # 忽略由于编码问题（如二进制文件）或其他原因无法读取的文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def check_end_of_line_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常规的代码文件类型，排除类似.log等非代码文件\\n code_exts = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs'}\\n \\n # 匹配在行尾出现的注释（允许注释后仅跟随空白字符）\\n # 使用了 (?:(?!\\\\*/).)* 和 (?:(?!-->).)* 以保证定位非贪婪的正确闭合区域，避免吃掉同一行中可能存在的内部有效代码\\n pattern = re.compile(r'(//.*|#.*|/\\\\*(?:(?!\\\\*/).)*\\\\*/|<!--(?:(?!-->).)*-->)\\\\s*$')\\n has_code_file = False\\n \\n def remove_strings(code_line: str) -> str:\\n res = []\\n in_string = False\\n string_char = ''\\n escape = False\\n for char in code_line:\\n if not in_string:\\n if char in (\\\"'\\\", '\\\"', '`'):\\n in_string = True\\n string_char = char\\n res.append(char)\\n else:\\n res.append(char)\\n else:\\n if escape:\\n res.append(' ')\\n escape = False\\n elif char == '\\\\\\\\':\\n escape = True\\n res.append(' ')\\n elif char == string_char:\\n in_string = False\\n res.append(char)\\n else:\\n res.append(' ')\\n return \\\"\\\".join(res)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except UnicodeDecodeError:\\n continue\\n \\n for line in lines:\\n if not line.strip():\\n # 跳过空白行，空行是否允许由其他的系统提示限制判断，此项特指“代码行”\\n continue\\n \\n # 避免字符串内容及常见未被带引号URL造成的 // 误判干扰\\n processed_line = remove_strings(line)\\n processed_line = processed_line.replace('http://', 'http: ').replace('https://', 'https: ')\\n \\n # 1. 检查当前行是否存在位于行尾的注释闭合\\n if not pattern.search(processed_line):\\n return False\\n \\n # 2. 检查是否为独立行注释（即去除一切符合规则的注释部分后，观察是否还有实质代码）\\n temp = processed_line\\n temp = re.sub(r'//.*', '', temp)\\n temp = re.sub(r'#.*', '', temp)\\n temp = re.sub(r'/\\\\*.*?\\\\*/', '', temp)\\n temp = re.sub(r'<!--.*?-->', '', temp)\\n \\n if not temp.strip():\\n # 去除所有注释后没有实质代码可见字符了，属于严格被禁的独立行注释\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"编写代码结束后，绝对禁止你做任何测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中不得出现while循环\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def has_while_loop(code_content: str) -> bool:\\n # 去除 HTML 注释\\n code_content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', code_content)\\n # 去除 多行注释 (/* ... */)\\n code_content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code_content)\\n # 去除 单行注释 (// ...)\\n code_content = re.sub(r'//.*', '', code_content)\\n \\n # 查找 while 关键字后面可能伴随空白符并紧跟左括号的情况\\n if re.search(r'\\\\bwhile\\\\s*\\\\(', code_content):\\n return True\\n return False\\n\\n code_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx'}\\n\\n # 1. 遍历并检查 workspace_path 下生成的实际代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if has_while_loop(content):\\n return False\\n except Exception:\\n continue\\n \\n # 2. 检查 response 中的代码块 (包括 markdown 格式下的各类前端代码块)\\n code_blocks = re.findall(r'```[a-zA-Z]*\\\\s*\\\\n([\\\\s\\\\S]*?)```', response)\\n for block in code_blocks:\\n if has_while_loop(block):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"对于你产出的代码，严禁使用ESLint进行校验\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的代码文件或日志文件后缀\\n code_extensions = {'.html', '.js', '.css', '.log'}\\n \\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_checked_files = False\\n \\n # 遍历项目工作区路径\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_checked_files = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码模式读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明该文件不是合法的 UTF-8 编码（例如使用了 GBK 并包含中文注释）\\n return False\\n except Exception:\\n # 若遇到其他读取异常，同样视为验证失败\\n return False\\n \\n # 只有当成功检查了至少一个代码文件且均无异常时，才返回 True\\n return has_checked_files\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在来完善玩家的生存和游戏失败机制。给玩家设定初始 3 点生命值，用几颗小红心或纯文字显示在画布右上角。如果有敌机直接撞击到玩家的飞机，或者有敌机成功逃脱飞出了屏幕下边界，玩家都要扣除 1 点生命值。玩家受伤扣血时，让整个画布背景短暂闪烁一下红色作为警告。当生命值降为 0 时，游戏暂停（停止所有的物理运动和生成），在画布正中央渲染大字号的『Game Over』，并在其下方绘制一个『重新开始』的按钮形状，点击该区域后可以完全重置游戏状态（清空实体、恢复生命、分数归零）。另外请你后续始终遵循以下调整和新增要求：原要求编写代码结束后绝对禁止你做任何测试调整为编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过；同时处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 可能的代码文件类型\\n code_extensions = {'.html', '.js', '.css'}\\n file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 忽略没有任何非空白字符的空行\\n if not line.strip():\\n continue\\n \\n # 提取前导空白字符\\n leading_spaces = \\\"\\\"\\n for char in line:\\n if char in [' ', '\\\\t']:\\n leading_spaces += char\\n else:\\n break\\n \\n # 规则1：不能出现制表符 (Tab) 缩进\\n if '\\\\t' in leading_spaces:\\n return False\\n \\n # 规则2：缩进量必须是 4 的倍数\\n if len(leading_spaces) % 4 != 0:\\n # 豁免多行注释的特殊情况（形如 \\\" * \\\" 开头的行通常会多出不满足倍数要求的单个空格）\\n if line.lstrip().startswith('*'):\\n continue\\n \\n return False\\n\\n return file_found\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def extract_js_from_html(html_content):\\n # 提取 HTML 文件中所有的 <script> 标签内容\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', html_content, flags=re.DOTALL | re.IGNORECASE)\\n return \\\"\\\\n\\\".join(scripts)\\n \\n def remove_strings_and_comments(code):\\n # 将 JS 中的字符串内容和注释替换为空格，但保留换行符以维持原有行号准确无误\\n new_code = []\\n i = 0\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < len(code):\\n c = code[i]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n new_code.append(c)\\n else:\\n new_code.append(' ')\\n i += 1\\n elif in_block_comment:\\n if c == '*' and i + 1 < len(code) and code[i+1] == '/':\\n in_block_comment = False\\n new_code.append(' ')\\n new_code.append(' ')\\n i += 2\\n elif c == '\\\\n':\\n new_code.append(c)\\n i += 1\\n else:\\n new_code.append(' ')\\n i += 1\\n elif in_string:\\n if c == '\\\\\\\\' and i + 1 < len(code):\\n new_code.append(' ')\\n new_code.append(' ')\\n if code[i+1] == '\\\\n':\\n new_code[-1] = '\\\\n'\\n i += 2\\n elif c == string_char:\\n in_string = False\\n new_code.append(' ')\\n i += 1\\n elif c == '\\\\n':\\n new_code.append(c)\\n i += 1\\n else:\\n new_code.append(' ')\\n i += 1\\n else:\\n if c == '/' and i + 1 < len(code) and code[i+1] == '/':\\n in_line_comment = True\\n new_code.append(' ')\\n new_code.append(' ')\\n i += 2\\n elif c == '/' and i + 1 < len(code) and code[i+1] == '*':\\n in_block_comment = True\\n new_code.append(' ')\\n new_code.append(' ')\\n i += 2\\n elif c in (\\\"'\\\", '\\\"', '`'):\\n in_string = True\\n string_char = c\\n new_code.append(' ')\\n i += 1\\n else:\\n new_code.append(c)\\n i += 1\\n return \\\"\\\".join(new_code)\\n\\n code_files = []\\n for root, dirs, files in os.walk(workspace_path):\\n for f in files:\\n if f.endswith('.js') or f.endswith('.html'):\\n code_files.append(os.path.join(root, f))\\n \\n if not code_files:\\n return True\\n \\n for f in code_files:\\n try:\\n with open(f, 'r', encoding='utf-8') as file:\\n content = file.read()\\n except Exception:\\n continue\\n \\n if f.endswith('.html'):\\n content = extract_js_from_html(content)\\n \\n if not content.strip():\\n continue\\n \\n # 清理字符串和注释内容，以防止对 `{` 和 `}` 的误判\\n clean_code = remove_strings_and_comments(content)\\n \\n i = 0\\n stack = []\\n \\n while i < len(clean_code):\\n # 检查是否有单行箭头函数 (例如 () => expression)\\n if clean_code[i:i+2] == '=>':\\n j = i + 2\\n while j < len(clean_code) and clean_code[j] in ' \\\\t\\\\r\\\\n':\\n j += 1\\n if j < len(clean_code) and clean_code[j] != '{':\\n # 如果不存在大括号且为单行箭头函数，其行数必然是1行，不满足 15 <= lines <= 30 的要求\\n return False\\n\\n c = clean_code[i]\\n if c == '{':\\n # 寻找紧挨着 { 前的非空白字符\\n j = i - 1\\n while j >= 0 and clean_code[j] in ' \\\\t\\\\r\\\\n':\\n j -= 1\\n \\n is_function = False\\n start_line_index = clean_code.count('\\\\n', 0, i)\\n \\n if j >= 0:\\n # 截取大括号前侧的文本作为判定依据\\n start_look = max(0, j - 100)\\n pre_text = clean_code[start_look:j+1]\\n \\n # 匹配普通函数 function ... ()\\n if re.search(r'\\\\bfunction\\\\b\\\\s*(\\\\w+\\\\s*)?\\\\(', pre_text):\\n is_function = True\\n match_func = re.search(r'\\\\bfunction\\\\b', pre_text)\\n if match_func:\\n start_pos = start_look + match_func.start()\\n start_line_index = clean_code.count('\\\\n', 0, start_pos)\\n # 匹配多行箭头函数 => {\\n elif re.search(r'=>$', pre_text):\\n is_function = True\\n match_arrow = re.search(r'=>', pre_text)\\n if match_arrow:\\n start_pos = start_look + match_arrow.start()\\n start_line_index = clean_code.count('\\\\n', 0, start_pos)\\n # 匹配对象的方法或是类方法 name() { \\n elif re.search(r'\\\\w+\\\\s*\\\\([^)]*\\\\)$', pre_text):\\n # 排除掉 if、for、while、switch、catch 等带括号的代码块\\n if not re.search(r'\\\\b(if|for|while|switch|catch)\\\\b\\\\s*\\\\([^)]*\\\\)$', pre_text):\\n is_function = True\\n match_method = re.search(r'\\\\b\\\\w+\\\\s*\\\\(', pre_text)\\n if match_method:\\n start_pos = start_look + match_method.start()\\n start_line_index = clean_code.count('\\\\n', 0, start_pos)\\n \\n # 入栈当前的起始行\\n stack.append((start_line_index, is_function))\\n \\n elif c == '}':\\n if stack:\\n start_line_index, is_function = stack.pop()\\n if is_function:\\n end_line_index = clean_code.count('\\\\n', 0, i)\\n # 计算函数的行数 (包含首尾两行)\\n func_lines = end_line_index - start_line_index + 1\\n if func_lines < 15 or func_lines > 30:\\n return False\\n i += 1\\n \\n # 若所有提取出的函数均在 [15, 30] 范围内，则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 若工作区路径不存在，则无法判定存在违规代码，默认返回 True\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n code_extensions = ['.html', '.js', '.css']\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 仅处理根据用户指令推断出的网页类型代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 使用正则表达式去除字符串面值内容（支持去除单引号、双引号和反引号所包裹的字符串）\\n # 这样可以避免字符串内部类似 http:// 的内容被错误判断为单行注释\\n try:\\n content_no_str = re.sub(r'([\\\"\\\\'`])(?:\\\\\\\\.|(?!\\\\1).)*?\\\\1', '', content)\\n except Exception:\\n # 如果正则处理出现异常，则降级为使用原始内容\\n content_no_str = content\\n \\n # 提取各种格式的注释文本\\n comments = []\\n comments.extend(re.findall(r'//(.*)', content_no_str))\\n comments.extend(re.findall(r'/\\\\*([\\\\s\\\\S]*?)\\\\*/', content_no_str))\\n comments.extend(re.findall(r'<!--([\\\\s\\\\S]*?)-->', content_no_str))\\n \\n # 校验提取出的注释文本中是否含有 emoji 符号\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀，涵盖HTML/JS项目以及其他常用语言\\n code_extensions = {\\n '.js', '.html', '.css', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.cs', '.php'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断文件是否为代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 判定空行：去除头尾空白字符后为空字符串\\n if line.strip() == \\\"\\\":\\n return False\\n except Exception:\\n # 若文件由于编码等问题无法读取，忽略该文件跳过\\n continue\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的代码文件扩展名，排除.log等非代码文件\\n code_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n # 构造综合正则表达式，用于匹配并抹除：\\n # 多行注释、HTML注释、单行注释、双引号字符串、单引号字符串、反引号模板字符串\\n # 使用三引号的原生字符串以避免Python中的转义和单双引号解析混乱\\n pattern = r\\\"\\\"\\\"/\\\\*[\\\\s\\\\S]*?\\\\*/|<!--[\\\\s\\\\S]*?-->|//.*|\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\[\\\\s\\\\S][^\\\"\\\\\\\\]*)*\\\"|'[^'\\\\\\\\]*(?:\\\\\\\\[\\\\s\\\\S][^'\\\\\\\\]*)*'|`[^`\\\\\\\\]*(?:\\\\\\\\[\\\\s\\\\S][^`\\\\\\\\]*)*`\\\"\\\"\\\"\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件存在格式异常无法用utf-8读取，跳过即可，这有可能导致漏判或假阴性\\n continue\\n \\n # 剔除所有的代码注释及字符串字面量片段，避免它们内部的'?'符号产生干扰\\n text = re.sub(pattern, '', content)\\n \\n # 统计剩余纯代码文本中独立的 '?'\\n # (?<!\\\\?) 排除前面的?，(?!\\\\.|\\\\?) 排除后面的.或?，从而有效避开 '?.' 和 '??'\\n matches = re.findall(r'(?<!\\\\?)\\\\?(?!\\\\.|\\\\?)', text)\\n \\n # 判断当前代码文件中的三元运算符数量是否精确为2\\n if len(matches) != 2:\\n return False\\n \\n # 所有代码文件都精确满足时返回True，前提是必须存在至少一个代码文件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 遍历项目地址下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以UTF-8格式读取文件内容\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n line_count = 0\\n for line in f:\\n line_count += 1\\n # 提早退出，一旦行数超过200行则直接返回False\\n if line_count > 200:\\n return False\\n except Exception:\\n # 忽略文件读取过程中可能出现的异常（如无读取权限或特殊文件）\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义我们要检查的代码文件扩展名\\n valid_extensions = {'.js', '.html', '.css'}\\n # 定义常见的注释标识符\\n comment_markers = ['//', '/*', '<!--']\\n \\n has_code_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于某种原因无法读取文件，可以跳过该文件处理\\n continue\\n \\n for line in lines:\\n stripped = line.strip()\\n \\n # 忽略空行\\n if not stripped:\\n continue\\n \\n # 1. 检查是否存在独立行注释 (即仅有注释没有代码)\\n # 如果字符串以任意注释符开头，说明是独立注释，不符合要求\\n is_independent = False\\n for marker in comment_markers:\\n if stripped.startswith(marker):\\n is_independent = True\\n break\\n \\n if is_independent:\\n return False\\n \\n # 2. 检查该行代码是否配有行尾注释\\n # 即该行不仅要有代码，还要在结尾或中间包含注释符\\n has_inline = False\\n for marker in comment_markers:\\n if marker in stripped:\\n has_inline = True\\n break\\n \\n if not has_inline:\\n return False\\n\\n # 必须有代码文件被解析到且全部满足条件，才返回True\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中不得出现while循环\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n # 如果项目目录不存在，默认未违反要求\\n return True\\n \\n # 定义需要检查的代码文件扩展名\\n valid_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 移除 HTML 注释 <!-- ... -->\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 移除 JS/TS 多行注释 /* ... */\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 移除 JS/TS 单行注释 // ...\\n content = re.sub(r'//.*', '', content)\\n \\n # 查找是否存在 while 关键字后紧跟可能的空格与左括号的结构\\n # 这可以准确识别 while (...) 和 do ... while (...)\\n if re.search(r'\\\\bwhile\\\\s*\\\\(', content):\\n return False\\n except Exception:\\n # 忽略读取异常跳过文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"对于你产出的代码，严禁使用ESLint进行校验\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推断出的潜在代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.ts'}\\n found_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行编码检测\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n \\n # 尝试严格使用 UTF-8 格式解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 只要有一个代码文件无法用 UTF-8 解码，说明不满足要求\\n return False\\n except Exception:\\n # 其他读取异常忽略，继续检查其他文件\\n pass\\n\\n # 如果存在代码文件且全部通过了 UTF-8 解码，则满足要求\\n return found_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"游戏目前比较单调，我们加入难度递增机制和特殊道具。第一，引入等级机制：分数每达到 100 分的整数倍时，游戏提升一级，并在屏幕中央闪烁提示『Level Up!』。等级越高，敌机的生成频率越快，整体下落速度也越快。第二，加入补给包：每隔约 15 秒，从顶部掉落一个蓝色的圆形『火力补给包』。如果玩家飞机接到了它，在接下来的 5 秒内，玩家飞机的射击模式临时变为『双发模式』（即每次按空格，会在机翼两侧平行发射两发子弹），5秒后恢复单发。另外请你后续始终遵循以下调整和新增要求：原要求严禁使用ESLint校验你产出的代码调整为你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过；同时，执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def is_valid_indent(line: str) -> bool:\\n # 匹配前导空白字符\\n match = re.match(r'^([ \\\\t]*)', line)\\n if not match:\\n return True\\n leading_ws = match.group(1)\\n\\n # 剥离换行符后的整行内容\\n content = line.rstrip('\\\\r\\\\n')\\n \\n # 如果整行都是空白字符（空行），只要没有制表符即认为合规\\n if len(leading_ws) == len(content):\\n return '\\\\t' not in leading_ws\\n\\n # 如果前导空白中含有制表符，直接判定为不合规\\n if '\\\\t' in leading_ws:\\n return False\\n\\n # 去除前导空白后剩下的可见内容\\n stripped = content.lstrip(' \\\\t')\\n\\n # 容错：多行注释中的中间行（通常以 '*' 开头对齐）可能会有一或额外几个空格\\n if stripped.startswith('*') or stripped.startswith('<!--') or stripped.startswith('-->'):\\n return True\\n\\n # 核心判断：前导空格数量必须是 4 的整数倍\\n if len(leading_ws) % 4 != 0:\\n return False\\n\\n return True\\n\\n valid_extensions = ['.html', '.css', '.js', '.jsx', '.ts', '.tsx']\\n has_checked_code = False\\n\\n # 1. 优先检查 workspace_path 中的文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file_name in files:\\n if any(file_name.endswith(ext) for ext in valid_extensions):\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n if not is_valid_indent(line):\\n return False\\n has_checked_code = True\\n except Exception:\\n continue\\n\\n # 2. 如果工作区没有生成有效代码文件，回退到检查 response 中的代码块\\n if not has_checked_code:\\n # 匹配 Markdown 中的代码块\\n code_blocks = re.findall(r'```(?:html|javascript|js|css|typescript|ts)?\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n if not code_blocks:\\n # 既没有物理文件也没有代码块，无法确认满足条件\\n return False\\n \\n for block in code_blocks:\\n for line in block.split('\\\\n'):\\n if not is_valid_indent(line):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n def mask_js(code: str) -> str:\\n \\\"\\\"\\\"\\n 屏蔽代码中的字符串和注释，将其替换为空格（保留换行符保证行号不变）\\n \\\"\\\"\\\"\\n masked = []\\n i = 0\\n n = len(code)\\n \\n in_single = False\\n in_double = False\\n in_backtick = False\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < n:\\n c = code[i]\\n if in_single:\\n if c == '\\\\\\\\':\\n masked.append(' ')\\n i += 1\\n if i < n:\\n masked.append('\\\\n' if code[i] == '\\\\n' else ' ')\\n i += 1\\n elif c == \\\"'\\\":\\n in_single = False\\n masked.append(' ')\\n i += 1\\n else:\\n masked.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n elif in_double:\\n if c == '\\\\\\\\':\\n masked.append(' ')\\n i += 1\\n if i < n:\\n masked.append('\\\\n' if code[i] == '\\\\n' else ' ')\\n i += 1\\n elif c == '\\\"':\\n in_double = False\\n masked.append(' ')\\n i += 1\\n else:\\n masked.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n elif in_backtick:\\n if c == '\\\\\\\\':\\n masked.append(' ')\\n i += 1\\n if i < n:\\n masked.append('\\\\n' if code[i] == '\\\\n' else ' ')\\n i += 1\\n elif c == \\\"`\\\":\\n in_backtick = False\\n masked.append(' ')\\n i += 1\\n else:\\n masked.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n elif in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n masked.append('\\\\n')\\n i += 1\\n else:\\n masked.append(' ')\\n i += 1\\n elif in_block_comment:\\n if c == '*' and i + 1 < n and code[i+1] == '/':\\n in_block_comment = False\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n else:\\n masked.append('\\\\n' if c == '\\\\n' else ' ')\\n i += 1\\n else:\\n if c == \\\"'\\\" :\\n in_single = True\\n masked.append(' ')\\n i += 1\\n elif c == '\\\"':\\n in_double = True\\n masked.append(' ')\\n i += 1\\n elif c == '`':\\n in_backtick = True\\n masked.append(' ')\\n i += 1\\n elif c == '/' and i + 1 < n and code[i+1] == '/':\\n in_line_comment = True\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n elif c == '/' and i + 1 < n and code[i+1] == '*':\\n in_block_comment = True\\n masked.append(' ')\\n masked.append(' ')\\n i += 2\\n else:\\n masked.append(c)\\n i += 1\\n return \\\"\\\".join(masked)\\n\\n def get_function_line_counts(code: str) -> list:\\n masked_code = mask_js(code)\\n \\n # 预计算行起始索引以加快行号查找\\n line_starts = [0]\\n for i, c in enumerate(code):\\n if c == '\\\\n':\\n line_starts.append(i + 1)\\n \\n def get_line_num(idx: int) -> int:\\n return bisect.bisect_right(line_starts, idx)\\n \\n def get_prev_non_ws(start_idx: int) -> int:\\n curr = start_idx\\n while curr >= 0 and masked_code[curr].isspace():\\n curr -= 1\\n return curr\\n\\n def is_function_start(idx: int) -> bool:\\n p1 = get_prev_non_ws(idx - 1)\\n if p1 < 0:\\n return False\\n \\n # 1. 遇到前一个有效字符是箭头函数的 '=>'\\n if masked_code[p1] == '>' and p1 > 0 and masked_code[p1-1] == '=':\\n return True\\n \\n # 2. 遇到前一个是 ')' 对应常规函数/方法结构\\n if masked_code[p1] == ')':\\n depth = 0\\n p2 = p1\\n while p2 >= 0:\\n if masked_code[p2] == ')':\\n depth += 1\\n elif masked_code[p2] == '(':\\n depth -= 1\\n if depth == 0:\\n break\\n p2 -= 1\\n \\n if p2 < 0:\\n return False\\n \\n p3 = get_prev_non_ws(p2 - 1)\\n if p3 < 0:\\n return True\\n \\n end_ident = p3\\n while p3 >= 0 and (masked_code[p3].isalnum() or masked_code[p3] in ('_', '$')):\\n p3 -= 1\\n \\n if p3 == end_ident:\\n # 匿名函数没有名称，如 `function() {` 或者 `(function() {`\\n return True\\n \\n ident = masked_code[p3+1 : end_ident+1]\\n # 排除由于控制流等引起的 `) {` 结构\\n reserved = {'if', 'for', 'while', 'catch', 'switch', 'with', 'await'}\\n if ident in reserved:\\n return False\\n return True\\n \\n return False\\n\\n function_lengths = []\\n stack = []\\n \\n i = 0\\n n = len(masked_code)\\n while i < n:\\n c = masked_code[i]\\n \\n # 检测并处理无花括号的单行箭头函数 (例如 `const f = a => a + 1;`)\\n if c == '=' and i + 1 < n and masked_code[i+1] == '>':\\n nxt = i + 2\\n while nxt < n and masked_code[nxt].isspace():\\n nxt += 1\\n if nxt < n and masked_code[nxt] != '{':\\n # 无花括号的箭头函数通常只有1行或极短，默认长度记为1（它必将违背 15-30 的要求，成功阻截）\\n function_lengths.append(1)\\n i += 2\\n continue\\n \\n if c == '{':\\n is_func = is_function_start(i)\\n start_line = get_line_num(i)\\n stack.append((is_func, start_line))\\n i += 1\\n continue\\n \\n if c == '}':\\n if stack:\\n is_func, start_line = stack.pop()\\n if is_func:\\n end_line = get_line_num(i)\\n function_lengths.append(end_line - start_line + 1)\\n i += 1\\n continue\\n \\n i += 1\\n \\n return function_lengths\\n\\n js_codes = []\\n \\n # 获取所有的 JS 和 HTML 文件内容\\n for root, dirs, files in os.walk(workspace_path):\\n # 避开不需要检测的三方库及构建缓存目录\\n for exclude_dir in ['node_modules', '.git', 'dist', 'build', 'coverage', '__pycache__']:\\n if exclude_dir in dirs:\\n dirs.remove(exclude_dir)\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n js_codes.append(f.read())\\n except Exception:\\n pass\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 使用正则提取出所有script块的内容\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n js_codes.extend(scripts)\\n except Exception:\\n pass\\n \\n all_lengths = []\\n for code in js_codes:\\n all_lengths.extend(get_function_line_counts(code))\\n \\n if not all_lengths:\\n # 如果项目中没有任何函数代码被解析出来，视为未达标准（通常一个合格的Canvas骨架肯定包含函数）\\n return False\\n \\n for length in all_lengths:\\n if not (15 <= length <= 30):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 构造正则表达式，用于安全、准确地提取注释片段。\\n # 核心原理：先匹配任何形式的字符串常量并消耗掉，避免字符串内部的 //, /* 等内容被误认为注释\\n pattern = re.compile(\\n r'(?P<string>\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'([^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`([^`\\\\\\\\]|\\\\\\\\.)*`)' # 匹配双引号、单引号和反引号内的字符串\\n r'|(?P<html_comment><!--[\\\\s\\\\S]*?-->)' # 匹配 HTML 注释\\n r'|(?P<multi_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)' # 匹配 JS/CSS 多行注释\\n r'|(?P<single_comment>//[^\\\\n]*)' # 匹配 JS/TS 单行注释\\n )\\n \\n # 根据用户指令中对HTML5 Canvas、JS以及ESLint等要素的描述，推断可能存在的代码文件类型\\n valid_extensions = ('.html', '.js', '.css', '.ts', '.jsx', '.tsx')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 在当前文件内容中查找所有模式\\n for match in pattern.finditer(content):\\n # 如果当前成功匹配的是某一种注释分组，则检查该组中的内容\\n if match.lastgroup in ('html_comment', 'multi_comment', 'single_comment'):\\n comment_text = match.group()\\n \\n # 遍历该注释内容中的每一个字符\\n for char in comment_text:\\n # 题目约定了可以直接使用提供的 is_emoji 函数\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据指令推测可能的代码文件后缀\\n code_extensions = {'.html', '.htm', '.js', '.css'}\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 判定是否为空行（仅包含空白字符或换行符的行 strip() 后都会变为空字符串）\\n if line.strip() == \\\"\\\":\\n return False\\n except Exception:\\n # 如果遇到文件编码或读取异常，说明存在生成问题或不符合UTF-8的要求，返回False\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码文件后缀类型\\n code_extensions = ['.js', '.html', '.css', '.ts', '.jsx', '.tsx']\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n # 1. 移除字符串字面量（避免字符串内部的 `?` 和 `:` 造成的统计干扰）\\n # 处理双引号、单引号和反引号，并在正则中匹配可能的转义字符\\n code = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '', code)\\n code = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", '', code)\\n code = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '', code, flags=re.DOTALL)\\n \\n # 2. 移除代码中的注释\\n # 移除 /* ... */ 块注释\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除 // ... 行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除 HTML <!-- ... --> 注释\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n \\n # 3. 排除正则表达式中非捕获组的 (?:\\n code = code.replace('(?:', '')\\n \\n # 4. 统计三元运算符数量\\n # 查找 `?`，并利用负向先行断言排除后面紧跟 `.` (可选链) 或 `?` (空值合并) 的情况\\n matches = re.findall(r'\\\\?(?!\\\\.|[?])', code)\\n \\n # 每个代码文件必须精确包含 2 个三元运算符\\n if len(matches) != 2:\\n return False\\n except Exception:\\n # 如果文件读取或解析抛出异常（如编码错误），视为未满足要求\\n return False\\n \\n # 如果没有找到任何代码文件，也是未满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def verify_file_line_count(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据多轮指令推断相关可能生成的代码和文件扩展名\\n allowed_extensions = {'.html', '.js', '.css', '.log', '.json'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的非AI直接生成的依赖库或代码控制相关目录\\n if 'node_modules' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 匹配推测的扩展名，或者特殊配置文件（如 .eslintrc 及其变体）\\n if ext in allowed_extensions or file.startswith('.eslintrc'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 迭代计算该文件内的行数\\n line_count = sum(1 for _ in f)\\n \\n # 检查行数是否大于200，超过则违反规定\\n if line_count > 200:\\n return False\\n except Exception:\\n # 如果由于读取非纯文本文件或编码问题导致读取异常，则跳过继续检查下一个文件\\n continue\\n \\n # 如果所有被检查的文件行数均 <= 200，则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 推断前端项目可能的代码文件类型\\n code_extensions = {'.html', '.htm', '.js', '.ts', '.css', '.jsx', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n line = line.strip()\\n if not line:\\n continue # 忽略完全空白的行\\n \\n # 剥离字符串内容，以避免将字符串内的 URL (`http://`) 或 注释符内容误判为真实注释\\n s_line = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', line)\\n s_line = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", s_line)\\n s_line = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', s_line)\\n \\n s_line_stripped = s_line.strip()\\n if not s_line_stripped:\\n continue\\n \\n # 确定对应文件支持的注释符号\\n syms = []\\n if ext in ['.js', '.ts', '.jsx', '.tsx']:\\n syms = ['//', '/*']\\n elif ext in ['.html', '.htm']:\\n # HTML 文件既能包含 HTML 注释，也可能包含内置脚本片段里的注释\\n syms = ['<!--', '//', '/*']\\n elif ext in ['.css']:\\n syms = ['/*']\\n \\n comment_idx = -1\\n first_sym = None\\n \\n # 寻找本行中最先出现的有效注释符号\\n for sym in syms:\\n idx = s_line.find(sym)\\n if idx != -1:\\n if comment_idx == -1 or idx < comment_idx:\\n comment_idx = idx\\n first_sym = sym\\n \\n if comment_idx == -1:\\n # 没有任何注释符，且此行非空，说明这是只有代码没有行尾注释的行\\n return False\\n \\n left_part = s_line[:comment_idx].strip()\\n if not left_part:\\n # 找到了注释符，但其左侧没有任何代码，属于独立行注释\\n return False\\n \\n # 确保多行/块级注释不会跨行（跨行的块级注释必然会在其他行构成独立注释行）\\n if first_sym == '/*':\\n if '*/' not in s_line[comment_idx + 2:]:\\n return False\\n elif first_sym == '<!--':\\n if '-->' not in s_line[comment_idx + 4:]:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码文件中不得出现while循环\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def remove_comments_and_strings(code: str) -> str:\\n # 正则表达式：匹配双引号、单引号、反引号字符串，以及块注释和行注释\\n pattern = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`|/\\\\*.*?\\\\*/|//[^\\\\n]*'\\n # 替换匹配到的内容为空，实现过滤\\n return re.sub(pattern, '', code, flags=re.DOTALL)\\n\\n def extract_script_content(html: str) -> str:\\n # 正则表达式：提取 <script> 标签内的内容\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', html, flags=re.IGNORECASE | re.DOTALL)\\n return \\\"\\\\n\\\".join(scripts)\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # 处理 JS / TS 类文件\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n cleaned_code = remove_comments_and_strings(content)\\n # 匹配 while 循环结构，即单词 while 后面跟着零或多个空格和左括号 (\\n if re.search(r'\\\\bwhile\\\\s*\\\\(', cleaned_code):\\n return False\\n except Exception:\\n # 忽略无法读取或编码有问题的文件\\n pass\\n \\n # 处理 HTML 类文件\\n elif file.endswith(('.html', '.htm')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n scripts = extract_script_content(content)\\n cleaned_code = remove_comments_and_strings(scripts)\\n if re.search(r'\\\\bwhile\\\\s*\\\\(', cleaned_code):\\n return False\\n except Exception:\\n pass\\n\\n # 若所有相关代码文件中均未发现 while 循环，则满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推断可能存在的代码文件类型\\n code_extensions = {'.html', '.js', '.css'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前工作空间的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 utf-8 对文件内容进行解码\\n # 因为包含中文注释，非 utf-8 编码必定会在此处抛出 UnicodeDecodeError\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 解码失败，说明存在非 UTF-8 编码的代码文件\\n return False\\n except Exception:\\n # 应对其他异常（如读取错误等）\\n return False\\n \\n # 如果找到了代码文件，并且全都没有抛出异常，则满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"最后，为了让视觉效果更现代、更炫酷，我们需要增加粒子爆炸特效。请编写一个粒子系统：当敌机被子弹击毁时，不要让它直接生硬地消失，而是要在敌机的中心坐标生成 10 到 15 个颜色在橙色和红色之间随机的小圆点粒子。这些粒子生成后，需要以随机的角度和速度向四面八方炸开，并且在飞散的过程中受到一点点『阻力』减速，同时逐渐缩小并变透明，直到生命周期结束完全消失。另外请你后续始终遵循以下调整和新增要求：原要求你提供的代码文件中不得出现while循环调整为在编写生成的代码文件时，请避免使用for循环；同时，请确保生成的代码文件中不包含多层if。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"生成代码文件的缩进量必须固定为4个空格\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.htm', '.js', '.css'}\\n has_code_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 去除行尾的换行符和空白字符，如果该行全空，则跳过\\n if not line.strip():\\n continue\\n \\n # 匹配行首的空白字符（仅匹配空格和制表符）\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_whitespace = match.group(1)\\n \\n # 检查1：不能包含制表符（Tab）\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 检查2：空格的数量必须是4的整数倍\\n if len(leading_whitespace) % 4 != 0:\\n return False\\n \\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所输出的代码文件里，任何一个函数的行数都必须介于15和30之间\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def mask_code(text: str) -> str:\\n \\\"\\\"\\\"\\n 将代码中的注释和字符串内容替换为空格，但保留换行符，\\n 从而在不改变任何代码行号的前提下，避免字符串和注释中的 '{' 和 '}' 干扰代码块解析。\\n \\\"\\\"\\\"\\n result = list(text)\\n i = 0\\n n = len(text)\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < n:\\n if in_line_comment:\\n if text[i] == '\\\\n':\\n in_line_comment = False\\n else:\\n result[i] = ' '\\n i += 1\\n elif in_block_comment:\\n if text[i:i+2] == '*/':\\n result[i] = ' '\\n if i + 1 < n:\\n result[i+1] = ' '\\n in_block_comment = False\\n i += 2\\n else:\\n if text[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n elif in_string:\\n if text[i] == '\\\\\\\\':\\n result[i] = ' '\\n if i + 1 < n:\\n result[i+1] = ' '\\n i += 2\\n elif text[i] == string_char:\\n result[i] = ' '\\n in_string = False\\n i += 1\\n else:\\n if text[i] != '\\\\n':\\n result[i] = ' '\\n i += 1\\n else:\\n if text[i:i+2] == '//':\\n in_line_comment = True\\n result[i] = ' '\\n if i + 1 < n:\\n result[i+1] = ' '\\n i += 2\\n elif text[i:i+2] == '/*':\\n in_block_comment = True\\n result[i] = ' '\\n if i + 1 < n:\\n result[i+1] = ' '\\n i += 2\\n elif text[i] in (\\\"'\\\", '\\\"', \\\"`\\\"):\\n in_string = True\\n string_char = text[i]\\n result[i] = ' '\\n i += 1\\n else:\\n i += 1\\n return \\\"\\\".join(result)\\n\\n def get_functions_line_counts(code: str, is_html: bool) -> list:\\n if is_html:\\n scripts = []\\n # 提取所有 <script> 标签中的内容\\n for match in re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', code, re.IGNORECASE | re.DOTALL):\\n scripts.append(match.group(1))\\n codes_to_check = scripts\\n else:\\n codes_to_check = [code]\\n \\n counts = []\\n for c in codes_to_check:\\n masked = mask_code(c)\\n \\n # 定义三种常见的JS函数入口特征正则表达式\\n # p1: function keyword (e.g. function name() { or function() { )\\n p1 = r'\\\\bfunction\\\\b[^{]*\\\\{'\\n # p2: arrow functions with braces (e.g. () => { or args => { )\\n p2 = r'(?:[a-zA-Z0-9_$]+|\\\\([^)]*\\\\))\\\\s*=>\\\\s*\\\\{'\\n # p3: class methods or object literal methods (e.g. methodName() { )\\n p3 = r'\\\\b(?!(?:if|while|for|switch|catch|return|with|function|new)\\\\b)[a-zA-Z0-9_$]+\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n \\n brace_indices = [i for i, char in enumerate(masked) if char == '{']\\n \\n for brace_idx in brace_indices:\\n # 截取大括号前的一段文本，判断是否以函数的特征结尾\\n prefix = masked[max(0, brace_idx - 300):brace_idx + 1]\\n \\n is_func = False\\n match_start_idx = -1\\n \\n m1 = re.search(p1 + r'$', prefix)\\n if m1:\\n is_func = True\\n match_start_idx = brace_idx - len(prefix) + m1.start()\\n else:\\n m2 = re.search(p2 + r'$', prefix)\\n if m2:\\n is_func = True\\n match_start_idx = brace_idx - len(prefix) + m2.start()\\n else:\\n m3 = re.search(p3 + r'$', prefix)\\n if m3:\\n is_func = True\\n match_start_idx = brace_idx - len(prefix) + m3.start()\\n \\n if is_func:\\n brace_count = 1\\n end_idx = brace_idx + 1\\n while end_idx < len(masked) and brace_count > 0:\\n if masked[end_idx] == '{':\\n brace_count += 1\\n elif masked[end_idx] == '}':\\n brace_count -= 1\\n end_idx += 1\\n \\n if brace_count == 0:\\n # 计算行数差值\\n start_line = masked[:match_start_idx].count('\\\\n')\\n end_line = masked[:end_idx].count('\\\\n')\\n counts.append(end_line - start_line + 1)\\n \\n # 单独匹配没有任何大括号 `{` 的单行箭头函数：例如 x => x * 2;\\n # 这种函数即使由于换行被拆成了两三行，也极难达到 15 行的下限，因此计入后将直接使验证失败\\n for m in re.finditer(r'(?<!=)=>', masked):\\n idx = m.end()\\n while idx < len(masked) and masked[idx].isspace():\\n idx += 1\\n if idx < len(masked) and masked[idx] != '{':\\n counts.append(1) # 简化处理，统一记录为1行，后续均会不达标\\n \\n return counts\\n\\n all_counts = []\\n files_checked = 0\\n \\n # 优先解析项目文件夹下生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for f in files:\\n if f.endswith('.js'):\\n with open(os.path.join(root, f), 'r', encoding='utf-8') as file:\\n code = file.read()\\n all_counts.extend(get_functions_line_counts(code, is_html=False))\\n files_checked += 1\\n elif f.endswith('.html'):\\n with open(os.path.join(root, f), 'r', encoding='utf-8') as file:\\n code = file.read()\\n all_counts.extend(get_functions_line_counts(code, is_html=True))\\n files_checked += 1\\n \\n # 如果项目文件夹内未找到，则尝试回退到解析 response 里提供的 Markdown 代码块\\n if files_checked == 0:\\n for match in re.finditer(r'```(?:javascript|js)\\\\s*(.*?)```', response, re.DOTALL | re.IGNORECASE):\\n all_counts.extend(get_functions_line_counts(match.group(1), is_html=False))\\n for match in re.finditer(r'```html\\\\s*(.*?)```', response, re.DOTALL | re.IGNORECASE):\\n all_counts.extend(get_functions_line_counts(match.group(1), is_html=True))\\n \\n # 如果没有找到任何函数，那么不能认为满足了要求（对于游戏项目来说这是不符合逻辑的）\\n if not all_counts:\\n return False\\n \\n # 检查所有的函数行数是否都符合 [15, 30] 闭区间的要求\\n for c in all_counts:\\n if c < 15 or c > 30:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"emoji表情符号 绝对不能出现在任何代码文件的注释内容中\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 构造正则表达式以匹配 字符串字面量 和 各种注释风格\\n # Group 1: 各种引号包裹的字符串（过滤掉字符串内的 // 等字符，避免误判）\\n # Group 2: 单行注释 (JS, TS, CSS 等)\\n # Group 3: 多行注释 (JS, TS, CSS 等)\\n # Group 4: HTML 注释\\n # Group 5: 单行注释 (Python, Shell 等)\\n pattern = re.compile(\\n r'(\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(//.*?$)|'\\n r'(/\\\\*[\\\\s\\\\S]*?\\\\*/)|'\\n r'(<!--[\\\\s\\\\S]*?-->)|'\\n r'(#.*?$)',\\n re.MULTILINE\\n )\\n\\n def has_emoji_in_comments(content: str) -> bool:\\n \\\"\\\"\\\"检查给定内容中的注释是否含有 emoji\\\"\\\"\\\"\\n for match in pattern.finditer(content):\\n # 获取注释的内容（如果在字符串内，则 group 2, 3, 4, 5 均为 None）\\n comment = match.group(2) or match.group(3) or match.group(4) or match.group(5)\\n if comment:\\n # 遍历注释中的每一个字符\\n for char in comment:\\n if is_emoji(char):\\n return True\\n return False\\n\\n valid_extensions = ('.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py', '.sh', '.json')\\n\\n # 1. 检查 workspace_path 下生成的代码文件中的注释\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若文件中注释含有 emoji，要求不被满足\\n if has_emoji_in_comments(content):\\n return False\\n except Exception:\\n continue\\n\\n # 2. 检查 response 文本中自带的代码块内的注释\\n code_blocks = re.findall(r'```(?:\\\\w*)\\\\s*\\\\n([\\\\s\\\\S]*?)```', response)\\n for block in code_blocks:\\n if has_emoji_in_comments(block):\\n return False\\n\\n # 没有发现任何违规情况，要求被满足\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"不要让你生成的任何代码文件里存在空行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据上下文推断可能生成的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx'}\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n def check_no_empty_lines_in_file(file_path: str) -> bool:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去除行首尾的空白字符（包含空格、制表符、回车换行等）\\n # 如果结果是空字符串，意味着当前行是空行\\n if line.strip() == '':\\n return False\\n except Exception:\\n # 若文件读取失败（例如编码不支持），则跳过\\n pass\\n return True\\n\\n # 如果传入的路径本身是文件\\n if os.path.isfile(workspace_path):\\n ext = os.path.splitext(workspace_path)[1].lower()\\n if ext in code_extensions:\\n return check_no_empty_lines_in_file(workspace_path)\\n return True\\n\\n # 如果传入的是目录，则递归遍历其下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n if not check_no_empty_lines_in_file(file_path):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"每一个代码文件的生成标准是：必须精确包含2个三元运算符(? :)\",\n \"validation_code\": \"def check_ternary_operator_in_files(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的代码文件类型\\n valid_extensions = {'.js', '.html', '.css', '.ts', '.jsx', '.tsx'}\\n files_checked = 0\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 1. 剔除字符串内容 (双引号、单引号、反引号)，避免匹配到字符串里的符号\\n # 处理转义字符，避免匹配提前结束\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', content)\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', content)\\n content = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '', content)\\n \\n # 2. 剔除注释内容\\n # 多行注释\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n # HTML 注释\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n # 单行注释 (在去除了字符串之后进行，避免误删 URL 中的 //)\\n content = re.sub(r'//.*', '', content)\\n \\n # 3. 统计三元运算符中的问号数量\\n # 排除可选链 `?.` 和空值合并 `??`\\n ternary_count = len(re.findall(r'\\\\?(?![?.])', content))\\n \\n # 如果任意一个代码文件不满足精确包含2个三元运算符的要求，直接返回 False\\n if ternary_count != 2:\\n return False\\n \\n files_checked += 1\\n \\n # 如果没有扫描到任何有效的代码文件，返回 False\\n if files_checked == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"你创建的每一个文件，其行数都应保持在200行或以下\",\n \"validation_code\": \"def check_file_length_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区不存在，无法判断是否有超行数的文件，直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令推断的可能代码及日志文件类型\\n valid_extensions = {'.html', '.js', '.css', '.log'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉非生成环境所产生的目录，防止干扰\\n for exclude_dir in ['node_modules', '.git']:\\n if exclude_dir in dirs:\\n dirs.remove(exclude_dir)\\n\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n # 若文件的行数大于200，则不满足要求\\n if len(lines) > 200:\\n return False\\n except Exception:\\n # 对于不可读或编码不符的文件跳过检查\\n pass\\n\\n # 遍历结束无超过200行的文件则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"请保证生成的代码文件中，每一行代码都配有行尾注释（即写在该行代码的右侧），不能采用位于代码上方或下方的独立行注释\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n valid_exts = ['.js', '.html', '.css', '.ts']\\n code_file_count = 0\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n\\n filepath = os.path.join(root, file)\\n code_file_count += 1\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n\\n for line in lines:\\n s_line = line.strip()\\n if not s_line:\\n continue\\n\\n # 检查是否为独立行注释（不允许存在位于代码上方或下方的独立行注释）\\n if s_line.startswith('//') or s_line.startswith('/*') or s_line.startswith('<!--'):\\n return False\\n\\n # 检查该行代码是否包含行尾注释\\n has_comment = False\\n if ext in ['.js', '.ts', '.html']:\\n if '/*' in s_line:\\n has_comment = True\\n elif '<!--' in s_line and ext == '.html':\\n has_comment = True\\n else:\\n # 查找 // 注释，避开类似 http:// 的非注释情况\\n idx = s_line.find('//')\\n while idx != -1:\\n if idx > 0 and s_line[idx-1] == ':':\\n idx = s_line.find('//', idx + 2)\\n else:\\n has_comment = True\\n break\\n elif ext == '.css':\\n if '/*' in s_line:\\n has_comment = True\\n\\n # 如果这行代码没有找到任何注释标识符，代表缺失了行尾注释\\n if not has_comment:\\n return False\\n\\n # 至少需要检查到一个代码文件且均满足要求时才认为任务成功\\n return code_file_count > 0\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在编写生成的代码文件时，请避免使用for循环\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def remove_comments_and_strings(code: str) -> str:\\n # 正则表达式用于匹配并提取可能的干扰项：\\n # 1. /* ... */ 多行注释\\n # 2. // ... 单行注释\\n # 3. \\\"...\\\" 双引号字符串\\n # 4. '...' 单引号字符串\\n # 5. `...` 反引号字符串\\n pattern = r'(/\\\\*.*?\\\\*/)|(//[^\\\\n]*)|(\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\")|(\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\')|(`(?:\\\\\\\\.|[^\\\\\\\\`])*`)'\\n \\n # 将上述匹配到的注释或字符串内容替换为空格，以防止其中的内容被错误地识别为代码关键字\\n return re.sub(pattern, ' ', code, flags=re.DOTALL)\\n\\n # 如果工作区路径不存在，直接返回 True（可视为未违反要求）\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 针对飞机大战游戏，主要检查 html 与 js 后缀的源码文件\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n js_content = \\\"\\\"\\n if file.endswith('.html'):\\n # 仅提取 HTML 文件中 <script> 标签内的 JS 代码\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, flags=re.DOTALL | re.IGNORECASE)\\n js_content = '\\\\n'.join(scripts)\\n else:\\n js_content = content\\n \\n # 去除注释和字符串后的代码片段\\n clean_js = remove_comments_and_strings(js_content)\\n \\n # 检查清理后的代码中是否出现 for 循环，包括 for(...) 和 for await(...)\\n # \\\\bfor\\\\b : 确保匹配的是完整关键字 for\\n # \\\\s* : 匹配可能存在的空格\\n # (?:await\\\\s*)? : 兼容 JavaScript 中的 for await 语法\\n # \\\\( : 匹配语法左括号\\n if re.search(r'\\\\bfor\\\\s*(?:await\\\\s*)?\\\\(', clean_js):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"对于项目的每一次改动，均要求新建立一个对应的 change_log_X.log 文件进行记录，此处的 X 从 0 开始，代表改动的次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你输出的代码必须接受 ESLint 的检查，并且需要修复所有检测出的报错，确保最终 ESLint 能够零错误通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 针对前端和常见JS项目，定义相关的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.js', '.css', '.ts', '.json'}\\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 以二进制模式读取整个文件\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试将其按照 utf-8 进行解码\\n # 如果不是 UTF-8 编码或 ASCII 编码（比如 GBK 等含中文的内容），此处将会抛出异常\\n raw_data.decode('utf-8')\\n \\n except UnicodeDecodeError:\\n # 只要有一个代码文件不是 utf-8 编码，就不满足要求\\n return False\\n except Exception:\\n # 其他读取相关的异常安全起见也视为验证失败\\n return False\\n \\n # 必须至少包含一个代码文件，并且所有代码文件均符合 UTF-8 编码，才算达成目标\\n return code_files_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"执行 bash 下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 60s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件中不包含多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["打开页面，确认页面中包含一个水平居中的黑色背景区域作为游戏画布，且尺寸大约为 400x600 像素。","观察画布底部中央，确认显示有一架代表玩家的飞机，它由简单的几何图形（如三角形、矩形）组合而成。","按下键盘的左方向键（或 A 键），确认玩家飞机能向左平移；按下右方向键（或 D 键），确认玩家飞机能向右平移。","长按右方向键将飞机移动至画布最右侧，长按左方向键移动至最左侧，确认飞机到达边缘后停止平移，不会飞出画布可视范围之外。","按下空格键，确认玩家飞机的机头位置生成一个黄色的细长矩形作为子弹，并且子弹向上方垂直飞行。","按住空格键不松开或极速连续按空格键，观察子弹的发射密度，确认子弹在垂直方向上保持明显的距离间隔，没有因发射过快而连成一条连续的实线，当子弹飞出画布上边缘时，应将它从代码的数组里销毁掉。","等待片刻观察画布，确认画布顶部能在随机的横坐标位置不断生成红色的方块表示敌机，并且敌机生成后持续向下方垂直移动，当敌机飞出画布下边缘时，也要从内存中清理。","当屏幕上同时出现多架敌机时截屏，对比不同的敌机，确认不同敌机的红色方块尺寸大小有轻微差异，并且下落移动的速度不完全相同。","发射子弹击中一架红色敌机，确认子弹碰到敌机的瞬间两者同时从画布上销毁消失，且画布左上角的白色得分面板上的分数有增加。","子弹击毁敌机时，观察被击毁的位置，确认短暂出现了一个『+10』的文字特效，且该文字会缓缓向上方飘移，在飘移过程中透明度降低直至消失。","游戏刚开始时，观察画布右上角区域，确认用文字或小红心图标正确显示了玩家的初始生命值为 3 点。","停止射击，任由一架敌机飞离出屏幕的下边界，确认此时右上角的生命值发生变动，扣减了 1 点。","移动玩家飞机故意触碰一架下落的红色敌机，确认发生接触碰撞时，右上角的生命值扣减了 1 点。","在发生敌机漏掉飞出下边界或者敌机撞击到玩家飞机的瞬间，观察游戏画面，确认整个画布背景会出现短暂闪烁一下红色的视觉特效作为受伤警告。","持续漏过或撞击敌机直到生命值降为 0，确认画布上所有的子弹和敌机停止移动和生成，画布中央显示大号的『Game Over』文字，并在下方显示一个『重新开始』的按钮图形。","在 Game Over 状态下，使用鼠标点击『重新开始』按钮区域，确认画布上遗留的敌机和子弹被清空，生命值恢复为 3，分数归 0，游戏能够重新正常开始游玩。","正常进行游戏直至左上角得分达到 100 分，确认在刚好达到 100 分的整数倍时，屏幕中央出现闪烁的『Level Up!』提示，之后观察确认敌机的下落速度或生成频率比游戏刚开始时更快。","在游戏进行中等待约 15 秒钟，观察画布顶部，确认从上方掉落了一个蓝色的圆形实体作为『火力补给包』。","移动玩家飞机接住蓝色圆形的火力补给包，随后按下空格键射击，确认玩家飞机的射击模式发生了改变，每次开火会在机翼的两侧平行发射出两发子弹。","接住补给包并进入双发模式后，继续游戏并等待超过 5 秒，再次按下空格射击，确认玩家飞机的射击模式自动恢复为机头中心的单发子弹射击。","射击并击毁一架红色敌机，观察敌机原本的中心位置，确认瞬间生成了约 10 到 15 个颜色介于橙色和红色之间的小圆点粒子。","击毁敌机后在不到1秒内连续观察爆炸产生的圆点粒子，确认粒子生成后以随机角度向四面八方飞散炸开，且在飞行途中速度减弱、体积逐渐缩小、颜色逐渐透明直到最终消失。"],"string":"[\n \"打开页面，确认页面中包含一个水平居中的黑色背景区域作为游戏画布，且尺寸大约为 400x600 像素。\",\n \"观察画布底部中央，确认显示有一架代表玩家的飞机，它由简单的几何图形（如三角形、矩形）组合而成。\",\n \"按下键盘的左方向键（或 A 键），确认玩家飞机能向左平移；按下右方向键（或 D 键），确认玩家飞机能向右平移。\",\n \"长按右方向键将飞机移动至画布最右侧，长按左方向键移动至最左侧，确认飞机到达边缘后停止平移，不会飞出画布可视范围之外。\",\n \"按下空格键，确认玩家飞机的机头位置生成一个黄色的细长矩形作为子弹，并且子弹向上方垂直飞行。\",\n \"按住空格键不松开或极速连续按空格键，观察子弹的发射密度，确认子弹在垂直方向上保持明显的距离间隔，没有因发射过快而连成一条连续的实线，当子弹飞出画布上边缘时，应将它从代码的数组里销毁掉。\",\n \"等待片刻观察画布，确认画布顶部能在随机的横坐标位置不断生成红色的方块表示敌机，并且敌机生成后持续向下方垂直移动，当敌机飞出画布下边缘时，也要从内存中清理。\",\n \"当屏幕上同时出现多架敌机时截屏，对比不同的敌机，确认不同敌机的红色方块尺寸大小有轻微差异，并且下落移动的速度不完全相同。\",\n \"发射子弹击中一架红色敌机，确认子弹碰到敌机的瞬间两者同时从画布上销毁消失，且画布左上角的白色得分面板上的分数有增加。\",\n \"子弹击毁敌机时，观察被击毁的位置，确认短暂出现了一个『+10』的文字特效，且该文字会缓缓向上方飘移，在飘移过程中透明度降低直至消失。\",\n \"游戏刚开始时，观察画布右上角区域，确认用文字或小红心图标正确显示了玩家的初始生命值为 3 点。\",\n \"停止射击，任由一架敌机飞离出屏幕的下边界，确认此时右上角的生命值发生变动，扣减了 1 点。\",\n \"移动玩家飞机故意触碰一架下落的红色敌机，确认发生接触碰撞时，右上角的生命值扣减了 1 点。\",\n \"在发生敌机漏掉飞出下边界或者敌机撞击到玩家飞机的瞬间，观察游戏画面，确认整个画布背景会出现短暂闪烁一下红色的视觉特效作为受伤警告。\",\n \"持续漏过或撞击敌机直到生命值降为 0，确认画布上所有的子弹和敌机停止移动和生成，画布中央显示大号的『Game Over』文字，并在下方显示一个『重新开始』的按钮图形。\",\n \"在 Game Over 状态下，使用鼠标点击『重新开始』按钮区域，确认画布上遗留的敌机和子弹被清空，生命值恢复为 3，分数归 0，游戏能够重新正常开始游玩。\",\n \"正常进行游戏直至左上角得分达到 100 分，确认在刚好达到 100 分的整数倍时，屏幕中央出现闪烁的『Level Up!』提示，之后观察确认敌机的下落速度或生成频率比游戏刚开始时更快。\",\n \"在游戏进行中等待约 15 秒钟，观察画布顶部，确认从上方掉落了一个蓝色的圆形实体作为『火力补给包』。\",\n \"移动玩家飞机接住蓝色圆形的火力补给包，随后按下空格键射击，确认玩家飞机的射击模式发生了改变，每次开火会在机翼的两侧平行发射出两发子弹。\",\n \"接住补给包并进入双发模式后，继续游戏并等待超过 5 秒，再次按下空格射击，确认玩家飞机的射击模式自动恢复为机头中心的单发子弹射击。\",\n \"射击并击毁一架红色敌机，观察敌机原本的中心位置，确认瞬间生成了约 10 到 15 个颜色介于橙色和红色之间的小圆点粒子。\",\n \"击毁敌机后在不到1秒内连续观察爆炸产生的圆点粒子，确认粒子生成后以随机角度向四面八方飞散炸开，且在飞行途中速度减弱、体积逐渐缩小、颜色逐渐透明直到最终消失。\"\n]","isConversational":false}}},{"rowIdx":81,"cells":{"task_id":{"kind":"number","value":82,"string":"82"},"repository_policy":{"kind":"string","value":"你在后续所有与用户的交互轮次中，必须始终严格遵循以下所有约束：\n1. 在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方。\n2. 进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复。\n3. 你的每一条回复都必须使用语气词\"嗨～\"来打头。\n4. 你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"。\n5. 对于你创建的所有代码文件，规定必须以UTF-8进行编码。\n6. 生成的代码文件必须满足：单个函数的规模不得大于 80 行。\n7. 每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数。\n8. 在你输出的任何一个代码文件里，务必出现 \"TODO\"。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含不被允许的称呼“你”\n if '你' in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n # 检查回复中是否出现了“我”字\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否以\"嗨～\"打头。\n \"\"\"\n if not response:\n return False\n \n # 去除回复内容中可能存在的前导空白字符，并判断是否以\"嗨～\"开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否在最末尾加上了语气词\"哦～\"\n \"\"\"\n if not response:\n return False\n \n # 去除回复末尾可能存在的空白字符（如换行符、空格）\n clean_response = response.rstrip()\n \n # 检查是否以\"哦～\"结尾\n return clean_response.endswith(\"哦～\")","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts', \n '.html', '.css', '.go', '.rs', '.php', '.rb', '.swift', '.kt', \n '.sh', '.bash', '.sql', '.json', '.xml', '.yaml', '.yml'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对代码文件进行校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以二进制读取并使用 utf-8 解码\n with open(file_path, 'rb') as f:\n content = f.read()\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果解码失败，说明该文件不是 UTF-8 编码\n return False\n except Exception:\n # 出现其他无法读取的异常也视为不满足要求\n return False\n\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import ast\n\n # 若工作区路径不存在，说明未生成代码，自然没有函数超过80行\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n\n # 遍历代码项目的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 此处推断代码文件主要为Python文件，仅针对Python类型进行精确检查\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n source_code = f.read()\n \n # 解析代码抽象语法树\n tree = ast.parse(source_code)\n \n # 遍历语法树的各个节点寻找函数定义\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start_lineno = node.lineno\n \n # 获取函数的结束行号\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n end_lineno = node.end_lineno\n else:\n # 兼容早期版本 Python（如 3.7 及更早），通过子节点推断最大行号\n end_lineno = start_lineno\n for child in ast.walk(node):\n if hasattr(child, \"lineno\"):\n end_lineno = max(end_lineno, child.lineno)\n \n # 计算函数包含的总行数\n function_length = end_lineno - start_lineno + 1\n \n # 若函数的行数大于 80，则判定为不满足要求\n if function_length > 80:\n return False\n except Exception:\n # 对于存在语法错误导致无法解析的文件直接跳过\n continue\n\n # 所有文件验证完毕，均未发现超过80行的函数\n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个数字是否为素数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 常见的代码文件扩展名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs',\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', '.rb', '.php',\n '.sh', '.swift', '.kt', '.rs'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历项目目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 使用 '\\n' 分割可以准确计算行数，并且包含所有空行\n lines = f.read().split('\\n')\n line_count = len(lines)\n \n # 判断总行数是否为素数\n if is_prime(line_count):\n return False\n except Exception:\n # 如果由于编码或其他原因无法读取，直接忽略跳过\n continue\n \n # 所有代码文件的行数均不为素数，满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件的扩展名集合\n code_extensions = {\n '.py', '.java', '.cpp', '.c', '.cc', '.h', '.hpp', '.cs', \n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.go', \n '.rs', '.php', '.rb', '.swift', '.sh', '.bash', '.kt', '.scala'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n if \"TODO\" not in content:\n return False\n except Exception:\n # 若文件读取出现异常，保守判定未满足\n return False\n\n return True","tags":["Content","Keyword"]}],"string":"[\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含不被允许的称呼“你”\\n if '你' in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否出现了“我”字\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否以\\\"嗨～\\\"打头。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复内容中可能存在的前导空白字符，并判断是否以\\\"嗨～\\\"开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否在最末尾加上了语气词\\\"哦～\\\"\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除回复末尾可能存在的空白字符（如换行符、空格）\\n clean_response = response.rstrip()\\n \\n # 检查是否以\\\"哦～\\\"结尾\\n return clean_response.endswith(\\\"哦～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts', \\n '.html', '.css', '.go', '.rs', '.php', '.rb', '.swift', '.kt', \\n '.sh', '.bash', '.sql', '.json', '.xml', '.yaml', '.yml'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对代码文件进行校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以二进制读取并使用 utf-8 解码\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果解码失败，说明该文件不是 UTF-8 编码\\n return False\\n except Exception:\\n # 出现其他无法读取的异常也视为不满足要求\\n return False\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n\\n # 若工作区路径不存在，说明未生成代码，自然没有函数超过80行\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历代码项目的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 此处推断代码文件主要为Python文件，仅针对Python类型进行精确检查\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n source_code = f.read()\\n \\n # 解析代码抽象语法树\\n tree = ast.parse(source_code)\\n \\n # 遍历语法树的各个节点寻找函数定义\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start_lineno = node.lineno\\n \\n # 获取函数的结束行号\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n end_lineno = node.end_lineno\\n else:\\n # 兼容早期版本 Python（如 3.7 及更早），通过子节点推断最大行号\\n end_lineno = start_lineno\\n for child in ast.walk(node):\\n if hasattr(child, \\\"lineno\\\"):\\n end_lineno = max(end_lineno, child.lineno)\\n \\n # 计算函数包含的总行数\\n function_length = end_lineno - start_lineno + 1\\n \\n # 若函数的行数大于 80，则判定为不满足要求\\n if function_length > 80:\\n return False\\n except Exception:\\n # 对于存在语法错误导致无法解析的文件直接跳过\\n continue\\n\\n # 所有文件验证完毕，均未发现超过80行的函数\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个数字是否为素数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 常见的代码文件扩展名\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs',\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.go', '.rb', '.php',\\n '.sh', '.swift', '.kt', '.rs'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历项目目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 使用 '\\\\n' 分割可以准确计算行数，并且包含所有空行\\n lines = f.read().split('\\\\n')\\n line_count = len(lines)\\n \\n # 判断总行数是否为素数\\n if is_prime(line_count):\\n return False\\n except Exception:\\n # 如果由于编码或其他原因无法读取，直接忽略跳过\\n continue\\n \\n # 所有代码文件的行数均不为素数，满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件的扩展名集合\\n code_extensions = {\\n '.py', '.java', '.cpp', '.c', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue', '.go', \\n '.rs', '.php', '.rb', '.swift', '.sh', '.bash', '.kt', '.scala'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 若文件读取出现异常，保守判定未满足\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个网页版的俄罗斯方块。第一步，先把页面的基本骨架搭出来。整个页面需要是纯黑背景。在页面中央放置一个游戏主网格区域（可以用 Canvas 也可以用 DOM），旁边预留一个信息面板区域。在页面底部加一个大号的“开始游戏”按钮。现在先不用实现任何方块的生成和下落逻辑，只要把暗黑风格的背景、空空的网格线画出来，让页面初步呈现出炫酷的视觉雏形即可。请注意后续所有轮次都需要遵循以下要求：要求输出代码时，统一使用2个空格作为缩进标准；你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径。","instruction_following_checklist":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复中是否使用了“阁下”代替“你”来称呼对方。\n 为了确保没有使用“你”，直接检查回复内容中是否不包含字符“你”。\n \"\"\"\n return \"你\" not in response","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复中是否出现了\"我\"字\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否以语气词\"嗨～\"打头。\n \"\"\"\n # 剔除可能存在的前导空白字符后，判断是否以“嗨～”开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容末尾可能存在的空白字符、换行符以及常见的中英文标点符号\n cleaned_response = response.rstrip(' \\t\\n\\r.!?。！？\"\\']}”’')\n \n # 判断清理后的字符串是否以\"哦～\"结尾\n return cleaned_response.endswith(\"哦～\")","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件扩展名（针对网页开发等情况）\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.xml', '.md', '.txt', '.py'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试将文件字节数据按 utf-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该代码文件不是合法的 UTF-8 编码\n return False\n except Exception:\n return False\n\n # 若未找到任何代码文件，则无法证明其满足了要求；若找到了且全都通过了验证，则返回 True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def is_function_block_valid(code_text: str) -> bool:\n # 将注释内容替换为空白但保留换行符，以免破坏行号\n code_text = re.sub(r'/\\*.*?\\*/', lambda m: '\\n' * m.group(0).count('\\n'), code_text, flags=re.DOTALL)\n code_text = re.sub(r'//.*', '', code_text)\n \n # 将模板字符串内部的花括号替换为方括号，避免干扰解析流程\n code_text = re.sub(r'`.*?`', lambda m: m.group(0).replace('{', '[').replace('}', ']'), code_text, flags=re.DOTALL)\n \n # 将一般的字符串内容简化掉\n code_text = re.sub(r'\"(?:[^\"\\\\]|\\\\.)*\"', '\"\"', code_text)\n code_text = re.sub(r\"'(?:[^'\\\\]|\\\\.)*'\", \"''\", code_text)\n\n lines = code_text.split('\\n')\n stack = []\n \n for i, line in enumerate(lines):\n # 扫描大括号\n for match in re.finditer(r'[{}]', line):\n char = match.group()\n if char == '{':\n prefix = line[:match.start()].strip()\n j = i - 1\n \n # 向上寻找非空代码行以判断大括号归属\n while not prefix and j >= 0:\n prefix = lines[j].strip()\n j -= 1\n \n is_func = False\n # 识别是否为函数定义\n if 'function' in prefix or '=>' in prefix:\n is_func = True\n else:\n # 识别类似 name(args) 的方法声明形式\n if re.search(r'\\b[A-Za-z_][A-Za-z0-9_]*\\s*\\([^)]*\\)$', prefix):\n # 排除常见的控制流关键字\n if not re.search(r'\\b(if|for|while|switch|catch|with)\\b', prefix):\n is_func = True\n \n stack.append((is_func, i))\n elif char == '}':\n if stack:\n is_func, start_line = stack.pop()\n if is_func:\n func_length = i - start_line + 1\n # 若函数的总行数超过 80，则判定为不符合要求\n if func_length > 80:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历所有代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 检测 JS 或 TS 脚本代码\n if ext in ['.js', '.ts']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not is_function_block_valid(code):\n return False\n except Exception:\n pass\n # 提取 HTML 文件中的 script 片段进行检测\n elif ext in ['.html', '.htm']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n html = f.read()\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', html, flags=re.IGNORECASE | re.DOTALL)\n for script in scripts:\n if not is_function_block_valid(script):\n return False\n except Exception:\n pass\n\n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n \"\"\"判断一个非负整数是否为素数\"\"\"\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 网页版俄罗斯方块涉及的可能代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n \n # 统计包含空行在内的总行数\n if not text:\n num_lines = 0\n else:\n num_lines = len(text.split('\\n'))\n \n # 如果有任何一个代码文件的行数为素数，则未满足要求\n if is_prime(num_lines):\n return False\n except Exception:\n # 忽略可能出现的文件读取异常（如编码格式不支持的情况等）\n continue\n\n # 所有代码文件的行数均不为素数，则满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对网页版俄罗斯方块项目推断可能的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_files = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_files = True\n file_path = os.path.join(root, file_name)\n \n try:\n # 依据系统提示要求，统一使用 utf-8 编码进行读取\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果读取失败（如编码不符合utf-8等原因），视为验证失败\n return False\n \n # 检查文件内容中是否包含 \"TODO\"\n if \"TODO\" not in content:\n return False\n \n # 确保至少检查了一个代码文件，并且所有代码文件都包含 \"TODO\"\n return has_code_files","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用2个空格作为缩进标准","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n code_contents = []\n \n # 1. 提取 workspace 中生成的前端代码文件内容\n if os.path.isdir(workspace_path):\n valid_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_contents.append(f.read())\n except Exception:\n pass\n \n # 2. 提取 response 中的前端/无标记 Markdown 代码块\n code_blocks = re.findall(r\"```([^\\n]*)\\n(.*?)```\", response, re.DOTALL)\n valid_langs = {'html', 'css', 'js', 'javascript', 'ts', 'typescript', 'vue', 'jsx', 'tsx', ''}\n \n for header, block in code_blocks:\n # 提取语言标识\n lang = header.split()[0].lower() if header.strip() else ''\n if lang in valid_langs:\n if block.endswith('\\n'):\n block = block[:-1]\n code_contents.append(block)\n \n # 如果既没有找到文件代码，也没有找到代码块，直接返回 False\n if not code_contents:\n return False\n \n all_indentations = set()\n \n # 3. 逐行分析缩进情况\n for code in code_contents:\n lines = code.split('\\n')\n for line in lines:\n stripped = line.lstrip()\n # 跳过空行\n if not stripped:\n continue\n \n # 跳过注释行 (主要防范 JSDoc 风格或者多行 CSS 注释导致的奇数缩进对齐)\n if stripped.startswith('*'):\n continue\n \n # 获取前导空白字符\n leading_whitespace = line[:len(line) - len(stripped)]\n \n # 若包含制表符(\\t)或其他非正常空格的前导字符，不符合统一空格的要求\n if leading_whitespace.replace(' ', '') != '':\n return False\n \n space_count = len(leading_whitespace)\n \n # 缩进不应为奇数\n if space_count % 2 != 0:\n return False\n \n if space_count > 0:\n all_indentations.add(space_count)\n \n # 如果代码完全没有任何层级的缩进（全是0缩进），无法判断为满足标准\n if not all_indentations:\n return False\n \n # 如果所有的非零缩进都是 4 的倍数，那么很可能该代码使用的是 4 个空格作为缩进标准，予以排除\n if all(x % 4 == 0 for x in all_indentations):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":1,"instruction":"不错，界面骨架有了。接下来我们引入方块并让它动起来。请定义标准的7种俄罗斯方块形状（I, J, L, O, S, T, Z），为了契合炫酷风格，请给每种形状赋予不同的高饱和度荧光色（比如霓虹粉、青色、亮黄等）。点击“开始游戏”按钮后，在网格顶部中心随机生成一个方块，并让它按照固定的时间间隔自动匀速下落。暂时不需要完整的左右移动和旋转，只要能碰到网格最底部停住不掉出去就行。另外请你注意，从本轮开始后续所有轮次要求输出代码时，统一使用4个空格作为缩进标准；同时后续所有开发环节都需要配套编写单元测试，你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率，以上两项要求后续所有轮次均需遵循。","instruction_following_checklist":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n # 如果回复中出现了“你”字，说明没有始终使用“阁下”进行代替，违反了要求\n if \"你\" in response:\n return False\n \n # 若没有出现“你”，则认为满足了该项约束\n return True","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复中是否没有使用\"我\"来进行自我指代。\n 这里通过判断回复内容中是否包含\"我\"字来进行严格验证。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除字符串左侧可能的空白字符，并判断是否以指定的语气词开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n # 清理字符串末尾的空白字符以及常见的中英文标点符号\n # 注意：不要将\"～\"或\"~\"加入清理列表，以免破坏目标语气词\n cleaned_response = re.sub(r'[\\s。！？，、；：\\.,!;:]+$', '', response)\n \n # 检查清理后的回复字符串是否以要求的语气词结尾\n return cleaned_response.endswith(\"哦～\")","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断可能的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.mjs', '.cjs', '.ts', '.jsx', '.tsx', '.py', '.json'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码打开并读取文件所有内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果发生 UnicodeDecodeError 异常，则说明存在不是以 UTF-8 编码的文件\n return False\n except Exception:\n # 其他任何导致文件无法正常读取的异常视作验证失败\n return False\n\n # 若至少存在一个代码文件，且全部能够通过 UTF-8 解码，则返回 True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n def get_py_functions_line_counts(code: str) -> list:\n try:\n tree = ast.parse(code)\n except Exception:\n return []\n counts = []\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n counts.append(node.end_lineno - node.lineno + 1)\n return counts\n\n def get_js_functions_line_counts(code: str) -> list:\n # 去除所有可能干扰分析的注释和字符串，同时用等量换行符占位以保证行号完全对应\n pattern = r'(//[^\\n]*)|(/\\*[\\s\\S]*?\\*/)|(\\'(?:\\\\[\\s\\S]|[^\\\\\\'])*\\')|(\"(?:\\\\[\\s\\S]|[^\\\\\"])*\")|(`(?:\\\\[\\s\\S]|[^\\\\`])*`)'\n def replacer(m):\n text = m.group(0)\n return ' ' + '\\n' * text.count('\\n')\n \n try:\n skeleton = re.sub(pattern, replacer, code)\n except Exception:\n return []\n \n lines = skeleton.split('\\n')\n # 扫描出关键标记（包含标识符、箭头函数符号、括号等）\n token_pattern = re.compile(r'[a-zA-Z_$][a-zA-Z0-9_$]*|=>|[{}]|[();]')\n tokens = []\n for i, line in enumerate(lines):\n for m in token_pattern.finditer(line):\n tokens.append((m.group(), i + 1))\n \n func_brace_indices = {}\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\n \n for i, (tok, line_no) in enumerate(tokens):\n if tok == '{':\n is_func = False\n func_start_line = line_no\n \n # 规则1: 往前找是否有 function 关键字\n for j in range(i-1, max(-1, i-10), -1):\n if tokens[j][0] == 'function':\n is_func = True\n func_start_line = tokens[j][1]\n break\n elif tokens[j][0] in [';', '{', '}']:\n break\n \n if not is_func:\n # 规则2: 往前找是否有箭头函数标识符 =>\n for j in range(i-1, max(-1, i-5), -1):\n if tokens[j][0] == '=>':\n is_func = True\n func_start_line = tokens[j][1]\n # 尽量向前延伸到括号处作为起点\n for k in range(j-1, max(-1, j-4), -1):\n if tokens[k][0] == '(':\n func_start_line = tokens[k][1]\n break\n break\n elif tokens[j][0] in [';', '{', '}']:\n break\n \n if not is_func:\n # 规则3: 判断是否属于 method(args) { ... } 的形式（类方法或对象函数定义）\n if i >= 4:\n if tokens[i-1][0] == ')':\n paren_depth = 1\n name_idx = -1\n for j in range(i-2, -1, -1):\n if tokens[j][0] == ')':\n paren_depth += 1\n elif tokens[j][0] == '(':\n paren_depth -= 1\n if paren_depth == 0:\n name_idx = j - 1\n break\n if name_idx >= 0:\n name_tok = tokens[name_idx][0]\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', name_tok) and name_tok not in keywords:\n is_func = True\n func_start_line = tokens[name_idx][1]\n \n if is_func:\n func_brace_indices[i] = func_start_line\n \n # 使用栈匹配闭合大括号，并计算已被标记为函数的块规模\n stack = []\n func_lines_counts = []\n for i, (tok, line_no) in enumerate(tokens):\n if tok == '{':\n if i in func_brace_indices:\n stack.append((func_brace_indices[i], True))\n else:\n stack.append((line_no, False))\n elif tok == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n func_lines_counts.append(line_no - start_line + 1)\n \n return func_lines_counts\n\n def get_html_script_line_counts(code: str) -> list:\n counts = []\n # 提取HTML文件内的独立脚本部分做逻辑检测\n scripts = re.finditer(r'<script\\b[^>]*>([\\s\\S]*?)</script>', code, re.IGNORECASE)\n for m in scripts:\n script_code = m.group(1)\n counts.extend(get_js_functions_line_counts(script_code))\n return counts\n\n # 检测开始\n if not os.path.isdir(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 只针对可能的代码载体文件进行分析\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n counts = []\n if ext == '.py':\n counts = get_py_functions_line_counts(code)\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n counts = get_js_functions_line_counts(code)\n elif ext in ['.html', '.htm']:\n counts = get_html_script_line_counts(code)\n \n for c in counts:\n if c > 80:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一个整数是否为素数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 网页项目常见的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 读取文件时忽略部分解码错误，防止直接崩溃影响验证逻辑\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 包含空行的总行数\n lines = content.split('\\n')\n line_count = len(lines)\n \n # 如果存在文件的总行数是素数，则不满足要求\n if is_prime(line_count):\n return False\n except Exception:\n continue\n \n # 所有代码文件的行数均不是素数，满足要求\n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n # 遍历工作区中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 检查 \"TODO\" 是否出现在代码文件中\n if \"TODO\" not in content:\n return False\n except Exception:\n # 如果由于某种原因文件无法正常读取，出于严格校验目的，返回False\n return False\n \n # 如果有代码文件并且所有代码文件都通过了检查，返回True，否则返回False\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"def check_indentation_4_spaces(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_fragments = []\n \n # 1. 从 response 中提取所有的 Markdown 代码块\n pattern = re.compile(r'```[^\\n]*\\n(.*?)```', re.DOTALL)\n code_fragments.extend(pattern.findall(response))\n \n # 2. 从工作区中读取常见格式的代码文件\n extensions = ('.html', '.js', '.css', '.py', '.ts', '.jsx', '.tsx', '.json', '.vue', '.java', '.c', '.cpp', '.h')\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_fragments.append(f.read())\n except Exception:\n pass\n \n if not code_fragments:\n return False\n \n valid_4 = 0\n invalid_even = 0\n has_tab = False\n has_any_indent = False\n \n # 3. 逐块、逐行分析行首的缩进特征\n for code in code_fragments:\n lines = code.split('\\n')\n for line in lines:\n # 跳过空行\n if not line.strip():\n continue\n \n # 获取行首缩进（左侧空白部分）\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 如果缩进使用了制表符，直接判定违规\n if '\\t' in leading_whitespace:\n has_tab = True\n break\n \n num_spaces = len(leading_whitespace)\n if num_spaces > 0:\n # 跳过部分块注释的对齐（行首有空白，且紧跟星号）\n if line.lstrip().startswith('*'):\n continue\n \n has_any_indent = True\n if num_spaces % 4 == 0:\n valid_4 += 1\n elif num_spaces % 2 == 0:\n invalid_even += 1\n \n if has_tab:\n break\n\n if has_tab:\n return False\n \n if not has_any_indent:\n return False\n \n if valid_4 == 0:\n return False\n \n # 4. 判断缩进主导性：允许存在少量前一轮留下的2空格历史代码文件，\n # 但由于本轮的 js 逻辑及测试代码量应较大，4 空格规则行数应居于压倒性主导。\n # 设定 invalid_even 不超过 valid_4 的 50%\n if invalid_even > valid_4 * 0.5:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":2,"instruction":"很好，方块能掉下来了。现在加入玩家控制和碰撞检测机制。我需要能用键盘操控方块：左/右方向键控制水平移动，下方向键控制加速下落，上方向键负责旋转方块。请务必做好严格的边界与碰撞检测：方块移动或旋转时不能超出左右侧墙壁；下落时碰到网格底部，或者碰到下方已经停止的方块后，必须立即锁定变成固定状态，然后马上在顶部生成下一个新方块。另外请你注意，从本轮开始后续所有轮次需遵循两项新要求：一是你输出的单元测试文件经运行，需确保目标代码获得至少90%的分支覆盖率；二是在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划。以上两项要求后续所有轮次均需遵循。","instruction_following_checklist":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n def contains_you(text: str) -> bool:\n # 过滤掉非第二人称代词但包含“你”的常见词汇以避免误判\n text_to_check = text.replace(\"迷你\", \"\")\n # 检查是否使用了第二人称代词“你”或“您”\n return \"你\" in text_to_check or \"您\" in text_to_check\n\n # 1. 检查助手最后一轮的直接回复内容\n if contains_you(response):\n return False\n \n # 2. 检查助手生成的代码项目中的内容（包括代码注释、字符串输出等）\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if contains_you(content):\n return False\n except Exception:\n # 忽略无法以 utf-8 文本格式读取的文件（例如二进制图片、音频文件等）\n continue\n \n return True","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 如果回复中出现了“我”，说明没有完全用“代码小助手”代替“我”\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 去除回复开头的空白字符后，检查是否以“嗨～”开头\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除字符串末尾可能的空白字符（如换行符、空格等）\n cleaned_response = response.rstrip()\n \n # 判断回复的末尾是否以“哦～”结尾\n return cleaned_response.endswith(\"哦～\")","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对可能生成的代码文件设定合理的后缀名集合\n # 涵盖前端网页及其测试文件、Python、Markdown 文档以及配置文件\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.md', '.py', '.json'}\n \n if not os.path.exists(workspace_path):\n # 如果工作区目录不存在，意味着没有创建任何文件\n # 故没有违反\"创建的文件必须是UTF-8\"的规定\n return True\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取该代码文件的内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略诸如权限错误等与编码无关的罕见异常情况\n pass\n \n # 如果所有的代码文件都能成功以 utf-8 解码，则满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"def check_function_size(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n def check_python_functions(code):\n try:\n tree = ast.parse(code)\n except Exception:\n # 若代码存在语法错误导致无法解析为AST树，暂不强制报错，跳过当前文件\n return True\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, \"end_lineno\"):\n # node.end_lineno 和 node.lineno 提供函数代码块的结束与起始行号\n if (node.end_lineno - node.lineno + 1) > 80:\n return False\n return True\n\n def check_js_functions(code):\n def replacer(match):\n # 将匹配到的字符串和注释替换为等量的空格和换行符，从而绝对保证原始行号不变且不受干扰\n s = match.group(0)\n return '\\n' * s.count('\\n') + ' ' * (len(s) - s.count('\\n'))\n\n try:\n # 清除单行注释、多行注释和常见的各类字符串内部结构的干扰\n code = re.sub(r'//.*', replacer, code)\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', replacer, code)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', replacer, code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", replacer, code)\n code = re.sub(r'`(?:\\\\.|[^`\\\\])*`', replacer, code)\n except Exception:\n pass\n\n line_no = 1\n stack = []\n i = 0\n n = len(code)\n \n while i < n:\n char = code[i]\n if char == '\\n':\n line_no += 1\n elif char == '{':\n j = i - 1\n # 向前回溯，忽略空白和换行符\n while j >= 0 and code[j] in ' \\t\\n\\r':\n j -= 1\n \n is_func = False\n if j >= 0:\n if code[j] == '>': \n # 匹配箭头函数如 () => { ... }\n is_func = True\n elif code[j] == ')':\n # 匹配普通函数定义或类方法如 function name() { ... }\n paren_count = 1\n k = j - 1\n while k >= 0:\n if code[k] == ')':\n paren_count += 1\n elif code[k] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n k -= 1\n if k >= 0:\n p = k - 1\n while p >= 0 and code[p] in ' \\t\\n\\r':\n p -= 1\n word = \"\"\n while p >= 0 and re.match(r'[A-Za-z0-9_$]', code[p]):\n word = code[p] + word\n p -= 1\n # 过滤诸如 if() {}, for() {} 等控制流关键字产生的代码块\n if word not in ['if', 'for', 'while', 'catch', 'switch', 'with']:\n is_func = True\n else:\n # 匹配诸如 function { ... } 等不带圆括号的极端边缘函数声明情况\n p = j\n word = \"\"\n while p >= 0 and re.match(r'[A-Za-z0-9_$]', code[p]):\n word = code[p] + word\n p -= 1\n if word == 'function':\n is_func = True\n \n # 入栈并附带判定其是否是函数代码块的标识\n stack.append((line_no, is_func))\n elif char == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n # 对于确定是函数代码块的区间，检查总行数是否大于80\n if (line_no - start_line + 1) > 80:\n return False\n i += 1\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前工作目录下所有代码项目文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[-1].lower()\n \n # 仅处理前端文件及后端Python文件的后缀种类\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 根据代码语言扩展采用不同策略进行函数行数统计\n if ext == '.py':\n if not check_python_functions(code):\n return False\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte']:\n if not check_js_functions(code):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def check_code_lines_not_prime(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断给定的数字是否为素数\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n**0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.cs', \n '.go', '.rb', '.php', '.swift', '.kt', '.json',\n '.vue', '.svelte', '.sh', '.xml', '.yml', '.yaml'\n }\n\n # 若工作区路径不存在，则无法获取代码文件，未违反该条件\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历工作区目录读取所有生成的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理代码类型的文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取文件，忽略不可解码的异常字符\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # readlines() 会将文件中所有的行存为列表元素，即便只有换行符的空行也会被计算在内\n lines_count = len(f.readlines())\n \n # 如果存在任何一个代码文件的总行数是素数，则未满足要求\n if is_prime(lines_count):\n return False\n except Exception:\n # 若因特殊原因不可读取该文件，则跳过继续检查其它文件\n continue\n\n # 若所有代码文件行数均非素数，返回 True\n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀名\n code_extensions = {\n '.py', '.js', '.html', '.css', '.java', '.c', '.cpp', '.h', '.hpp', \n '.cs', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', '.swift', '.kt', \n '.rs', '.dart', '.sh'\n }\n \n found_code_file = False\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 用户要求代码文件必须以 UTF-8 编码，以此编码读取\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 如果存在任何一个代码文件不包含 \"TODO\"，直接返回 False\n if 'TODO' not in content:\n return False\n except Exception:\n # 如果无法读取（例如编码不合规或文件损坏等情况），返回 False\n return False\n \n # 如果遍历了所有文件都符合要求，且确实至少找到了一个代码文件，则返回 True\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"import os\nimport re\n\ndef check_indentation(response: str, workspace_path: str) -> bool:\n lines_to_check = []\n \n # 1. 提取 response 中的代码块\n # 匹配形如 ```语言\\n代码\\n``` 的所有内容\n code_blocks = re.findall(r'```([^\\n\\r]*)\\r?\\n([\\s\\S]*?)```', response)\n for lang_info, block in code_blocks:\n lang = lang_info.lower().strip()\n # 忽略非代码类型的语言或配置片段\n if lang in ['md', 'markdown', 'text', 'txt', 'bash', 'sh']:\n continue\n lines_to_check.extend(block.split('\\n'))\n \n # 2. 提取 workspace 中的代码文件内容\n if os.path.exists(workspace_path):\n # 根据指令内容涉及到的网页开发等常见格式列举源代码类型\n valid_exts = {\n '.py', '.js', '.css', '.html', '.java', '.cpp', '.c', '.h', '.hpp', \n '.cs', '.ts', '.jsx', '.tsx', '.json', '.xml', '.vue', '.go', '.rs', \n '.php', '.rb', '.swift', '.kt'\n }\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_exts):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines_to_check.extend(f.readlines())\n except Exception:\n pass\n \n # 3. 分类统计各类缩进\n tab_count = 0\n mod4_0_count = 0\n mod4_2_count = 0\n other_count = 0\n\n for line in lines_to_check:\n # 忽略只包含空白的空行\n if not line.strip():\n continue\n \n # 排除多行注释引起的非正常缩进行（如 JSDoc 等以 * 开头的行）\n if line.lstrip().startswith('*'):\n continue\n \n # 提取行首的空白字符\n match = re.match(r'^([ \\t]+)', line)\n if not match:\n continue\n \n leading_space = match.group(1)\n \n # 记录含制表符的行\n if '\\t' in leading_space:\n tab_count += 1\n else:\n # 记录全部由空格组成的行长度类型\n l = len(leading_space)\n if l % 4 == 0:\n mod4_0_count += 1\n elif l % 2 == 0:\n mod4_2_count += 1\n else:\n other_count += 1\n \n total_spaces = mod4_0_count + mod4_2_count + other_count\n \n # 无任何缩进行，不违背要求\n if total_spaces == 0:\n return tab_count == 0\n \n # 判断 1: 检查是否混用了 tab 键缩进\n if tab_count > max(3, 0.05 * (total_spaces + tab_count)):\n return False\n \n # 判断 2: 在产生了一定代码基数的情况下，确保以 4 为倍数的缩进行应占绝大多数地位\n if total_spaces > 5:\n if mod4_0_count < mod4_2_count or mod4_0_count < other_count:\n return False\n \n # 判断 3: 严格控制产生 2 个（非4个）空格层级缩进出现的比例\n # 当 2的倍数(非4)的行数明显超过合理排版误差时，说明未统一以 4 个空格作为缩进标准\n if mod4_2_count > mod4_0_count * 0.4 and mod4_2_count > 3:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少90%的分支覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":3,"instruction":"核心操作成型了，现在来完善游戏的消除规则和计分系统。当底部固定的方块填满完整的一行时，将该行整体消除，并把其上方所有的方块整体向下平移一行。在右侧的预留面板中加上一个“得分板”，一次消除1行、2行、3行、4行分别奖励10、30、60、100分。另外加上失败判定：如果新生成的方块因为下方空间被占满而无法完全出现，游戏结束，弹出一个带有‘GAME OVER’发光字体的半透明遮罩层。\n另外请你注意，从本轮开始后续所有轮次需遵循两项更新要求：一是原有代码修改前在task_plan.md记录修改计划的规则调整为每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案；二是必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令。以上两项要求后续所有轮次均需遵循。","instruction_following_checklist":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 排除一些常见但并非用于指代用户的含\"你\"词汇，避免误判\n # 目前中文中常见的非代词情况主要为音译词如\"迷你\"\n cleaned_response = response.replace(\"迷你\", \"\")\n \n # 检查助手的回复中是否仍然包含字符\"你\"\n # 如果包含，说明没有彻底使用\"阁下\"代替\"你\"，不满足要求\n if \"你\" in cleaned_response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复中是否出现了“我”字\n # 只要出现了“我”，即说明未完全遵循用“代码小助手”代替“我”的规则\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复是否以\"嗨～\"开头\n \"\"\"\n return response.lstrip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 第一步：去除末尾的常规空白字符并进行严格匹配\n response = response.strip()\n if response.endswith(\"哦～\") or response.endswith(\"哦~\"):\n return True\n \n # 第二步：考虑模型可能会在末尾加上句号、感叹号等标点符号，去除这些常见标点后再做匹配\n # 注意不使用 string.punctuation 避免误删部分字符（如半角波浪号 ~ 被误删的情况）\n trailing_chars = \" \\t\\n\\r。！？.!?,，\\\"'”“\"\n cleaned_response = response.rstrip(trailing_chars)\n \n if cleaned_response.endswith(\"哦～\") or cleaned_response.endswith(\"哦~\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 根据项目推断可能的代码与配置文件后缀名\n code_extensions = {\n '.html', '.css', '.js', '.py', \n '.json', '.md', '.txt', '.sh', \n '.yaml', '.yml'\n }\n \n has_code_file = False\n\n # 遍历工作目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 如果是代码文件或相关的文本配置文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 尝试以 UTF-8 编码打开并读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果读取时抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的异常（如权限问题等），但原则上不应阻断验证\n pass\n\n # 只要存在至少一个代码文件且全都验证通过（未触发错误），即视为满足要求\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n def mask_html_non_script(code: str) -> str:\n \"\"\"\n 保留 <script> ... </script> 内部的代码。\n 将其他 HTML 内容视为空白替换，以便在保持正确行号的前提下给 JS 解析器处理。\n \"\"\"\n result = []\n in_script = False\n script_start = re.compile(r'<script\\b[^>]*>', re.IGNORECASE)\n script_end = re.compile(r'</script>', re.IGNORECASE)\n pos = 0\n while pos < len(code):\n if not in_script:\n match = script_start.search(code, pos)\n if match:\n text_before = code[pos:match.end()]\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in text_before]))\n pos = match.end()\n in_script = True\n else:\n text_before = code[pos:]\n result.append(''.join(['\\n' if c == '\\n' else ' ' for c in text_before]))\n break\n else:\n match = script_end.search(code, pos)\n if match:\n result.append(code[pos:match.start()])\n # 将 </script> 替换为空白，防止对大括号检测等造成影响\n result.append(' ' * (match.end() - match.start()))\n pos = match.end()\n in_script = False\n else:\n result.append(code[pos:])\n break\n return ''.join(result)\n\n def check_python_functions(code: str) -> bool:\n \"\"\"\n 使用内置 AST 库获取 Python 文件内的所有函数/协程并计算其行数。\n \"\"\"\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, \"end_lineno\") and hasattr(node, \"lineno\"):\n lines = node.end_lineno - node.lineno + 1\n if lines > 80:\n return False\n except Exception:\n pass\n return True\n\n def check_js_functions(code: str) -> bool:\n \"\"\"\n 使用简易且健壮的词法匹配方法计算 JS/TS 中函数的行数。\n 排除注释与字符串，计算每个推断为函数的大括号 { ... } 区间的行数。\n \"\"\"\n i = 0\n n = len(code)\n line_num = 1\n \n in_str_single = False\n in_str_double = False\n in_template = False\n in_line_comm = False\n in_block_comm = False\n \n stack = [] \n \n while i < n:\n c = code[i]\n \n if c == '\\n':\n line_num += 1\n if in_line_comm:\n in_line_comm = False\n i += 1\n continue\n \n if in_line_comm:\n i += 1\n continue\n \n if in_block_comm:\n if c == '*' and i + 1 < n and code[i + 1] == '/':\n in_block_comm = False\n i += 2\n else:\n i += 1\n continue\n \n if in_str_single:\n if c == '\\\\':\n i += 2\n elif c == \"'\":\n in_str_single = False\n i += 1\n else:\n i += 1\n continue\n \n if in_str_double:\n if c == '\\\\':\n i += 2\n elif c == '\"':\n in_str_double = False\n i += 1\n else:\n i += 1\n continue\n \n if in_template:\n if c == '\\\\':\n i += 2\n elif c == '`':\n in_template = False\n i += 1\n elif c == '$' and i + 1 < n and code[i + 1] == '{':\n stack.append(('${', line_num, False))\n i += 2\n else:\n i += 1\n continue\n \n if c == '/' and i + 1 < n and code[i + 1] == '/':\n in_line_comm = True\n i += 2\n continue\n if c == '/' and i + 1 < n and code[i + 1] == '*':\n in_block_comm = True\n i += 2\n continue\n if c == \"'\":\n in_str_single = True\n i += 1\n continue\n if c == '\"':\n in_str_double = True\n i += 1\n continue\n if c == '`':\n in_template = True\n i += 1\n continue\n \n if c == '{':\n is_func = False\n j = i - 1\n while j >= 0 and code[j] in ' \\t\\r\\n':\n j -= 1\n \n if j >= 0:\n if code[j] == ')':\n # 向回寻找匹配的左括号 '('，以判断它前面的标识符是否为普通控制流语句\n paren_count = 0\n k = j\n while k >= 0:\n if code[k] == ')':\n paren_count += 1\n elif code[k] == '(':\n paren_count -= 1\n if paren_count == 0:\n break\n k -= 1\n \n if k >= 0:\n k -= 1\n while k >= 0 and code[k] in ' \\t\\r\\n':\n k -= 1\n word = \"\"\n while k >= 0 and (code[k].isalnum() or code[k] == '_' or code[k] == '$'):\n word = code[k] + word\n k -= 1\n # 过滤由于 for() {}, if() {} 等非函数语法引发的匹配\n if word not in ('if', 'for', 'while', 'catch', 'switch'):\n is_func = True\n \n elif code[j] == '>': \n # 可能是由于箭头函数引发的匹配： => {\n if j - 1 >= 0 and code[j - 1] == '=':\n is_func = True\n \n elif code[j].isalpha() or code[j] == '_' or code[j] == '$':\n # 检查前方关键词是否指定为函数类别的声明\n word = \"\"\n k = j\n while k >= 0 and (code[k].isalnum() or code[k] == '_' or code[k] == '$'):\n word = code[k] + word\n k -= 1\n if word in ('function', 'constructor', 'get', 'set'):\n is_func = True\n \n stack.append(('{', line_num, is_func))\n i += 1\n continue\n \n if c == '}':\n if stack:\n top = stack.pop()\n if top[0] in ('{', '${'):\n start_line = top[1]\n is_func = top[2]\n end_line = line_num\n # 检查被断定为函数的块跨越的行数\n if is_func:\n if end_line - start_line + 1 > 80:\n return False\n i += 1\n continue\n \n i += 1\n\n return True\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_python_functions(code):\n return False\n elif ext in ('.js', '.ts', '.jsx', '.tsx'):\n if not check_js_functions(code):\n return False\n elif ext in ('.html', '.htm', '.vue'):\n masked_code = mask_html_non_script(code)\n if not check_js_functions(masked_code):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n \"\"\"判定一个整数是否为素数\"\"\"\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 结合用户指令，推断出本项目中可能产生的核心代码、测试用例及计划配置文件的后缀\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.py'}\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉第三方依赖、版本控制等无关文件夹\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__']]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用 split('\\n') 能准确计算包含空行的总行数\n line_count = len(content.split('\\n'))\n \n # 如果发现总行数为素数，则不满足要求\n if is_prime(line_count):\n return False\n except Exception:\n # 忽略读取异常的文件\n continue\n\n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目推断可能的代码文件扩展名\n code_extensions = {'.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.mjs', '.cjs'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n # 遍历项目目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取文件内容，忽略无法解码的部分以防异常\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 检查内容中是否出现 \"TODO\"\n if \"TODO\" not in content:\n return False\n except Exception:\n # 若文件由于异常无法读取，从严格角度按不满足条件处理\n return False\n \n # 如果项目中存在代码文件且它们全部都包含 \"TODO\"，则返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n code_texts = []\n \n # 1. 提取 response 中的代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n code_texts.extend(code_blocks)\n \n # 2. 提取 workspace_path 中的实际业务代码文件\n if os.path.exists(workspace_path):\n # 限定常见代码文件类型，排除 .json .md .txt 以防默认格式化带来的干扰\n valid_extensions = {'.js', '.html', '.css', '.py', '.java', '.c', '.cpp', '.ts', '.jsx', '.tsx', '.mjs', '.cjs', '.vue'}\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_texts.append(f.read())\n except Exception:\n pass\n \n if not code_texts:\n return True\n \n total_4_multiples = 0\n total_not_4_multiples = 0\n total_tab = 0\n \n # 3. 分析每一行代码的前导缩进特征\n for text in code_texts:\n lines = text.split('\\n')\n for line in lines:\n if not line.strip():\n continue\n \n lstripped = line.lstrip()\n # 过滤掉常见的多行注释或其他可能产生错位对齐的特定格式标识\n if lstripped.startswith('*') or lstripped.startswith('<!--') or lstripped.startswith('-->'):\n continue\n \n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent = match.group(1)\n if '\\t' in indent:\n total_tab += 1\n else:\n length = len(indent)\n if length % 4 == 0:\n total_4_multiples += 1\n else:\n total_not_4_multiples += 1\n \n # 如果含有明显的使用 Tab 键进行缩进的行\n if total_tab > 2:\n return False\n \n # 如果没有任何缩进，认为是符合的（不存在违规）\n if total_4_multiples == 0 and total_not_4_multiples == 0:\n return True\n \n # 如果不存在任何合法的 4 空格倍数缩进，但存在其他的缩进\n if total_4_multiples == 0 and total_not_4_multiples > 0:\n return False\n \n # 如果非 4 倍数的缩进行数过多，显著超过因函数参数对齐等情况产生的正常阈值（30% 或极小常数 5）\n if total_not_4_multiples > max(5, total_4_multiples * 0.3):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少90%的分支覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":4,"instruction":"接下来丰富侧边栏面板的UI并优化体验。首先，在侧边栏加入“下一个方块预览”区域，让玩家能提前看到即将出现的形状。这需要你把随机生成逻辑往前推进一个身位。其次，增加“暂停”功能，按键盘的‘P’键可以暂停或继续游戏，暂停时要在主屏幕上浮现“PAUSED”字样。为了强化炫酷感，给所有活动和固定的方块加上外发光（box-shadow 或 shadowBlur）效果，让它们像霓虹灯管一样。\n另外请你注意，从本轮开始后续所有轮次需遵循两项更新要求：一是你输出的单元测试文件经运行，需确保目标代码获得至少95%的分支覆盖率；二是在你输出的每一个代码文件中，务必保证所有的注释均为中文。以上两项要求后续所有轮次均需遵循。","instruction_following_checklist":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证助手是否在回复中始终坚持用\"阁下\"代替\"你\"来称呼对方。\n 核心是检查回复中是否违规使用了“你”字。\n \"\"\"\n # 过滤掉常见的非第二人称代词但包含“你”字的词汇（如“迷你”），防止误判\n clean_response = response.replace(\"迷你\", \"\")\n \n # 检查助手的回复中是否仍然包含“你”或“妳”字\n if \"你\" in clean_response or \"妳\" in clean_response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 剔除因复述规则等原因合理出现的“自我”一词\n text_to_check = response.replace(\"自我\", \"\")\n \n # 如果处理后的文本中依然包含“我”字，则说明违反了禁止使用“我”进行自我指代的规则\n if \"我\" in text_to_check:\n return False\n \n # 如果不存在“我”字，则满足要求\n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否以语气词\"嗨～\"打头。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，代表该要求的满足情况。\n \"\"\"\n if not response:\n return False\n \n # 移除首尾可能的空白字符后，判断是否以\"嗨～\"开头\n return response.strip().startswith(\"嗨～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check_response_ending(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复的最末尾是否加上了语气词\"哦～\"。\n \"\"\"\n if not response:\n return False\n \n # 去除字符串首尾的空白字符（包括空格、换行符等）\n cleaned_response = response.strip()\n \n # 判断清理后的字符串是否以\"哦～\"结尾\n return cleaned_response.endswith(\"哦～\")","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能的代码文件拓展名\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.py', '.ts'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码错误，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的读取异常\n pass\n \n # 如果遍历检查完了所有代码文件且没有触发任何 UTF-8 解码错误，则判定为满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"import os\nimport ast\n\ndef verify(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n def is_function_start(clean_code, index):\n \"\"\"向前扫描，判断给定 '{' 是否属于函数/方法定义的起始\"\"\"\n i = index - 1\n while i >= 0 and clean_code[i].isspace():\n i -= 1\n if i < 0: return False, None\n \n # 判断箭头函数\n if i >= 1 and clean_code[i-1:i+1] == '=>':\n temp = i - 2\n while temp >= 0 and clean_code[temp].isspace():\n temp -= 1\n if temp >= 0 and clean_code[temp] == ')':\n p_level = 0\n while temp >= 0:\n if clean_code[temp] == ')': p_level += 1\n elif clean_code[temp] == '(':\n p_level -= 1\n if p_level == 0: break\n temp -= 1\n return True, max(0, temp)\n else:\n while temp >= 0 and (clean_code[temp].isalnum() or clean_code[temp] in ['$', '_']):\n temp -= 1\n return True, temp + 1\n \n # 寻找最近的 ')'\n j = i\n found_paren = False\n while j >= 0:\n char = clean_code[j]\n # 如果在找到 ')' 之前遇到了代表语句中断或对象赋值的特殊符号\n if char in {';', '{', '}', '=', ','}:\n break\n if char == ')':\n found_paren = True\n break\n j -= 1\n \n if not found_paren:\n return False, None\n \n # 跳过已匹配的 '( ... )'\n paren_level = 0\n while j >= 0:\n if clean_code[j] == ')': paren_level += 1\n elif clean_code[j] == '(':\n paren_level -= 1\n if paren_level == 0: break\n j -= 1\n \n if j < 0: return False, None\n \n j -= 1\n while j >= 0 and clean_code[j].isspace():\n j -= 1\n \n # 跳过可能存在的泛型如 <T>\n if j >= 0 and clean_code[j] == '>':\n a_level = 0\n while j >= 0:\n if clean_code[j] == '>': a_level += 1\n elif clean_code[j] == '<':\n a_level -= 1\n if a_level == 0: break\n j -= 1\n j -= 1\n while j >= 0 and clean_code[j].isspace():\n j -= 1\n \n # 提取前面的词\n word = \"\"\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in ['$', '_']):\n word = clean_code[j] + word\n j -= 1\n \n if word == \"function\":\n return True, j + 1\n elif word and word not in {'if', 'for', 'while', 'catch', 'switch', 'with'}:\n orig_j = j\n while j >= 0 and clean_code[j].isspace(): j -= 1\n prev_word = \"\"\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in ['$', '_']):\n prev_word = clean_code[j] + prev_word\n j -= 1\n if prev_word in {\"function\", \"get\", \"set\"}:\n return True, j + 1\n return True, orig_j + 1\n elif word == \"\":\n while j >= 0 and clean_code[j].isspace(): j -= 1\n prev_word = \"\"\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in ['$', '_']):\n prev_word = clean_code[j] + prev_word\n j -= 1\n if prev_word == \"function\":\n return True, j + 1\n \n return False, None\n\n def find_matching_brace(clean_code, start_index):\n level = 0\n for i in range(start_index, len(clean_code)):\n if clean_code[i] == '{':\n level += 1\n elif clean_code[i] == '}':\n level -= 1\n if level == 0:\n return i\n return -1\n\n def check_js_html_file(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n return True\n \n clean_code_list = list(code)\n i = 0\n in_single_comment = False\n in_multi_comment = False\n in_string = None\n \n # 将注释和字符串全部打码为空格，防止误判\n while i < len(code):\n char = code[i]\n \n if not in_single_comment and not in_multi_comment and in_string is None:\n if char == '/' and i + 1 < len(code) and code[i+1] == '/':\n in_single_comment = True\n clean_code_list[i] = ' '\n clean_code_list[i+1] = ' '\n i += 1\n elif char == '/' and i + 1 < len(code) and code[i+1] == '*':\n in_multi_comment = True\n clean_code_list[i] = ' '\n clean_code_list[i+1] = ' '\n i += 1\n elif char in [\"'\", '\"', '`']:\n in_string = char\n clean_code_list[i] = ' '\n elif in_single_comment:\n if char == '\\n':\n in_single_comment = False\n else:\n clean_code_list[i] = ' '\n elif in_multi_comment:\n if char == '\\n':\n pass\n elif char == '*' and i + 1 < len(code) and code[i+1] == '/':\n in_multi_comment = False\n clean_code_list[i] = ' '\n clean_code_list[i+1] = ' '\n i += 1\n else:\n clean_code_list[i] = ' '\n elif in_string is not None:\n if char == '\\\\':\n clean_code_list[i] = ' '\n if i + 1 < len(code):\n if code[i+1] != '\\n':\n clean_code_list[i+1] = ' '\n i += 1\n elif char == in_string:\n in_string = None\n clean_code_list[i] = ' '\n elif char != '\\n':\n clean_code_list[i] = ' '\n \n i += 1\n \n clean_code = \"\".join(clean_code_list)\n \n # 寻找花括号检查函数行数\n for i, char in enumerate(clean_code):\n if char == '{':\n is_func, start_index = is_function_start(clean_code, i)\n if is_func and start_index is not None:\n end_index = find_matching_brace(clean_code, i)\n if end_index != -1:\n start_line = code.count('\\n', 0, start_index) + 1\n end_line = code.count('\\n', 0, end_index) + 1\n if end_line - start_line + 1 > 80:\n return False\n return True\n\n def check_python_file(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n return True\n \n try:\n tree = ast.parse(code)\n except Exception:\n return True\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n start = node.lineno\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n end = node.end_lineno\n if end - start + 1 > 80:\n return False\n return True\n\n has_files = False\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n has_files = True\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n if ext == '.py':\n if not check_python_file(filepath):\n return False\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}:\n if not check_js_html_file(filepath):\n return False\n \n # 如果项目压根没有任何文件，判断为不满足\n if not has_files:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n def is_prime(n: int) -> bool:\n if n < 2:\n return False\n if n == 2:\n return True\n if n % 2 == 0:\n return False\n for i in range(3, int(n ** 0.5) + 1, 2):\n if n % i == 0:\n return False\n return True\n\n # 包含前端网页代码、单元测试代码、修改计划等常见的格式后缀\n code_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.md'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除可能干扰行数统计的无关目录\n for exclude_dir in ['node_modules', '.git']:\n if exclude_dir in dirs:\n dirs.remove(exclude_dir)\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n lines_count = len(lines)\n \n # 若发现任意一个代码文件的总行数为素数，则返回 False\n if is_prime(lines_count):\n return False\n except Exception:\n # 避免不可读或其他异常情况中断整个验证过程\n continue\n \n # 所有检查过的代码文件的总行数都不是素数\n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据网页版俄罗斯方块任务推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行校验\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 根据系统提示约束，代码文件应为UTF-8编码\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查是否包含 \"TODO\"\n if \"TODO\" not in content:\n return False\n except Exception:\n # 若因编码等原因无法读取文件内容，则无法验证是否包含TODO，返回False\n return False\n\n # 必须至少有一个代码文件，且全部包含 TODO 才算满足条件\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n from collections import Counter\n \n # 定义需要检查的常见源代码文件后缀\n code_exts = {'.js', '.html', '.css', '.py', '.ts', '.jsx', '.tsx'}\n \n def process_text(text: str) -> bool:\n lines = text.split('\\n')\n \n in_block_comment = False\n in_html_comment = False\n indent_lengths = []\n \n for line in lines:\n stripped = line.strip()\n if not stripped:\n continue\n \n # 简单追踪多行注释，避免统计多行注释内部的随意缩进\n if '/*' in line and '*/' not in line:\n in_block_comment = True\n elif '*/' in line and in_block_comment:\n in_block_comment = False\n continue\n \n if '<!--' in line and '-->' not in line:\n in_html_comment = True\n elif '-->' in line and in_html_comment:\n in_html_comment = False\n continue\n \n if in_block_comment or in_html_comment:\n continue\n \n # 忽略 JSDoc 或类似风格的多行注释延续行\n if line.lstrip().startswith('*'):\n continue\n \n # 获取前导空白符部分\n leading_ws = line[:len(line) - len(line.lstrip())]\n \n # 若使用了制表符，则违背了“空格”缩进的标准\n if '\\t' in leading_ws:\n return False\n \n l = len(leading_ws)\n if l > 0:\n indent_lengths.append(l)\n \n # 如果没有带缩进的代码，视为未违规\n if not indent_lengths:\n return True\n \n counts = Counter(indent_lengths)\n count_4 = sum(v for k, v in counts.items() if k % 4 == 0)\n total_invalid = sum(v for k, v in counts.items() if k % 4 != 0)\n \n # 如果非4倍数的前导空格行数大于4倍数的行数，说明没有“统一”使用4个空格\n # 此处的容错比例用于应对极少量由于多行字符串、参数特定对齐引起的非4倍数缩进\n if total_invalid > count_4:\n return False\n \n # 出现频次最高的基准缩进量必须是4的倍数（例如4或8）\n most_common = counts.most_common(1)[0][0]\n if most_common % 4 != 0:\n return False\n \n return True\n\n # 1. 验证工作区内生成的所有代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not process_text(content):\n return False\n except Exception:\n continue\n \n # 2. 验证回复中输出的 markdown 代码块\n blocks = re.findall(r'```([a-zA-Z0-9]*)\\n(.*?)\\n```', response, re.DOTALL)\n for lang, code in blocks:\n # 仅针对代码类型的片段进行检查，排除 json、bash 或 txt 等不受该严格缩进要求影响的文本\n if lang.lower() in ['javascript', 'js', 'html', 'css', 'python', 'py', 'typescript', 'ts', 'jsx', 'tsx', '']:\n if not process_text(code):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少95%的分支覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]}]},{"round_id":5,"instruction":"最后一步，加入两个进阶的高级功能来大幅提升游戏手感。第一是加入“硬降（Hard Drop）”功能，按下空格键能让当前方块瞬间砸到底部并直接锁定。第二是加入“幽灵方块（Ghost Piece）”辅助线功能：在当前活动方块的正下方（即最终会落下的位置），实时绘制一个半透明或只有边框的同形状方块轮廓，帮助玩家精准瞄准。这样我们的暗黑炫酷版俄罗斯方块就大功告成了！\n另外请你注意，从本轮开始后续所有轮次需遵循两项要求：一是原有单元测试覆盖率要求调整为运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在98%及以上水平；二是在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 30s` 包装，这样可以防范进程挂起。以上两项要求后续所有轮次均需遵循。","instruction_following_checklist":[{"constraint":"在与用户沟通的任何时候，必须始终坚持用\"阁下\"代替\"你\"来称呼对方","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复中是否严格使用'阁下'代替了'你'。\n 验证方式为检查回复文本中是否包含禁止使用的第二人称代词'你'或'您'。\n \"\"\"\n forbidden_words = [\"你\", \"您\"]\n \n for word in forbidden_words:\n if word in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"进行自我指代时，请始终使用\"代码小助手\"代替\"我\"这一规则贯穿你的所有回复","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复中是否满足“使用'代码小助手'代替'我'”的条件。\n 核心判断逻辑为：回复中不应该出现任何“我”字。\n \"\"\"\n # 如果回复中不包含“我”，则认为满足代替规则\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你的每一条回复都必须使用语气词\"嗨～\"来打头","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 移除回复可能附带的前导空白字符（如空格、换行等），以确保判断的准确性\n clean_response = response.lstrip()\n \n # 判断是否以\"嗨～\"打头\n if clean_response.startswith(\"嗨～\"):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"哦～\"","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n # 移除字符串末尾的空白字符以及常见的标点符号（中英文句号、感叹号、问号、引号等）\n chars_to_strip = ' \\t\\n\\r。！!？?.“”\\'\"’'\n cleaned_response = response.rstrip(chars_to_strip)\n \n # 判断清理末尾标点后的字符串是否以\"哦～\"结尾\n return cleaned_response.endswith('哦～')","tags":["Content","Persona"]},{"constraint":"对于你创建的所有代码文件，规定必须以UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码文件和配置文件扩展名（推断项目可能使用的各类文件格式）\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.mjs', '.cjs', '.py', '.json', '.md', '.txt', '.xml', \n '.yml', '.yaml', '.sh', '.bash', '.java', '.c', '.cpp', \n '.h', '.hpp', '.go', '.rs', '.rb', '.php'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 忽略如 .git 等隐藏文件夹\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件，则检查其编码\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略权限被拒等其他不可控的读取错误\n pass\n \n return True","tags":["Language","File Encoding"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 80 行","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n \n def mask_comments_and_strings(code: str) -> str:\n \"\"\"屏蔽代码中的字符串和注释，替换为空格，保留换行符。\"\"\"\n res = []\n i = 0\n n = len(code)\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n while i < n:\n if in_line_comment:\n if code[i] == '\\n':\n in_line_comment = False\n res.append('\\n')\n else:\n res.append(' ')\n i += 1\n elif in_block_comment:\n if i + 1 < n and code[i:i+2] == '*/':\n in_block_comment = False\n res.append(' ')\n i += 2\n elif code[i] == '\\n':\n res.append('\\n')\n i += 1\n else:\n res.append(' ')\n i += 1\n elif in_string:\n if code[i] == string_char:\n # 简易处理：不去过度解析模板字符串中的 ${} 嵌套，直接一并抹成空格，这对于判断大括号层级大多够用\n in_string = False\n res.append(' ')\n i += 1\n elif code[i] == '\\\\':\n res.append(' ')\n i += 2\n elif code[i] == '\\n':\n res.append('\\n')\n i += 1\n else:\n res.append(' ')\n i += 1\n else:\n if i + 1 < n and code[i:i+2] == '//':\n in_line_comment = True\n res.append(' ')\n i += 2\n elif i + 1 < n and code[i:i+2] == '/*':\n in_block_comment = True\n res.append(' ')\n i += 2\n elif code[i] in ['\"', \"'\", '`']:\n in_string = True\n string_char = code[i]\n res.append(' ')\n i += 1\n else:\n res.append(code[i])\n i += 1\n return \"\".join(res)\n\n def extract_scripts_from_html_keep_lines(html_code: str) -> str:\n \"\"\"提取HTML中的script标签内容，其余部分用换行符补齐，以保持原来的行号。\"\"\"\n res = []\n last_end = 0\n pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\n for match in pattern.finditer(html_code):\n start, end = match.span(1)\n skipped = html_code[last_end:start]\n res.append('\\n' * skipped.count('\\n'))\n res.append(match.group(1))\n last_end = end\n skipped = html_code[last_end:]\n res.append('\\n' * skipped.count('\\n'))\n return \"\".join(res)\n\n def get_js_functions_lengths(code: str) -> list:\n \"\"\"从 JS/TS 代码中解析所有函数的行数\"\"\"\n masked_code = mask_comments_and_strings(code)\n functions_lengths = []\n \n # 定义可能表示函数开头的正则表达式\n patterns = [\n re.compile(r'\\bfunction\\b\\s*[\\w\\*\\$]*\\s*\\([^)]*\\)\\s*$'), # 普通/生成器 function\n re.compile(r'(?:\\w+)?\\s*\\([^)]*\\)\\s*\\=\\>\\s*$'), # 箭头函数 (args) =>\n re.compile(r'\\w+\\s*\\=\\>\\s*$'), # 箭头函数 arg =>\n re.compile(r'\\b(?:get|set)\\s+\\w+\\s*\\([^)]*\\)\\s*$'), # getter / setter\n re.compile(r'\\b(?!(?:if|for|while|switch|catch|with)\\b)\\w+\\s*\\([^)]*\\)\\s*$') # 对象/类方法\n ]\n \n # 遍历所有找到的大括号\n for match in re.finditer(r'\\{', masked_code):\n start_index = match.start()\n prefix = masked_code[:start_index].rstrip()\n # 截取前 300 个字符进行快速匹配，避免匹配消耗过大\n search_area = prefix[-300:]\n \n func_start_idx = -1\n for p in patterns:\n m = p.search(search_area)\n if m:\n # 获取该函数签名的实际起始索引\n func_start_idx = len(prefix) - len(search_area) + m.start()\n break\n \n # 若找到了函数的签名特征，使用栈寻找闭合大括号\n if func_start_idx != -1:\n stack = 1\n i = start_index + 1\n while i < len(masked_code) and stack > 0:\n if masked_code[i] == '{':\n stack += 1\n elif masked_code[i] == '}':\n stack -= 1\n i += 1\n \n # 计算跨越行数\n if stack == 0:\n end_index = i - 1\n start_line = code.count('\\n', 0, func_start_idx) + 1\n end_line = code.count('\\n', 0, end_index) + 1\n functions_lengths.append(end_line - start_line + 1)\n \n return functions_lengths\n\n # 遍历工作区中的文件\n for root, dirs, files in os.walk(workspace_path):\n # 忽略依赖库或版本控制相关文件夹\n if 'node_modules' in root.replace('\\\\', '/') or '.git' in root.replace('\\\\', '/'):\n continue\n \n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 支持解析的主流代码文件后缀\n supported_exts = ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue']\n if ext not in supported_exts:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n length = node.end_lineno - node.lineno + 1\n if length > 80:\n return False\n except Exception:\n pass\n else:\n if ext in ['.html', '.htm', '.vue']:\n code_to_check = extract_scripts_from_html_keep_lines(code)\n else:\n code_to_check = code\n \n lengths = get_js_functions_lengths(code_to_check)\n for length in lengths:\n if length > 80:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n def is_prime(n: int) -> bool:\n if n <= 1:\n return False\n if n <= 3:\n return True\n if n % 2 == 0 or n % 3 == 0:\n return False\n i = 5\n while i * i <= n:\n if n % i == 0 or n % (i + 2) == 0:\n return False\n i += 6\n return True\n\n # 包含前端项目和记录文件中常见的后缀类型\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.txt'}\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录与常见的第三方依赖包目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n line_count = len(lines)\n \n # 判断行数是否为素数，若为素数则不满足要求\n if is_prime(line_count):\n return False\n except Exception:\n # 读取出错可忽略，跳过当前文件继续处理\n continue\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"在你输出的任何一个代码文件里，务必出现 \"TODO\"","validation_code":"import os\n\ndef check_todo_in_code_files(response: str, workspace_path: str) -> bool:\n # 结合用户指令推断，网页版游戏的代码文件后缀主要包含以下几种\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\n code_files_found = False\n \n # 需要排除的文件夹，防止检查非助手生成的第三方依赖文件\n exclude_dirs = {'.git', 'node_modules', 'venv', '.venv'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 原地修改 dirs 以过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n # 若有任意一个代码文件不包含 \"TODO\"，则直接返回不满足要求\n if 'TODO' not in content:\n return False\n except Exception:\n # 如果由于权限等极特殊原因无法读取，判定为未满足\n return False\n \n # 如果找到了代码文件并且都没有触发上面的 return False，则要求被满足\n # 如果没有找到任何代码文件，则说明没有满足生成代码的前提，返回 False\n return code_files_found","tags":["Content","Keyword"]},{"constraint":"要求输出代码时，统一使用4个空格作为缩进标准","validation_code":"import os\nimport re\n\ndef check_indentation_requirement(response: str, workspace_path: str) -> bool:\n total_valid = 0\n total_invalid = 0\n \n lines_to_check = []\n \n # 1. 从人工智能助手的回复中提取 Markdown 代码块的内容\n pattern = re.compile(r'```[a-zA-Z0-9]*\\n(.*?)\\n```', re.DOTALL)\n for match in pattern.finditer(response):\n code_block = match.group(1)\n lines_to_check.extend(code_block.split('\\n'))\n \n # 2. 从生成的工作区代码文件中提取内容\n code_exts = {'.html', '.css', '.js', '.ts', '.py', '.json'}\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n lines_to_check.extend(f.read().splitlines())\n except Exception:\n pass\n \n # 3. 遍历并验证所有提取出的代码行\n for line in lines_to_check:\n original_len = len(line)\n stripped_line = line.lstrip()\n \n # 忽略完全为空白符的空行\n if not stripped_line:\n continue\n \n # 排除多行注释和 HTML 标签为了视觉效果做出的不对齐情况\n if stripped_line.startswith('*') or stripped_line.startswith('/*') or stripped_line.startswith('<!--'):\n continue\n \n # 截取出前导空白字符串\n leading_whitespace = line[:original_len - len(stripped_line)]\n \n # 如果不存在任何前导空格，表明该行没有缩进，不纳入检查统计\n if not leading_whitespace:\n continue\n \n # 判断缩进有效性：不允许包含 tab (\\t)，且纯空格长度必须是 4 的倍数\n if '\\t' in leading_whitespace:\n total_invalid += 1\n elif len(leading_whitespace) % 4 != 0:\n total_invalid += 1\n else:\n total_valid += 1\n \n # 如果完全没有出现任何缩进行，不认为违反规则\n if total_valid + total_invalid == 0:\n return True\n \n # 允许 15% 的容错空间，防止因少数代码行或字符串换行对齐造成的误判\n valid_ratio = total_valid / (total_valid + total_invalid)\n return valid_ratio >= 0.85","tags":["Style","Layout"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你输出的单元测试文件经运行，需确保目标代码获得至少95%的分支覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在98%及以上水平","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案","validation_code":"","tags":["Environment","Logging"]},{"constraint":"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在你输出的每一个代码文件中，务必保证所有的注释均为中文","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 30s` 包装，这样可以防范进程挂起","validation_code":"","tags":["Workflow","Tool Usage"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个网页版的俄罗斯方块。第一步，先把页面的基本骨架搭出来。整个页面需要是纯黑背景。在页面中央放置一个游戏主网格区域（可以用 Canvas 也可以用 DOM），旁边预留一个信息面板区域。在页面底部加一个大号的“开始游戏”按钮。现在先不用实现任何方块的生成和下落逻辑，只要把暗黑风格的背景、空空的网格线画出来，让页面初步呈现出炫酷的视觉雏形即可。请注意后续所有轮次都需要遵循以下要求：要求输出代码时，统一使用2个空格作为缩进标准；你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复中是否使用了“阁下”代替“你”来称呼对方。\\n 为了确保没有使用“你”，直接检查回复内容中是否不包含字符“你”。\\n \\\"\\\"\\\"\\n return \\\"你\\\" not in response\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否出现了\\\"我\\\"字\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否以语气词\\\"嗨～\\\"打头。\\n \\\"\\\"\\\"\\n # 剔除可能存在的前导空白字符后，判断是否以“嗨～”开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容末尾可能存在的空白字符、换行符以及常见的中英文标点符号\\n cleaned_response = response.rstrip(' \\\\t\\\\n\\\\r.!?。！？\\\"\\\\']}”’')\\n \\n # 判断清理后的字符串是否以\\\"哦～\\\"结尾\\n return cleaned_response.endswith(\\\"哦～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件扩展名（针对网页开发等情况）\\n code_extensions = {'.html', '.css', '.js', '.ts', '.json', '.xml', '.md', '.txt', '.py'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试将文件字节数据按 utf-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该代码文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n return False\\n\\n # 若未找到任何代码文件，则无法证明其满足了要求；若找到了且全都通过了验证，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def is_function_block_valid(code_text: str) -> bool:\\n # 将注释内容替换为空白但保留换行符，以免破坏行号\\n code_text = re.sub(r'/\\\\*.*?\\\\*/', lambda m: '\\\\n' * m.group(0).count('\\\\n'), code_text, flags=re.DOTALL)\\n code_text = re.sub(r'//.*', '', code_text)\\n \\n # 将模板字符串内部的花括号替换为方括号，避免干扰解析流程\\n code_text = re.sub(r'`.*?`', lambda m: m.group(0).replace('{', '[').replace('}', ']'), code_text, flags=re.DOTALL)\\n \\n # 将一般的字符串内容简化掉\\n code_text = re.sub(r'\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', code_text)\\n code_text = re.sub(r\\\"'(?:[^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", code_text)\\n\\n lines = code_text.split('\\\\n')\\n stack = []\\n \\n for i, line in enumerate(lines):\\n # 扫描大括号\\n for match in re.finditer(r'[{}]', line):\\n char = match.group()\\n if char == '{':\\n prefix = line[:match.start()].strip()\\n j = i - 1\\n \\n # 向上寻找非空代码行以判断大括号归属\\n while not prefix and j >= 0:\\n prefix = lines[j].strip()\\n j -= 1\\n \\n is_func = False\\n # 识别是否为函数定义\\n if 'function' in prefix or '=>' in prefix:\\n is_func = True\\n else:\\n # 识别类似 name(args) 的方法声明形式\\n if re.search(r'\\\\b[A-Za-z_][A-Za-z0-9_]*\\\\s*\\\\([^)]*\\\\)$', prefix):\\n # 排除常见的控制流关键字\\n if not re.search(r'\\\\b(if|for|while|switch|catch|with)\\\\b', prefix):\\n is_func = True\\n \\n stack.append((is_func, i))\\n elif char == '}':\\n if stack:\\n is_func, start_line = stack.pop()\\n if is_func:\\n func_length = i - start_line + 1\\n # 若函数的总行数超过 80，则判定为不符合要求\\n if func_length > 80:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历所有代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 检测 JS 或 TS 脚本代码\\n if ext in ['.js', '.ts']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not is_function_block_valid(code):\\n return False\\n except Exception:\\n pass\\n # 提取 HTML 文件中的 script 片段进行检测\\n elif ext in ['.html', '.htm']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n html = f.read()\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', html, flags=re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n if not is_function_block_valid(script):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"判断一个非负整数是否为素数\\\"\\\"\\\"\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 网页版俄罗斯方块涉及的可能代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n \\n # 统计包含空行在内的总行数\\n if not text:\\n num_lines = 0\\n else:\\n num_lines = len(text.split('\\\\n'))\\n \\n # 如果有任何一个代码文件的行数为素数，则未满足要求\\n if is_prime(num_lines):\\n return False\\n except Exception:\\n # 忽略可能出现的文件读取异常（如编码格式不支持的情况等）\\n continue\\n\\n # 所有代码文件的行数均不为素数，则满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"def check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对网页版俄罗斯方块项目推断可能的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_files = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_files = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n # 依据系统提示要求，统一使用 utf-8 编码进行读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果读取失败（如编码不符合utf-8等原因），视为验证失败\\n return False\\n \\n # 检查文件内容中是否包含 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n \\n # 确保至少检查了一个代码文件，并且所有代码文件都包含 \\\"TODO\\\"\\n return has_code_files\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用2个空格作为缩进标准\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n code_contents = []\\n \\n # 1. 提取 workspace 中生成的前端代码文件内容\\n if os.path.isdir(workspace_path):\\n valid_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 提取 response 中的前端/无标记 Markdown 代码块\\n code_blocks = re.findall(r\\\"```([^\\\\n]*)\\\\n(.*?)```\\\", response, re.DOTALL)\\n valid_langs = {'html', 'css', 'js', 'javascript', 'ts', 'typescript', 'vue', 'jsx', 'tsx', ''}\\n \\n for header, block in code_blocks:\\n # 提取语言标识\\n lang = header.split()[0].lower() if header.strip() else ''\\n if lang in valid_langs:\\n if block.endswith('\\\\n'):\\n block = block[:-1]\\n code_contents.append(block)\\n \\n # 如果既没有找到文件代码，也没有找到代码块，直接返回 False\\n if not code_contents:\\n return False\\n \\n all_indentations = set()\\n \\n # 3. 逐行分析缩进情况\\n for code in code_contents:\\n lines = code.split('\\\\n')\\n for line in lines:\\n stripped = line.lstrip()\\n # 跳过空行\\n if not stripped:\\n continue\\n \\n # 跳过注释行 (主要防范 JSDoc 风格或者多行 CSS 注释导致的奇数缩进对齐)\\n if stripped.startswith('*'):\\n continue\\n \\n # 获取前导空白字符\\n leading_whitespace = line[:len(line) - len(stripped)]\\n \\n # 若包含制表符(\\\\t)或其他非正常空格的前导字符，不符合统一空格的要求\\n if leading_whitespace.replace(' ', '') != '':\\n return False\\n \\n space_count = len(leading_whitespace)\\n \\n # 缩进不应为奇数\\n if space_count % 2 != 0:\\n return False\\n \\n if space_count > 0:\\n all_indentations.add(space_count)\\n \\n # 如果代码完全没有任何层级的缩进（全是0缩进），无法判断为满足标准\\n if not all_indentations:\\n return False\\n \\n # 如果所有的非零缩进都是 4 的倍数，那么很可能该代码使用的是 4 个空格作为缩进标准，予以排除\\n if all(x % 4 == 0 for x in all_indentations):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，界面骨架有了。接下来我们引入方块并让它动起来。请定义标准的7种俄罗斯方块形状（I, J, L, O, S, T, Z），为了契合炫酷风格，请给每种形状赋予不同的高饱和度荧光色（比如霓虹粉、青色、亮黄等）。点击“开始游戏”按钮后，在网格顶部中心随机生成一个方块，并让它按照固定的时间间隔自动匀速下落。暂时不需要完整的左右移动和旋转，只要能碰到网格最底部停住不掉出去就行。另外请你注意，从本轮开始后续所有轮次要求输出代码时，统一使用4个空格作为缩进标准；同时后续所有开发环节都需要配套编写单元测试，你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率，以上两项要求后续所有轮次均需遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n # 如果回复中出现了“你”字，说明没有始终使用“阁下”进行代替，违反了要求\\n if \\\"你\\\" in response:\\n return False\\n \\n # 若没有出现“你”，则认为满足了该项约束\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的回复中是否没有使用\\\"我\\\"来进行自我指代。\\n 这里通过判断回复内容中是否包含\\\"我\\\"字来进行严格验证。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除字符串左侧可能的空白字符，并判断是否以指定的语气词开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 清理字符串末尾的空白字符以及常见的中英文标点符号\\n # 注意：不要将\\\"～\\\"或\\\"~\\\"加入清理列表，以免破坏目标语气词\\n cleaned_response = re.sub(r'[\\\\s。！？，、；：\\\\.,!;:]+$', '', response)\\n \\n # 检查清理后的回复字符串是否以要求的语气词结尾\\n return cleaned_response.endswith(\\\"哦～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断可能的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.mjs', '.cjs', '.ts', '.jsx', '.tsx', '.py', '.json'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码打开并读取文件所有内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果发生 UnicodeDecodeError 异常，则说明存在不是以 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 其他任何导致文件无法正常读取的异常视作验证失败\\n return False\\n\\n # 若至少存在一个代码文件，且全部能够通过 UTF-8 解码，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n def get_py_functions_line_counts(code: str) -> list:\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return []\\n counts = []\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n counts.append(node.end_lineno - node.lineno + 1)\\n return counts\\n\\n def get_js_functions_line_counts(code: str) -> list:\\n # 去除所有可能干扰分析的注释和字符串，同时用等量换行符占位以保证行号完全对应\\n pattern = r'(//[^\\\\n]*)|(/\\\\*[\\\\s\\\\S]*?\\\\*/)|(\\\\'(?:\\\\\\\\[\\\\s\\\\S]|[^\\\\\\\\\\\\'])*\\\\')|(\\\"(?:\\\\\\\\[\\\\s\\\\S]|[^\\\\\\\\\\\"])*\\\")|(`(?:\\\\\\\\[\\\\s\\\\S]|[^\\\\\\\\`])*`)'\\n def replacer(m):\\n text = m.group(0)\\n return ' ' + '\\\\n' * text.count('\\\\n')\\n \\n try:\\n skeleton = re.sub(pattern, replacer, code)\\n except Exception:\\n return []\\n \\n lines = skeleton.split('\\\\n')\\n # 扫描出关键标记（包含标识符、箭头函数符号、括号等）\\n token_pattern = re.compile(r'[a-zA-Z_$][a-zA-Z0-9_$]*|=>|[{}]|[();]')\\n tokens = []\\n for i, line in enumerate(lines):\\n for m in token_pattern.finditer(line):\\n tokens.append((m.group(), i + 1))\\n \\n func_brace_indices = {}\\n keywords = {'if', 'for', 'while', 'switch', 'catch', 'with'}\\n \\n for i, (tok, line_no) in enumerate(tokens):\\n if tok == '{':\\n is_func = False\\n func_start_line = line_no\\n \\n # 规则1: 往前找是否有 function 关键字\\n for j in range(i-1, max(-1, i-10), -1):\\n if tokens[j][0] == 'function':\\n is_func = True\\n func_start_line = tokens[j][1]\\n break\\n elif tokens[j][0] in [';', '{', '}']:\\n break\\n \\n if not is_func:\\n # 规则2: 往前找是否有箭头函数标识符 =>\\n for j in range(i-1, max(-1, i-5), -1):\\n if tokens[j][0] == '=>':\\n is_func = True\\n func_start_line = tokens[j][1]\\n # 尽量向前延伸到括号处作为起点\\n for k in range(j-1, max(-1, j-4), -1):\\n if tokens[k][0] == '(':\\n func_start_line = tokens[k][1]\\n break\\n break\\n elif tokens[j][0] in [';', '{', '}']:\\n break\\n \\n if not is_func:\\n # 规则3: 判断是否属于 method(args) { ... } 的形式（类方法或对象函数定义）\\n if i >= 4:\\n if tokens[i-1][0] == ')':\\n paren_depth = 1\\n name_idx = -1\\n for j in range(i-2, -1, -1):\\n if tokens[j][0] == ')':\\n paren_depth += 1\\n elif tokens[j][0] == '(':\\n paren_depth -= 1\\n if paren_depth == 0:\\n name_idx = j - 1\\n break\\n if name_idx >= 0:\\n name_tok = tokens[name_idx][0]\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', name_tok) and name_tok not in keywords:\\n is_func = True\\n func_start_line = tokens[name_idx][1]\\n \\n if is_func:\\n func_brace_indices[i] = func_start_line\\n \\n # 使用栈匹配闭合大括号，并计算已被标记为函数的块规模\\n stack = []\\n func_lines_counts = []\\n for i, (tok, line_no) in enumerate(tokens):\\n if tok == '{':\\n if i in func_brace_indices:\\n stack.append((func_brace_indices[i], True))\\n else:\\n stack.append((line_no, False))\\n elif tok == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n func_lines_counts.append(line_no - start_line + 1)\\n \\n return func_lines_counts\\n\\n def get_html_script_line_counts(code: str) -> list:\\n counts = []\\n # 提取HTML文件内的独立脚本部分做逻辑检测\\n scripts = re.finditer(r'<script\\\\b[^>]*>([\\\\s\\\\S]*?)</script>', code, re.IGNORECASE)\\n for m in scripts:\\n script_code = m.group(1)\\n counts.extend(get_js_functions_line_counts(script_code))\\n return counts\\n\\n # 检测开始\\n if not os.path.isdir(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 只针对可能的代码载体文件进行分析\\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n counts = []\\n if ext == '.py':\\n counts = get_py_functions_line_counts(code)\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n counts = get_js_functions_line_counts(code)\\n elif ext in ['.html', '.htm']:\\n counts = get_html_script_line_counts(code)\\n \\n for c in counts:\\n if c > 80:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一个整数是否为素数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 网页项目常见的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 读取文件时忽略部分解码错误，防止直接崩溃影响验证逻辑\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 包含空行的总行数\\n lines = content.split('\\\\n')\\n line_count = len(lines)\\n \\n # 如果存在文件的总行数是素数，则不满足要求\\n if is_prime(line_count):\\n return False\\n except Exception:\\n continue\\n \\n # 所有代码文件的行数均不是素数，满足要求\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n # 遍历工作区中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 检查 \\\"TODO\\\" 是否出现在代码文件中\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 如果由于某种原因文件无法正常读取，出于严格校验目的，返回False\\n return False\\n \\n # 如果有代码文件并且所有代码文件都通过了检查，返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"def check_indentation_4_spaces(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_fragments = []\\n \\n # 1. 从 response 中提取所有的 Markdown 代码块\\n pattern = re.compile(r'```[^\\\\n]*\\\\n(.*?)```', re.DOTALL)\\n code_fragments.extend(pattern.findall(response))\\n \\n # 2. 从工作区中读取常见格式的代码文件\\n extensions = ('.html', '.js', '.css', '.py', '.ts', '.jsx', '.tsx', '.json', '.vue', '.java', '.c', '.cpp', '.h')\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_fragments.append(f.read())\\n except Exception:\\n pass\\n \\n if not code_fragments:\\n return False\\n \\n valid_4 = 0\\n invalid_even = 0\\n has_tab = False\\n has_any_indent = False\\n \\n # 3. 逐块、逐行分析行首的缩进特征\\n for code in code_fragments:\\n lines = code.split('\\\\n')\\n for line in lines:\\n # 跳过空行\\n if not line.strip():\\n continue\\n \\n # 获取行首缩进（左侧空白部分）\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 如果缩进使用了制表符，直接判定违规\\n if '\\\\t' in leading_whitespace:\\n has_tab = True\\n break\\n \\n num_spaces = len(leading_whitespace)\\n if num_spaces > 0:\\n # 跳过部分块注释的对齐（行首有空白，且紧跟星号）\\n if line.lstrip().startswith('*'):\\n continue\\n \\n has_any_indent = True\\n if num_spaces % 4 == 0:\\n valid_4 += 1\\n elif num_spaces % 2 == 0:\\n invalid_even += 1\\n \\n if has_tab:\\n break\\n\\n if has_tab:\\n return False\\n \\n if not has_any_indent:\\n return False\\n \\n if valid_4 == 0:\\n return False\\n \\n # 4. 判断缩进主导性：允许存在少量前一轮留下的2空格历史代码文件，\\n # 但由于本轮的 js 逻辑及测试代码量应较大，4 空格规则行数应居于压倒性主导。\\n # 设定 invalid_even 不超过 valid_4 的 50%\\n if invalid_even > valid_4 * 0.5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"很好，方块能掉下来了。现在加入玩家控制和碰撞检测机制。我需要能用键盘操控方块：左/右方向键控制水平移动，下方向键控制加速下落，上方向键负责旋转方块。请务必做好严格的边界与碰撞检测：方块移动或旋转时不能超出左右侧墙壁；下落时碰到网格底部，或者碰到下方已经停止的方块后，必须立即锁定变成固定状态，然后马上在顶部生成下一个新方块。另外请你注意，从本轮开始后续所有轮次需遵循两项新要求：一是你输出的单元测试文件经运行，需确保目标代码获得至少90%的分支覆盖率；二是在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划。以上两项要求后续所有轮次均需遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def contains_you(text: str) -> bool:\\n # 过滤掉非第二人称代词但包含“你”的常见词汇以避免误判\\n text_to_check = text.replace(\\\"迷你\\\", \\\"\\\")\\n # 检查是否使用了第二人称代词“你”或“您”\\n return \\\"你\\\" in text_to_check or \\\"您\\\" in text_to_check\\n\\n # 1. 检查助手最后一轮的直接回复内容\\n if contains_you(response):\\n return False\\n \\n # 2. 检查助手生成的代码项目中的内容（包括代码注释、字符串输出等）\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if contains_you(content):\\n return False\\n except Exception:\\n # 忽略无法以 utf-8 文本格式读取的文件（例如二进制图片、音频文件等）\\n continue\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 如果回复中出现了“我”，说明没有完全用“代码小助手”代替“我”\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 去除回复开头的空白字符后，检查是否以“嗨～”开头\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除字符串末尾可能的空白字符（如换行符、空格等）\\n cleaned_response = response.rstrip()\\n \\n # 判断回复的末尾是否以“哦～”结尾\\n return cleaned_response.endswith(\\\"哦～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对可能生成的代码文件设定合理的后缀名集合\\n # 涵盖前端网页及其测试文件、Python、Markdown 文档以及配置文件\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.md', '.py', '.json'}\\n \\n if not os.path.exists(workspace_path):\\n # 如果工作区目录不存在，意味着没有创建任何文件\\n # 故没有违反\\\"创建的文件必须是UTF-8\\\"的规定\\n return True\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取该代码文件的内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略诸如权限错误等与编码无关的罕见异常情况\\n pass\\n \\n # 如果所有的代码文件都能成功以 utf-8 解码，则满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"def check_function_size(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n def check_python_functions(code):\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n # 若代码存在语法错误导致无法解析为AST树，暂不强制报错，跳过当前文件\\n return True\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, \\\"end_lineno\\\"):\\n # node.end_lineno 和 node.lineno 提供函数代码块的结束与起始行号\\n if (node.end_lineno - node.lineno + 1) > 80:\\n return False\\n return True\\n\\n def check_js_functions(code):\\n def replacer(match):\\n # 将匹配到的字符串和注释替换为等量的空格和换行符，从而绝对保证原始行号不变且不受干扰\\n s = match.group(0)\\n return '\\\\n' * s.count('\\\\n') + ' ' * (len(s) - s.count('\\\\n'))\\n\\n try:\\n # 清除单行注释、多行注释和常见的各类字符串内部结构的干扰\\n code = re.sub(r'//.*', replacer, code)\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', replacer, code)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', replacer, code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", replacer, code)\\n code = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', replacer, code)\\n except Exception:\\n pass\\n\\n line_no = 1\\n stack = []\\n i = 0\\n n = len(code)\\n \\n while i < n:\\n char = code[i]\\n if char == '\\\\n':\\n line_no += 1\\n elif char == '{':\\n j = i - 1\\n # 向前回溯，忽略空白和换行符\\n while j >= 0 and code[j] in ' \\\\t\\\\n\\\\r':\\n j -= 1\\n \\n is_func = False\\n if j >= 0:\\n if code[j] == '>': \\n # 匹配箭头函数如 () => { ... }\\n is_func = True\\n elif code[j] == ')':\\n # 匹配普通函数定义或类方法如 function name() { ... }\\n paren_count = 1\\n k = j - 1\\n while k >= 0:\\n if code[k] == ')':\\n paren_count += 1\\n elif code[k] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n k -= 1\\n if k >= 0:\\n p = k - 1\\n while p >= 0 and code[p] in ' \\\\t\\\\n\\\\r':\\n p -= 1\\n word = \\\"\\\"\\n while p >= 0 and re.match(r'[A-Za-z0-9_$]', code[p]):\\n word = code[p] + word\\n p -= 1\\n # 过滤诸如 if() {}, for() {} 等控制流关键字产生的代码块\\n if word not in ['if', 'for', 'while', 'catch', 'switch', 'with']:\\n is_func = True\\n else:\\n # 匹配诸如 function { ... } 等不带圆括号的极端边缘函数声明情况\\n p = j\\n word = \\\"\\\"\\n while p >= 0 and re.match(r'[A-Za-z0-9_$]', code[p]):\\n word = code[p] + word\\n p -= 1\\n if word == 'function':\\n is_func = True\\n \\n # 入栈并附带判定其是否是函数代码块的标识\\n stack.append((line_no, is_func))\\n elif char == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n # 对于确定是函数代码块的区间，检查总行数是否大于80\\n if (line_no - start_line + 1) > 80:\\n return False\\n i += 1\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前工作目录下所有代码项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[-1].lower()\\n \\n # 仅处理前端文件及后端Python文件的后缀种类\\n if ext not in ['.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 根据代码语言扩展采用不同策略进行函数行数统计\\n if ext == '.py':\\n if not check_python_functions(code):\\n return False\\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue', '.svelte']:\\n if not check_js_functions(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def check_code_lines_not_prime(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断给定的数字是否为素数\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n**0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.cs', \\n '.go', '.rb', '.php', '.swift', '.kt', '.json',\\n '.vue', '.svelte', '.sh', '.xml', '.yml', '.yaml'\\n }\\n\\n # 若工作区路径不存在，则无法获取代码文件，未违反该条件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历工作区目录读取所有生成的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理代码类型的文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取文件，忽略不可解码的异常字符\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # readlines() 会将文件中所有的行存为列表元素，即便只有换行符的空行也会被计算在内\\n lines_count = len(f.readlines())\\n \\n # 如果存在任何一个代码文件的总行数是素数，则未满足要求\\n if is_prime(lines_count):\\n return False\\n except Exception:\\n # 若因特殊原因不可读取该文件，则跳过继续检查其它文件\\n continue\\n\\n # 若所有代码文件行数均非素数，返回 True\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀名\\n code_extensions = {\\n '.py', '.js', '.html', '.css', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.cs', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', '.swift', '.kt', \\n '.rs', '.dart', '.sh'\\n }\\n \\n found_code_file = False\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 用户要求代码文件必须以 UTF-8 编码，以此编码读取\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 如果存在任何一个代码文件不包含 \\\"TODO\\\"，直接返回 False\\n if 'TODO' not in content:\\n return False\\n except Exception:\\n # 如果无法读取（例如编码不合规或文件损坏等情况），返回 False\\n return False\\n \\n # 如果遍历了所有文件都符合要求，且确实至少找到了一个代码文件，则返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_indentation(response: str, workspace_path: str) -> bool:\\n lines_to_check = []\\n \\n # 1. 提取 response 中的代码块\\n # 匹配形如 ```语言\\\\n代码\\\\n``` 的所有内容\\n code_blocks = re.findall(r'```([^\\\\n\\\\r]*)\\\\r?\\\\n([\\\\s\\\\S]*?)```', response)\\n for lang_info, block in code_blocks:\\n lang = lang_info.lower().strip()\\n # 忽略非代码类型的语言或配置片段\\n if lang in ['md', 'markdown', 'text', 'txt', 'bash', 'sh']:\\n continue\\n lines_to_check.extend(block.split('\\\\n'))\\n \\n # 2. 提取 workspace 中的代码文件内容\\n if os.path.exists(workspace_path):\\n # 根据指令内容涉及到的网页开发等常见格式列举源代码类型\\n valid_exts = {\\n '.py', '.js', '.css', '.html', '.java', '.cpp', '.c', '.h', '.hpp', \\n '.cs', '.ts', '.jsx', '.tsx', '.json', '.xml', '.vue', '.go', '.rs', \\n '.php', '.rb', '.swift', '.kt'\\n }\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_exts):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines_to_check.extend(f.readlines())\\n except Exception:\\n pass\\n \\n # 3. 分类统计各类缩进\\n tab_count = 0\\n mod4_0_count = 0\\n mod4_2_count = 0\\n other_count = 0\\n\\n for line in lines_to_check:\\n # 忽略只包含空白的空行\\n if not line.strip():\\n continue\\n \\n # 排除多行注释引起的非正常缩进行（如 JSDoc 等以 * 开头的行）\\n if line.lstrip().startswith('*'):\\n continue\\n \\n # 提取行首的空白字符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if not match:\\n continue\\n \\n leading_space = match.group(1)\\n \\n # 记录含制表符的行\\n if '\\\\t' in leading_space:\\n tab_count += 1\\n else:\\n # 记录全部由空格组成的行长度类型\\n l = len(leading_space)\\n if l % 4 == 0:\\n mod4_0_count += 1\\n elif l % 2 == 0:\\n mod4_2_count += 1\\n else:\\n other_count += 1\\n \\n total_spaces = mod4_0_count + mod4_2_count + other_count\\n \\n # 无任何缩进行，不违背要求\\n if total_spaces == 0:\\n return tab_count == 0\\n \\n # 判断 1: 检查是否混用了 tab 键缩进\\n if tab_count > max(3, 0.05 * (total_spaces + tab_count)):\\n return False\\n \\n # 判断 2: 在产生了一定代码基数的情况下，确保以 4 为倍数的缩进行应占绝大多数地位\\n if total_spaces > 5:\\n if mod4_0_count < mod4_2_count or mod4_0_count < other_count:\\n return False\\n \\n # 判断 3: 严格控制产生 2 个（非4个）空格层级缩进出现的比例\\n # 当 2的倍数(非4)的行数明显超过合理排版误差时，说明未统一以 4 个空格作为缩进标准\\n if mod4_2_count > mod4_0_count * 0.4 and mod4_2_count > 3:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少90%的分支覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在执行任何代码修改动作之前，要求你必先于task_plan.md记录修改计划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"核心操作成型了，现在来完善游戏的消除规则和计分系统。当底部固定的方块填满完整的一行时，将该行整体消除，并把其上方所有的方块整体向下平移一行。在右侧的预留面板中加上一个“得分板”，一次消除1行、2行、3行、4行分别奖励10、30、60、100分。另外加上失败判定：如果新生成的方块因为下方空间被占满而无法完全出现，游戏结束，弹出一个带有‘GAME OVER’发光字体的半透明遮罩层。\\n另外请你注意，从本轮开始后续所有轮次需遵循两项更新要求：一是原有代码修改前在task_plan.md记录修改计划的规则调整为每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案；二是必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令。以上两项要求后续所有轮次均需遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 排除一些常见但并非用于指代用户的含\\\"你\\\"词汇，避免误判\\n # 目前中文中常见的非代词情况主要为音译词如\\\"迷你\\\"\\n cleaned_response = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 检查助手的回复中是否仍然包含字符\\\"你\\\"\\n # 如果包含，说明没有彻底使用\\\"阁下\\\"代替\\\"你\\\"，不满足要求\\n if \\\"你\\\" in cleaned_response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否出现了“我”字\\n # 只要出现了“我”，即说明未完全遵循用“代码小助手”代替“我”的规则\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复是否以\\\"嗨～\\\"开头\\n \\\"\\\"\\\"\\n return response.lstrip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 第一步：去除末尾的常规空白字符并进行严格匹配\\n response = response.strip()\\n if response.endswith(\\\"哦～\\\") or response.endswith(\\\"哦~\\\"):\\n return True\\n \\n # 第二步：考虑模型可能会在末尾加上句号、感叹号等标点符号，去除这些常见标点后再做匹配\\n # 注意不使用 string.punctuation 避免误删部分字符（如半角波浪号 ~ 被误删的情况）\\n trailing_chars = \\\" \\\\t\\\\n\\\\r。！？.!?,，\\\\\\\"'”“\\\"\\n cleaned_response = response.rstrip(trailing_chars)\\n \\n if cleaned_response.endswith(\\\"哦～\\\") or cleaned_response.endswith(\\\"哦~\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据项目推断可能的代码与配置文件后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.py', \\n '.json', '.md', '.txt', '.sh', \\n '.yaml', '.yml'\\n }\\n \\n has_code_file = False\\n\\n # 遍历工作目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 如果是代码文件或相关的文本配置文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试以 UTF-8 编码打开并读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果读取时抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的异常（如权限问题等），但原则上不应阻断验证\\n pass\\n\\n # 只要存在至少一个代码文件且全都验证通过（未触发错误），即视为满足要求\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n def mask_html_non_script(code: str) -> str:\\n \\\"\\\"\\\"\\n 保留 <script> ... </script> 内部的代码。\\n 将其他 HTML 内容视为空白替换，以便在保持正确行号的前提下给 JS 解析器处理。\\n \\\"\\\"\\\"\\n result = []\\n in_script = False\\n script_start = re.compile(r'<script\\\\b[^>]*>', re.IGNORECASE)\\n script_end = re.compile(r'</script>', re.IGNORECASE)\\n pos = 0\\n while pos < len(code):\\n if not in_script:\\n match = script_start.search(code, pos)\\n if match:\\n text_before = code[pos:match.end()]\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text_before]))\\n pos = match.end()\\n in_script = True\\n else:\\n text_before = code[pos:]\\n result.append(''.join(['\\\\n' if c == '\\\\n' else ' ' for c in text_before]))\\n break\\n else:\\n match = script_end.search(code, pos)\\n if match:\\n result.append(code[pos:match.start()])\\n # 将 </script> 替换为空白，防止对大括号检测等造成影响\\n result.append(' ' * (match.end() - match.start()))\\n pos = match.end()\\n in_script = False\\n else:\\n result.append(code[pos:])\\n break\\n return ''.join(result)\\n\\n def check_python_functions(code: str) -> bool:\\n \\\"\\\"\\\"\\n 使用内置 AST 库获取 Python 文件内的所有函数/协程并计算其行数。\\n \\\"\\\"\\\"\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, \\\"end_lineno\\\") and hasattr(node, \\\"lineno\\\"):\\n lines = node.end_lineno - node.lineno + 1\\n if lines > 80:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n def check_js_functions(code: str) -> bool:\\n \\\"\\\"\\\"\\n 使用简易且健壮的词法匹配方法计算 JS/TS 中函数的行数。\\n 排除注释与字符串，计算每个推断为函数的大括号 { ... } 区间的行数。\\n \\\"\\\"\\\"\\n i = 0\\n n = len(code)\\n line_num = 1\\n \\n in_str_single = False\\n in_str_double = False\\n in_template = False\\n in_line_comm = False\\n in_block_comm = False\\n \\n stack = [] \\n \\n while i < n:\\n c = code[i]\\n \\n if c == '\\\\n':\\n line_num += 1\\n if in_line_comm:\\n in_line_comm = False\\n i += 1\\n continue\\n \\n if in_line_comm:\\n i += 1\\n continue\\n \\n if in_block_comm:\\n if c == '*' and i + 1 < n and code[i + 1] == '/':\\n in_block_comm = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n if in_str_single:\\n if c == '\\\\\\\\':\\n i += 2\\n elif c == \\\"'\\\":\\n in_str_single = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n if in_str_double:\\n if c == '\\\\\\\\':\\n i += 2\\n elif c == '\\\"':\\n in_str_double = False\\n i += 1\\n else:\\n i += 1\\n continue\\n \\n if in_template:\\n if c == '\\\\\\\\':\\n i += 2\\n elif c == '`':\\n in_template = False\\n i += 1\\n elif c == '$' and i + 1 < n and code[i + 1] == '{':\\n stack.append(('${', line_num, False))\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n if c == '/' and i + 1 < n and code[i + 1] == '/':\\n in_line_comm = True\\n i += 2\\n continue\\n if c == '/' and i + 1 < n and code[i + 1] == '*':\\n in_block_comm = True\\n i += 2\\n continue\\n if c == \\\"'\\\":\\n in_str_single = True\\n i += 1\\n continue\\n if c == '\\\"':\\n in_str_double = True\\n i += 1\\n continue\\n if c == '`':\\n in_template = True\\n i += 1\\n continue\\n \\n if c == '{':\\n is_func = False\\n j = i - 1\\n while j >= 0 and code[j] in ' \\\\t\\\\r\\\\n':\\n j -= 1\\n \\n if j >= 0:\\n if code[j] == ')':\\n # 向回寻找匹配的左括号 '('，以判断它前面的标识符是否为普通控制流语句\\n paren_count = 0\\n k = j\\n while k >= 0:\\n if code[k] == ')':\\n paren_count += 1\\n elif code[k] == '(':\\n paren_count -= 1\\n if paren_count == 0:\\n break\\n k -= 1\\n \\n if k >= 0:\\n k -= 1\\n while k >= 0 and code[k] in ' \\\\t\\\\r\\\\n':\\n k -= 1\\n word = \\\"\\\"\\n while k >= 0 and (code[k].isalnum() or code[k] == '_' or code[k] == '$'):\\n word = code[k] + word\\n k -= 1\\n # 过滤由于 for() {}, if() {} 等非函数语法引发的匹配\\n if word not in ('if', 'for', 'while', 'catch', 'switch'):\\n is_func = True\\n \\n elif code[j] == '>': \\n # 可能是由于箭头函数引发的匹配： => {\\n if j - 1 >= 0 and code[j - 1] == '=':\\n is_func = True\\n \\n elif code[j].isalpha() or code[j] == '_' or code[j] == '$':\\n # 检查前方关键词是否指定为函数类别的声明\\n word = \\\"\\\"\\n k = j\\n while k >= 0 and (code[k].isalnum() or code[k] == '_' or code[k] == '$'):\\n word = code[k] + word\\n k -= 1\\n if word in ('function', 'constructor', 'get', 'set'):\\n is_func = True\\n \\n stack.append(('{', line_num, is_func))\\n i += 1\\n continue\\n \\n if c == '}':\\n if stack:\\n top = stack.pop()\\n if top[0] in ('{', '${'):\\n start_line = top[1]\\n is_func = top[2]\\n end_line = line_num\\n # 检查被断定为函数的块跨越的行数\\n if is_func:\\n if end_line - start_line + 1 > 80:\\n return False\\n i += 1\\n continue\\n \\n i += 1\\n\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_python_functions(code):\\n return False\\n elif ext in ('.js', '.ts', '.jsx', '.tsx'):\\n if not check_js_functions(code):\\n return False\\n elif ext in ('.html', '.htm', '.vue'):\\n masked_code = mask_html_non_script(code)\\n if not check_js_functions(masked_code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n \\\"\\\"\\\"判定一个整数是否为素数\\\"\\\"\\\"\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 结合用户指令，推断出本项目中可能产生的核心代码、测试用例及计划配置文件的后缀\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.py'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉第三方依赖、版本控制等无关文件夹\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', '__pycache__']]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用 split('\\\\n') 能准确计算包含空行的总行数\\n line_count = len(content.split('\\\\n'))\\n \\n # 如果发现总行数为素数，则不满足要求\\n if is_prime(line_count):\\n return False\\n except Exception:\\n # 忽略读取异常的文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目推断可能的代码文件扩展名\\n code_extensions = {'.py', '.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.mjs', '.cjs'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 遍历项目目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取文件内容，忽略无法解码的部分以防异常\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 检查内容中是否出现 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 若文件由于异常无法读取，从严格角度按不满足条件处理\\n return False\\n \\n # 如果项目中存在代码文件且它们全部都包含 \\\"TODO\\\"，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n code_texts = []\\n \\n # 1. 提取 response 中的代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n code_texts.extend(code_blocks)\\n \\n # 2. 提取 workspace_path 中的实际业务代码文件\\n if os.path.exists(workspace_path):\\n # 限定常见代码文件类型，排除 .json .md .txt 以防默认格式化带来的干扰\\n valid_extensions = {'.js', '.html', '.css', '.py', '.java', '.c', '.cpp', '.ts', '.jsx', '.tsx', '.mjs', '.cjs', '.vue'}\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_texts.append(f.read())\\n except Exception:\\n pass\\n \\n if not code_texts:\\n return True\\n \\n total_4_multiples = 0\\n total_not_4_multiples = 0\\n total_tab = 0\\n \\n # 3. 分析每一行代码的前导缩进特征\\n for text in code_texts:\\n lines = text.split('\\\\n')\\n for line in lines:\\n if not line.strip():\\n continue\\n \\n lstripped = line.lstrip()\\n # 过滤掉常见的多行注释或其他可能产生错位对齐的特定格式标识\\n if lstripped.startswith('*') or lstripped.startswith('<!--') or lstripped.startswith('-->'):\\n continue\\n \\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent = match.group(1)\\n if '\\\\t' in indent:\\n total_tab += 1\\n else:\\n length = len(indent)\\n if length % 4 == 0:\\n total_4_multiples += 1\\n else:\\n total_not_4_multiples += 1\\n \\n # 如果含有明显的使用 Tab 键进行缩进的行\\n if total_tab > 2:\\n return False\\n \\n # 如果没有任何缩进，认为是符合的（不存在违规）\\n if total_4_multiples == 0 and total_not_4_multiples == 0:\\n return True\\n \\n # 如果不存在任何合法的 4 空格倍数缩进，但存在其他的缩进\\n if total_4_multiples == 0 and total_not_4_multiples > 0:\\n return False\\n \\n # 如果非 4 倍数的缩进行数过多，显著超过因函数参数对齐等情况产生的正常阈值（30% 或极小常数 5）\\n if total_not_4_multiples > max(5, total_4_multiples * 0.3):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少90%的分支覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"接下来丰富侧边栏面板的UI并优化体验。首先，在侧边栏加入“下一个方块预览”区域，让玩家能提前看到即将出现的形状。这需要你把随机生成逻辑往前推进一个身位。其次，增加“暂停”功能，按键盘的‘P’键可以暂停或继续游戏，暂停时要在主屏幕上浮现“PAUSED”字样。为了强化炫酷感，给所有活动和固定的方块加上外发光（box-shadow 或 shadowBlur）效果，让它们像霓虹灯管一样。\\n另外请你注意，从本轮开始后续所有轮次需遵循两项更新要求：一是你输出的单元测试文件经运行，需确保目标代码获得至少95%的分支覆盖率；二是在你输出的每一个代码文件中，务必保证所有的注释均为中文。以上两项要求后续所有轮次均需遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证助手是否在回复中始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方。\\n 核心是检查回复中是否违规使用了“你”字。\\n \\\"\\\"\\\"\\n # 过滤掉常见的非第二人称代词但包含“你”字的词汇（如“迷你”），防止误判\\n clean_response = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 检查助手的回复中是否仍然包含“你”或“妳”字\\n if \\\"你\\\" in clean_response or \\\"妳\\\" in clean_response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 剔除因复述规则等原因合理出现的“自我”一词\\n text_to_check = response.replace(\\\"自我\\\", \\\"\\\")\\n \\n # 如果处理后的文本中依然包含“我”字，则说明违反了禁止使用“我”进行自我指代的规则\\n if \\\"我\\\" in text_to_check:\\n return False\\n \\n # 如果不存在“我”字，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否以语气词\\\"嗨～\\\"打头。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，代表该要求的满足情况。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 移除首尾可能的空白字符后，判断是否以\\\"嗨～\\\"开头\\n return response.strip().startswith(\\\"嗨～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check_response_ending(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复的最末尾是否加上了语气词\\\"哦～\\\"。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除字符串首尾的空白字符（包括空格、换行符等）\\n cleaned_response = response.strip()\\n \\n # 判断清理后的字符串是否以\\\"哦～\\\"结尾\\n return cleaned_response.endswith(\\\"哦～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能的代码文件拓展名\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.py', '.ts'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码错误，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的读取异常\\n pass\\n \\n # 如果遍历检查完了所有代码文件且没有触发任何 UTF-8 解码错误，则判定为满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"import os\\nimport ast\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n def is_function_start(clean_code, index):\\n \\\"\\\"\\\"向前扫描，判断给定 '{' 是否属于函数/方法定义的起始\\\"\\\"\\\"\\n i = index - 1\\n while i >= 0 and clean_code[i].isspace():\\n i -= 1\\n if i < 0: return False, None\\n \\n # 判断箭头函数\\n if i >= 1 and clean_code[i-1:i+1] == '=>':\\n temp = i - 2\\n while temp >= 0 and clean_code[temp].isspace():\\n temp -= 1\\n if temp >= 0 and clean_code[temp] == ')':\\n p_level = 0\\n while temp >= 0:\\n if clean_code[temp] == ')': p_level += 1\\n elif clean_code[temp] == '(':\\n p_level -= 1\\n if p_level == 0: break\\n temp -= 1\\n return True, max(0, temp)\\n else:\\n while temp >= 0 and (clean_code[temp].isalnum() or clean_code[temp] in ['$', '_']):\\n temp -= 1\\n return True, temp + 1\\n \\n # 寻找最近的 ')'\\n j = i\\n found_paren = False\\n while j >= 0:\\n char = clean_code[j]\\n # 如果在找到 ')' 之前遇到了代表语句中断或对象赋值的特殊符号\\n if char in {';', '{', '}', '=', ','}:\\n break\\n if char == ')':\\n found_paren = True\\n break\\n j -= 1\\n \\n if not found_paren:\\n return False, None\\n \\n # 跳过已匹配的 '( ... )'\\n paren_level = 0\\n while j >= 0:\\n if clean_code[j] == ')': paren_level += 1\\n elif clean_code[j] == '(':\\n paren_level -= 1\\n if paren_level == 0: break\\n j -= 1\\n \\n if j < 0: return False, None\\n \\n j -= 1\\n while j >= 0 and clean_code[j].isspace():\\n j -= 1\\n \\n # 跳过可能存在的泛型如 <T>\\n if j >= 0 and clean_code[j] == '>':\\n a_level = 0\\n while j >= 0:\\n if clean_code[j] == '>': a_level += 1\\n elif clean_code[j] == '<':\\n a_level -= 1\\n if a_level == 0: break\\n j -= 1\\n j -= 1\\n while j >= 0 and clean_code[j].isspace():\\n j -= 1\\n \\n # 提取前面的词\\n word = \\\"\\\"\\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in ['$', '_']):\\n word = clean_code[j] + word\\n j -= 1\\n \\n if word == \\\"function\\\":\\n return True, j + 1\\n elif word and word not in {'if', 'for', 'while', 'catch', 'switch', 'with'}:\\n orig_j = j\\n while j >= 0 and clean_code[j].isspace(): j -= 1\\n prev_word = \\\"\\\"\\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in ['$', '_']):\\n prev_word = clean_code[j] + prev_word\\n j -= 1\\n if prev_word in {\\\"function\\\", \\\"get\\\", \\\"set\\\"}:\\n return True, j + 1\\n return True, orig_j + 1\\n elif word == \\\"\\\":\\n while j >= 0 and clean_code[j].isspace(): j -= 1\\n prev_word = \\\"\\\"\\n while j >= 0 and (clean_code[j].isalnum() or clean_code[j] in ['$', '_']):\\n prev_word = clean_code[j] + prev_word\\n j -= 1\\n if prev_word == \\\"function\\\":\\n return True, j + 1\\n \\n return False, None\\n\\n def find_matching_brace(clean_code, start_index):\\n level = 0\\n for i in range(start_index, len(clean_code)):\\n if clean_code[i] == '{':\\n level += 1\\n elif clean_code[i] == '}':\\n level -= 1\\n if level == 0:\\n return i\\n return -1\\n\\n def check_js_html_file(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n return True\\n \\n clean_code_list = list(code)\\n i = 0\\n in_single_comment = False\\n in_multi_comment = False\\n in_string = None\\n \\n # 将注释和字符串全部打码为空格，防止误判\\n while i < len(code):\\n char = code[i]\\n \\n if not in_single_comment and not in_multi_comment and in_string is None:\\n if char == '/' and i + 1 < len(code) and code[i+1] == '/':\\n in_single_comment = True\\n clean_code_list[i] = ' '\\n clean_code_list[i+1] = ' '\\n i += 1\\n elif char == '/' and i + 1 < len(code) and code[i+1] == '*':\\n in_multi_comment = True\\n clean_code_list[i] = ' '\\n clean_code_list[i+1] = ' '\\n i += 1\\n elif char in [\\\"'\\\", '\\\"', '`']:\\n in_string = char\\n clean_code_list[i] = ' '\\n elif in_single_comment:\\n if char == '\\\\n':\\n in_single_comment = False\\n else:\\n clean_code_list[i] = ' '\\n elif in_multi_comment:\\n if char == '\\\\n':\\n pass\\n elif char == '*' and i + 1 < len(code) and code[i+1] == '/':\\n in_multi_comment = False\\n clean_code_list[i] = ' '\\n clean_code_list[i+1] = ' '\\n i += 1\\n else:\\n clean_code_list[i] = ' '\\n elif in_string is not None:\\n if char == '\\\\\\\\':\\n clean_code_list[i] = ' '\\n if i + 1 < len(code):\\n if code[i+1] != '\\\\n':\\n clean_code_list[i+1] = ' '\\n i += 1\\n elif char == in_string:\\n in_string = None\\n clean_code_list[i] = ' '\\n elif char != '\\\\n':\\n clean_code_list[i] = ' '\\n \\n i += 1\\n \\n clean_code = \\\"\\\".join(clean_code_list)\\n \\n # 寻找花括号检查函数行数\\n for i, char in enumerate(clean_code):\\n if char == '{':\\n is_func, start_index = is_function_start(clean_code, i)\\n if is_func and start_index is not None:\\n end_index = find_matching_brace(clean_code, i)\\n if end_index != -1:\\n start_line = code.count('\\\\n', 0, start_index) + 1\\n end_line = code.count('\\\\n', 0, end_index) + 1\\n if end_line - start_line + 1 > 80:\\n return False\\n return True\\n\\n def check_python_file(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n return True\\n \\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return True\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n start = node.lineno\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n end = node.end_lineno\\n if end - start + 1 > 80:\\n return False\\n return True\\n\\n has_files = False\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n has_files = True\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n if ext == '.py':\\n if not check_python_file(filepath):\\n return False\\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}:\\n if not check_js_html_file(filepath):\\n return False\\n \\n # 如果项目压根没有任何文件，判断为不满足\\n if not has_files:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n def is_prime(n: int) -> bool:\\n if n < 2:\\n return False\\n if n == 2:\\n return True\\n if n % 2 == 0:\\n return False\\n for i in range(3, int(n ** 0.5) + 1, 2):\\n if n % i == 0:\\n return False\\n return True\\n\\n # 包含前端网页代码、单元测试代码、修改计划等常见的格式后缀\\n code_exts = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.json', '.md'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除可能干扰行数统计的无关目录\\n for exclude_dir in ['node_modules', '.git']:\\n if exclude_dir in dirs:\\n dirs.remove(exclude_dir)\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n lines_count = len(lines)\\n \\n # 若发现任意一个代码文件的总行数为素数，则返回 False\\n if is_prime(lines_count):\\n return False\\n except Exception:\\n # 避免不可读或其他异常情况中断整个验证过程\\n continue\\n \\n # 所有检查过的代码文件的总行数都不是素数\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据网页版俄罗斯方块任务推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行校验\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 根据系统提示约束，代码文件应为UTF-8编码\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查是否包含 \\\"TODO\\\"\\n if \\\"TODO\\\" not in content:\\n return False\\n except Exception:\\n # 若因编码等原因无法读取文件内容，则无法验证是否包含TODO，返回False\\n return False\\n\\n # 必须至少有一个代码文件，且全部包含 TODO 才算满足条件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n from collections import Counter\\n \\n # 定义需要检查的常见源代码文件后缀\\n code_exts = {'.js', '.html', '.css', '.py', '.ts', '.jsx', '.tsx'}\\n \\n def process_text(text: str) -> bool:\\n lines = text.split('\\\\n')\\n \\n in_block_comment = False\\n in_html_comment = False\\n indent_lengths = []\\n \\n for line in lines:\\n stripped = line.strip()\\n if not stripped:\\n continue\\n \\n # 简单追踪多行注释，避免统计多行注释内部的随意缩进\\n if '/*' in line and '*/' not in line:\\n in_block_comment = True\\n elif '*/' in line and in_block_comment:\\n in_block_comment = False\\n continue\\n \\n if '<!--' in line and '-->' not in line:\\n in_html_comment = True\\n elif '-->' in line and in_html_comment:\\n in_html_comment = False\\n continue\\n \\n if in_block_comment or in_html_comment:\\n continue\\n \\n # 忽略 JSDoc 或类似风格的多行注释延续行\\n if line.lstrip().startswith('*'):\\n continue\\n \\n # 获取前导空白符部分\\n leading_ws = line[:len(line) - len(line.lstrip())]\\n \\n # 若使用了制表符，则违背了“空格”缩进的标准\\n if '\\\\t' in leading_ws:\\n return False\\n \\n l = len(leading_ws)\\n if l > 0:\\n indent_lengths.append(l)\\n \\n # 如果没有带缩进的代码，视为未违规\\n if not indent_lengths:\\n return True\\n \\n counts = Counter(indent_lengths)\\n count_4 = sum(v for k, v in counts.items() if k % 4 == 0)\\n total_invalid = sum(v for k, v in counts.items() if k % 4 != 0)\\n \\n # 如果非4倍数的前导空格行数大于4倍数的行数，说明没有“统一”使用4个空格\\n # 此处的容错比例用于应对极少量由于多行字符串、参数特定对齐引起的非4倍数缩进\\n if total_invalid > count_4:\\n return False\\n \\n # 出现频次最高的基准缩进量必须是4的倍数（例如4或8）\\n most_common = counts.most_common(1)[0][0]\\n if most_common % 4 != 0:\\n return False\\n \\n return True\\n\\n # 1. 验证工作区内生成的所有代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not process_text(content):\\n return False\\n except Exception:\\n continue\\n \\n # 2. 验证回复中输出的 markdown 代码块\\n blocks = re.findall(r'```([a-zA-Z0-9]*)\\\\n(.*?)\\\\n```', response, re.DOTALL)\\n for lang, code in blocks:\\n # 仅针对代码类型的片段进行检查，排除 json、bash 或 txt 等不受该严格缩进要求影响的文本\\n if lang.lower() in ['javascript', 'js', 'html', 'css', 'python', 'py', 'typescript', 'ts', 'jsx', 'tsx', '']:\\n if not process_text(code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少95%的分支覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后一步，加入两个进阶的高级功能来大幅提升游戏手感。第一是加入“硬降（Hard Drop）”功能，按下空格键能让当前方块瞬间砸到底部并直接锁定。第二是加入“幽灵方块（Ghost Piece）”辅助线功能：在当前活动方块的正下方（即最终会落下的位置），实时绘制一个半透明或只有边框的同形状方块轮廓，帮助玩家精准瞄准。这样我们的暗黑炫酷版俄罗斯方块就大功告成了！\\n另外请你注意，从本轮开始后续所有轮次需遵循两项要求：一是原有单元测试覆盖率要求调整为运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在98%及以上水平；二是在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 30s` 包装，这样可以防范进程挂起。以上两项要求后续所有轮次均需遵循。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在与用户沟通的任何时候，必须始终坚持用\\\"阁下\\\"代替\\\"你\\\"来称呼对方\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查人工智能助手的回复中是否严格使用'阁下'代替了'你'。\\n 验证方式为检查回复文本中是否包含禁止使用的第二人称代词'你'或'您'。\\n \\\"\\\"\\\"\\n forbidden_words = [\\\"你\\\", \\\"您\\\"]\\n \\n for word in forbidden_words:\\n if word in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"进行自我指代时，请始终使用\\\"代码小助手\\\"代替\\\"我\\\"这一规则贯穿你的所有回复\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复中是否满足“使用'代码小助手'代替'我'”的条件。\\n 核心判断逻辑为：回复中不应该出现任何“我”字。\\n \\\"\\\"\\\"\\n # 如果回复中不包含“我”，则认为满足代替规则\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的每一条回复都必须使用语气词\\\"嗨～\\\"来打头\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 移除回复可能附带的前导空白字符（如空格、换行等），以确保判断的准确性\\n clean_response = response.lstrip()\\n \\n # 判断是否以\\\"嗨～\\\"打头\\n if clean_response.startswith(\\\"嗨～\\\"):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"哦～\\\"\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n # 移除字符串末尾的空白字符以及常见的标点符号（中英文句号、感叹号、问号、引号等）\\n chars_to_strip = ' \\\\t\\\\n\\\\r。！!？?.“”\\\\'\\\"’'\\n cleaned_response = response.rstrip(chars_to_strip)\\n \\n # 判断清理末尾标点后的字符串是否以\\\"哦～\\\"结尾\\n return cleaned_response.endswith('哦～')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"对于你创建的所有代码文件，规定必须以UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码文件和配置文件扩展名（推断项目可能使用的各类文件格式）\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.mjs', '.cjs', '.py', '.json', '.md', '.txt', '.xml', \\n '.yml', '.yaml', '.sh', '.bash', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.go', '.rs', '.rb', '.php'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略如 .git 等隐藏文件夹\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件，则检查其编码\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略权限被拒等其他不可控的读取错误\\n pass\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足：单个函数的规模不得大于 80 行\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n \\n def mask_comments_and_strings(code: str) -> str:\\n \\\"\\\"\\\"屏蔽代码中的字符串和注释，替换为空格，保留换行符。\\\"\\\"\\\"\\n res = []\\n i = 0\\n n = len(code)\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n while i < n:\\n if in_line_comment:\\n if code[i] == '\\\\n':\\n in_line_comment = False\\n res.append('\\\\n')\\n else:\\n res.append(' ')\\n i += 1\\n elif in_block_comment:\\n if i + 1 < n and code[i:i+2] == '*/':\\n in_block_comment = False\\n res.append(' ')\\n i += 2\\n elif code[i] == '\\\\n':\\n res.append('\\\\n')\\n i += 1\\n else:\\n res.append(' ')\\n i += 1\\n elif in_string:\\n if code[i] == string_char:\\n # 简易处理：不去过度解析模板字符串中的 ${} 嵌套，直接一并抹成空格，这对于判断大括号层级大多够用\\n in_string = False\\n res.append(' ')\\n i += 1\\n elif code[i] == '\\\\\\\\':\\n res.append(' ')\\n i += 2\\n elif code[i] == '\\\\n':\\n res.append('\\\\n')\\n i += 1\\n else:\\n res.append(' ')\\n i += 1\\n else:\\n if i + 1 < n and code[i:i+2] == '//':\\n in_line_comment = True\\n res.append(' ')\\n i += 2\\n elif i + 1 < n and code[i:i+2] == '/*':\\n in_block_comment = True\\n res.append(' ')\\n i += 2\\n elif code[i] in ['\\\"', \\\"'\\\", '`']:\\n in_string = True\\n string_char = code[i]\\n res.append(' ')\\n i += 1\\n else:\\n res.append(code[i])\\n i += 1\\n return \\\"\\\".join(res)\\n\\n def extract_scripts_from_html_keep_lines(html_code: str) -> str:\\n \\\"\\\"\\\"提取HTML中的script标签内容，其余部分用换行符补齐，以保持原来的行号。\\\"\\\"\\\"\\n res = []\\n last_end = 0\\n pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.IGNORECASE | re.DOTALL)\\n for match in pattern.finditer(html_code):\\n start, end = match.span(1)\\n skipped = html_code[last_end:start]\\n res.append('\\\\n' * skipped.count('\\\\n'))\\n res.append(match.group(1))\\n last_end = end\\n skipped = html_code[last_end:]\\n res.append('\\\\n' * skipped.count('\\\\n'))\\n return \\\"\\\".join(res)\\n\\n def get_js_functions_lengths(code: str) -> list:\\n \\\"\\\"\\\"从 JS/TS 代码中解析所有函数的行数\\\"\\\"\\\"\\n masked_code = mask_comments_and_strings(code)\\n functions_lengths = []\\n \\n # 定义可能表示函数开头的正则表达式\\n patterns = [\\n re.compile(r'\\\\bfunction\\\\b\\\\s*[\\\\w\\\\*\\\\$]*\\\\s*\\\\([^)]*\\\\)\\\\s*$'), # 普通/生成器 function\\n re.compile(r'(?:\\\\w+)?\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\=\\\\>\\\\s*$'), # 箭头函数 (args) =>\\n re.compile(r'\\\\w+\\\\s*\\\\=\\\\>\\\\s*$'), # 箭头函数 arg =>\\n re.compile(r'\\\\b(?:get|set)\\\\s+\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*$'), # getter / setter\\n re.compile(r'\\\\b(?!(?:if|for|while|switch|catch|with)\\\\b)\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*$') # 对象/类方法\\n ]\\n \\n # 遍历所有找到的大括号\\n for match in re.finditer(r'\\\\{', masked_code):\\n start_index = match.start()\\n prefix = masked_code[:start_index].rstrip()\\n # 截取前 300 个字符进行快速匹配，避免匹配消耗过大\\n search_area = prefix[-300:]\\n \\n func_start_idx = -1\\n for p in patterns:\\n m = p.search(search_area)\\n if m:\\n # 获取该函数签名的实际起始索引\\n func_start_idx = len(prefix) - len(search_area) + m.start()\\n break\\n \\n # 若找到了函数的签名特征，使用栈寻找闭合大括号\\n if func_start_idx != -1:\\n stack = 1\\n i = start_index + 1\\n while i < len(masked_code) and stack > 0:\\n if masked_code[i] == '{':\\n stack += 1\\n elif masked_code[i] == '}':\\n stack -= 1\\n i += 1\\n \\n # 计算跨越行数\\n if stack == 0:\\n end_index = i - 1\\n start_line = code.count('\\\\n', 0, func_start_idx) + 1\\n end_line = code.count('\\\\n', 0, end_index) + 1\\n functions_lengths.append(end_line - start_line + 1)\\n \\n return functions_lengths\\n\\n # 遍历工作区中的文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略依赖库或版本控制相关文件夹\\n if 'node_modules' in root.replace('\\\\\\\\', '/') or '.git' in root.replace('\\\\\\\\', '/'):\\n continue\\n \\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 支持解析的主流代码文件后缀\\n supported_exts = ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.htm', '.vue']\\n if ext not in supported_exts:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n length = node.end_lineno - node.lineno + 1\\n if length > 80:\\n return False\\n except Exception:\\n pass\\n else:\\n if ext in ['.html', '.htm', '.vue']:\\n code_to_check = extract_scripts_from_html_keep_lines(code)\\n else:\\n code_to_check = code\\n \\n lengths = get_js_functions_lengths(code_to_check)\\n for length in lengths:\\n if length > 80:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"每一个由你生成的代码文件，其总行数（把空行计算在内）绝对不能是素数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_prime(n: int) -> bool:\\n if n <= 1:\\n return False\\n if n <= 3:\\n return True\\n if n % 2 == 0 or n % 3 == 0:\\n return False\\n i = 5\\n while i * i <= n:\\n if n % i == 0 or n % (i + 2) == 0:\\n return False\\n i += 6\\n return True\\n\\n # 包含前端项目和记录文件中常见的后缀类型\\n code_extensions = {'.html', '.css', '.js', '.json', '.md', '.ts', '.txt'}\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录与常见的第三方依赖包目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n line_count = len(lines)\\n \\n # 判断行数是否为素数，若为素数则不满足要求\\n if is_prime(line_count):\\n return False\\n except Exception:\\n # 读取出错可忽略，跳过当前文件继续处理\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在你输出的任何一个代码文件里，务必出现 \\\"TODO\\\"\",\n \"validation_code\": \"import os\\n\\ndef check_todo_in_code_files(response: str, workspace_path: str) -> bool:\\n # 结合用户指令推断，网页版游戏的代码文件后缀主要包含以下几种\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n code_files_found = False\\n \\n # 需要排除的文件夹，防止检查非助手生成的第三方依赖文件\\n exclude_dirs = {'.git', 'node_modules', 'venv', '.venv'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 原地修改 dirs 以过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n # 若有任意一个代码文件不包含 \\\"TODO\\\"，则直接返回不满足要求\\n if 'TODO' not in content:\\n return False\\n except Exception:\\n # 如果由于权限等极特殊原因无法读取，判定为未满足\\n return False\\n \\n # 如果找到了代码文件并且都没有触发上面的 return False，则要求被满足\\n # 如果没有找到任何代码文件，则说明没有满足生成代码的前提，返回 False\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n total_valid = 0\\n total_invalid = 0\\n \\n lines_to_check = []\\n \\n # 1. 从人工智能助手的回复中提取 Markdown 代码块的内容\\n pattern = re.compile(r'```[a-zA-Z0-9]*\\\\n(.*?)\\\\n```', re.DOTALL)\\n for match in pattern.finditer(response):\\n code_block = match.group(1)\\n lines_to_check.extend(code_block.split('\\\\n'))\\n \\n # 2. 从生成的工作区代码文件中提取内容\\n code_exts = {'.html', '.css', '.js', '.ts', '.py', '.json'}\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n lines_to_check.extend(f.read().splitlines())\\n except Exception:\\n pass\\n \\n # 3. 遍历并验证所有提取出的代码行\\n for line in lines_to_check:\\n original_len = len(line)\\n stripped_line = line.lstrip()\\n \\n # 忽略完全为空白符的空行\\n if not stripped_line:\\n continue\\n \\n # 排除多行注释和 HTML 标签为了视觉效果做出的不对齐情况\\n if stripped_line.startswith('*') or stripped_line.startswith('/*') or stripped_line.startswith('<!--'):\\n continue\\n \\n # 截取出前导空白字符串\\n leading_whitespace = line[:original_len - len(stripped_line)]\\n \\n # 如果不存在任何前导空格，表明该行没有缩进，不纳入检查统计\\n if not leading_whitespace:\\n continue\\n \\n # 判断缩进有效性：不允许包含 tab (\\\\t)，且纯空格长度必须是 4 的倍数\\n if '\\\\t' in leading_whitespace:\\n total_invalid += 1\\n elif len(leading_whitespace) % 4 != 0:\\n total_invalid += 1\\n else:\\n total_valid += 1\\n \\n # 如果完全没有出现任何缩进行，不认为违反规则\\n if total_valid + total_invalid == 0:\\n return True\\n \\n # 允许 15% 的容错空间，防止因少数代码行或字符串换行对齐造成的误判\\n valid_ratio = total_valid / (total_valid + total_invalid)\\n return valid_ratio >= 0.85\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少85%的行覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你输出的单元测试文件经运行，需确保目标代码获得至少95%的分支覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"运行你提供的单元测试文件后，目标代码的语句覆盖率必须保持在98%及以上水平\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次准备改动代码时，都必须提前在implementation_plan.json里写明此次的修改方案\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"必须尽最大努力选用专用工具（如Read, Write, Edit）进行文件操作，从而避免使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在你输出的每一个代码文件中，务必保证所有的注释均为中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"在运行任何 bash 测试、安装或构建命令时，强制要求在最前面添加 `timeout 30s` 包装，这样可以防范进程挂起\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["初始页面加载验证：打开页面，确认页面背景为纯黑色，中央显示一个带有网格线的暗黑风格游戏主网格区域，网格右侧预留了信息面板区域，页面底部显示一个大号的“开始游戏”按钮。","方块生成与外观验证：点击“开始游戏”按钮，观察网格顶部中心是否成功生成一个俄罗斯方块。确认方块颜色为高饱和度荧光色（如霓虹粉、青色、亮黄等），并带有明显的外发光（霓虹灯管）视觉效果。","自动下落功能验证：游戏开始且方块生成后，不进行任何键盘操作，观察并截图确认活动方块随时间自动匀速向下移动。","下一个方块预览功能验证：观察右侧的信息面板，确认存在“下一个方块预览”区域，并且该区域内正确显示了即将出现的方块形状（同样具有高饱和颜色和发光效果）。","幽灵方块（辅助线）验证：观察网格区域，确认在当前活动方块的正下方（即垂直落到底部的最终位置），实时显示着一个形状完全相同、半透明或仅有边框的“幽灵方块”辅助线。","水平移动与边界检测验证：按键盘左/右方向键，确认方块能够向左/右平移；将方块一直移动到网格边缘并继续按相同方向键，确认方块停留在边缘，严格没有穿透或越出左右侧墙壁。","加速下落验证：按住键盘下方向键，观察活动方块，确认其向下的下落速度比不按键时明显加快。","旋转与边界保护验证：按键盘上方向键，确认方块能够顺利改变方向（旋转）；将方块紧贴侧边缘并尝试旋转，确认方块旋转后不会越出左右边界墙壁。","硬降（Hard Drop）功能验证：按下键盘空格键，确认当前活动方块瞬间砸到底部位置（即幽灵方块所在位置）并立即变为锁定状态，同时网格顶部立刻生成出下一个新方块。","触底堆叠与状态锁定验证：让方块自然下落接触网格底部或下方已经停止的方块，确认其碰撞后立即停止下落变为固定状态并保持外发光效果，随后网格顶部正常生成新方块。","满行消除与平移逻辑验证：通过操作方块将网格底部的某一行完整填满，确认该行方块瞬间被消除，且其上方所有处于锁定状态的方块整体向下平移一行。","计分系统验证：观察右侧面板的“得分板”，在成功完整消除方块后，确认得分板上的分数对应增加。需验证一次性消除1行、2行、3行、4行时，分数是否分别增加10分、30分、60分和100分。","暂停功能验证：在游戏进行中按键盘的‘P’键，确认活动方块立即停止下落，同时在游戏主屏幕中央浮现出明确的“PAUSED”字样；再次按‘P’键，字样消失且游戏恢复正常下落。","游戏结束（GAME OVER）判定验证：故意将方块在某一列堆叠至网格最顶部，当新生成的方块因下方空间被占满无法完全出现时，确认游戏结束操作锁定，并在界面上弹出一个带有“GAME OVER”发光字体的半透明遮罩层。"],"string":"[\n \"初始页面加载验证：打开页面，确认页面背景为纯黑色，中央显示一个带有网格线的暗黑风格游戏主网格区域，网格右侧预留了信息面板区域，页面底部显示一个大号的“开始游戏”按钮。\",\n \"方块生成与外观验证：点击“开始游戏”按钮，观察网格顶部中心是否成功生成一个俄罗斯方块。确认方块颜色为高饱和度荧光色（如霓虹粉、青色、亮黄等），并带有明显的外发光（霓虹灯管）视觉效果。\",\n \"自动下落功能验证：游戏开始且方块生成后，不进行任何键盘操作，观察并截图确认活动方块随时间自动匀速向下移动。\",\n \"下一个方块预览功能验证：观察右侧的信息面板，确认存在“下一个方块预览”区域，并且该区域内正确显示了即将出现的方块形状（同样具有高饱和颜色和发光效果）。\",\n \"幽灵方块（辅助线）验证：观察网格区域，确认在当前活动方块的正下方（即垂直落到底部的最终位置），实时显示着一个形状完全相同、半透明或仅有边框的“幽灵方块”辅助线。\",\n \"水平移动与边界检测验证：按键盘左/右方向键，确认方块能够向左/右平移；将方块一直移动到网格边缘并继续按相同方向键，确认方块停留在边缘，严格没有穿透或越出左右侧墙壁。\",\n \"加速下落验证：按住键盘下方向键，观察活动方块，确认其向下的下落速度比不按键时明显加快。\",\n \"旋转与边界保护验证：按键盘上方向键，确认方块能够顺利改变方向（旋转）；将方块紧贴侧边缘并尝试旋转，确认方块旋转后不会越出左右边界墙壁。\",\n \"硬降（Hard Drop）功能验证：按下键盘空格键，确认当前活动方块瞬间砸到底部位置（即幽灵方块所在位置）并立即变为锁定状态，同时网格顶部立刻生成出下一个新方块。\",\n \"触底堆叠与状态锁定验证：让方块自然下落接触网格底部或下方已经停止的方块，确认其碰撞后立即停止下落变为固定状态并保持外发光效果，随后网格顶部正常生成新方块。\",\n \"满行消除与平移逻辑验证：通过操作方块将网格底部的某一行完整填满，确认该行方块瞬间被消除，且其上方所有处于锁定状态的方块整体向下平移一行。\",\n \"计分系统验证：观察右侧面板的“得分板”，在成功完整消除方块后，确认得分板上的分数对应增加。需验证一次性消除1行、2行、3行、4行时，分数是否分别增加10分、30分、60分和100分。\",\n \"暂停功能验证：在游戏进行中按键盘的‘P’键，确认活动方块立即停止下落，同时在游戏主屏幕中央浮现出明确的“PAUSED”字样；再次按‘P’键，字样消失且游戏恢复正常下落。\",\n \"游戏结束（GAME OVER）判定验证：故意将方块在某一列堆叠至网格最顶部，当新生成的方块因下方空间被占满无法完全出现时，确认游戏结束操作锁定，并在界面上弹出一个带有“GAME OVER”发光字体的半透明遮罩层。\"\n]","isConversational":false}}},{"rowIdx":82,"cells":{"task_id":{"kind":"number","value":83,"string":"83"},"repository_policy":{"kind":"string","value":"以下是你后续所有轮次回复都必须严格遵循的约束条件：\n1. 凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况。\n2. 在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称。\n3. 你对用户作出的任何回复，都需在最末尾加上语气词“汪～”。\n4. 无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码。\n5. 在你生成的所有代码文件中，注释内容必须统一使用中文编写。\n6. 你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n7. 一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。\n8. 在为各个代码文件编写注释时，切勿使用 emoji表情符号。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否符合“称呼用户不得出现以‘你’相称的情况”的要求。\n \"\"\"\n # 如果回复中出现了“你”字，则违反了“不得出现以‘你’相称的情况”的要求\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def verify_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复中是否全程以“小助手”作为第一人称，切勿使用“我”来自称。\n \"\"\"\n # 检查回复中是否包含“我”字\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除回复内容首尾的空白字符（包括换行符）\n clean_response = response.strip()\n \n # 使用正则表达式匹配末尾是否为“汪～”或“汪~”，允许其后跟随常见的结束标点符号\n pattern = r'汪[～~][。！？\\.\\!\\?]*$'\n \n # 如果正则匹配成功，说明在最末尾加上了该语气词\n if re.search(pattern, clean_response):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.java', '.cpp', '.c', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.html', '.css', '.sh', '.rb', '.go', '.rs', \n '.php', '.swift', '.kt', '.m', '.scala', '.dart'\n }\n \n # 如果工作区路径不存在，说明未生成任何代码，故没有违反规则\n if not os.path.exists(workspace_path):\n return True\n\n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline().strip()\n # 检查第一行是否包含要求中严格规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果遇到无法读取（如编码错误或权限问题）的情况，判定为未满足要求\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 预定义主流的代码文件后缀集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cxx', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', '.swift', \n '.kt', '.rs', '.m', '.mm', '.scala', '.sh', '.bash', '.html', '.css'\n }\n \n # 1. 验证 workspace_path 目录下的所有落盘代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 提取前10行代表“开头处”\n lines = [f.readline() for _ in range(10)]\n \n found = False\n for line in lines:\n if target_statement in line:\n found = True\n break\n \n # 若某代码文件开头不存在要求声明，则直接判否\n if not found:\n return False\n except Exception:\n pass\n \n # 2. 验证 response 中的 Markdown 代码块（防止生成在文本回复中）\n # 预定义主流代码语言标签\n lang_set = {\n 'python', 'py', 'java', 'c', 'cpp', 'c++', 'cxx', 'cs', 'c#',\n 'js', 'javascript', 'ts', 'typescript', 'go', 'ruby', 'rb', \n 'php', 'swift', 'kotlin', 'rust', 'rs', 'scala', 'sh', 'bash', \n 'html', 'css'\n }\n \n # 正则提取带有语言标识的 Markdown 代码块\n code_blocks = re.findall(r'```([a-zA-Z0-9+#\\-]+)?[ \\t]*\\n(.*?)```', response, re.DOTALL)\n for lang, block in code_blocks:\n if lang and lang.lower() in lang_set:\n lines = block.strip().split('\\n')[:10]\n found = False\n for line in lines:\n if target_statement in line:\n found = True\n break\n \n if not found:\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import tokenize\n import io\n\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.py', '.js', '.ts', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.go', '.rs', '.swift', '.php', '.rb', '.sh', '.bash', '.html', \n '.css', '.scss', '.xml', '.yml', '.yaml', '.sql'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 若不是常见的代码文件类型，则跳过\n if ext and ext not in code_extensions:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 若文件无法以utf-8读取，则跳过\n continue\n\n comments = []\n \n # 提取 Python 注释\n if ext == '.py':\n try:\n tokens = tokenize.generate_tokens(io.StringIO(content).readline)\n for tok_type, tok_string, _, _, _ in tokens:\n if tok_type == tokenize.COMMENT:\n comments.append(tok_string)\n elif tok_type == tokenize.STRING:\n # 捕获 Python 的 Docstring 作为注释检查\n if tok_string.startswith('\"\"\"') or tok_string.startswith(\"'''\"):\n comments.append(tok_string)\n except Exception:\n # tokenize失败时，回退到正则匹配\n comments.extend(re.findall(r'#.*', content))\n comments.extend(re.findall(r'\"\"\"[\\s\\S]*?\"\"\"', content))\n comments.extend(re.findall(r\"'''[\\s\\S]*?'''\", content))\n # 提取其他语言注释\n else:\n # 匹配 Python/Shell/Ruby 等单行注释\n comments.extend(re.findall(r'#.*', content))\n # 匹配 C/Java/JS 等单行注释，并避开 `http://` 等格式\n comments.extend(re.findall(r'(?<!:)\\/\\/.*', content))\n # 匹配 C/Java/JS/CSS 等多行注释\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 匹配 HTML/XML 注释\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n\n # 遍历检查提取到的所有注释内容\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]}],"string":"[\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否符合“称呼用户不得出现以‘你’相称的情况”的要求。\\n \\\"\\\"\\\"\\n # 如果回复中出现了“你”字，则违反了“不得出现以‘你’相称的情况”的要求\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def verify_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的回复中是否全程以“小助手”作为第一人称，切勿使用“我”来自称。\\n \\\"\\\"\\\"\\n # 检查回复中是否包含“我”字\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除回复内容首尾的空白字符（包括换行符）\\n clean_response = response.strip()\\n \\n # 使用正则表达式匹配末尾是否为“汪～”或“汪~”，允许其后跟随常见的结束标点符号\\n pattern = r'汪[～~][。！？\\\\.\\\\!\\\\?]*$'\\n \\n # 如果正则匹配成功，说明在最末尾加上了该语气词\\n if re.search(pattern, clean_response):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.java', '.cpp', '.c', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.html', '.css', '.sh', '.rb', '.go', '.rs', \\n '.php', '.swift', '.kt', '.m', '.scala', '.dart'\\n }\\n \\n # 如果工作区路径不存在，说明未生成任何代码，故没有违反规则\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行检查\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline().strip()\\n # 检查第一行是否包含要求中严格规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果遇到无法读取（如编码错误或权限问题）的情况，判定为未满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 预定义主流的代码文件后缀集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cxx', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.go', '.rb', '.php', '.swift', \\n '.kt', '.rs', '.m', '.mm', '.scala', '.sh', '.bash', '.html', '.css'\\n }\\n \\n # 1. 验证 workspace_path 目录下的所有落盘代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 提取前10行代表“开头处”\\n lines = [f.readline() for _ in range(10)]\\n \\n found = False\\n for line in lines:\\n if target_statement in line:\\n found = True\\n break\\n \\n # 若某代码文件开头不存在要求声明，则直接判否\\n if not found:\\n return False\\n except Exception:\\n pass\\n \\n # 2. 验证 response 中的 Markdown 代码块（防止生成在文本回复中）\\n # 预定义主流代码语言标签\\n lang_set = {\\n 'python', 'py', 'java', 'c', 'cpp', 'c++', 'cxx', 'cs', 'c#',\\n 'js', 'javascript', 'ts', 'typescript', 'go', 'ruby', 'rb', \\n 'php', 'swift', 'kotlin', 'rust', 'rs', 'scala', 'sh', 'bash', \\n 'html', 'css'\\n }\\n \\n # 正则提取带有语言标识的 Markdown 代码块\\n code_blocks = re.findall(r'```([a-zA-Z0-9+#\\\\-]+)?[ \\\\t]*\\\\n(.*?)```', response, re.DOTALL)\\n for lang, block in code_blocks:\\n if lang and lang.lower() in lang_set:\\n lines = block.strip().split('\\\\n')[:10]\\n found = False\\n for line in lines:\\n if target_statement in line:\\n found = True\\n break\\n \\n if not found:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import tokenize\\n import io\\n\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.py', '.js', '.ts', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.go', '.rs', '.swift', '.php', '.rb', '.sh', '.bash', '.html', \\n '.css', '.scss', '.xml', '.yml', '.yaml', '.sql'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 若不是常见的代码文件类型，则跳过\\n if ext and ext not in code_extensions:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 若文件无法以utf-8读取，则跳过\\n continue\\n\\n comments = []\\n \\n # 提取 Python 注释\\n if ext == '.py':\\n try:\\n tokens = tokenize.generate_tokens(io.StringIO(content).readline)\\n for tok_type, tok_string, _, _, _ in tokens:\\n if tok_type == tokenize.COMMENT:\\n comments.append(tok_string)\\n elif tok_type == tokenize.STRING:\\n # 捕获 Python 的 Docstring 作为注释检查\\n if tok_string.startswith('\\\"\\\"\\\"') or tok_string.startswith(\\\"'''\\\"):\\n comments.append(tok_string)\\n except Exception:\\n # tokenize失败时，回退到正则匹配\\n comments.extend(re.findall(r'#.*', content))\\n comments.extend(re.findall(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', content))\\n comments.extend(re.findall(r\\\"'''[\\\\s\\\\S]*?'''\\\", content))\\n # 提取其他语言注释\\n else:\\n # 匹配 Python/Shell/Ruby 等单行注释\\n comments.extend(re.findall(r'#.*', content))\\n # 匹配 C/Java/JS 等单行注释，并避开 `http://` 等格式\\n comments.extend(re.findall(r'(?<!:)\\\\/\\\\/.*', content))\\n # 匹配 C/Java/JS/CSS 等多行注释\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 匹配 HTML/XML 注释\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n\\n # 遍历检查提取到的所有注释内容\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"我想用HTML、CSS和纯JS开发一个简单的单机版记账APP，完全在浏览器本地运行。第一步，请帮我把基础骨架搭出来。要求适配手机端屏幕大小，页面主色调使用清新的绿色。顶部是一个绿色背景的统计卡片，显示固定的占位文本：“总支出：¥0.00”和“结余：¥0.00”。页面底部固定一个悬浮的绿色圆形“+”添加按钮。中间部分留白，作为未来展示账单列表的区域。这一轮只需做好静态UI和响应式布局，不需要实现任何记账逻辑。请你后续始终遵循以下要求：第一，代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）；第二，务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 如果回复中出现了“你”字，则说明违背了“不得出现以‘你’相称的情况”的约束\n if '你' in response:\n return False\n \n # 未出现“你”字，视为满足要求\n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否未使用“我”来自称。\n 通过直接判断字符串中是否包含“我”字来进行严格限制。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response or not response.strip():\n return False\n \n # 定义可能出现在句末的常见标点符号和空白字符\n # 注意：不包含波浪号 ~ 和 ～ 以防误删关键内容，也不包含尖括号 < 和 > 以防破坏闭合标签的匹配结构\n remove_chars = '。，！？；：“”‘’（）《》〈〉【】『』「」.!,?;\"\\'()[]{}\\n\\r\\t '\n \n cleaned_response = response\n \n # 循环去除末尾的干扰性字符和结构化闭合标签，露出语义上的有效末尾内容\n while True:\n prev = cleaned_response\n \n # 1. 去除末尾的标点符号和空白字符\n cleaned_response = cleaned_response.rstrip(remove_chars)\n \n # 2. 去除末尾可能存在的 XML/HTML 闭合标签（例如 </description>）\n cleaned_response = re.sub(r'</[^>]+>$', '', cleaned_response)\n \n # 3. 去除末尾可能存在的 Markdown 代码块闭合标记（例如 ``` 或 ```javascript）\n cleaned_response = re.sub(r'```[^`\\n]*$', '', cleaned_response)\n \n # 如果经过一轮剥离后字符串没有任何变化，说明已经处理到了纯文本的有效末尾，结束循环\n if prev == cleaned_response:\n break\n \n # 最终判断清理后的有效文本末尾是否是以“汪～”或“汪~”结尾\n return cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~')","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check_coding_declaration(response: str, workspace_path: str) -> bool:\n # 根据用户指令，预期生成的代码文件主要是HTML、CSS和JS\n target_extensions = {'.html', '.css', '.js'}\n file_found = False\n\n # 确保给定的项目地址是一个有效的目录\n if not os.path.isdir(workspace_path):\n return False\n\n # 遍历当前项目文件夹下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查读取到的第一行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 若文件为空、无法读取或编码错误，统一视为不满足要求\n return False\n \n # 若至少存在一个代码文件且全都满足首行包含声明的要求，则返回True\n return file_found","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义需要检查的目标声明文字\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 根据用户指令推断，可能涉及的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js'}\n has_code_file = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历当前代码项目工作区\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果读取遇到编码或其他错误，视为不符合规范或无法验证\n return False\n \n statement_found = False\n # 检查文件开头处（取前15行以避免第一行的编码声明或其他doctype标签影响）\n for line in lines[:15]:\n if target_statement in line:\n statement_found = True\n break\n \n # 若存在任何一个代码文件不满足此开头声明，即返回False\n if not statement_found:\n return False\n\n # 要求必须有生成代码文件才算完成了\"搭出基础骨架\"的第一步\n return has_code_file","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区目录不存在，则无法验证文件，直接默认为True\n if not os.path.isdir(workspace_path):\n return True\n\n # 遍历当前工作区内所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选针对前端项目的核心代码文件类型\n if not file.lower().endswith(('.html', '.css', '.js')):\n continue\n\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n comments = []\n \n # 1. 提取 HTML 注释 <!-- ... -->，支持跨行\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\n \n # 2. 提取 CSS/JS 块级注释 /* ... */，支持跨行\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, flags=re.DOTALL))\n \n # 3. 提取 JS 行级注释 // ...，使用负向后视断言排除 http:// 和 https:// 中的双斜杠\n comments.extend(re.findall(r'(?<!:)//(.*)', content))\n\n # 遍历所有提取到的注释，逐个字符判断是否存在 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":1,"instruction":"不错，页面看起来很清爽。接下来完善底部“+”按钮的交互。点击它后，要从底部滑动弹出一个“添加账单”的面板。面板里需要包含一个表单：顶部是“支出”和“收入”的单选切换标签；然后是分类选择（下拉框或排列的图标，比如餐饮、交通、购物、工资等）；一个金额输入框（需限制只能输入数字和小数点）；一个日期选择器（默认当天）；以及备注输入框。底部放置“保存”和“取消”按钮。点击取消或者面板外部可以关闭该面板。请用原生JS完成这些UI的交互效果。另外请你后续始终遵循以下更新后的要求：一是对项目所做的每一次改动，都需要通过更新progress.md文件来记录本轮完成的工作；二是必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率，替换此前的并行执行相关要求。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复内容中是否包含“你”字\n if '你' in response:\n return False\n \n # 也可以扩展检查“您”字，以确保没有使用其他第二人称代词，\n # 但根据严格的字面要求，至少不能包含“你”\n if '您' in response:\n return False\n\n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def verify_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n # 去除多行代码块，避免代码中合理的“我”字干扰判断\n text_without_multiline_code = re.sub(r'```[\\s\\S]*?```', '', response)\n \n # 去除行内代码块\n pure_text = re.sub(r'`[^`]*`', '', text_without_multiline_code)\n \n # 检查纯文本回复中是否使用了“我”\n if '我' in pure_text:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符\n text = response.strip()\n \n # 匹配末尾是否为“汪～”或“汪~”，允许后面跟随任意数量的非单词字符（如标点符号）\n match = re.search(r'汪[～~]\\W*$', text)\n \n return bool(match)","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历项目路径下所有的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 读取文件第一行并判断\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 判断给定的编码声明字符串是否包含在第一行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于编码或其他读取错误，视为不符合要求\n return False\n \n # 如果找到了至少一个代码文件且全部满足要求，返回True；否则返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 依据用户指令推断代码文件类型为 html, css, js\n code_extensions = {\".html\", \".css\", \".js\"}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n # 检查开头处（前10行）是否包含该声明\n found_statement = False\n for line in lines[:10]:\n if target_statement in line:\n found_statement = True\n break\n \n if not found_statement:\n return False\n except Exception:\n # 如果由于某种原因无法读取某代码文件，视为验证失败\n return False\n\n return has_code_file","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 编译正则表达式以匹配各种注释形式\n # 1. HTML注释: <!-- 匹配内容 -->\n html_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\n # 2. 块注释 (CSS和JS): /* 匹配内容 */\n block_pattern = re.compile(r'/\\*(.*?)\\*/', re.DOTALL)\n # 3. 单行注释 (JS): // 匹配内容，并使用负向后视断言(?<!:)来避免匹配 http:// 或 https:// \n line_pattern = re.compile(r'(?<!:)\\/\\/(.*?)$', re.MULTILINE)\n \n # 遍历当前项目工作区中的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 根据用户指令，推断只检查 .html, .css, .js\n if ext in ['.html', '.css', '.js']:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n \n # 根据不同文件类型应用对应的正则表达式来提取注释内容\n if ext == '.html':\n comments.extend(html_pattern.findall(content))\n # HTML中可能包含内嵌的CSS和JS脚本，也同时提取相应注释\n comments.extend(block_pattern.findall(content))\n comments.extend(line_pattern.findall(content))\n elif ext == '.css':\n comments.extend(block_pattern.findall(content))\n elif ext == '.js':\n comments.extend(block_pattern.findall(content))\n comments.extend(line_pattern.findall(content))\n \n # 遍历所有提取到的注释文本\n for comment in comments:\n for char in comment:\n # 引用外部已经实现的 is_emoji 函数，判断是否有emoji\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新progress.md文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":2,"instruction":"很好，表单交互很顺滑。现在把核心的数据存储逻辑串起来。点击“保存”时，需要校验金额不能为0或空，验证通过后，将这条记录（类型、分类、金额、日期、备注）以JSON格式存入浏览器的 localStorage 中，并关闭面板。同时，在中部的空白区域根据 localStorage 的数据实时渲染出每一条账单。并且，需要计算所有数据的“收入”减去“支出”，实时更新顶部卡片里的“结余”和“总支出”金额。另外请你后续始终遵循以下要求：一是对项目所做的每一次改动，都需要通过更新 process-log.md 文件来记录本轮完成的工作，替换此前的进度记录相关要求；二是在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断一段文本中是否包含违规的第二人称称呼“你”或“您”\n def contains_second_person_pronoun(text: str) -> bool:\n if not text:\n return False\n # 将带有“你”但通常不作为人称代词使用的常见特例词汇剔除，避免产生误判\n clean_text = text.replace(\"迷你\", \"\").replace(\"迪士尼\", \"\")\n # 检查剔除后是否依然包含“你”或“您”\n if \"你\" in clean_text or \"您\" in clean_text:\n return True\n return False\n\n # 1. 检查 AI 助手的最后一轮回复文本中是否违规使用了“你”或“您”\n if contains_second_person_pronoun(response):\n return False\n\n # 2. 检查工作区生成的代码文件与进度日志文件是否违规\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n # 根据用户指令中提到的 Web 三剑客和进度日志推断目标文件类型\n target_extensions = ('.html', '.css', '.js', '.md', '.txt', '.json')\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if file_name.endswith(target_extensions):\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if contains_second_person_pronoun(content):\n return False\n except Exception:\n # 对于无法以 utf-8 读取或发生其他 IO 异常的文件，直接跳过\n continue\n\n # 如果所有被检查的内容里均未出现“你”或“您”，则认为满足了给定要求\n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"import re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否切勿使用“我”来自称。\n \"\"\"\n # 使用正则表达式去除 Markdown 格式的块级代码，避免代码内部的注释或字符串导致误判\n text_without_code_blocks = re.sub(r'```[\\s\\S]*?```', '', response)\n \n # 使用正则表达式去除 Markdown 格式的行内代码\n pure_text = re.sub(r'`[^`]+`', '', text_without_code_blocks)\n \n # 检查纯文本中是否包含“我”字\n if '我' in pure_text:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def check_response(response: str, workspace_path: str) -> bool:\n import re\n if not response:\n return False\n \n # 去除回复末尾可能包含的空白字符以及常见的中英文标点符号\n cleaned_response = re.sub(r'[\\s。，！、？\\.,!\\?]*$', '', response)\n \n # 判断剔除尾部干扰字符后，是否以“汪～”或“汪~”结尾\n return cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~')","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 筛选出代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行\n first_line = f.readline().strip()\n # 检查第一行是否包含指定声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件读取失败，也视为不符合要求\n return False\n \n # 如果项目中没有任何代码文件，则无法证明满足要求，返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n code_extensions = {\".html\", \".css\", \".js\"}\n has_code_file = False\n \n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n found_statement = False\n # 检查文件开头处（前10行以内）是否包含目标声明\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n if target_statement in line:\n found_statement = True\n break\n \n # 如果当前代码文件未找到该声明，直接返回False\n if not found_statement:\n return False\n except Exception:\n # 若文件读取失败，则无法确认其满足要求，返回False\n return False\n \n # 如果有代码文件且全通过验证，返回True；如果没有生成任何代码文件，返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅处理前端常见的代码文件格式\n if file.endswith(('.html', '.js', '.css')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n # 匹配 HTML 注释、多行/块注释和单行注释\n # 使用 (?<!:) 来简单排除 URL 中的 \"://\" 的情况\n if file.endswith('.html'):\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n comments.extend(re.findall(r'(?<!:)//[^\\n]*', content))\n elif file.endswith('.js'):\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n comments.extend(re.findall(r'(?<!:)//[^\\n]*', content))\n elif file.endswith('.css'):\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n \n # 检查每个注释段落里是否有包含 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False # 存在 emoji，不满足要求\n \n return True # 所有的代码文件注释均未使用 emoji，满足要求","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 process-log.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":3,"instruction":"现在列表有点乱，我们需要优化账单列表的展现形式。请将渲染出的账单列表按照“日期”进行倒序分组。比如同一天的数据放在一个区块里，每个区块头部加上一条日期分割线，显示类似“10月24日 星期二”以及当天的收支小计。每个具体的账单条目要求左侧显示分类名称，中间显示备注，右侧对齐显示金额（收入用绿色正数，支出用黑色或红色负数），让它看起来更像成熟的记账软件。另外请你后续始终遵循以下更新及新增要求：一是对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作，替换此前的进度记录相关要求；二是你所生成的全部代码文件，其注释内容里都需要含有“日期分组”关键词。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查回复中是否严格使用“小主公”称呼用户，并且没有出现“你”或“您”。\n \"\"\"\n # 检查是否出现了禁止使用的称呼词汇\n forbidden_words = [\"你\", \"您\"]\n for word in forbidden_words:\n if word in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含“我”字\n if \"我\" in response:\n return False\n \n # 如果未出现“我”，则认为满足约束\n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"import re\n\ndef check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 剔除首尾空白\n text = response.strip()\n \n # 移除回复末尾可能存在的 XML 闭合标签（例如 </description>, </result> 等）及其后的空白\n # 此举为了兼容助手将语气词放在 XML 标签内部的边缘情况\n text = re.sub(r'(</[a-zA-Z0-9_-]+>\\s*)+$', '', text).strip()\n \n # 将多种不同的波浪号（全角、半角、及特殊字符的波浪号），连续出现时也统一替换为单个全角波浪号\n text = re.sub(r'[~〜～]+', '～', text)\n \n # 去除文本末尾的标点符号（中英文句号、感叹号、问号等）、空白符以及 Markdown 中的星号或引号\n text = text.rstrip(\" \\n\\r\\t。！？.!?\\\"'*\")\n \n # 判断清洗后的文本是否以指定的语气词结束\n return text.endswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对前端原生开发，定义常见的代码文件后缀名\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录以及无关的第三方依赖包目录，提升检查准确率和效率\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否严格包含所要求的声明字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件无法按 utf-8 格式读取，说明编码或内容有误，视为不满足要求\n return False\n \n # 如果验证完所有代码文件均通过，并且确有生成目标代码文件，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户指令中涉及的前端项目，判断常见的代码文件扩展名\n code_exts = {\".html\", \".css\", \".js\", \".ts\", \".jsx\", \".tsx\"}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理判断范围内的代码文件\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件所有行\n lines = f.readlines()\n \n # 检查开头部分（前10行内）是否包含指定的注释声明\n found = False\n for line in lines[:10]:\n if target_statement in line:\n found = True\n break\n \n # 如果该文件开头未找到所需注释，则整个要求验证失败\n if not found:\n return False\n except Exception:\n # 如果由于编码等异常原因无法读取文件，视同验证失败\n return False\n\n # 若所有读取到的代码文件都通过了检查，则满足要求\n return True","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 匹配各类注释的正则表达式\n # 匹配 HTML 注释：<!-- 内容 -->\n html_comment_pattern = re.compile(r'<!--[\\s\\S]*?-->')\n # 匹配 CSS 和 JS 的多行注释：/* 内容 */\n block_comment_pattern = re.compile(r'/\\*[\\s\\S]*?\\*/')\n # 匹配 JS 单行注释：// 内容（排除 url 里的 ://）\n inline_comment_pattern = re.compile(r'(?<!:)//.*')\n\n target_extensions = ['.html', '.css', '.js']\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果有文件读取错误直接跳过\n continue\n\n comments = []\n # 根据不同文件类型匹配对应的注释类型\n # HTML 中可能会嵌套 CSS 和 JS，所以需要应用所有的模式\n if ext == '.html':\n comments.extend(html_comment_pattern.findall(content))\n comments.extend(block_comment_pattern.findall(content))\n comments.extend(inline_comment_pattern.findall(content))\n elif ext == '.css':\n comments.extend(block_comment_pattern.findall(content))\n elif ext == '.js':\n comments.extend(block_comment_pattern.findall(content))\n comments.extend(inline_comment_pattern.findall(content))\n\n # 遍历提取出的所有注释字符串，逐个字符判断是否存在 emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False # 一旦发现在注释内使用了 emoji，立刻返回 False\n\n # 若所有匹配的文件注释中都没有 emoji，则满足给定要求\n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“日期分组”关键词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令，推断单机纯前端Web应用可能的代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n # 如果文件无法读取，视作未能验证通过\n return False\n \n # 提取HTML注释 <!-- ... -->\n html_comments = re.findall(r'<!--(.*?)-->', content, re.DOTALL)\n # 提取CSS/JS多行块注释 /* ... */\n block_comments = re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL)\n # 提取JS单行注释 // ...\n line_comments = re.findall(r'//(.*)$', content, re.MULTILINE)\n \n # 将该文件中提取出来的所有注释拼接在一起\n all_comments = html_comments + block_comments + line_comments\n all_comments_text = \"\\n\".join(all_comments)\n \n # 检查注释内容中是否含有“日期分组”关键词\n if \"日期分组\" not in all_comments_text:\n return False\n \n # 确保至少存在一个代码文件，且所有代码文件的注释均包含目标关键词\n return has_code_file","tags":["Content","Keyword"]}]},{"round_id":4,"instruction":"越来越有模有样了。如果我记错账了需要修改怎么办？接下来增加编辑和删除功能。请给每个账单条目加上左滑手势（或者右侧加上一个小删除图标），点击删除时弹出原生确认框，确认后从 localStorage 中移除并刷新列表和顶部统计。另外，点击列表中的任意账单条目，可以重新唤起“添加账单”的面板，并自动将这条账单的数据回填到表单中，此时点击保存应该更新这条记录，而不是新增一条。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“编辑删除、本地存储操作”关键词，替换此前的注释关键词相关要求；二是请确保生成的代码文件采用4个空格进行缩进。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI是否严格使用“小主公”且未出现以“你”或“您”相称的情况。\n \"\"\"\n def contains_you(text: str) -> bool:\n # 排除可能包含“你”的常见非人称词汇干扰（本场景下记账APP可能极少出现，但为了严谨做简单过滤）\n text = text.replace(\"迷你\", \"\")\n # 检查是否包含“你”或敬称“您”\n return \"你\" in text or \"您\" in text\n\n # 1. 检查助手的直接回复内容\n if contains_you(response):\n return False\n\n # 2. 检查工作空间内的生成文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # Markdown 文件通常用于记录日志和工作进度，属于向用户汇报的文本，直接进行全文检查\n if file.endswith('.md'):\n if contains_you(content):\n return False\n \n # 代码文件（.js, .css, .html）中，AI 可能在注释里称呼用户，因此提取注释进行检查\n elif file.endswith(('.js', '.css', '.html')):\n comments = []\n if file.endswith('.html'):\n # 提取 HTML 注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n else:\n # 提取 CSS / JS 多行注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n # 提取 JS 单行注释\n if file.endswith('.js'):\n comments.extend(re.findall(r'//(.*)', content))\n \n for comment in comments:\n if contains_you(comment):\n return False\n except Exception:\n # 忽略不可读的文件\n continue\n\n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否满足第一人称使用的约束。\n 如果回复内容中包含了“我”，则违反了“切勿使用‘我’来自称”的规定，返回False；\n 否则视为满足规定，返回True。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"import re\n\ndef check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 定义匹配“汪～”结尾的逻辑（兼容全半角波浪号）\n def ends_with_wang(text: str) -> bool:\n # 移除末尾的空白字符和常见的结束标点符号\n clean_text = re.sub(r'[\\s。！!\\.，,]*$', '', text)\n return clean_text.endswith('汪～') or clean_text.endswith('汪~')\n\n # 1. 直接检查整个回复末尾\n if ends_with_wang(response):\n return True\n \n # 2. 考虑语气词可能被包含在最后一个XML闭合标签内的特殊情况\n # 比如: <description>xxxx汪～</description>\n clean_text = re.sub(r'[\\s。！!\\.，,]*$', '', response)\n clean_text_without_xml = re.sub(r'</[a-zA-Z0-9_-]+>[\\s]*$', '', clean_text)\n \n if clean_text != clean_text_without_xml and ends_with_wang(clean_text_without_xml):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 依据用户指令推断可能生成的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅检查代码文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 防止部分带有 BOM 头的文件影响第一行读取判断\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline()\n \n # 检查第一行是否包含指定的编码声明字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果由于任何原因（如编码问题导致读取失败），判定为不满足\n return False\n \n # 如果找到了至少一个代码文件且都满足条件则返回True，否则返回False\n return found_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n code_extensions = {\".html\", \".css\", \".js\"}\n has_code_file = False\n \n # 检查工作路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为目标代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, \"r\", encoding=\"utf-8\") as f:\n # 读取所有行（在绝大部分前端项目中文件大小可控）\n lines = f.readlines()\n \n found_statement = False\n # 检查开头处，这里放宽至前10行以适应编码声明或空行\n for line in lines[:10]:\n if target_statement in line:\n found_statement = True\n break\n \n # 如果该代码文件未包含该声明，则不满足要求\n if not found_statement:\n return False\n \n except Exception:\n # 如果遇到文件不可读等异常情况，判定为验证失败\n return False\n \n # 如果遍历完成且所有代码文件均包含声明，返回是否存在至少一个代码文件\n return has_code_file","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 限制需要检查的代码文件类型\n valid_extensions = {'.html', '.css', '.js'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果遇到文件读取异常，跳过该文件继续处理\n continue\n \n # 收集文件中所有的注释内容\n comments = []\n \n # 1. 匹配 HTML 注释: <!-- comment -->\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\n # 2. 匹配 CSS/JS 多行注释: /* comment */\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, flags=re.DOTALL))\n # 3. 匹配 JS 单行注释: // comment (不加 re.DOTALL 意味着只匹配到行末)\n comments.extend(re.findall(r'//(.*)', content))\n \n # 遍历提取出的每一段注释\n for comment in comments:\n # 确保 comment 是字符串类型\n if isinstance(comment, tuple):\n comment = \"\".join(comment)\n \n # 遍历注释中的每一个字符\n for char in comment:\n # 使用人类专家提供的 is_emoji 函数判断是否为 emoji\n if is_emoji(char):\n return False # 如果找到任何 emoji 符号，表示违反了要求，返回 False\n \n # 如果遍历完毕，均没有在任何代码注释中发现 emoji，则返回 True\n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“编辑删除、本地存储操作”关键词","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 记账APP相关的代码文件扩展名\n code_extensions = ['.html', '.css', '.js']\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果读取遇到非UTF-8编码等问题，尝试以其他方式忽略或报错\n # 依据系统提示，所有代码文件第一行必须是 utf-8，因此正常应该能读出\n continue\n \n # 使用正则表达式提取文件内容中的各类注释\n comments = []\n # 1. 提取 HTML 注释 <!-- ... -->\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # 2. 提取 JS/CSS 多行注释 /* ... */\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 3. 提取 JS 单行注释 // ... ，并使用否定逆序环视防止匹配到 http:// 等URL协议\n comments.extend(re.findall(r'(?<!:)//.*', content))\n \n # 将该文件的所有注释拼接到一起\n comments_text = \"\\n\".join(comments)\n \n # 检查指定的关键词是否包含在提取出的注释中\n if \"编辑删除、本地存储操作\" not in comments_text:\n return False\n \n # 如果没有找到任何代码文件，则也没有满足要求\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保生成的代码文件采用4个空格进行缩进","validation_code":"def check_indentation(response: str, workspace_path: str) -> bool:\n import os\n\n valid_extensions = {'.html', '.css', '.js'}\n tab_lines = 0\n space_lines = 0\n indent_counts = {}\n has_code_files = False\n\n if not os.path.exists(workspace_path):\n return False\n\n # 遍历工作区，查找代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in valid_extensions):\n has_code_files = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n # 分析每一行的缩进\n for line in lines:\n if not line.strip(): # 跳过空行\n continue\n \n stripped_line = line.lstrip()\n if stripped_line == line: # 跳过没有缩进的顶头行\n continue\n \n indent_str = line[:len(line)-len(stripped_line)]\n \n # 检查是否包含制表符\n if '\\t' in indent_str:\n tab_lines += 1\n else:\n spaces = len(indent_str)\n indent_counts[spaces] = indent_counts.get(spaces, 0) + 1\n space_lines += 1\n\n # 如果没有找到任何相关的代码文件，无法判断是否满足，按不通过处理\n if not has_code_files:\n return False\n\n even_spaces = 0\n mod4_spaces = 0\n \n # 统计偶数空格缩进和4的倍数空格缩进的情况\n for spaces, count in indent_counts.items():\n if spaces % 2 == 0:\n even_spaces += count\n if spaces % 4 == 0:\n mod4_spaces += count\n\n # 如果用 Tab 缩进的行数比用空格缩进的行数还多，判定不通过\n if tab_lines > space_lines:\n return False\n\n # 如果代码完全没有偶数空格缩进（可能均无缩进或格式异常），判定不通过\n if even_spaces == 0:\n return False\n\n # 判断4的倍数空格是否在偶数空格缩进中占据主导地位（阈值设为 80%）\n if mod4_spaces / even_spaces >= 0.8:\n return True\n\n return False","tags":["Style","Layout"]}]},{"round_id":5,"instruction":"接下来挑战一下预算控制功能。在顶部卡片的“总支出”下方，增加一个“设置月度预算”的文字按钮。点击后弹出一个简单的输入框设定每月预算并保存在本地。设定好后，该文字变成一个水平进度条，直观展示当前“总支出”占“月度预算”的百分比。当支出超过预算的80%时，进度条颜色变为橙色警告；超过100%时变为红色，并在旁边提示超支金额。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“预算控制、进度条阈值判断”关键词，替换此前的注释关键词相关要求；二是确保在代码开发完成之后，测试所有的代码并保证它们均通过测试。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证助手回复中是否没有出现“你”或“您”，以满足必须用“小主公”称呼用户的要求。\n \"\"\"\n if not response:\n return False\n \n # 排除掉非代词的固定词汇对判断的干扰（例如“迷你”）\n clean_response = response.replace(\"迷你\", \"\")\n \n # 检查是否包含“你”或“您”\n if \"你\" in clean_response or \"您\" in clean_response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含禁止使用的自称“我”\n if '我' in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 查找回复文本中所有 \"汪～\" 或 \"汪~\" 的位置\n matches = list(re.finditer(r'汪[～~]', response))\n if not matches:\n return False\n \n # 获取最后一次出现的位置\n last_match = matches[-1]\n \n # 截取出最后一次 \"汪～\" 出现后的所有文本\n after_text = response[last_match.end():]\n \n # 去除尾部可能由系统其他约束生成的 XML 闭合标签，防止发生误判冲突\n after_text_clean = re.sub(r'</[a-zA-Z0-9_]+>', '', after_text)\n \n # 检查清理后的剩余文本是否只含有非实质性字符（\\W 会匹配中英文标点、空格、换行和特殊符号等）\n # 如果仅包含这些符号，表示 \"汪～\" 后面再无额外有效文本输出，可以认为它处于最末尾。\n if re.fullmatch(r'\\W*', after_text_clean):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 根据用户指令，推断可能的代码文件类型为 HTML, CSS, JS\n code_extensions = {'.html', '.htm', '.css', '.js'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明字符串\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件读取发生异常（如编码错误等），视作不符合要求\n return False\n \n # 如果找到了代码文件且均通过了首行检查，则返回 True；否则返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 针对用户指令背景，锁定前端开发最常涉及的核心代码文件类型\n code_extensions = {\".html\", \".css\", \".js\"}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，防止检查到类似 .git 等不相关路径\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n statement_found = False\n \n try:\n # 使用 utf-8 编码读取文件内容，忽略潜在的解码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n # 检查开头部分（前20行），确认是否包含目标声明注释\n for _ in range(20):\n line = f.readline()\n if not line:\n break\n if target_statement in line:\n statement_found = True\n break\n except Exception:\n # 如果读取异常，则认为该文件无法验证通过\n return False\n \n # 若当前代码文件的开头没有找到要求的声明语句，则立即判定为未满足要求\n if not statement_found:\n return False\n \n # 如果找到了代码文件且均满足声明要求，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check_emoji_in_code_comments(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 若工作区路径不存在，视作满足要求\n if not os.path.exists(workspace_path):\n return True\n \n # 定义匹配各类注释的正则表达式\n html_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\n block_pattern = re.compile(r'/\\*(.*?)\\*/', re.DOTALL)\n # 使用负向后视断言 (?<!:) 避免匹配到 http:// 等网址链接\n js_line_pattern = re.compile(r'(?<!:)//(.*?)$', re.MULTILINE)\n py_line_pattern = re.compile(r'#(.*?)$', re.MULTILINE)\n \n # 遍历项目下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅处理前端项目可能的代码文件类型以及Python文件\n if ext not in ['.html', '.css', '.js', '.py']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n # 根据文件扩展名提取对应的注释内容\n if ext == '.html':\n comments.extend(html_pattern.findall(content))\n # HTML文件中可能包含 <script> 或 <style> 标签，故需支持JS和CSS的注释写法\n comments.extend(block_pattern.findall(content))\n comments.extend(js_line_pattern.findall(content))\n elif ext == '.css':\n comments.extend(block_pattern.findall(content))\n elif ext == '.js':\n comments.extend(block_pattern.findall(content))\n comments.extend(js_line_pattern.findall(content))\n elif ext == '.py':\n comments.extend(py_line_pattern.findall(content))\n \n # 遍历所有提取到的注释内容\n for comment in comments:\n # 遍历注释中的每个字符，判断是否为 emoji\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“预算控制、进度条阈值判断”关键词","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n \n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.tsx', '.vue', '.jsx'}\n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 收集当前文件中提取到的所有注释内容\n comments = []\n \n # HTML 注释 <!-- ... -->\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n # 多行注释 /* ... */\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 单行注释 // ...\n comments.extend(re.findall(r'//.*', content))\n # Python/Shell 单行注释 # ...\n if ext == '.py':\n comments.extend(re.findall(r'#.*', content))\n \n comments_text = '\\n'.join(comments)\n \n # 检查是否同时包含所要求的关键词信息\n # 处理 AI 直接将含顿号的短语写进注释，或将这两个关键词拆分写入的多种可能\n if \"预算控制\" not in comments_text or \"进度条阈值判断\" not in comments_text:\n if \"预算控制、进度条阈值判断\" not in comments_text:\n return False\n\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保生成的代码文件采用4个空格进行缩进","validation_code":"import os\n\ndef verify(response: str, workspace_path: str) -> bool:\n has_code_file = False\n \n # 遍历当前项目的工作区文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx')):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n has_code_file = True\n indent_counts = {}\n has_tab = False\n \n for line in lines:\n # 检查是否使用了 Tab 键缩进\n if line.startswith('\\t'):\n has_tab = True\n break\n \n leading_spaces = len(line) - len(line.lstrip(' '))\n stripped_line = line.lstrip(' \\t\\n\\r')\n \n if not stripped_line:\n continue\n \n if leading_spaces > 0:\n # 跳过多行注释中以 '*' 符号开头的换行对齐\n if stripped_line.startswith('*'):\n continue\n \n indent_counts[leading_spaces] = indent_counts.get(leading_spaces, 0) + 1\n \n if has_tab:\n return False\n \n if indent_counts:\n total_indents = sum(indent_counts.values())\n valid_indents = sum(v for k, v in indent_counts.items() if k % 4 == 0)\n \n # 统计 2/6 空格 和 4/8 空格的行数分布\n count_2 = indent_counts.get(2, 0) + indent_counts.get(6, 0)\n count_4 = indent_counts.get(4, 0) + indent_counts.get(8, 0)\n \n # 规则1：如果采用2空格缩进的特征行数多于4空格缩进的行数，说明不符合要求\n if count_2 > count_4 and count_2 > 0:\n return False\n \n # 规则2：容错处理：确保整个文件中至少有很大比例(>=70%)的缩进行符合4的倍数（规避偶尔长属性对齐导致非4倍数的影响）\n if total_indents > 5 and valid_indents / total_indents < 0.7:\n return False\n \n return has_code_file","tags":["Style","Layout"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":6,"instruction":"随着记账越来越多，我们需要加入时间过滤功能。在顶部统计卡片和下方账单列表之间，加一个“月份选择器”（例如 `< 2023年10月 >`，带左右切换箭头）。当用户切换月份时，下方的账单列表、顶部的总支出、结余，以及预算进度条，都必须严格依据所选月份的数据进行过滤、重新计算和渲染。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“月份筛选、时间过滤、数据联动刷新”关键词，替换此前的注释关键词相关要求；二是每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n forbidden_words = [\"你\", \"您\"]\n \n # 检查回复中是否包含违禁称呼字眼\n for word in forbidden_words:\n if word in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否满足：全程以“小助手”作为第一人称，切勿使用“我”来自称。\n \"\"\"\n # 检查回复内容中是否包含了禁止使用的“我”字\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 剔除首尾的空白字符\n cleaned_response = response.strip()\n \n # 定义常见的末尾标点符号，避免标点符号干扰结尾的判断\n punctuations = ['。', '！', '!', '.', '\"', \"'\", '”', '’', '…']\n \n # 循环剔除末尾的标点符号和随后的空白字符\n while cleaned_response and cleaned_response[-1] in punctuations:\n cleaned_response = cleaned_response[:-1].strip()\n \n # 判断最终清理后的回复内容是否以“汪～”或“汪~”结尾\n if cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~'):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 根据用户指令，推断可能的代码文件类型为 HTML, CSS, JS\n code_extensions = {'.html', '.css', '.js'}\n code_files = []\n \n # 检查工作空间路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 遍历项目目录，寻找代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果没有找到任何相关的代码文件，视为未满足要求\n if not code_files:\n return False\n \n # 逐一检查每个代码文件的第一行\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明格式\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果文件读取发生异常，同样视为验证失败\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 目标要求声明的字符串\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 考虑到项目是个单机版记账APP，检查HTML、CSS、JS这三种代码文件\n code_extensions = {'.html', '.css', '.js'}\n \n if not os.path.exists(workspace_path):\n return False\n \n code_files_found = 0\n \n # 遍历工作目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的前5行以检查开头处的声明\n lines = [f.readline() for _ in range(5)]\n \n found = False\n for line in lines:\n if not line: # 文件结束\n break\n if target_statement in line:\n found = True\n break\n \n # 如果在这前5行中没有发现目标声明，直接判定未满足要求\n if not found:\n return False\n except Exception:\n # 如果有文件无法读取，返回验证失败\n return False\n \n # 如果没找到任何代码文件，则判定不满足\n if code_files_found == 0:\n return False\n \n # 如果所有的代码文件前几行中都包含了目标声明，则满足要求\n return True","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 定义需要检查的代码文件扩展名\n target_extensions = {'.html', '.htm', '.js', '.css'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in target_extensions:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n except Exception:\n continue\n \n comments = []\n \n # HTML 文件注释提取\n if ext in ['.html', '.htm']:\n # 提取HTML原生注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 提取内联JS中的注释\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n comments.extend(re.findall(r'//(.*?)$', script, re.MULTILINE))\n comments.extend(re.findall(r'/\\*(.*?)\\*/', script, re.DOTALL))\n # 提取内联CSS中的注释\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, re.DOTALL | re.IGNORECASE)\n for style in styles:\n comments.extend(re.findall(r'/\\*(.*?)\\*/', style, re.DOTALL))\n \n # JS 文件注释提取\n elif ext == '.js':\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n \n # CSS 文件注释提取\n elif ext == '.css':\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n \n # 遍历提取出的所有注释，判断是否含有emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n\n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“月份筛选、时间过滤、数据联动刷新”关键词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义前端项目中常见的代码文件扩展名\n code_exts = {'.html', '.htm', '.css', '.js'}\n \n # 确保给定的工作区路径存在\n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n # 需要满足的注释关键词\n keywords = [\"月份筛选\", \"时间过滤\", \"数据联动刷新\"]\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于某些编码或权限原因无法读取，这里可以选择跳过或作为未通过处理\n # 为严谨起见，无法读取即无法验证注释，直接视为不满足\n return False\n \n # 存储该文件中的所有注释内容\n comments = []\n # 1. 提取 JavaScript 和 CSS 中的块注释 /* ... */\n comments.extend(re.findall(r'/\\*[\\s\\S]*?\\*/', content))\n # 2. 提取 JavaScript 中的单行注释 // ...\n comments.extend(re.findall(r'//.*', content))\n # 3. 提取 HTML 文件中的注释 <!-- ... -->\n comments.extend(re.findall(r'<!--[\\s\\S]*?-->', content))\n \n # 将该文件中的所有注释合并为一个长字符串以统一检查\n comments_text = \" \".join(comments)\n \n # 验证是否包含要求的三个关键词\n for kw in keywords:\n if kw not in comments_text:\n return False\n\n # 若至少存在一个代码文件，且每一个代码文件均无遗漏地在注释中包含了相关关键词，则验证通过\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保生成的代码文件采用4个空格进行缩进","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据项目背景，推断主要的代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n \n valid_indent_count = 0\n invalid_indent_count = 0\n tab_indent_count = 0\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n for line in f:\n stripped_line = line.lstrip()\n if not stripped_line:\n continue # 忽略空行\n \n leading_whitespace = line[:len(line) - len(stripped_line)]\n if not leading_whitespace:\n continue # 忽略没有缩进（顶格）的行\n \n if '\\t' in leading_whitespace:\n tab_indent_count += 1\n else:\n space_len = len(leading_whitespace)\n if space_len % 4 == 0:\n valid_indent_count += 1\n elif space_len % 2 == 0:\n # 偶数但非4的倍数（如2, 6, 10），通常是采用了2空格缩进风格\n invalid_indent_count += 1\n # 奇数个空格可能是多行注释（如 /* \\n * ）或简单的折行对齐，为了容错不计入非法缩进\n except Exception:\n continue\n\n if not has_code_file:\n return False\n \n total_indents = valid_indent_count + invalid_indent_count + tab_indent_count\n \n # 如果全篇代码没有任何缩进，则认为不符合\"采用4个空格缩进\"的约定\n if total_indents == 0:\n return False\n \n # 判断4个空格缩进是否在项目中占绝对主导地位（设定容错阈值为80%）\n if valid_indent_count / total_indents > 0.8:\n return True\n \n return False","tags":["Style","Layout"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":7,"instruction":"找特定账单还是有些麻烦，现在引入多维度筛选与搜索。在月份选择器下方增加一行过滤选项卡：全部、支出、收入，点击后列表动态过滤显示对应类型。同时，在这个区域的右侧加上一个搜索输入框，用户输入文字时，实时模糊匹配当前列表中的“备注”或“分类”字段，隐藏不匹配的条目，实现秒搜体验。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“多维度筛选、模糊搜索、类型过滤”关键词，替换此前的注释关键词相关要求；二是请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径，后续所有轮次均需遵守该要求。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否严格使用了“小主公”，且未出现“你”或“您”。\n \"\"\"\n # 检查是否包含“你”或“您”\n if \"你\" in response or \"您\" in response:\n return False\n \n # 只要未出现“你”或“您”，我们认为该要求得到满足\n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否满足：以“小助手”为第一人称，且未出现“我”来自称。\n \"\"\"\n # 如果回复中出现了“我”，则认为违反了“切勿使用'我'来自称”的要求\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符（包括换行符等）\n cleaned_response = response.strip()\n \n # 去除回复最末尾可能存在的常见中英文句末标点符号，防止误判（如“汪～。”）\n cleaned_response = re.sub(r'[。！？.!?]+$', '', cleaned_response).strip()\n \n # 判断最终清洗后的回复是否严格以“汪～”结尾\n return cleaned_response.endswith('汪～')","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n\n # 针对用户指令推断的常见代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n files_checked = 0\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取第一行数据\n first_line = f.readline()\n # 验证第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n files_checked += 1\n except Exception:\n # 如果文件读取发生异常，则视为验证失败\n return False\n\n # 如果没有任何代码文件被检查过，则视为未完成代码生成任务\n if files_checked == 0:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n \n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户项目需求，定义可能存在的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录，避免检查到如 .git, .vscode 等无关文件夹内的内容\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件的前20行，\"开头处\"通常在这一范围内\n lines = [f.readline() for _ in range(20)]\n \n found = False\n for line in lines:\n if not line: # 已经读取到文件末尾\n break\n if target_statement in line:\n found = True\n break\n \n # 只要有一个代码文件未在其开头包含该注释声明，即判定为不满足要求\n if not found:\n return False\n except Exception:\n # 如果文件无法正常读取或存在编码问题，判定为验证失败\n return False\n \n # 所有被找到的代码文件均符合要求，或不存在任何代码文件时，视为满足\n return True","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查代码项目中的各个代码文件，判断它们的注释内容中是否含有emoji。\n 如果在任何一个代码文件的注释里发现了emoji，则返回False，否则返回True。\n \"\"\"\n # 常用代码文件后缀分类\n html_exts = {'.html', '.htm'}\n c_style_exts = {'.js', '.css', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cs', '.go', '.swift', '.kt', '.php'}\n hash_exts = {'.py', '.rb', '.sh', '.yaml', '.yml'}\n \n # 注释的正则表达式提取\n html_comment_re = re.compile(r'<!--[\\s\\S]*?-->')\n multiline_comment_re = re.compile(r'/\\*[\\s\\S]*?\\*/')\n # 使用负向后行断言 (?<!:) 避免匹配到 http:// 或 https:// \n single_line_comment_re = re.compile(r'(?<!:)//[^\\r\\n]*')\n hash_comment_re = re.compile(r'#[^\\r\\n]*')\n \n # 用于移除字符串字面量（避免字符串内部被误解析为注释）\n string_re = re.compile(r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅处理常见的代码文件，跳过Markdown等文档文件\n if ext not in html_exts and ext not in c_style_exts and ext not in hash_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 首先剔除字符串字面量，防止误匹配\n content_no_strings = string_re.sub('', content)\n comments = []\n \n # 根据文件类型收集可能的注释区块\n if ext in html_exts:\n comments.extend(html_comment_re.findall(content_no_strings))\n comments.extend(multiline_comment_re.findall(content_no_strings))\n comments.extend(single_line_comment_re.findall(content_no_strings))\n \n if ext in c_style_exts:\n comments.extend(multiline_comment_re.findall(content_no_strings))\n comments.extend(single_line_comment_re.findall(content_no_strings))\n if ext == '.php':\n comments.extend(html_comment_re.findall(content_no_strings))\n comments.extend(hash_comment_re.findall(content_no_strings))\n \n if ext in hash_exts:\n comments.extend(hash_comment_re.findall(content_no_strings))\n \n # 遍历所有被提取出来的注释文本，判断是否含有emoji\n for comment in comments:\n for char in comment:\n if is_emoji(char):\n return False\n \n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“多维度筛选、模糊搜索、类型过滤”关键词","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证项目中的所有代码文件（HTML/CSS/JS）的注释中是否包含指定的三个关键词：\n “多维度筛选”、“模糊搜索”、“类型过滤”。\n \"\"\"\n keywords = [\"多维度筛选\", \"模糊搜索\", \"类型过滤\"]\n code_extensions = ['.html', '.css', '.js']\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 遇到无法以 utf-8 读取的文件，可视为不满足规范\n return False\n \n comments = []\n # 提取不同类型文件中的注释\n if ext == '.html':\n # HTML注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 考虑到HTML中可能内嵌JS和CSS代码，一并尝试提取对应的注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\n elif ext == '.css':\n # CSS注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n elif ext == '.js':\n # JS的多行和单行注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\n \n # 拼接该文件的所有注释为一段文本\n all_comments = \" \".join(comments)\n \n # 检查注释中是否同时包含这三个关键词\n for kw in keywords:\n if kw not in all_comments:\n return False\n \n # 至少需要有一个代码文件并被验证通过\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保生成的代码文件采用4个空格进行缩进","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n from collections import Counter\n\n if not os.path.exists(workspace_path):\n return False\n \n # 推断可能的代码文件类型\n valid_exts = {'.html', '.css', '.js', '.json'}\n checked_files = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n checked_files += 1\n \n indent_steps = []\n prev_spaces = 0\n \n for line in lines:\n # 忽略空行或完全由空白符组成的行\n if not line.strip():\n continue\n \n # 提取行首空白字符\n leading_whitespace = line[:len(line) - len(line.lstrip())]\n \n # 如果检测到 Tab 缩进，则直接判定不符合 4 个空格的规则\n if '\\t' in leading_whitespace:\n return False\n \n stripped = line.strip()\n # 忽略常见注释行，以免注释对齐格式（如 block comment 里的 *）干扰正常的缩进判断\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('//') or stripped.startswith('<!--'):\n continue\n \n # 计算前导空格数量\n spaces = len(leading_whitespace)\n \n # 若当前行缩进增加，则记录新增的缩进步长\n if spaces > prev_spaces:\n indent_steps.append(spaces - prev_spaces)\n prev_spaces = spaces\n \n # 若文件存在缩进变化，分析最常使用的缩进步长\n if indent_steps:\n counter = Counter(indent_steps)\n max_count = max(counter.values())\n # 找出所有出现次数等于最大频次的步长\n most_common_steps = [k for k, v in counter.items() if v == max_count]\n \n # 若4不在最常见缩进步长中，则判定该文件没有主要采用4个空格缩进\n if 4 not in most_common_steps:\n return False\n\n # 若未找到任何可验证的代码文件，则视为不满足条件\n if checked_files == 0:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":8,"instruction":"只有流水账不够直观，我们需要数据可视化。请在底部导航栏进行改造，原本只有一个“+”按钮，现在变成经典的底部Tab栏：“明细”（左）、“添加”（中间突出的+号）、“统计”（右）。点击“统计”时，隐藏账单列表，展示统计视图。在统计视图里，请使用原生 HTML5 Canvas（不依赖第三方库）绘制一个当前选中月份的“支出分类占比”饼图，并在饼图下方列出各分类支出的金额排行榜。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“底部Tab栏切换、Canvas饼图绘制、分类支出统计排名”关键词，替换此前的注释关键词相关要求；二是每行代码都要求写注释，并且在每个生成的代码文件里，所有的注释都只能追加在代码所在行的末尾（右边），不允许写在代码行之上或之下。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否满足称呼用户的相关限制。\n 条件: 不得出现“你”或“您”。\n \"\"\"\n if \"你\" in response or \"您\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否满足：以“小助手”作为第一人称，不使用“我”。\n 验证逻辑主要是检查回复内容中是否包含“我”字。\n \"\"\"\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否在最末尾加上了语气词“汪～”。\n \"\"\"\n if not response:\n return False\n \n # 去除字符串末尾的空白字符以及常见的中英文结束标点符号\n cleaned_response = response.rstrip(\" \\n\\r\\t.。!！\")\n \n # 判断清洗后的字符串是否以“汪～”结尾\n return cleaned_response.endswith(\"汪～\")","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义目标代码文件的后缀名\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历代码项目路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 筛选出属于前端代码类型的文件\n if any(file_name.lower().endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file_name)\n \n try:\n # 读取代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n \n # 检查第一行是否包含指定的编码声明文本\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 遇到无法读取的情况判定为不满足要求\n return False\n \n # 如果至少解析到一个代码文件，且全部通过了检查，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户项目描述，推断出的代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n \n if not os.path.exists(workspace_path):\n return False\n\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = []\n # 读取前10行，代表文件的\"开头处\"\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n lines.append(line)\n \n # 检查是否有任何一行包含目标注释声明\n statement_found = any(target_statement in line for line in lines)\n \n if not statement_found:\n return False\n except Exception:\n # 如果读取文件出现异常，保守判定为未满足要求\n return False\n\n # 必须至少存在一个代码文件，且全部符合要求才算True\n return found_code_file","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def check_comments_for_emoji(text: str) -> bool:\n comments = []\n # 1. 匹配 JS 单行注释 (排除 http:// 类的 url)\n comments.extend(re.findall(r'(?<!:)//(.*?)$', text, flags=re.MULTILINE))\n # 2. 匹配 JS / CSS 多行注释\n comments.extend(re.findall(r'/\\*(.*?)\\*/', text, flags=re.DOTALL))\n # 3. 匹配 HTML 注释\n comments.extend(re.findall(r'<!--(.*?)-->', text, flags=re.DOTALL))\n \n # 遍历所有提取出的注释文本\n for comment in comments:\n for char in comment:\n # 使用题目预置的人类专家函数 is_emoji() 校验每个字符\n if is_emoji(char):\n return False\n return True\n\n # 步骤一：检查 response 中的 markdown 代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9_-]*\\n(.*?)```', response, flags=re.DOTALL)\n for block in code_blocks:\n if not check_comments_for_emoji(block):\n return False\n\n # 步骤二：检查 workspace_path 下生成的代码文件\n code_extensions = {'.html', '.css', '.js'}\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对前端代码文件进行校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if not check_comments_for_emoji(content):\n return False\n except Exception:\n pass\n\n return True","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“底部Tab栏切换、Canvas饼图绘制、分类支出统计排名”关键词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 从指令中推断可能生成的代码文件扩展名\n code_extensions = ['.html', '.css', '.js']\n keywords = [\"底部Tab栏切换\", \"Canvas饼图绘制\", \"分类支出统计排名\"]\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n comments = []\n # 提取 HTML 注释 <!-- ... -->\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 提取 JS/CSS 多行注释 /* ... */\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n # 提取 JS 单行注释 // ...，排除类似 http:// 的情况\n comments.extend(re.findall(r'(?<!:)\\/\\/(.*)', content))\n\n # 将当前文件中所有的注释内容拼接为一个字符串以供检查\n comment_text = \" \".join(comments)\n\n # 验证当前代码文件的注释中是否包含所有被要求的关键字\n for kw in keywords:\n if kw not in comment_text:\n return False\n\n # 若至少存在一个代码文件，并且所有代码文件都经过验证包含了给定的关键词\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保生成的代码文件采用4个空格进行缩进","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 针对前端和常见代码文件的扩展名\n code_extensions = {'.js', '.html', '.css', '.py', '.json'}\n code_files_found = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于编码等问题无法读取，则跳过\n continue\n \n code_files_found += 1\n file_indents = []\n \n for line in lines:\n # 滤除结尾换行和空白，若当前行为空则跳过\n if not line.strip():\n continue\n \n # 匹配行首前导空白字符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n whitespace = match.group(1)\n \n # 过滤掉以星号开头用于对齐的多行注释内部行 (如 JSDoc / CSS 注释)\n if re.match(r'^[ \\t]+\\*(\\s|$)', line):\n continue\n \n # 只要发现存在制表符缩进，即违背了“采用空格”的规则\n if '\\t' in whitespace:\n return False\n \n # 记录具体的空格缩进数\n file_indents.append(len(whitespace))\n \n # 对文件的缩进情况进行统计分析\n if file_indents:\n # 统计缩进量为4的倍数（即符合4个空格缩进层级）的代码行数\n count_multiples_of_4 = sum(1 for x in file_indents if x % 4 == 0)\n total_indents = len(file_indents)\n \n if total_indents >= 5:\n # 对于超过一定行数的文件，若4的倍数缩进行占比小于60%，很可能是2空格或其他缩进\n if count_multiples_of_4 < total_indents * 0.6:\n return False\n else:\n # 文件内缩进行较少时（如极短的代码文件），若存在偶数且无4的倍数缩进（如2，6），判定失败\n if count_multiples_of_4 == 0 and sum(1 for x in file_indents if x % 2 == 0) > 0:\n return False\n\n # 若工作区内连一个合规类型的代码文件都没有，则无法验证成功\n if code_files_found == 0:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每行代码都要求写注释，并且在每个生成的代码文件里，所有的注释都只能追加在代码所在行的末尾（右边），不允许写在代码行之上或之下","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 确保项目路径存在\n if not os.path.exists(workspace_path):\n return False\n \n # 定义目标代码文件的扩展名\n code_extensions = ['.html', '.css', '.js']\n has_code_files = False\n \n # 定义对应语言中常见的注释标识符\n comment_tokens = ['//', '/*', '<!--']\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅针对代码文件进行检查\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_files = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n stripped = line.strip()\n # 跳过完全为空的行\n if not stripped:\n continue\n \n # 剔除常见的网络协议以防 URL 中的 '//' 干扰误判\n stripped_clean = stripped.replace('http://', '').replace('https://', '')\n \n # 1. 检查注释是否单独成行或位于代码行最前端（违反了“只能追加在末尾/右侧”）\n starts_with_comment = False\n for token in comment_tokens:\n if stripped_clean.startswith(token):\n starts_with_comment = True\n break\n \n if starts_with_comment:\n return False\n \n # 2. 检查本行是否包含注释标识符（违反了“每行代码都要求写注释”）\n has_comment = False\n for token in comment_tokens:\n if token in stripped_clean:\n has_comment = True\n break\n \n if not has_comment:\n return False\n \n # 只要存在代码文件并且全都没有违反规定，则返回True\n return has_code_files","tags":["Style","Layout"]}]},{"round_id":9,"instruction":"最后一步，为了防止用户清理浏览器缓存导致数据丢失，我们来实现本地数据备份功能。在页面右上角添加一个“设置”齿轮图标，点击弹出一个侧边抽屉，里面提供“导出备份”和“导入恢复”两个按钮。点击导出时，将 localStorage 中的全量账单数据转成 JSON 文件并触发浏览器下载；点击导入时，允许用户选择 JSON 文件，读取解析后覆盖本地数据并重新刷新整个页面视图。同时，如果当前没有任何账单数据，记得在列表区显示一个友好的“暂无账单，快去记一笔吧”的缺省提示图文。\n另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“本地数据备份导出、JSON文件导入恢复、空数据缺省提示”关键词，替换此前的注释关键词相关要求；二是编辑已有文件时，要求必须通过输出文件的全量代码来进行覆盖替换，不可使用Unified Diff或其他类型的补丁格式，后续所有轮次均需遵守该要求。","instruction_following_checklist":[{"constraint":"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况","validation_code":"import re\n\ndef verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复内容是否满足：凡是需要称呼用户的地方均使用“小主公”，不得出现“你”。\n \"\"\"\n # 移除 markdown 代码块，以防代码字符串或注释中存在不具备称谓意义的匹配干扰\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\n \n # 检查助手的对话文本中是否出现了“你”或“您”\n if '你' in text_without_code or '您' in text_without_code:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import re\n \n # 1. 移除 Markdown 代码块（支持跨行）\n clean_text = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\n \n # 2. 移除行内代码\n clean_text = re.sub(r'`.*?`', '', clean_text)\n \n # 3. 移除被双引号包裹的引用内容（中英文引号）\n clean_text = re.sub(r'[\"“].*?[\"”]', '', clean_text)\n \n # 4. 移除被单引号包裹的引用内容（中英文引号）\n clean_text = re.sub(r\"['‘].*?['’]\", '', clean_text)\n \n # 5. 检查清理后的文本中是否仍然包含“我”字\n if \"我\" in clean_text:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import re\n if not response:\n return False\n \n # 去除回复首尾的空白字符\n clean_response = response.strip()\n \n # 匹配末尾是否为“汪～”或“汪~”，允许跟随常见的句末标点（如感叹号、句号）和空白字符\n pattern = r'汪[～~][\\s\\.\\!。！]*$'\n \n # 1. 优先检查绝对的文本末尾\n if re.search(pattern, clean_response):\n return True\n \n # 2. 考虑到助手可能将语气词置于要求输出的 XML 标签内部（如 </description> 之前）\n # 尝试循环剥离位于字符串最末尾的 XML 闭合标签\n while re.search(r'</[a-zA-Z0-9]+>\\s*$', clean_response):\n clean_response = re.sub(r'</[a-zA-Z0-9]+>\\s*$', '', clean_response).strip()\n \n # 剥离标签后再次检查末尾是否符合要求\n if re.search(pattern, clean_response):\n return True\n \n return False","tags":["Content","Persona"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 针对前端项目场景，定义需要检查的代码文件扩展名\n code_exts = {'.html', '.js', '.css'}\n has_code_file = False\n \n # 如果工作区路径不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选指定的代码文件\n if any(file.endswith(ext) for ext in code_exts):\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取文件的起始第一行\n first_line = f.readline()\n # 检查特定子串是否包含在第一行中（允许由于不同语言包裹不同注释符）\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果由于读取等异常（如编码不支持或权限问题）则判定为不满足要求\n return False\n \n # 如果找到了代码文件且所有文件第一行均满足要求，则返回True，否则返回False\n return has_code_file","tags":["Content","Format"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 需要匹配的目标注释文本\n target_text = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n # 根据用户指令推断，项目为HTML、CSS和纯JS开发\n code_extensions = ['.html', '.css', '.js']\n \n code_files = []\n # 遍历工作区，收集所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n # 如果工作区内没有任何代码文件，通常意味着未生成文件，无法验证满足该要求\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n found = False\n # 检查文件开头前10行是否包含该目标注释声明\n for line in lines[:10]:\n if target_text in line:\n found = True\n break\n \n if not found:\n return False\n except Exception:\n # 若文件读取失败出现异常，则视为无法满足要求\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"在为各个代码文件编写注释时，切勿使用 emoji表情符号","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 匹配 HTML 注释\n html_comment_pattern = re.compile(r'<!--[\\s\\S]*?-->')\n \n # 匹配 HTML 中的 script 和 style 标签区块\n script_style_pattern = re.compile(r'<(script|style)[^>]*>([\\s\\S]*?)</\\1>', re.IGNORECASE)\n \n # 匹配 JS/CSS/Python 等代码中的字符串字面量和注释\n # group(1) 捕获到的才是注释，没捕获到的则是字符串，借此跳过字符串内的干扰内容\n code_comment_pattern = re.compile(\n r'\"(?:\\\\.|[^\"\\\\])*\"' # 双引号字符串\n r'|\\'(?:\\\\.|[^\\'\\\\])*\\'' # 单引号字符串\n r'|`(?:\\\\.|[^`\\\\])*`' # 反引号字符串\n r'|(/\\*[\\s\\S]*?\\*/|//.*|#.*)' # 块注释或行注释\n )\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤前端及常见的代码类型\n if ext not in ['.html', '.htm', '.js', '.css', '.py', '.ts', '.jsx', '.tsx']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n comments = []\n \n if ext in ['.html', '.htm']:\n # 收集 HTML 自身的原生注释\n comments.extend(html_comment_pattern.findall(content))\n \n # 收集内嵌 JS / CSS 中的代码注释\n for match in script_style_pattern.finditer(content):\n inner_content = match.group(2)\n for m in code_comment_pattern.finditer(inner_content):\n if m.group(1):\n comments.append(m.group(1))\n else:\n # 直接收集纯代码文件中的注释\n for m in code_comment_pattern.finditer(content):\n if m.group(1):\n comments.append(m.group(1))\n \n # 遍历所有提取出来的注释文本片段\n for comment in comments:\n for char in comment:\n # 使用人类专家提供的 is_emoji 函数判断该字符\n if is_emoji(char):\n return False # 如果在任何一处注释中发现 emoji，判定为未满足要求，返回 False\n \n return True # 所有的代码文件注释均无 emoji，满足给定要求","tags":["Content","Keyword"]},{"constraint":"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有“本地数据备份导出、JSON文件导入恢复、空数据缺省提示”关键词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的代码文件扩展名（基于前端单页面Web应用的需求）\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\n \n code_files_found = False\n # 需要验证的关键词，拆分成三个子关键词验证能同时兼容拼接逗号或顿号的不同情况\n keywords = [\"本地数据备份导出\", \"JSON文件导入恢复\", \"空数据缺省提示\"]\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用正则表达式提取各类可能的注释文本\n # 1. HTML/XML 注释\n html_comments = re.findall(r'<!--[\\s\\S]*?-->', content)\n # 2. JS/CSS 多行注释\n multi_comments = re.findall(r'/\\*[\\s\\S]*?\\*/', content)\n # 3. JS 单行注释\n single_comments = re.findall(r'//.*', content)\n # 4. 其他潜在的单行注释 (如 #)\n hash_comments = re.findall(r'#.*', content)\n \n # 拼接同一文件中所有的注释内容\n all_comments = \" \".join(html_comments + multi_comments + single_comments + hash_comments)\n \n # 检查该文件的注释是否包含所有要求的关键词\n for kw in keywords:\n if kw not in all_comments:\n return False\n except Exception:\n # 若文件无法正常读取或解码，判定验证失败\n return False\n \n # 必须至少包含一个代码文件，且所有代码文件的注释均通过了关键词检查\n return code_files_found","tags":["Content","Keyword"]},{"constraint":"请确保生成的代码文件采用4个空格进行缩进","validation_code":"import os\nfrom collections import Counter\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 定义该项目相关的代码文件后缀\n code_extensions = {'.html', '.css', '.js'}\n code_files = []\n \n # 遍历获取工作区下的全部代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if os.path.splitext(file)[1].lower() in code_extensions:\n code_files.append(os.path.join(root, file))\n \n # 如果没找到任何代码文件，则无法验证是否满足要求\n if not code_files:\n return False\n \n for filepath in code_files:\n with open(filepath, 'r', encoding='utf-8') as f:\n try:\n lines = f.readlines()\n except UnicodeDecodeError:\n # 忽略编码异常的非文本文件\n continue\n \n indents = []\n for line in lines:\n # 过滤空行\n if not line.strip():\n continue\n \n stripped = line.lstrip()\n # 截取前导空白字符\n indent_str = line[:len(line) - len(stripped)]\n \n # 如果缩进中存在制表符（Tab），直接判定失败\n if '\\t' in indent_str:\n return False\n \n # 忽略部分可能为了对齐而出现异常缩进的注释格式\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('<!--'):\n continue\n \n indents.append(len(indent_str))\n \n # 计算相邻行的缩进步长\n steps = []\n for i in range(1, len(indents)):\n diff = abs(indents[i] - indents[i-1])\n # 仅在缩进发生变化时记录步长\n if diff > 0:\n steps.append(diff)\n \n # 分析主导的缩进风格\n if steps:\n c = Counter(steps)\n # 如果 2 空格或 3 空格的缩进变化频率高于 4 空格，说明不符合 4 空格缩进的要求\n if c[2] > c[4] or c[3] > c[4]:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"每行代码都要求写注释，并且在每个生成的代码文件里，所有的注释都只能追加在代码所在行的末尾（右边），不允许写在代码行之上或之下","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 辅助函数：替换字符串内容和URL，防止内部碰巧出现的 '//' 或 '/*' 被误识别为注释\n def remove_strings_and_urls(text: str) -> str:\n # 替换常见 URL 协议头，防止 http:// 中的 // 被识别为注释\n text = text.replace('http://', 'http__').replace('https://', 'https__')\n # 将双引号、单引号、反引号内的字符串内容替换为下划线 '_'\n text = re.sub(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"', '_', text)\n text = re.sub(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\", '_', text)\n text = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '_', text)\n return text\n\n # 核心检查逻辑：验证一组代码行是否符合要求\n def check_lines(lines: list, ext_or_lang: str) -> bool:\n # 根据文件扩展名或语言类型分配可能出现的合法注释符\n if ext_or_lang in ['.js', '.ts', '.jsx', '.tsx', 'js', 'javascript', 'ts', 'typescript']:\n tokens = ['//', '/*']\n elif ext_or_lang in ['.css', 'css']:\n tokens = ['/*']\n elif ext_or_lang in ['.html', '.htm', 'html', 'htm']:\n # HTML 文件中可能内嵌 JS 和 CSS，所以放宽注释符校验\n tokens = ['<!--', '//', '/*']\n elif ext_or_lang in ['.py', 'python']:\n tokens = ['#']\n else:\n tokens = ['//', '/*', '<!--', '#']\n \n for line in lines:\n stripped = line.strip()\n # 忽略纯空行，空行不属于“代码行”范畴\n if not stripped:\n continue\n \n # 1. 检查是否为独立注释行（注释写在代码行之上或之下）\n is_standalone = False\n for token in tokens:\n if stripped.startswith(token):\n is_standalone = True\n break\n \n if is_standalone:\n return False # 发现独立注释行，不符合“不允许写在代码行之上或之下”\n \n # 2. 检查是否有追加注释（每行代码都必须有注释）\n line_no_str = remove_strings_and_urls(line)\n has_comment = False\n for token in tokens:\n if token in line_no_str:\n has_comment = True\n break\n \n if not has_comment:\n return False # 发现没有注释的代码行，不符合“每行代码都要求写注释”\n \n return True\n\n # 1. 优先扫描工作区中的代码文件\n code_extensions = ['.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.py']\n files_checked = 0\n\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n if not check_lines(lines, ext):\n return False\n files_checked += 1\n except Exception:\n pass\n \n if files_checked > 0:\n return True\n \n # 2. 如果工作区未找到文件，作为回退方案，检查 response 中的 Markdown 代码块\n blocks = re.findall(r'```(.*?)\\n(.*?)```', response, re.DOTALL)\n if not blocks:\n return False # 未发现任何代码\n \n for lang, code in blocks:\n lang = lang.lower().strip()\n if not lang:\n lang = 'txt'\n if not check_lines(code.split('\\n'), lang):\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"编辑已有文件时，要求必须通过输出文件的全量代码来进行覆盖替换，不可使用Unified Diff或其他类型的补丁格式","validation_code":"","tags":["Environment","File Operation"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"我想用HTML、CSS和纯JS开发一个简单的单机版记账APP，完全在浏览器本地运行。第一步，请帮我把基础骨架搭出来。要求适配手机端屏幕大小，页面主色调使用清新的绿色。顶部是一个绿色背景的统计卡片，显示固定的占位文本：“总支出：¥0.00”和“结余：¥0.00”。页面底部固定一个悬浮的绿色圆形“+”添加按钮。中间部分留白，作为未来展示账单列表的区域。这一轮只需做好静态UI和响应式布局，不需要实现任何记账逻辑。请你后续始终遵循以下要求：第一，代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）；第二，务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 如果回复中出现了“你”字，则说明违背了“不得出现以‘你’相称的情况”的约束\\n if '你' in response:\\n return False\\n \\n # 未出现“你”字，视为满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否未使用“我”来自称。\\n 通过直接判断字符串中是否包含“我”字来进行严格限制。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response or not response.strip():\\n return False\\n \\n # 定义可能出现在句末的常见标点符号和空白字符\\n # 注意：不包含波浪号 ~ 和 ～ 以防误删关键内容，也不包含尖括号 < 和 > 以防破坏闭合标签的匹配结构\\n remove_chars = '。，！？；：“”‘’（）《》〈〉【】『』「」.!,?;\\\"\\\\'()[]{}\\\\n\\\\r\\\\t '\\n \\n cleaned_response = response\\n \\n # 循环去除末尾的干扰性字符和结构化闭合标签，露出语义上的有效末尾内容\\n while True:\\n prev = cleaned_response\\n \\n # 1. 去除末尾的标点符号和空白字符\\n cleaned_response = cleaned_response.rstrip(remove_chars)\\n \\n # 2. 去除末尾可能存在的 XML/HTML 闭合标签（例如 </description>）\\n cleaned_response = re.sub(r'</[^>]+>$', '', cleaned_response)\\n \\n # 3. 去除末尾可能存在的 Markdown 代码块闭合标记（例如 ``` 或 ```javascript）\\n cleaned_response = re.sub(r'```[^`\\\\n]*$', '', cleaned_response)\\n \\n # 如果经过一轮剥离后字符串没有任何变化，说明已经处理到了纯文本的有效末尾，结束循环\\n if prev == cleaned_response:\\n break\\n \\n # 最终判断清理后的有效文本末尾是否是以“汪～”或“汪~”结尾\\n return cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check_coding_declaration(response: str, workspace_path: str) -> bool:\\n # 根据用户指令，预期生成的代码文件主要是HTML、CSS和JS\\n target_extensions = {'.html', '.css', '.js'}\\n file_found = False\\n\\n # 确保给定的项目地址是一个有效的目录\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历当前项目文件夹下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查读取到的第一行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 若文件为空、无法读取或编码错误，统一视为不满足要求\\n return False\\n \\n # 若至少存在一个代码文件且全都满足首行包含声明的要求，则返回True\\n return file_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义需要检查的目标声明文字\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 根据用户指令推断，可能涉及的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n has_code_file = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历当前代码项目工作区\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果读取遇到编码或其他错误，视为不符合规范或无法验证\\n return False\\n \\n statement_found = False\\n # 检查文件开头处（取前15行以避免第一行的编码声明或其他doctype标签影响）\\n for line in lines[:15]:\\n if target_statement in line:\\n statement_found = True\\n break\\n \\n # 若存在任何一个代码文件不满足此开头声明，即返回False\\n if not statement_found:\\n return False\\n\\n # 要求必须有生成代码文件才算完成了\\\"搭出基础骨架\\\"的第一步\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区目录不存在，则无法验证文件，直接默认为True\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 遍历当前工作区内所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选针对前端项目的核心代码文件类型\\n if not file.lower().endswith(('.html', '.css', '.js')):\\n continue\\n\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n comments = []\\n \\n # 1. 提取 HTML 注释 <!-- ... -->，支持跨行\\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\\n \\n # 2. 提取 CSS/JS 块级注释 /* ... */，支持跨行\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, flags=re.DOTALL))\\n \\n # 3. 提取 JS 行级注释 // ...，使用负向后视断言排除 http:// 和 https:// 中的双斜杠\\n comments.extend(re.findall(r'(?<!:)//(.*)', content))\\n\\n # 遍历所有提取到的注释，逐个字符判断是否存在 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"务必在条件允许的最大范围内，并行执行多个独立的工具调用以提高效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，页面看起来很清爽。接下来完善底部“+”按钮的交互。点击它后，要从底部滑动弹出一个“添加账单”的面板。面板里需要包含一个表单：顶部是“支出”和“收入”的单选切换标签；然后是分类选择（下拉框或排列的图标，比如餐饮、交通、购物、工资等）；一个金额输入框（需限制只能输入数字和小数点）；一个日期选择器（默认当天）；以及备注输入框。底部放置“保存”和“取消”按钮。点击取消或者面板外部可以关闭该面板。请用原生JS完成这些UI的交互效果。另外请你后续始终遵循以下更新后的要求：一是对项目所做的每一次改动，都需要通过更新progress.md文件来记录本轮完成的工作；二是必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率，替换此前的并行执行相关要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复内容中是否包含“你”字\\n if '你' in response:\\n return False\\n \\n # 也可以扩展检查“您”字，以确保没有使用其他第二人称代词，\\n # 但根据严格的字面要求，至少不能包含“你”\\n if '您' in response:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def verify_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 去除多行代码块，避免代码中合理的“我”字干扰判断\\n text_without_multiline_code = re.sub(r'```[\\\\s\\\\S]*?```', '', response)\\n \\n # 去除行内代码块\\n pure_text = re.sub(r'`[^`]*`', '', text_without_multiline_code)\\n \\n # 检查纯文本回复中是否使用了“我”\\n if '我' in pure_text:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符\\n text = response.strip()\\n \\n # 匹配末尾是否为“汪～”或“汪~”，允许后面跟随任意数量的非单词字符（如标点符号）\\n match = re.search(r'汪[～~]\\\\W*$', text)\\n \\n return bool(match)\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历项目路径下所有的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 读取文件第一行并判断\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 判断给定的编码声明字符串是否包含在第一行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于编码或其他读取错误，视为不符合要求\\n return False\\n \\n # 如果找到了至少一个代码文件且全部满足要求，返回True；否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 依据用户指令推断代码文件类型为 html, css, js\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\"}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n # 检查开头处（前10行）是否包含该声明\\n found_statement = False\\n for line in lines[:10]:\\n if target_statement in line:\\n found_statement = True\\n break\\n \\n if not found_statement:\\n return False\\n except Exception:\\n # 如果由于某种原因无法读取某代码文件，视为验证失败\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check_no_emoji_in_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 编译正则表达式以匹配各种注释形式\\n # 1. HTML注释: <!-- 匹配内容 -->\\n html_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\\n # 2. 块注释 (CSS和JS): /* 匹配内容 */\\n block_pattern = re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL)\\n # 3. 单行注释 (JS): // 匹配内容，并使用负向后视断言(?<!:)来避免匹配 http:// 或 https:// \\n line_pattern = re.compile(r'(?<!:)\\\\/\\\\/(.*?)$', re.MULTILINE)\\n \\n # 遍历当前项目工作区中的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 根据用户指令，推断只检查 .html, .css, .js\\n if ext in ['.html', '.css', '.js']:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n \\n # 根据不同文件类型应用对应的正则表达式来提取注释内容\\n if ext == '.html':\\n comments.extend(html_pattern.findall(content))\\n # HTML中可能包含内嵌的CSS和JS脚本，也同时提取相应注释\\n comments.extend(block_pattern.findall(content))\\n comments.extend(line_pattern.findall(content))\\n elif ext == '.css':\\n comments.extend(block_pattern.findall(content))\\n elif ext == '.js':\\n comments.extend(block_pattern.findall(content))\\n comments.extend(line_pattern.findall(content))\\n \\n # 遍历所有提取到的注释文本\\n for comment in comments:\\n for char in comment:\\n # 引用外部已经实现的 is_emoji 函数，判断是否有emoji\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新progress.md文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"很好，表单交互很顺滑。现在把核心的数据存储逻辑串起来。点击“保存”时，需要校验金额不能为0或空，验证通过后，将这条记录（类型、分类、金额、日期、备注）以JSON格式存入浏览器的 localStorage 中，并关闭面板。同时，在中部的空白区域根据 localStorage 的数据实时渲染出每一条账单。并且，需要计算所有数据的“收入”减去“支出”，实时更新顶部卡片里的“结余”和“总支出”金额。另外请你后续始终遵循以下要求：一是对项目所做的每一次改动，都需要通过更新 process-log.md 文件来记录本轮完成的工作，替换此前的进度记录相关要求；二是在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 辅助函数：判断一段文本中是否包含违规的第二人称称呼“你”或“您”\\n def contains_second_person_pronoun(text: str) -> bool:\\n if not text:\\n return False\\n # 将带有“你”但通常不作为人称代词使用的常见特例词汇剔除，避免产生误判\\n clean_text = text.replace(\\\"迷你\\\", \\\"\\\").replace(\\\"迪士尼\\\", \\\"\\\")\\n # 检查剔除后是否依然包含“你”或“您”\\n if \\\"你\\\" in clean_text or \\\"您\\\" in clean_text:\\n return True\\n return False\\n\\n # 1. 检查 AI 助手的最后一轮回复文本中是否违规使用了“你”或“您”\\n if contains_second_person_pronoun(response):\\n return False\\n\\n # 2. 检查工作区生成的代码文件与进度日志文件是否违规\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n # 根据用户指令中提到的 Web 三剑客和进度日志推断目标文件类型\\n target_extensions = ('.html', '.css', '.js', '.md', '.txt', '.json')\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if file_name.endswith(target_extensions):\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if contains_second_person_pronoun(content):\\n return False\\n except Exception:\\n # 对于无法以 utf-8 读取或发生其他 IO 异常的文件，直接跳过\\n continue\\n\\n # 如果所有被检查的内容里均未出现“你”或“您”，则认为满足了给定要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"import re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否切勿使用“我”来自称。\\n \\\"\\\"\\\"\\n # 使用正则表达式去除 Markdown 格式的块级代码，避免代码内部的注释或字符串导致误判\\n text_without_code_blocks = re.sub(r'```[\\\\s\\\\S]*?```', '', response)\\n \\n # 使用正则表达式去除 Markdown 格式的行内代码\\n pure_text = re.sub(r'`[^`]+`', '', text_without_code_blocks)\\n \\n # 检查纯文本中是否包含“我”字\\n if '我' in pure_text:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def check_response(response: str, workspace_path: str) -> bool:\\n import re\\n if not response:\\n return False\\n \\n # 去除回复末尾可能包含的空白字符以及常见的中英文标点符号\\n cleaned_response = re.sub(r'[\\\\s。，！、？\\\\.,!\\\\?]*$', '', response)\\n \\n # 判断剔除尾部干扰字符后，是否以“汪～”或“汪~”结尾\\n return cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 筛选出代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行\\n first_line = f.readline().strip()\\n # 检查第一行是否包含指定声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取失败，也视为不符合要求\\n return False\\n \\n # 如果项目中没有任何代码文件，则无法证明满足要求，返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\"}\\n has_code_file = False\\n \\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n found_statement = False\\n # 检查文件开头处（前10行以内）是否包含目标声明\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n if target_statement in line:\\n found_statement = True\\n break\\n \\n # 如果当前代码文件未找到该声明，直接返回False\\n if not found_statement:\\n return False\\n except Exception:\\n # 若文件读取失败，则无法确认其满足要求，返回False\\n return False\\n \\n # 如果有代码文件且全通过验证，返回True；如果没有生成任何代码文件，返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅处理前端常见的代码文件格式\\n if file.endswith(('.html', '.js', '.css')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n # 匹配 HTML 注释、多行/块注释和单行注释\\n # 使用 (?<!:) 来简单排除 URL 中的 \\\"://\\\" 的情况\\n if file.endswith('.html'):\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n comments.extend(re.findall(r'(?<!:)//[^\\\\n]*', content))\\n elif file.endswith('.js'):\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n comments.extend(re.findall(r'(?<!:)//[^\\\\n]*', content))\\n elif file.endswith('.css'):\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n \\n # 检查每个注释段落里是否有包含 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False # 存在 emoji，不满足要求\\n \\n return True # 所有的代码文件注释均未使用 emoji，满足要求\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 process-log.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在列表有点乱，我们需要优化账单列表的展现形式。请将渲染出的账单列表按照“日期”进行倒序分组。比如同一天的数据放在一个区块里，每个区块头部加上一条日期分割线，显示类似“10月24日 星期二”以及当天的收支小计。每个具体的账单条目要求左侧显示分类名称，中间显示备注，右侧对齐显示金额（收入用绿色正数，支出用黑色或红色负数），让它看起来更像成熟的记账软件。另外请你后续始终遵循以下更新及新增要求：一是对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作，替换此前的进度记录相关要求；二是你所生成的全部代码文件，其注释内容里都需要含有“日期分组”关键词。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查回复中是否严格使用“小主公”称呼用户，并且没有出现“你”或“您”。\\n \\\"\\\"\\\"\\n # 检查是否出现了禁止使用的称呼词汇\\n forbidden_words = [\\\"你\\\", \\\"您\\\"]\\n for word in forbidden_words:\\n if word in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含“我”字\\n if \\\"我\\\" in response:\\n return False\\n \\n # 如果未出现“我”，则认为满足约束\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"import re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 剔除首尾空白\\n text = response.strip()\\n \\n # 移除回复末尾可能存在的 XML 闭合标签（例如 </description>, </result> 等）及其后的空白\\n # 此举为了兼容助手将语气词放在 XML 标签内部的边缘情况\\n text = re.sub(r'(</[a-zA-Z0-9_-]+>\\\\s*)+$', '', text).strip()\\n \\n # 将多种不同的波浪号（全角、半角、及特殊字符的波浪号），连续出现时也统一替换为单个全角波浪号\\n text = re.sub(r'[~〜～]+', '～', text)\\n \\n # 去除文本末尾的标点符号（中英文句号、感叹号、问号等）、空白符以及 Markdown 中的星号或引号\\n text = text.rstrip(\\\" \\\\n\\\\r\\\\t。！？.!?\\\\\\\"'*\\\")\\n \\n # 判断清洗后的文本是否以指定的语气词结束\\n return text.endswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对前端原生开发，定义常见的代码文件后缀名\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录以及无关的第三方依赖包目录，提升检查准确率和效率\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否严格包含所要求的声明字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件无法按 utf-8 格式读取，说明编码或内容有误，视为不满足要求\\n return False\\n \\n # 如果验证完所有代码文件均通过，并且确有生成目标代码文件，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户指令中涉及的前端项目，判断常见的代码文件扩展名\\n code_exts = {\\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\"}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理判断范围内的代码文件\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件所有行\\n lines = f.readlines()\\n \\n # 检查开头部分（前10行内）是否包含指定的注释声明\\n found = False\\n for line in lines[:10]:\\n if target_statement in line:\\n found = True\\n break\\n \\n # 如果该文件开头未找到所需注释，则整个要求验证失败\\n if not found:\\n return False\\n except Exception:\\n # 如果由于编码等异常原因无法读取文件，视同验证失败\\n return False\\n\\n # 若所有读取到的代码文件都通过了检查，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 匹配各类注释的正则表达式\\n # 匹配 HTML 注释：<!-- 内容 -->\\n html_comment_pattern = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n # 匹配 CSS 和 JS 的多行注释：/* 内容 */\\n block_comment_pattern = re.compile(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n # 匹配 JS 单行注释：// 内容（排除 url 里的 ://）\\n inline_comment_pattern = re.compile(r'(?<!:)//.*')\\n\\n target_extensions = ['.html', '.css', '.js']\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果有文件读取错误直接跳过\\n continue\\n\\n comments = []\\n # 根据不同文件类型匹配对应的注释类型\\n # HTML 中可能会嵌套 CSS 和 JS，所以需要应用所有的模式\\n if ext == '.html':\\n comments.extend(html_comment_pattern.findall(content))\\n comments.extend(block_comment_pattern.findall(content))\\n comments.extend(inline_comment_pattern.findall(content))\\n elif ext == '.css':\\n comments.extend(block_comment_pattern.findall(content))\\n elif ext == '.js':\\n comments.extend(block_comment_pattern.findall(content))\\n comments.extend(inline_comment_pattern.findall(content))\\n\\n # 遍历提取出的所有注释字符串，逐个字符判断是否存在 emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False # 一旦发现在注释内使用了 emoji，立刻返回 False\\n\\n # 若所有匹配的文件注释中都没有 emoji，则满足给定要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“日期分组”关键词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令，推断单机纯前端Web应用可能的代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n # 如果文件无法读取，视作未能验证通过\\n return False\\n \\n # 提取HTML注释 <!-- ... -->\\n html_comments = re.findall(r'<!--(.*?)-->', content, re.DOTALL)\\n # 提取CSS/JS多行块注释 /* ... */\\n block_comments = re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL)\\n # 提取JS单行注释 // ...\\n line_comments = re.findall(r'//(.*)$', content, re.MULTILINE)\\n \\n # 将该文件中提取出来的所有注释拼接在一起\\n all_comments = html_comments + block_comments + line_comments\\n all_comments_text = \\\"\\\\n\\\".join(all_comments)\\n \\n # 检查注释内容中是否含有“日期分组”关键词\\n if \\\"日期分组\\\" not in all_comments_text:\\n return False\\n \\n # 确保至少存在一个代码文件，且所有代码文件的注释均包含目标关键词\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"越来越有模有样了。如果我记错账了需要修改怎么办？接下来增加编辑和删除功能。请给每个账单条目加上左滑手势（或者右侧加上一个小删除图标），点击删除时弹出原生确认框，确认后从 localStorage 中移除并刷新列表和顶部统计。另外，点击列表中的任意账单条目，可以重新唤起“添加账单”的面板，并自动将这条账单的数据回填到表单中，此时点击保存应该更新这条记录，而不是新增一条。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“编辑删除、本地存储操作”关键词，替换此前的注释关键词相关要求；二是请确保生成的代码文件采用4个空格进行缩进。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI是否严格使用“小主公”且未出现以“你”或“您”相称的情况。\\n \\\"\\\"\\\"\\n def contains_you(text: str) -> bool:\\n # 排除可能包含“你”的常见非人称词汇干扰（本场景下记账APP可能极少出现，但为了严谨做简单过滤）\\n text = text.replace(\\\"迷你\\\", \\\"\\\")\\n # 检查是否包含“你”或敬称“您”\\n return \\\"你\\\" in text or \\\"您\\\" in text\\n\\n # 1. 检查助手的直接回复内容\\n if contains_you(response):\\n return False\\n\\n # 2. 检查工作空间内的生成文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # Markdown 文件通常用于记录日志和工作进度，属于向用户汇报的文本，直接进行全文检查\\n if file.endswith('.md'):\\n if contains_you(content):\\n return False\\n \\n # 代码文件（.js, .css, .html）中，AI 可能在注释里称呼用户，因此提取注释进行检查\\n elif file.endswith(('.js', '.css', '.html')):\\n comments = []\\n if file.endswith('.html'):\\n # 提取 HTML 注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n else:\\n # 提取 CSS / JS 多行注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n # 提取 JS 单行注释\\n if file.endswith('.js'):\\n comments.extend(re.findall(r'//(.*)', content))\\n \\n for comment in comments:\\n if contains_you(comment):\\n return False\\n except Exception:\\n # 忽略不可读的文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否满足第一人称使用的约束。\\n 如果回复内容中包含了“我”，则违反了“切勿使用‘我’来自称”的规定，返回False；\\n 否则视为满足规定，返回True。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"import re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 定义匹配“汪～”结尾的逻辑（兼容全半角波浪号）\\n def ends_with_wang(text: str) -> bool:\\n # 移除末尾的空白字符和常见的结束标点符号\\n clean_text = re.sub(r'[\\\\s。！!\\\\.，,]*$', '', text)\\n return clean_text.endswith('汪～') or clean_text.endswith('汪~')\\n\\n # 1. 直接检查整个回复末尾\\n if ends_with_wang(response):\\n return True\\n \\n # 2. 考虑语气词可能被包含在最后一个XML闭合标签内的特殊情况\\n # 比如: <description>xxxx汪～</description>\\n clean_text = re.sub(r'[\\\\s。！!\\\\.，,]*$', '', response)\\n clean_text_without_xml = re.sub(r'</[a-zA-Z0-9_-]+>[\\\\s]*$', '', clean_text)\\n \\n if clean_text != clean_text_without_xml and ends_with_wang(clean_text_without_xml):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 依据用户指令推断可能生成的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅检查代码文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 防止部分带有 BOM 头的文件影响第一行读取判断\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline()\\n \\n # 检查第一行是否包含指定的编码声明字符串\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何原因（如编码问题导致读取失败），判定为不满足\\n return False\\n \\n # 如果找到了至少一个代码文件且都满足条件则返回True，否则返回False\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\"}\\n has_code_file = False\\n \\n # 检查工作路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为目标代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, \\\"r\\\", encoding=\\\"utf-8\\\") as f:\\n # 读取所有行（在绝大部分前端项目中文件大小可控）\\n lines = f.readlines()\\n \\n found_statement = False\\n # 检查开头处，这里放宽至前10行以适应编码声明或空行\\n for line in lines[:10]:\\n if target_statement in line:\\n found_statement = True\\n break\\n \\n # 如果该代码文件未包含该声明，则不满足要求\\n if not found_statement:\\n return False\\n \\n except Exception:\\n # 如果遇到文件不可读等异常情况，判定为验证失败\\n return False\\n \\n # 如果遍历完成且所有代码文件均包含声明，返回是否存在至少一个代码文件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 限制需要检查的代码文件类型\\n valid_extensions = {'.html', '.css', '.js'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果遇到文件读取异常，跳过该文件继续处理\\n continue\\n \\n # 收集文件中所有的注释内容\\n comments = []\\n \\n # 1. 匹配 HTML 注释: <!-- comment -->\\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\\n # 2. 匹配 CSS/JS 多行注释: /* comment */\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, flags=re.DOTALL))\\n # 3. 匹配 JS 单行注释: // comment (不加 re.DOTALL 意味着只匹配到行末)\\n comments.extend(re.findall(r'//(.*)', content))\\n \\n # 遍历提取出的每一段注释\\n for comment in comments:\\n # 确保 comment 是字符串类型\\n if isinstance(comment, tuple):\\n comment = \\\"\\\".join(comment)\\n \\n # 遍历注释中的每一个字符\\n for char in comment:\\n # 使用人类专家提供的 is_emoji 函数判断是否为 emoji\\n if is_emoji(char):\\n return False # 如果找到任何 emoji 符号，表示违反了要求，返回 False\\n \\n # 如果遍历完毕，均没有在任何代码注释中发现 emoji，则返回 True\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“编辑删除、本地存储操作”关键词\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 记账APP相关的代码文件扩展名\\n code_extensions = ['.html', '.css', '.js']\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果读取遇到非UTF-8编码等问题，尝试以其他方式忽略或报错\\n # 依据系统提示，所有代码文件第一行必须是 utf-8，因此正常应该能读出\\n continue\\n \\n # 使用正则表达式提取文件内容中的各类注释\\n comments = []\\n # 1. 提取 HTML 注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # 2. 提取 JS/CSS 多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 3. 提取 JS 单行注释 // ... ，并使用否定逆序环视防止匹配到 http:// 等URL协议\\n comments.extend(re.findall(r'(?<!:)//.*', content))\\n \\n # 将该文件的所有注释拼接到一起\\n comments_text = \\\"\\\\n\\\".join(comments)\\n \\n # 检查指定的关键词是否包含在提取出的注释中\\n if \\\"编辑删除、本地存储操作\\\" not in comments_text:\\n return False\\n \\n # 如果没有找到任何代码文件，则也没有满足要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件采用4个空格进行缩进\",\n \"validation_code\": \"def check_indentation(response: str, workspace_path: str) -> bool:\\n import os\\n\\n valid_extensions = {'.html', '.css', '.js'}\\n tab_lines = 0\\n space_lines = 0\\n indent_counts = {}\\n has_code_files = False\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 遍历工作区，查找代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in valid_extensions):\\n has_code_files = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n # 分析每一行的缩进\\n for line in lines:\\n if not line.strip(): # 跳过空行\\n continue\\n \\n stripped_line = line.lstrip()\\n if stripped_line == line: # 跳过没有缩进的顶头行\\n continue\\n \\n indent_str = line[:len(line)-len(stripped_line)]\\n \\n # 检查是否包含制表符\\n if '\\\\t' in indent_str:\\n tab_lines += 1\\n else:\\n spaces = len(indent_str)\\n indent_counts[spaces] = indent_counts.get(spaces, 0) + 1\\n space_lines += 1\\n\\n # 如果没有找到任何相关的代码文件，无法判断是否满足，按不通过处理\\n if not has_code_files:\\n return False\\n\\n even_spaces = 0\\n mod4_spaces = 0\\n \\n # 统计偶数空格缩进和4的倍数空格缩进的情况\\n for spaces, count in indent_counts.items():\\n if spaces % 2 == 0:\\n even_spaces += count\\n if spaces % 4 == 0:\\n mod4_spaces += count\\n\\n # 如果用 Tab 缩进的行数比用空格缩进的行数还多，判定不通过\\n if tab_lines > space_lines:\\n return False\\n\\n # 如果代码完全没有偶数空格缩进（可能均无缩进或格式异常），判定不通过\\n if even_spaces == 0:\\n return False\\n\\n # 判断4的倍数空格是否在偶数空格缩进中占据主导地位（阈值设为 80%）\\n if mod4_spaces / even_spaces >= 0.8:\\n return True\\n\\n return False\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"接下来挑战一下预算控制功能。在顶部卡片的“总支出”下方，增加一个“设置月度预算”的文字按钮。点击后弹出一个简单的输入框设定每月预算并保存在本地。设定好后，该文字变成一个水平进度条，直观展示当前“总支出”占“月度预算”的百分比。当支出超过预算的80%时，进度条颜色变为橙色警告；超过100%时变为红色，并在旁边提示超支金额。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“预算控制、进度条阈值判断”关键词，替换此前的注释关键词相关要求；二是确保在代码开发完成之后，测试所有的代码并保证它们均通过测试。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证助手回复中是否没有出现“你”或“您”，以满足必须用“小主公”称呼用户的要求。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 排除掉非代词的固定词汇对判断的干扰（例如“迷你”）\\n clean_response = response.replace(\\\"迷你\\\", \\\"\\\")\\n \\n # 检查是否包含“你”或“您”\\n if \\\"你\\\" in clean_response or \\\"您\\\" in clean_response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含禁止使用的自称“我”\\n if '我' in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 查找回复文本中所有 \\\"汪～\\\" 或 \\\"汪~\\\" 的位置\\n matches = list(re.finditer(r'汪[～~]', response))\\n if not matches:\\n return False\\n \\n # 获取最后一次出现的位置\\n last_match = matches[-1]\\n \\n # 截取出最后一次 \\\"汪～\\\" 出现后的所有文本\\n after_text = response[last_match.end():]\\n \\n # 去除尾部可能由系统其他约束生成的 XML 闭合标签，防止发生误判冲突\\n after_text_clean = re.sub(r'</[a-zA-Z0-9_]+>', '', after_text)\\n \\n # 检查清理后的剩余文本是否只含有非实质性字符（\\\\W 会匹配中英文标点、空格、换行和特殊符号等）\\n # 如果仅包含这些符号，表示 \\\"汪～\\\" 后面再无额外有效文本输出，可以认为它处于最末尾。\\n if re.fullmatch(r'\\\\W*', after_text_clean):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 根据用户指令，推断可能的代码文件类型为 HTML, CSS, JS\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明字符串\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取发生异常（如编码错误等），视作不符合要求\\n return False\\n \\n # 如果找到了代码文件且均通过了首行检查，则返回 True；否则返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 针对用户指令背景，锁定前端开发最常涉及的核心代码文件类型\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\"}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，防止检查到类似 .git 等不相关路径\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n statement_found = False\\n \\n try:\\n # 使用 utf-8 编码读取文件内容，忽略潜在的解码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n # 检查开头部分（前20行），确认是否包含目标声明注释\\n for _ in range(20):\\n line = f.readline()\\n if not line:\\n break\\n if target_statement in line:\\n statement_found = True\\n break\\n except Exception:\\n # 如果读取异常，则认为该文件无法验证通过\\n return False\\n \\n # 若当前代码文件的开头没有找到要求的声明语句，则立即判定为未满足要求\\n if not statement_found:\\n return False\\n \\n # 如果找到了代码文件且均满足声明要求，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check_emoji_in_code_comments(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 若工作区路径不存在，视作满足要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 定义匹配各类注释的正则表达式\\n html_pattern = re.compile(r'<!--(.*?)-->', re.DOTALL)\\n block_pattern = re.compile(r'/\\\\*(.*?)\\\\*/', re.DOTALL)\\n # 使用负向后视断言 (?<!:) 避免匹配到 http:// 等网址链接\\n js_line_pattern = re.compile(r'(?<!:)//(.*?)$', re.MULTILINE)\\n py_line_pattern = re.compile(r'#(.*?)$', re.MULTILINE)\\n \\n # 遍历项目下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅处理前端项目可能的代码文件类型以及Python文件\\n if ext not in ['.html', '.css', '.js', '.py']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n # 根据文件扩展名提取对应的注释内容\\n if ext == '.html':\\n comments.extend(html_pattern.findall(content))\\n # HTML文件中可能包含 <script> 或 <style> 标签，故需支持JS和CSS的注释写法\\n comments.extend(block_pattern.findall(content))\\n comments.extend(js_line_pattern.findall(content))\\n elif ext == '.css':\\n comments.extend(block_pattern.findall(content))\\n elif ext == '.js':\\n comments.extend(block_pattern.findall(content))\\n comments.extend(js_line_pattern.findall(content))\\n elif ext == '.py':\\n comments.extend(py_line_pattern.findall(content))\\n \\n # 遍历所有提取到的注释内容\\n for comment in comments:\\n # 遍历注释中的每个字符，判断是否为 emoji\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“预算控制、进度条阈值判断”关键词\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.tsx', '.vue', '.jsx'}\\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 收集当前文件中提取到的所有注释内容\\n comments = []\\n \\n # HTML 注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n # 多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 单行注释 // ...\\n comments.extend(re.findall(r'//.*', content))\\n # Python/Shell 单行注释 # ...\\n if ext == '.py':\\n comments.extend(re.findall(r'#.*', content))\\n \\n comments_text = '\\\\n'.join(comments)\\n \\n # 检查是否同时包含所要求的关键词信息\\n # 处理 AI 直接将含顿号的短语写进注释，或将这两个关键词拆分写入的多种可能\\n if \\\"预算控制\\\" not in comments_text or \\\"进度条阈值判断\\\" not in comments_text:\\n if \\\"预算控制、进度条阈值判断\\\" not in comments_text:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件采用4个空格进行缩进\",\n \"validation_code\": \"import os\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n has_code_file = False\\n \\n # 遍历当前项目的工作区文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx')):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n has_code_file = True\\n indent_counts = {}\\n has_tab = False\\n \\n for line in lines:\\n # 检查是否使用了 Tab 键缩进\\n if line.startswith('\\\\t'):\\n has_tab = True\\n break\\n \\n leading_spaces = len(line) - len(line.lstrip(' '))\\n stripped_line = line.lstrip(' \\\\t\\\\n\\\\r')\\n \\n if not stripped_line:\\n continue\\n \\n if leading_spaces > 0:\\n # 跳过多行注释中以 '*' 符号开头的换行对齐\\n if stripped_line.startswith('*'):\\n continue\\n \\n indent_counts[leading_spaces] = indent_counts.get(leading_spaces, 0) + 1\\n \\n if has_tab:\\n return False\\n \\n if indent_counts:\\n total_indents = sum(indent_counts.values())\\n valid_indents = sum(v for k, v in indent_counts.items() if k % 4 == 0)\\n \\n # 统计 2/6 空格 和 4/8 空格的行数分布\\n count_2 = indent_counts.get(2, 0) + indent_counts.get(6, 0)\\n count_4 = indent_counts.get(4, 0) + indent_counts.get(8, 0)\\n \\n # 规则1：如果采用2空格缩进的特征行数多于4空格缩进的行数，说明不符合要求\\n if count_2 > count_4 and count_2 > 0:\\n return False\\n \\n # 规则2：容错处理：确保整个文件中至少有很大比例(>=70%)的缩进行符合4的倍数（规避偶尔长属性对齐导致非4倍数的影响）\\n if total_indents > 5 and valid_indents / total_indents < 0.7:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"随着记账越来越多，我们需要加入时间过滤功能。在顶部统计卡片和下方账单列表之间，加一个“月份选择器”（例如 `< 2023年10月 >`，带左右切换箭头）。当用户切换月份时，下方的账单列表、顶部的总支出、结余，以及预算进度条，都必须严格依据所选月份的数据进行过滤、重新计算和渲染。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“月份筛选、时间过滤、数据联动刷新”关键词，替换此前的注释关键词相关要求；二是每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n forbidden_words = [\\\"你\\\", \\\"您\\\"]\\n \\n # 检查回复中是否包含违禁称呼字眼\\n for word in forbidden_words:\\n if word in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否满足：全程以“小助手”作为第一人称，切勿使用“我”来自称。\\n \\\"\\\"\\\"\\n # 检查回复内容中是否包含了禁止使用的“我”字\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 剔除首尾的空白字符\\n cleaned_response = response.strip()\\n \\n # 定义常见的末尾标点符号，避免标点符号干扰结尾的判断\\n punctuations = ['。', '！', '!', '.', '\\\"', \\\"'\\\", '”', '’', '…']\\n \\n # 循环剔除末尾的标点符号和随后的空白字符\\n while cleaned_response and cleaned_response[-1] in punctuations:\\n cleaned_response = cleaned_response[:-1].strip()\\n \\n # 判断最终清理后的回复内容是否以“汪～”或“汪~”结尾\\n if cleaned_response.endswith('汪～') or cleaned_response.endswith('汪~'):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 根据用户指令，推断可能的代码文件类型为 HTML, CSS, JS\\n code_extensions = {'.html', '.css', '.js'}\\n code_files = []\\n \\n # 检查工作空间路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 遍历项目目录，寻找代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何相关的代码文件，视为未满足要求\\n if not code_files:\\n return False\\n \\n # 逐一检查每个代码文件的第一行\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明格式\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果文件读取发生异常，同样视为验证失败\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 目标要求声明的字符串\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n \\n # 考虑到项目是个单机版记账APP，检查HTML、CSS、JS这三种代码文件\\n code_extensions = {'.html', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n code_files_found = 0\\n \\n # 遍历工作目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的前5行以检查开头处的声明\\n lines = [f.readline() for _ in range(5)]\\n \\n found = False\\n for line in lines:\\n if not line: # 文件结束\\n break\\n if target_statement in line:\\n found = True\\n break\\n \\n # 如果在这前5行中没有发现目标声明，直接判定未满足要求\\n if not found:\\n return False\\n except Exception:\\n # 如果有文件无法读取，返回验证失败\\n return False\\n \\n # 如果没找到任何代码文件，则判定不满足\\n if code_files_found == 0:\\n return False\\n \\n # 如果所有的代码文件前几行中都包含了目标声明，则满足要求\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 定义需要检查的代码文件扩展名\\n target_extensions = {'.html', '.htm', '.js', '.css'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in target_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comments = []\\n \\n # HTML 文件注释提取\\n if ext in ['.html', '.htm']:\\n # 提取HTML原生注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 提取内联JS中的注释\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n comments.extend(re.findall(r'//(.*?)$', script, re.MULTILINE))\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', script, re.DOTALL))\\n # 提取内联CSS中的注释\\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, re.DOTALL | re.IGNORECASE)\\n for style in styles:\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', style, re.DOTALL))\\n \\n # JS 文件注释提取\\n elif ext == '.js':\\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n \\n # CSS 文件注释提取\\n elif ext == '.css':\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n \\n # 遍历提取出的所有注释，判断是否含有emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“月份筛选、时间过滤、数据联动刷新”关键词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义前端项目中常见的代码文件扩展名\\n code_exts = {'.html', '.htm', '.css', '.js'}\\n \\n # 确保给定的工作区路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n # 需要满足的注释关键词\\n keywords = [\\\"月份筛选\\\", \\\"时间过滤\\\", \\\"数据联动刷新\\\"]\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于某些编码或权限原因无法读取，这里可以选择跳过或作为未通过处理\\n # 为严谨起见，无法读取即无法验证注释，直接视为不满足\\n return False\\n \\n # 存储该文件中的所有注释内容\\n comments = []\\n # 1. 提取 JavaScript 和 CSS 中的块注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content))\\n # 2. 提取 JavaScript 中的单行注释 // ...\\n comments.extend(re.findall(r'//.*', content))\\n # 3. 提取 HTML 文件中的注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--[\\\\s\\\\S]*?-->', content))\\n \\n # 将该文件中的所有注释合并为一个长字符串以统一检查\\n comments_text = \\\" \\\".join(comments)\\n \\n # 验证是否包含要求的三个关键词\\n for kw in keywords:\\n if kw not in comments_text:\\n return False\\n\\n # 若至少存在一个代码文件，且每一个代码文件均无遗漏地在注释中包含了相关关键词，则验证通过\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件采用4个空格进行缩进\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据项目背景，推断主要的代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n \\n valid_indent_count = 0\\n invalid_indent_count = 0\\n tab_indent_count = 0\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n for line in f:\\n stripped_line = line.lstrip()\\n if not stripped_line:\\n continue # 忽略空行\\n \\n leading_whitespace = line[:len(line) - len(stripped_line)]\\n if not leading_whitespace:\\n continue # 忽略没有缩进（顶格）的行\\n \\n if '\\\\t' in leading_whitespace:\\n tab_indent_count += 1\\n else:\\n space_len = len(leading_whitespace)\\n if space_len % 4 == 0:\\n valid_indent_count += 1\\n elif space_len % 2 == 0:\\n # 偶数但非4的倍数（如2, 6, 10），通常是采用了2空格缩进风格\\n invalid_indent_count += 1\\n # 奇数个空格可能是多行注释（如 /* \\\\n * ）或简单的折行对齐，为了容错不计入非法缩进\\n except Exception:\\n continue\\n\\n if not has_code_file:\\n return False\\n \\n total_indents = valid_indent_count + invalid_indent_count + tab_indent_count\\n \\n # 如果全篇代码没有任何缩进，则认为不符合\\\"采用4个空格缩进\\\"的约定\\n if total_indents == 0:\\n return False\\n \\n # 判断4个空格缩进是否在项目中占绝对主导地位（设定容错阈值为80%）\\n if valid_indent_count / total_indents > 0.8:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"找特定账单还是有些麻烦，现在引入多维度筛选与搜索。在月份选择器下方增加一行过滤选项卡：全部、支出、收入，点击后列表动态过滤显示对应类型。同时，在这个区域的右侧加上一个搜索输入框，用户输入文字时，实时模糊匹配当前列表中的“备注”或“分类”字段，隐藏不匹配的条目，实现秒搜体验。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“多维度筛选、模糊搜索、类型过滤”关键词，替换此前的注释关键词相关要求；二是请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径，后续所有轮次均需遵守该要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否严格使用了“小主公”，且未出现“你”或“您”。\\n \\\"\\\"\\\"\\n # 检查是否包含“你”或“您”\\n if \\\"你\\\" in response or \\\"您\\\" in response:\\n return False\\n \\n # 只要未出现“你”或“您”，我们认为该要求得到满足\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否满足：以“小助手”为第一人称，且未出现“我”来自称。\\n \\\"\\\"\\\"\\n # 如果回复中出现了“我”，则认为违反了“切勿使用'我'来自称”的要求\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（包括换行符等）\\n cleaned_response = response.strip()\\n \\n # 去除回复最末尾可能存在的常见中英文句末标点符号，防止误判（如“汪～。”）\\n cleaned_response = re.sub(r'[。！？.!?]+$', '', cleaned_response).strip()\\n \\n # 判断最终清洗后的回复是否严格以“汪～”结尾\\n return cleaned_response.endswith('汪～')\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 针对用户指令推断的常见代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n files_checked = 0\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取第一行数据\\n first_line = f.readline()\\n # 验证第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n files_checked += 1\\n except Exception:\\n # 如果文件读取发生异常，则视为验证失败\\n return False\\n\\n # 如果没有任何代码文件被检查过，则视为未完成代码生成任务\\n if files_checked == 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n \\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户项目需求，定义可能存在的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录，避免检查到如 .git, .vscode 等无关文件夹内的内容\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件的前20行，\\\"开头处\\\"通常在这一范围内\\n lines = [f.readline() for _ in range(20)]\\n \\n found = False\\n for line in lines:\\n if not line: # 已经读取到文件末尾\\n break\\n if target_statement in line:\\n found = True\\n break\\n \\n # 只要有一个代码文件未在其开头包含该注释声明，即判定为不满足要求\\n if not found:\\n return False\\n except Exception:\\n # 如果文件无法正常读取或存在编码问题，判定为验证失败\\n return False\\n \\n # 所有被找到的代码文件均符合要求，或不存在任何代码文件时，视为满足\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查代码项目中的各个代码文件，判断它们的注释内容中是否含有emoji。\\n 如果在任何一个代码文件的注释里发现了emoji，则返回False，否则返回True。\\n \\\"\\\"\\\"\\n # 常用代码文件后缀分类\\n html_exts = {'.html', '.htm'}\\n c_style_exts = {'.js', '.css', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.cs', '.go', '.swift', '.kt', '.php'}\\n hash_exts = {'.py', '.rb', '.sh', '.yaml', '.yml'}\\n \\n # 注释的正则表达式提取\\n html_comment_re = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n multiline_comment_re = re.compile(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n # 使用负向后行断言 (?<!:) 避免匹配到 http:// 或 https:// \\n single_line_comment_re = re.compile(r'(?<!:)//[^\\\\r\\\\n]*')\\n hash_comment_re = re.compile(r'#[^\\\\r\\\\n]*')\\n \\n # 用于移除字符串字面量（避免字符串内部被误解析为注释）\\n string_re = re.compile(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅处理常见的代码文件，跳过Markdown等文档文件\\n if ext not in html_exts and ext not in c_style_exts and ext not in hash_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 首先剔除字符串字面量，防止误匹配\\n content_no_strings = string_re.sub('', content)\\n comments = []\\n \\n # 根据文件类型收集可能的注释区块\\n if ext in html_exts:\\n comments.extend(html_comment_re.findall(content_no_strings))\\n comments.extend(multiline_comment_re.findall(content_no_strings))\\n comments.extend(single_line_comment_re.findall(content_no_strings))\\n \\n if ext in c_style_exts:\\n comments.extend(multiline_comment_re.findall(content_no_strings))\\n comments.extend(single_line_comment_re.findall(content_no_strings))\\n if ext == '.php':\\n comments.extend(html_comment_re.findall(content_no_strings))\\n comments.extend(hash_comment_re.findall(content_no_strings))\\n \\n if ext in hash_exts:\\n comments.extend(hash_comment_re.findall(content_no_strings))\\n \\n # 遍历所有被提取出来的注释文本，判断是否含有emoji\\n for comment in comments:\\n for char in comment:\\n if is_emoji(char):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“多维度筛选、模糊搜索、类型过滤”关键词\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证项目中的所有代码文件（HTML/CSS/JS）的注释中是否包含指定的三个关键词：\\n “多维度筛选”、“模糊搜索”、“类型过滤”。\\n \\\"\\\"\\\"\\n keywords = [\\\"多维度筛选\\\", \\\"模糊搜索\\\", \\\"类型过滤\\\"]\\n code_extensions = ['.html', '.css', '.js']\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 遇到无法以 utf-8 读取的文件，可视为不满足规范\\n return False\\n \\n comments = []\\n # 提取不同类型文件中的注释\\n if ext == '.html':\\n # HTML注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 考虑到HTML中可能内嵌JS和CSS代码，一并尝试提取对应的注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\\n elif ext == '.css':\\n # CSS注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n elif ext == '.js':\\n # JS的多行和单行注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n comments.extend(re.findall(r'//(.*?)$', content, re.MULTILINE))\\n \\n # 拼接该文件的所有注释为一段文本\\n all_comments = \\\" \\\".join(comments)\\n \\n # 检查注释中是否同时包含这三个关键词\\n for kw in keywords:\\n if kw not in all_comments:\\n return False\\n \\n # 至少需要有一个代码文件并被验证通过\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件采用4个空格进行缩进\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n from collections import Counter\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 推断可能的代码文件类型\\n valid_exts = {'.html', '.css', '.js', '.json'}\\n checked_files = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n checked_files += 1\\n \\n indent_steps = []\\n prev_spaces = 0\\n \\n for line in lines:\\n # 忽略空行或完全由空白符组成的行\\n if not line.strip():\\n continue\\n \\n # 提取行首空白字符\\n leading_whitespace = line[:len(line) - len(line.lstrip())]\\n \\n # 如果检测到 Tab 缩进，则直接判定不符合 4 个空格的规则\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n stripped = line.strip()\\n # 忽略常见注释行，以免注释对齐格式（如 block comment 里的 *）干扰正常的缩进判断\\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('//') or stripped.startswith('<!--'):\\n continue\\n \\n # 计算前导空格数量\\n spaces = len(leading_whitespace)\\n \\n # 若当前行缩进增加，则记录新增的缩进步长\\n if spaces > prev_spaces:\\n indent_steps.append(spaces - prev_spaces)\\n prev_spaces = spaces\\n \\n # 若文件存在缩进变化，分析最常使用的缩进步长\\n if indent_steps:\\n counter = Counter(indent_steps)\\n max_count = max(counter.values())\\n # 找出所有出现次数等于最大频次的步长\\n most_common_steps = [k for k, v in counter.items() if v == max_count]\\n \\n # 若4不在最常见缩进步长中，则判定该文件没有主要采用4个空格缩进\\n if 4 not in most_common_steps:\\n return False\\n\\n # 若未找到任何可验证的代码文件，则视为不满足条件\\n if checked_files == 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"只有流水账不够直观，我们需要数据可视化。请在底部导航栏进行改造，原本只有一个“+”按钮，现在变成经典的底部Tab栏：“明细”（左）、“添加”（中间突出的+号）、“统计”（右）。点击“统计”时，隐藏账单列表，展示统计视图。在统计视图里，请使用原生 HTML5 Canvas（不依赖第三方库）绘制一个当前选中月份的“支出分类占比”饼图，并在饼图下方列出各分类支出的金额排行榜。另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“底部Tab栏切换、Canvas饼图绘制、分类支出统计排名”关键词，替换此前的注释关键词相关要求；二是每行代码都要求写注释，并且在每个生成的代码文件里，所有的注释都只能追加在代码所在行的末尾（右边），不允许写在代码行之上或之下。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否满足称呼用户的相关限制。\\n 条件: 不得出现“你”或“您”。\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response or \\\"您\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否满足：以“小助手”作为第一人称，不使用“我”。\\n 验证逻辑主要是检查回复内容中是否包含“我”字。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否在最末尾加上了语气词“汪～”。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除字符串末尾的空白字符以及常见的中英文结束标点符号\\n cleaned_response = response.rstrip(\\\" \\\\n\\\\r\\\\t.。!！\\\")\\n \\n # 判断清洗后的字符串是否以“汪～”结尾\\n return cleaned_response.endswith(\\\"汪～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义目标代码文件的后缀名\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历代码项目路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 筛选出属于前端代码类型的文件\\n if any(file_name.lower().endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n # 读取代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n \\n # 检查第一行是否包含指定的编码声明文本\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 遇到无法读取的情况判定为不满足要求\\n return False\\n \\n # 如果至少解析到一个代码文件，且全部通过了检查，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_statement = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户项目描述，推断出的代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = []\\n # 读取前10行，代表文件的\\\"开头处\\\"\\n for _ in range(10):\\n line = f.readline()\\n if not line:\\n break\\n lines.append(line)\\n \\n # 检查是否有任何一行包含目标注释声明\\n statement_found = any(target_statement in line for line in lines)\\n \\n if not statement_found:\\n return False\\n except Exception:\\n # 如果读取文件出现异常，保守判定为未满足要求\\n return False\\n\\n # 必须至少存在一个代码文件，且全部符合要求才算True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def check_comments_for_emoji(text: str) -> bool:\\n comments = []\\n # 1. 匹配 JS 单行注释 (排除 http:// 类的 url)\\n comments.extend(re.findall(r'(?<!:)//(.*?)$', text, flags=re.MULTILINE))\\n # 2. 匹配 JS / CSS 多行注释\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', text, flags=re.DOTALL))\\n # 3. 匹配 HTML 注释\\n comments.extend(re.findall(r'<!--(.*?)-->', text, flags=re.DOTALL))\\n \\n # 遍历所有提取出的注释文本\\n for comment in comments:\\n for char in comment:\\n # 使用题目预置的人类专家函数 is_emoji() 校验每个字符\\n if is_emoji(char):\\n return False\\n return True\\n\\n # 步骤一：检查 response 中的 markdown 代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9_-]*\\\\n(.*?)```', response, flags=re.DOTALL)\\n for block in code_blocks:\\n if not check_comments_for_emoji(block):\\n return False\\n\\n # 步骤二：检查 workspace_path 下生成的代码文件\\n code_extensions = {'.html', '.css', '.js'}\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对前端代码文件进行校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if not check_comments_for_emoji(content):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“底部Tab栏切换、Canvas饼图绘制、分类支出统计排名”关键词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 从指令中推断可能生成的代码文件扩展名\\n code_extensions = ['.html', '.css', '.js']\\n keywords = [\\\"底部Tab栏切换\\\", \\\"Canvas饼图绘制\\\", \\\"分类支出统计排名\\\"]\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n comments = []\\n # 提取 HTML 注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 提取 JS/CSS 多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n # 提取 JS 单行注释 // ...，排除类似 http:// 的情况\\n comments.extend(re.findall(r'(?<!:)\\\\/\\\\/(.*)', content))\\n\\n # 将当前文件中所有的注释内容拼接为一个字符串以供检查\\n comment_text = \\\" \\\".join(comments)\\n\\n # 验证当前代码文件的注释中是否包含所有被要求的关键字\\n for kw in keywords:\\n if kw not in comment_text:\\n return False\\n\\n # 若至少存在一个代码文件，并且所有代码文件都经过验证包含了给定的关键词\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件采用4个空格进行缩进\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 针对前端和常见代码文件的扩展名\\n code_extensions = {'.js', '.html', '.css', '.py', '.json'}\\n code_files_found = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于编码等问题无法读取，则跳过\\n continue\\n \\n code_files_found += 1\\n file_indents = []\\n \\n for line in lines:\\n # 滤除结尾换行和空白，若当前行为空则跳过\\n if not line.strip():\\n continue\\n \\n # 匹配行首前导空白字符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n whitespace = match.group(1)\\n \\n # 过滤掉以星号开头用于对齐的多行注释内部行 (如 JSDoc / CSS 注释)\\n if re.match(r'^[ \\\\t]+\\\\*(\\\\s|$)', line):\\n continue\\n \\n # 只要发现存在制表符缩进，即违背了“采用空格”的规则\\n if '\\\\t' in whitespace:\\n return False\\n \\n # 记录具体的空格缩进数\\n file_indents.append(len(whitespace))\\n \\n # 对文件的缩进情况进行统计分析\\n if file_indents:\\n # 统计缩进量为4的倍数（即符合4个空格缩进层级）的代码行数\\n count_multiples_of_4 = sum(1 for x in file_indents if x % 4 == 0)\\n total_indents = len(file_indents)\\n \\n if total_indents >= 5:\\n # 对于超过一定行数的文件，若4的倍数缩进行占比小于60%，很可能是2空格或其他缩进\\n if count_multiples_of_4 < total_indents * 0.6:\\n return False\\n else:\\n # 文件内缩进行较少时（如极短的代码文件），若存在偶数且无4的倍数缩进（如2，6），判定失败\\n if count_multiples_of_4 == 0 and sum(1 for x in file_indents if x % 2 == 0) > 0:\\n return False\\n\\n # 若工作区内连一个合规类型的代码文件都没有，则无法验证成功\\n if code_files_found == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每行代码都要求写注释，并且在每个生成的代码文件里，所有的注释都只能追加在代码所在行的末尾（右边），不允许写在代码行之上或之下\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 确保项目路径存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义目标代码文件的扩展名\\n code_extensions = ['.html', '.css', '.js']\\n has_code_files = False\\n \\n # 定义对应语言中常见的注释标识符\\n comment_tokens = ['//', '/*', '<!--']\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅针对代码文件进行检查\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_files = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n stripped = line.strip()\\n # 跳过完全为空的行\\n if not stripped:\\n continue\\n \\n # 剔除常见的网络协议以防 URL 中的 '//' 干扰误判\\n stripped_clean = stripped.replace('http://', '').replace('https://', '')\\n \\n # 1. 检查注释是否单独成行或位于代码行最前端（违反了“只能追加在末尾/右侧”）\\n starts_with_comment = False\\n for token in comment_tokens:\\n if stripped_clean.startswith(token):\\n starts_with_comment = True\\n break\\n \\n if starts_with_comment:\\n return False\\n \\n # 2. 检查本行是否包含注释标识符（违反了“每行代码都要求写注释”）\\n has_comment = False\\n for token in comment_tokens:\\n if token in stripped_clean:\\n has_comment = True\\n break\\n \\n if not has_comment:\\n return False\\n \\n # 只要存在代码文件并且全都没有违反规定，则返回True\\n return has_code_files\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 9,\n \"instruction\": \"最后一步，为了防止用户清理浏览器缓存导致数据丢失，我们来实现本地数据备份功能。在页面右上角添加一个“设置”齿轮图标，点击弹出一个侧边抽屉，里面提供“导出备份”和“导入恢复”两个按钮。点击导出时，将 localStorage 中的全量账单数据转成 JSON 文件并触发浏览器下载；点击导入时，允许用户选择 JSON 文件，读取解析后覆盖本地数据并重新刷新整个页面视图。同时，如果当前没有任何账单数据，记得在列表区显示一个友好的“暂无账单，快去记一笔吧”的缺省提示图文。\\n另外请你后续始终遵循以下更新及新增要求：一是你所生成的全部代码文件，其注释内容里都需要含有“本地数据备份导出、JSON文件导入恢复、空数据缺省提示”关键词，替换此前的注释关键词相关要求；二是编辑已有文件时，要求必须通过输出文件的全量代码来进行覆盖替换，不可使用Unified Diff或其他类型的补丁格式，后续所有轮次均需遵守该要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"凡是需要称呼用户的地方，均须严格使用“小主公”，不得出现以“你”相称的情况\",\n \"validation_code\": \"import re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证AI助手的回复内容是否满足：凡是需要称呼用户的地方均使用“小主公”，不得出现“你”。\\n \\\"\\\"\\\"\\n # 移除 markdown 代码块，以防代码字符串或注释中存在不具备称谓意义的匹配干扰\\n text_without_code = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\\n \\n # 检查助手的对话文本中是否出现了“你”或“您”\\n if '你' in text_without_code or '您' in text_without_code:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"在回复过程中，请务必全程以“小助手”作为第一人称，切勿使用“我”来自称\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 1. 移除 Markdown 代码块（支持跨行）\\n clean_text = re.sub(r'```.*?```', '', response, flags=re.DOTALL)\\n \\n # 2. 移除行内代码\\n clean_text = re.sub(r'`.*?`', '', clean_text)\\n \\n # 3. 移除被双引号包裹的引用内容（中英文引号）\\n clean_text = re.sub(r'[\\\"“].*?[\\\"”]', '', clean_text)\\n \\n # 4. 移除被单引号包裹的引用内容（中英文引号）\\n clean_text = re.sub(r\\\"['‘].*?['’]\\\", '', clean_text)\\n \\n # 5. 检查清理后的文本中是否仍然包含“我”字\\n if \\\"我\\\" in clean_text:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词“汪～”\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import re\\n if not response:\\n return False\\n \\n # 去除回复首尾的空白字符\\n clean_response = response.strip()\\n \\n # 匹配末尾是否为“汪～”或“汪~”，允许跟随常见的句末标点（如感叹号、句号）和空白字符\\n pattern = r'汪[～~][\\\\s\\\\.\\\\!。！]*$'\\n \\n # 1. 优先检查绝对的文本末尾\\n if re.search(pattern, clean_response):\\n return True\\n \\n # 2. 考虑到助手可能将语气词置于要求输出的 XML 标签内部（如 </description> 之前）\\n # 尝试循环剥离位于字符串最末尾的 XML 闭合标签\\n while re.search(r'</[a-zA-Z0-9]+>\\\\s*$', clean_response):\\n clean_response = re.sub(r'</[a-zA-Z0-9]+>\\\\s*$', '', clean_response).strip()\\n \\n # 剥离标签后再次检查末尾是否符合要求\\n if re.search(pattern, clean_response):\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 针对前端项目场景，定义需要检查的代码文件扩展名\\n code_exts = {'.html', '.js', '.css'}\\n has_code_file = False\\n \\n # 如果工作区路径不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选指定的代码文件\\n if any(file.endswith(ext) for ext in code_exts):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取文件的起始第一行\\n first_line = f.readline()\\n # 检查特定子串是否包含在第一行中（允许由于不同语言包裹不同注释符）\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果由于读取等异常（如编码不支持或权限问题）则判定为不满足要求\\n return False\\n \\n # 如果找到了代码文件且所有文件第一行均满足要求，则返回True，否则返回False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 需要匹配的目标注释文本\\n target_text = \\\"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\\\"\\n # 根据用户指令推断，项目为HTML、CSS和纯JS开发\\n code_extensions = ['.html', '.css', '.js']\\n \\n code_files = []\\n # 遍历工作区，收集所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n # 如果工作区内没有任何代码文件，通常意味着未生成文件，无法验证满足该要求\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n found = False\\n # 检查文件开头前10行是否包含该目标注释声明\\n for line in lines[:10]:\\n if target_text in line:\\n found = True\\n break\\n \\n if not found:\\n return False\\n except Exception:\\n # 若文件读取失败出现异常，则视为无法满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"一旦对生成的代码测试完成，必须在回复中严格按照此XML格式，分项列出执行结果与描述内容：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"在为各个代码文件编写注释时，切勿使用 emoji表情符号\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 匹配 HTML 注释\\n html_comment_pattern = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n \\n # 匹配 HTML 中的 script 和 style 标签区块\\n script_style_pattern = re.compile(r'<(script|style)[^>]*>([\\\\s\\\\S]*?)</\\\\1>', re.IGNORECASE)\\n \\n # 匹配 JS/CSS/Python 等代码中的字符串字面量和注释\\n # group(1) 捕获到的才是注释，没捕获到的则是字符串，借此跳过字符串内的干扰内容\\n code_comment_pattern = re.compile(\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"' # 双引号字符串\\n r'|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'' # 单引号字符串\\n r'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`' # 反引号字符串\\n r'|(/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*|#.*)' # 块注释或行注释\\n )\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤前端及常见的代码类型\\n if ext not in ['.html', '.htm', '.js', '.css', '.py', '.ts', '.jsx', '.tsx']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n comments = []\\n \\n if ext in ['.html', '.htm']:\\n # 收集 HTML 自身的原生注释\\n comments.extend(html_comment_pattern.findall(content))\\n \\n # 收集内嵌 JS / CSS 中的代码注释\\n for match in script_style_pattern.finditer(content):\\n inner_content = match.group(2)\\n for m in code_comment_pattern.finditer(inner_content):\\n if m.group(1):\\n comments.append(m.group(1))\\n else:\\n # 直接收集纯代码文件中的注释\\n for m in code_comment_pattern.finditer(content):\\n if m.group(1):\\n comments.append(m.group(1))\\n \\n # 遍历所有提取出来的注释文本片段\\n for comment in comments:\\n for char in comment:\\n # 使用人类专家提供的 is_emoji 函数判断该字符\\n if is_emoji(char):\\n return False # 如果在任何一处注释中发现 emoji，判定为未满足要求，返回 False\\n \\n return True # 所有的代码文件注释均无 emoji，满足给定要求\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码必须是最短路径实现，不得进行过度设计，也不准自行补充需求未提及的方案（例如额外的动画、交互效果等）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"对项目所做的每一次改动，都需要通过更新 bill-record-progress.md 文件来记录本轮完成的工作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在涉及文件操作的场景中，必须尽可能地利用专用工具（如Read, Write, Edit），而不是依赖 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有“本地数据备份导出、JSON文件导入恢复、空数据缺省提示”关键词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的代码文件扩展名（基于前端单页面Web应用的需求）\\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n \\n code_files_found = False\\n # 需要验证的关键词，拆分成三个子关键词验证能同时兼容拼接逗号或顿号的不同情况\\n keywords = [\\\"本地数据备份导出\\\", \\\"JSON文件导入恢复\\\", \\\"空数据缺省提示\\\"]\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用正则表达式提取各类可能的注释文本\\n # 1. HTML/XML 注释\\n html_comments = re.findall(r'<!--[\\\\s\\\\S]*?-->', content)\\n # 2. JS/CSS 多行注释\\n multi_comments = re.findall(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', content)\\n # 3. JS 单行注释\\n single_comments = re.findall(r'//.*', content)\\n # 4. 其他潜在的单行注释 (如 #)\\n hash_comments = re.findall(r'#.*', content)\\n \\n # 拼接同一文件中所有的注释内容\\n all_comments = \\\" \\\".join(html_comments + multi_comments + single_comments + hash_comments)\\n \\n # 检查该文件的注释是否包含所有要求的关键词\\n for kw in keywords:\\n if kw not in all_comments:\\n return False\\n except Exception:\\n # 若文件无法正常读取或解码，判定验证失败\\n return False\\n \\n # 必须至少包含一个代码文件，且所有代码文件的注释均通过了关键词检查\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保生成的代码文件采用4个空格进行缩进\",\n \"validation_code\": \"import os\\nfrom collections import Counter\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 定义该项目相关的代码文件后缀\\n code_extensions = {'.html', '.css', '.js'}\\n code_files = []\\n \\n # 遍历获取工作区下的全部代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if os.path.splitext(file)[1].lower() in code_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没找到任何代码文件，则无法验证是否满足要求\\n if not code_files:\\n return False\\n \\n for filepath in code_files:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n try:\\n lines = f.readlines()\\n except UnicodeDecodeError:\\n # 忽略编码异常的非文本文件\\n continue\\n \\n indents = []\\n for line in lines:\\n # 过滤空行\\n if not line.strip():\\n continue\\n \\n stripped = line.lstrip()\\n # 截取前导空白字符\\n indent_str = line[:len(line) - len(stripped)]\\n \\n # 如果缩进中存在制表符（Tab），直接判定失败\\n if '\\\\t' in indent_str:\\n return False\\n \\n # 忽略部分可能为了对齐而出现异常缩进的注释格式\\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('<!--'):\\n continue\\n \\n indents.append(len(indent_str))\\n \\n # 计算相邻行的缩进步长\\n steps = []\\n for i in range(1, len(indents)):\\n diff = abs(indents[i] - indents[i-1])\\n # 仅在缩进发生变化时记录步长\\n if diff > 0:\\n steps.append(diff)\\n \\n # 分析主导的缩进风格\\n if steps:\\n c = Counter(steps)\\n # 如果 2 空格或 3 空格的缩进变化频率高于 4 空格，说明不符合 4 空格缩进的要求\\n if c[2] > c[4] or c[3] > c[4]:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每次有新的内容更新时，都需要更新 Review.md 文件来Review已有代码，格式请参考：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请确保在输出的代码文件中，所有的文件引用均使用相对路径，而非绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"每行代码都要求写注释，并且在每个生成的代码文件里，所有的注释都只能追加在代码所在行的末尾（右边），不允许写在代码行之上或之下\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 辅助函数：替换字符串内容和URL，防止内部碰巧出现的 '//' 或 '/*' 被误识别为注释\\n def remove_strings_and_urls(text: str) -> str:\\n # 替换常见 URL 协议头，防止 http:// 中的 // 被识别为注释\\n text = text.replace('http://', 'http__').replace('https://', 'https__')\\n # 将双引号、单引号、反引号内的字符串内容替换为下划线 '_'\\n text = re.sub(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '_', text)\\n text = re.sub(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\", '_', text)\\n text = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '_', text)\\n return text\\n\\n # 核心检查逻辑：验证一组代码行是否符合要求\\n def check_lines(lines: list, ext_or_lang: str) -> bool:\\n # 根据文件扩展名或语言类型分配可能出现的合法注释符\\n if ext_or_lang in ['.js', '.ts', '.jsx', '.tsx', 'js', 'javascript', 'ts', 'typescript']:\\n tokens = ['//', '/*']\\n elif ext_or_lang in ['.css', 'css']:\\n tokens = ['/*']\\n elif ext_or_lang in ['.html', '.htm', 'html', 'htm']:\\n # HTML 文件中可能内嵌 JS 和 CSS，所以放宽注释符校验\\n tokens = ['<!--', '//', '/*']\\n elif ext_or_lang in ['.py', 'python']:\\n tokens = ['#']\\n else:\\n tokens = ['//', '/*', '<!--', '#']\\n \\n for line in lines:\\n stripped = line.strip()\\n # 忽略纯空行，空行不属于“代码行”范畴\\n if not stripped:\\n continue\\n \\n # 1. 检查是否为独立注释行（注释写在代码行之上或之下）\\n is_standalone = False\\n for token in tokens:\\n if stripped.startswith(token):\\n is_standalone = True\\n break\\n \\n if is_standalone:\\n return False # 发现独立注释行，不符合“不允许写在代码行之上或之下”\\n \\n # 2. 检查是否有追加注释（每行代码都必须有注释）\\n line_no_str = remove_strings_and_urls(line)\\n has_comment = False\\n for token in tokens:\\n if token in line_no_str:\\n has_comment = True\\n break\\n \\n if not has_comment:\\n return False # 发现没有注释的代码行，不符合“每行代码都要求写注释”\\n \\n return True\\n\\n # 1. 优先扫描工作区中的代码文件\\n code_extensions = ['.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx', '.py']\\n files_checked = 0\\n\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n if not check_lines(lines, ext):\\n return False\\n files_checked += 1\\n except Exception:\\n pass\\n \\n if files_checked > 0:\\n return True\\n \\n # 2. 如果工作区未找到文件，作为回退方案，检查 response 中的 Markdown 代码块\\n blocks = re.findall(r'```(.*?)\\\\n(.*?)```', response, re.DOTALL)\\n if not blocks:\\n return False # 未发现任何代码\\n \\n for lang, code in blocks:\\n lang = lang.lower().strip()\\n if not lang:\\n lang = 'txt'\\n if not check_lines(code.split('\\\\n'), lang):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"编辑已有文件时，要求必须通过输出文件的全量代码来进行覆盖替换，不可使用Unified Diff或其他类型的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["验证页面基础UI样式：不进行任何操作，直接查看页面布局。预期：页面宽度适配手机端，主色调为清新的绿色。页面顶部包含绿色背景的统计卡片，显示有'总支出'和'结余'的金额信息。若布局变形、主色调非绿色或缺少对应文字元素则视为错误。","验证添加面板的唤出与基础内容：点击页面底部的'+'按钮。预期：屏幕从底部滑动弹出一个面板。截屏验证面板内容需包含：顶部的'支出'和'收入'单选切换标签、分类选择区域、金额输入框、日期选择器（默认显示当前当天日期）以及备注输入框，底部包含'保存'和'取消'按钮。缺少任何一个元素或未从底部滑出即为不通过。","验证添加面板的关闭交互：在唤出面板后，点击面板的'取消'按钮，或点击面板外部的空白遮罩区域。预期：截屏验证，面板向下收起并隐藏，页面恢复至初始状态，不会额外添加任何空账单。","验证金额输入框的类型限制：唤出添加面板，尝试在金额输入框内输入英文字母（如abc）或特殊符号。预期：键盘输入被拦截或自动清空，输入框内只允许显示阿拉伯数字及小数点。","验证金额空值或零值阻断校验：唤出添加面板，不在金额框输入任何值（或输入数字0），直接点击底部的'保存'按钮。预期：界面应出现金额不能为0或空的错误提示（弹窗或文字提示），并且面板保持打开状态，本地localStorage和列表均未生成新记录。","验证单条完整记账表单的提交流程：在添加面板正确填写一笔账单数据（例如金额100，选择某个分类并填写备注），点击'保存'。预期：面板成功关闭隐藏，在页面中部的空白列表区域正确渲染出该笔100元的记账条目展示。","验证顶部统计金额自动计算联动功能：在空数据下依次添加一笔100元的'收入'账单和一笔30元的'支出'账单。预期：截取顶部统计卡片查看，'总支出'的数值应变为30.00，'结余'的数值应被正确更新为70.00（100-30）。若数值计算错误或未实时刷新即为不通过。","验证账单列表的按日期倒序分组功能：添加至少两条属于不同日期（例如今天和昨天）的账单记录。预期：截屏查看列表区结构，数据按日期从近到远排序，并分为独立的区块，每个区块头部存在明显的日期分割线（如包含'10月24日 星期二'字样）以及当前日期的收支小计数值展示。","验证单条账单内部详细UI的排版与颜色规范：查看列表中某一条具体的已添加账单。预期：条目左侧必须为分类名称，中间显示用户输入的备注文字，右侧右对齐显示金额。如果是收入，金额文本颜色须为绿色且带正号（如+100）；如果是支出，颜色应为黑色或红色，且带负号（如-30）。","验证账单删除功能及数据同步联动：对列表中任一已存在条目进行向左滑动操作（或点击条目上的删除小图标）。预期：首先弹出一个浏览器的原生确认对话框，点击'确认'后该账单条目从页面彻底消失，同时截屏核对顶部的'结余'和'总支出'金额是否已经同步扣除了该笔删除的数额。","验证账单编辑更新功能：点击列表中某条已经存在的账单条目。预期：自动唤起'添加账单'面板，且面板内的金额、分类、日期、备注等字段应自动回填了该账单的原有数据。将金额修改为另一数值后点击'保存'，截屏确认列表中原有的该条目数值已被更新，且没有另外新增第二条雷同记录。","验证月度预算设置的UI入口与状态切换：点击顶部统计卡片中'总支出'下方的'设置月度预算'文字。预期：弹出一个输入框供设置预算数值，输入大于当前支出的金额（例如1000）并保存，原有的设置文字位置随即被替换显示为一个水平进度的进度条组件。","验证预算进度条在安全比例下的正常状态展示：设置月度预算为1000后，添加一笔金额为500的支出账单（占比为50%）。预期：截屏查看顶部预算进度条，进度条走到正中间50%位置，颜色应呈现为非警告类型的正常颜色（如绿色或蓝色等），不具备预警提示表现。","验证预算进度条在超过80%时的橙色警告状态：继续添加一笔支出，使得当月总支出达到总预算的80%至100%区间（例如总预算1000，当前累计总支出达到850）。预期：截屏查看预算进度条的UI呈现，进度条本体颜色必须变更为橙色作为即将超支的警告。","验证预算进度条在超过100%时的红色超支状态：再次添加一笔大额支出，使当月总支出彻底突破设定预算总额（例如预算1000，当前总支出1100）。预期：截屏查看预算进度条，进度条颜色必须变更为醒目的红色，且在进度条附近必须显式展示出超支金额的提示性文字。","验证月份选择组件的基础外观与交互呈现：查看统计卡片与账单列表交界区域。预期：页面必须在此区域提供一个月份选择器（呈现形式类似 '< 2023年10月 >'），具有显示当前选定月份文字及左右两个可点击的切换箭头图标。","验证切换月份联动页面全局数据刷新功能：点击月份选择器的左箭头图标切换到一个未产生任何记账记录的上一月份。预期：截屏确认下方账单列表被完全清空，顶部的'总支出'和'结余'数值重置归零，且预算进度条的消耗百分比亦重新计算归零，页面全局必须严格依附该目标月份的数据表现。","验证收支类型维度的选项卡筛选功能：在同时包含收入与支出的多记录当前月份列表中，点击列表上方的'支出'选项卡。预期：列表执行动态过滤，视觉上立即隐藏所有'收入'性质的条目，仅显示'支出'类型的条目；再次点击'全部'选项卡后，列表恢复显示全部记录。","验证文本模糊搜索过滤的实时功能：在过滤选项卡右侧的搜索输入框内，输入某个特定账单条目的特有'备注'关键词或'分类'名称。预期：随着键盘字符的输入，下方账单列表须实时响应过滤，仅保留并展示其备注或分类含有该输入关键词的账单条目，其余不相关的记录即刻隐藏。","验证底部Tab导航栏及统计视图切换交互：观察底部的区域，原单调的'+'按钮区域应改造为具备'明细'、中间突出'+'按钮、'统计'三个入口的底部导航Tab栏。此时点击右侧的'统计'Tab。预期：下方的明细账单列表被立刻隐藏，屏幕的主体部分随即展现出全新的统计分析视图界面。","验证统计视图内原生Canvas饼图渲染是否实现：确保在有当月支出数据的前提下进入'统计'Tab视图。预期：查看屏幕能成功展示出一个依据当前选中月份各支出分类金额占比绘制的饼状图。该饼图须通过DOM审查证明其为一个原生的<canvas>元素实现绘制，且未使用诸如Echarts的第三方外部库。","验证统计视图饼图下方的排行榜UI呈现：在展示饼图的'统计'Tab视图中，向下滚动查看图表下方的区域。预期：能够看到按当月各分类支出金额从大到小向下排列的排行榜列表展现。榜单需清楚标明对应的分类名称及总计金额，且名次排序由大至小逻辑正确。","验证设置页侧边抽屉组件的基础呈现：点击页面右上角的'设置'（齿轮样式）图标。预期：页面从侧边滑动弹出一个抽屉面板（或侧边栏），面板内部清晰列出'导出备份'和'导入恢复'两个功能交互按钮项。","验证全量数据以JSON文件导出备份的功能：在记账本内已存在多条账单数据的前提下，打开设置抽屉并点击'导出备份'按钮。预期：浏览器会成功触发本地文件下载行为，并将全量本地账单数据持久化保存为一个格式规范的.json后缀文件至用户设备中。","验证本地JSON数据文件导入及全局页面恢复功能：先清空当前页面的所有账单数据至0状态，打开设置抽屉点击'导入恢复'按钮，选中刚刚备份导出的合法JSON记录文件。预期：文件被正确读取与解析后，当前页面的总支出、结余卡片及下方账单列表被立刻全局刷新重绘，完整呈现刚导入文件内所含有的账单明细记录。","验证应用在无数据状态下的缺省页友好提示逻辑：在全新初次加载无记录或是手动彻底清空所有账单的场景下，查看页面的主账单明细列表区域。预期：该片中间留白区域展现一段内容类似'暂无账单，快去记一笔吧'的缺省图文或文字提示信息；一旦存在任何账单，该提示必须消失不可见。"],"string":"[\n \"验证页面基础UI样式：不进行任何操作，直接查看页面布局。预期：页面宽度适配手机端，主色调为清新的绿色。页面顶部包含绿色背景的统计卡片，显示有'总支出'和'结余'的金额信息。若布局变形、主色调非绿色或缺少对应文字元素则视为错误。\",\n \"验证添加面板的唤出与基础内容：点击页面底部的'+'按钮。预期：屏幕从底部滑动弹出一个面板。截屏验证面板内容需包含：顶部的'支出'和'收入'单选切换标签、分类选择区域、金额输入框、日期选择器（默认显示当前当天日期）以及备注输入框，底部包含'保存'和'取消'按钮。缺少任何一个元素或未从底部滑出即为不通过。\",\n \"验证添加面板的关闭交互：在唤出面板后，点击面板的'取消'按钮，或点击面板外部的空白遮罩区域。预期：截屏验证，面板向下收起并隐藏，页面恢复至初始状态，不会额外添加任何空账单。\",\n \"验证金额输入框的类型限制：唤出添加面板，尝试在金额输入框内输入英文字母（如abc）或特殊符号。预期：键盘输入被拦截或自动清空，输入框内只允许显示阿拉伯数字及小数点。\",\n \"验证金额空值或零值阻断校验：唤出添加面板，不在金额框输入任何值（或输入数字0），直接点击底部的'保存'按钮。预期：界面应出现金额不能为0或空的错误提示（弹窗或文字提示），并且面板保持打开状态，本地localStorage和列表均未生成新记录。\",\n \"验证单条完整记账表单的提交流程：在添加面板正确填写一笔账单数据（例如金额100，选择某个分类并填写备注），点击'保存'。预期：面板成功关闭隐藏，在页面中部的空白列表区域正确渲染出该笔100元的记账条目展示。\",\n \"验证顶部统计金额自动计算联动功能：在空数据下依次添加一笔100元的'收入'账单和一笔30元的'支出'账单。预期：截取顶部统计卡片查看，'总支出'的数值应变为30.00，'结余'的数值应被正确更新为70.00（100-30）。若数值计算错误或未实时刷新即为不通过。\",\n \"验证账单列表的按日期倒序分组功能：添加至少两条属于不同日期（例如今天和昨天）的账单记录。预期：截屏查看列表区结构，数据按日期从近到远排序，并分为独立的区块，每个区块头部存在明显的日期分割线（如包含'10月24日 星期二'字样）以及当前日期的收支小计数值展示。\",\n \"验证单条账单内部详细UI的排版与颜色规范：查看列表中某一条具体的已添加账单。预期：条目左侧必须为分类名称，中间显示用户输入的备注文字，右侧右对齐显示金额。如果是收入，金额文本颜色须为绿色且带正号（如+100）；如果是支出，颜色应为黑色或红色，且带负号（如-30）。\",\n \"验证账单删除功能及数据同步联动：对列表中任一已存在条目进行向左滑动操作（或点击条目上的删除小图标）。预期：首先弹出一个浏览器的原生确认对话框，点击'确认'后该账单条目从页面彻底消失，同时截屏核对顶部的'结余'和'总支出'金额是否已经同步扣除了该笔删除的数额。\",\n \"验证账单编辑更新功能：点击列表中某条已经存在的账单条目。预期：自动唤起'添加账单'面板，且面板内的金额、分类、日期、备注等字段应自动回填了该账单的原有数据。将金额修改为另一数值后点击'保存'，截屏确认列表中原有的该条目数值已被更新，且没有另外新增第二条雷同记录。\",\n \"验证月度预算设置的UI入口与状态切换：点击顶部统计卡片中'总支出'下方的'设置月度预算'文字。预期：弹出一个输入框供设置预算数值，输入大于当前支出的金额（例如1000）并保存，原有的设置文字位置随即被替换显示为一个水平进度的进度条组件。\",\n \"验证预算进度条在安全比例下的正常状态展示：设置月度预算为1000后，添加一笔金额为500的支出账单（占比为50%）。预期：截屏查看顶部预算进度条，进度条走到正中间50%位置，颜色应呈现为非警告类型的正常颜色（如绿色或蓝色等），不具备预警提示表现。\",\n \"验证预算进度条在超过80%时的橙色警告状态：继续添加一笔支出，使得当月总支出达到总预算的80%至100%区间（例如总预算1000，当前累计总支出达到850）。预期：截屏查看预算进度条的UI呈现，进度条本体颜色必须变更为橙色作为即将超支的警告。\",\n \"验证预算进度条在超过100%时的红色超支状态：再次添加一笔大额支出，使当月总支出彻底突破设定预算总额（例如预算1000，当前总支出1100）。预期：截屏查看预算进度条，进度条颜色必须变更为醒目的红色，且在进度条附近必须显式展示出超支金额的提示性文字。\",\n \"验证月份选择组件的基础外观与交互呈现：查看统计卡片与账单列表交界区域。预期：页面必须在此区域提供一个月份选择器（呈现形式类似 '< 2023年10月 >'），具有显示当前选定月份文字及左右两个可点击的切换箭头图标。\",\n \"验证切换月份联动页面全局数据刷新功能：点击月份选择器的左箭头图标切换到一个未产生任何记账记录的上一月份。预期：截屏确认下方账单列表被完全清空，顶部的'总支出'和'结余'数值重置归零，且预算进度条的消耗百分比亦重新计算归零，页面全局必须严格依附该目标月份的数据表现。\",\n \"验证收支类型维度的选项卡筛选功能：在同时包含收入与支出的多记录当前月份列表中，点击列表上方的'支出'选项卡。预期：列表执行动态过滤，视觉上立即隐藏所有'收入'性质的条目，仅显示'支出'类型的条目；再次点击'全部'选项卡后，列表恢复显示全部记录。\",\n \"验证文本模糊搜索过滤的实时功能：在过滤选项卡右侧的搜索输入框内，输入某个特定账单条目的特有'备注'关键词或'分类'名称。预期：随着键盘字符的输入，下方账单列表须实时响应过滤，仅保留并展示其备注或分类含有该输入关键词的账单条目，其余不相关的记录即刻隐藏。\",\n \"验证底部Tab导航栏及统计视图切换交互：观察底部的区域，原单调的'+'按钮区域应改造为具备'明细'、中间突出'+'按钮、'统计'三个入口的底部导航Tab栏。此时点击右侧的'统计'Tab。预期：下方的明细账单列表被立刻隐藏，屏幕的主体部分随即展现出全新的统计分析视图界面。\",\n \"验证统计视图内原生Canvas饼图渲染是否实现：确保在有当月支出数据的前提下进入'统计'Tab视图。预期：查看屏幕能成功展示出一个依据当前选中月份各支出分类金额占比绘制的饼状图。该饼图须通过DOM审查证明其为一个原生的<canvas>元素实现绘制，且未使用诸如Echarts的第三方外部库。\",\n \"验证统计视图饼图下方的排行榜UI呈现：在展示饼图的'统计'Tab视图中，向下滚动查看图表下方的区域。预期：能够看到按当月各分类支出金额从大到小向下排列的排行榜列表展现。榜单需清楚标明对应的分类名称及总计金额，且名次排序由大至小逻辑正确。\",\n \"验证设置页侧边抽屉组件的基础呈现：点击页面右上角的'设置'（齿轮样式）图标。预期：页面从侧边滑动弹出一个抽屉面板（或侧边栏），面板内部清晰列出'导出备份'和'导入恢复'两个功能交互按钮项。\",\n \"验证全量数据以JSON文件导出备份的功能：在记账本内已存在多条账单数据的前提下，打开设置抽屉并点击'导出备份'按钮。预期：浏览器会成功触发本地文件下载行为，并将全量本地账单数据持久化保存为一个格式规范的.json后缀文件至用户设备中。\",\n \"验证本地JSON数据文件导入及全局页面恢复功能：先清空当前页面的所有账单数据至0状态，打开设置抽屉点击'导入恢复'按钮，选中刚刚备份导出的合法JSON记录文件。预期：文件被正确读取与解析后，当前页面的总支出、结余卡片及下方账单列表被立刻全局刷新重绘，完整呈现刚导入文件内所含有的账单明细记录。\",\n \"验证应用在无数据状态下的缺省页友好提示逻辑：在全新初次加载无记录或是手动彻底清空所有账单的场景下，查看页面的主账单明细列表区域。预期：该片中间留白区域展现一段内容类似'暂无账单，快去记一笔吧'的缺省图文或文字提示信息；一旦存在任何账单，该提示必须消失不可见。\"\n]","isConversational":false}}},{"rowIdx":83,"cells":{"task_id":{"kind":"number","value":84,"string":"84"},"repository_policy":{"kind":"string","value":"以下是你在后续所有轮次的交互中必须始终严格遵循的约束条件：\n1. 你生成给用户的回复内容，必须一直使用中文。\n2. 你所生成的代码文件中绝对不能使用多层if。\n3. 对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层。\n4. 要求输出代码时，统一使用 4 个空格作为缩进标准。\n5. 生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式。\n6. 请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数。\n7. 除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释。\n8. 生成代码文件时，务必让每一个文件名都以 my_ 开始。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"要求输出代码时，统一使用 4 个空格作为缩进标准","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n import os\n\n codes = []\n \n # 提取回复中的Markdown代码块\n pattern = re.compile(r'```[^\\n]*\\n(.*?)```', re.DOTALL)\n codes.extend(pattern.findall(response))\n \n # 常见代码文件后缀名推断\n valid_extensions = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts',\n '.jsx', '.tsx', '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.html', '.css'\n }\n \n # 如果存在项目路径，则读取生成的相关代码文件\n if workspace_path and os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n codes.append(f.read())\n except Exception:\n pass\n\n # 如果没有任何代码，则默认满足要求\n if not codes:\n return True\n \n for code in codes:\n lines = code.splitlines()\n for line in lines:\n stripped = line.strip()\n \n # 忽略空行\n if not stripped:\n continue\n \n # 忽略Java/C/C++等语言中常见的多行块注释的中间行，以防误判\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('*/'):\n continue\n \n # 提取该行的前导空白字符\n leading_chars = line[:len(line) - len(line.lstrip())]\n \n # 如果存在制表符，直接判定不符合要求\n if '\\t' in leading_chars:\n return False\n \n # 验证空格数是否严格为4的倍数\n if len(leading_chars) % 4 != 0:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def is_lower_camel_case(name: str) -> bool:\n # 小驼峰命名规则：首字母小写，只能包含字母和数字，不得包含下划线等\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\n \n def get_python_functions(content: str) -> list:\n try:\n tree = ast.parse(content)\n return [\n node.name for node in ast.walk(tree) \n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef))\n ]\n except SyntaxError:\n # 当代码存在语法错误无法正常通过 AST 解析时，回退到正则提取\n return re.findall(r'^[ \\t]*def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(', content, re.MULTILINE)\n\n # 兼容多种语言函数的正则表达式匹配模式\n func_patterns = [\n # JavaScript, PHP, TypeScript: function name(\n r'\\bfunction\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(',\n # Go: func name(\n r'\\bfunc\\s+(?:\\([^)]+\\)\\s+)?([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(',\n # JavaScript 箭头/匿名函数赋值: const name = function() 或 const name = () =>\n r'\\b(?:const|let|var)\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*=\\s*(?:function\\b|\\([^)]*\\)\\s*=>)',\n # C/C++/Java/C# 等: [修饰符] 返回类型 函数名(参数) {\n r'^[ \\t]*(?:(?:public|private|protected|static|virtual|inline|explicit|friend|async)\\s+)*[a-zA-Z_][a-zA-Z0-9_<>\\[\\]\\*&]+\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\([^)]*\\)\\s*(?:const\\s*)?(?:throws\\s+[a-zA-Z0-9_, \\t\\n]+\\s*)?\\{'\n ]\n \n # 忽略控制流关键字（防止正则表达式由于宽泛导致误匹配）\n ignore_keywords = {'if', 'for', 'while', 'switch', 'catch', 'return'}\n \n # 常见的代码文件后缀\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.cs', \n '.js', '.ts', '.go', '.rb', '.php', '.swift', \n '.kt', '.rs', '.dart'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n funcs = []\n if ext == '.py':\n funcs.extend(get_python_functions(content))\n else:\n for pattern in func_patterns:\n # 针对以 ^ 开头锚定的正则，使用 MULTILINE 多行模式匹配\n if pattern.startswith('^'):\n funcs.extend(re.findall(pattern, content, re.MULTILINE))\n else:\n funcs.extend(re.findall(pattern, content))\n \n for func_name in funcs:\n if func_name in ignore_keywords:\n continue\n # 若存在任何一个函数名称不符合小驼峰格式要求，则立即返回 False\n if not is_lower_camel_case(func_name):\n return False\n \n # 如果项目中所有的函数都被成功检查且皆符合小驼峰格式（或者没有定义函数）\n return True","tags":["Style","Naming"]},{"constraint":"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数","validation_code":"def check_class_name_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码文件后缀名，以便过滤提取\n code_extensions = {\n '.py', '.java', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs', \n '.js', '.ts', '.php', '.rb', '.swift', '.kt', '.dart'\n }\n \n # 定义正则表达式模式以匹配类名：匹配 \"class\" 关键字，后跟空白字符，随后是类名\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_]\\w*)')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找当前代码文件中的所有类名\n class_names = class_pattern.findall(content)\n for class_name in class_names:\n # 根据要求统计“字符长度”并验证是否为偶数\n if len(class_name) % 2 != 0:\n return False\n except Exception:\n # 忽略编码问题或其他无法读取的文件\n continue\n \n return True","tags":["Style","Naming"]},{"constraint":"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 建立常见代码文件扩展名及其对应的注释符号映射\n ext_comment_symbols = {\n '.py': ['#', '\"\"\"', \"'''\"],\n '.rb': ['#'],\n '.sh': ['#'],\n '.bash': ['#'],\n '.java': ['//', '/*', '*/'],\n '.js': ['//', '/*', '*/'],\n '.ts': ['//', '/*', '*/'],\n '.jsx': ['//', '/*', '*/'],\n '.tsx': ['//', '/*', '*/'],\n '.c': ['//', '/*', '*/'],\n '.cpp': ['//', '/*', '*/'],\n '.h': ['//', '/*', '*/'],\n '.hpp': ['//', '/*', '*/'],\n '.cs': ['//', '/*', '*/'],\n '.go': ['//', '/*', '*/'],\n '.swift': ['//', '/*', '*/'],\n '.kt': ['//', '/*', '*/'],\n '.rs': ['//', '/*', '*/'],\n '.php': ['//', '/*', '*/', '#'],\n '.sql': ['--', '/*', '*/'],\n '.html': ['<!--', '-->', '//', '/*', '*/'],\n '.css': ['/*', '*/'],\n '.vue': ['<!--', '-->', '//', '/*', '*/'],\n '.m': ['//', '/*', '*/'],\n '.lua': ['--']\n }\n \n def remove_strings(text: str) -> str:\n # 保护常见的网址形式中的 // 不被当做注释符误判\n text = text.replace('http://', 'http_URL_')\n text = text.replace('https://', 'https_URL_')\n # 去除双引号内的字符串字面量（兼容包含转义的写法）\n text = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', text)\n # 去除单引号内的字符串字面量\n text = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", text)\n # 去除反引号内的字符串字面量 (如 JS 模板字符串)\n text = re.sub(r'`(?:\\\\.|[^`\\\\])*`', '``', text)\n return text\n\n # 遍历 workspace_path 下所有的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in ext_comment_symbols:\n symbols = ext_comment_symbols[ext]\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n # 逐行检查是否含有注释\n for i, line in enumerate(lines):\n # 第 5 行（索引为 4）允许包含注释，跳过检查\n if i == 4: \n continue\n \n # 剔除行内的字符串内容后，检查是否仍存在注释符号\n cleaned_line = remove_strings(line)\n for sym in symbols:\n if sym in cleaned_line:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成代码文件时，务必让每一个文件名都以 my_ 开始","validation_code":"def check_file_names(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n has_valid_files = False\n \n # 遍历代码项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 排除隐藏文件（如 .gitignore 等）\n if file_name.startswith('.'):\n continue\n \n has_valid_files = True\n \n # 判断文件名是否以 'my_' 开始\n if not file_name.startswith('my_'):\n return False\n \n # 只要存在有效文件且均以 'my_' 开头即返回 True\n return has_valid_files","tags":["Style","Naming"]}],"string":"[\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用 4 个空格作为缩进标准\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n import os\\n\\n codes = []\\n \\n # 提取回复中的Markdown代码块\\n pattern = re.compile(r'```[^\\\\n]*\\\\n(.*?)```', re.DOTALL)\\n codes.extend(pattern.findall(response))\\n \\n # 常见代码文件后缀名推断\\n valid_extensions = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.js', '.ts',\\n '.jsx', '.tsx', '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.html', '.css'\\n }\\n \\n # 如果存在项目路径，则读取生成的相关代码文件\\n if workspace_path and os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n codes.append(f.read())\\n except Exception:\\n pass\\n\\n # 如果没有任何代码，则默认满足要求\\n if not codes:\\n return True\\n \\n for code in codes:\\n lines = code.splitlines()\\n for line in lines:\\n stripped = line.strip()\\n \\n # 忽略空行\\n if not stripped:\\n continue\\n \\n # 忽略Java/C/C++等语言中常见的多行块注释的中间行，以防误判\\n if stripped.startswith('*') or stripped.startswith('/*') or stripped.startswith('*/'):\\n continue\\n \\n # 提取该行的前导空白字符\\n leading_chars = line[:len(line) - len(line.lstrip())]\\n \\n # 如果存在制表符，直接判定不符合要求\\n if '\\\\t' in leading_chars:\\n return False\\n \\n # 验证空格数是否严格为4的倍数\\n if len(leading_chars) % 4 != 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def is_lower_camel_case(name: str) -> bool:\\n # 小驼峰命名规则：首字母小写，只能包含字母和数字，不得包含下划线等\\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\\n \\n def get_python_functions(content: str) -> list:\\n try:\\n tree = ast.parse(content)\\n return [\\n node.name for node in ast.walk(tree) \\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef))\\n ]\\n except SyntaxError:\\n # 当代码存在语法错误无法正常通过 AST 解析时，回退到正则提取\\n return re.findall(r'^[ \\\\t]*def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(', content, re.MULTILINE)\\n\\n # 兼容多种语言函数的正则表达式匹配模式\\n func_patterns = [\\n # JavaScript, PHP, TypeScript: function name(\\n r'\\\\bfunction\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(',\\n # Go: func name(\\n r'\\\\bfunc\\\\s+(?:\\\\([^)]+\\\\)\\\\s+)?([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(',\\n # JavaScript 箭头/匿名函数赋值: const name = function() 或 const name = () =>\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=\\\\s*(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>)',\\n # C/C++/Java/C# 等: [修饰符] 返回类型 函数名(参数) {\\n r'^[ \\\\t]*(?:(?:public|private|protected|static|virtual|inline|explicit|friend|async)\\\\s+)*[a-zA-Z_][a-zA-Z0-9_<>\\\\[\\\\]\\\\*&]+\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\([^)]*\\\\)\\\\s*(?:const\\\\s*)?(?:throws\\\\s+[a-zA-Z0-9_, \\\\t\\\\n]+\\\\s*)?\\\\{'\\n ]\\n \\n # 忽略控制流关键字（防止正则表达式由于宽泛导致误匹配）\\n ignore_keywords = {'if', 'for', 'while', 'switch', 'catch', 'return'}\\n \\n # 常见的代码文件后缀\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.cs', \\n '.js', '.ts', '.go', '.rb', '.php', '.swift', \\n '.kt', '.rs', '.dart'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n funcs = []\\n if ext == '.py':\\n funcs.extend(get_python_functions(content))\\n else:\\n for pattern in func_patterns:\\n # 针对以 ^ 开头锚定的正则，使用 MULTILINE 多行模式匹配\\n if pattern.startswith('^'):\\n funcs.extend(re.findall(pattern, content, re.MULTILINE))\\n else:\\n funcs.extend(re.findall(pattern, content))\\n \\n for func_name in funcs:\\n if func_name in ignore_keywords:\\n continue\\n # 若存在任何一个函数名称不符合小驼峰格式要求，则立即返回 False\\n if not is_lower_camel_case(func_name):\\n return False\\n \\n # 如果项目中所有的函数都被成功检查且皆符合小驼峰格式（或者没有定义函数）\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数\",\n \"validation_code\": \"def check_class_name_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码文件后缀名，以便过滤提取\\n code_extensions = {\\n '.py', '.java', '.cpp', '.cc', '.cxx', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.php', '.rb', '.swift', '.kt', '.dart'\\n }\\n \\n # 定义正则表达式模式以匹配类名：匹配 \\\"class\\\" 关键字，后跟空白字符，随后是类名\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找当前代码文件中的所有类名\\n class_names = class_pattern.findall(content)\\n for class_name in class_names:\\n # 根据要求统计“字符长度”并验证是否为偶数\\n if len(class_name) % 2 != 0:\\n return False\\n except Exception:\\n # 忽略编码问题或其他无法读取的文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 建立常见代码文件扩展名及其对应的注释符号映射\\n ext_comment_symbols = {\\n '.py': ['#', '\\\"\\\"\\\"', \\\"'''\\\"],\\n '.rb': ['#'],\\n '.sh': ['#'],\\n '.bash': ['#'],\\n '.java': ['//', '/*', '*/'],\\n '.js': ['//', '/*', '*/'],\\n '.ts': ['//', '/*', '*/'],\\n '.jsx': ['//', '/*', '*/'],\\n '.tsx': ['//', '/*', '*/'],\\n '.c': ['//', '/*', '*/'],\\n '.cpp': ['//', '/*', '*/'],\\n '.h': ['//', '/*', '*/'],\\n '.hpp': ['//', '/*', '*/'],\\n '.cs': ['//', '/*', '*/'],\\n '.go': ['//', '/*', '*/'],\\n '.swift': ['//', '/*', '*/'],\\n '.kt': ['//', '/*', '*/'],\\n '.rs': ['//', '/*', '*/'],\\n '.php': ['//', '/*', '*/', '#'],\\n '.sql': ['--', '/*', '*/'],\\n '.html': ['<!--', '-->', '//', '/*', '*/'],\\n '.css': ['/*', '*/'],\\n '.vue': ['<!--', '-->', '//', '/*', '*/'],\\n '.m': ['//', '/*', '*/'],\\n '.lua': ['--']\\n }\\n \\n def remove_strings(text: str) -> str:\\n # 保护常见的网址形式中的 // 不被当做注释符误判\\n text = text.replace('http://', 'http_URL_')\\n text = text.replace('https://', 'https_URL_')\\n # 去除双引号内的字符串字面量（兼容包含转义的写法）\\n text = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', text)\\n # 去除单引号内的字符串字面量\\n text = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", text)\\n # 去除反引号内的字符串字面量 (如 JS 模板字符串)\\n text = re.sub(r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`', '``', text)\\n return text\\n\\n # 遍历 workspace_path 下所有的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ext_comment_symbols:\\n symbols = ext_comment_symbols[ext]\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n # 逐行检查是否含有注释\\n for i, line in enumerate(lines):\\n # 第 5 行（索引为 4）允许包含注释，跳过检查\\n if i == 4: \\n continue\\n \\n # 剔除行内的字符串内容后，检查是否仍存在注释符号\\n cleaned_line = remove_strings(line)\\n for sym in symbols:\\n if sym in cleaned_line:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，务必让每一个文件名都以 my_ 开始\",\n \"validation_code\": \"def check_file_names(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_valid_files = False\\n \\n # 遍历代码项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 排除隐藏文件（如 .gitignore 等）\\n if file_name.startswith('.'):\\n continue\\n \\n has_valid_files = True\\n \\n # 判断文件名是否以 'my_' 开始\\n if not file_name.startswith('my_'):\\n return False\\n \\n # 只要存在有效文件且均以 'my_' 开头即返回 True\\n return has_valid_files\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Application Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个刷题系统的管理后台基础页面骨架。需求如下：1. 整体布局分为左、上、右三部分。左侧是导航侧边栏，包含“题库管理”、“用户管理”、“权限管理”三个主菜单项；顶部是Header栏，左侧放系统Logo标题，右侧放一个“当前用户”的占位显示；右侧是主内容区域。2. 请实现基本的路由切换或视图切换逻辑：点击左侧不同菜单时，右侧内容区相应地切换标题（如显示“题库管理内容区”、“用户管理内容区”即可），暂时不需要实现任何具体的业务功能，留出这三个页面的代码扩展位置。\n另外请你后续所有操作始终遵循以下要求：请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）；生成代码时，请将各函数的行数严格控制在 50 行及以内。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"要求输出代码时，统一使用 4 个空格作为缩进标准","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n all_codes = []\n \n # 1. 提取 response 中的 Markdown 代码块\n in_code = False\n current_block = []\n for line in response.split('\\n'):\n if line.strip().startswith('```'):\n if in_code:\n all_codes.append('\\n'.join(current_block))\n current_block = []\n in_code = False\n else:\n in_code = True\n elif in_code:\n current_block.append(line)\n \n # 2. 读取 workspace_path 下生成的代码文件\n if os.path.isdir(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 排除日志文件\n if file.endswith('.log'):\n continue\n \n # 筛选通常的项目代码文件\n ext = os.path.splitext(file)[1].lower()\n valid_exts = ['.html', '.css', '.js', '.py', '.vue', '.ts', '.jsx', '.tsx', '.json', '.java', '.c', '.cpp', '.cs', '.php', '.go']\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n all_codes.append(f.read())\n except Exception:\n pass\n\n # 如果没有任何代码被发现，默认为未违背规则\n if not all_codes:\n return True\n\n invalid_lines_count = 0\n total_indented_lines = 0\n\n # 3. 逐行判断前导缩进\n for code in all_codes:\n lines = code.split('\\n')\n for line in lines:\n # 提取行首空白（同时保证该行不全为空白）\n match = re.match(r'^([ \\t]+)(?=\\S)', line)\n if match:\n ws = match.group(1)\n total_indented_lines += 1\n \n # 如果包含 Tab，直接不符合要求\n if '\\t' in ws:\n return False\n \n # 如果空格数量不是 4 的倍数，记录为不规范行\n if len(ws) % 4 != 0:\n invalid_lines_count += 1\n\n # 4. 判断违规行数占比\n # 允许少数由意外折行造成的对齐误差（阈值设为最多3行或者占比不超过5%），超过则视为未统一使用 4 个空格\n if invalid_lines_count > 0:\n if invalid_lines_count > 3 or (invalid_lines_count / total_indented_lines > 0.05):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_lower_camel(name: str) -> bool:\n # 排除可能被正则误当作方法名抓取的语言保留字或关键字\n BLACKLIST = {\n 'if', 'for', 'while', 'switch', 'catch', 'return', \n 'else', 'elseif', 'elif', 'constructor', 'super', \n 'with', 'function', 'try', 'finally', 'do', 'yield', 'await',\n 'get', 'set'\n }\n if name in BLACKLIST:\n return True\n # 检查是否满足小驼峰格式：以小写字母开头，由大小写字母及数字组成，不能有下划线或其他特殊字符\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\n\n if not os.path.exists(workspace_path):\n return False\n \n # 常用代码文件的后缀集合，过滤掉类似 .log 的非代码文件\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.java', '.c', '.cpp', '.cs', '.go'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n funcs = []\n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n funcs.append(node.name)\n except SyntaxError:\n # 如果代码结构不完整导致 ast 解析失败，则回退到正则提取\n for m in re.finditer(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(', content):\n funcs.append(m.group(1))\n else:\n # 提取 function functionName(...) 形式，包含 Generator function\n for m in re.finditer(r'\\bfunction\\s*\\*?\\s*([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(', content):\n funcs.append(m.group(1))\n \n # 提取 const/let/var funcName = (...) => 或者 = function(...) 形式\n for m in re.finditer(r'\\b(?:const|let|var)\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*=\\s*(?:async\\s+)?(?:function\\s*\\*?\\s*\\(|\\([^)]*\\)\\s*=>|[a-zA-Z_][a-zA-Z0-9_]*\\s*=>)', content):\n funcs.append(m.group(1))\n \n # 提取类的方法或对象的方法 methodName(...) { \n # 采用行首及缩进开始限制，防止误匹配部分不相关的中间代码\n for m in re.finditer(r'^\\s*(?:async\\s+)?(?:get\\s+|set\\s+)?([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\([^)]*\\)\\s*\\{', content, re.MULTILINE):\n funcs.append(m.group(1))\n \n # 提取 Vue 等对象属性风格的方法声明 propertyName: function(...) {\n for m in re.finditer(r'\\b([a-zA-Z_][a-zA-Z0-9_]*)\\s*:\\s*(?:async\\s+)?function\\s*\\*?\\s*\\(', content):\n funcs.append(m.group(1))\n \n # 依次对当前文件提取到的所有函数名进行判定\n for func_name in funcs:\n if not check_lower_camel(func_name):\n return False\n\n # 若所有提取出的函数都符合小驼峰要求，或未包含函数，均视为满足要求\n return True","tags":["Style","Naming"]},{"constraint":"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 验证工作区路径是否存在且为目录\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义正则表达式匹配面向对象语言中的类声明\n # 匹配诸如 `class MyClass`, `class MyClass(object):`, `class MyClass {` 中的 `MyClass`\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_]\\w*)')\n \n # 定义需要扫描的代码文件后缀集合\n valid_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.cpp', '.c', '.h', '.cs',\n '.php', '.rb', '.go', '.swift', '.html', '.htm', '.vue'\n }\n\n # 遍历当前项目的工作区文件\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）以及常见的依赖库和构建目录，防止误判第三方代码库的类名\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in {'node_modules', 'venv', 'env', 'dist', 'build', '__pycache__'}]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找当前文件中的所有类名声明\n for match in class_pattern.finditer(content):\n cls_name = match.group(1)\n # 如果存在的任意一个类名的字符长度不是偶数，则判定不满足要求\n if len(cls_name) % 2 != 0:\n return False\n except Exception:\n # 忽略因编码等问题无法读取的文件\n pass\n\n # 如果代码中所有被提取的类名长度都恰好是偶数（或者完全没有写任何类名），均满足题目要求\n return True","tags":["Style","Naming"]},{"constraint":"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 针对 JavaScript, Java, C, C++, CSS 等使用的 C-Style 解析器\n def get_comment_lines_c_style(text):\n comment_lines = set()\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n escape = False\n line_num = 1\n \n i = 0\n n = len(text)\n while i < n:\n char = text[i]\n is_escaped = escape\n \n if escape:\n escape = False\n else:\n if char == '\\\\':\n escape = True\n else:\n if not in_string and not in_line_comment and not in_block_comment:\n if char in ('\"', \"'\", \"`\"):\n in_string = True\n string_char = char\n elif char == '/' and i + 1 < n:\n if text[i+1] == '/':\n in_line_comment = True\n comment_lines.add(line_num)\n i += 1\n elif text[i+1] == '*':\n in_block_comment = True\n comment_lines.add(line_num)\n i += 1\n elif in_string:\n if char == string_char:\n in_string = False\n elif in_line_comment:\n if char != '\\n':\n comment_lines.add(line_num)\n elif in_block_comment:\n comment_lines.add(line_num)\n if char == '*' and i + 1 < n and text[i+1] == '/':\n in_block_comment = False\n i += 1\n \n if in_line_comment and char == '\\n':\n in_line_comment = False\n\n if char == '\\n':\n # 单行字符串换行自动跳出，增强容错能力\n if in_string and string_char != '`' and not is_escaped:\n in_string = False\n line_num += 1\n \n i += 1\n return comment_lines\n\n # 针对 Python, Ruby 等使用的解析器\n def get_comment_lines_python(text):\n comment_lines = set()\n in_string = False\n string_char = ''\n in_comment = False\n escape = False\n line_num = 1\n \n i = 0\n n = len(text)\n while i < n:\n char = text[i]\n is_escaped = escape\n \n if escape:\n escape = False\n else:\n if char == '\\\\':\n escape = True\n else:\n if not in_string and not in_comment:\n if text[i:i+3] in ('\"\"\"', \"'''\"):\n in_string = True\n string_char = text[i:i+3]\n i += 2\n elif char in ('\"', \"'\"):\n in_string = True\n string_char = char\n elif char == '#':\n in_comment = True\n comment_lines.add(line_num)\n elif in_string:\n if string_char in ('\"\"\"', \"'''\"):\n if text[i:i+3] == string_char:\n in_string = False\n i += 2\n elif char == string_char:\n in_string = False\n elif in_comment:\n if char != '\\n':\n comment_lines.add(line_num)\n \n if in_comment and char == '\\n':\n in_comment = False\n \n if char == '\\n':\n if in_string and string_char not in ('\"\"\"', \"'''\") and not is_escaped:\n in_string = False\n line_num += 1\n \n i += 1\n return comment_lines\n\n # 针对 HTML 标签使用的解析器\n def get_comment_lines_html(text):\n comment_lines = set()\n in_comment = False\n line_num = 1\n i = 0\n n = len(text)\n while i < n:\n if not in_comment:\n if text[i:i+4] == '<!--':\n in_comment = True\n comment_lines.add(line_num)\n i += 3\n else:\n comment_lines.add(line_num)\n if text[i:i+3] == '-->':\n in_comment = False\n i += 2\n \n if text[i] == '\\n':\n line_num += 1\n \n i += 1\n return comment_lines\n\n # 检查工作空间路径\n if not os.path.exists(workspace_path):\n return False\n \n # 定义常见的代码扩展名\n CODE_EXTS = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.c', '.cpp', '.cs', \n '.css', '.html', '.htm', '.vue', '.go', '.rs', '.rb', '.php', '.swift', '.kt'\n }\n \n # 遍历项目下所有文件进行判断\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in CODE_EXTS:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n comment_lines = set()\n \n # 针对不同扩展名进行各自适应的解析判断\n if ext in {'.py', '.rb'}:\n comment_lines.update(get_comment_lines_python(content))\n elif ext in {'.html', '.htm'}:\n comment_lines.update(get_comment_lines_html(content))\n comment_lines.update(get_comment_lines_c_style(content))\n elif ext in {'.vue', '.php'}:\n comment_lines.update(get_comment_lines_html(content))\n comment_lines.update(get_comment_lines_c_style(content))\n comment_lines.update(get_comment_lines_python(content))\n else:\n comment_lines.update(get_comment_lines_c_style(content))\n \n # 若发现任何一行不是第5行的代码包含注释，证明未满足要求\n for line in comment_lines:\n if line != 5:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成代码文件时，务必让每一个文件名都以 my_ 开始","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名，用于识别哪些文件是代码文件\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', \n '.ts', '.tsx', '.vue', '.py', '.java', \n '.c', '.cpp', '.cc', '.h', '.hpp', \n '.go', '.rs', '.php', '.rb', '.swift'\n }\n \n # 记录是否实际生成了代码文件\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 获取文件扩展名并统一转换为小写\n ext = os.path.splitext(file_name)[1].lower()\n \n # 如果该文件是代码文件\n if ext in code_extensions:\n has_code_file = True\n # 检查文件名是否以 my_ 开始\n if not file_name.startswith('my_'):\n return False\n \n # 确保生成了代码文件，且所有代码文件均满足要求\n return has_code_file","tags":["Style","Naming"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在 50 行及以内","validation_code":"import os\nimport re\nimport ast\n\ndef check(response: str, workspace_path: str) -> bool:\n def check_py_file(file_path: str) -> bool:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\n lines_count = node.end_lineno - node.lineno + 1\n if lines_count > 50:\n return False\n except Exception:\n pass\n return True\n\n def check_js_content(content: str) -> bool:\n # 正则表达式匹配可能属于函数的特征起始\n pattern = re.compile(\n r'(\\bfunction\\b|=>\\s*\\{|\\b(?!(?:if|for|while|switch|catch)\\b)[a-zA-Z_$][\\w$]*\\s*\\([^)]*\\)\\s*\\{)'\n )\n \n length = len(content)\n i = 0\n while i < length:\n match = pattern.search(content, i)\n if not match:\n break\n \n start = match.start()\n brace_index = content.find('{', start)\n if brace_index == -1:\n i = start + 1\n continue\n \n brace_count = 0\n in_str = False\n str_char = ''\n in_line_comment = False\n in_block_comment = False\n \n end_index = -1\n j = brace_index\n while j < length:\n c = content[j]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n elif in_block_comment:\n if c == '*' and j + 1 < length and content[j+1] == '/':\n in_block_comment = False\n j += 1\n elif in_str:\n if c == '\\\\':\n j += 1\n elif c == str_char:\n in_str = False\n else:\n if c == '/' and j + 1 < length:\n if content[j+1] == '/':\n in_line_comment = True\n j += 1\n elif content[j+1] == '*':\n in_block_comment = True\n j += 1\n elif c in ('\"', \"'\", '`'):\n in_str = True\n str_char = c\n elif c == '{':\n brace_count += 1\n elif c == '}':\n brace_count -= 1\n if brace_count == 0:\n end_index = j\n break\n j += 1\n \n if end_index != -1:\n func_text = content[start:end_index+1]\n lines_count = len(func_text.split('\\n'))\n if lines_count > 50:\n return False\n # 继续在原区块内部或者接下来的位置进行寻找嵌套函数\n i = start + 1\n else:\n i = brace_index + 1\n\n return True\n\n def check_file(file_path: str) -> bool:\n if file_path.endswith('.py'):\n return check_py_file(file_path)\n else:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file_path.endswith(('.html', '.htm', '.vue')):\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n for script in scripts:\n if not check_js_content(script):\n return False\n return True\n else:\n return check_js_content(content)\n except Exception:\n pass\n return True\n\n if not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 扫描常见的后端、前端代码文件\n if file.endswith(('.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\n if not check_file(file_path):\n return False\n\n return True","tags":["Quantity","Range"]}]},{"round_id":1,"instruction":"好，接下来在这个基础上丰富“题库管理”页面的界面。请在题库管理内容区实现一个完整的数据操作界面：1. 顶部需要一个操作栏，左边包含一个搜索框和一个题目类型下拉筛选器（选项：单选题、多选题、简答题），右边有一个“上传/新增题目”按钮。2. 下方是一个题目列表表格，包含表头：题目ID、题干简述、题型、难度、状态（已上架/已下架）。3. 表格每一行的最右侧有一列“操作”，里面提供三个文字按钮：“编辑”、“切换上下架”、“删除”。请使用Mock数据渲染出几行表格内容，当点击“删除”按钮时，能把这一行从视图中移除。\n另外请你后续所有操作始终遵循以下新要求：生成代码时，请将各函数的行数严格控制在30行及以内；操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"要求输出代码时，统一使用 4 个空格作为缩进标准","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的代码文件扩展名\n code_exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue', '.java', '.cpp', '.c', '.h', '.hpp', '.cs', '.php', '.rb', '.go', '.rs', '.swift', '.kt'}\n \n total_valid = 0\n total_invalid = 0\n has_tab = False\n \n # 首先检查工作区中的代码文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n stripped = line.strip()\n # 忽略空行\n if not stripped:\n continue\n # 忽略可能是多行注释或其他注释导致的特殊对齐\n if stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('#') or stripped.startswith('<!--'):\n continue\n \n # 匹配行首的空白字符\n match = re.match(r'^([\\t ]+)', line)\n if match:\n indent = match.group(1)\n if '\\t' in indent:\n has_tab = True\n # 判断空格数是否为 4 的倍数\n if len(indent) % 4 == 0:\n total_valid += 1\n else:\n total_invalid += 1\n\n # 如果工作区中有进行过有效缩进统计，直接得出结论\n if total_valid + total_invalid > 0:\n if has_tab:\n return False\n # 考虑到代码换行对齐时的非4倍数空格，允许有一定的容错(阈值设为0.8)\n return (total_valid / (total_valid + total_invalid)) >= 0.8\n \n # 如果工作区没有生成文件或者未找到任何缩进行，退化到检查 response 中的代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n lines = block.split('\\n')\n for line in lines:\n stripped = line.strip()\n if not stripped:\n continue\n if stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('#') or stripped.startswith('<!--'):\n continue\n \n match = re.match(r'^([\\t ]+)', line)\n if match:\n indent = match.group(1)\n if '\\t' in indent:\n has_tab = True\n if len(indent) % 4 == 0:\n total_valid += 1\n else:\n total_invalid += 1\n\n if total_valid + total_invalid > 0:\n if has_tab:\n return False\n return (total_valid / (total_valid + total_invalid)) >= 0.8\n \n # 如果不存在任何具有缩进的代码（比如全部顶格），默认视为没有违反约束\n return True","tags":["Style","Layout"]},{"constraint":"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式","validation_code":"def check_function_names_lower_camel_case(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 判断是否为小驼峰命名：必须以小写字母开头，后续字符只能是字母和数字\n def is_lower_camel_case(name: str) -> bool:\n if not name:\n return False\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\n\n # 去除代码中的注释，防止注释中的内容导致误判\n def remove_comments(content: str) -> str:\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n return content\n\n # 定义前端及常见类C语法的函数匹配正则\n # 匹配：function myFunc (\n pattern_func1 = re.compile(r'\\bfunction\\s+([a-zA-Z_$][\\w$]*)\\s*\\(')\n # 匹配：const myFunc = () => 或者 const myFunc = function() \n pattern_func2 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n # 匹配类方法或对象方法：myMethod() {\n pattern_func4 = re.compile(r'\\b([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{')\n # 匹配对象属性赋值函数：myFunc: () =>\n pattern_func5 = re.compile(r'\\b([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n # 匹配对象属性绑定函数：obj.myFunc = () =>\n pattern_func6 = re.compile(r'\\.\\s*([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)')\n\n # 需要排除的 JS/TS/Java 等的控制流或内置关键字\n keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', 'finally', \n 'return', 'super', 'constructor', 'function', 'with', 'typeof', 'instanceof'\n }\n \n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.go'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n\n if ext == '.py':\n # Python 文件直接使用 AST 分析，绝对精准\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n tree = ast.parse(f.read())\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if not is_lower_camel_case(node.name):\n return False\n except Exception:\n pass\n\n elif ext in code_extensions:\n # 其他代码文件使用正则匹配提取函数名\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n\n content = remove_comments(content)\n functions = []\n \n functions.extend(pattern_func1.findall(content))\n functions.extend(pattern_func2.findall(content))\n functions.extend(pattern_func5.findall(content))\n functions.extend(pattern_func6.findall(content))\n \n methods = pattern_func4.findall(content)\n for m in methods:\n if m not in keywords:\n functions.append(m)\n\n for func in functions:\n if not is_lower_camel_case(func):\n return False\n except Exception:\n pass\n \n return True","tags":["Style","Naming"]},{"constraint":"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数","validation_code":"def check_class_name_length_is_even(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义需要扫描的代码文件后缀集合\n valid_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.cpp', \n '.c', '.h', '.hpp', '.cs', '.php', '.go', '.rs', \n '.swift', '.kt', '.rb', '.html', '.htm', '.vue'\n }\n \n # 匹配 class 定义的正则表达式\n # \\bclass\\s+ : 匹配 'class ' 及其后跟随的空白字符\n # ([A-Za-z_][A-Za-z0-9_]*) : 匹配合法的标识符作为类名\n # \\b : 确保类名被完整匹配并排除相连符号\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z_][A-Za-z0-9_]*)\\b')\n \n # 排除语法关键字，防止在匿名类中将其误认为类名（如 'class extends Base'）\n excluded_keywords = {'extends', 'implements'}\n \n # 遍历当前项目工作空间的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_extensions:\n continue\n \n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 格式读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果文件读取失败（例如编码问题或不可读），则跳过\n continue\n \n # 提取文件中的所有类名\n matches = class_pattern.findall(content)\n for class_name in matches:\n # 跳过关键字\n if class_name in excluded_keywords:\n continue\n \n # 校验提取出的类名字符长度是否恰好是偶数\n if len(class_name) % 2 != 0:\n return False\n \n # 如果项目中没有任何一个类名的长度为奇数（或没有定义类），则符合要求\n return True","tags":["Style","Naming"]},{"constraint":"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码扩展名，用来过滤并仅检查代码文件\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \n '.java', '.c', '.cpp', '.h', '.hpp', '.py', '.sh', \n '.yml', '.yaml', '.css', '.scss', '.less'\n }\n \n # 针对前端、类 C 语言的代码注释匹配模式 (含 HTML/Vue, JS/TS, C/Java 等)\n # 优先匹配字符串字面量以避开引号内注释符；对于 // 单行注释使用 (?<!:) 规避对 http:// 的误判\n pattern_c_like = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)|'\n r'(?P<html_comment><!--[\\s\\S]*?-->)|'\n r'(?P<multi_comment>/\\*[\\s\\S]*?\\*/)|'\n r'(?P<single_comment>(?<!:)//.*)'\n )\n \n # 针对 Python 或脚本语言代码注释匹配模式\n pattern_py = re.compile(\n r'(?P<py_multi_string>\\\"\\\"\\\"[\\s\\S]*?\\\"\\\"\\\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\')|'\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')|'\n r'(?P<hash_comment>#.*)'\n )\n \n # 针对 CSS 等只支持 /* ... */ 注释模式\n pattern_css = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')|'\n r'(?P<multi_comment>/\\*[\\s\\S]*?\\*/)'\n )\n \n # 将正则匹配跨越的文本索引，转化为包含行号的集合（1-based）\n def get_lines_covered(text: str, start: int, end: int):\n lines = text.split('\\n')\n line_starts = []\n current_pos = 0\n for line in lines:\n line_starts.append(current_pos)\n current_pos += len(line) + 1 # 包含 \\n 的长度\n \n start_line = 1\n end_line = 1\n \n for i, pos in enumerate(line_starts):\n if start >= pos:\n start_line = i + 1\n if end > pos:\n end_line = i + 1\n \n return set(range(start_line, end_line + 1))\n\n # 遍历所有项目文件\n for root_dir, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_path = os.path.join(root_dir, file)\n try:\n # 默认使用 Universal Newlines 模式读取，将 \\r\\n 转换成 \\n 使得索引精确映射\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n continue\n \n # 根据文件类型选择对应的正则处理器\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.java', '.c', '.cpp', '.h', '.hpp', '.scss', '.less']:\n pattern = pattern_c_like\n elif ext in ['.py', '.sh', '.yml', '.yaml']:\n pattern = pattern_py\n elif ext in ['.css']:\n pattern = pattern_css\n else:\n continue\n \n # 查找并计算所有注释所在的行号范围\n for match in pattern.finditer(text):\n group_name = match.lastgroup\n # 若捕获组名为注释组（忽略掉前面用来吃掉字符串的 string 组）\n if group_name in ['html_comment', 'multi_comment', 'single_comment', 'hash_comment']:\n covered_lines = get_lines_covered(text, match.start(), match.end())\n for line_num in covered_lines:\n # 发现不仅在第5行出现了注释，或注释覆盖跨度包含了非第5行的位置，则违反要求\n if line_num != 5:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成代码文件时，务必让每一个文件名都以 my_ 开始","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令背景推测，生成的主要是前端页面及 Mock 逻辑文件\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件的后缀名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为推断的代码文件类型\n if ext in code_extensions:\n # 校验文件名是否以 my_ 开头\n if not file.startswith(\"my_\"):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在 30 行及以内","validation_code":"def check_function_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 判断大括号 '{' 所在位置的前缀是否是函数的开头\n def check_is_func(clean_text: str, pos: int) -> bool:\n p = pos - 1\n \n # 辅助跳过空白字符\n def skip_whitespace(idx: int) -> int:\n while idx >= 0 and clean_text[idx] in ' \\t\\n\\r':\n idx -= 1\n return idx\n\n p = skip_whitespace(p)\n if p < 0:\n return False\n \n # 判断是否为箭头函数 '=>'\n if clean_text[p] == '>' and p > 0 and clean_text[p-1] == '=':\n return True\n \n # 判断是否带有参数括号 ')'\n if clean_text[p] == ')':\n depth = 1\n p -= 1\n # 跳过成对的括号组合 (...)\n while p >= 0 and depth > 0:\n if clean_text[p] == ')':\n depth += 1\n elif clean_text[p] == '(':\n depth -= 1\n p -= 1\n \n if depth > 0:\n return False\n \n p = skip_whitespace(p)\n if p < 0:\n return False\n \n # 提取紧跟着括号前面的单词\n word = ''\n while p >= 0 and re.match(r'[a-zA-Z0-9_$]', clean_text[p]):\n word = clean_text[p] + word\n p -= 1\n \n if word:\n # 排除常见的非函数作用域的控制流关键字\n if word in ['if', 'for', 'while', 'switch', 'catch', 'new', 'with']:\n return False\n return True\n else:\n return False\n \n return False\n\n # 解析包含大括号特性的常规前端及后端语言代码\n def check_braces_file(file_path: str) -> bool:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n masked = []\n in_str = False\n str_char = ''\n in_line_comment = False\n in_block_comment = False\n \n i = 0\n length = len(content)\n # 第一阶段：掩码化屏蔽字符串内容与注释，避免误匹配大括号\n while i < length:\n c = content[i]\n if in_line_comment:\n if c == '\\n':\n in_line_comment = False\n masked.append(c)\n else:\n masked.append(' ')\n elif in_block_comment:\n if c == '*' and i + 1 < length and content[i+1] == '/':\n in_block_comment = False\n masked.append(' ')\n masked.append(' ')\n i += 1\n else:\n if c == '\\n':\n masked.append('\\n')\n else:\n masked.append(' ')\n elif in_str:\n if c == '\\\\':\n masked.append(' ')\n if i + 1 < length:\n masked.append(' ' if content[i+1] != '\\n' else '\\n')\n i += 1\n elif c == str_char:\n in_str = False\n masked.append(' ')\n else:\n if c == '\\n':\n masked.append('\\n')\n else:\n masked.append(' ')\n else:\n if c == '/' and i + 1 < length and content[i+1] == '/':\n in_line_comment = True\n masked.append(' ')\n masked.append(' ')\n i += 1\n elif c == '/' and i + 1 < length and content[i+1] == '*':\n in_block_comment = True\n masked.append(' ')\n masked.append(' ')\n i += 1\n elif c in ('\"', \"'\", \"`\"):\n in_str = True\n str_char = c\n masked.append(' ')\n else:\n masked.append(c)\n i += 1\n \n clean_text = ''.join(masked)\n \n # 第二阶段：依靠清理后的安全文本记录大括号深度并计算块的行数\n stack = []\n line_num = 1\n \n for pos, c in enumerate(clean_text):\n if c == '\\n':\n line_num += 1\n elif c == '{':\n is_func = check_is_func(clean_text, pos)\n stack.append((line_num, is_func))\n elif c == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n if line_num - start_line + 1 > 30:\n return False\n return True\n\n # 针对 Python 文件的精确行数解析\n def check_python_file(file_path: str) -> bool:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 判断行数是否严格控制在 30 行以内\n length = node.end_lineno - node.lineno + 1\n if length > 30:\n return False\n except Exception:\n # 存在不能解析的语法错误，予以跳过\n pass\n return True\n\n # 遍历当前生成项目所有的代码文件\n for root, dirs, files in os.walk(workspace_path):\n # 排除掉常见的生成目录或者无关的包依赖文件夹\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build', '.venv', 'venv', '__pycache__']]\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 判断对应后缀的代码文件\n if ext == '.py':\n if not check_python_file(file_path):\n return False\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.java', '.cpp', '.c', '.cs', '.go']:\n if not check_braces_file(file_path):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":2,"instruction":"现在来完善“上传/新增题目”和“编辑”的核心交互。请实现：点击“新增题目”或某行的“编辑”按钮时，弹出一个居中表单弹窗。这个表单需要有一套具有挑战性的动态联动逻辑：1. 包含基础字段（题干、解析、难度）。2. 包含一个“题目类型”选择器。当选择“单选题”时，下方渲染出一组选项输入框（A, B, C, D等），并带有一个单选按钮用于标记哪个是正确答案，同时允许新增和删除选项；当选择“多选题”时，逻辑类似，但标记正确答案的地方变为多选框；当切换为“简答题”时，之前的选项区全部隐藏，替换为一个大的多行文本输入框用于填写参考答案。填写完弹窗后点击“确定”，要求关闭弹窗并在后面的表格顶端新增这条题目记录。\n另外请你后续所有操作始终遵循以下要求：生成代码时，请将各函数的行数严格控制在25行及以内；执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"要求输出代码时，统一使用 4 个空格作为缩进标准","validation_code":"def check_indentation_standard(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义可能包含缩进格式的常见代码文件扩展名\n code_extensions = {'.py', '.js', '.html', '.css', '.vue', '.ts', '.jsx', '.tsx', '.json', '.java', '.cpp', '.c', '.h'}\n \n total_indented_lines = 0\n valid_4_space_lines = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 忽略空行或全是空白字符的行\n if line.strip() == '':\n continue\n \n # 仅匹配行首的半角空格和制表符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n leading_whitespace = match.group(1)\n \n # 如果使用了制表符缩进，则直接不满足要求\n if '\\t' in leading_whitespace:\n return False\n \n num_spaces = len(leading_whitespace)\n total_indented_lines += 1\n \n # 判断空格数是否为 4 的倍数\n if num_spaces % 4 == 0:\n valid_4_space_lines += 1\n\n # 如果存在缩进行，判断 4 个空格缩进的比例是否占据绝大多数（容错比例设为20%，用来包容少量的代码参数对齐换行）\n if total_indented_lines > 0:\n if valid_4_space_lines / total_indented_lines < 0.8:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n if not os.path.exists(workspace_path):\n return False\n\n # 小驼峰命名格式正则：首字母小写，只允许包含字母和数字\n is_lower_camel_case = re.compile(r'^[a-z][a-zA-Z0-9]*$')\n \n # 需要排除的常用语法关键字，避免被正则误认为函数名\n keywords = {\n 'if', 'for', 'while', 'catch', 'switch', 'return', \n 'constructor', 'super', 'function', 'else', 'import', \n 'export', 'try', 'finally', 'with', 'await', 'yield', \n 'typeof', 'instanceof', 'new', 'delete', 'void', \n 'class', 'interface', 'implements', 'extends'\n }\n\n # 1. 常规 function / func 声明\n func_decl_pattern = re.compile(r'\\b(?:function|func)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n\n # 2. JS/TS 等函数表达式 / 箭头函数赋值\n func_expr_pattern = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s*)?(?:function\\s*\\(|\\(.*?\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)')\n\n # 3. 对象字面量中的方法赋值\n obj_method_pattern = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s*)?(?:function\\s*\\(|\\(.*?\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)')\n\n # 4. 类中的方法定义（兼容 TS/JS, Java, C#, C++ 等大括号语言）\n method_pattern = re.compile(\n r'^\\s*(?:public\\s+|private\\s+|protected\\s+|static\\s+|async\\s+|final\\s+|abstract\\s+|override\\s+|virtual\\s+)*'\n r'(?:[a-zA-Z_$][a-zA-Z0-9_$<>\\[\\]]*\\s+)?'\n r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*(?::\\s*[^{]+)?\\s*\\{', \n re.MULTILINE\n )\n\n # 5. Go 带有接收者的方法定义\n go_method_pattern = re.compile(r'\\bfunc\\s+\\([^)]+\\)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 常见代码文件\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.go']:\n found_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 汇总当前文件提取到的所有函数/方法名\n all_funcs = (\n func_decl_pattern.findall(content) +\n func_expr_pattern.findall(content) +\n obj_method_pattern.findall(content) +\n method_pattern.findall(content) +\n go_method_pattern.findall(content)\n )\n \n for fn in all_funcs:\n if fn in keywords:\n continue\n if not is_lower_camel_case.match(fn):\n return False\n \n # Python 文件通过 AST 安全提取\n elif ext in ['.py']:\n found_code_file = True\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 包括类方法和 __init__ 在内，如果未满足小驼峰则直接判定失败\n if not is_lower_camel_case.match(node.name):\n return False\n except Exception:\n continue\n\n # 如果正确解析到了代码文件且中途没有被判定为False，说明全面符合要求\n return found_code_file","tags":["Style","Naming"]},{"constraint":"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数","validation_code":"import os\nimport re\n\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\n # 结合用户指令推断可能的前后端代码文件扩展名\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\n \n # 匹配类定义中类名的正则表达式\n # 能够匹配 \"class MyClass \" 等常见类的定义形式，且过滤掉HTML中 \"class=\" 的干扰\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z_][A-Za-z0-9_]*)\\b')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找当前文件内所有符合类名定义的字符串\n class_names = class_pattern.findall(content)\n \n for name in class_names:\n # 针对要求中明确指出的“字符长度”要求，使用 len 函数进行统计判断\n if len(name) % 2 != 0:\n return False\n except Exception:\n # 读取解析文件出现异常则忽略当前文件\n continue\n \n return True","tags":["Style","Naming"]},{"constraint":"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n code_exts = {\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', \n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.php', '.sh'\n }\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令中“务必让每一个文件名都以 my_ 开始”过滤文件\n if not file.startswith('my_'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext not in code_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n comment_lines = set()\n in_string = False\n string_char = None\n in_multiline_comment = False\n multiline_type = None \n \n # 根据不同的文件扩展名设定该文件可能包含的注释类型\n allow_c_multi = ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', '.java', '.c', '.cpp', '.cs', '.go', '.php'}\n allow_html_multi = ext in {'.vue', '.html', '.xml', '.svg', '.php'}\n allow_c_single = ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.go', '.php'}\n allow_hash_single = ext in {'.py', '.sh', '.bash', '.rb', '.yml', '.yaml', '.dockerfile'}\n \n for i, line in enumerate(lines):\n j = 0\n line_has_comment = False\n \n while j < len(line):\n # 当既不在多行注释内，也不在字符串内时\n if not in_multiline_comment and not in_string:\n # 检查是否进入了字符串区域\n if line[j] in ['\"', \"'\", '`']:\n in_string = True\n string_char = line[j]\n j += 1\n continue\n \n # 检查多行注释起点 /*\n if allow_c_multi and line[j:j+2] == '/*':\n in_multiline_comment = True\n multiline_type = 'C'\n line_has_comment = True\n j += 2\n continue\n \n # 检查多行注释起点 <!--\n if allow_html_multi and line[j:j+4] == '<!--':\n in_multiline_comment = True\n multiline_type = 'HTML'\n line_has_comment = True\n j += 4\n continue\n \n # 检查单行注释起点 //\n if allow_c_single and line[j:j+2] == '//':\n line_has_comment = True\n break\n \n # 检查单行注释起点 #\n if allow_hash_single and line[j] == '#':\n line_has_comment = True\n break\n \n elif in_string:\n # 跳过转义字符，避免遇到 \\\" 或 \\' 误退出字符串状态\n if line[j] == '\\\\':\n j += 2\n continue\n # 检查是否结束了字符串区域\n elif line[j] == string_char:\n in_string = False\n j += 1\n continue\n \n elif in_multiline_comment:\n line_has_comment = True\n # 检查 C 风格多行注释是否结束\n if multiline_type == 'C' and line[j:j+2] == '*/':\n in_multiline_comment = False\n j += 2\n continue\n # 检查 HTML 风格多行注释是否结束\n if multiline_type == 'HTML' and line[j:j+3] == '-->':\n in_multiline_comment = False\n j += 3\n continue\n \n j += 1\n \n # 如果当前行结束时仍处于多行注释中，本行也应视为有注释\n if in_multiline_comment:\n line_has_comment = True\n \n if line_has_comment:\n comment_lines.add(i + 1)\n \n # 针对常规单引号和双引号的不规范跨行情况，行末直接重置字符串状态避免误判\n if in_string and string_char in ['\"', \"'\"]:\n in_string = False\n \n # 移除被允许的第 5 行注释（如果存在）\n comment_lines.discard(5)\n \n # 若仍存在非第 5 行的注释，则说明未满足条件\n if len(comment_lines) > 0:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成代码文件时，务必让每一个文件名都以 my_ 开始","validation_code":"def check_code_filename_prefix(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推断，主要为前端相关代码文件，也可以加入部分其他语言的代码文件后缀\n code_extensions = {\n '.html', '.htm', '.css', '.scss', '.less',\n '.js', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.php', '.go'\n }\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀名并转换为小写，以进行统一匹配\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n \n # 检查文件名是否以 my_ 开始\n if not file.startswith('my_'):\n return False\n\n # 要求必须生成了符合规范的代码文件\n return has_code_file","tags":["Style","Naming"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在25行及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 辅助状态机函数，将JS/TS/Vue等文件中的字符串和注释清空为空格，以防止干扰函数正文的识别，同时保持原有行数不变。\n def clean_code(code: str) -> str:\n result = []\n state = 0\n i = 0\n n = len(code)\n while i < n:\n c = code[i]\n if state == 0:\n if c == '/' and i+1 < n and code[i+1] == '/':\n state = 1\n result.append(' ')\n result.append(' ')\n i += 2\n continue\n elif c == '/' and i+1 < n and code[i+1] == '*':\n state = 2\n result.append(' ')\n result.append(' ')\n i += 2\n continue\n elif c == \"'\":\n state = 3\n result.append(' ')\n i += 1\n continue\n elif c == '\"':\n state = 4\n result.append(' ')\n i += 1\n continue\n elif c == '`':\n state = 5\n result.append(' ')\n i += 1\n continue\n else:\n result.append(c)\n i += 1\n elif state == 1:\n if c == '\\n':\n state = 0\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 2:\n if c == '*' and i+1 < n and code[i+1] == '/':\n state = 0\n result.append(' ')\n result.append(' ')\n i += 2\n else:\n if c == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 3:\n if c == '\\\\':\n result.append(' ')\n if i+1 < n:\n result.append('\\n' if code[i+1] == '\\n' else ' ')\n i += 2\n elif c == \"'\":\n state = 0\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 4:\n if c == '\\\\':\n result.append(' ')\n if i+1 < n:\n result.append('\\n' if code[i+1] == '\\n' else ' ')\n i += 2\n elif c == '\"':\n state = 0\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n elif state == 5:\n if c == '\\\\':\n result.append(' ')\n if i+1 < n:\n result.append('\\n' if code[i+1] == '\\n' else ' ')\n i += 2\n elif c == '`':\n state = 0\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append('\\n')\n else:\n result.append(' ')\n i += 1\n return ''.join(result)\n\n if not os.path.exists(workspace_path):\n return False\n \n # 正则表达式寻找前端函数定义的开头特征\n pattern = re.compile(\n r'(?<!\\w)(?!(?:if|for|while|switch|catch|with|else|return|new|import|export)\\b)'\n r'(?:function(?:\\s+\\w+)?\\s*\\([^)]*\\)|(?:async\\s+)?(?:\\([^)]*\\)|\\w+)\\s*=>|(?:async\\s+)?(?:\\w+)\\s*\\([^)]*\\)|(?:\\w+)\\s*:\\s*function\\s*\\([^)]*\\))\\s*\\{'\n )\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据要求过滤日志和无关文件\n if file.endswith('.log'):\n continue\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n if ext in ['.png', '.jpg', '.jpeg', '.gif', '.zip', '.tar', '.gz', '.pdf', '.docx', '.mp4']:\n continue\n \n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n # 使用 AST 解析 Python 代码\n if ext == '.py':\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n line_count = node.end_lineno - node.lineno + 1\n if line_count > 25:\n return False\n except Exception:\n pass\n else:\n # 对待各类前端页面/逻辑代码进行匹配运算\n cleaned_code = clean_code(code)\n for m in pattern.finditer(cleaned_code):\n start_pos = m.start()\n bracket_pos = cleaned_code.find('{', start_pos)\n if bracket_pos == -1:\n continue\n \n # 使用栈匹配大括号，准确找到函数结束位置\n count = 0\n end_pos = bracket_pos\n for i in range(bracket_pos, len(cleaned_code)):\n if cleaned_code[i] == '{':\n count += 1\n elif cleaned_code[i] == '}':\n count -= 1\n if count == 0:\n end_pos = i\n break\n if count != 0:\n continue\n \n # 包含换行符的数量即可转换为对应的代码行数跨度\n line_count = cleaned_code.count('\\n', start_pos, end_pos) + 1\n if line_count > 25:\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"不错，题库这块可以了。下一步我们做“用户管理”页面。页面主体同样是一个数据表格，展示几个测试用户的基本信息：用户ID、用户昵称、注册时间、总刷题数、整体正确率。在表格的“操作”列中，放置一个“查看答题记录”按钮。点击该按钮后，请从屏幕右侧滑出一个抽屉（Drawer）组件。在这个抽屉里面，要求使用垂直时间轴（Timeline）组件的形式，展示该用户最近的答题流水记录，每条时间轴节点需要清晰地显示：答题时间、对应的题目名称概要，以及一个醒目的Tag标签（绿色表示“回答正确”，红色表示“回答错误”）。\n另外请你后续所有操作始终遵循以下新要求：生成代码时，请将各函数的行数严格控制在20行及以内；只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"要求输出代码时，统一使用 4 个空格作为缩进标准","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 基于多轮指令预测可能涉及的代码文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.json', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.cs', '.go', '.rb', '.php'\n }\n \n space_4n_lines = 0\n space_not_4n_lines = 0\n tab_indent_lines = 0\n \n def process_lines(lines):\n nonlocal space_4n_lines, space_not_4n_lines, tab_indent_lines\n for line in lines:\n line_str = line.rstrip('\\n\\r')\n if not line_str.strip():\n continue\n \n # 匹配每一行的前导空白字符及后续内容\n m = re.match(r'^([ \\t]+)(\\S.*)', line_str)\n if m:\n spaces = m.group(1)\n content = m.group(2)\n \n # 排除可能的不规则注释对齐（避免误杀因多行注释导致的前导不规则空格）\n if content.startswith('*') or content.startswith('/*') or content.startswith('<!--'):\n continue\n \n if '\\t' in spaces:\n tab_indent_lines += 1\n else:\n if len(spaces) % 4 == 0:\n space_4n_lines += 1\n else:\n space_not_4n_lines += 1\n\n # 1. 检查 workspace_path 下生成的项目代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 忽略隐藏目录与编译依赖目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'dist', 'build', 'out')]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n file_lines = f.readlines()\n \n # 判断是否为直接保存进去的 Diff 文件格式内容（防 LLM 出错存储成Diff文件）\n content_start = \"\".join(file_lines[:15])\n if '@@ -' in content_start and ' +' in content_start:\n continue\n \n process_lines(file_lines)\n except Exception:\n continue\n\n # 2. 检查最后一轮回复内容 (response) 中可能存在的代码块\n pattern = re.compile(r'```([a-zA-Z0-9+\\-]*)\\n(.*?)\\n```', re.DOTALL)\n for match in pattern.finditer(response):\n lang = match.group(1).lower()\n # 跳过非代码块的 diff, patch 或 log（因 Unified Diff 中空格被赋予了特殊增加与删除标记的含义）\n if lang in ('diff', 'patch', 'log'):\n continue\n \n code_text = match.group(2)\n # 即便未显式声明语言类型，如果特征匹配 Unified Diff，也应跳过验证\n if '@@ -' in code_text and ' +' in code_text:\n continue\n \n code_lines = code_text.split('\\n')\n process_lines(code_lines)\n\n total_indents = space_4n_lines + space_not_4n_lines + tab_indent_lines\n \n # 若全文中未检查到任何行首进行缩进的行，理论上并未违反此约束\n if total_indents == 0:\n return True\n \n # 计算严格符合 4 的倍数的空格缩进行在全部缩进行之中的占比\n ratio = space_4n_lines / total_indents\n \n # 遵循 4 个空格缩进规范的项目，绝大部分缩进行必然是 4 的倍数，但可能存在极少量的方法传参换行对齐；\n # 相比之下，若采用 2 空格做标准，此比例仅徘徊在 40% ~ 50% 区间，若用 Tab 缩进则占比趋近 0；\n # 因此设定 0.75（75%）作为极高鲁棒性的分水岭阈值进行评判。\n return ratio >= 0.75","tags":["Style","Layout"]},{"constraint":"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 若工作空间路径不存在直接返回 False\n if not os.path.exists(workspace_path):\n return False\n \n # 定义待扫描和解析的代码文件后缀（主要覆盖前端及部分常见后端）\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\n \n # 检测名称是否满足小驼峰命名\n def check_camel_case(name):\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\n \n # 去除代码中的注释以防止正则误匹配\n def remove_comments(code, ext):\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'<!--[\\s\\S]*?-->', '', code)\n elif ext == '.py':\n code = re.sub(r'#.*', '', code)\n code = re.sub(r\"'''[\\s\\S]*?'''\", '', code)\n code = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '', code)\n return code\n\n # 遍历项目文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n content = remove_comments(content, ext)\n names = []\n \n # 解析不同语言中的函数名称\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\n # 1. 明确的 function 关键字声明（含 generator 函数）\n names.extend(re.findall(r'\\bfunction\\s*\\*?\\s+([a-zA-Z0-9_]+)\\s*\\(', content))\n # 2. 箭头函数赋值表达式\n names.extend(re.findall(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z0-9_]+)\\s*=>', content))\n # 3. 匿名函数赋值表达式\n names.extend(re.findall(r'\\b(?:const|let|var)\\s+([a-zA-Z0-9_]+)\\s*=\\s*(?:async\\s+)?function\\b', content))\n \n # 4. 对象方法、类方法定义等\n pattern = r'^[ \\t]*(?:public\\s+|private\\s+|protected\\s+|static\\s+|async\\s+|get\\s+|set\\s+)*([a-zA-Z0-9_]+)\\s*(?:<[^>]+>)?\\s*\\(([^)]*)\\)(?::\\s*[^{=;]+)?\\s*\\{'\n for match in re.finditer(pattern, content, re.MULTILINE):\n name = match.group(1)\n args = match.group(2)\n # 排除控制流等关键字\n if name in {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'with'}:\n continue\n # 避免误伤带有回调的外部函数调用形式\n if 'function' in args:\n continue\n names.append(name)\n \n elif ext == '.py':\n names.extend(re.findall(r'\\bdef\\s+([a-zA-Z0-9_]+)\\s*\\(', content))\n \n # 针对所有提取出的函数名称进行合规校验\n for name in names:\n if not check_camel_case(name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数","validation_code":"import os\nimport re\n\ndef verify_class_name_length(response: str, workspace_path: str) -> bool:\n # 若项目目录尚不存在，说明未生成违规文件\n if not os.path.isdir(workspace_path):\n return True\n\n # 常见可能包含类定义的代码文件扩展名\n code_extensions = {\n '.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.cpp', '.cc', '.cxx', \n '.h', '.hpp', '.cs', '.php', '.rb', '.go', '.swift', '.kt', '.vue'\n }\n \n # 匹配 'class' 关键字，然后是一个或多个空白符，接着是类名标识符（不能以数字开头）\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_]\\w*)')\n \n # 在某些语言（如JS/TS等）中可能使用匿名类继承，导致直接捕获到保留字，在此做过滤\n ignore_keywords = {'extends', 'implements'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 查找当前代码文件中所有的类名\n classes = class_pattern.findall(content)\n for cls_name in classes:\n if cls_name in ignore_keywords:\n continue\n \n # 检查该类名的字符长度是否为偶数\n if len(cls_name) % 2 != 0:\n return False\n except Exception:\n # 如果存在不可读文件或编码异常，直接忽略该文件以防阻断验证\n continue\n \n return True","tags":["Style","Naming"]},{"constraint":"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n def is_code_file(filename):\n # 定义需要被检查的常见代码文件后缀名\n exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', \n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', '.hpp',\n '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.cs', '.m'}\n _, ext = os.path.splitext(filename)\n return ext.lower() in exts\n\n def get_comment_lines(text, file_ext):\n comment_lines = set()\n state = 'NORMAL'\n lines = text.split('\\n')\n \n for line_idx, line in enumerate(lines):\n i = 0\n n = len(line)\n # 单行注释如果在换行时结束，则重置状态为 NORMAL\n if state in ['IN_SINGLE_LINE_COMMENT_C', 'IN_SINGLE_LINE_COMMENT_SHELL']:\n state = 'NORMAL'\n \n while i < n:\n if state == 'NORMAL':\n if file_ext in {'.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.h', '.hpp', \n '.cs', '.go', '.rs', '.php', '.swift', '.kt', '.vue', '.html', \n '.htm', '.css', '.scss', '.less'}:\n # 对于 js、ts 等支持 // 注释。注意标准的 CSS 不支持 // 作为注释，所以予以排除。\n if line[i:i+2] == '//' and file_ext not in {'.css'}:\n state = 'IN_SINGLE_LINE_COMMENT_C'\n comment_lines.add(line_idx + 1)\n i += 2\n continue\n elif line[i:i+2] == '/*':\n state = 'IN_MULTILINE_COMMENT_C'\n comment_lines.add(line_idx + 1)\n i += 2\n continue\n elif line[i:i+4] == '<!--' and file_ext in {'.html', '.htm', '.vue'}:\n state = 'IN_MULTILINE_COMMENT_HTML'\n comment_lines.add(line_idx + 1)\n i += 4\n continue\n \n if file_ext in {'.py', '.rb', '.php', '.sh'}:\n if line[i] == '#':\n state = 'IN_SINGLE_LINE_COMMENT_SHELL'\n comment_lines.add(line_idx + 1)\n i += 1\n continue\n \n if line[i] == '\"':\n state = 'IN_STRING_DOUBLE'\n elif line[i] == \"'\":\n state = 'IN_STRING_SINGLE'\n elif line[i] == '`' and file_ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.go'}:\n state = 'IN_STRING_BACKTICK'\n \n i += 1\n \n elif state == 'IN_STRING_DOUBLE':\n if line[i] == '\\\\':\n i += 2\n elif line[i] == '\"':\n state = 'NORMAL'\n i += 1\n else:\n i += 1\n elif state == 'IN_STRING_SINGLE':\n if line[i] == '\\\\':\n i += 2\n elif line[i] == \"'\":\n state = 'NORMAL'\n i += 1\n else:\n i += 1\n elif state == 'IN_STRING_BACKTICK':\n if line[i] == '\\\\':\n i += 2\n elif line[i] == '`':\n state = 'NORMAL'\n i += 1\n else:\n i += 1\n elif state == 'IN_MULTILINE_COMMENT_C':\n comment_lines.add(line_idx + 1)\n if line[i:i+2] == '*/':\n state = 'NORMAL'\n i += 2\n else:\n i += 1\n elif state == 'IN_MULTILINE_COMMENT_HTML':\n comment_lines.add(line_idx + 1)\n if line[i:i+3] == '-->':\n state = 'NORMAL'\n i += 3\n else:\n i += 1\n elif state in ['IN_SINGLE_LINE_COMMENT_C', 'IN_SINGLE_LINE_COMMENT_SHELL']:\n comment_lines.add(line_idx + 1)\n i += 1\n \n # 处理完该行后，如果仍处于多行注释状态中，证明整行都属于多行注释的覆盖范围\n if state in ['IN_MULTILINE_COMMENT_C', 'IN_MULTILINE_COMMENT_HTML']:\n comment_lines.add(line_idx + 1)\n \n return comment_lines\n\n # 遍历工作区目录，核对所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if is_code_file(file):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n continue\n \n _, ext = os.path.splitext(file)\n ext = ext.lower()\n \n # 获取该文件所有存在注释的行号集合\n comments = get_comment_lines(text, ext)\n \n # 规则约束：除了第 5 行之外，其余代码行均不得添加任何注释\n illegal_comments = comments - {5}\n if illegal_comments:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"生成代码文件时，务必让每一个文件名都以 my_ 开始","validation_code":"def check_code_filenames(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的扩展名集合，根据指令推测大概率为前端及常见编程语言代码\n code_extensions = {\n \".html\", \".css\", \".js\", \".ts\", \".jsx\", \".tsx\", \".vue\", \n \".py\", \".java\", \".c\", \".cpp\", \".go\", \".rs\", \".php\", \".swift\"\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 忽略隐藏文件或系统生成文件\n if file.startswith('.'):\n continue\n \n # 获取文件的后缀名\n _, ext = os.path.splitext(file)\n ext = ext.lower()\n \n # 如果是代码文件，验证其是否以 my_ 开头\n if ext in code_extensions:\n if not file.startswith(\"my_\"):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在20行及以内","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n # 解析 Python 文件中的函数行数\n def get_func_length_in_py(content: str):\n funcs = []\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\n funcs.append((node.lineno, node.end_lineno))\n except Exception:\n pass\n return funcs\n\n # 解析 类C/前端 文件中的函数行数（通过自建的微型词法解析大括号匹配）\n def get_func_length_brace_lang(content: str):\n funcs = []\n in_single_quote = False\n in_double_quote = False\n in_backtick = False\n in_single_comment = False\n in_multi_comment = False\n escape = False\n \n depth = 0\n line_num = 1\n func_starts = {}\n \n i = 0\n n = len(content)\n while i < n:\n char = content[i]\n \n # 换行处理\n if char == '\\n':\n line_num += 1\n in_single_comment = False\n escape = False\n i += 1\n continue\n \n if escape:\n escape = False\n i += 1\n continue\n \n if in_single_comment:\n i += 1\n continue\n \n if in_multi_comment:\n if char == '*' and i + 1 < n and content[i+1] == '/':\n in_multi_comment = False\n i += 2\n continue\n i += 1\n continue\n \n if in_single_quote:\n if char == '\\\\':\n escape = True\n elif char == \"'\":\n in_single_quote = False\n i += 1\n continue\n \n if in_double_quote:\n if char == '\\\\':\n escape = True\n elif char == '\"':\n in_double_quote = False\n i += 1\n continue\n \n if in_backtick:\n if char == '\\\\':\n escape = True\n elif char == '`':\n in_backtick = False\n i += 1\n continue\n \n # 判断是否进入注释或字符串\n if char == '/' and i + 1 < n:\n if content[i+1] == '/':\n in_single_comment = True\n i += 2\n continue\n elif content[i+1] == '*':\n in_multi_comment = True\n i += 2\n continue\n elif char == \"'\":\n in_single_quote = True\n elif char == '\"':\n in_double_quote = True\n elif char == '`':\n in_backtick = True\n elif char == '{':\n depth += 1\n # 获取大括号前面的内容片段，去除上一条语句的影响\n prefix = content[max(0, i-200):i]\n last_idx = max(prefix.rfind(';'), prefix.rfind('{'), prefix.rfind('}'))\n if last_idx != -1:\n prefix = prefix[last_idx+1:]\n \n is_func = False\n # 判断是否属于 function funcName(...) 或者 function(...)\n if re.search(r'\\bfunction\\b\\s*[A-Za-z0-9_]*\\s*\\([^)]*\\)\\s*$', prefix):\n is_func = True\n # 判断是否属于箭头函数 => \n elif re.search(r'=>\\s*$', prefix):\n is_func = True\n else:\n # 判断是否是常见的类方法声明 methodName(...)\n match = re.search(r'\\b([A-Za-z0-9_]+)\\s*\\([^)]*\\)\\s*$', prefix)\n # 排除掉常见的控制流关键字\n if match and match.group(1) not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\n is_func = True\n \n if is_func:\n func_starts[depth] = line_num\n \n elif char == '}':\n if depth in func_starts:\n start_line = func_starts.pop(depth)\n funcs.append((start_line, line_num))\n depth -= 1\n if depth < 0:\n depth = 0\n func_starts.clear()\n \n i += 1\n \n return funcs\n\n # 支持扫描的目标文件扩展名（包括前端常见扩展名与常规后台语言）\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.php', '.rb', '.rs'\n }\n \n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'dist', 'build', 'venv', '.env'}\n \n # 遍历项目文件\n for root, dirs, files in os.walk(workspace_path):\n # 屏蔽不相干的构建和依赖文件夹\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 根据语言执行相应的行数提取\n if ext == '.py':\n funcs = get_func_length_in_py(content)\n else:\n funcs = get_func_length_brace_lang(content)\n \n # 验证提取的每一个函数\n for start_line, end_line in funcs:\n length = end_line - start_line + 1\n if length > 20:\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":4,"instruction":"最后一步，我们把“权限管理”概念融入进去，实现全局的角色控制功能。请在顶部Header最右侧的“当前用户”占位处，改为一个下拉菜单，允许在两个角色之间来回切换：“超级管理员”和“普通运营者”。这需要实现跨组件的权限联动：1. 当切换为“普通运营者”时，左侧导航栏的“权限管理”菜单项需要直接隐藏；同时，在“题库管理”页面的表格中，所有行的“删除”按钮要变为置灰不可点击的禁用状态。2. 当切换回“超级管理员”时，左侧隐藏的菜单恢复显示，“删除”按钮也恢复可点击。请确保这种角色切换在UI上有即时的视觉反馈。\n另外请你后续所有操作始终遵循以下要求：生成代码时，请将各函数的行数严格控制在15行及以内；提高效率的必要前提是，必须尽可能并行触发多个独立的工具调用。","instruction_following_checklist":[{"constraint":"你生成给用户的回复内容，必须一直使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件中绝对不能使用多层if","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"要求输出代码时，统一使用 4 个空格作为缩进标准","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 推断可能的代码文件类型集合（包含常见的前端和后端扩展名）\n code_extensions = {\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \n '.css', '.scss', '.less', '.py', '.java', '.c', \n '.cpp', '.go', '.rs', '.php'\n }\n \n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 忽略无法以纯文本/UTF-8读取的异常文件\n continue\n\n for i, line in enumerate(lines):\n # 根据其余要求，第5行允许存在注释，以防特殊格式导致不规则缩进，故跳过第5行\n if i == 4:\n continue\n \n # 忽略完全是空行、无内容的行\n if not line.strip():\n continue\n \n # 匹配当前行行首的空白字符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n indent = match.group(1)\n \n # 不能包含任何制表符\n if '\\t' in indent:\n return False\n \n # 缩进的空格数必须恰好是4的整数倍\n if len(indent) % 4 != 0:\n return False\n\n return has_code_file","tags":["Style","Layout"]},{"constraint":"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式","validation_code":"def check_function_names_lower_camel(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 验证是否为小驼峰命名\n def is_lower_camel_case(s: str) -> bool:\n if not s:\n return False\n # 小驼峰: 必须以小写字母开头，后续只能包含字母或数字（不包含下划线，不包含 $）\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', s))\n \n # 去除代码中的各类注释以避免干扰\n def remove_comments(code: str) -> str:\n # 移除HTML注释 <!-- ... -->\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\n # 移除多行注释 /* ... */\n code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)\n # 移除JS单行注释 // ...\n code = re.sub(r'//.*', '', code)\n # 移除Python单行注释 # ...\n code = re.sub(r'#.*', '', code)\n return code\n\n # 常规控制流关键字等白名单（防止被当做函数名处理）\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'with', 'return', \n 'function', 'yield', 'await', 'import', 'export', 'typeof',\n 'instanceof', 'elif', 'except', 'finally', 'try'\n }\n\n # 用于捕获主流语言常见函数声明的正则表达式集合\n patterns = [\n # JS/TS 传统函数: function funcName (\n r'function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(',\n # JS/TS 箭头函数: const/let/var funcName = (...) => 或者 = async (...) =>\n r'(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>',\n # JS/TS 对象方法: funcName: function (\n r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*function\\s*\\(',\n # Python 函数: def funcName (\n r'def\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('\n ]\n \n # 单独行匹配，用来捕获类方法、Vue/React生命周期等（例如： methodName(args) { ）\n line_pattern = r'^\\s*(?:async\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{'\n \n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅处理前端/后端主流代码文件，略过静态资源等\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.js', '.ts', '.vue', '.html', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.cs', '.go']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n content_no_comments = remove_comments(content)\n \n # 1. 对整个文件内容做全局模式的捕获\n for pat in patterns:\n matches = re.finditer(pat, content_no_comments)\n for m in matches:\n func_name = m.group(1)\n if func_name in keywords:\n continue\n if not is_lower_camel_case(func_name):\n return False\n \n # 2. 逐行匹配类/对象方法的定义\n for line in content_no_comments.split('\\n'):\n m = re.match(line_pattern, line)\n if m:\n func_name = m.group(1)\n if func_name in keywords:\n continue\n if not is_lower_camel_case(func_name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区路径不存在，判定为不满足（或根据实际测试台环境视作异常）\n if not os.path.exists(workspace_path):\n return False\n\n # 定义涉及的前后端主流代码文件扩展名集合\n valid_exts = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \n '.java', '.cs', '.cpp', '.cc', '.h', '.hpp', '.php', \n '.rb', '.go', '.swift', '.kt', '.scala', '.dart'\n }\n \n # 正则表达式匹配类定义：\n # \\bclass\\s+ 匹配 \"class\" 关键字后紧跟至少一个空白字符\n # ([A-Za-z_][A-Za-z0-9_]*) 捕获组提取类名的标识符\n class_pattern = re.compile(r'\\bclass\\s+([A-Za-z_][A-Za-z0-9_]*)')\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n\n # 简单去除代码注释及多行字符串的干扰，防止误匹配到注释里的文本\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'\"\"\".*?\"\"\"', '', content, flags=re.DOTALL)\n content = re.sub(r\"'''.*?'''\", '', content, flags=re.DOTALL)\n\n # 提取出该代码文件中的所有类名\n matches = class_pattern.findall(content)\n for class_name in matches:\n # 指令中明确提到“字符长度”，直接使用原生的 len() 函数\n if len(class_name) % 2 != 0:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见代码文件扩展名及对应的注释风格分类\n code_extensions = {\n '.py': 'python',\n '.js': 'c_style',\n '.jsx': 'c_style',\n '.ts': 'c_style',\n '.tsx': 'c_style',\n '.java': 'c_style',\n '.c': 'c_style',\n '.cpp': 'c_style',\n '.go': 'c_style',\n '.css': 'c_style',\n '.html': 'html_style',\n '.vue': 'html_style'\n }\n\n # 预编译各类语言下匹配“字符串常量”与“注释”的正则表达式组合\n # 巧妙利用正则表达式先匹配字符串组、后匹配注释组，确保字符串内部的注释符号不会被当作注释捕获\n patterns = {\n 'python': re.compile(\n r'(?P<string>\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\')|'\n r'(?P<comment>#.*?$)',\n re.MULTILINE\n ),\n 'c_style': re.compile(\n r'(?P<string>\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\'|`[\\s\\S]*?`)|'\n r'(?P<comment>//.*?$|/\\*[\\s\\S]*?\\*/)',\n re.MULTILINE\n ),\n 'html_style': re.compile(\n r'(?P<string>\"(?:[^\"\\\\]|\\\\.)*\"|\\'(?:[^\\'\\\\]|\\\\.)*\\'|`[\\s\\S]*?`)|'\n r'(?P<comment><!--[\\s\\S]*?-->|//.*?$|/\\*[\\s\\S]*?\\*/)',\n re.MULTILINE\n )\n }\n\n # 遍历代码工作区下所有的文件\n for root, dirs, files in os.walk(workspace_path):\n # 忽略可能的隐藏目录（如 .git）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 严格依据用户第8条系统指令，只读取以 my_ 命名的生成代码文件\n if not file.startswith('my_'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n file_type = code_extensions[ext]\n pattern = patterns[file_type]\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 使用正则迭代查找文件中的所有匹配项\n for match in pattern.finditer(content):\n # 只要存在属于注释分组的内容\n if match.group('comment'):\n start_pos = match.start('comment')\n end_pos = match.end('comment')\n \n # 基于当前匹配位置前面的换行符数量来计算其代码所在行号（从 1 开始计行）\n start_line = content.count('\\n', 0, start_pos) + 1\n end_line = content.count('\\n', 0, end_pos) + 1\n \n # 验证注释是否越权出现在了不该出现的位置\n for line_num in range(start_line, end_line + 1):\n # 一旦注释出现在除第 5 行外的其它行，立刻返回失败\n if line_num != 5:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"生成代码文件时，务必让每一个文件名都以 my_ 开始","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名集合\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', \n '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.swift', \n '.kt', '.cs', '.sh', '.dart'\n }\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名\n _, ext = os.path.splitext(file)\n \n # 判断是否为代码文件\n if ext.lower() in code_extensions:\n # 检查文件名是否以 my_ 开头\n if not file.startswith('my_'):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成代码时，请将各函数的行数严格控制在15行及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def lex(code: str) -> list:\n tokens = []\n i = 0\n line_num = 1\n n = len(code)\n \n stack = []\n \n while i < n:\n c = code[i]\n \n # 处理模板字符串的内部状态\n if stack and stack[-1] == 'template':\n if code[i] == '\\\\':\n i += 2\n elif code[i] == '`':\n stack.pop()\n i += 1\n elif code[i:i+2] == '${':\n stack.append('interpolation')\n tokens.append(('{', line_num)) # 使用 { 保证插值大括号平衡\n i += 2\n else:\n if code[i] == '\\n': line_num += 1\n i += 1\n continue\n\n if c == '\\n':\n line_num += 1\n i += 1\n elif c in ' \\t\\r':\n i += 1\n elif c == '/' and i + 1 < n and code[i+1] == '/':\n i += 2\n while i < n and code[i] != '\\n':\n i += 1\n elif c == '/' and i + 1 < n and code[i+1] == '*':\n i += 2\n while i < n:\n if code[i] == '\\n':\n line_num += 1\n i += 1\n elif code[i] == '*' and i + 1 < n and code[i+1] == '/':\n i += 2\n break\n else:\n i += 1\n elif c == \"'\":\n i += 1\n while i < n:\n if code[i] == '\\\\':\n i += 2\n elif code[i] == \"'\":\n i += 1\n break\n else:\n if code[i] == '\\n': line_num += 1\n i += 1\n elif c == '\"':\n i += 1\n while i < n:\n if code[i] == '\\\\':\n i += 2\n elif code[i] == '\"':\n i += 1\n break\n else:\n if code[i] == '\\n': line_num += 1\n i += 1\n elif c == '`':\n stack.append('template')\n i += 1\n elif code[i:i+2] == '=>':\n tokens.append(('=>', line_num))\n i += 2\n elif c == '{':\n tokens.append(('{', line_num))\n i += 1\n elif c == '}':\n if stack and stack[-1] == 'interpolation':\n stack.pop()\n tokens.append(('}', line_num))\n else:\n tokens.append(('}', line_num))\n i += 1\n else:\n m = re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*', code[i:])\n if m:\n tokens.append((m.group(0), line_num))\n i += m.end()\n elif c in '()':\n tokens.append((c, line_num))\n i += 1\n else:\n i += 1\n return tokens\n\n def check_functions(tokens: list) -> bool:\n idx = 0\n blacklist = {\n 'if', 'for', 'while', 'switch', 'catch', 'return', 'new', \n 'import', 'export', 'super', 'else', 'with', 'typeof', \n 'instanceof', 'await', 'yield'\n }\n \n while idx < len(tokens):\n t, ln = tokens[idx]\n start_idx = -1\n brace_idx = -1\n \n # 模式1: 传统 function() {}\n if t == 'function':\n start_idx = idx\n paren_count = 0\n for j in range(idx + 1, len(tokens)):\n if tokens[j][0] == '(':\n paren_count += 1\n elif tokens[j][0] == ')':\n paren_count -= 1\n elif tokens[j][0] == '{' and paren_count == 0:\n brace_idx = j\n break\n \n # 模式2: 箭头函数 () => {}\n elif t == '=>':\n start_idx = idx\n if idx + 1 < len(tokens) and tokens[idx+1][0] == '{':\n brace_idx = idx + 1\n \n # 模式3: 类方法或对象方法 identifier() {}\n elif re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t) and t not in blacklist:\n if idx + 1 < len(tokens) and tokens[idx+1][0] == '(':\n paren_count = 0\n found_close_paren = False\n for j in range(idx + 1, len(tokens)):\n if tokens[j][0] == '(':\n paren_count += 1\n elif tokens[j][0] == ')':\n paren_count -= 1\n if paren_count == 0:\n found_close_paren = True\n elif tokens[j][0] == '{' and paren_count == 0 and found_close_paren:\n start_idx = idx\n brace_idx = j\n break\n \n # 若找到了函数的开端和它的左侧大括号\n if brace_idx != -1:\n brace_count = 0\n end_brace_idx = -1\n for j in range(brace_idx, len(tokens)):\n if tokens[j][0] == '{':\n brace_count += 1\n elif tokens[j][0] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_brace_idx = j\n break\n \n if end_brace_idx != -1:\n # 避免误抓 TypeScript 等返回类型里的大括号 \n if end_brace_idx + 1 < len(tokens) and tokens[end_brace_idx + 1][0] == '{':\n brace_idx = end_brace_idx + 1\n brace_count = 0\n for j in range(brace_idx, len(tokens)):\n if tokens[j][0] == '{':\n brace_count += 1\n elif tokens[j][0] == '}':\n brace_count -= 1\n if brace_count == 0:\n end_brace_idx = j\n break\n \n start_line = tokens[start_idx][1]\n end_line = tokens[end_brace_idx][1]\n func_lines = end_line - start_line + 1\n \n if func_lines > 15:\n return False\n \n idx += 1\n \n return True\n\n def extract_scripts(code: str, ext: str) -> str:\n pattern = re.compile(r'<script\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\n if not pattern.search(code):\n return \"\"\n \n result = []\n last_idx = 0\n for m in pattern.finditer(code):\n start = m.start(1)\n end = m.end(1)\n newlines = code[last_idx:start].count('\\n')\n result.append('\\n' * newlines)\n result.append(code[start:end])\n last_idx = end\n return \"\".join(result)\n\n def check_python_functions(code: str) -> bool:\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno:\n line_count = node.end_lineno - node.lineno + 1\n if line_count > 15:\n return False\n except Exception:\n pass\n return True\n\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n if not check_python_functions(code):\n return False\n continue\n \n if ext in {'.vue', '.html'}:\n code_to_check = extract_scripts(code, ext)\n else:\n code_to_check = code\n \n if not code_to_check.strip():\n continue\n \n tokens = lex(code_to_check)\n if not check_functions(tokens):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"提高效率的必要前提是，必须尽可能并行触发多个独立的工具调用","validation_code":"","tags":["Workflow","Orchestration"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个刷题系统的管理后台基础页面骨架。需求如下：1. 整体布局分为左、上、右三部分。左侧是导航侧边栏，包含“题库管理”、“用户管理”、“权限管理”三个主菜单项；顶部是Header栏，左侧放系统Logo标题，右侧放一个“当前用户”的占位显示；右侧是主内容区域。2. 请实现基本的路由切换或视图切换逻辑：点击左侧不同菜单时，右侧内容区相应地切换标题（如显示“题库管理内容区”、“用户管理内容区”即可），暂时不需要实现任何具体的业务功能，留出这三个页面的代码扩展位置。\\n另外请你后续所有操作始终遵循以下要求：请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）；生成代码时，请将各函数的行数严格控制在 50 行及以内。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用 4 个空格作为缩进标准\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n all_codes = []\\n \\n # 1. 提取 response 中的 Markdown 代码块\\n in_code = False\\n current_block = []\\n for line in response.split('\\\\n'):\\n if line.strip().startswith('```'):\\n if in_code:\\n all_codes.append('\\\\n'.join(current_block))\\n current_block = []\\n in_code = False\\n else:\\n in_code = True\\n elif in_code:\\n current_block.append(line)\\n \\n # 2. 读取 workspace_path 下生成的代码文件\\n if os.path.isdir(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 排除日志文件\\n if file.endswith('.log'):\\n continue\\n \\n # 筛选通常的项目代码文件\\n ext = os.path.splitext(file)[1].lower()\\n valid_exts = ['.html', '.css', '.js', '.py', '.vue', '.ts', '.jsx', '.tsx', '.json', '.java', '.c', '.cpp', '.cs', '.php', '.go']\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n all_codes.append(f.read())\\n except Exception:\\n pass\\n\\n # 如果没有任何代码被发现，默认为未违背规则\\n if not all_codes:\\n return True\\n\\n invalid_lines_count = 0\\n total_indented_lines = 0\\n\\n # 3. 逐行判断前导缩进\\n for code in all_codes:\\n lines = code.split('\\\\n')\\n for line in lines:\\n # 提取行首空白（同时保证该行不全为空白）\\n match = re.match(r'^([ \\\\t]+)(?=\\\\S)', line)\\n if match:\\n ws = match.group(1)\\n total_indented_lines += 1\\n \\n # 如果包含 Tab，直接不符合要求\\n if '\\\\t' in ws:\\n return False\\n \\n # 如果空格数量不是 4 的倍数，记录为不规范行\\n if len(ws) % 4 != 0:\\n invalid_lines_count += 1\\n\\n # 4. 判断违规行数占比\\n # 允许少数由意外折行造成的对齐误差（阈值设为最多3行或者占比不超过5%），超过则视为未统一使用 4 个空格\\n if invalid_lines_count > 0:\\n if invalid_lines_count > 3 or (invalid_lines_count / total_indented_lines > 0.05):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_lower_camel(name: str) -> bool:\\n # 排除可能被正则误当作方法名抓取的语言保留字或关键字\\n BLACKLIST = {\\n 'if', 'for', 'while', 'switch', 'catch', 'return', \\n 'else', 'elseif', 'elif', 'constructor', 'super', \\n 'with', 'function', 'try', 'finally', 'do', 'yield', 'await',\\n 'get', 'set'\\n }\\n if name in BLACKLIST:\\n return True\\n # 检查是否满足小驼峰格式：以小写字母开头，由大小写字母及数字组成，不能有下划线或其他特殊字符\\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 常用代码文件的后缀集合，过滤掉类似 .log 的非代码文件\\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.java', '.c', '.cpp', '.cs', '.go'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n funcs = []\\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n funcs.append(node.name)\\n except SyntaxError:\\n # 如果代码结构不完整导致 ast 解析失败，则回退到正则提取\\n for m in re.finditer(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(', content):\\n funcs.append(m.group(1))\\n else:\\n # 提取 function functionName(...) 形式，包含 Generator function\\n for m in re.finditer(r'\\\\bfunction\\\\s*\\\\*?\\\\s*([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(', content):\\n funcs.append(m.group(1))\\n \\n # 提取 const/let/var funcName = (...) => 或者 = function(...) 形式\\n for m in re.finditer(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\s*\\\\*?\\\\s*\\\\(|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_][a-zA-Z0-9_]*\\\\s*=>)', content):\\n funcs.append(m.group(1))\\n \\n # 提取类的方法或对象的方法 methodName(...) { \\n # 采用行首及缩进开始限制，防止误匹配部分不相关的中间代码\\n for m in re.finditer(r'^\\\\s*(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+)?([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', content, re.MULTILINE):\\n funcs.append(m.group(1))\\n \\n # 提取 Vue 等对象属性风格的方法声明 propertyName: function(...) {\\n for m in re.finditer(r'\\\\b([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*:\\\\s*(?:async\\\\s+)?function\\\\s*\\\\*?\\\\s*\\\\(', content):\\n funcs.append(m.group(1))\\n \\n # 依次对当前文件提取到的所有函数名进行判定\\n for func_name in funcs:\\n if not check_lower_camel(func_name):\\n return False\\n\\n # 若所有提取出的函数都符合小驼峰要求，或未包含函数，均视为满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 验证工作区路径是否存在且为目录\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义正则表达式匹配面向对象语言中的类声明\\n # 匹配诸如 `class MyClass`, `class MyClass(object):`, `class MyClass {` 中的 `MyClass`\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)')\\n \\n # 定义需要扫描的代码文件后缀集合\\n valid_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.cpp', '.c', '.h', '.cs',\\n '.php', '.rb', '.go', '.swift', '.html', '.htm', '.vue'\\n }\\n\\n # 遍历当前项目的工作区文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）以及常见的依赖库和构建目录，防止误判第三方代码库的类名\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in {'node_modules', 'venv', 'env', 'dist', 'build', '__pycache__'}]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找当前文件中的所有类名声明\\n for match in class_pattern.finditer(content):\\n cls_name = match.group(1)\\n # 如果存在的任意一个类名的字符长度不是偶数，则判定不满足要求\\n if len(cls_name) % 2 != 0:\\n return False\\n except Exception:\\n # 忽略因编码等问题无法读取的文件\\n pass\\n\\n # 如果代码中所有被提取的类名长度都恰好是偶数（或者完全没有写任何类名），均满足题目要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 针对 JavaScript, Java, C, C++, CSS 等使用的 C-Style 解析器\\n def get_comment_lines_c_style(text):\\n comment_lines = set()\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n escape = False\\n line_num = 1\\n \\n i = 0\\n n = len(text)\\n while i < n:\\n char = text[i]\\n is_escaped = escape\\n \\n if escape:\\n escape = False\\n else:\\n if char == '\\\\\\\\':\\n escape = True\\n else:\\n if not in_string and not in_line_comment and not in_block_comment:\\n if char in ('\\\"', \\\"'\\\", \\\"`\\\"):\\n in_string = True\\n string_char = char\\n elif char == '/' and i + 1 < n:\\n if text[i+1] == '/':\\n in_line_comment = True\\n comment_lines.add(line_num)\\n i += 1\\n elif text[i+1] == '*':\\n in_block_comment = True\\n comment_lines.add(line_num)\\n i += 1\\n elif in_string:\\n if char == string_char:\\n in_string = False\\n elif in_line_comment:\\n if char != '\\\\n':\\n comment_lines.add(line_num)\\n elif in_block_comment:\\n comment_lines.add(line_num)\\n if char == '*' and i + 1 < n and text[i+1] == '/':\\n in_block_comment = False\\n i += 1\\n \\n if in_line_comment and char == '\\\\n':\\n in_line_comment = False\\n\\n if char == '\\\\n':\\n # 单行字符串换行自动跳出，增强容错能力\\n if in_string and string_char != '`' and not is_escaped:\\n in_string = False\\n line_num += 1\\n \\n i += 1\\n return comment_lines\\n\\n # 针对 Python, Ruby 等使用的解析器\\n def get_comment_lines_python(text):\\n comment_lines = set()\\n in_string = False\\n string_char = ''\\n in_comment = False\\n escape = False\\n line_num = 1\\n \\n i = 0\\n n = len(text)\\n while i < n:\\n char = text[i]\\n is_escaped = escape\\n \\n if escape:\\n escape = False\\n else:\\n if char == '\\\\\\\\':\\n escape = True\\n else:\\n if not in_string and not in_comment:\\n if text[i:i+3] in ('\\\"\\\"\\\"', \\\"'''\\\"):\\n in_string = True\\n string_char = text[i:i+3]\\n i += 2\\n elif char in ('\\\"', \\\"'\\\"):\\n in_string = True\\n string_char = char\\n elif char == '#':\\n in_comment = True\\n comment_lines.add(line_num)\\n elif in_string:\\n if string_char in ('\\\"\\\"\\\"', \\\"'''\\\"):\\n if text[i:i+3] == string_char:\\n in_string = False\\n i += 2\\n elif char == string_char:\\n in_string = False\\n elif in_comment:\\n if char != '\\\\n':\\n comment_lines.add(line_num)\\n \\n if in_comment and char == '\\\\n':\\n in_comment = False\\n \\n if char == '\\\\n':\\n if in_string and string_char not in ('\\\"\\\"\\\"', \\\"'''\\\") and not is_escaped:\\n in_string = False\\n line_num += 1\\n \\n i += 1\\n return comment_lines\\n\\n # 针对 HTML 标签使用的解析器\\n def get_comment_lines_html(text):\\n comment_lines = set()\\n in_comment = False\\n line_num = 1\\n i = 0\\n n = len(text)\\n while i < n:\\n if not in_comment:\\n if text[i:i+4] == '<!--':\\n in_comment = True\\n comment_lines.add(line_num)\\n i += 3\\n else:\\n comment_lines.add(line_num)\\n if text[i:i+3] == '-->':\\n in_comment = False\\n i += 2\\n \\n if text[i] == '\\\\n':\\n line_num += 1\\n \\n i += 1\\n return comment_lines\\n\\n # 检查工作空间路径\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义常见的代码扩展名\\n CODE_EXTS = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.c', '.cpp', '.cs', \\n '.css', '.html', '.htm', '.vue', '.go', '.rs', '.rb', '.php', '.swift', '.kt'\\n }\\n \\n # 遍历项目下所有文件进行判断\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in CODE_EXTS:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n comment_lines = set()\\n \\n # 针对不同扩展名进行各自适应的解析判断\\n if ext in {'.py', '.rb'}:\\n comment_lines.update(get_comment_lines_python(content))\\n elif ext in {'.html', '.htm'}:\\n comment_lines.update(get_comment_lines_html(content))\\n comment_lines.update(get_comment_lines_c_style(content))\\n elif ext in {'.vue', '.php'}:\\n comment_lines.update(get_comment_lines_html(content))\\n comment_lines.update(get_comment_lines_c_style(content))\\n comment_lines.update(get_comment_lines_python(content))\\n else:\\n comment_lines.update(get_comment_lines_c_style(content))\\n \\n # 若发现任何一行不是第5行的代码包含注释，证明未满足要求\\n for line in comment_lines:\\n if line != 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，务必让每一个文件名都以 my_ 开始\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名，用于识别哪些文件是代码文件\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', \\n '.ts', '.tsx', '.vue', '.py', '.java', \\n '.c', '.cpp', '.cc', '.h', '.hpp', \\n '.go', '.rs', '.php', '.rb', '.swift'\\n }\\n \\n # 记录是否实际生成了代码文件\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 获取文件扩展名并统一转换为小写\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 如果该文件是代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n # 检查文件名是否以 my_ 开始\\n if not file_name.startswith('my_'):\\n return False\\n \\n # 确保生成了代码文件，且所有代码文件均满足要求\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在 50 行及以内\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n def check_py_file(file_path: str) -> bool:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\\n lines_count = node.end_lineno - node.lineno + 1\\n if lines_count > 50:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n def check_js_content(content: str) -> bool:\\n # 正则表达式匹配可能属于函数的特征起始\\n pattern = re.compile(\\n r'(\\\\bfunction\\\\b|=>\\\\s*\\\\{|\\\\b(?!(?:if|for|while|switch|catch)\\\\b)[a-zA-Z_$][\\\\w$]*\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{)'\\n )\\n \\n length = len(content)\\n i = 0\\n while i < length:\\n match = pattern.search(content, i)\\n if not match:\\n break\\n \\n start = match.start()\\n brace_index = content.find('{', start)\\n if brace_index == -1:\\n i = start + 1\\n continue\\n \\n brace_count = 0\\n in_str = False\\n str_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n end_index = -1\\n j = brace_index\\n while j < length:\\n c = content[j]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n elif in_block_comment:\\n if c == '*' and j + 1 < length and content[j+1] == '/':\\n in_block_comment = False\\n j += 1\\n elif in_str:\\n if c == '\\\\\\\\':\\n j += 1\\n elif c == str_char:\\n in_str = False\\n else:\\n if c == '/' and j + 1 < length:\\n if content[j+1] == '/':\\n in_line_comment = True\\n j += 1\\n elif content[j+1] == '*':\\n in_block_comment = True\\n j += 1\\n elif c in ('\\\"', \\\"'\\\", '`'):\\n in_str = True\\n str_char = c\\n elif c == '{':\\n brace_count += 1\\n elif c == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_index = j\\n break\\n j += 1\\n \\n if end_index != -1:\\n func_text = content[start:end_index+1]\\n lines_count = len(func_text.split('\\\\n'))\\n if lines_count > 50:\\n return False\\n # 继续在原区块内部或者接下来的位置进行寻找嵌套函数\\n i = start + 1\\n else:\\n i = brace_index + 1\\n\\n return True\\n\\n def check_file(file_path: str) -> bool:\\n if file_path.endswith('.py'):\\n return check_py_file(file_path)\\n else:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file_path.endswith(('.html', '.htm', '.vue')):\\n scripts = re.findall(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n for script in scripts:\\n if not check_js_content(script):\\n return False\\n return True\\n else:\\n return check_js_content(content)\\n except Exception:\\n pass\\n return True\\n\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 扫描常见的后端、前端代码文件\\n if file.endswith(('.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue')):\\n if not check_file(file_path):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来在这个基础上丰富“题库管理”页面的界面。请在题库管理内容区实现一个完整的数据操作界面：1. 顶部需要一个操作栏，左边包含一个搜索框和一个题目类型下拉筛选器（选项：单选题、多选题、简答题），右边有一个“上传/新增题目”按钮。2. 下方是一个题目列表表格，包含表头：题目ID、题干简述、题型、难度、状态（已上架/已下架）。3. 表格每一行的最右侧有一列“操作”，里面提供三个文字按钮：“编辑”、“切换上下架”、“删除”。请使用Mock数据渲染出几行表格内容，当点击“删除”按钮时，能把这一行从视图中移除。\\n另外请你后续所有操作始终遵循以下新要求：生成代码时，请将各函数的行数严格控制在30行及以内；操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用 4 个空格作为缩进标准\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的代码文件扩展名\\n code_exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.vue', '.java', '.cpp', '.c', '.h', '.hpp', '.cs', '.php', '.rb', '.go', '.rs', '.swift', '.kt'}\\n \\n total_valid = 0\\n total_invalid = 0\\n has_tab = False\\n \\n # 首先检查工作区中的代码文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n stripped = line.strip()\\n # 忽略空行\\n if not stripped:\\n continue\\n # 忽略可能是多行注释或其他注释导致的特殊对齐\\n if stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('#') or stripped.startswith('<!--'):\\n continue\\n \\n # 匹配行首的空白字符\\n match = re.match(r'^([\\\\t ]+)', line)\\n if match:\\n indent = match.group(1)\\n if '\\\\t' in indent:\\n has_tab = True\\n # 判断空格数是否为 4 的倍数\\n if len(indent) % 4 == 0:\\n total_valid += 1\\n else:\\n total_invalid += 1\\n\\n # 如果工作区中有进行过有效缩进统计，直接得出结论\\n if total_valid + total_invalid > 0:\\n if has_tab:\\n return False\\n # 考虑到代码换行对齐时的非4倍数空格，允许有一定的容错(阈值设为0.8)\\n return (total_valid / (total_valid + total_invalid)) >= 0.8\\n \\n # 如果工作区没有生成文件或者未找到任何缩进行，退化到检查 response 中的代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n for line in lines:\\n stripped = line.strip()\\n if not stripped:\\n continue\\n if stripped.startswith('*') or stripped.startswith('//') or stripped.startswith('#') or stripped.startswith('<!--'):\\n continue\\n \\n match = re.match(r'^([\\\\t ]+)', line)\\n if match:\\n indent = match.group(1)\\n if '\\\\t' in indent:\\n has_tab = True\\n if len(indent) % 4 == 0:\\n total_valid += 1\\n else:\\n total_invalid += 1\\n\\n if total_valid + total_invalid > 0:\\n if has_tab:\\n return False\\n return (total_valid / (total_valid + total_invalid)) >= 0.8\\n \\n # 如果不存在任何具有缩进的代码（比如全部顶格），默认视为没有违反约束\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式\",\n \"validation_code\": \"def check_function_names_lower_camel_case(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 判断是否为小驼峰命名：必须以小写字母开头，后续字符只能是字母和数字\\n def is_lower_camel_case(name: str) -> bool:\\n if not name:\\n return False\\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\\n\\n # 去除代码中的注释，防止注释中的内容导致误判\\n def remove_comments(content: str) -> str:\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n return content\\n\\n # 定义前端及常见类C语法的函数匹配正则\\n # 匹配：function myFunc (\\n pattern_func1 = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\(')\\n # 匹配：const myFunc = () => 或者 const myFunc = function() \\n pattern_func2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n # 匹配类方法或对象方法：myMethod() {\\n pattern_func4 = re.compile(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n # 匹配对象属性赋值函数：myFunc: () =>\\n pattern_func5 = re.compile(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n # 匹配对象属性绑定函数：obj.myFunc = () =>\\n pattern_func6 = re.compile(r'\\\\.\\\\s*([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)')\\n\\n # 需要排除的 JS/TS/Java 等的控制流或内置关键字\\n keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'catch', 'finally', \\n 'return', 'super', 'constructor', 'function', 'with', 'typeof', 'instanceof'\\n }\\n \\n code_extensions = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.go'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n\\n if ext == '.py':\\n # Python 文件直接使用 AST 分析，绝对精准\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n tree = ast.parse(f.read())\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if not is_lower_camel_case(node.name):\\n return False\\n except Exception:\\n pass\\n\\n elif ext in code_extensions:\\n # 其他代码文件使用正则匹配提取函数名\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n\\n content = remove_comments(content)\\n functions = []\\n \\n functions.extend(pattern_func1.findall(content))\\n functions.extend(pattern_func2.findall(content))\\n functions.extend(pattern_func5.findall(content))\\n functions.extend(pattern_func6.findall(content))\\n \\n methods = pattern_func4.findall(content)\\n for m in methods:\\n if m not in keywords:\\n functions.append(m)\\n\\n for func in functions:\\n if not is_lower_camel_case(func):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数\",\n \"validation_code\": \"def check_class_name_length_is_even(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义需要扫描的代码文件后缀集合\\n valid_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', '.cpp', \\n '.c', '.h', '.hpp', '.cs', '.php', '.go', '.rs', \\n '.swift', '.kt', '.rb', '.html', '.htm', '.vue'\\n }\\n \\n # 匹配 class 定义的正则表达式\\n # \\\\bclass\\\\s+ : 匹配 'class ' 及其后跟随的空白字符\\n # ([A-Za-z_][A-Za-z0-9_]*) : 匹配合法的标识符作为类名\\n # \\\\b : 确保类名被完整匹配并排除相连符号\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_][A-Za-z0-9_]*)\\\\b')\\n \\n # 排除语法关键字，防止在匿名类中将其误认为类名（如 'class extends Base'）\\n excluded_keywords = {'extends', 'implements'}\\n \\n # 遍历当前项目工作空间的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_extensions:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 格式读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果文件读取失败（例如编码问题或不可读），则跳过\\n continue\\n \\n # 提取文件中的所有类名\\n matches = class_pattern.findall(content)\\n for class_name in matches:\\n # 跳过关键字\\n if class_name in excluded_keywords:\\n continue\\n \\n # 校验提取出的类名字符长度是否恰好是偶数\\n if len(class_name) % 2 != 0:\\n return False\\n \\n # 如果项目中没有任何一个类名的长度为奇数（或没有定义类），则符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码扩展名，用来过滤并仅检查代码文件\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', \\n '.java', '.c', '.cpp', '.h', '.hpp', '.py', '.sh', \\n '.yml', '.yaml', '.css', '.scss', '.less'\\n }\\n \\n # 针对前端、类 C 语言的代码注释匹配模式 (含 HTML/Vue, JS/TS, C/Java 等)\\n # 优先匹配字符串字面量以避开引号内注释符；对于 // 单行注释使用 (?<!:) 规避对 http:// 的误判\\n pattern_c_like = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(?P<html_comment><!--[\\\\s\\\\S]*?-->)|'\\n r'(?P<multi_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)|'\\n r'(?P<single_comment>(?<!:)//.*)'\\n )\\n \\n # 针对 Python 或脚本语言代码注释匹配模式\\n pattern_py = re.compile(\\n r'(?P<py_multi_string>\\\\\\\"\\\\\\\"\\\\\\\"[\\\\s\\\\S]*?\\\\\\\"\\\\\\\"\\\\\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\')|'\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|'\\n r'(?P<hash_comment>#.*)'\\n )\\n \\n # 针对 CSS 等只支持 /* ... */ 注释模式\\n pattern_css = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')|'\\n r'(?P<multi_comment>/\\\\*[\\\\s\\\\S]*?\\\\*/)'\\n )\\n \\n # 将正则匹配跨越的文本索引，转化为包含行号的集合（1-based）\\n def get_lines_covered(text: str, start: int, end: int):\\n lines = text.split('\\\\n')\\n line_starts = []\\n current_pos = 0\\n for line in lines:\\n line_starts.append(current_pos)\\n current_pos += len(line) + 1 # 包含 \\\\n 的长度\\n \\n start_line = 1\\n end_line = 1\\n \\n for i, pos in enumerate(line_starts):\\n if start >= pos:\\n start_line = i + 1\\n if end > pos:\\n end_line = i + 1\\n \\n return set(range(start_line, end_line + 1))\\n\\n # 遍历所有项目文件\\n for root_dir, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_path = os.path.join(root_dir, file)\\n try:\\n # 默认使用 Universal Newlines 模式读取，将 \\\\r\\\\n 转换成 \\\\n 使得索引精确映射\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n # 根据文件类型选择对应的正则处理器\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.htm', '.java', '.c', '.cpp', '.h', '.hpp', '.scss', '.less']:\\n pattern = pattern_c_like\\n elif ext in ['.py', '.sh', '.yml', '.yaml']:\\n pattern = pattern_py\\n elif ext in ['.css']:\\n pattern = pattern_css\\n else:\\n continue\\n \\n # 查找并计算所有注释所在的行号范围\\n for match in pattern.finditer(text):\\n group_name = match.lastgroup\\n # 若捕获组名为注释组（忽略掉前面用来吃掉字符串的 string 组）\\n if group_name in ['html_comment', 'multi_comment', 'single_comment', 'hash_comment']:\\n covered_lines = get_lines_covered(text, match.start(), match.end())\\n for line_num in covered_lines:\\n # 发现不仅在第5行出现了注释，或注释覆盖跨度包含了非第5行的位置，则违反要求\\n if line_num != 5:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，务必让每一个文件名都以 my_ 开始\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令背景推测，生成的主要是前端页面及 Mock 逻辑文件\\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue'}\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件的后缀名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为推断的代码文件类型\\n if ext in code_extensions:\\n # 校验文件名是否以 my_ 开头\\n if not file.startswith(\\\"my_\\\"):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在 30 行及以内\",\n \"validation_code\": \"def check_function_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 判断大括号 '{' 所在位置的前缀是否是函数的开头\\n def check_is_func(clean_text: str, pos: int) -> bool:\\n p = pos - 1\\n \\n # 辅助跳过空白字符\\n def skip_whitespace(idx: int) -> int:\\n while idx >= 0 and clean_text[idx] in ' \\\\t\\\\n\\\\r':\\n idx -= 1\\n return idx\\n\\n p = skip_whitespace(p)\\n if p < 0:\\n return False\\n \\n # 判断是否为箭头函数 '=>'\\n if clean_text[p] == '>' and p > 0 and clean_text[p-1] == '=':\\n return True\\n \\n # 判断是否带有参数括号 ')'\\n if clean_text[p] == ')':\\n depth = 1\\n p -= 1\\n # 跳过成对的括号组合 (...)\\n while p >= 0 and depth > 0:\\n if clean_text[p] == ')':\\n depth += 1\\n elif clean_text[p] == '(':\\n depth -= 1\\n p -= 1\\n \\n if depth > 0:\\n return False\\n \\n p = skip_whitespace(p)\\n if p < 0:\\n return False\\n \\n # 提取紧跟着括号前面的单词\\n word = ''\\n while p >= 0 and re.match(r'[a-zA-Z0-9_$]', clean_text[p]):\\n word = clean_text[p] + word\\n p -= 1\\n \\n if word:\\n # 排除常见的非函数作用域的控制流关键字\\n if word in ['if', 'for', 'while', 'switch', 'catch', 'new', 'with']:\\n return False\\n return True\\n else:\\n return False\\n \\n return False\\n\\n # 解析包含大括号特性的常规前端及后端语言代码\\n def check_braces_file(file_path: str) -> bool:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n masked = []\\n in_str = False\\n str_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n i = 0\\n length = len(content)\\n # 第一阶段：掩码化屏蔽字符串内容与注释，避免误匹配大括号\\n while i < length:\\n c = content[i]\\n if in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n masked.append(c)\\n else:\\n masked.append(' ')\\n elif in_block_comment:\\n if c == '*' and i + 1 < length and content[i+1] == '/':\\n in_block_comment = False\\n masked.append(' ')\\n masked.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n masked.append('\\\\n')\\n else:\\n masked.append(' ')\\n elif in_str:\\n if c == '\\\\\\\\':\\n masked.append(' ')\\n if i + 1 < length:\\n masked.append(' ' if content[i+1] != '\\\\n' else '\\\\n')\\n i += 1\\n elif c == str_char:\\n in_str = False\\n masked.append(' ')\\n else:\\n if c == '\\\\n':\\n masked.append('\\\\n')\\n else:\\n masked.append(' ')\\n else:\\n if c == '/' and i + 1 < length and content[i+1] == '/':\\n in_line_comment = True\\n masked.append(' ')\\n masked.append(' ')\\n i += 1\\n elif c == '/' and i + 1 < length and content[i+1] == '*':\\n in_block_comment = True\\n masked.append(' ')\\n masked.append(' ')\\n i += 1\\n elif c in ('\\\"', \\\"'\\\", \\\"`\\\"):\\n in_str = True\\n str_char = c\\n masked.append(' ')\\n else:\\n masked.append(c)\\n i += 1\\n \\n clean_text = ''.join(masked)\\n \\n # 第二阶段：依靠清理后的安全文本记录大括号深度并计算块的行数\\n stack = []\\n line_num = 1\\n \\n for pos, c in enumerate(clean_text):\\n if c == '\\\\n':\\n line_num += 1\\n elif c == '{':\\n is_func = check_is_func(clean_text, pos)\\n stack.append((line_num, is_func))\\n elif c == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n if line_num - start_line + 1 > 30:\\n return False\\n return True\\n\\n # 针对 Python 文件的精确行数解析\\n def check_python_file(file_path: str) -> bool:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 判断行数是否严格控制在 30 行以内\\n length = node.end_lineno - node.lineno + 1\\n if length > 30:\\n return False\\n except Exception:\\n # 存在不能解析的语法错误，予以跳过\\n pass\\n return True\\n\\n # 遍历当前生成项目所有的代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉常见的生成目录或者无关的包依赖文件夹\\n dirs[:] = [d for d in dirs if d not in ['.git', 'node_modules', 'dist', 'build', '.venv', 'venv', '__pycache__']]\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断对应后缀的代码文件\\n if ext == '.py':\\n if not check_python_file(file_path):\\n return False\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.java', '.cpp', '.c', '.cs', '.go']:\\n if not check_braces_file(file_path):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来完善“上传/新增题目”和“编辑”的核心交互。请实现：点击“新增题目”或某行的“编辑”按钮时，弹出一个居中表单弹窗。这个表单需要有一套具有挑战性的动态联动逻辑：1. 包含基础字段（题干、解析、难度）。2. 包含一个“题目类型”选择器。当选择“单选题”时，下方渲染出一组选项输入框（A, B, C, D等），并带有一个单选按钮用于标记哪个是正确答案，同时允许新增和删除选项；当选择“多选题”时，逻辑类似，但标记正确答案的地方变为多选框；当切换为“简答题”时，之前的选项区全部隐藏，替换为一个大的多行文本输入框用于填写参考答案。填写完弹窗后点击“确定”，要求关闭弹窗并在后面的表格顶端新增这条题目记录。\\n另外请你后续所有操作始终遵循以下要求：生成代码时，请将各函数的行数严格控制在25行及以内；执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用 4 个空格作为缩进标准\",\n \"validation_code\": \"def check_indentation_standard(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义可能包含缩进格式的常见代码文件扩展名\\n code_extensions = {'.py', '.js', '.html', '.css', '.vue', '.ts', '.jsx', '.tsx', '.json', '.java', '.cpp', '.c', '.h'}\\n \\n total_indented_lines = 0\\n valid_4_space_lines = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 忽略空行或全是空白字符的行\\n if line.strip() == '':\\n continue\\n \\n # 仅匹配行首的半角空格和制表符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n leading_whitespace = match.group(1)\\n \\n # 如果使用了制表符缩进，则直接不满足要求\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n num_spaces = len(leading_whitespace)\\n total_indented_lines += 1\\n \\n # 判断空格数是否为 4 的倍数\\n if num_spaces % 4 == 0:\\n valid_4_space_lines += 1\\n\\n # 如果存在缩进行，判断 4 个空格缩进的比例是否占据绝大多数（容错比例设为20%，用来包容少量的代码参数对齐换行）\\n if total_indented_lines > 0:\\n if valid_4_space_lines / total_indented_lines < 0.8:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 小驼峰命名格式正则：首字母小写，只允许包含字母和数字\\n is_lower_camel_case = re.compile(r'^[a-z][a-zA-Z0-9]*$')\\n \\n # 需要排除的常用语法关键字，避免被正则误认为函数名\\n keywords = {\\n 'if', 'for', 'while', 'catch', 'switch', 'return', \\n 'constructor', 'super', 'function', 'else', 'import', \\n 'export', 'try', 'finally', 'with', 'await', 'yield', \\n 'typeof', 'instanceof', 'new', 'delete', 'void', \\n 'class', 'interface', 'implements', 'extends'\\n }\\n\\n # 1. 常规 function / func 声明\\n func_decl_pattern = re.compile(r'\\\\b(?:function|func)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n\\n # 2. JS/TS 等函数表达式 / 箭头函数赋值\\n func_expr_pattern = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s*)?(?:function\\\\s*\\\\(|\\\\(.*?\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)')\\n\\n # 3. 对象字面量中的方法赋值\\n obj_method_pattern = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s*)?(?:function\\\\s*\\\\(|\\\\(.*?\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)')\\n\\n # 4. 类中的方法定义（兼容 TS/JS, Java, C#, C++ 等大括号语言）\\n method_pattern = re.compile(\\n r'^\\\\s*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+|async\\\\s+|final\\\\s+|abstract\\\\s+|override\\\\s+|virtual\\\\s+)*'\\n r'(?:[a-zA-Z_$][a-zA-Z0-9_$<>\\\\[\\\\]]*\\\\s+)?'\\n r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^{]+)?\\\\s*\\\\{', \\n re.MULTILINE\\n )\\n\\n # 5. Go 带有接收者的方法定义\\n go_method_pattern = re.compile(r'\\\\bfunc\\\\s+\\\\([^)]+\\\\)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 常见代码文件\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.go']:\\n found_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 汇总当前文件提取到的所有函数/方法名\\n all_funcs = (\\n func_decl_pattern.findall(content) +\\n func_expr_pattern.findall(content) +\\n obj_method_pattern.findall(content) +\\n method_pattern.findall(content) +\\n go_method_pattern.findall(content)\\n )\\n \\n for fn in all_funcs:\\n if fn in keywords:\\n continue\\n if not is_lower_camel_case.match(fn):\\n return False\\n \\n # Python 文件通过 AST 安全提取\\n elif ext in ['.py']:\\n found_code_file = True\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 包括类方法和 __init__ 在内，如果未满足小驼峰则直接判定失败\\n if not is_lower_camel_case.match(node.name):\\n return False\\n except Exception:\\n continue\\n\\n # 如果正确解析到了代码文件且中途没有被判定为False，说明全面符合要求\\n return found_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\\n # 结合用户指令推断可能的前后端代码文件扩展名\\n code_extensions = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\\n \\n # 匹配类定义中类名的正则表达式\\n # 能够匹配 \\\"class MyClass \\\" 等常见类的定义形式，且过滤掉HTML中 \\\"class=\\\" 的干扰\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_][A-Za-z0-9_]*)\\\\b')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找当前文件内所有符合类名定义的字符串\\n class_names = class_pattern.findall(content)\\n \\n for name in class_names:\\n # 针对要求中明确指出的“字符长度”要求，使用 len 函数进行统计判断\\n if len(name) % 2 != 0:\\n return False\\n except Exception:\\n # 读取解析文件出现异常则忽略当前文件\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n code_exts = {\\n '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', \\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.php', '.sh'\\n }\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令中“务必让每一个文件名都以 my_ 开始”过滤文件\\n if not file.startswith('my_'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n comment_lines = set()\\n in_string = False\\n string_char = None\\n in_multiline_comment = False\\n multiline_type = None \\n \\n # 根据不同的文件扩展名设定该文件可能包含的注释类型\\n allow_c_multi = ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.css', '.java', '.c', '.cpp', '.cs', '.go', '.php'}\\n allow_html_multi = ext in {'.vue', '.html', '.xml', '.svg', '.php'}\\n allow_c_single = ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.java', '.c', '.cpp', '.cs', '.go', '.php'}\\n allow_hash_single = ext in {'.py', '.sh', '.bash', '.rb', '.yml', '.yaml', '.dockerfile'}\\n \\n for i, line in enumerate(lines):\\n j = 0\\n line_has_comment = False\\n \\n while j < len(line):\\n # 当既不在多行注释内，也不在字符串内时\\n if not in_multiline_comment and not in_string:\\n # 检查是否进入了字符串区域\\n if line[j] in ['\\\"', \\\"'\\\", '`']:\\n in_string = True\\n string_char = line[j]\\n j += 1\\n continue\\n \\n # 检查多行注释起点 /*\\n if allow_c_multi and line[j:j+2] == '/*':\\n in_multiline_comment = True\\n multiline_type = 'C'\\n line_has_comment = True\\n j += 2\\n continue\\n \\n # 检查多行注释起点 <!--\\n if allow_html_multi and line[j:j+4] == '<!--':\\n in_multiline_comment = True\\n multiline_type = 'HTML'\\n line_has_comment = True\\n j += 4\\n continue\\n \\n # 检查单行注释起点 //\\n if allow_c_single and line[j:j+2] == '//':\\n line_has_comment = True\\n break\\n \\n # 检查单行注释起点 #\\n if allow_hash_single and line[j] == '#':\\n line_has_comment = True\\n break\\n \\n elif in_string:\\n # 跳过转义字符，避免遇到 \\\\\\\" 或 \\\\' 误退出字符串状态\\n if line[j] == '\\\\\\\\':\\n j += 2\\n continue\\n # 检查是否结束了字符串区域\\n elif line[j] == string_char:\\n in_string = False\\n j += 1\\n continue\\n \\n elif in_multiline_comment:\\n line_has_comment = True\\n # 检查 C 风格多行注释是否结束\\n if multiline_type == 'C' and line[j:j+2] == '*/':\\n in_multiline_comment = False\\n j += 2\\n continue\\n # 检查 HTML 风格多行注释是否结束\\n if multiline_type == 'HTML' and line[j:j+3] == '-->':\\n in_multiline_comment = False\\n j += 3\\n continue\\n \\n j += 1\\n \\n # 如果当前行结束时仍处于多行注释中，本行也应视为有注释\\n if in_multiline_comment:\\n line_has_comment = True\\n \\n if line_has_comment:\\n comment_lines.add(i + 1)\\n \\n # 针对常规单引号和双引号的不规范跨行情况，行末直接重置字符串状态避免误判\\n if in_string and string_char in ['\\\"', \\\"'\\\"]:\\n in_string = False\\n \\n # 移除被允许的第 5 行注释（如果存在）\\n comment_lines.discard(5)\\n \\n # 若仍存在非第 5 行的注释，则说明未满足条件\\n if len(comment_lines) > 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，务必让每一个文件名都以 my_ 开始\",\n \"validation_code\": \"def check_code_filename_prefix(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推断，主要为前端相关代码文件，也可以加入部分其他语言的代码文件后缀\\n code_extensions = {\\n '.html', '.htm', '.css', '.scss', '.less',\\n '.js', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.php', '.go'\\n }\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀名并转换为小写，以进行统一匹配\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n \\n # 检查文件名是否以 my_ 开始\\n if not file.startswith('my_'):\\n return False\\n\\n # 要求必须生成了符合规范的代码文件\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在25行及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 辅助状态机函数，将JS/TS/Vue等文件中的字符串和注释清空为空格，以防止干扰函数正文的识别，同时保持原有行数不变。\\n def clean_code(code: str) -> str:\\n result = []\\n state = 0\\n i = 0\\n n = len(code)\\n while i < n:\\n c = code[i]\\n if state == 0:\\n if c == '/' and i+1 < n and code[i+1] == '/':\\n state = 1\\n result.append(' ')\\n result.append(' ')\\n i += 2\\n continue\\n elif c == '/' and i+1 < n and code[i+1] == '*':\\n state = 2\\n result.append(' ')\\n result.append(' ')\\n i += 2\\n continue\\n elif c == \\\"'\\\":\\n state = 3\\n result.append(' ')\\n i += 1\\n continue\\n elif c == '\\\"':\\n state = 4\\n result.append(' ')\\n i += 1\\n continue\\n elif c == '`':\\n state = 5\\n result.append(' ')\\n i += 1\\n continue\\n else:\\n result.append(c)\\n i += 1\\n elif state == 1:\\n if c == '\\\\n':\\n state = 0\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 2:\\n if c == '*' and i+1 < n and code[i+1] == '/':\\n state = 0\\n result.append(' ')\\n result.append(' ')\\n i += 2\\n else:\\n if c == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 3:\\n if c == '\\\\\\\\':\\n result.append(' ')\\n if i+1 < n:\\n result.append('\\\\n' if code[i+1] == '\\\\n' else ' ')\\n i += 2\\n elif c == \\\"'\\\":\\n state = 0\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 4:\\n if c == '\\\\\\\\':\\n result.append(' ')\\n if i+1 < n:\\n result.append('\\\\n' if code[i+1] == '\\\\n' else ' ')\\n i += 2\\n elif c == '\\\"':\\n state = 0\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n elif state == 5:\\n if c == '\\\\\\\\':\\n result.append(' ')\\n if i+1 < n:\\n result.append('\\\\n' if code[i+1] == '\\\\n' else ' ')\\n i += 2\\n elif c == '`':\\n state = 0\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append('\\\\n')\\n else:\\n result.append(' ')\\n i += 1\\n return ''.join(result)\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 正则表达式寻找前端函数定义的开头特征\\n pattern = re.compile(\\n r'(?<!\\\\w)(?!(?:if|for|while|switch|catch|with|else|return|new|import|export)\\\\b)'\\n r'(?:function(?:\\\\s+\\\\w+)?\\\\s*\\\\([^)]*\\\\)|(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|\\\\w+)\\\\s*=>|(?:async\\\\s+)?(?:\\\\w+)\\\\s*\\\\([^)]*\\\\)|(?:\\\\w+)\\\\s*:\\\\s*function\\\\s*\\\\([^)]*\\\\))\\\\s*\\\\{'\\n )\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据要求过滤日志和无关文件\\n if file.endswith('.log'):\\n continue\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ['.png', '.jpg', '.jpeg', '.gif', '.zip', '.tar', '.gz', '.pdf', '.docx', '.mp4']:\\n continue\\n \\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n # 使用 AST 解析 Python 代码\\n if ext == '.py':\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n line_count = node.end_lineno - node.lineno + 1\\n if line_count > 25:\\n return False\\n except Exception:\\n pass\\n else:\\n # 对待各类前端页面/逻辑代码进行匹配运算\\n cleaned_code = clean_code(code)\\n for m in pattern.finditer(cleaned_code):\\n start_pos = m.start()\\n bracket_pos = cleaned_code.find('{', start_pos)\\n if bracket_pos == -1:\\n continue\\n \\n # 使用栈匹配大括号，准确找到函数结束位置\\n count = 0\\n end_pos = bracket_pos\\n for i in range(bracket_pos, len(cleaned_code)):\\n if cleaned_code[i] == '{':\\n count += 1\\n elif cleaned_code[i] == '}':\\n count -= 1\\n if count == 0:\\n end_pos = i\\n break\\n if count != 0:\\n continue\\n \\n # 包含换行符的数量即可转换为对应的代码行数跨度\\n line_count = cleaned_code.count('\\\\n', start_pos, end_pos) + 1\\n if line_count > 25:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"不错，题库这块可以了。下一步我们做“用户管理”页面。页面主体同样是一个数据表格，展示几个测试用户的基本信息：用户ID、用户昵称、注册时间、总刷题数、整体正确率。在表格的“操作”列中，放置一个“查看答题记录”按钮。点击该按钮后，请从屏幕右侧滑出一个抽屉（Drawer）组件。在这个抽屉里面，要求使用垂直时间轴（Timeline）组件的形式，展示该用户最近的答题流水记录，每条时间轴节点需要清晰地显示：答题时间、对应的题目名称概要，以及一个醒目的Tag标签（绿色表示“回答正确”，红色表示“回答错误”）。\\n另外请你后续所有操作始终遵循以下新要求：生成代码时，请将各函数的行数严格控制在20行及以内；只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用 4 个空格作为缩进标准\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 基于多轮指令预测可能涉及的代码文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.json', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.cs', '.go', '.rb', '.php'\\n }\\n \\n space_4n_lines = 0\\n space_not_4n_lines = 0\\n tab_indent_lines = 0\\n \\n def process_lines(lines):\\n nonlocal space_4n_lines, space_not_4n_lines, tab_indent_lines\\n for line in lines:\\n line_str = line.rstrip('\\\\n\\\\r')\\n if not line_str.strip():\\n continue\\n \\n # 匹配每一行的前导空白字符及后续内容\\n m = re.match(r'^([ \\\\t]+)(\\\\S.*)', line_str)\\n if m:\\n spaces = m.group(1)\\n content = m.group(2)\\n \\n # 排除可能的不规则注释对齐（避免误杀因多行注释导致的前导不规则空格）\\n if content.startswith('*') or content.startswith('/*') or content.startswith('<!--'):\\n continue\\n \\n if '\\\\t' in spaces:\\n tab_indent_lines += 1\\n else:\\n if len(spaces) % 4 == 0:\\n space_4n_lines += 1\\n else:\\n space_not_4n_lines += 1\\n\\n # 1. 检查 workspace_path 下生成的项目代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略隐藏目录与编译依赖目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', 'dist', 'build', 'out')]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n file_lines = f.readlines()\\n \\n # 判断是否为直接保存进去的 Diff 文件格式内容（防 LLM 出错存储成Diff文件）\\n content_start = \\\"\\\".join(file_lines[:15])\\n if '@@ -' in content_start and ' +' in content_start:\\n continue\\n \\n process_lines(file_lines)\\n except Exception:\\n continue\\n\\n # 2. 检查最后一轮回复内容 (response) 中可能存在的代码块\\n pattern = re.compile(r'```([a-zA-Z0-9+\\\\-]*)\\\\n(.*?)\\\\n```', re.DOTALL)\\n for match in pattern.finditer(response):\\n lang = match.group(1).lower()\\n # 跳过非代码块的 diff, patch 或 log（因 Unified Diff 中空格被赋予了特殊增加与删除标记的含义）\\n if lang in ('diff', 'patch', 'log'):\\n continue\\n \\n code_text = match.group(2)\\n # 即便未显式声明语言类型，如果特征匹配 Unified Diff，也应跳过验证\\n if '@@ -' in code_text and ' +' in code_text:\\n continue\\n \\n code_lines = code_text.split('\\\\n')\\n process_lines(code_lines)\\n\\n total_indents = space_4n_lines + space_not_4n_lines + tab_indent_lines\\n \\n # 若全文中未检查到任何行首进行缩进的行，理论上并未违反此约束\\n if total_indents == 0:\\n return True\\n \\n # 计算严格符合 4 的倍数的空格缩进行在全部缩进行之中的占比\\n ratio = space_4n_lines / total_indents\\n \\n # 遵循 4 个空格缩进规范的项目，绝大部分缩进行必然是 4 的倍数，但可能存在极少量的方法传参换行对齐；\\n # 相比之下，若采用 2 空格做标准，此比例仅徘徊在 40% ~ 50% 区间，若用 Tab 缩进则占比趋近 0；\\n # 因此设定 0.75（75%）作为极高鲁棒性的分水岭阈值进行评判。\\n return ratio >= 0.75\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 若工作空间路径不存在直接返回 False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义待扫描和解析的代码文件后缀（主要覆盖前端及部分常见后端）\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py'}\\n \\n # 检测名称是否满足小驼峰命名\\n def check_camel_case(name):\\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', name))\\n \\n # 去除代码中的注释以防止正则误匹配\\n def remove_comments(code, ext):\\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', code)\\n elif ext == '.py':\\n code = re.sub(r'#.*', '', code)\\n code = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", '', code)\\n code = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '', code)\\n return code\\n\\n # 遍历项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n content = remove_comments(content, ext)\\n names = []\\n \\n # 解析不同语言中的函数名称\\n if ext in {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html'}:\\n # 1. 明确的 function 关键字声明（含 generator 函数）\\n names.extend(re.findall(r'\\\\bfunction\\\\s*\\\\*?\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(', content))\\n # 2. 箭头函数赋值表达式\\n names.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z0-9_]+)\\\\s*=>', content))\\n # 3. 匿名函数赋值表达式\\n names.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z0-9_]+)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\b', content))\\n \\n # 4. 对象方法、类方法定义等\\n pattern = r'^[ \\\\t]*(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+|async\\\\s+|get\\\\s+|set\\\\s+)*([a-zA-Z0-9_]+)\\\\s*(?:<[^>]+>)?\\\\s*\\\\(([^)]*)\\\\)(?::\\\\s*[^{=;]+)?\\\\s*\\\\{'\\n for match in re.finditer(pattern, content, re.MULTILINE):\\n name = match.group(1)\\n args = match.group(2)\\n # 排除控制流等关键字\\n if name in {'if', 'for', 'while', 'switch', 'catch', 'function', 'return', 'with'}:\\n continue\\n # 避免误伤带有回调的外部函数调用形式\\n if 'function' in args:\\n continue\\n names.append(name)\\n \\n elif ext == '.py':\\n names.extend(re.findall(r'\\\\bdef\\\\s+([a-zA-Z0-9_]+)\\\\s*\\\\(', content))\\n \\n # 针对所有提取出的函数名称进行合规校验\\n for name in names:\\n if not check_camel_case(name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify_class_name_length(response: str, workspace_path: str) -> bool:\\n # 若项目目录尚不存在，说明未生成违规文件\\n if not os.path.isdir(workspace_path):\\n return True\\n\\n # 常见可能包含类定义的代码文件扩展名\\n code_extensions = {\\n '.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.cpp', '.cc', '.cxx', \\n '.h', '.hpp', '.cs', '.php', '.rb', '.go', '.swift', '.kt', '.vue'\\n }\\n \\n # 匹配 'class' 关键字，然后是一个或多个空白符，接着是类名标识符（不能以数字开头）\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)')\\n \\n # 在某些语言（如JS/TS等）中可能使用匿名类继承，导致直接捕获到保留字，在此做过滤\\n ignore_keywords = {'extends', 'implements'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 查找当前代码文件中所有的类名\\n classes = class_pattern.findall(content)\\n for cls_name in classes:\\n if cls_name in ignore_keywords:\\n continue\\n \\n # 检查该类名的字符长度是否为偶数\\n if len(cls_name) % 2 != 0:\\n return False\\n except Exception:\\n # 如果存在不可读文件或编码异常，直接忽略该文件以防阻断验证\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n def is_code_file(filename):\\n # 定义需要被检查的常见代码文件后缀名\\n exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.htm', \\n '.css', '.scss', '.less', '.java', '.c', '.cpp', '.h', '.hpp',\\n '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.cs', '.m'}\\n _, ext = os.path.splitext(filename)\\n return ext.lower() in exts\\n\\n def get_comment_lines(text, file_ext):\\n comment_lines = set()\\n state = 'NORMAL'\\n lines = text.split('\\\\n')\\n \\n for line_idx, line in enumerate(lines):\\n i = 0\\n n = len(line)\\n # 单行注释如果在换行时结束，则重置状态为 NORMAL\\n if state in ['IN_SINGLE_LINE_COMMENT_C', 'IN_SINGLE_LINE_COMMENT_SHELL']:\\n state = 'NORMAL'\\n \\n while i < n:\\n if state == 'NORMAL':\\n if file_ext in {'.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.cs', '.go', '.rs', '.php', '.swift', '.kt', '.vue', '.html', \\n '.htm', '.css', '.scss', '.less'}:\\n # 对于 js、ts 等支持 // 注释。注意标准的 CSS 不支持 // 作为注释，所以予以排除。\\n if line[i:i+2] == '//' and file_ext not in {'.css'}:\\n state = 'IN_SINGLE_LINE_COMMENT_C'\\n comment_lines.add(line_idx + 1)\\n i += 2\\n continue\\n elif line[i:i+2] == '/*':\\n state = 'IN_MULTILINE_COMMENT_C'\\n comment_lines.add(line_idx + 1)\\n i += 2\\n continue\\n elif line[i:i+4] == '<!--' and file_ext in {'.html', '.htm', '.vue'}:\\n state = 'IN_MULTILINE_COMMENT_HTML'\\n comment_lines.add(line_idx + 1)\\n i += 4\\n continue\\n \\n if file_ext in {'.py', '.rb', '.php', '.sh'}:\\n if line[i] == '#':\\n state = 'IN_SINGLE_LINE_COMMENT_SHELL'\\n comment_lines.add(line_idx + 1)\\n i += 1\\n continue\\n \\n if line[i] == '\\\"':\\n state = 'IN_STRING_DOUBLE'\\n elif line[i] == \\\"'\\\":\\n state = 'IN_STRING_SINGLE'\\n elif line[i] == '`' and file_ext in {'.js', '.ts', '.jsx', '.tsx', '.vue', '.go'}:\\n state = 'IN_STRING_BACKTICK'\\n \\n i += 1\\n \\n elif state == 'IN_STRING_DOUBLE':\\n if line[i] == '\\\\\\\\':\\n i += 2\\n elif line[i] == '\\\"':\\n state = 'NORMAL'\\n i += 1\\n else:\\n i += 1\\n elif state == 'IN_STRING_SINGLE':\\n if line[i] == '\\\\\\\\':\\n i += 2\\n elif line[i] == \\\"'\\\":\\n state = 'NORMAL'\\n i += 1\\n else:\\n i += 1\\n elif state == 'IN_STRING_BACKTICK':\\n if line[i] == '\\\\\\\\':\\n i += 2\\n elif line[i] == '`':\\n state = 'NORMAL'\\n i += 1\\n else:\\n i += 1\\n elif state == 'IN_MULTILINE_COMMENT_C':\\n comment_lines.add(line_idx + 1)\\n if line[i:i+2] == '*/':\\n state = 'NORMAL'\\n i += 2\\n else:\\n i += 1\\n elif state == 'IN_MULTILINE_COMMENT_HTML':\\n comment_lines.add(line_idx + 1)\\n if line[i:i+3] == '-->':\\n state = 'NORMAL'\\n i += 3\\n else:\\n i += 1\\n elif state in ['IN_SINGLE_LINE_COMMENT_C', 'IN_SINGLE_LINE_COMMENT_SHELL']:\\n comment_lines.add(line_idx + 1)\\n i += 1\\n \\n # 处理完该行后，如果仍处于多行注释状态中，证明整行都属于多行注释的覆盖范围\\n if state in ['IN_MULTILINE_COMMENT_C', 'IN_MULTILINE_COMMENT_HTML']:\\n comment_lines.add(line_idx + 1)\\n \\n return comment_lines\\n\\n # 遍历工作区目录，核对所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if is_code_file(file):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n _, ext = os.path.splitext(file)\\n ext = ext.lower()\\n \\n # 获取该文件所有存在注释的行号集合\\n comments = get_comment_lines(text, ext)\\n \\n # 规则约束：除了第 5 行之外，其余代码行均不得添加任何注释\\n illegal_comments = comments - {5}\\n if illegal_comments:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，务必让每一个文件名都以 my_ 开始\",\n \"validation_code\": \"def check_code_filenames(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的扩展名集合，根据指令推测大概率为前端及常见编程语言代码\\n code_extensions = {\\n \\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".vue\\\", \\n \\\".py\\\", \\\".java\\\", \\\".c\\\", \\\".cpp\\\", \\\".go\\\", \\\".rs\\\", \\\".php\\\", \\\".swift\\\"\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 忽略隐藏文件或系统生成文件\\n if file.startswith('.'):\\n continue\\n \\n # 获取文件的后缀名\\n _, ext = os.path.splitext(file)\\n ext = ext.lower()\\n \\n # 如果是代码文件，验证其是否以 my_ 开头\\n if ext in code_extensions:\\n if not file.startswith(\\\"my_\\\"):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在20行及以内\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n # 解析 Python 文件中的函数行数\\n def get_func_length_in_py(content: str):\\n funcs = []\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'lineno') and hasattr(node, 'end_lineno'):\\n funcs.append((node.lineno, node.end_lineno))\\n except Exception:\\n pass\\n return funcs\\n\\n # 解析 类C/前端 文件中的函数行数（通过自建的微型词法解析大括号匹配）\\n def get_func_length_brace_lang(content: str):\\n funcs = []\\n in_single_quote = False\\n in_double_quote = False\\n in_backtick = False\\n in_single_comment = False\\n in_multi_comment = False\\n escape = False\\n \\n depth = 0\\n line_num = 1\\n func_starts = {}\\n \\n i = 0\\n n = len(content)\\n while i < n:\\n char = content[i]\\n \\n # 换行处理\\n if char == '\\\\n':\\n line_num += 1\\n in_single_comment = False\\n escape = False\\n i += 1\\n continue\\n \\n if escape:\\n escape = False\\n i += 1\\n continue\\n \\n if in_single_comment:\\n i += 1\\n continue\\n \\n if in_multi_comment:\\n if char == '*' and i + 1 < n and content[i+1] == '/':\\n in_multi_comment = False\\n i += 2\\n continue\\n i += 1\\n continue\\n \\n if in_single_quote:\\n if char == '\\\\\\\\':\\n escape = True\\n elif char == \\\"'\\\":\\n in_single_quote = False\\n i += 1\\n continue\\n \\n if in_double_quote:\\n if char == '\\\\\\\\':\\n escape = True\\n elif char == '\\\"':\\n in_double_quote = False\\n i += 1\\n continue\\n \\n if in_backtick:\\n if char == '\\\\\\\\':\\n escape = True\\n elif char == '`':\\n in_backtick = False\\n i += 1\\n continue\\n \\n # 判断是否进入注释或字符串\\n if char == '/' and i + 1 < n:\\n if content[i+1] == '/':\\n in_single_comment = True\\n i += 2\\n continue\\n elif content[i+1] == '*':\\n in_multi_comment = True\\n i += 2\\n continue\\n elif char == \\\"'\\\":\\n in_single_quote = True\\n elif char == '\\\"':\\n in_double_quote = True\\n elif char == '`':\\n in_backtick = True\\n elif char == '{':\\n depth += 1\\n # 获取大括号前面的内容片段，去除上一条语句的影响\\n prefix = content[max(0, i-200):i]\\n last_idx = max(prefix.rfind(';'), prefix.rfind('{'), prefix.rfind('}'))\\n if last_idx != -1:\\n prefix = prefix[last_idx+1:]\\n \\n is_func = False\\n # 判断是否属于 function funcName(...) 或者 function(...)\\n if re.search(r'\\\\bfunction\\\\b\\\\s*[A-Za-z0-9_]*\\\\s*\\\\([^)]*\\\\)\\\\s*$', prefix):\\n is_func = True\\n # 判断是否属于箭头函数 => \\n elif re.search(r'=>\\\\s*$', prefix):\\n is_func = True\\n else:\\n # 判断是否是常见的类方法声明 methodName(...)\\n match = re.search(r'\\\\b([A-Za-z0-9_]+)\\\\s*\\\\([^)]*\\\\)\\\\s*$', prefix)\\n # 排除掉常见的控制流关键字\\n if match and match.group(1) not in ['if', 'for', 'while', 'switch', 'catch', 'with']:\\n is_func = True\\n \\n if is_func:\\n func_starts[depth] = line_num\\n \\n elif char == '}':\\n if depth in func_starts:\\n start_line = func_starts.pop(depth)\\n funcs.append((start_line, line_num))\\n depth -= 1\\n if depth < 0:\\n depth = 0\\n func_starts.clear()\\n \\n i += 1\\n \\n return funcs\\n\\n # 支持扫描的目标文件扩展名（包括前端常见扩展名与常规后台语言）\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.php', '.rb', '.rs'\\n }\\n \\n ignore_dirs = {'.git', 'node_modules', '__pycache__', 'dist', 'build', 'venv', '.env'}\\n \\n # 遍历项目文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 屏蔽不相干的构建和依赖文件夹\\n dirs[:] = [d for d in dirs if d not in ignore_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 根据语言执行相应的行数提取\\n if ext == '.py':\\n funcs = get_func_length_in_py(content)\\n else:\\n funcs = get_func_length_brace_lang(content)\\n \\n # 验证提取的每一个函数\\n for start_line, end_line in funcs:\\n length = end_line - start_line + 1\\n if length > 20:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后一步，我们把“权限管理”概念融入进去，实现全局的角色控制功能。请在顶部Header最右侧的“当前用户”占位处，改为一个下拉菜单，允许在两个角色之间来回切换：“超级管理员”和“普通运营者”。这需要实现跨组件的权限联动：1. 当切换为“普通运营者”时，左侧导航栏的“权限管理”菜单项需要直接隐藏；同时，在“题库管理”页面的表格中，所有行的“删除”按钮要变为置灰不可点击的禁用状态。2. 当切换回“超级管理员”时，左侧隐藏的菜单恢复显示，“删除”按钮也恢复可点击。请确保这种角色切换在UI上有即时的视觉反馈。\\n另外请你后续所有操作始终遵循以下要求：生成代码时，请将各函数的行数严格控制在15行及以内；提高效率的必要前提是，必须尽可能并行触发多个独立的工具调用。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"你生成给用户的回复内容，必须一直使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件中绝对不能使用多层if\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，其循环嵌套的最高层数被限制为 2 层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"要求输出代码时，统一使用 4 个空格作为缩进标准\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 推断可能的代码文件类型集合（包含常见的前端和后端扩展名）\\n code_extensions = {\\n '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \\n '.css', '.scss', '.less', '.py', '.java', '.c', \\n '.cpp', '.go', '.rs', '.php'\\n }\\n \\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 忽略无法以纯文本/UTF-8读取的异常文件\\n continue\\n\\n for i, line in enumerate(lines):\\n # 根据其余要求，第5行允许存在注释，以防特殊格式导致不规则缩进，故跳过第5行\\n if i == 4:\\n continue\\n \\n # 忽略完全是空行、无内容的行\\n if not line.strip():\\n continue\\n \\n # 匹配当前行行首的空白字符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n indent = match.group(1)\\n \\n # 不能包含任何制表符\\n if '\\\\t' in indent:\\n return False\\n \\n # 缩进的空格数必须恰好是4的整数倍\\n if len(indent) % 4 != 0:\\n return False\\n\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件内，任何一个函数名都需采用 小驼峰 命名格式\",\n \"validation_code\": \"def check_function_names_lower_camel(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 验证是否为小驼峰命名\\n def is_lower_camel_case(s: str) -> bool:\\n if not s:\\n return False\\n # 小驼峰: 必须以小写字母开头，后续只能包含字母或数字（不包含下划线，不包含 $）\\n return bool(re.match(r'^[a-z][a-zA-Z0-9]*$', s))\\n \\n # 去除代码中的各类注释以避免干扰\\n def remove_comments(code: str) -> str:\\n # 移除HTML注释 <!-- ... -->\\n code = re.sub(r'<!--.*?-->', '', code, flags=re.DOTALL)\\n # 移除多行注释 /* ... */\\n code = re.sub(r'/\\\\*.*?\\\\*/', '', code, flags=re.DOTALL)\\n # 移除JS单行注释 // ...\\n code = re.sub(r'//.*', '', code)\\n # 移除Python单行注释 # ...\\n code = re.sub(r'#.*', '', code)\\n return code\\n\\n # 常规控制流关键字等白名单（防止被当做函数名处理）\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'with', 'return', \\n 'function', 'yield', 'await', 'import', 'export', 'typeof',\\n 'instanceof', 'elif', 'except', 'finally', 'try'\\n }\\n\\n # 用于捕获主流语言常见函数声明的正则表达式集合\\n patterns = [\\n # JS/TS 传统函数: function funcName (\\n r'function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(',\\n # JS/TS 箭头函数: const/let/var funcName = (...) => 或者 = async (...) =>\\n r'(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>',\\n # JS/TS 对象方法: funcName: function (\\n r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*function\\\\s*\\\\(',\\n # Python 函数: def funcName (\\n r'def\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('\\n ]\\n \\n # 单独行匹配，用来捕获类方法、Vue/React生命周期等（例如： methodName(args) { ）\\n line_pattern = r'^\\\\s*(?:async\\\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{'\\n \\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅处理前端/后端主流代码文件，略过静态资源等\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.js', '.ts', '.vue', '.html', '.jsx', '.tsx', '.py', '.java', '.c', '.cpp', '.cs', '.go']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n content_no_comments = remove_comments(content)\\n \\n # 1. 对整个文件内容做全局模式的捕获\\n for pat in patterns:\\n matches = re.finditer(pat, content_no_comments)\\n for m in matches:\\n func_name = m.group(1)\\n if func_name in keywords:\\n continue\\n if not is_lower_camel_case(func_name):\\n return False\\n \\n # 2. 逐行匹配类/对象方法的定义\\n for line in content_no_comments.split('\\\\n'):\\n m = re.match(line_pattern, line)\\n if m:\\n func_name = m.group(1)\\n if func_name in keywords:\\n continue\\n if not is_lower_camel_case(func_name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请确保你在代码文件中生成的每一个类名，其字符长度都恰好是偶数\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区路径不存在，判定为不满足（或根据实际测试台环境视作异常）\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义涉及的前后端主流代码文件扩展名集合\\n valid_exts = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.vue', '.html', \\n '.java', '.cs', '.cpp', '.cc', '.h', '.hpp', '.php', \\n '.rb', '.go', '.swift', '.kt', '.scala', '.dart'\\n }\\n \\n # 正则表达式匹配类定义：\\n # \\\\bclass\\\\s+ 匹配 \\\"class\\\" 关键字后紧跟至少一个空白字符\\n # ([A-Za-z_][A-Za-z0-9_]*) 捕获组提取类名的标识符\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_][A-Za-z0-9_]*)')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n\\n # 简单去除代码注释及多行字符串的干扰，防止误匹配到注释里的文本\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'''.*?'''\\\", '', content, flags=re.DOTALL)\\n\\n # 提取出该代码文件中的所有类名\\n matches = class_pattern.findall(content)\\n for class_name in matches:\\n # 指令中明确提到“字符长度”，直接使用原生的 len() 函数\\n if len(class_name) % 2 != 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"除了你生成的每个代码文件中的第 5 行之外，其余代码行均不得添加任何注释\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见代码文件扩展名及对应的注释风格分类\\n code_extensions = {\\n '.py': 'python',\\n '.js': 'c_style',\\n '.jsx': 'c_style',\\n '.ts': 'c_style',\\n '.tsx': 'c_style',\\n '.java': 'c_style',\\n '.c': 'c_style',\\n '.cpp': 'c_style',\\n '.go': 'c_style',\\n '.css': 'c_style',\\n '.html': 'html_style',\\n '.vue': 'html_style'\\n }\\n\\n # 预编译各类语言下匹配“字符串常量”与“注释”的正则表达式组合\\n # 巧妙利用正则表达式先匹配字符串组、后匹配注释组，确保字符串内部的注释符号不会被当作注释捕获\\n patterns = {\\n 'python': re.compile(\\n r'(?P<string>\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\')|'\\n r'(?P<comment>#.*?$)',\\n re.MULTILINE\\n ),\\n 'c_style': re.compile(\\n r'(?P<string>\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`[\\\\s\\\\S]*?`)|'\\n r'(?P<comment>//.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/)',\\n re.MULTILINE\\n ),\\n 'html_style': re.compile(\\n r'(?P<string>\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"|\\\\'(?:[^\\\\'\\\\\\\\]|\\\\\\\\.)*\\\\'|`[\\\\s\\\\S]*?`)|'\\n r'(?P<comment><!--[\\\\s\\\\S]*?-->|//.*?$|/\\\\*[\\\\s\\\\S]*?\\\\*/)',\\n re.MULTILINE\\n )\\n }\\n\\n # 遍历代码工作区下所有的文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略可能的隐藏目录（如 .git）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 严格依据用户第8条系统指令，只读取以 my_ 命名的生成代码文件\\n if not file.startswith('my_'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n file_type = code_extensions[ext]\\n pattern = patterns[file_type]\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 使用正则迭代查找文件中的所有匹配项\\n for match in pattern.finditer(content):\\n # 只要存在属于注释分组的内容\\n if match.group('comment'):\\n start_pos = match.start('comment')\\n end_pos = match.end('comment')\\n \\n # 基于当前匹配位置前面的换行符数量来计算其代码所在行号（从 1 开始计行）\\n start_line = content.count('\\\\n', 0, start_pos) + 1\\n end_line = content.count('\\\\n', 0, end_pos) + 1\\n \\n # 验证注释是否越权出现在了不该出现的位置\\n for line_num in range(start_line, end_line + 1):\\n # 一旦注释出现在除第 5 行外的其它行，立刻返回失败\\n if line_num != 5:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时，务必让每一个文件名都以 my_ 开始\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名集合\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', \\n '.vue', '.svelte', '.py', '.java', '.c', '.cpp', \\n '.h', '.hpp', '.go', '.rs', '.php', '.rb', '.swift', \\n '.kt', '.cs', '.sh', '.dart'\\n }\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名\\n _, ext = os.path.splitext(file)\\n \\n # 判断是否为代码文件\\n if ext.lower() in code_extensions:\\n # 检查文件名是否以 my_ 开头\\n if not file.startswith('my_'):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在15行及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def lex(code: str) -> list:\\n tokens = []\\n i = 0\\n line_num = 1\\n n = len(code)\\n \\n stack = []\\n \\n while i < n:\\n c = code[i]\\n \\n # 处理模板字符串的内部状态\\n if stack and stack[-1] == 'template':\\n if code[i] == '\\\\\\\\':\\n i += 2\\n elif code[i] == '`':\\n stack.pop()\\n i += 1\\n elif code[i:i+2] == '${':\\n stack.append('interpolation')\\n tokens.append(('{', line_num)) # 使用 { 保证插值大括号平衡\\n i += 2\\n else:\\n if code[i] == '\\\\n': line_num += 1\\n i += 1\\n continue\\n\\n if c == '\\\\n':\\n line_num += 1\\n i += 1\\n elif c in ' \\\\t\\\\r':\\n i += 1\\n elif c == '/' and i + 1 < n and code[i+1] == '/':\\n i += 2\\n while i < n and code[i] != '\\\\n':\\n i += 1\\n elif c == '/' and i + 1 < n and code[i+1] == '*':\\n i += 2\\n while i < n:\\n if code[i] == '\\\\n':\\n line_num += 1\\n i += 1\\n elif code[i] == '*' and i + 1 < n and code[i+1] == '/':\\n i += 2\\n break\\n else:\\n i += 1\\n elif c == \\\"'\\\":\\n i += 1\\n while i < n:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n elif code[i] == \\\"'\\\":\\n i += 1\\n break\\n else:\\n if code[i] == '\\\\n': line_num += 1\\n i += 1\\n elif c == '\\\"':\\n i += 1\\n while i < n:\\n if code[i] == '\\\\\\\\':\\n i += 2\\n elif code[i] == '\\\"':\\n i += 1\\n break\\n else:\\n if code[i] == '\\\\n': line_num += 1\\n i += 1\\n elif c == '`':\\n stack.append('template')\\n i += 1\\n elif code[i:i+2] == '=>':\\n tokens.append(('=>', line_num))\\n i += 2\\n elif c == '{':\\n tokens.append(('{', line_num))\\n i += 1\\n elif c == '}':\\n if stack and stack[-1] == 'interpolation':\\n stack.pop()\\n tokens.append(('}', line_num))\\n else:\\n tokens.append(('}', line_num))\\n i += 1\\n else:\\n m = re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*', code[i:])\\n if m:\\n tokens.append((m.group(0), line_num))\\n i += m.end()\\n elif c in '()':\\n tokens.append((c, line_num))\\n i += 1\\n else:\\n i += 1\\n return tokens\\n\\n def check_functions(tokens: list) -> bool:\\n idx = 0\\n blacklist = {\\n 'if', 'for', 'while', 'switch', 'catch', 'return', 'new', \\n 'import', 'export', 'super', 'else', 'with', 'typeof', \\n 'instanceof', 'await', 'yield'\\n }\\n \\n while idx < len(tokens):\\n t, ln = tokens[idx]\\n start_idx = -1\\n brace_idx = -1\\n \\n # 模式1: 传统 function() {}\\n if t == 'function':\\n start_idx = idx\\n paren_count = 0\\n for j in range(idx + 1, len(tokens)):\\n if tokens[j][0] == '(':\\n paren_count += 1\\n elif tokens[j][0] == ')':\\n paren_count -= 1\\n elif tokens[j][0] == '{' and paren_count == 0:\\n brace_idx = j\\n break\\n \\n # 模式2: 箭头函数 () => {}\\n elif t == '=>':\\n start_idx = idx\\n if idx + 1 < len(tokens) and tokens[idx+1][0] == '{':\\n brace_idx = idx + 1\\n \\n # 模式3: 类方法或对象方法 identifier() {}\\n elif re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t) and t not in blacklist:\\n if idx + 1 < len(tokens) and tokens[idx+1][0] == '(':\\n paren_count = 0\\n found_close_paren = False\\n for j in range(idx + 1, len(tokens)):\\n if tokens[j][0] == '(':\\n paren_count += 1\\n elif tokens[j][0] == ')':\\n paren_count -= 1\\n if paren_count == 0:\\n found_close_paren = True\\n elif tokens[j][0] == '{' and paren_count == 0 and found_close_paren:\\n start_idx = idx\\n brace_idx = j\\n break\\n \\n # 若找到了函数的开端和它的左侧大括号\\n if brace_idx != -1:\\n brace_count = 0\\n end_brace_idx = -1\\n for j in range(brace_idx, len(tokens)):\\n if tokens[j][0] == '{':\\n brace_count += 1\\n elif tokens[j][0] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_brace_idx = j\\n break\\n \\n if end_brace_idx != -1:\\n # 避免误抓 TypeScript 等返回类型里的大括号 \\n if end_brace_idx + 1 < len(tokens) and tokens[end_brace_idx + 1][0] == '{':\\n brace_idx = end_brace_idx + 1\\n brace_count = 0\\n for j in range(brace_idx, len(tokens)):\\n if tokens[j][0] == '{':\\n brace_count += 1\\n elif tokens[j][0] == '}':\\n brace_count -= 1\\n if brace_count == 0:\\n end_brace_idx = j\\n break\\n \\n start_line = tokens[start_idx][1]\\n end_line = tokens[end_brace_idx][1]\\n func_lines = end_line - start_line + 1\\n \\n if func_lines > 15:\\n return False\\n \\n idx += 1\\n \\n return True\\n\\n def extract_scripts(code: str, ext: str) -> str:\\n pattern = re.compile(r'<script\\\\b[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\\n if not pattern.search(code):\\n return \\\"\\\"\\n \\n result = []\\n last_idx = 0\\n for m in pattern.finditer(code):\\n start = m.start(1)\\n end = m.end(1)\\n newlines = code[last_idx:start].count('\\\\n')\\n result.append('\\\\n' * newlines)\\n result.append(code[start:end])\\n last_idx = end\\n return \\\"\\\".join(result)\\n\\n def check_python_functions(code: str) -> bool:\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno:\\n line_count = node.end_lineno - node.lineno + 1\\n if line_count > 15:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n if not check_python_functions(code):\\n return False\\n continue\\n \\n if ext in {'.vue', '.html'}:\\n code_to_check = extract_scripts(code, ext)\\n else:\\n code_to_check = code\\n \\n if not code_to_check.strip():\\n continue\\n \\n tokens = lex(code_to_check)\\n if not check_functions(tokens):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"操作文件时，要求尽可能优先使用专用工具（如Read, Write, Edit），而不是去执行 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"执行任何文件相关操作前，请务必先运行`pwd`以明确当前所在的目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"只要是修改已经存在的文件，就必须严格输出Unified Diff格式，不可直接输出全量代码进行替换\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"提高效率的必要前提是，必须尽可能并行触发多个独立的工具调用\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["验证页面整体骨架与路由切换：观察系统整体布局，在超级管理员角色下分别点击左侧导航栏的'题库管理'、'用户管理'和'权限管理'菜单。预期：页面呈现左（侧边栏）、上（Header栏带Logo和角色/当前用户下拉区域）、右（主内容区）布局；点击左侧菜单时，右侧主内容区能够随之切换页面标题或内容区视图。","验证题库管理界面的静态渲染：进入'题库管理'页面，观察界面布局。预期：顶部操作栏包含搜索框、题目类型下拉筛选器（含单选题、多选题、简答题选项）以及'上传/新增题目'按钮；下方表格正确显示表头（题目ID、题干简述、题型、难度、状态）及Mock数据渲染的列表；每行最右侧操作列固定显示'编辑'、'切换上下架'、'删除'三个文字按钮。","验证题库表格的删除交互逻辑（超级管理员状态）：确认当前处于超级管理员角色，点击题库表格中任意一行的'删除'按钮。预期：该行数据立即从表格视图中被移除消失。","验证题库管理新增弹窗及保存逻辑：点击'上传/新增题目'按钮，在弹窗中填写基础字段并点击'确定'按钮。预期：点击新增时弹出一个居中的表单弹窗，弹窗内包含题干、解析、难度字段；点击确定后，弹窗关闭，且在表格最顶端新增出一条刚填写的题目记录行。","验证表单弹窗动态联动（单选题）：在题库弹窗中，将题目类型选择为'单选题'。预期：表单下方动态渲染出一组选项输入框（如A、B、C、D等），并且每一个选项附带一个单选按钮用于标记正确答案，同时界面上包含可以新增和删除某个选项的功能按钮。","验证表单弹窗动态联动（多选题）：在题库弹窗中，将题目类型切换为'多选题'。预期：选项区域仍保留并支持增删操作，但选项旁边的标记正确答案组件变为多选框组件，允许操作者同时勾选多个选项作为正确答案。","验证表单弹窗动态联动（简答题）：在题库弹窗中，将题目类型切换为'简答题'。预期：此前展示的所有选项输入框和增删选项按钮全部隐藏消失，对应区域替换渲染为一个巨大的多行文本输入框，用于填写参考答案。","验证用户管理表格与抽屉滑出交互：点击左侧进入'用户管理'页面，观察表格并点击任意一行的'查看答题记录'按钮。预期：表格正确展示用户ID、昵称、注册时间、总刷题数、整体正确率等数据；点击按钮后，从屏幕右侧平滑滑出一个抽屉（Drawer）面板。","验证抽屉内部的时间轴组件渲染：打开答题记录抽屉，观察内部数据展示结构。预期：抽屉内部采用垂直时间轴（Timeline）组件展示答题流水记录，每一个时间节点清晰包含答题时间、题目名称概要，并且配有视觉醒目的正确性Tag标签（绿色代表'回答正确'，红色代表'回答错误'）。","验证全局权限联动（普通运营者角色）：点击顶部Header最右侧的下拉菜单，将角色切换为'普通运营者'，观察左侧导航栏和题库管理表格。预期：左侧导航栏中的'权限管理'主菜单项立即隐藏消失；题库管理表格内所有数据行的'删除'按钮均变为置灰状态，且模拟点击无效（禁用状态）。","验证全局权限联动（恢复超级管理员角色）：再次点击顶部Header下拉菜单，将角色切换回'超级管理员'，观察左侧导航栏和题库管理表格。预期：左侧被隐藏的'权限管理'菜单重新恢复显示；题库表格中所有行的'删除'按钮视觉上恢复正常，且变为可点击状态。","验证题库管理编辑弹窗及保存逻辑：点击题库表格任意一行的'编辑'按钮。预期：弹出与新增题目一致的居中表单弹窗，表单中可编辑题干、解析、难度和题目类型相关内容；修改后点击'确定'，弹窗关闭，且表格中对应题目记录更新为修改后的内容。"],"string":"[\n \"验证页面整体骨架与路由切换：观察系统整体布局，在超级管理员角色下分别点击左侧导航栏的'题库管理'、'用户管理'和'权限管理'菜单。预期：页面呈现左（侧边栏）、上（Header栏带Logo和角色/当前用户下拉区域）、右（主内容区）布局；点击左侧菜单时，右侧主内容区能够随之切换页面标题或内容区视图。\",\n \"验证题库管理界面的静态渲染：进入'题库管理'页面，观察界面布局。预期：顶部操作栏包含搜索框、题目类型下拉筛选器（含单选题、多选题、简答题选项）以及'上传/新增题目'按钮；下方表格正确显示表头（题目ID、题干简述、题型、难度、状态）及Mock数据渲染的列表；每行最右侧操作列固定显示'编辑'、'切换上下架'、'删除'三个文字按钮。\",\n \"验证题库表格的删除交互逻辑（超级管理员状态）：确认当前处于超级管理员角色，点击题库表格中任意一行的'删除'按钮。预期：该行数据立即从表格视图中被移除消失。\",\n \"验证题库管理新增弹窗及保存逻辑：点击'上传/新增题目'按钮，在弹窗中填写基础字段并点击'确定'按钮。预期：点击新增时弹出一个居中的表单弹窗，弹窗内包含题干、解析、难度字段；点击确定后，弹窗关闭，且在表格最顶端新增出一条刚填写的题目记录行。\",\n \"验证表单弹窗动态联动（单选题）：在题库弹窗中，将题目类型选择为'单选题'。预期：表单下方动态渲染出一组选项输入框（如A、B、C、D等），并且每一个选项附带一个单选按钮用于标记正确答案，同时界面上包含可以新增和删除某个选项的功能按钮。\",\n \"验证表单弹窗动态联动（多选题）：在题库弹窗中，将题目类型切换为'多选题'。预期：选项区域仍保留并支持增删操作，但选项旁边的标记正确答案组件变为多选框组件，允许操作者同时勾选多个选项作为正确答案。\",\n \"验证表单弹窗动态联动（简答题）：在题库弹窗中，将题目类型切换为'简答题'。预期：此前展示的所有选项输入框和增删选项按钮全部隐藏消失，对应区域替换渲染为一个巨大的多行文本输入框，用于填写参考答案。\",\n \"验证用户管理表格与抽屉滑出交互：点击左侧进入'用户管理'页面，观察表格并点击任意一行的'查看答题记录'按钮。预期：表格正确展示用户ID、昵称、注册时间、总刷题数、整体正确率等数据；点击按钮后，从屏幕右侧平滑滑出一个抽屉（Drawer）面板。\",\n \"验证抽屉内部的时间轴组件渲染：打开答题记录抽屉，观察内部数据展示结构。预期：抽屉内部采用垂直时间轴（Timeline）组件展示答题流水记录，每一个时间节点清晰包含答题时间、题目名称概要，并且配有视觉醒目的正确性Tag标签（绿色代表'回答正确'，红色代表'回答错误'）。\",\n \"验证全局权限联动（普通运营者角色）：点击顶部Header最右侧的下拉菜单，将角色切换为'普通运营者'，观察左侧导航栏和题库管理表格。预期：左侧导航栏中的'权限管理'主菜单项立即隐藏消失；题库管理表格内所有数据行的'删除'按钮均变为置灰状态，且模拟点击无效（禁用状态）。\",\n \"验证全局权限联动（恢复超级管理员角色）：再次点击顶部Header下拉菜单，将角色切换回'超级管理员'，观察左侧导航栏和题库管理表格。预期：左侧被隐藏的'权限管理'菜单重新恢复显示；题库表格中所有行的'删除'按钮视觉上恢复正常，且变为可点击状态。\",\n \"验证题库管理编辑弹窗及保存逻辑：点击题库表格任意一行的'编辑'按钮。预期：弹出与新增题目一致的居中表单弹窗，表单中可编辑题干、解析、难度和题目类型相关内容；修改后点击'确定'，弹窗关闭，且表格中对应题目记录更新为修改后的内容。\"\n]","isConversational":false}}},{"rowIdx":84,"cells":{"task_id":{"kind":"number","value":85,"string":"85"},"repository_policy":{"kind":"string","value":"后续所有交互轮次你都必须严格遵循以下约束：\n1. 不论何时，你生成用户回复时都必须使用中文。\n2. 你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径。\n3. 凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式。\n4. 在你生成的所有代码文件中，注释内容必须统一使用中文编写。\n5. 保持你生成的代码文件里每个函数的圈复杂度均在 3 以内。\n6. 你在生成代码文件时，需将循环嵌套层数控制在 2 层以内。\n7. 你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符。\n8. 当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 如果项目路径不存在或不是目录，无法进行文件级别的验证，默认返回 True\n if not workspace_path or not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return True\n \n # 定义常见的代码及配置文件扩展名集合\n code_extensions = {\n '.py', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \n '.java', '.js', '.ts', '.jsx', '.tsx', \n '.html', '.css', '.scss', '.less',\n '.go', '.rs', '.cs', '.php', '.rb', '.sh', '.bash',\n '.swift', '.kt', '.kts', '.m', '.dart', '.r', '.lua',\n '.sql', '.json', '.yaml', '.yml', '.xml'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 忽略以点开头的隐藏文件（如 .gitignore 等系统级或配置级文件）\n if file_name.startswith('.'):\n continue\n \n # 分离主文件名和扩展名\n base_name, ext = os.path.splitext(file_name)\n \n # 如果是预期的代码文件\n if ext.lower() in code_extensions:\n # 检查主文件名是否仅包含小写字母和下划线\n # 不能包含任何其他字符（包括数字、大写字母、中划线、额外的点等）\n if not re.match(r'^[a-z_]+$', base_name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]}],"string":"[\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 如果项目路径不存在或不是目录，无法进行文件级别的验证，默认返回 True\\n if not workspace_path or not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return True\\n \\n # 定义常见的代码及配置文件扩展名集合\\n code_extensions = {\\n '.py', '.c', '.cpp', '.cc', '.cxx', '.h', '.hpp', \\n '.java', '.js', '.ts', '.jsx', '.tsx', \\n '.html', '.css', '.scss', '.less',\\n '.go', '.rs', '.cs', '.php', '.rb', '.sh', '.bash',\\n '.swift', '.kt', '.kts', '.m', '.dart', '.r', '.lua',\\n '.sql', '.json', '.yaml', '.yml', '.xml'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 忽略以点开头的隐藏文件（如 .gitignore 等系统级或配置级文件）\\n if file_name.startswith('.'):\\n continue\\n \\n # 分离主文件名和扩展名\\n base_name, ext = os.path.splitext(file_name)\\n \\n # 如果是预期的代码文件\\n if ext.lower() in code_extensions:\\n # 检查主文件名是否仅包含小写字母和下划线\\n # 不能包含任何其他字符（包括数字、大写字母、中划线、额外的点等）\\n if not re.match(r'^[a-z_]+$', base_name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Build & Deployment"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个“Markdown批量转换工具”的Web页面。第一步，先搭个基础骨架。页面需要一个顶部导航栏，左边是侧边栏用于放文件列表，右边是主工作区。主工作区里暂时放一个简单的占位提示“请选择文件”，左侧栏放一个“添加文件”的按钮。整体风格要简洁现代一点，暂不实现具体的业务逻辑。请你后续始终遵循两个要求：一是生成的代码文件必须满足完全不使用多层if的要求，二是在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return True\n\n pattern = re.compile(r'^[a-z_]+(\\.[a-z_]+)?$')\n\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n if not pattern.match(file_name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":1,"instruction":"不错，现在的界面有模有样了。接下来完善左侧的文件列表。我希望用户可以通过点击“添加文件”按钮一次性选择多个 `.md` 文件，或者直接把文件拖拽到左侧区域来添加。添加后，文件会以列表形式展示在左侧栏中，显示文件名。每个文件项旁边要有一个删除图标可以单独移除，列表最下方再加一个“清空列表”的按钮。请你后续始终遵循以下新增要求：一是无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划；二是每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端与配置相关的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.md', '.py', '.sh'}\n \n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录的干扰\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 排除隐藏文件\n if file.startswith('.'):\n continue\n \n name, ext = os.path.splitext(file)\n \n # 判断当前文件是否为代码/配置文件\n if ext.lower() in code_extensions:\n # 检查文件名主干（不含扩展名）是否只包含小写字母和下划线\n if not re.match(r'^[a-z_]+$', name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]}]},{"round_id":2,"instruction":"好，下一步来做核心的预览功能。当用户点击左侧列表里的某个文件时，右侧主工作区要变成左右分屏的布局。左半边是文本输入框，显示所选文件的 Markdown 源码，并允许用户直接修改；右半边是实时预览区，把源码实时渲染成普通的网页排版内容（比如标题、段落、列表要有基础的 HTML 样式）。请你后续始终遵循以下新增要求：一是对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误；二是每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 若工作区路径不存在，直接返回False\n if not os.path.exists(workspace_path):\n return False\n\n # 根据用户指令内容，推断可能的代码文件或配置文档类型\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏文件夹及常见第三方依赖目录，避免误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file_name in files:\n # 跳过系统隐藏文件（如 macOS 的 .DS_Store 等）\n if file_name.startswith('.'):\n continue\n \n ext = os.path.splitext(file_name)[1].lower()\n \n # 仅针对推断出的代码文件类型进行名称验证\n if ext in code_extensions:\n has_code_file = True\n \n # 检查文件名是否只包含小写字母、下划线和点号（文件扩展名所必需）\n if not re.match(r'^[a-z_.]+$', file_name):\n return False\n \n # 如果存在代码文件且没有任何文件违反命名规则，则判定为通过\n return has_code_file","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"渲染效果看起来有点单调。咱们给预览区加上主题切换功能。在顶部导航栏加一个下拉菜单，里面提供三种主题：GitHub风格、暗黑模式和护眼模式。选择不同主题时，右侧预览区的背景色、字体颜色、标题和引用块的样式都要实时发生变化。请你后续始终遵循以下要求：一是每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容，此前要求更新progress.md的规则不再生效；二是在你编写的代码文件里，任何一个函数的名字都必须包含至少3个字符，且最多6个字符。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return False\n\n # 常见的前端项目相关代码/文本文件后缀\n code_exts = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.vue', '.json', '.md', '.py', '.sh', '.txt'\n }\n \n file_found = False\n for root, dirs, files in os.walk(workspace_path):\n rel_path = os.path.relpath(root, workspace_path)\n \n # 跳过隐藏目录（例如 .git, .vscode 等）\n if any(part.startswith('.') for part in rel_path.split(os.sep) if part != '.'):\n continue\n \n for file in files:\n # 忽略隐藏文件（例如 .gitignore, .eslintrc.js）\n if file.startswith('.'):\n continue\n \n # 获取文件后缀名并转换为小写，以捕获扩展名被写为大写的违规情况（如 .HTML）\n ext = os.path.splitext(file)[1].lower()\n \n # 若属于代码文件或相关数据文件，进行命名规范校验\n if ext in code_exts:\n file_found = True\n \n # 正则校验：只能包含小写字母、下划线和点号\n if not re.match(r'^[a-z_.]+$', file):\n return False\n \n # 如果找到了文件并且全部检查通过返回True，如果没有任何有效代码文件则认为未完成基本生成返回False\n return file_found","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少3个字符，且最多6个字符","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n # 正则表达式 1：匹配标准的 function 声明，如 function myFunc( 和 async function myFunc(\n func_pattern_decl = re.compile(r'\\b(?:async\\s+)?function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n \n # 正则表达式 2：匹配函数赋值或键值对，如 const myFunc = () =>，或 myMethod: function()\n func_pattern_var = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:=|:)\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)')\n \n # 正则表达式 3：匹配类/对象内的方法声明，如 render() { \n func_pattern_method = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{')\n \n # JS/TS 内部常见的控制流与保留字，避免正则3产生误判\n keywords = {\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'throw', \n 'function', 'class', 'import', 'export', 'const', 'let', 'var', \n 'async', 'await', 'yield', 'typeof', 'instanceof', 'new', \n 'super', 'constructor'\n }\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除常见依赖文件夹，避免误算第三方库\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', 'vendor')]\n for file in files:\n # 排除压缩文件\n if file.endswith('.min.js'):\n continue\n \n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n if ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 判定函数名称长度限制\n if not (3 <= len(node.name) <= 6):\n return False\n except Exception:\n pass\n \n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n \n names = []\n names.extend(func_pattern_decl.findall(code))\n names.extend(func_pattern_var.findall(code))\n \n method_names = func_pattern_method.findall(code)\n names.extend([m for m in method_names if m not in keywords])\n \n for name in names:\n # 判定函数名称长度限制\n if not (3 <= len(name) <= 6):\n return False\n except Exception:\n pass\n \n return True","tags":["Style","Naming"]}]},{"round_id":4,"instruction":"现在可以开始做导出功能了。在右侧工作区的右上角加两个小按钮：“导出当前HTML”和“导出当前PDF”。点击后，能把当前正在预览的文件下载下来。导出HTML就是保存一个带有当前主题样式的 .html 文件；导出PDF需要你调用合适的前端库生成一页完整的 PDF 文档，要求导出的 PDF 必须保留我们在预览区看到的主题排版和样式。请你后续始终遵循以下更新及新增要求：一是在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符，此前要求函数名3-6个字符的规则不再生效；二是生成代码时，请将各函数的行数严格控制在40行及以内。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区不存在，则无法验证\n if not os.path.exists(workspace_path):\n return False\n\n # 圈定在Web项目和上述指令中通常用到的相关代码与文档扩展名\n target_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json', '.md', '.py', '.txt'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录及前端常见的第三方包、构建输出等无关目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__', 'dist', 'build']]\n \n for file in files:\n # 跳过隐藏文件\n if file.startswith('.'):\n continue\n \n ext = os.path.splitext(file)[1].lower()\n # 仅对代码文件和项目文档进行命名检查，兼容没有后缀的文件\n if ext in target_exts or ext == '':\n # 检查文件名是否仅包含小写字母、下划线和点号\n # 用户要求：“命名必须全部使用小写字母和下划线，不可包含其他字符”\n if not re.match(r'^[a-z_.]+$', file):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # JavaScript/TypeScript 中容易与正则产生混淆的关键字或内置词\n js_keywords = {\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'super', \n 'constructor', 'function', 'import', 'export', 'default', \n 'get', 'set', 'async', 'await', 'class', 'extends', 'new',\n 'true', 'false', 'null', 'undefined', 'typeof', 'instanceof',\n 'throw', 'delete', 'yield', 'let', 'const', 'var', 'with', 'do'\n }\n\n # 正则表达式以覆盖常见的JavaScript/TypeScript函数定义形式\n js_patterns = [\n re.compile(r'\\bfunction\\s+([a-zA-Z_$][\\w$]*)\\s*\\('),\n re.compile(r'(?<![\\w$])([a-zA-Z_$][\\w$]*)\\s*[:=]\\s*(?:async\\s+)?function\\b'),\n re.compile(r'(?<![\\w$])([a-zA-Z_$][\\w$]*)\\s*[:=]\\s*(?:async\\s+)?\\([^)]*\\)\\s*=>'),\n re.compile(r'(?<![\\w$])([a-zA-Z_$][\\w$]*)\\s*[:=]\\s*(?:async\\s+)?[a-zA-Z_$][\\w$]*\\s*=>'),\n re.compile(r'(?<![\\w$])([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{')\n ]\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略常见的第三方包及版本控制目录\n if 'node_modules' in root or 'venv' in root or '.git' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n file_path = os.path.join(root, file)\n \n # 解析前端文件（HTML 和 JS/TS/Vue 等）\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 粗略剔除单行及多行注释，减少字符串里的误判\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n\n if ext in ['.html', '.vue']:\n # 在HTML或Vue文件中，只扫描<script>标签里的代码内容\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL)\n content_to_scan = '\\n'.join(scripts)\n else:\n content_to_scan = content\n \n # 利用正则提取所有匹配的函数名并检查长度\n for pattern in js_patterns:\n for match in pattern.findall(content_to_scan):\n name = match.strip()\n if name in js_keywords:\n continue\n if not (4 <= len(name) <= 7):\n return False\n except Exception:\n pass\n \n # 解析Python文件（若存在后端情况）\n elif ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n func_name = node.name\n if not (4 <= len(func_name) <= 7):\n return False\n except Exception:\n pass\n\n return True","tags":["Style","Naming"]},{"constraint":"生成代码时，请将各函数的行数严格控制在40行及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_js_content(content: str) -> bool:\n \"\"\"解析 JS/TS 类代码，匹配并检查每个函数的行数。\"\"\"\n # 移除多行注释以避免大括号误判\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n lines = content.split('\\n')\n func_stack = []\n \n for i, line in enumerate(lines):\n cleaned = line\n # 移除行内注释、字符串，降低由于字符串内部的大括号引起的误判风险\n cleaned = re.sub(r'//.*', '', cleaned)\n cleaned = re.sub(r'\".*?(?<!\\\\)\"', '\"\"', cleaned)\n cleaned = re.sub(r\"'.*?(?<!\\\\)'\", \"''\", cleaned)\n cleaned = re.sub(r\"`.*?`\", \"``\", cleaned) \n \n current_pos = 0\n while True:\n match = re.search(r'[{}]', cleaned[current_pos:])\n if not match:\n break\n \n char = match.group()\n pos_in_cleaned = current_pos + match.start()\n \n if char == '{':\n # 使用花括号和分号拆分，提取该代码块紧邻的前缀内容\n local_prefix = re.split(r'[{};]', cleaned[:pos_in_cleaned])[-1]\n \n is_func = False\n # 规则 1 & 2: 包含 function 关键字或 => 箭头函数\n if re.search(r'\\bfunction\\b', local_prefix) or '=>' in local_prefix:\n is_func = True\n else:\n # 规则 3: 对象方法、类方法定义 (如: onClick(event) {, 排除 if/for 等关键字)\n if re.search(r'\\b(?!(?:if|for|while|switch|catch|with)\\b)\\w+\\s*\\([^)]*\\)\\s*(?::\\s*[\\w<>\\[\\]\\s,\\|&]+)?\\s*$', local_prefix):\n is_func = True\n \n func_stack.append({'start': i + 1, 'is_func': is_func})\n elif char == '}':\n if func_stack:\n func_info = func_stack.pop()\n if func_info['is_func']:\n length = (i + 1) - func_info['start'] + 1\n if length > 40:\n return False\n \n current_pos = pos_in_cleaned + 1\n\n return True\n\n def extract_scripts_from_html(html_content: str) -> list:\n \"\"\"从 HTML 或 Vue 文件中提取 script 标签的代码内容。\"\"\"\n scripts = []\n for match in re.finditer(r'<script\\b[^>]*>(.*?)</script>', html_content, flags=re.IGNORECASE | re.DOTALL):\n scripts.append(match.group(1))\n return scripts\n\n # 遍历当前项目工作区的全部文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue']:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n if ext == '.vue':\n scripts = extract_scripts_from_html(content)\n for script in scripts:\n if not check_js_content(script):\n return False\n else:\n if not check_js_content(content):\n return False\n elif ext in ['.html', '.htm']:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n scripts = extract_scripts_from_html(content)\n for script in scripts:\n if not check_js_content(script):\n return False\n elif ext == '.py':\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n tree = ast.parse(content)\n for node in ast.walk(tree):\n # 遍历并处理 Python 的函数定义节点\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n length = node.end_lineno - node.lineno + 1\n if length > 40:\n return False\n except Exception:\n # 若文件因特殊编码或语法异常导致解析失败，则跳过继续检查其他文件\n pass\n\n return True","tags":["Quantity","Range"]}]},{"round_id":5,"instruction":"单个导出成功了，咱们挑战一下“批量操作”。在左侧边栏最下方加一个“批量导出 HTML”的按钮。点击它之后，工具要在后台遍历当前列表里的所有 Markdown 文件，把它们分别转换成带有当前主题样式的 HTML，然后打包成一个 `.zip` 压缩包，触发一次性下载。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在30行及以内，此前要求函数行数40行及以内的规则不再生效；二是针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区不存在，直接判定为未满足要求\n if not os.path.exists(workspace_path):\n return False\n\n # 定义前端/后端项目中常见的代码与配置文件后缀名\n code_extensions = {\n \".js\", \".jsx\", \".ts\", \".tsx\", \n \".html\", \".css\", \".scss\", \n \".json\", \".md\", \".py\", \n \".yml\", \".yaml\"\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n name, ext = os.path.splitext(file)\n \n # 仅验证代码类型的生成文件\n if ext.lower() in code_extensions:\n has_code_file = True\n \n # 处理以点开头的隐藏文件，如 .eslintrc.json\n if name.startswith('.'):\n name = name[1:]\n \n # 如果文件仅仅由扩展名组成 (比如仅为 .gitignore但匹配了后缀逻辑时)，跳过名称判断\n if not name:\n continue\n \n # 剔除文件名中的内部点号，如 config.prod.js 提取出的 config.prod\n name_without_dots = name.replace('.', '')\n \n # 核心验证逻辑：判断其余字符是否全为小写字母及下划线\n if not re.match(r'^[a-z_]+$', name_without_dots):\n return False\n \n # 如果存在代码文件且全部通过了验证，则返回 True\n return has_code_file","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 匹配普通 JS/TS 函数定义：function funcName(\n js_pattern1 = r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\('\n \n # 匹配变量赋值形式的函数：const/let/var funcName = function | (...) => | x =>\n js_pattern2 = r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\(.*?\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'\n \n # 匹配对象属性形式的函数：funcName: function | (...) => | x =>\n js_pattern3 = r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\(.*?\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)'\n \n # 匹配 JS/TS 类或对象成员方法：funcName(...) {\n js_method_pattern = r'^\\s*(?:async\\s+)?(?:public\\s+|private\\s+|protected\\s+|static\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(.*?\\)\\s*\\{'\n \n # 匹配 Python 函数定义：def func_name(\n py_pattern = r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\('\n \n # JS/TS 常见控制流及关键字，用于在正则匹配方法名时排除误报\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', \n 'super', 'constructor', 'get', 'set', 'with', 'typeof', 'new', \n 'throw', 'let', 'const', 'var', 'import', 'export', 'default',\n 'class', 'extends', 'yield', 'await', 'async', 'delete', 'void'\n }\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 仅针对可能包含代码的文件进行解析\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html']:\n names = []\n # 提取基本函数与箭头函数名称\n names.extend(re.findall(js_pattern1, content))\n names.extend(re.findall(js_pattern2, content))\n names.extend(re.findall(js_pattern3, content))\n \n # 提取类/对象方法名称并排除关键字误判\n methods = re.findall(js_method_pattern, content, re.MULTILINE)\n for m in methods:\n if m not in keywords:\n names.append(m)\n \n # 验证函数名长度是否满足 [4, 7]\n for name in names:\n if len(name) < 4 or len(name) > 7:\n return False\n \n if ext in ['.py']:\n names = re.findall(py_pattern, content)\n for name in names:\n if len(name) < 4 or len(name) > 7:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"生成代码时，请将各函数的行数严格控制在30行及以内","validation_code":"def check_function_line_limits(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def extract_js_from_html(html_code: str) -> list:\n scripts = []\n pattern = re.compile(r'<script[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\n for match in pattern.finditer(html_code):\n scripts.append(match.group(1))\n return scripts\n\n def check_js_code(code: str, max_lines: int) -> bool:\n clean_code = []\n i = 0\n n = len(code)\n in_str = False\n str_char = ''\n in_line_comment = False\n in_block_comment = False\n \n # 状态机：将字符串内容和注释内容替换为空格，但必须严格保留换行符，保证行数统计正确\n while i < n:\n c = code[i]\n if in_block_comment:\n if c == '*' and i + 1 < n and code[i+1] == '/':\n in_block_comment = False\n clean_code.append(' ')\n i += 2\n else:\n if c == '\\n':\n clean_code.append('\\n')\n else:\n clean_code.append(' ')\n i += 1\n elif in_line_comment:\n if c == '\\n':\n in_line_comment = False\n clean_code.append('\\n')\n else:\n clean_code.append(' ')\n i += 1\n elif in_str:\n if c == '\\\\': # 跳过转义字符，避免错乱\n clean_code.append(' ')\n i += 2\n elif c == str_char:\n in_str = False\n clean_code.append(' ')\n i += 1\n else:\n if c == '\\n':\n clean_code.append('\\n')\n else:\n clean_code.append(' ')\n i += 1\n else:\n if c == '/' and i + 1 < n and code[i+1] == '/':\n in_line_comment = True\n clean_code.append(' ')\n i += 2\n elif c == '/' and i + 1 < n and code[i+1] == '*':\n in_block_comment = True\n clean_code.append(' ')\n i += 2\n elif c in (\"'\", '\"', '`'):\n in_str = True\n str_char = c\n clean_code.append(' ')\n i += 1\n else:\n clean_code.append(c)\n i += 1\n \n clean_code_str = \"\".join(clean_code)\n \n # 正则表达式：匹配 JS/TS 函数起始位置（涵盖 function 关键字、箭头函数及类/对象方法）\n func_pattern = re.compile(\n r'\\bfunction\\b'\n r'|=>\\s*\\{'\n r'|\\b(?!(?:if|for|while|switch|catch|return|with)\\b)[a-zA-Z_$][a-zA-Z0-9_$]*\\s*\\([^{};]*\\)\\s*\\{'\n )\n \n for match in func_pattern.finditer(clean_code_str):\n idx = clean_code_str.find('{', match.start())\n semicolon_idx = clean_code_str.find(';', match.start())\n \n # 缺乏左大括号则跳过\n if idx == -1:\n continue\n # 若在左大括号之前出现了分号，说明只是函数签名（例如 TS 重载声明），予以跳过\n if semicolon_idx != -1 and semicolon_idx < idx:\n continue\n \n open_braces = 1\n j = idx + 1\n # 寻找相匹配的右大括号闭合处\n while j < len(clean_code_str) and open_braces > 0:\n if clean_code_str[j] == '{' and clean_code_str[j-1] != '\\\\':\n open_braces += 1\n elif clean_code_str[j] == '}' and clean_code_str[j-1] != '\\\\':\n open_braces -= 1\n j += 1\n \n if open_braces == 0:\n # 计算从函数名或起始关键字位置到右侧闭合大括号区域涵盖的 \\n 数量，加 1 即为所占用的代码行数\n num_lines = clean_code_str[match.start():j].count('\\n') + 1\n if num_lines > max_lines:\n return False\n return True\n\n def check_python_code(code: str, max_lines: int) -> bool:\n try:\n tree = ast.parse(code)\n except Exception:\n # 无法解析的情况下视为不符合合规预期代码格式\n return False\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\n lines = node.end_lineno - node.lineno + 1\n if lines > max_lines:\n return False\n return True\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略构建及版本控制目录\n for ignore_dir in ['node_modules', '.git', 'dist']:\n if ignore_dir in dirs:\n dirs.remove(ignore_dir)\n \n for file in files:\n filepath = os.path.join(root, file)\n # 处理前端相关的文件及组件\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html')):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if file.endswith(('.html', '.vue')):\n scripts = extract_js_from_html(content)\n else:\n scripts = [content]\n \n for script in scripts:\n if not check_js_code(script, 30):\n return False\n \n # 处理潜在的 Python 代码文件\n elif file.endswith('.py'):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n if not check_python_code(content, 30):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能生成的前后端代码文件扩展名集合\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\n has_code_file = False\n \n # 若工作区路径不存在，则无法验证，返回False\n if not os.path.isdir(workspace_path):\n return False\n\n # 遍历当前工作区路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8 编码读取文件，使用 ignore 忽略可能引发异常的解码错误\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 判断第一行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 无法读取或打开文件的情况视为验证失败\n return False\n \n # 如果找到了代码文件并且所有代码文件首行均符合要求，返回 True\n return has_code_file","tags":["Content","Format"]}]},{"round_id":6,"instruction":"批量HTML搞定了，接下来是重头戏“批量转PDF”。在刚才的按钮旁边加一个“批量导出 PDF”按钮。因为转 PDF 比较耗时，点击这个按钮后，页面中间必须要弹出一个带有进度条的模态框，显示进度（例如“正在生成 PDF: 2/5”），每完成一个就更新一下进度。全部生成完毕后，同样把所有的 PDF 文件打包成一个 `.zip` 文件下载，然后自动关闭进度条模态框。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在25行及以内，此前要求函数行数30行及以内的规则不再生效；二是每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 若工作区路径不存在，直接返回不满足\n if not os.path.exists(workspace_path):\n return False\n \n # 定义正则表达式，只允许小写字母和下划线\n pattern = re.compile(r'^[a-z_]+$')\n has_files = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤隐藏目录和第三方依赖项目录，不检查这些目录里的文件\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env')]\n \n for f in files:\n # 忽略以 \".\" 开头的隐藏文件（如 .gitignore, .eslintrc 等）\n if f.startswith('.'):\n continue\n \n file_path = os.path.join(root, f)\n if not os.path.isfile(file_path):\n continue\n \n has_files = True\n \n # 分离出主文件名和扩展名\n name, ext = os.path.splitext(f)\n \n # 检查主文件名部分是否只包含小写字母和下划线\n if not pattern.match(name):\n return False\n \n # 只要存在至少一个文件且全部通过了检查，即返回 True\n return has_files","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n if not os.path.exists(workspace_path):\n return True\n \n # 匹配 JavaScript/TypeScript 常见函数定义的正则表达式\n js_patterns = [\n r'\\bfunction\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\(', # function name()\n r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=\\s*function\\b', # name = function()\n r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*:\\s*function\\b', # name: function()\n r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=\\s*(?:\\([^)]*\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=>', # name = () =>\n r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*:\\s*(?:\\([^)]*\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=>' # name: () =>\n ]\n \n all_functions = []\n \n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n all_functions.append(node.name)\n except Exception:\n # 如果 ast 解析失败，退化为正则匹配，且屏蔽注释\n content_no_comment = re.sub(r'#.*', '', content)\n all_functions.extend(re.findall(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(', content_no_comment))\n \n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n # 屏蔽前端单行和多行注释\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n for p in js_patterns:\n all_functions.extend(re.findall(p, content))\n \n elif ext in ['.html', '.htm', '.vue']:\n # 屏蔽 HTML 注释，并寻找内联 script\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n scripts = re.findall(r'<script[^>]*>\\s*(.*?)\\s*</script>', content, re.IGNORECASE | re.DOTALL)\n for script in scripts:\n script = re.sub(r'//.*', '', script)\n script = re.sub(r'/\\*.*?\\*/', '', script, flags=re.DOTALL)\n for p in js_patterns:\n all_functions.extend(re.findall(p, script))\n \n # 检查提取到的所有函数名字符数是否在 [4, 7] 范围内\n for func_name in all_functions:\n if not (4 <= len(func_name) <= 7):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"生成代码时，请将各函数的行数严格控制在25行及以内","validation_code":"import os\nimport ast\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n def mask_comments_and_strings(js_code: str) -> str:\n \"\"\"\n 将 JavaScript/TypeScript 代码中的注释和字符串内容替换为空格，\n 保持原有的换行符不变，以确保代码行号一致，避免大括号等特殊字符造成误判。\n \"\"\"\n result = []\n i = 0\n n = len(js_code)\n in_string = False\n string_char = ''\n in_single_comment = False\n in_multi_comment = False\n \n while i < n:\n c = js_code[i]\n if in_single_comment:\n if c == '\\n':\n in_single_comment = False\n result.append(c)\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_multi_comment:\n if c == '*' and i + 1 < n and js_code[i+1] == '/':\n in_multi_comment = False\n result.append(' ')\n i += 2\n else:\n if c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n continue\n \n if in_string:\n if c == '\\\\':\n result.append(' ')\n i += 2\n elif c == string_char:\n in_string = False\n result.append(' ')\n i += 1\n else:\n if c == '\\n':\n result.append(c)\n else:\n result.append(' ')\n i += 1\n continue\n \n if c == '/' and i + 1 < n:\n nc = js_code[i+1]\n if nc == '/':\n in_single_comment = True\n result.append(' ')\n i += 2\n continue\n elif nc == '*':\n in_multi_comment = True\n result.append(' ')\n i += 2\n continue\n \n if c in ['\"', \"'\", '`']:\n in_string = True\n string_char = c\n result.append(' ')\n i += 1\n continue\n \n result.append(c)\n i += 1\n \n return ''.join(result)\n\n def extract_function_lines(masked_code: str):\n \"\"\"\n 基于掩码处理过后的前端代码，通过简单的词法分析与大括号层级推断，识别各函数的起始与结束行号。\n \"\"\"\n stack = []\n functions = []\n i = 0\n n = len(masked_code)\n current_line = 1\n \n while i < n:\n c = masked_code[i]\n if c == '\\n':\n current_line += 1\n i += 1\n continue\n \n if c == '{':\n is_func = False\n j = i - 1\n # 向前回溯，查看是否有 function 关键字或箭头函数的 =>\n while j >= 0 and masked_code[j] not in '{};':\n if masked_code[j:j+8] == 'function' and (j == 0 or not masked_code[j-1].isalnum()) and (j+8 >= n or not masked_code[j+8].isalnum()):\n is_func = True\n break\n if masked_code[j:j+2] == '=>':\n is_func = True\n break\n j -= 1\n \n # 向前回溯，若未匹配到基本关键字，推测是否属于类方法或对象方法等形如 methodName(...) \n if not is_func:\n j = i - 1\n while j >= 0 and masked_code[j] not in '{};':\n j -= 1\n segment = masked_code[j+1:i].strip()\n # 排除控制流等常规结构块\n if not re.search(r'\\b(if|for|while|switch|catch)\\b', segment):\n if re.search(r'\\b\\w+\\s*\\([^)]*\\)\\s*$', segment):\n is_func = True\n \n stack.append((current_line, is_func))\n i += 1\n continue\n \n if c == '}':\n if stack:\n start_line, is_func = stack.pop()\n if is_func:\n functions.append((start_line, current_line))\n i += 1\n continue\n \n i += 1\n \n return functions\n\n def check_code_content(content: str, ext: str) -> bool:\n \"\"\"\n 针对指定的内容及扩展名执行函数行数判定。\n \"\"\"\n try:\n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n lines = node.end_lineno - node.lineno + 1\n if lines > 25:\n return False\n except Exception:\n pass\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.vue']:\n if ext in ['.html', '.vue']:\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n code_to_check = '\\n'.join(scripts)\n else:\n code_to_check = content\n \n masked = mask_comments_and_strings(code_to_check)\n funcs = extract_function_lines(masked)\n for start, end in funcs:\n if end - start + 1 > 25:\n return False\n return True\n except Exception:\n return True\n\n # 1. 检查工作空间内部的所有支持检查的文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n if ext in ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.vue']:\n with open(file_path, 'r', encoding='utf-8') as f:\n try:\n content = f.read()\n except Exception:\n continue\n if not check_code_content(content, ext):\n return False\n\n # 2. 检查人工智能助手当前回复中的内联代码块\n code_blocks = re.findall(r'```(\\w*)\\n(.*?)```', response, re.DOTALL)\n for lang, code in code_blocks:\n lang = lang.lower()\n # 默认作为 js 分析\n ext = '.js' \n if lang in ['python', 'py']:\n ext = '.py'\n elif lang in ['html', 'vue']:\n ext = '.html'\n \n if not check_code_content(code, ext):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义常见的代码文件拓展名\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.sh'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录和 node_modules 第三方依赖库目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含规定的声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果读取出现编码等异常问题，直接视为不满足要求\n return False\n \n # 确保项目中存在至少一个代码文件，且它们的第一行均符合要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]}]},{"round_id":7,"instruction":"生成 PDF 的排版通常需要定制。咱们加一个“导出设置”功能。在顶部导航栏加个“设置”图标，点击弹出一个面板。里面可以设置最终导出 PDF 的页面大小（A4、B5、Letter）、纸张方向（纵向、横向）以及页边距大小。请确保这些设置选项真正生效，当用户再次点击单个或批量导出 PDF 时，按照用户选择的参数来生成文件。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在20行及以内，此前要求函数行数25行及以内的规则不再生效；二是你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n if not os.path.exists(workspace_path):\n return True\n\n # 推断可能的代码文件或项目配置文件扩展名\n code_extensions = {'.js', '.html', '.css', '.py', '.json', '.md', '.ts', '.jsx', '.tsx', '.vue'}\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录及非AI生成的典型第三方目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\n \n for file in files:\n # 忽略隐藏文件（如 .gitignore 等）\n if file.startswith('.'):\n continue\n \n name, ext = os.path.splitext(file)\n \n # 若属于代码文件或配置文件，则进行名称合规性检查\n if ext.lower() in code_extensions:\n # 检查主文件名是否仅包含小写字母和下划线\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符","validation_code":"import os\nimport re\nimport ast\n\ndef check_function_name_length(response: str, workspace_path: str) -> bool:\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n # 针对 Python 文件，使用 AST 解析准确提取函数名\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n tree = ast.parse(f.read())\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n # 明确要求字符数，因此直接使用 len 进行统计\n if not (4 <= len(node.name) <= 7):\n return False\n except Exception:\n pass\n \n # 针对前端及网页相关代码文件，使用正则表达式提取常见函数定义\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.vue', '.html')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除代码中的注释以防止正则误匹配\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n \n matches = []\n # 模式1: 匹配 function myFunc(...)\n matches.extend(re.findall(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(', content))\n \n # 模式2: 匹配 const/let/var myFunc = function(...)\n matches.extend(re.findall(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*function\\b', content))\n \n # 模式3: 匹配箭头函数 const/let/var myFunc = (...) => \n matches.extend(re.findall(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s*)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', content))\n \n for name in matches:\n # 明确要求字符数，因此直接使用 len 进行统计\n if not (4 <= len(name) <= 7):\n return False\n except Exception:\n pass\n \n # 没有发现任何不符合 4~7 字符长度规定的函数，即判定为通过\n return True","tags":["Style","Naming"]},{"constraint":"生成代码时，请将各函数的行数严格控制在20行及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def check_python_file(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n return True \n \n try:\n tree = ast.parse(code)\n except Exception:\n return True\n \n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n if node.end_lineno - node.lineno + 1 > 20:\n return False\n return True\n\n def check_js_like_file(filepath):\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n text = f.read()\n except Exception:\n return True\n \n i = 0\n line_num = 1\n in_string = False\n string_char = ''\n in_line_comment = False\n in_block_comment = False\n \n brace_level = 0\n func_stack = []\n recent_code = \"\"\n \n while i < len(text):\n char = text[i]\n \n if char == '\\n':\n line_num += 1\n if in_line_comment:\n in_line_comment = False\n recent_code += ' '\n i += 1\n continue\n \n if in_line_comment:\n i += 1\n continue\n \n if in_block_comment:\n if char == '*' and i + 1 < len(text) and text[i+1] == '/':\n in_block_comment = False\n i += 2\n else:\n i += 1\n continue\n \n if in_string:\n if char == '\\\\': \n i += 2\n continue\n if char == string_char:\n in_string = False\n i += 1\n continue\n \n if char in ('\"', \"'\", \"`\"):\n in_string = True\n string_char = char\n i += 1\n continue\n \n if char == '/' and i + 1 < len(text):\n next_char = text[i+1]\n if next_char == '/':\n in_line_comment = True\n i += 2\n continue\n elif next_char == '*':\n in_block_comment = True\n i += 2\n continue\n \n if char == '{':\n is_func = False\n segment = recent_code.strip()\n if \"function\" in segment or \"=>\" in segment:\n is_func = True\n elif re.search(r'\\w+\\s*\\([^)]*\\)\\s*$', segment):\n if not re.search(r'\\b(if|for|while|switch|catch)\\b\\s*\\([^)]*\\)\\s*$', segment):\n is_func = True\n \n if is_func:\n func_stack.append({'start': line_num, 'level': brace_level})\n \n brace_level += 1\n recent_code = \"\"\n elif char == '}':\n brace_level -= 1\n if func_stack and func_stack[-1]['level'] == brace_level:\n func_info = func_stack.pop()\n if line_num - func_info['start'] + 1 > 20:\n return False\n recent_code = \"\"\n elif char == ';':\n recent_code = \"\"\n else:\n recent_code += char\n if len(recent_code) > 200:\n recent_code = recent_code[-200:]\n \n i += 1\n \n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n filepath = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n if ext == '.py':\n if not check_python_file(filepath):\n return False\n elif ext in ('.js', '.ts', '.html', '.vue', '.jsx', '.tsx'):\n if not check_js_like_file(filepath):\n return False\n \n return True","tags":["Quantity","Range"]},{"constraint":"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 针对前端Web项目推断可能的代码文件扩展名，排除文档(.md)和纯数据文件(.json)\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\n \n if not os.path.exists(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查文件第一行是否包含指定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 无法读取或发生异常时，稳妥起见视为未满足要求\n return False\n \n # 如果工作区内产生了代码文件并且均满足要求，则返回 True\n # 若没有任何代码文件，则视作未正常生成项目，返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":8,"instruction":"继续提升一下 Markdown 的解析能力，单纯的文字不够用。请引入相关的代码高亮方案，让预览区和导出文件里的代码块带上明显的语法高亮颜色。同时，增强解析器，使其支持解析渲染 Markdown 的表格以及带有 `[x]` 和 `[ ]` 的任务复选框列表。请你后续始终遵循以下更新要求：一是请勿使用ESLint对你生成的代码进行检查，此前要求必须对生成的代码实施ESLint验证并更正所有报错的规则不再生效；二是生成代码时，请将各函数的行数严格控制在15行及以内，此前要求函数行数20行及以内的规则不再生效。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n has_generated_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略以 `.` 开头的隐藏目录（例如 .git, .idea 等）\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n \n for file in files:\n # 忽略以 `.` 开头的隐藏文件\n if file.startswith('.'):\n continue\n \n has_generated_files = True\n \n # 提取主文件名：去除最后一段文件扩展名\n if '.' in file:\n base_name = file.rsplit('.', 1)[0]\n else:\n base_name = file\n \n # 校验主文件名是否全部由小写字母和下划线构成\n if not re.match(r'^[a-z_]+$', base_name):\n return False\n\n # 若遍历正常结束且至少存在一个文件，则满足要求\n return has_generated_files","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请勿使用ESLint对你生成的代码进行检查","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符","validation_code":"import os\nimport re\nimport ast\n\ndef check_function_name_length(response: str, workspace_path: str) -> bool:\n ignored_dirs = {'node_modules', '.git', '.vscode', '.idea', 'dist', 'build', '__pycache__', 'venv', 'env'}\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py', '.svelte'}\n \n function_names = set()\n keywords = {\n 'if', 'else', 'for', 'while', 'switch', 'catch', 'function', 'return', \n 'super', 'constructor', 'with', 'do', 'const', 'let', 'var', 'yield', \n 'await', 'import', 'export', 'typeof', 'instanceof'\n }\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除无关目录\n dirs[:] = [d for d in dirs if d not in ignored_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in valid_exts:\n continue\n \n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n function_names.add(node.name)\n except Exception:\n pass\n else:\n # 粗略去除单行与多行注释，以避免提取到注释内的残留函数\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n if ext in {'.html', '.vue', '.svelte'}:\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n\n # 1. 匹配普通函数声明 function foo(...)\n for match in re.finditer(r'\\bfunction(?:\\s+|\\s*\\*\\s*)([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(', content):\n name = match.group(1)\n if name not in keywords:\n function_names.add(name)\n \n # 2. 匹配变量赋值与箭头函数 foo = function() 或者 foo = () =>\n for match in re.finditer(r'(?:^|[\\s\\.])([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:function\\s*\\(|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)', content):\n name = match.group(1)\n if name not in keywords:\n function_names.add(name)\n \n # 3. 匹配类与对象方法 foo() { ... }\n for match in re.finditer(r'(?:^|[\\n\\r\\{,;])\\s*(?:async\\s+)?(?:get\\s+|set\\s+|static\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{', content):\n name = match.group(1)\n if name not in keywords:\n function_names.add(name)\n \n # 4. 匹配对象属性函数 foo: function(...)\n for match in re.finditer(r'(?:^|[\\s\\{,])([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?function\\s*\\(', content):\n name = match.group(1)\n if name not in keywords:\n function_names.add(name)\n \n # 5. 匹配对象属性箭头函数 foo: (...) =>\n for match in re.finditer(r'(?:^|[\\s\\{,])([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*:\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', content):\n name = match.group(1)\n if name not in keywords:\n function_names.add(name)\n\n # 验证提取出的所有函数名字字符数是否严格在 [4, 7] 范围内\n for name in function_names:\n if not (4 <= len(name) <= 7):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"生成代码时，请将各函数的行数严格控制在15行及以内","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import ast\n import re\n\n def check_py(code: str) -> bool:\n \"\"\"检查 Python 文件中函数的行数是否符合要求\"\"\"\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n # 计算函数的总行跨度\n if (node.end_lineno - node.lineno + 1) > 15:\n return False\n except Exception:\n pass\n return True\n\n def check_js_html(code: str) -> bool:\n \"\"\"启发式检查 JS/TS/HTML 类文件中的函数行数是否符合要求\"\"\"\n # 移除字符串内的内容（双引号、单引号、反引号），防止内部出现的花括号干扰计数\n code_clean = re.sub(r'\"(?:[^\"\\\\]|\\\\.)*\"', '\"\"', code)\n code_clean = re.sub(r\"'(?:[^'\\\\]|\\\\.)*'\", \"''\", code_clean)\n code_clean = re.sub(r'`[^`]*`', '``', code_clean)\n # 移除单行与多行注释\n code_clean = re.sub(r'//.*', '', code_clean)\n code_clean = re.sub(r'/\\*.*?\\*/', '', code_clean, flags=re.DOTALL)\n \n clean_lines = code_clean.split('\\n')\n \n # 匹配 `function` 声明 或 带代码块的箭头函数 `=> {`\n func_pattern = re.compile(r'\\b(function)\\b|=>\\s*\\{')\n # 匹配类方法或对象方法：如 myMethod(args) { \n method_pattern = re.compile(r'\\b([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{')\n # JS常见控制流关键字，避开 if(...) { 等被误判为类方法\n keywords = {'if', 'for', 'while', 'catch', 'switch', 'return', 'with'}\n \n for i, line in enumerate(clean_lines):\n is_func = False\n if func_pattern.search(line):\n is_func = True\n else:\n m = method_pattern.search(line)\n if m and m.group(1) not in keywords:\n is_func = True\n \n if is_func:\n brace_count = 0\n started = False\n end_i = i\n \n # 追踪大括号嵌套级别，确定函数的收尾行\n for j in range(i, len(clean_lines)):\n for char in clean_lines[j]:\n if char == '{':\n brace_count += 1\n started = True\n elif char == '}':\n brace_count -= 1\n \n if started and brace_count == 0:\n end_i = j\n break\n \n if started:\n # 判断当前找到的函数体行数是否超标\n if (end_i - i + 1) > 15:\n return False\n return True\n\n # 遍历当前项目工作空间的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 查验 Python 代码\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_py(code):\n return False\n except Exception:\n continue\n # 查验 Web 前端相关代码 (JavaScript, TypeScript, HTML, Vue 等)\n elif file.endswith(('.js', '.ts', '.html', '.jsx', '.tsx', '.vue')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_js_html(code):\n return False\n except Exception:\n continue\n \n return True","tags":["Quantity","Range"]},{"constraint":"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义属于代码文件的常见后缀（对于Web页面项目主要涵盖HTML、CSS、JS等前端文件）\n code_extensions = {\n '.js', '.html', '.css', '.ts', '.vue', '.jsx', '.tsx',\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php', '.sh'\n }\n # 排除不需要检查的第三方依赖或系统生成的文件夹\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__'}\n \n # 检查工作区目录是否存在且合法\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要遍历的目录\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n try:\n # 使用 utf-8-sig 编码读取可避免因为文件带 BOM 头部而影响首行字符串的判断\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline()\n # 检查规定的声明是否存在于第一行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 无法读取或发生其它异常，直接视为校验失败\n return False\n \n # 只要存在至少一个代码文件，并且所有代码文件全部通过校验，即可判定为成功\n return has_code_file","tags":["Content","Format"]},{"constraint":"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字","validation_code":"","tags":["Style","Paradigm"]}]},{"round_id":9,"instruction":"最后，做一下体验优化。利用浏览器的本地存储功能自动保存当前左侧的文件列表以及每个文件修改后的源码内容，保证用户刷新页面后数据不丢失。另外加上文件类型拦截，如果用户通过拖拽或者选择器尝试添加非 `.md` 或 `.txt` 的文件，要在页面上方弹出一个红色的 Toast 报错提示框阻止其加入列表。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在12行及以内，此前要求函数行数15行及以内的规则不再生效；二是生成代码时，请在“=”号的左侧和右侧各放置两个空格。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"在你生成的所有代码文件中，注释内容必须统一使用中文编写","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符","validation_code":"def check_filename_convention(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端和相关项目的代码文件、文本文件扩展名\n code_extensions = {\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', \n '.scss', '.less', '.json', '.md', '.txt', '.py', '.sh'\n }\n \n # 定义需要忽略的目录，避免误判第三方依赖库或构建产物里的文件\n ignore_dirs = {'node_modules', 'dist', 'build', '__pycache__'}\n \n file_found = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏目录（如 .git）和不需要检查的目录\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ignore_dirs]\n \n for file in files:\n # 跳过隐藏文件\n if file.startswith('.'):\n continue\n \n # 提取文件扩展名和主文件名\n basename, ext = os.path.splitext(file)\n ext = ext.lower()\n \n # 只检查代码文件\n if ext in code_extensions:\n file_found = True\n \n # 检查主文件名是否仅由小写字母和下划线构成\n if not re.match(r'^[a-z_]+$', basename):\n return False\n \n # 如果找到了属于代码类型的文件，且全部通过了正则检查，则返回 True\n # 如果没找到任何对应文件，返回 False，代表无有效产出供验证\n return file_found","tags":["Style","Naming"]},{"constraint":"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"生成的代码文件必须满足完全不使用多层if的要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划","validation_code":"","tags":["Environment","Logging"]},{"constraint":"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请勿使用ESLint对你生成的代码进行检查","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\"pass\"、\"fail\"和\"error\"中选择其一）","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符","validation_code":"def check_function_name_length(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 用于去除代码内的注释，避免注释内的内容影响函数名的正则匹配\n def remove_comments(code: str) -> str:\n # 移除 JS/TS 单行注释\n code = re.sub(r'//.*', '', code)\n # 移除 JS/TS 多行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n # 移除 Python 单行注释\n code = re.sub(r'#.*', '', code)\n # 移除 HTML 注释\n code = re.sub(r'<!--[\\s\\S]*?-->', '', code)\n return code\n\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py'}\n \n # 匹配 Python 函数 (def funcName)\n py_func = re.compile(r'\\b(?:async\\s+)?def\\s+([a-zA-Z_]\\w*)\\s*\\(')\n \n # 匹配 JS/TS 传统函数 (function funcName)\n js_func1 = re.compile(r'\\b(?:async\\s+)?function\\s+([a-zA-Z_]\\w*)\\s*\\(')\n \n # 匹配 JS/TS 变量赋值形式的函数 (const funcName = () => {} 或 const funcName = function() {})\n js_func2 = re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_]\\w*)\\s*(?::\\s*[^=]+)?\\s*=\\s*(?:async\\s*)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_]\\w*\\s*=>)')\n \n # 匹配 JS/TS 类方法或对象内的方法 (funcName() {})\n js_func3 = re.compile(r'(?<!\\.)\\b([a-zA-Z_]\\w*)\\s*\\([^)]*\\)\\s*(?::\\s*[^={;]+)?\\s*\\{')\n \n # 匹配 JS/TS 对象属性形式的函数赋值 (funcName: () => {})\n js_func4 = re.compile(r'\\b([a-zA-Z_]\\w*)\\s*:\\s*(?:async\\s*)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_]\\w*\\s*=>)')\n \n # 常见可能产生误匹配的控制流关键字或保留字集合\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'return', \n 'super', 'constructor', 'else', 'import', 'export', \n 'try', 'finally', 'do', 'with', 'await', 'function', \n 'async', 'get', 'set', 'class', 'yield'\n }\n\n func_names = set()\n has_files = False\n\n # 1. 尝试从 workspace_path 中读取项目源码文件并提取函数名\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n has_files = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n content = remove_comments(content)\n \n if ext == '.py':\n for match in py_func.finditer(content):\n func_names.add(match.group(1))\n else:\n for match in js_func1.finditer(content):\n func_names.add(match.group(1))\n for match in js_func2.finditer(content):\n func_names.add(match.group(1))\n for match in js_func3.finditer(content):\n name = match.group(1)\n if name not in keywords:\n func_names.add(name)\n for match in js_func4.finditer(content):\n name = match.group(1)\n if name not in keywords:\n func_names.add(name)\n\n # 2. 如果项目路径中没有检索到相关代码文件，则兜底在最后一轮 response 中查找代码块\n if not has_files:\n code_blocks = re.findall(r'```(?:\\w+)?\\s*([\\s\\S]*?)```', response, re.IGNORECASE)\n for content in code_blocks:\n content = remove_comments(content)\n for match in py_func.finditer(content):\n func_names.add(match.group(1))\n for match in js_func1.finditer(content):\n func_names.add(match.group(1))\n for match in js_func2.finditer(content):\n func_names.add(match.group(1))\n for match in js_func3.finditer(content):\n name = match.group(1)\n if name not in keywords:\n func_names.add(name)\n for match in js_func4.finditer(content):\n name = match.group(1)\n if name not in keywords:\n func_names.add(name)\n\n # 3. 验证提取到的所有函数名称的字符长度是否满足 [4, 7] 的硬性范围\n for name in func_names:\n if not (4 <= len(name) <= 7):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"生成代码时，请将各函数的行数严格控制在12行及以内","validation_code":"import os\nimport re\nimport ast\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 用于检查 Python 代码中的函数行数\n def check_python_functions(code_str: str) -> bool:\n try:\n tree = ast.parse(code_str)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\n if node.end_lineno - node.lineno + 1 > 12:\n return False\n except:\n pass\n return True\n\n # 用于检查 JS/TS/HTML 等前端代码中的函数行数\n def check_js_functions(code_str: str) -> bool:\n processed = []\n i = 0\n length = len(code_str)\n \n # 1. 过滤掉字符串和注释中的内容，替换为空格，但保留换行符 \\n\n while i < length:\n char = code_str[i]\n \n # 处理字符串\n if char in ('\"', \"'\", '`'):\n quote = char\n processed.append(' ')\n i += 1\n while i < length:\n if code_str[i] == '\\\\' and i + 1 < length:\n processed.append(' ')\n processed.append(' ')\n i += 2\n continue\n if code_str[i] == quote:\n processed.append(' ')\n i += 1\n break\n if code_str[i] == '\\n':\n processed.append('\\n')\n else:\n processed.append(' ')\n i += 1\n continue\n \n # 处理单行注释\n if char == '/' and i + 1 < length and code_str[i+1] == '/':\n processed.append(' ')\n processed.append(' ')\n i += 2\n while i < length and code_str[i] != '\\n':\n processed.append(' ')\n i += 1\n continue\n \n # 处理多行注释\n if char == '/' and i + 1 < length and code_str[i+1] == '*':\n processed.append(' ')\n processed.append(' ')\n i += 2\n while i < length:\n if code_str[i] == '*' and i + 1 < length and code_str[i+1] == '/':\n processed.append(' ')\n processed.append(' ')\n i += 2\n break\n if code_str[i] == '\\n':\n processed.append('\\n')\n else:\n processed.append(' ')\n i += 1\n continue\n \n processed.append(char)\n i += 1\n\n clean_code = ''.join(processed)\n func_starts = []\n \n # 2. 定位函数的可能起点\n # function 关键字定义\n for m in re.finditer(r'\\bfunction\\b[^\\{;]*\\{', clean_code):\n func_starts.append(m.start())\n \n # 箭头函数定义\n for m in re.finditer(r'=>\\s*\\{', clean_code):\n func_starts.append(m.start())\n \n # 类方法/对象方法定义 (例如: myFunc(a, b) { )\n for m in re.finditer(r'\\b([a-zA-Z_$][\\w$]*)\\s*(?:<[^>]+>)?\\s*\\([^)]*\\)[^\\{;]*\\{', clean_code):\n name = m.group(1)\n # 排除前端内置的流程控制和错误处理等关键字\n if name not in ('if', 'for', 'while', 'switch', 'catch', 'return', 'with'):\n func_starts.append(m.start())\n \n func_starts = sorted(list(set(func_starts)))\n \n # 3. 计算行号辅助逻辑\n lines = clean_code.split('\\n')\n line_starts = []\n curr = 0\n for line in lines:\n line_starts.append(curr)\n curr += len(line) + 1\n \n def get_line_num(index):\n low = 0\n high = len(line_starts) - 1\n while low <= high:\n mid = (low + high) // 2\n if line_starts[mid] <= index:\n if mid == len(line_starts) - 1 or line_starts[mid+1] > index:\n return mid + 1\n else:\n low = mid + 1\n else:\n high = mid - 1\n return 1\n\n # 4. 基于大括号配对计算每个函数的行数\n for start_idx in func_starts:\n open_brace_idx = -1\n for i in range(start_idx, len(clean_code)):\n if clean_code[i] == '{':\n open_brace_idx = i\n break\n elif clean_code[i] in ';}':\n break\n \n if open_brace_idx != -1:\n depth = 0\n close_brace_idx = -1\n for i in range(open_brace_idx, len(clean_code)):\n if clean_code[i] == '{':\n depth += 1\n elif clean_code[i] == '}':\n depth -= 1\n if depth == 0:\n close_brace_idx = i\n break\n \n # 如果匹配成功闭合，则校验总行数\n if close_brace_idx != -1:\n start_line = get_line_num(start_idx)\n end_line = get_line_num(close_brace_idx)\n if end_line - start_line + 1 > 12:\n return False\n return True\n\n # 对生成的项目文件进行函数行数检验\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_content = f.read()\n except:\n continue\n \n if ext == '.py':\n if not check_python_functions(code_content):\n return False\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.svelte', '.css']:\n if not check_js_functions(code_content):\n return False\n else:\n # 对于未标明扩展名但看似前端代码的文件同样做兜底检查\n if 'function ' in code_content or '=>' in code_content:\n if not check_js_functions(code_content):\n return False\n\n # 对回复中的代码块进行兜底检查\n code_blocks = re.findall(r'```(?:.*?)\\n(.*?)```', response, re.DOTALL)\n for block in code_blocks:\n if not check_python_functions(block):\n return False\n if not check_js_functions(block):\n return False\n\n return True","tags":["Quantity","Range"]},{"constraint":"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-","validation_code":"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.exists(workspace_path):\n return False\n\n # 根据Web页面项目推测可能生成的前端及相关代码文件扩展名\n code_extensions = {\n '.html', '.js', '.css', '.ts', \n '.jsx', '.tsx', '.vue', '.py', \n '.mjs', '.cjs'\n }\n \n code_files = []\n for root, dirs, files in os.walk(workspace_path):\n # 排除通常不需要人为添加编码声明的依赖目录或版本控制目录\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n if '.git' in dirs:\n dirs.remove('.git')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files.append(os.path.join(root, file))\n \n # 如果项目中没有任何代码文件，则无法判定满足了要求\n if not code_files:\n return False\n \n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行中是否包含规定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 文件读取发生异常（如编码错误或没有权限），视为不符合要求\n return False\n \n return True","tags":["Content","Format"]},{"constraint":"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"生成代码时，请在“=”号的左侧和右侧各放置两个空格","validation_code":"def check_assignment_spaces(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 匹配单独的等号，并捕获其前后的连续空白字符\n # (?<![<>=!\\+\\-\\*/%&\\|\\^~:\\?]) 排除了等号前面紧连的常见组合操作符\n # (?![<>=!~]) 排除了等号后面紧连的操作符\n pattern = re.compile(r'([ \\t]*)(?<![<>=!\\+\\-\\*/%&\\|\\^~:\\?])=(?![<>=!~])([ \\t]*)')\n \n code_extensions = {\n '.py', '.js', '.ts', '.html', '.css', '.vue', '.java', \n '.cpp', '.c', '.h', '.cs', '.php', '.rb', '.go', '.rs', \n '.swift', '.kt', '.sh', '.bash', '.xml', '.yml', '.yaml'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 剔除 Python 的多行字符串\n content = re.sub(r'\"\"\"(.*?)\"\"\"', '\"\"\"\"\"\"', content, flags=re.DOTALL)\n content = re.sub(r\"'''(.*?)'''\", \"''''''\", content, flags=re.DOTALL)\n \n # 剔除普通字符串和模板字符串（避免字符串里的 \"=\" 号引发误判）\n content = re.sub(r'\"[^\"\\\\]*(\\\\.[^\"\\\\]*)*\"', '\"\"', content)\n content = re.sub(r\"'[^'\\\\]*(\\\\.[^'\\\\]*)*'\", \"''\", content)\n content = re.sub(r'`[^`\\\\]*(\\\\.[^`\\\\]*)*`', '``', content)\n \n # 剔除常见的块级/多行注释\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n lines = content.split('\\n')\n for line in lines:\n # 针对不同语言特性剔除单行注释\n if ext in {'.py', '.rb', '.sh', '.bash', '.yaml', '.yml'}:\n line = line.split('#')[0]\n elif ext in {'.js', '.ts', '.cpp', '.c', '.h', '.cs', '.java', '.php', '.go', '.rs', '.swift', '.kt', '.vue'}:\n line = line.split('//')[0]\n elif ext in {'.lua', '.sql'}:\n line = line.split('--')[0]\n \n # 寻找独立的等号\n for match in pattern.finditer(line):\n left_spaces = match.group(1)\n right_spaces = match.group(2)\n \n # 检查等号两侧是否恰好为两个空格\n if left_spaces != \" \" or right_spaces != \" \":\n return False\n\n # 若未找到任何可扫描的代码文件，则无法证实该要求被满足\n return has_code_file","tags":["Style","Layout"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个“Markdown批量转换工具”的Web页面。第一步，先搭个基础骨架。页面需要一个顶部导航栏，左边是侧边栏用于放文件列表，右边是主工作区。主工作区里暂时放一个简单的占位提示“请选择文件”，左侧栏放一个“添加文件”的按钮。整体风格要简洁现代一点，暂不实现具体的业务逻辑。请你后续始终遵循两个要求：一是生成的代码文件必须满足完全不使用多层if的要求，二是在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n pattern = re.compile(r'^[a-z_]+(\\\\.[a-z_]+)?$')\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n if not pattern.match(file_name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，现在的界面有模有样了。接下来完善左侧的文件列表。我希望用户可以通过点击“添加文件”按钮一次性选择多个 `.md` 文件，或者直接把文件拖拽到左侧区域来添加。添加后，文件会以列表形式展示在左侧栏中，显示文件名。每个文件项旁边要有一个删除图标可以单独移除，列表最下方再加一个“清空列表”的按钮。请你后续始终遵循以下新增要求：一是无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划；二是每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端与配置相关的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.md', '.py', '.sh'}\\n \\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录的干扰\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 排除隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n name, ext = os.path.splitext(file)\\n \\n # 判断当前文件是否为代码/配置文件\\n if ext.lower() in code_extensions:\\n # 检查文件名主干（不含扩展名）是否只包含小写字母和下划线\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"好，下一步来做核心的预览功能。当用户点击左侧列表里的某个文件时，右侧主工作区要变成左右分屏的布局。左半边是文本输入框，显示所选文件的 Markdown 源码，并允许用户直接修改；右半边是实时预览区，把源码实时渲染成普通的网页排版内容（比如标题、段落、列表要有基础的 HTML 样式）。请你后续始终遵循以下新增要求：一是对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误；二是每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 若工作区路径不存在，直接返回False\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据用户指令内容，推断可能的代码文件或配置文档类型\\n code_extensions = {'.html', '.css', '.js', '.json', '.md'}\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏文件夹及常见第三方依赖目录，避免误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file_name in files:\\n # 跳过系统隐藏文件（如 macOS 的 .DS_Store 等）\\n if file_name.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 仅针对推断出的代码文件类型进行名称验证\\n if ext in code_extensions:\\n has_code_file = True\\n \\n # 检查文件名是否只包含小写字母、下划线和点号（文件扩展名所必需）\\n if not re.match(r'^[a-z_.]+$', file_name):\\n return False\\n \\n # 如果存在代码文件且没有任何文件违反命名规则，则判定为通过\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新progress.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"渲染效果看起来有点单调。咱们给预览区加上主题切换功能。在顶部导航栏加一个下拉菜单，里面提供三种主题：GitHub风格、暗黑模式和护眼模式。选择不同主题时，右侧预览区的背景色、字体颜色、标题和引用块的样式都要实时发生变化。请你后续始终遵循以下要求：一是每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容，此前要求更新progress.md的规则不再生效；二是在你编写的代码文件里，任何一个函数的名字都必须包含至少3个字符，且最多6个字符。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 常见的前端项目相关代码/文本文件后缀\\n code_exts = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.vue', '.json', '.md', '.py', '.sh', '.txt'\\n }\\n \\n file_found = False\\n for root, dirs, files in os.walk(workspace_path):\\n rel_path = os.path.relpath(root, workspace_path)\\n \\n # 跳过隐藏目录（例如 .git, .vscode 等）\\n if any(part.startswith('.') for part in rel_path.split(os.sep) if part != '.'):\\n continue\\n \\n for file in files:\\n # 忽略隐藏文件（例如 .gitignore, .eslintrc.js）\\n if file.startswith('.'):\\n continue\\n \\n # 获取文件后缀名并转换为小写，以捕获扩展名被写为大写的违规情况（如 .HTML）\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 若属于代码文件或相关数据文件，进行命名规范校验\\n if ext in code_exts:\\n file_found = True\\n \\n # 正则校验：只能包含小写字母、下划线和点号\\n if not re.match(r'^[a-z_.]+$', file):\\n return False\\n \\n # 如果找到了文件并且全部检查通过返回True，如果没有任何有效代码文件则认为未完成基本生成返回False\\n return file_found\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少3个字符，且最多6个字符\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 正则表达式 1：匹配标准的 function 声明，如 function myFunc( 和 async function myFunc(\\n func_pattern_decl = re.compile(r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n \\n # 正则表达式 2：匹配函数赋值或键值对，如 const myFunc = () =>，或 myMethod: function()\\n func_pattern_var = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:=|:)\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)')\\n \\n # 正则表达式 3：匹配类/对象内的方法声明，如 render() { \\n func_pattern_method = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # JS/TS 内部常见的控制流与保留字，避免正则3产生误判\\n keywords = {\\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'throw', \\n 'function', 'class', 'import', 'export', 'const', 'let', 'var', \\n 'async', 'await', 'yield', 'typeof', 'instanceof', 'new', \\n 'super', 'constructor'\\n }\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见依赖文件夹，避免误算第三方库\\n dirs[:] = [d for d in dirs if d not in ('node_modules', '.git', 'dist', 'build', 'vendor')]\\n for file in files:\\n # 排除压缩文件\\n if file.endswith('.min.js'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 判定函数名称长度限制\\n if not (3 <= len(node.name) <= 6):\\n return False\\n except Exception:\\n pass\\n \\n elif ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n \\n names = []\\n names.extend(func_pattern_decl.findall(code))\\n names.extend(func_pattern_var.findall(code))\\n \\n method_names = func_pattern_method.findall(code)\\n names.extend([m for m in method_names if m not in keywords])\\n \\n for name in names:\\n # 判定函数名称长度限制\\n if not (3 <= len(name) <= 6):\\n return False\\n except Exception:\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在可以开始做导出功能了。在右侧工作区的右上角加两个小按钮：“导出当前HTML”和“导出当前PDF”。点击后，能把当前正在预览的文件下载下来。导出HTML就是保存一个带有当前主题样式的 .html 文件；导出PDF需要你调用合适的前端库生成一页完整的 PDF 文档，要求导出的 PDF 必须保留我们在预览区看到的主题排版和样式。请你后续始终遵循以下更新及新增要求：一是在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符，此前要求函数名3-6个字符的规则不再生效；二是生成代码时，请将各函数的行数严格控制在40行及以内。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区不存在，则无法验证\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 圈定在Web项目和上述指令中通常用到的相关代码与文档扩展名\\n target_exts = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.json', '.md', '.py', '.txt'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录及前端常见的第三方包、构建输出等无关目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['node_modules', '__pycache__', 'dist', 'build']]\\n \\n for file in files:\\n # 跳过隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n # 仅对代码文件和项目文档进行命名检查，兼容没有后缀的文件\\n if ext in target_exts or ext == '':\\n # 检查文件名是否仅包含小写字母、下划线和点号\\n # 用户要求：“命名必须全部使用小写字母和下划线，不可包含其他字符”\\n if not re.match(r'^[a-z_.]+$', file):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # JavaScript/TypeScript 中容易与正则产生混淆的关键字或内置词\\n js_keywords = {\\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'super', \\n 'constructor', 'function', 'import', 'export', 'default', \\n 'get', 'set', 'async', 'await', 'class', 'extends', 'new',\\n 'true', 'false', 'null', 'undefined', 'typeof', 'instanceof',\\n 'throw', 'delete', 'yield', 'let', 'const', 'var', 'with', 'do'\\n }\\n\\n # 正则表达式以覆盖常见的JavaScript/TypeScript函数定义形式\\n js_patterns = [\\n re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\('),\\n re.compile(r'(?<![\\\\w$])([a-zA-Z_$][\\\\w$]*)\\\\s*[:=]\\\\s*(?:async\\\\s+)?function\\\\b'),\\n re.compile(r'(?<![\\\\w$])([a-zA-Z_$][\\\\w$]*)\\\\s*[:=]\\\\s*(?:async\\\\s+)?\\\\([^)]*\\\\)\\\\s*=>'),\\n re.compile(r'(?<![\\\\w$])([a-zA-Z_$][\\\\w$]*)\\\\s*[:=]\\\\s*(?:async\\\\s+)?[a-zA-Z_$][\\\\w$]*\\\\s*=>'),\\n re.compile(r'(?<![\\\\w$])([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n ]\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略常见的第三方包及版本控制目录\\n if 'node_modules' in root or 'venv' in root or '.git' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n file_path = os.path.join(root, file)\\n \\n # 解析前端文件（HTML 和 JS/TS/Vue 等）\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 粗略剔除单行及多行注释，减少字符串里的误判\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n\\n if ext in ['.html', '.vue']:\\n # 在HTML或Vue文件中，只扫描<script>标签里的代码内容\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.IGNORECASE | re.DOTALL)\\n content_to_scan = '\\\\n'.join(scripts)\\n else:\\n content_to_scan = content\\n \\n # 利用正则提取所有匹配的函数名并检查长度\\n for pattern in js_patterns:\\n for match in pattern.findall(content_to_scan):\\n name = match.strip()\\n if name in js_keywords:\\n continue\\n if not (4 <= len(name) <= 7):\\n return False\\n except Exception:\\n pass\\n \\n # 解析Python文件（若存在后端情况）\\n elif ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n func_name = node.name\\n if not (4 <= len(func_name) <= 7):\\n return False\\n except Exception:\\n pass\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在40行及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_js_content(content: str) -> bool:\\n \\\"\\\"\\\"解析 JS/TS 类代码，匹配并检查每个函数的行数。\\\"\\\"\\\"\\n # 移除多行注释以避免大括号误判\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n lines = content.split('\\\\n')\\n func_stack = []\\n \\n for i, line in enumerate(lines):\\n cleaned = line\\n # 移除行内注释、字符串，降低由于字符串内部的大括号引起的误判风险\\n cleaned = re.sub(r'//.*', '', cleaned)\\n cleaned = re.sub(r'\\\".*?(?<!\\\\\\\\)\\\"', '\\\"\\\"', cleaned)\\n cleaned = re.sub(r\\\"'.*?(?<!\\\\\\\\)'\\\", \\\"''\\\", cleaned)\\n cleaned = re.sub(r\\\"`.*?`\\\", \\\"``\\\", cleaned) \\n \\n current_pos = 0\\n while True:\\n match = re.search(r'[{}]', cleaned[current_pos:])\\n if not match:\\n break\\n \\n char = match.group()\\n pos_in_cleaned = current_pos + match.start()\\n \\n if char == '{':\\n # 使用花括号和分号拆分，提取该代码块紧邻的前缀内容\\n local_prefix = re.split(r'[{};]', cleaned[:pos_in_cleaned])[-1]\\n \\n is_func = False\\n # 规则 1 & 2: 包含 function 关键字或 => 箭头函数\\n if re.search(r'\\\\bfunction\\\\b', local_prefix) or '=>' in local_prefix:\\n is_func = True\\n else:\\n # 规则 3: 对象方法、类方法定义 (如: onClick(event) {, 排除 if/for 等关键字)\\n if re.search(r'\\\\b(?!(?:if|for|while|switch|catch|with)\\\\b)\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[\\\\w<>\\\\[\\\\]\\\\s,\\\\|&]+)?\\\\s*$', local_prefix):\\n is_func = True\\n \\n func_stack.append({'start': i + 1, 'is_func': is_func})\\n elif char == '}':\\n if func_stack:\\n func_info = func_stack.pop()\\n if func_info['is_func']:\\n length = (i + 1) - func_info['start'] + 1\\n if length > 40:\\n return False\\n \\n current_pos = pos_in_cleaned + 1\\n\\n return True\\n\\n def extract_scripts_from_html(html_content: str) -> list:\\n \\\"\\\"\\\"从 HTML 或 Vue 文件中提取 script 标签的代码内容。\\\"\\\"\\\"\\n scripts = []\\n for match in re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', html_content, flags=re.IGNORECASE | re.DOTALL):\\n scripts.append(match.group(1))\\n return scripts\\n\\n # 遍历当前项目工作区的全部文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n if ext in ['.js', '.jsx', '.ts', '.tsx', '.vue']:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n if ext == '.vue':\\n scripts = extract_scripts_from_html(content)\\n for script in scripts:\\n if not check_js_content(script):\\n return False\\n else:\\n if not check_js_content(content):\\n return False\\n elif ext in ['.html', '.htm']:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n scripts = extract_scripts_from_html(content)\\n for script in scripts:\\n if not check_js_content(script):\\n return False\\n elif ext == '.py':\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n # 遍历并处理 Python 的函数定义节点\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n length = node.end_lineno - node.lineno + 1\\n if length > 40:\\n return False\\n except Exception:\\n # 若文件因特殊编码或语法异常导致解析失败，则跳过继续检查其他文件\\n pass\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"单个导出成功了，咱们挑战一下“批量操作”。在左侧边栏最下方加一个“批量导出 HTML”的按钮。点击它之后，工具要在后台遍历当前列表里的所有 Markdown 文件，把它们分别转换成带有当前主题样式的 HTML，然后打包成一个 `.zip` 压缩包，触发一次性下载。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在30行及以内，此前要求函数行数40行及以内的规则不再生效；二是针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区不存在，直接判定为未满足要求\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 定义前端/后端项目中常见的代码与配置文件后缀名\\n code_extensions = {\\n \\\".js\\\", \\\".jsx\\\", \\\".ts\\\", \\\".tsx\\\", \\n \\\".html\\\", \\\".css\\\", \\\".scss\\\", \\n \\\".json\\\", \\\".md\\\", \\\".py\\\", \\n \\\".yml\\\", \\\".yaml\\\"\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n name, ext = os.path.splitext(file)\\n \\n # 仅验证代码类型的生成文件\\n if ext.lower() in code_extensions:\\n has_code_file = True\\n \\n # 处理以点开头的隐藏文件，如 .eslintrc.json\\n if name.startswith('.'):\\n name = name[1:]\\n \\n # 如果文件仅仅由扩展名组成 (比如仅为 .gitignore但匹配了后缀逻辑时)，跳过名称判断\\n if not name:\\n continue\\n \\n # 剔除文件名中的内部点号，如 config.prod.js 提取出的 config.prod\\n name_without_dots = name.replace('.', '')\\n \\n # 核心验证逻辑：判断其余字符是否全为小写字母及下划线\\n if not re.match(r'^[a-z_]+$', name_without_dots):\\n return False\\n \\n # 如果存在代码文件且全部通过了验证，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 匹配普通 JS/TS 函数定义：function funcName(\\n js_pattern1 = r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\('\\n \\n # 匹配变量赋值形式的函数：const/let/var funcName = function | (...) => | x =>\\n js_pattern2 = r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\(.*?\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'\\n \\n # 匹配对象属性形式的函数：funcName: function | (...) => | x =>\\n js_pattern3 = r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\(.*?\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)'\\n \\n # 匹配 JS/TS 类或对象成员方法：funcName(...) {\\n js_method_pattern = r'^\\\\s*(?:async\\\\s+)?(?:public\\\\s+|private\\\\s+|protected\\\\s+|static\\\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(.*?\\\\)\\\\s*\\\\{'\\n \\n # 匹配 Python 函数定义：def func_name(\\n py_pattern = r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\('\\n \\n # JS/TS 常见控制流及关键字，用于在正则匹配方法名时排除误报\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'return', \\n 'super', 'constructor', 'get', 'set', 'with', 'typeof', 'new', \\n 'throw', 'let', 'const', 'var', 'import', 'export', 'default',\\n 'class', 'extends', 'yield', 'await', 'async', 'delete', 'void'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅针对可能包含代码的文件进行解析\\n if ext not in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext in ['.js', '.ts', '.jsx', '.tsx', '.vue', '.html']:\\n names = []\\n # 提取基本函数与箭头函数名称\\n names.extend(re.findall(js_pattern1, content))\\n names.extend(re.findall(js_pattern2, content))\\n names.extend(re.findall(js_pattern3, content))\\n \\n # 提取类/对象方法名称并排除关键字误判\\n methods = re.findall(js_method_pattern, content, re.MULTILINE)\\n for m in methods:\\n if m not in keywords:\\n names.append(m)\\n \\n # 验证函数名长度是否满足 [4, 7]\\n for name in names:\\n if len(name) < 4 or len(name) > 7:\\n return False\\n \\n if ext in ['.py']:\\n names = re.findall(py_pattern, content)\\n for name in names:\\n if len(name) < 4 or len(name) > 7:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在30行及以内\",\n \"validation_code\": \"def check_function_line_limits(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def extract_js_from_html(html_code: str) -> list:\\n scripts = []\\n pattern = re.compile(r'<script[^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)\\n for match in pattern.finditer(html_code):\\n scripts.append(match.group(1))\\n return scripts\\n\\n def check_js_code(code: str, max_lines: int) -> bool:\\n clean_code = []\\n i = 0\\n n = len(code)\\n in_str = False\\n str_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n # 状态机：将字符串内容和注释内容替换为空格，但必须严格保留换行符，保证行数统计正确\\n while i < n:\\n c = code[i]\\n if in_block_comment:\\n if c == '*' and i + 1 < n and code[i+1] == '/':\\n in_block_comment = False\\n clean_code.append(' ')\\n i += 2\\n else:\\n if c == '\\\\n':\\n clean_code.append('\\\\n')\\n else:\\n clean_code.append(' ')\\n i += 1\\n elif in_line_comment:\\n if c == '\\\\n':\\n in_line_comment = False\\n clean_code.append('\\\\n')\\n else:\\n clean_code.append(' ')\\n i += 1\\n elif in_str:\\n if c == '\\\\\\\\': # 跳过转义字符，避免错乱\\n clean_code.append(' ')\\n i += 2\\n elif c == str_char:\\n in_str = False\\n clean_code.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n clean_code.append('\\\\n')\\n else:\\n clean_code.append(' ')\\n i += 1\\n else:\\n if c == '/' and i + 1 < n and code[i+1] == '/':\\n in_line_comment = True\\n clean_code.append(' ')\\n i += 2\\n elif c == '/' and i + 1 < n and code[i+1] == '*':\\n in_block_comment = True\\n clean_code.append(' ')\\n i += 2\\n elif c in (\\\"'\\\", '\\\"', '`'):\\n in_str = True\\n str_char = c\\n clean_code.append(' ')\\n i += 1\\n else:\\n clean_code.append(c)\\n i += 1\\n \\n clean_code_str = \\\"\\\".join(clean_code)\\n \\n # 正则表达式：匹配 JS/TS 函数起始位置（涵盖 function 关键字、箭头函数及类/对象方法）\\n func_pattern = re.compile(\\n r'\\\\bfunction\\\\b'\\n r'|=>\\\\s*\\\\{'\\n r'|\\\\b(?!(?:if|for|while|switch|catch|return|with)\\\\b)[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*\\\\([^{};]*\\\\)\\\\s*\\\\{'\\n )\\n \\n for match in func_pattern.finditer(clean_code_str):\\n idx = clean_code_str.find('{', match.start())\\n semicolon_idx = clean_code_str.find(';', match.start())\\n \\n # 缺乏左大括号则跳过\\n if idx == -1:\\n continue\\n # 若在左大括号之前出现了分号，说明只是函数签名（例如 TS 重载声明），予以跳过\\n if semicolon_idx != -1 and semicolon_idx < idx:\\n continue\\n \\n open_braces = 1\\n j = idx + 1\\n # 寻找相匹配的右大括号闭合处\\n while j < len(clean_code_str) and open_braces > 0:\\n if clean_code_str[j] == '{' and clean_code_str[j-1] != '\\\\\\\\':\\n open_braces += 1\\n elif clean_code_str[j] == '}' and clean_code_str[j-1] != '\\\\\\\\':\\n open_braces -= 1\\n j += 1\\n \\n if open_braces == 0:\\n # 计算从函数名或起始关键字位置到右侧闭合大括号区域涵盖的 \\\\n 数量，加 1 即为所占用的代码行数\\n num_lines = clean_code_str[match.start():j].count('\\\\n') + 1\\n if num_lines > max_lines:\\n return False\\n return True\\n\\n def check_python_code(code: str, max_lines: int) -> bool:\\n try:\\n tree = ast.parse(code)\\n except Exception:\\n # 无法解析的情况下视为不符合合规预期代码格式\\n return False\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and node.end_lineno is not None:\\n lines = node.end_lineno - node.lineno + 1\\n if lines > max_lines:\\n return False\\n return True\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略构建及版本控制目录\\n for ignore_dir in ['node_modules', '.git', 'dist']:\\n if ignore_dir in dirs:\\n dirs.remove(ignore_dir)\\n \\n for file in files:\\n filepath = os.path.join(root, file)\\n # 处理前端相关的文件及组件\\n if file.endswith(('.js', '.jsx', '.ts', '.tsx', '.vue', '.html')):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if file.endswith(('.html', '.vue')):\\n scripts = extract_js_from_html(content)\\n else:\\n scripts = [content]\\n \\n for script in scripts:\\n if not check_js_code(script, 30):\\n return False\\n \\n # 处理潜在的 Python 代码文件\\n elif file.endswith('.py'):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n if not check_python_code(content, 30):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能生成的前后端代码文件扩展名集合\\n code_extensions = {'.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.css', '.vue'}\\n has_code_file = False\\n \\n # 若工作区路径不存在，则无法验证，返回False\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历当前工作区路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8 编码读取文件，使用 ignore 忽略可能引发异常的解码错误\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 判断第一行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 无法读取或打开文件的情况视为验证失败\\n return False\\n \\n # 如果找到了代码文件并且所有代码文件首行均符合要求，返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"批量HTML搞定了，接下来是重头戏“批量转PDF”。在刚才的按钮旁边加一个“批量导出 PDF”按钮。因为转 PDF 比较耗时，点击这个按钮后，页面中间必须要弹出一个带有进度条的模态框，显示进度（例如“正在生成 PDF: 2/5”），每完成一个就更新一下进度。全部生成完毕后，同样把所有的 PDF 文件打包成一个 `.zip` 文件下载，然后自动关闭进度条模态框。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在25行及以内，此前要求函数行数30行及以内的规则不再生效；二是每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 若工作区路径不存在，直接返回不满足\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 定义正则表达式，只允许小写字母和下划线\\n pattern = re.compile(r'^[a-z_]+$')\\n has_files = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤隐藏目录和第三方依赖项目录，不检查这些目录里的文件\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__', 'venv', 'env')]\\n \\n for f in files:\\n # 忽略以 \\\".\\\" 开头的隐藏文件（如 .gitignore, .eslintrc 等）\\n if f.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, f)\\n if not os.path.isfile(file_path):\\n continue\\n \\n has_files = True\\n \\n # 分离出主文件名和扩展名\\n name, ext = os.path.splitext(f)\\n \\n # 检查主文件名部分是否只包含小写字母和下划线\\n if not pattern.match(name):\\n return False\\n \\n # 只要存在至少一个文件且全部通过了检查，即返回 True\\n return has_files\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n if not os.path.exists(workspace_path):\\n return True\\n \\n # 匹配 JavaScript/TypeScript 常见函数定义的正则表达式\\n js_patterns = [\\n r'\\\\bfunction\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\(', # function name()\\n r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=\\\\s*function\\\\b', # name = function()\\n r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*:\\\\s*function\\\\b', # name: function()\\n r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=\\\\s*(?:\\\\([^)]*\\\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=>', # name = () =>\\n r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*:\\\\s*(?:\\\\([^)]*\\\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=>' # name: () =>\\n ]\\n \\n all_functions = []\\n \\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n all_functions.append(node.name)\\n except Exception:\\n # 如果 ast 解析失败，退化为正则匹配，且屏蔽注释\\n content_no_comment = re.sub(r'#.*', '', content)\\n all_functions.extend(re.findall(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(', content_no_comment))\\n \\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n # 屏蔽前端单行和多行注释\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n for p in js_patterns:\\n all_functions.extend(re.findall(p, content))\\n \\n elif ext in ['.html', '.htm', '.vue']:\\n # 屏蔽 HTML 注释，并寻找内联 script\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n scripts = re.findall(r'<script[^>]*>\\\\s*(.*?)\\\\s*</script>', content, re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n script = re.sub(r'//.*', '', script)\\n script = re.sub(r'/\\\\*.*?\\\\*/', '', script, flags=re.DOTALL)\\n for p in js_patterns:\\n all_functions.extend(re.findall(p, script))\\n \\n # 检查提取到的所有函数名字符数是否在 [4, 7] 范围内\\n for func_name in all_functions:\\n if not (4 <= len(func_name) <= 7):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在25行及以内\",\n \"validation_code\": \"import os\\nimport ast\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n def mask_comments_and_strings(js_code: str) -> str:\\n \\\"\\\"\\\"\\n 将 JavaScript/TypeScript 代码中的注释和字符串内容替换为空格，\\n 保持原有的换行符不变，以确保代码行号一致，避免大括号等特殊字符造成误判。\\n \\\"\\\"\\\"\\n result = []\\n i = 0\\n n = len(js_code)\\n in_string = False\\n string_char = ''\\n in_single_comment = False\\n in_multi_comment = False\\n \\n while i < n:\\n c = js_code[i]\\n if in_single_comment:\\n if c == '\\\\n':\\n in_single_comment = False\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_multi_comment:\\n if c == '*' and i + 1 < n and js_code[i+1] == '/':\\n in_multi_comment = False\\n result.append(' ')\\n i += 2\\n else:\\n if c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if in_string:\\n if c == '\\\\\\\\':\\n result.append(' ')\\n i += 2\\n elif c == string_char:\\n in_string = False\\n result.append(' ')\\n i += 1\\n else:\\n if c == '\\\\n':\\n result.append(c)\\n else:\\n result.append(' ')\\n i += 1\\n continue\\n \\n if c == '/' and i + 1 < n:\\n nc = js_code[i+1]\\n if nc == '/':\\n in_single_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n elif nc == '*':\\n in_multi_comment = True\\n result.append(' ')\\n i += 2\\n continue\\n \\n if c in ['\\\"', \\\"'\\\", '`']:\\n in_string = True\\n string_char = c\\n result.append(' ')\\n i += 1\\n continue\\n \\n result.append(c)\\n i += 1\\n \\n return ''.join(result)\\n\\n def extract_function_lines(masked_code: str):\\n \\\"\\\"\\\"\\n 基于掩码处理过后的前端代码，通过简单的词法分析与大括号层级推断，识别各函数的起始与结束行号。\\n \\\"\\\"\\\"\\n stack = []\\n functions = []\\n i = 0\\n n = len(masked_code)\\n current_line = 1\\n \\n while i < n:\\n c = masked_code[i]\\n if c == '\\\\n':\\n current_line += 1\\n i += 1\\n continue\\n \\n if c == '{':\\n is_func = False\\n j = i - 1\\n # 向前回溯，查看是否有 function 关键字或箭头函数的 =>\\n while j >= 0 and masked_code[j] not in '{};':\\n if masked_code[j:j+8] == 'function' and (j == 0 or not masked_code[j-1].isalnum()) and (j+8 >= n or not masked_code[j+8].isalnum()):\\n is_func = True\\n break\\n if masked_code[j:j+2] == '=>':\\n is_func = True\\n break\\n j -= 1\\n \\n # 向前回溯，若未匹配到基本关键字，推测是否属于类方法或对象方法等形如 methodName(...) \\n if not is_func:\\n j = i - 1\\n while j >= 0 and masked_code[j] not in '{};':\\n j -= 1\\n segment = masked_code[j+1:i].strip()\\n # 排除控制流等常规结构块\\n if not re.search(r'\\\\b(if|for|while|switch|catch)\\\\b', segment):\\n if re.search(r'\\\\b\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*$', segment):\\n is_func = True\\n \\n stack.append((current_line, is_func))\\n i += 1\\n continue\\n \\n if c == '}':\\n if stack:\\n start_line, is_func = stack.pop()\\n if is_func:\\n functions.append((start_line, current_line))\\n i += 1\\n continue\\n \\n i += 1\\n \\n return functions\\n\\n def check_code_content(content: str, ext: str) -> bool:\\n \\\"\\\"\\\"\\n 针对指定的内容及扩展名执行函数行数判定。\\n \\\"\\\"\\\"\\n try:\\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n lines = node.end_lineno - node.lineno + 1\\n if lines > 25:\\n return False\\n except Exception:\\n pass\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.html', '.vue']:\\n if ext in ['.html', '.vue']:\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n code_to_check = '\\\\n'.join(scripts)\\n else:\\n code_to_check = content\\n \\n masked = mask_comments_and_strings(code_to_check)\\n funcs = extract_function_lines(masked)\\n for start, end in funcs:\\n if end - start + 1 > 25:\\n return False\\n return True\\n except Exception:\\n return True\\n\\n # 1. 检查工作空间内部的所有支持检查的文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n if ext in ['.py', '.js', '.jsx', '.ts', '.tsx', '.html', '.vue']:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n try:\\n content = f.read()\\n except Exception:\\n continue\\n if not check_code_content(content, ext):\\n return False\\n\\n # 2. 检查人工智能助手当前回复中的内联代码块\\n code_blocks = re.findall(r'```(\\\\w*)\\\\n(.*?)```', response, re.DOTALL)\\n for lang, code in code_blocks:\\n lang = lang.lower()\\n # 默认作为 js 分析\\n ext = '.js' \\n if lang in ['python', 'py']:\\n ext = '.py'\\n elif lang in ['html', 'vue']:\\n ext = '.html'\\n \\n if not check_code_content(code, ext):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义常见的代码文件拓展名\\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.sh'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录和 node_modules 第三方依赖库目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含规定的声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果读取出现编码等异常问题，直接视为不满足要求\\n return False\\n \\n # 确保项目中存在至少一个代码文件，且它们的第一行均符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"生成 PDF 的排版通常需要定制。咱们加一个“导出设置”功能。在顶部导航栏加个“设置”图标，点击弹出一个面板。里面可以设置最终导出 PDF 的页面大小（A4、B5、Letter）、纸张方向（纵向、横向）以及页边距大小。请确保这些设置选项真正生效，当用户再次点击单个或批量导出 PDF 时，按照用户选择的参数来生成文件。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在20行及以内，此前要求函数行数25行及以内的规则不再生效；二是你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 推断可能的代码文件或项目配置文件扩展名\\n code_extensions = {'.js', '.html', '.css', '.py', '.json', '.md', '.ts', '.jsx', '.tsx', '.vue'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录及非AI生成的典型第三方目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('node_modules', '__pycache__')]\\n \\n for file in files:\\n # 忽略隐藏文件（如 .gitignore 等）\\n if file.startswith('.'):\\n continue\\n \\n name, ext = os.path.splitext(file)\\n \\n # 若属于代码文件或配置文件，则进行名称合规性检查\\n if ext.lower() in code_extensions:\\n # 检查主文件名是否仅包含小写字母和下划线\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_function_name_length(response: str, workspace_path: str) -> bool:\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n # 针对 Python 文件，使用 AST 解析准确提取函数名\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n tree = ast.parse(f.read())\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n # 明确要求字符数，因此直接使用 len 进行统计\\n if not (4 <= len(node.name) <= 7):\\n return False\\n except Exception:\\n pass\\n \\n # 针对前端及网页相关代码文件，使用正则表达式提取常见函数定义\\n elif file.endswith(('.js', '.ts', '.jsx', '.tsx', '.vue', '.html')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除代码中的注释以防止正则误匹配\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n \\n matches = []\\n # 模式1: 匹配 function myFunc(...)\\n matches.extend(re.findall(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(', content))\\n \\n # 模式2: 匹配 const/let/var myFunc = function(...)\\n matches.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*function\\\\b', content))\\n \\n # 模式3: 匹配箭头函数 const/let/var myFunc = (...) => \\n matches.extend(re.findall(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s*)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', content))\\n \\n for name in matches:\\n # 明确要求字符数，因此直接使用 len 进行统计\\n if not (4 <= len(name) <= 7):\\n return False\\n except Exception:\\n pass\\n \\n # 没有发现任何不符合 4~7 字符长度规定的函数，即判定为通过\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在20行及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def check_python_file(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n return True \\n \\n try:\\n tree = ast.parse(code)\\n except Exception:\\n return True\\n \\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n if node.end_lineno - node.lineno + 1 > 20:\\n return False\\n return True\\n\\n def check_js_like_file(filepath):\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n text = f.read()\\n except Exception:\\n return True\\n \\n i = 0\\n line_num = 1\\n in_string = False\\n string_char = ''\\n in_line_comment = False\\n in_block_comment = False\\n \\n brace_level = 0\\n func_stack = []\\n recent_code = \\\"\\\"\\n \\n while i < len(text):\\n char = text[i]\\n \\n if char == '\\\\n':\\n line_num += 1\\n if in_line_comment:\\n in_line_comment = False\\n recent_code += ' '\\n i += 1\\n continue\\n \\n if in_line_comment:\\n i += 1\\n continue\\n \\n if in_block_comment:\\n if char == '*' and i + 1 < len(text) and text[i+1] == '/':\\n in_block_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n if in_string:\\n if char == '\\\\\\\\': \\n i += 2\\n continue\\n if char == string_char:\\n in_string = False\\n i += 1\\n continue\\n \\n if char in ('\\\"', \\\"'\\\", \\\"`\\\"):\\n in_string = True\\n string_char = char\\n i += 1\\n continue\\n \\n if char == '/' and i + 1 < len(text):\\n next_char = text[i+1]\\n if next_char == '/':\\n in_line_comment = True\\n i += 2\\n continue\\n elif next_char == '*':\\n in_block_comment = True\\n i += 2\\n continue\\n \\n if char == '{':\\n is_func = False\\n segment = recent_code.strip()\\n if \\\"function\\\" in segment or \\\"=>\\\" in segment:\\n is_func = True\\n elif re.search(r'\\\\w+\\\\s*\\\\([^)]*\\\\)\\\\s*$', segment):\\n if not re.search(r'\\\\b(if|for|while|switch|catch)\\\\b\\\\s*\\\\([^)]*\\\\)\\\\s*$', segment):\\n is_func = True\\n \\n if is_func:\\n func_stack.append({'start': line_num, 'level': brace_level})\\n \\n brace_level += 1\\n recent_code = \\\"\\\"\\n elif char == '}':\\n brace_level -= 1\\n if func_stack and func_stack[-1]['level'] == brace_level:\\n func_info = func_stack.pop()\\n if line_num - func_info['start'] + 1 > 20:\\n return False\\n recent_code = \\\"\\\"\\n elif char == ';':\\n recent_code = \\\"\\\"\\n else:\\n recent_code += char\\n if len(recent_code) > 200:\\n recent_code = recent_code[-200:]\\n \\n i += 1\\n \\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n filepath = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n if ext == '.py':\\n if not check_python_file(filepath):\\n return False\\n elif ext in ('.js', '.ts', '.html', '.vue', '.jsx', '.tsx'):\\n if not check_js_like_file(filepath):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 针对前端Web项目推断可能的代码文件扩展名，排除文档(.md)和纯数据文件(.json)\\n code_extensions = {'.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', '.py'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查文件第一行是否包含指定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 无法读取或发生异常时，稳妥起见视为未满足要求\\n return False\\n \\n # 如果工作区内产生了代码文件并且均满足要求，则返回 True\\n # 若没有任何代码文件，则视作未正常生成项目，返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"继续提升一下 Markdown 的解析能力，单纯的文字不够用。请引入相关的代码高亮方案，让预览区和导出文件里的代码块带上明显的语法高亮颜色。同时，增强解析器，使其支持解析渲染 Markdown 的表格以及带有 `[x]` 和 `[ ]` 的任务复选框列表。请你后续始终遵循以下更新要求：一是请勿使用ESLint对你生成的代码进行检查，此前要求必须对生成的代码实施ESLint验证并更正所有报错的规则不再生效；二是生成代码时，请将各函数的行数严格控制在15行及以内，此前要求函数行数20行及以内的规则不再生效。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_generated_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略以 `.` 开头的隐藏目录（例如 .git, .idea 等）\\n dirs[:] = [d for d in dirs if not d.startswith('.')]\\n \\n for file in files:\\n # 忽略以 `.` 开头的隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n has_generated_files = True\\n \\n # 提取主文件名：去除最后一段文件扩展名\\n if '.' in file:\\n base_name = file.rsplit('.', 1)[0]\\n else:\\n base_name = file\\n \\n # 校验主文件名是否全部由小写字母和下划线构成\\n if not re.match(r'^[a-z_]+$', base_name):\\n return False\\n\\n # 若遍历正常结束且至少存在一个文件，则满足要求\\n return has_generated_files\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请勿使用ESLint对你生成的代码进行检查\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_function_name_length(response: str, workspace_path: str) -> bool:\\n ignored_dirs = {'node_modules', '.git', '.vscode', '.idea', 'dist', 'build', '__pycache__', 'venv', 'env'}\\n valid_exts = {'.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.py', '.svelte'}\\n \\n function_names = set()\\n keywords = {\\n 'if', 'else', 'for', 'while', 'switch', 'catch', 'function', 'return', \\n 'super', 'constructor', 'with', 'do', 'const', 'let', 'var', 'yield', \\n 'await', 'import', 'export', 'typeof', 'instanceof'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除无关目录\\n dirs[:] = [d for d in dirs if d not in ignored_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in valid_exts:\\n continue\\n \\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n function_names.add(node.name)\\n except Exception:\\n pass\\n else:\\n # 粗略去除单行与多行注释，以避免提取到注释内的残留函数\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n if ext in {'.html', '.vue', '.svelte'}:\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n\\n # 1. 匹配普通函数声明 function foo(...)\\n for match in re.finditer(r'\\\\bfunction(?:\\\\s+|\\\\s*\\\\*\\\\s*)([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(', content):\\n name = match.group(1)\\n if name not in keywords:\\n function_names.add(name)\\n \\n # 2. 匹配变量赋值与箭头函数 foo = function() 或者 foo = () =>\\n for match in re.finditer(r'(?:^|[\\\\s\\\\.])([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\s*\\\\(|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)', content):\\n name = match.group(1)\\n if name not in keywords:\\n function_names.add(name)\\n \\n # 3. 匹配类与对象方法 foo() { ... }\\n for match in re.finditer(r'(?:^|[\\\\n\\\\r\\\\{,;])\\\\s*(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+|static\\\\s+)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', content):\\n name = match.group(1)\\n if name not in keywords:\\n function_names.add(name)\\n \\n # 4. 匹配对象属性函数 foo: function(...)\\n for match in re.finditer(r'(?:^|[\\\\s\\\\{,])([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?function\\\\s*\\\\(', content):\\n name = match.group(1)\\n if name not in keywords:\\n function_names.add(name)\\n \\n # 5. 匹配对象属性箭头函数 foo: (...) =>\\n for match in re.finditer(r'(?:^|[\\\\s\\\\{,])([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', content):\\n name = match.group(1)\\n if name not in keywords:\\n function_names.add(name)\\n\\n # 验证提取出的所有函数名字字符数是否严格在 [4, 7] 范围内\\n for name in function_names:\\n if not (4 <= len(name) <= 7):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在15行及以内\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import ast\\n import re\\n\\n def check_py(code: str) -> bool:\\n \\\"\\\"\\\"检查 Python 文件中函数的行数是否符合要求\\\"\\\"\\\"\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n # 计算函数的总行跨度\\n if (node.end_lineno - node.lineno + 1) > 15:\\n return False\\n except Exception:\\n pass\\n return True\\n\\n def check_js_html(code: str) -> bool:\\n \\\"\\\"\\\"启发式检查 JS/TS/HTML 类文件中的函数行数是否符合要求\\\"\\\"\\\"\\n # 移除字符串内的内容（双引号、单引号、反引号），防止内部出现的花括号干扰计数\\n code_clean = re.sub(r'\\\"(?:[^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', code)\\n code_clean = re.sub(r\\\"'(?:[^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", code_clean)\\n code_clean = re.sub(r'`[^`]*`', '``', code_clean)\\n # 移除单行与多行注释\\n code_clean = re.sub(r'//.*', '', code_clean)\\n code_clean = re.sub(r'/\\\\*.*?\\\\*/', '', code_clean, flags=re.DOTALL)\\n \\n clean_lines = code_clean.split('\\\\n')\\n \\n # 匹配 `function` 声明 或 带代码块的箭头函数 `=> {`\\n func_pattern = re.compile(r'\\\\b(function)\\\\b|=>\\\\s*\\\\{')\\n # 匹配类方法或对象方法：如 myMethod(args) { \\n method_pattern = re.compile(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n # JS常见控制流关键字，避开 if(...) { 等被误判为类方法\\n keywords = {'if', 'for', 'while', 'catch', 'switch', 'return', 'with'}\\n \\n for i, line in enumerate(clean_lines):\\n is_func = False\\n if func_pattern.search(line):\\n is_func = True\\n else:\\n m = method_pattern.search(line)\\n if m and m.group(1) not in keywords:\\n is_func = True\\n \\n if is_func:\\n brace_count = 0\\n started = False\\n end_i = i\\n \\n # 追踪大括号嵌套级别，确定函数的收尾行\\n for j in range(i, len(clean_lines)):\\n for char in clean_lines[j]:\\n if char == '{':\\n brace_count += 1\\n started = True\\n elif char == '}':\\n brace_count -= 1\\n \\n if started and brace_count == 0:\\n end_i = j\\n break\\n \\n if started:\\n # 判断当前找到的函数体行数是否超标\\n if (end_i - i + 1) > 15:\\n return False\\n return True\\n\\n # 遍历当前项目工作空间的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 查验 Python 代码\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_py(code):\\n return False\\n except Exception:\\n continue\\n # 查验 Web 前端相关代码 (JavaScript, TypeScript, HTML, Vue 等)\\n elif file.endswith(('.js', '.ts', '.html', '.jsx', '.tsx', '.vue')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_js_html(code):\\n return False\\n except Exception:\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义属于代码文件的常见后缀（对于Web页面项目主要涵盖HTML、CSS、JS等前端文件）\\n code_extensions = {\\n '.js', '.html', '.css', '.ts', '.vue', '.jsx', '.tsx',\\n '.py', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php', '.sh'\\n }\\n # 排除不需要检查的第三方依赖或系统生成的文件夹\\n exclude_dirs = {'.git', 'node_modules', 'dist', 'build', '__pycache__'}\\n \\n # 检查工作区目录是否存在且合法\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要遍历的目录\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n try:\\n # 使用 utf-8-sig 编码读取可避免因为文件带 BOM 头部而影响首行字符串的判断\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline()\\n # 检查规定的声明是否存在于第一行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 无法读取或发生其它异常，直接视为校验失败\\n return False\\n \\n # 只要存在至少一个代码文件，并且所有代码文件全部通过校验，即可判定为成功\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 9,\n \"instruction\": \"最后，做一下体验优化。利用浏览器的本地存储功能自动保存当前左侧的文件列表以及每个文件修改后的源码内容，保证用户刷新页面后数据不丢失。另外加上文件类型拦截，如果用户通过拖拽或者选择器尝试添加非 `.md` 或 `.txt` 的文件，要在页面上方弹出一个红色的 Toast 报错提示框阻止其加入列表。请你后续始终遵循以下更新及新增要求：一是生成代码时，请将各函数的行数严格控制在12行及以内，此前要求函数行数15行及以内的规则不再生效；二是生成代码时，请在“=”号的左侧和右侧各放置两个空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"你所生成的代码文件内，凡涉及到文件引用，必须且只能运用相对路径，避免使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及修改现有文件的情况，均需输出文件的完整代码来覆盖原文件，坚决禁止使用 Unified Diff 及其他任何补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"在你生成的所有代码文件中，注释内容必须统一使用中文编写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"保持你生成的代码文件里每个函数的圈复杂度均在 3 以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你在生成代码文件时，需将循环嵌套层数控制在 2 层以内\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件，其命名必须全部使用小写字母和下划线，不可包含其他字符\",\n \"validation_code\": \"def check_filename_convention(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端和相关项目的代码文件、文本文件扩展名\\n code_extensions = {\\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', \\n '.scss', '.less', '.json', '.md', '.txt', '.py', '.sh'\\n }\\n \\n # 定义需要忽略的目录，避免误判第三方依赖库或构建产物里的文件\\n ignore_dirs = {'node_modules', 'dist', 'build', '__pycache__'}\\n \\n file_found = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除隐藏目录（如 .git）和不需要检查的目录\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ignore_dirs]\\n \\n for file in files:\\n # 跳过隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n # 提取文件扩展名和主文件名\\n basename, ext = os.path.splitext(file)\\n ext = ext.lower()\\n \\n # 只检查代码文件\\n if ext in code_extensions:\\n file_found = True\\n \\n # 检查主文件名是否仅由小写字母和下划线构成\\n if not re.match(r'^[a-z_]+$', basename):\\n return False\\n \\n # 如果找到了属于代码类型的文件，且全部通过了正则检查，则返回 True\\n # 如果没找到任何对应文件，返回 False，代表无有效产出供验证\\n return file_found\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"当你准备更改已有文件时，切勿在当前同级目录下生成该文件修改前的备份文件\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件必须满足完全不使用多层if的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"在进行文件操作时，必须尽可能依赖专用工具（如Read, Write, Edit），而不是使用 bash 命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"无论何时修改代码，首要步骤都必须是在implementation_plan.json中写入本次修改的规划\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"每当项目发生变更时，请务必更新dev_records.md文件，以记录本轮所完成的内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请勿使用ESLint对你生成的代码进行检查\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"每当生成代码的测试完成时，务必将结果更新到test_results.json文件中。该文件的格式必须为List[Dict]，其中每个字典元素需具备两个键：一个是记录本次测试内容的description，另一个是状态status（其值必须在\\\"pass\\\"、\\\"fail\\\"和\\\"error\\\"中选择其一）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在你编写的代码文件里，任何一个函数的名字都必须包含至少4个字符，且最多7个字符\",\n \"validation_code\": \"def check_function_name_length(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 用于去除代码内的注释，避免注释内的内容影响函数名的正则匹配\\n def remove_comments(code: str) -> str:\\n # 移除 JS/TS 单行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除 JS/TS 多行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n # 移除 Python 单行注释\\n code = re.sub(r'#.*', '', code)\\n # 移除 HTML 注释\\n code = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', code)\\n return code\\n\\n valid_exts = {'.js', '.ts', '.jsx', '.tsx', '.vue', '.html', '.py'}\\n \\n # 匹配 Python 函数 (def funcName)\\n py_func = re.compile(r'\\\\b(?:async\\\\s+)?def\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n \\n # 匹配 JS/TS 传统函数 (function funcName)\\n js_func1 = re.compile(r'\\\\b(?:async\\\\s+)?function\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n \\n # 匹配 JS/TS 变量赋值形式的函数 (const funcName = () => {} 或 const funcName = function() {})\\n js_func2 = re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_]\\\\w*)\\\\s*(?::\\\\s*[^=]+)?\\\\s*=\\\\s*(?:async\\\\s*)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_]\\\\w*\\\\s*=>)')\\n \\n # 匹配 JS/TS 类方法或对象内的方法 (funcName() {})\\n js_func3 = re.compile(r'(?<!\\\\.)\\\\b([a-zA-Z_]\\\\w*)\\\\s*\\\\([^)]*\\\\)\\\\s*(?::\\\\s*[^={;]+)?\\\\s*\\\\{')\\n \\n # 匹配 JS/TS 对象属性形式的函数赋值 (funcName: () => {})\\n js_func4 = re.compile(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*:\\\\s*(?:async\\\\s*)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_]\\\\w*\\\\s*=>)')\\n \\n # 常见可能产生误匹配的控制流关键字或保留字集合\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'return', \\n 'super', 'constructor', 'else', 'import', 'export', \\n 'try', 'finally', 'do', 'with', 'await', 'function', \\n 'async', 'get', 'set', 'class', 'yield'\\n }\\n\\n func_names = set()\\n has_files = False\\n\\n # 1. 尝试从 workspace_path 中读取项目源码文件并提取函数名\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n has_files = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n content = remove_comments(content)\\n \\n if ext == '.py':\\n for match in py_func.finditer(content):\\n func_names.add(match.group(1))\\n else:\\n for match in js_func1.finditer(content):\\n func_names.add(match.group(1))\\n for match in js_func2.finditer(content):\\n func_names.add(match.group(1))\\n for match in js_func3.finditer(content):\\n name = match.group(1)\\n if name not in keywords:\\n func_names.add(name)\\n for match in js_func4.finditer(content):\\n name = match.group(1)\\n if name not in keywords:\\n func_names.add(name)\\n\\n # 2. 如果项目路径中没有检索到相关代码文件，则兜底在最后一轮 response 中查找代码块\\n if not has_files:\\n code_blocks = re.findall(r'```(?:\\\\w+)?\\\\s*([\\\\s\\\\S]*?)```', response, re.IGNORECASE)\\n for content in code_blocks:\\n content = remove_comments(content)\\n for match in py_func.finditer(content):\\n func_names.add(match.group(1))\\n for match in js_func1.finditer(content):\\n func_names.add(match.group(1))\\n for match in js_func2.finditer(content):\\n func_names.add(match.group(1))\\n for match in js_func3.finditer(content):\\n name = match.group(1)\\n if name not in keywords:\\n func_names.add(name)\\n for match in js_func4.finditer(content):\\n name = match.group(1)\\n if name not in keywords:\\n func_names.add(name)\\n\\n # 3. 验证提取到的所有函数名称的字符长度是否满足 [4, 7] 的硬性范围\\n for name in func_names:\\n if not (4 <= len(name) <= 7):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请将各函数的行数严格控制在12行及以内\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 用于检查 Python 代码中的函数行数\\n def check_python_functions(code_str: str) -> bool:\\n try:\\n tree = ast.parse(code_str)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n if hasattr(node, 'end_lineno') and hasattr(node, 'lineno'):\\n if node.end_lineno - node.lineno + 1 > 12:\\n return False\\n except:\\n pass\\n return True\\n\\n # 用于检查 JS/TS/HTML 等前端代码中的函数行数\\n def check_js_functions(code_str: str) -> bool:\\n processed = []\\n i = 0\\n length = len(code_str)\\n \\n # 1. 过滤掉字符串和注释中的内容，替换为空格，但保留换行符 \\\\n\\n while i < length:\\n char = code_str[i]\\n \\n # 处理字符串\\n if char in ('\\\"', \\\"'\\\", '`'):\\n quote = char\\n processed.append(' ')\\n i += 1\\n while i < length:\\n if code_str[i] == '\\\\\\\\' and i + 1 < length:\\n processed.append(' ')\\n processed.append(' ')\\n i += 2\\n continue\\n if code_str[i] == quote:\\n processed.append(' ')\\n i += 1\\n break\\n if code_str[i] == '\\\\n':\\n processed.append('\\\\n')\\n else:\\n processed.append(' ')\\n i += 1\\n continue\\n \\n # 处理单行注释\\n if char == '/' and i + 1 < length and code_str[i+1] == '/':\\n processed.append(' ')\\n processed.append(' ')\\n i += 2\\n while i < length and code_str[i] != '\\\\n':\\n processed.append(' ')\\n i += 1\\n continue\\n \\n # 处理多行注释\\n if char == '/' and i + 1 < length and code_str[i+1] == '*':\\n processed.append(' ')\\n processed.append(' ')\\n i += 2\\n while i < length:\\n if code_str[i] == '*' and i + 1 < length and code_str[i+1] == '/':\\n processed.append(' ')\\n processed.append(' ')\\n i += 2\\n break\\n if code_str[i] == '\\\\n':\\n processed.append('\\\\n')\\n else:\\n processed.append(' ')\\n i += 1\\n continue\\n \\n processed.append(char)\\n i += 1\\n\\n clean_code = ''.join(processed)\\n func_starts = []\\n \\n # 2. 定位函数的可能起点\\n # function 关键字定义\\n for m in re.finditer(r'\\\\bfunction\\\\b[^\\\\{;]*\\\\{', clean_code):\\n func_starts.append(m.start())\\n \\n # 箭头函数定义\\n for m in re.finditer(r'=>\\\\s*\\\\{', clean_code):\\n func_starts.append(m.start())\\n \\n # 类方法/对象方法定义 (例如: myFunc(a, b) { )\\n for m in re.finditer(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*(?:<[^>]+>)?\\\\s*\\\\([^)]*\\\\)[^\\\\{;]*\\\\{', clean_code):\\n name = m.group(1)\\n # 排除前端内置的流程控制和错误处理等关键字\\n if name not in ('if', 'for', 'while', 'switch', 'catch', 'return', 'with'):\\n func_starts.append(m.start())\\n \\n func_starts = sorted(list(set(func_starts)))\\n \\n # 3. 计算行号辅助逻辑\\n lines = clean_code.split('\\\\n')\\n line_starts = []\\n curr = 0\\n for line in lines:\\n line_starts.append(curr)\\n curr += len(line) + 1\\n \\n def get_line_num(index):\\n low = 0\\n high = len(line_starts) - 1\\n while low <= high:\\n mid = (low + high) // 2\\n if line_starts[mid] <= index:\\n if mid == len(line_starts) - 1 or line_starts[mid+1] > index:\\n return mid + 1\\n else:\\n low = mid + 1\\n else:\\n high = mid - 1\\n return 1\\n\\n # 4. 基于大括号配对计算每个函数的行数\\n for start_idx in func_starts:\\n open_brace_idx = -1\\n for i in range(start_idx, len(clean_code)):\\n if clean_code[i] == '{':\\n open_brace_idx = i\\n break\\n elif clean_code[i] in ';}':\\n break\\n \\n if open_brace_idx != -1:\\n depth = 0\\n close_brace_idx = -1\\n for i in range(open_brace_idx, len(clean_code)):\\n if clean_code[i] == '{':\\n depth += 1\\n elif clean_code[i] == '}':\\n depth -= 1\\n if depth == 0:\\n close_brace_idx = i\\n break\\n \\n # 如果匹配成功闭合，则校验总行数\\n if close_brace_idx != -1:\\n start_line = get_line_num(start_idx)\\n end_line = get_line_num(close_brace_idx)\\n if end_line - start_line + 1 > 12:\\n return False\\n return True\\n\\n # 对生成的项目文件进行函数行数检验\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_content = f.read()\\n except:\\n continue\\n \\n if ext == '.py':\\n if not check_python_functions(code_content):\\n return False\\n elif ext in ['.js', '.jsx', '.ts', '.tsx', '.vue', '.html', '.svelte', '.css']:\\n if not check_js_functions(code_content):\\n return False\\n else:\\n # 对于未标明扩展名但看似前端代码的文件同样做兜底检查\\n if 'function ' in code_content or '=>' in code_content:\\n if not check_js_functions(code_content):\\n return False\\n\\n # 对回复中的代码块进行兜底检查\\n code_blocks = re.findall(r'```(?:.*?)\\\\n(.*?)```', response, re.DOTALL)\\n for block in code_blocks:\\n if not check_python_functions(block):\\n return False\\n if not check_js_functions(block):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"针对你提供的全部代码文件，必须在第一行插入规定的编码声明：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据Web页面项目推测可能生成的前端及相关代码文件扩展名\\n code_extensions = {\\n '.html', '.js', '.css', '.ts', \\n '.jsx', '.tsx', '.vue', '.py', \\n '.mjs', '.cjs'\\n }\\n \\n code_files = []\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除通常不需要人为添加编码声明的依赖目录或版本控制目录\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n if '.git' in dirs:\\n dirs.remove('.git')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果项目中没有任何代码文件，则无法判定满足了要求\\n if not code_files:\\n return False\\n \\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行中是否包含规定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 文件读取发生异常（如编码错误或没有权限），视为不符合要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"每当测试完生成的代码之后，你的回复必须严格使用指定的XML格式来分项呈现执行结果与相关描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"你生成的代码文件中，所有的数字必须通过定义常量来体现，坚决禁止直接使用具体的数字\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"生成代码时，请在“=”号的左侧和右侧各放置两个空格\",\n \"validation_code\": \"def check_assignment_spaces(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 匹配单独的等号，并捕获其前后的连续空白字符\\n # (?<![<>=!\\\\+\\\\-\\\\*/%&\\\\|\\\\^~:\\\\?]) 排除了等号前面紧连的常见组合操作符\\n # (?![<>=!~]) 排除了等号后面紧连的操作符\\n pattern = re.compile(r'([ \\\\t]*)(?<![<>=!\\\\+\\\\-\\\\*/%&\\\\|\\\\^~:\\\\?])=(?![<>=!~])([ \\\\t]*)')\\n \\n code_extensions = {\\n '.py', '.js', '.ts', '.html', '.css', '.vue', '.java', \\n '.cpp', '.c', '.h', '.cs', '.php', '.rb', '.go', '.rs', \\n '.swift', '.kt', '.sh', '.bash', '.xml', '.yml', '.yaml'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 剔除 Python 的多行字符串\\n content = re.sub(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', '\\\"\\\"\\\"\\\"\\\"\\\"', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'''(.*?)'''\\\", \\\"''''''\\\", content, flags=re.DOTALL)\\n \\n # 剔除普通字符串和模板字符串（避免字符串里的 \\\"=\\\" 号引发误判）\\n content = re.sub(r'\\\"[^\\\"\\\\\\\\]*(\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"', '\\\"\\\"', content)\\n content = re.sub(r\\\"'[^'\\\\\\\\]*(\\\\\\\\.[^'\\\\\\\\]*)*'\\\", \\\"''\\\", content)\\n content = re.sub(r'`[^`\\\\\\\\]*(\\\\\\\\.[^`\\\\\\\\]*)*`', '``', content)\\n \\n # 剔除常见的块级/多行注释\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n lines = content.split('\\\\n')\\n for line in lines:\\n # 针对不同语言特性剔除单行注释\\n if ext in {'.py', '.rb', '.sh', '.bash', '.yaml', '.yml'}:\\n line = line.split('#')[0]\\n elif ext in {'.js', '.ts', '.cpp', '.c', '.h', '.cs', '.java', '.php', '.go', '.rs', '.swift', '.kt', '.vue'}:\\n line = line.split('//')[0]\\n elif ext in {'.lua', '.sql'}:\\n line = line.split('--')[0]\\n \\n # 寻找独立的等号\\n for match in pattern.finditer(line):\\n left_spaces = match.group(1)\\n right_spaces = match.group(2)\\n \\n # 检查等号两侧是否恰好为两个空格\\n if left_spaces != \\\" \\\" or right_spaces != \\\" \\\":\\n return False\\n\\n # 若未找到任何可扫描的代码文件，则无法证实该要求被满足\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看初始页面基础骨架：截屏检查页面是否包含顶部导航栏、左侧侧边栏，右侧主工作区。确认左侧栏包含'添加文件'按钮，主工作区正中间显示'请选择文件'的占位提示。","测试点击添加文件功能：点击左侧边栏的'添加文件'按钮，在弹出的文件选择器中一次性选择两个 '.md' 文件，并额外选择一个允许的 '.txt' 文件，确认这些文件均被加入并在左侧列表中清晰展示出文件名。","测试拖拽添加文件功能：用鼠标将一个允许的 '.md' 或 '.txt' 文件拖拽进入左侧侧边栏区域并松开，确认该文件被成功添加至左侧列表底部并展示文件名。","测试单个文件删除功能：在左侧文件列表中，点击任意一个文件名称旁边附带的'删除图标'，确认该对应文件项目从左侧列表中立刻消失移除。","测试清空列表功能：当左侧文件列表中存在至少两个文件时，滑动到底部点击'清空列表'按钮，确认左侧列表瞬间变为空白状态。","测试主工作区分屏布局展现：在左侧列表中点击任意一个已添加的 Markdown 文件，截屏确认右侧主工作区从占位提示变更为左右对半分屏的布局，左边显示所选文件的 Markdown 源码并允许编辑，右侧显示渲染排版的预览区。","测试Markdown实时渲染效果：在左侧源码输入框中输入包含标题、段落和列表的 Markdown 内容，观察右半边预览区，确认无需点击刷新按钮即可实时渲染为具有基础 HTML 样式的标题、段落和列表排版。","查看主题切换菜单展现：点击顶部导航栏中的主题切换下拉菜单，截屏确认展开的菜单中完整包含了'GitHub风格'、'暗黑模式'和'护眼模式'三个选项。","测试主题切换功能及样式生效：分别选择'GitHub风格'、'暗黑模式'和'护眼模式'，截屏观察右侧预览区，确认背景色、字体颜色、标题样式和引用块样式都会随所选主题实时变化。","查看右侧工作区按钮展现：截屏查看右侧主工作区的右上角位置，确认界面在此处展示了'导出当前HTML'和'导出当前PDF'两个操作按钮。","测试导出当前HTML功能：点击右上角的'导出当前HTML'按钮，确认浏览器触发下载动作，下载的 `.html` 格式文件在本地双击打开后，内容及主题样式排版与页面内的预览区保持一致。","测试导出当前PDF功能：点击右上角的'导出当前PDF'按钮，确认浏览器触发下载单页 PDF 文档的操作，且下载后的 PDF 成功保留了界面预览区当前看到的主题排版、文字和背景样式。","测试批量导出HTML功能：保证左侧列表中有至少两个文件，点击左下角的'批量导出 HTML'按钮，确认后台将列表中所有 Markdown 文件分别转换为带有当前主题样式的 HTML，并触发下载包含所有HTML文件的 `.zip` 压缩包。","测试批量导出PDF进度条展现：点击左边栏下方的'批量导出 PDF'按钮，截屏确认页面中间弹出一个带有进度条的模态框，且模态框内部明确显示类似'正在生成 PDF: 1/x'的进度更新文本。","测试批量导出PDF闭环逻辑：等待批量导出PDF的进度条跑到终点满后，确认进度模态框自动从页面消失，并且浏览器随即自动触发包含列表所有PDF的 `.zip` 压缩包下载操作。","查看导出设置面板展现：点击顶部导航栏的'设置'图标，截屏确认页面中间或侧边弹出设置面板，该面板内包含能够选择'页面大小'(A4/B5/Letter)、'纸张方向'(纵向/横向)以及输入修改'页边距大小'的表单项。","测试导出设置对PDF的生效控制：在设置面板中分别调整页面大小、纸张方向和页边距后，再次点击单个或批量导出 PDF，打开最新下载的 PDF 文件，确认页面尺寸、方向和页边距参数均按设置生效。","测试代码块高亮解析功能：在左侧源码输入框键入包含语言声明标志（如 ```javascript）的代码块，观察右侧预览区，并导出当前 HTML/PDF 文件，确认预览区和导出文件中的代码块均具有明显的语法高亮颜色。","测试表格与任务列表增强解析：在源码输入框以 Markdown 语法写入一个基础表格，再写入含有 `[x]` 和 `[ ]` 的任务列表，观察预览区并导出当前文件，确认预览区和导出文件均成功渲染出表格以及带勾选/空复选框的列表项。","测试浏览器缓存不丢失机制：在左侧添加一个文件并在源码区输入一些自定义内容后，使用浏览器功能强制刷新页面，刷新结束后确认左侧之前添加的文件列表仍然存在，点击该文件确认你之前输入的自定义内容完全还原，未丢失。","测试文件类型严格拦截功能：通过拖拽或文件选择器尝试添加普通图片等非 `.md`、非 `.txt` 文件，确认页面上方弹出红色 Toast 报错提示框，且该文件名称没有出现在左侧列表内。"],"string":"[\n \"查看初始页面基础骨架：截屏检查页面是否包含顶部导航栏、左侧侧边栏，右侧主工作区。确认左侧栏包含'添加文件'按钮，主工作区正中间显示'请选择文件'的占位提示。\",\n \"测试点击添加文件功能：点击左侧边栏的'添加文件'按钮，在弹出的文件选择器中一次性选择两个 '.md' 文件，并额外选择一个允许的 '.txt' 文件，确认这些文件均被加入并在左侧列表中清晰展示出文件名。\",\n \"测试拖拽添加文件功能：用鼠标将一个允许的 '.md' 或 '.txt' 文件拖拽进入左侧侧边栏区域并松开，确认该文件被成功添加至左侧列表底部并展示文件名。\",\n \"测试单个文件删除功能：在左侧文件列表中，点击任意一个文件名称旁边附带的'删除图标'，确认该对应文件项目从左侧列表中立刻消失移除。\",\n \"测试清空列表功能：当左侧文件列表中存在至少两个文件时，滑动到底部点击'清空列表'按钮，确认左侧列表瞬间变为空白状态。\",\n \"测试主工作区分屏布局展现：在左侧列表中点击任意一个已添加的 Markdown 文件，截屏确认右侧主工作区从占位提示变更为左右对半分屏的布局，左边显示所选文件的 Markdown 源码并允许编辑，右侧显示渲染排版的预览区。\",\n \"测试Markdown实时渲染效果：在左侧源码输入框中输入包含标题、段落和列表的 Markdown 内容，观察右半边预览区，确认无需点击刷新按钮即可实时渲染为具有基础 HTML 样式的标题、段落和列表排版。\",\n \"查看主题切换菜单展现：点击顶部导航栏中的主题切换下拉菜单，截屏确认展开的菜单中完整包含了'GitHub风格'、'暗黑模式'和'护眼模式'三个选项。\",\n \"测试主题切换功能及样式生效：分别选择'GitHub风格'、'暗黑模式'和'护眼模式'，截屏观察右侧预览区，确认背景色、字体颜色、标题样式和引用块样式都会随所选主题实时变化。\",\n \"查看右侧工作区按钮展现：截屏查看右侧主工作区的右上角位置，确认界面在此处展示了'导出当前HTML'和'导出当前PDF'两个操作按钮。\",\n \"测试导出当前HTML功能：点击右上角的'导出当前HTML'按钮，确认浏览器触发下载动作，下载的 `.html` 格式文件在本地双击打开后，内容及主题样式排版与页面内的预览区保持一致。\",\n \"测试导出当前PDF功能：点击右上角的'导出当前PDF'按钮，确认浏览器触发下载单页 PDF 文档的操作，且下载后的 PDF 成功保留了界面预览区当前看到的主题排版、文字和背景样式。\",\n \"测试批量导出HTML功能：保证左侧列表中有至少两个文件，点击左下角的'批量导出 HTML'按钮，确认后台将列表中所有 Markdown 文件分别转换为带有当前主题样式的 HTML，并触发下载包含所有HTML文件的 `.zip` 压缩包。\",\n \"测试批量导出PDF进度条展现：点击左边栏下方的'批量导出 PDF'按钮，截屏确认页面中间弹出一个带有进度条的模态框，且模态框内部明确显示类似'正在生成 PDF: 1/x'的进度更新文本。\",\n \"测试批量导出PDF闭环逻辑：等待批量导出PDF的进度条跑到终点满后，确认进度模态框自动从页面消失，并且浏览器随即自动触发包含列表所有PDF的 `.zip` 压缩包下载操作。\",\n \"查看导出设置面板展现：点击顶部导航栏的'设置'图标，截屏确认页面中间或侧边弹出设置面板，该面板内包含能够选择'页面大小'(A4/B5/Letter)、'纸张方向'(纵向/横向)以及输入修改'页边距大小'的表单项。\",\n \"测试导出设置对PDF的生效控制：在设置面板中分别调整页面大小、纸张方向和页边距后，再次点击单个或批量导出 PDF，打开最新下载的 PDF 文件，确认页面尺寸、方向和页边距参数均按设置生效。\",\n \"测试代码块高亮解析功能：在左侧源码输入框键入包含语言声明标志（如 ```javascript）的代码块，观察右侧预览区，并导出当前 HTML/PDF 文件，确认预览区和导出文件中的代码块均具有明显的语法高亮颜色。\",\n \"测试表格与任务列表增强解析：在源码输入框以 Markdown 语法写入一个基础表格，再写入含有 `[x]` 和 `[ ]` 的任务列表，观察预览区并导出当前文件，确认预览区和导出文件均成功渲染出表格以及带勾选/空复选框的列表项。\",\n \"测试浏览器缓存不丢失机制：在左侧添加一个文件并在源码区输入一些自定义内容后，使用浏览器功能强制刷新页面，刷新结束后确认左侧之前添加的文件列表仍然存在，点击该文件确认你之前输入的自定义内容完全还原，未丢失。\",\n \"测试文件类型严格拦截功能：通过拖拽或文件选择器尝试添加普通图片等非 `.md`、非 `.txt` 文件，确认页面上方弹出红色 Toast 报错提示框，且该文件名称没有出现在左侧列表内。\"\n]","isConversational":false}}},{"rowIdx":85,"cells":{"task_id":{"kind":"number","value":86,"string":"86"},"repository_policy":{"kind":"string","value":"以下是你在后续所有轮次中必须始终严格遵循的约束条件：\n1. 不论何时，你生成用户回复时都必须使用中文。\n2. 生成的任何一个代码文件中，都必须通篇使用中文作为注释语言。\n3. 你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径。\n4. 在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符。\n5. 凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md。\n6. 只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中。\n7. 请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行。\n8. 只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]}],"string":"[\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Machine Learning"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"后续所有操作都需要遵循以下新增规则，且规则长期有效：1. 处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令；2. 输出代码时，统一使用4个空格作为缩进标准。现在我要开发一个基于Python和PyQt5的OCR图片识别与标注工具。首先帮我搭建一个基础的GUI界面骨架。界面需要包含：顶部一个文本输入框（用于输入图片URL，请将默认文字设为“https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_002.png”），旁边有一个“加载图片”按钮；主区域分为左右两部分，左边是一个较大的图片显示区域（暂时用灰色背景占位，并显示文字“图片显示区”），右边是一个占据三分之一宽度的空白文本多行显示框。点击“加载图片”时，暂时只要在控制台打印出输入的URL即可，不需要真正的下载网络请求。请给出完整的Python代码。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时，统一使用4个空格作为缩进标准","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n import tokenize\n import io\n\n def is_valid_indentation(code: str) -> bool:\n # 1. 精确检查: 使用 tokenize 解析合法的 Python 缩进块\n try:\n tokens = tokenize.generate_tokens(io.StringIO(code).readline)\n for tok in tokens:\n if tok.type == tokenize.INDENT:\n indent_str = tok.string\n # 缩进必须只包含空格，不能有制表符等\n if not re.match(r'^ +$', indent_str):\n return False\n # 缩进空格长度必须严格是 4 的倍数\n if len(indent_str) % 4 != 0:\n return False\n except IndentationError:\n # 捕获 Python 自身检测到的缩进对齐错误，直接判定不满足\n return False\n except Exception:\n # 忽略 TokenError 等其他语法错误 (例如回复中的代码块不完整等情况)\n # 此时依靠已成功解析的 INDENT 检查，以及后续的逐行兜底检查\n pass\n \n # 2. 兜底检查: 逐行检查以避免有任何制表符 (Tab) 作为缩进混入\n for line in code.split('\\n'):\n # 匹配行首的空白字符\n m = re.match(r'^([ \\t]+)', line)\n if m:\n indent = m.group(1)\n # 严禁出现制表符\n if '\\t' in indent:\n return False\n \n return True\n\n codes_to_check = []\n \n # 从 response 中提取所有的 Markdown 代码块\n code_blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n codes_to_check.extend(code_blocks)\n \n # 获取 workspace_path 下生成的 Python 代码文件\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n codes_to_check.append(f.read())\n except Exception:\n pass\n \n # 根据用户本轮指令要求，必须至少产出一段界面骨架代码\n if not codes_to_check:\n return False\n \n # 对收集到的全部代码片段进行严格缩进校验\n for code in codes_to_check:\n if code.strip():\n if not is_valid_indentation(code):\n return False\n \n return True","tags":["Style","Layout"]}]},{"round_id":1,"instruction":"好，接下来完善图片下载和校验功能。当点击“加载图片”按钮时，程序需要发起请求下载文本框中的URL内容。注意核心要求：下载完成后，必须通过读取数据的二进制文件头（Magic Number）或者尝试使用图像库解码，来严谨判断下载的内容是否真的是一张图片文件，而不能仅仅依靠URL的后缀名。如果判断不是图片，弹出一个错误提示对话框；如果是有效图片，则将其自适应比例缩放，并显示在左侧的图片区域中。另外，后续所有操作长期遵循两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n from collections import Counter\n \n code_snippets = []\n \n # 1. 尝试从工作区路径中获取所有的 Python 代码文件内容\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n code_snippets.append(f.read())\n except Exception:\n pass\n \n # 2. 从 AI 的回复中提取所有的 Markdown 代码块\n for match in re.finditer(r'```(.*?)```', response, re.DOTALL):\n block = match.group(1)\n if '\\n' in block:\n first_line, rest = block.split('\\n', 1)\n # 过滤掉语言声明标志（如 python, py）\n if first_line.strip().isalpha():\n code_snippets.append(rest)\n else:\n code_snippets.append(block)\n else:\n code_snippets.append(block)\n \n if not code_snippets:\n return False\n \n indent_diffs = []\n all_indents = []\n \n # 3. 逐行分析所有提取到的代码片段，精准计算语法带来的缩进增量\n for code in code_snippets:\n lines = code.split('\\n')\n prev_indent = 0\n prev_line = \"\"\n for line in lines:\n stripped = line.strip()\n \n # 跳过空行和纯注释行，因为它们不会作为有效逻辑行影响缩进层级关联\n if not stripped or stripped.startswith('#'):\n continue\n \n # 计算当前行的前导空格数量\n match = re.match(r'^( +)', line)\n current_indent = len(match.group(1)) if match else 0\n \n if current_indent > 0:\n all_indents.append(current_indent)\n \n # 当上一逻辑有效行是以冒号结尾时，下一行的缩进理应递增，记录此时的增量\n if prev_line.endswith(':'):\n if current_indent > prev_indent:\n indent_diffs.append(current_indent - prev_indent)\n \n prev_line = stripped\n prev_indent = current_indent\n \n # 4. 判定规则：只要在基础语法块结构中出现 2 个空格的递增，即判定为助手的输出满足了替换缩进要求\n if indent_diffs:\n counter = Counter(indent_diffs)\n if counter.get(2, 0) > 0:\n return True\n return False\n \n # 回退判断（防止没有任何冒号结构的极端少代码情况）\n if all_indents:\n counter = Counter(all_indents)\n if counter.get(2, 0) > 0:\n return True\n \n return False","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n code_extensions = ['.py']\n code_files = []\n \n # 查找项目工作区中的所有代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n code_files.append(os.path.join(root, file))\n \n has_checked_code = False\n \n # 遍历验证工作区里的代码文件首行（顶行）\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 如果文件为空或者不包含特定声明，直接返回 False\n if not first_line or \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n has_checked_code = True\n except Exception:\n # 遇到编码问题导致无法读取的文件也视作不符合（要求为utf-8）\n return False\n \n # 如果工作区中存在实质生成的代码文件且验证全部通过，则直接返回 True\n if has_checked_code:\n return True\n \n # 兜底策略：如果工作区中没有发现代码文件，则检查 response 中的 Python 代码块\n python_blocks = re.findall(r'```python[ \\t]*\\n(.*?)```', response, re.DOTALL)\n for block in python_blocks:\n lines = block.split('\\n')\n # 获取第一行代码进行检验\n if not lines or \"-*- coding: utf-8 -*-\" not in lines[0]:\n return False\n has_checked_code = True\n \n # 如果找到了相应的代码文件或代码块并且验证通过，则返回 True，否则返回 False（未生成代码）\n return has_checked_code","tags":["Content","Format"]}]},{"round_id":2,"instruction":"现在来引入OCR核心功能。请在界面顶部的“加载图片”按钮旁边，新增一个“提取文字”按钮。当用户点击这个按钮时，请在后台使用 PaddleOCR（或类似的高精度中文OCR库）对当前已经加载显示的图片进行文字识别。将识别出来的所有文本内容按顺序提取出来，逐行追加显示在界面右侧的文本显示框中。为了防止界面卡死，请确保在处理OCR时界面不会假死。另外后续所有操作长期遵循两条规则：第一，输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求；第二，执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求","validation_code":"import os\nimport re\nimport tokenize\nfrom io import BytesIO\n\ndef verify(response: str, workspace_path: str) -> bool:\n def check_code_indent(code_str: str) -> bool:\n try:\n # 首选方案：尝试使用 tokenize 分析逻辑行的缩进，这是最严谨的判断方式\n tokens = list(tokenize.tokenize(BytesIO(code_str.encode('utf-8')).readline))\n for tok in tokens:\n if tok.type == tokenize.INDENT:\n indent_str = tok.string\n # 必须全部是空格，不包含制表符等\n if not all(c == ' ' for c in indent_str):\n return False\n # 缩进长度必须是 4 的倍数\n if len(indent_str) % 4 != 0:\n return False\n return True\n except Exception:\n # 降级方案：针对不完整代码片段、diff patch 等引起 tokenize 解析异常的情况\n lines = code_str.split('\\n')\n prev_colon_indent = None\n for line in lines:\n if not line.strip():\n continue\n \n leading_spaces = len(line) - len(line.lstrip(' '))\n if '\\t' in line[:leading_spaces]:\n return False\n \n is_comment = line.lstrip().startswith('#')\n \n if prev_colon_indent is not None:\n if not is_comment:\n # 遇到冒号后的非空非注释行，如果产生新的缩进级别，应当严格等于之前冒号行缩进 + 4 个空格\n if leading_spaces > prev_colon_indent:\n if leading_spaces != prev_colon_indent + 4:\n return False\n prev_colon_indent = None\n \n if not is_comment:\n stripped = line.strip()\n # 检查典型代码块关键字开头，其缩进长度需要严格满足 4 的倍数要求\n if stripped.startswith(('def ', 'class ', 'if ', 'elif ', 'else:', 'for ', 'while ', \n 'try:', 'except', 'finally:', 'with ', 'return ', 'import ', \n 'from ', 'print(', 'self.', 'pass', 'continue', 'break')):\n if leading_spaces > 0 and leading_spaces % 4 != 0:\n return False\n \n # 剥离行内注释后，寻找开启新缩进块的冒号\n code_part = line.split('#')[0].rstrip()\n if code_part.endswith(':'):\n prev_colon_indent = leading_spaces\n return True\n\n # 1. 检查 workspace_path 下生成的 Python 代码文件和 Markdown 文档里的代码块\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.py'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_code_indent(code):\n return False\n except Exception:\n pass\n elif file.endswith('.md'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n md_content = f.read()\n md_blocks = re.findall(r'```(?:python|py)\\s*\\n(.*?)\\n```', md_content, re.DOTALL | re.IGNORECASE)\n for block in md_blocks:\n if not check_code_indent(block):\n return False\n except Exception:\n pass\n \n # 2. 检查最后一轮 response 中输出的 Python 代码块\n python_blocks = re.findall(r'```(?:python|py)\\s*\\n(.*?)\\n```', response, re.DOTALL | re.IGNORECASE)\n for block in python_blocks:\n if not check_code_indent(block):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n has_code_file = False\n \n # 遍历项目地址下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令背景，推断主要代码文件类型为 Python (.py)\n if file.endswith('.py'):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取代码文件顶行（第一行）\n first_line = f.readline()\n \n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果读取文件失败（例如编码不正确等），则视为不满足要求\n return False\n \n # 如果没找到任何代码文件，则要求未满足；否则满足了所有代码文件都具备该声明\n return has_code_file","tags":["Content","Format"]},{"constraint":"执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]}]},{"round_id":3,"instruction":"很好。下一步是实现空间位置的标注功能。我希望在识别出文字的同时，能够直观地看到它们在图中的具体位置。请修改“提取文字”后的处理逻辑：不仅在右侧输出文字，还要使用OpenCV或PIL在原始图片上，用红色的矩形框（BBox）将所有识别出的文字区域框选出来。画好红色标注框后，将这张带有标注的图片更新显示到左侧的图片区域中。另外后续所有操作长期遵循两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，你输出的代码中，禁止出现超过3层的循环嵌套。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import re\n import os\n import io\n import tokenize\n\n code_snippets = []\n\n # 1. 提取 response 中以 Markdown 格式输出的代码片段\n # 匹配形如 ```lang ... ``` 的代码块\n for match in re.finditer(r'```([a-zA-Z]*)\\n(.*?)```', response, re.DOTALL):\n lang = match.group(1).lower()\n # 排除纯文本和一般文档（由于要求是“输出代码时”），对疑似代码的数据统统收集\n if lang not in ('md', 'markdown', 'txt', 'text'):\n code_snippets.append(match.group(2))\n\n # 2. 提取 workspace_path 中实际保存的代码文件\n if workspace_path and os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 结合用户指令推断主要为 Python 代码，并囊括少量可能存在的辅助脚本配置\n if file.endswith(('.py', '.sh', '.json')):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n code_snippets.append(f.read())\n except Exception:\n pass\n\n # 如果没有任何代码片段，则未触发检查，默认满足规则\n if not code_snippets:\n return True\n\n def check_snippet(code_str: str) -> bool:\n # 为了防止缺少末尾换行符导致 tokenize 解析抛出异常，末尾补齐换行符\n code_str = code_str + '\\n'\n \n # 尝试使用 python 原生的 tokenize 进行逻辑缩进检测\n try:\n tokens = tokenize.tokenize(io.BytesIO(code_str.encode('utf-8')).readline)\n indent_levels = [0]\n for tok in tokens:\n if tok.type == tokenize.INDENT:\n # 不允许出现制表符缩进\n if '\\t' in tok.string:\n return False\n indent_len = len(tok.string)\n step = indent_len - indent_levels[-1]\n # 严格的2空格缩进要求，每次逻辑缩进增加的步长必须为2\n if step != 2:\n return False\n indent_levels.append(indent_len)\n elif tok.type == tokenize.DEDENT:\n if len(indent_levels) > 1:\n indent_levels.pop()\n return True\n \n except Exception:\n # 降级验证策略：针对非 Python 代码或因不完整而报错的代码，采用正则按行验证\n lines = code_str.split('\\n')\n indents = []\n for line in lines:\n stripped = line.strip()\n # 忽略空行和常见注释行\n if not stripped or stripped.startswith('#') or stripped.startswith('//'):\n continue\n m = re.match(r'^( +)\\S', line)\n if m:\n indents.append(len(m.group(1)))\n\n # 如果不存在任何缩进，则不视为违规\n if not indents:\n return True\n\n # 如果所有缩进均满足是4的倍数，且存在有效的缩进，那么代码极有可能仍然使用的是4空格缩进，不符合要求\n if all(ind % 4 == 0 for ind in indents) and any(ind > 0 for ind in indents):\n return False\n\n return True\n\n # 3. 遍历检查搜集到的所有代码片段，如果存在未通过检测的代码，则返回 False\n for snippet in code_snippets:\n if not snippet.strip():\n continue\n if not check_snippet(snippet):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉隐藏目录和常见的环境/缓存目录，确保只检查助手生成的代码文件\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['venv', 'env', '__pycache__']]\n \n for file_name in files:\n # 推测此项目的代码文件主要为 Python 脚本文件\n if file_name.endswith('.py'):\n found_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果文件不可读或读取过程中发生异常，判定为不满足要求\n return False\n \n # 确保至少检查过一个生成的代码文件，若是则表示全部满足要求\n return found_code_file","tags":["Content","Format"]},{"constraint":"执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码中，禁止出现超过3层的循环嵌套","validation_code":"","tags":["Quantity","Complexity"]}]},{"round_id":4,"instruction":"现在来做一个交互增强功能。请将右侧的多行文本框替换为一个列表组件（如 QListWidget）。每次OCR识别后，将每一句识别出的文字作为列表的一项填入。当用户在右侧列表中鼠标点击选中某一行文字时，左侧图片上对应的那个红色标注框需要变成绿色并且线条加粗，而其他的框保持红色不变。这样可以方便用户核对具体的文字对应原图上的哪个位置。另外，后续所有操作长期遵循以下两条规则：第一，你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求；第二，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n import tokenize\n import io\n\n indent_sizes = []\n\n def process_code(code_str: str) -> bool:\n \"\"\"\n 利用 tokenize 解析代码片段，提取每一次发生逻辑缩进时的增量大小。\n 如果包含了 Tab 缩进则直接返回 False，否则返回 True。\n \"\"\"\n try:\n # 补齐尾部换行符以防部分代码因缺少换行导致 EOF TokenError\n code_str = code_str.strip() + '\\n'\n tokens = tokenize.generate_tokens(io.StringIO(code_str).readline)\n indent_stack = [0]\n \n for tok in tokens:\n if tok.type == tokenize.INDENT:\n s = tok.string\n if '\\t' in s:\n return False\n \n curr_indent = len(s)\n diff = curr_indent - indent_stack[-1]\n if diff > 0:\n indent_sizes.append(diff)\n indent_stack.append(curr_indent)\n elif tok.type == tokenize.DEDENT:\n if len(indent_stack) > 1:\n indent_stack.pop()\n except Exception:\n # 忽略因解析不完整的代码段抛出的 TokenError 等异常\n pass\n return True\n\n # 1. 处理 response 中的 Python 代码块\n # 匹配诸如 ```python, ```py, 或者纯 ``` 包裹的代码块\n code_blocks = re.findall(r'```(?:python|py)?[ \\t]*\\n(.*?)\\n```', response, re.DOTALL)\n for code in code_blocks:\n if not process_code(code):\n return False\n\n # 2. 处理 workspace_path 下生成的 .py 文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for f in files:\n if f.endswith('.py'):\n file_path = os.path.join(root, f)\n try:\n with open(file_path, 'r', encoding='utf-8') as file:\n content = file.read()\n if not process_code(content):\n return False\n except Exception:\n pass\n\n # 如果未能提取到任何缩进记录，意味着没有涉及任何层级的缩进\n # 在没有提供不符合要求证据的情况下，判定为没有违背规则\n if not indent_sizes:\n return True\n\n # 3. 统计并分析收集到的缩进增量情况\n count_2 = sum(1 for d in indent_sizes if d == 2)\n count_4 = sum(1 for d in indent_sizes if d == 4)\n\n # 如果4个空格的缩进占了主导位置（数量多于2个空格），说明主要使用了4空格缩进\n if count_4 > count_2:\n return False\n \n # 如果完全没有2个空格的逻辑缩进，但在记录中又存在其他的增量缩进（如3, 4, 8等）\n if count_2 == 0 and len(indent_sizes) > 0:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n has_code_file = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令，该项目主要生成的代码文件为 Python 文件\n if file.endswith('.py'):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n # 读取顶行（第一行）\n first_line = f.readline()\n \n # 判断顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果无法读取文件，视作不满足要求\n return False\n \n # 如果没有找到任何代码文件，则无法证明满足了要求；反之如果全部检查通过，则返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":5,"instruction":"非常棒，接下来实现本地持久化保存功能。在界面底部增加一个“保存结果”按钮。点击后，打开系统的保存文件对话框，让用户选择保存目录和主文件名（如 result）。程序需要做两件事：1. 将当前界面上带有所有标注框的图片保存为 `[文件名].png`；2. 把识别出的所有数据保存为一个 `[文件名].json` 文件，JSON数据需包含每个文本片段的内容、对应的四点坐标边界以及识别置信度。另外后续所有操作长期遵循以下两条规则：第一，输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求；第二，运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求","validation_code":"import os\nimport re\nimport tokenize\nimport io\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n python_codes = []\n \n # 1. 从工作区路径中提取所有 Python 文件的代码\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8', errors='ignore') as f:\n python_codes.append(f.read())\n except Exception:\n pass\n \n # 2. 从回复内容中提取所有的 Python 代码块 (支持 python/python3/py)\n python_blocks = re.findall(r'```(?:python|python3|py)\\n(.*?)```', response, re.DOTALL)\n python_codes.extend(python_blocks)\n \n # 如果没有任何代码，无法验证其是否满足“输出代码时”的前提，视为不满足\n if not python_codes:\n return False\n \n has_code = False\n \n # 3. 验证所有的 Python 代码缩进\n for code in python_codes:\n if not code.strip():\n continue\n has_code = True\n \n try:\n token_generator = tokenize.generate_tokens(io.StringIO(code).readline)\n while True:\n try:\n tok = next(token_generator)\n # 仅在产生新的代码块缩进时，tokenizer 才会生成 INDENT token\n if tok.type == tokenize.INDENT:\n indent_str = tok.string\n # 缩进不能包含制表符\n if '\\t' in indent_str:\n return False\n # 空格数量必须是 4 的整数倍 (拒绝2空格或其他非标准缩进)\n if len(indent_str) % 4 != 0:\n return False\n except StopIteration:\n break\n # 捕获 TokenError, IndentationError 等因 AI 生成了片段化或不完整代码引发的错误\n # 停止当前代码的解析，直接跳出处理下一个代码块\n except (tokenize.TokenError, IndentationError, SyntaxError):\n break\n except Exception:\n pass\n\n return has_code","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据项目需求，主要排查 Python 代码文件\n if file.endswith('.py'):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 若文件无法正常打开或读取（如编码等异常），则视为不满足条件\n return False\n \n # 如果遍历完成没有遇到不符合规则的文件，并且确实找到了代码文件，则返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":6,"instruction":"我们继续扩展。增加一个基于置信度的动态过滤功能。在右侧文字列表的上方增加一个滑块（Slider）和一个Label，滑块范围代表 0.00 到 1.00 的置信度（保留两位小数），默认值为0.00。当用户拖动滑块时，界面需要实时动态更新：右侧列表中自动隐藏掉置信度低于滑块设定值的文字项，同时左侧图片上也自动消失对应低置信度的矩形标注框（不需要重新调用OCR模型，直接操作当前的识别结果数据即可）。另外后续所有操作长期遵循两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求","validation_code":"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n import tokenize\n import io\n\n def evaluate_code(code_str, is_python):\n diffs_2 = 0\n diffs_4 = 0\n has_tab = False\n \n parsed = False\n if is_python:\n try:\n # 针对 Python 代码，使用 tokenize 获取准确的缩进层次，忽略悬挂对齐等\n tokens = tokenize.generate_tokens(io.StringIO(code_str).readline)\n levels = [0]\n for tok in tokens:\n if tok.type == tokenize.INDENT:\n if '\\t' in tok.string:\n has_tab = True\n curr = len(tok.string)\n diff = curr - levels[-1]\n if diff == 2:\n diffs_2 += 1\n elif diff == 4:\n diffs_4 += 1\n levels.append(curr)\n elif tok.type == tokenize.DEDENT:\n if len(levels) > 1:\n levels.pop()\n parsed = True\n except Exception:\n pass\n \n # 非Python代码或Python代码解析失败，回退到基于文本行的前导空格差值计算\n if not parsed:\n lines = code_str.split('\\n')\n levels = []\n for line in lines:\n # 过滤空行以及单行注释行，减少对缩进统计的干扰\n if line.strip() == '' or line.lstrip().startswith('#') or line.lstrip().startswith('//'):\n continue\n leading = len(line) - len(line.lstrip(' '))\n if '\\t' in line[:leading]:\n has_tab = True\n levels.append(leading)\n \n for i in range(1, len(levels)):\n diff = levels[i] - levels[i-1]\n if diff == 2:\n diffs_2 += 1\n elif diff == 4:\n diffs_4 += 1\n \n return diffs_2, diffs_4, has_tab\n\n # 1. 优先检查 response 中直接包含的 Markdown 代码块\n code_blocks = re.findall(r'```(\\w*)\\n(.*?)```', response, re.DOTALL)\n response_has_2 = False\n response_has_code = False\n \n for lang, code in code_blocks:\n is_py = lang.lower() in ['python', 'py', '']\n d2, d4, ht = evaluate_code(code, is_py)\n if ht:\n return False # 如果生成了包含 tab 的代码，必然不符合2个空格的要求\n \n if d2 > 0 or d4 > 0:\n response_has_code = True\n if d4 > d2:\n # 如果某个代码块明显以 4 个空格作为主力缩进，说明没有统一遵守新指令\n return False\n if d2 >= d4 and d2 > 0:\n response_has_2 = True\n\n if response_has_2:\n return True\n \n # 2. 如果 response 中没有足够的代码体现缩进规则，去检查工作区文件（涵盖通过沙箱工具进行的操作）\n total_d2 = 0\n total_d4 = 0\n if os.path.isdir(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1]\n if ext in ['.py', '.json', '.js', '.html', '.css', '.vue', '.ts', '.tsx']:\n is_py = (ext == '.py')\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n code = f.read()\n d2, d4, ht = evaluate_code(code, is_py)\n if ht:\n continue # 在历史或第三方文件中如果存在tab，忽略之以防止误伤\n total_d2 += d2\n total_d4 += d4\n except Exception:\n pass\n \n # 3. 如果没找到任何带有有效缩进的代码块和文件，默认认为未违背\n if not response_has_code and (total_d2 == 0 and total_d4 == 0):\n return True \n\n # 在所有代码文件的综合数据中，要求 2 空格缩进占主体地位，取代原来的 4 空格\n return total_d2 >= total_d4 and total_d2 > 0","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n has_checked_any = False\n \n # 优先检查工作目录中实际生成的代码文件\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 排除常见的无关隐藏文件夹及虚拟环境文件夹，避免误判\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('venv', 'env', '__pycache__', 'node_modules')]\n for file in files:\n if file.endswith('.py'):\n has_checked_any = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n lines = f.readlines()\n if not lines:\n return False # 代码文件为空，不满足“包含声明”的条件\n first_line = lines[0].strip()\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n return False\n \n # 如果在文件系统中成功找到了对应的代码文件，并全部通过了上述要求，则验证成功\n if has_checked_any:\n return True\n \n # 后备方案：如果工作目录中并未找到 .py 文件，检查回复中是否包含 Python 代码块\n code_blocks = re.findall(r'```(?:python|py|Python)[ \\t]*\\n(.*?)```', response, re.DOTALL)\n if code_blocks:\n has_checked_any = True\n for block in code_blocks:\n lines = block.split('\\n')\n if not lines:\n return False\n first_line = lines[0].strip()\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n \n # 只要进行了有效的检查且未在上面拦截返回False，则认为通过验证\n if has_checked_any:\n return True\n \n return False","tags":["Content","Format"]},{"constraint":"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n def check_equals_spacing(code: str) -> bool:\n # 1. 移除多行字符串 (匹配 '''...''' 和 \"\"\"...\"\"\")\n code = re.sub(r\"'''[\\s\\S]*?'''\", '', code)\n code = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '', code)\n \n # 2. 移除单行字符串 (包含对转义符号 \\ 的处理，防止引号被意外转义)\n code = re.sub(r\"'([^'\\\\]|\\\\.)*'\", '', code)\n code = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '', code)\n \n # 3. 移除注释\n code = re.sub(r'#.*', '', code)\n \n # 4. 检查纯代码逻辑中，= 的前面或后面是否含有空格或制表符\n if re.search(r'[ \\t]=|=[ \\t]', code):\n return False\n return True\n\n # 步骤 A：检查最后一轮回复中包含的 Python 代码块\n blocks = re.findall(r'```(?:python|py)\\s*\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\n for block in blocks:\n if not check_equals_spacing(block):\n return False\n \n # 步骤 B：检查项目工作区中生成的所有 Python 代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 若发现文件中存在等号前后带有空格的现象，直接判定为未满足要求\n if not check_equals_spacing(content):\n return False\n except Exception:\n # 忽略无法读取的异常文件\n pass\n \n return True","tags":["Style","Layout"]}]},{"round_id":7,"instruction":"下一步，增加本地缓存以提升体验。为了防止用户对同一个URL反复点击“提取文字”造成资源的浪费，请增加一个内存缓存机制（如使用字典）。以图片的哈希值（或URL本身）作为Key，缓存它的完整OCR识别结果。如果用户输入相同的URL并点击提取，先检查缓存，如果有，则直接从缓存读取数据渲染文字列表和标注框，并在控制台打印“命中缓存，跳过模型推理”。另外后续所有操作长期遵循以下两条规则：第一，输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求；第二，在整个回复过程中，任何文件路径或代码位置都必须始终仅以双引号将其包裹起来。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n import os\n from collections import Counter\n \n def analyze_snippet(snippet: str):\n diffs = []\n indents = []\n lines = snippet.split('\\n')\n prev_indent = 0\n \n for line in lines:\n if not line.strip():\n continue\n # 忽略纯注释行，防止注释的不规则缩进影响判断\n if line.lstrip().startswith('#'):\n continue\n \n match = re.match(r'^([ \\t]*)', line)\n leading = match.group(1) if match else \"\"\n \n # 为计算缩进长度，将制表符视作4个空格\n curr_indent = len(leading.replace('\\t', ' '))\n indents.append(curr_indent)\n \n # 记录缩进的增量\n if curr_indent > prev_indent:\n diffs.append(curr_indent - prev_indent)\n prev_indent = curr_indent\n \n # 如果没有产生缩进增加（例如只有单行代码或全局变量平铺）\n if not diffs:\n valid_indents = [i for i in indents if i > 0]\n if not valid_indents:\n return None # 无法判断该片段的缩进类型\n c = Counter(valid_indents)\n most_common = c.most_common(1)[0][0]\n # 检查主要存在的绝对缩进是否是4的倍数\n return most_common % 4 == 0\n \n c = Counter(diffs)\n count_4 = c.get(4, 0)\n count_2 = c.get(2, 0)\n \n # 核心判断逻辑，通过最直观的缩进步长对比\n if count_4 > count_2:\n return True\n elif count_2 > count_4:\n return False\n else:\n if count_4 == 0:\n most_common = c.most_common(1)[0][0]\n return most_common % 4 == 0\n else:\n return True\n\n # 1. 尝试从 response 提取 Python 或无高亮类型的代码块\n pattern = re.compile(r'```(?:python|py)?[ \\t]*\\n(.*?)```', re.DOTALL | re.IGNORECASE)\n matches = pattern.findall(response)\n \n response_results = []\n for snippet in matches:\n res = analyze_snippet(snippet)\n if res is not None:\n response_results.append(res)\n \n # 如果回复中能找到任何应用了 4 空格的代码块，判定为遵循要求\n if any(response_results):\n return True\n \n # 2. 如果 response 无法判定，读取工作区所有的 .py 文件进行检查\n file_results = []\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n snippet = f.read()\n res = analyze_snippet(snippet)\n if res is not None:\n file_results.append(res)\n except Exception:\n pass\n \n # 只要存在被修改或新生成的文件体现了 4 空格，就判定为符合要求\n if any(file_results):\n return True\n \n # 只有当包含可以判定的代码，且所有代码段皆违背4空格缩进要求时，才判定为 False\n if response_results or file_results:\n return False\n \n return True","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n has_py_files = False\n\n # 1. 检查 workspace_path 下生成的所有 .py 文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 根据任务背景，推断代码文件类型为 Python 脚本\n if file_name.endswith('.py'):\n has_py_files = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 验证顶行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 无法正常打开/读取的代码文件安全起见视为不满足要求\n return False\n\n # 若在工作区找到了生成的 .py 代码文件，且均通过了上方检查，则说明满足要求\n if has_py_files:\n return True\n\n # 2. 作为后备方案：如果未在工作区找到本地 .py 文件，则检查回复内容里的 python 代码块\n code_blocks = re.findall(r'```(?:python|py)[ \\t]*\\n(.*?)\\n[ \\t]*```', response, re.DOTALL | re.IGNORECASE)\n for block in code_blocks:\n lines = block.split('\\n')\n if lines:\n first_line = lines[0]\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格","validation_code":"def verify_no_space_around_equals(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n code_contents = []\n\n # 1. 提取 response 中以 Markdown 代码块形式返回的代码\n blocks = re.findall(r'```[^\\n]*\\n(.*?)```', response, re.DOTALL)\n if blocks:\n code_contents.extend(blocks)\n\n # 2. 遍历 workspace_path 中生成的 .py 代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n try:\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\n code_contents.append(f.read())\n except Exception:\n pass\n \n # 若没有检测到任何代码，则默认没有违规行为\n if not code_contents:\n return True\n \n for code in code_contents:\n # 3. 过滤掉注释，避免注释内存在的“= ”造成误判\n code = re.sub(r'#.*', '', code)\n \n # 4. 过滤掉多行和单行字符串字面量，防止字符串内部的等于号文本造成误判\n code = re.sub(r'\"\"\"[\\s\\S]*?\"\"\"', '\"\"', code)\n code = re.sub(r\"'''[\\s\\S]*?'''\", \"''\", code)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", code)\n \n # 5. 检查“=”前后是否存有空格或水平制表符\n if re.search(r'[ \\t]=|=[ \\t]', code):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"在整个回复过程中，任何文件路径或代码位置都必须始终仅以双引号将其包裹起来","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":8,"instruction":"最后一轮，增加一个数据脱敏场景下的高级图片处理功能。在底部再增加一个“文字涂抹遮盖”按钮。点击后，程序要利用当前保存的识别坐标框，将原图中所包含文字的矩形区域全部用纯黑色（或图片均值颜色）填充涂抹掉，生成一张“脱敏后的无字图片”，并将这张图片展示在左侧区域。右侧的文本列表同时清空，此时滑块交互暂停生效。另外后续所有操作长期遵循以下两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份。","instruction_following_checklist":[{"constraint":"不论何时，你生成用户回复时都必须使用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言","validation_code":"","tags":["Language","Comment Language"]},{"constraint":"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符","validation_code":"","tags":["Environment","File Path"]},{"constraint":"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行","validation_code":"","tags":["Environment","Logging"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n from collections import Counter\n\n def is_2_space_indent(content: str) -> bool:\n lines = content.split('\\n')\n # 简单检查是否有制表符作为缩进，有则直接判违规\n for line in lines:\n if line.startswith('\\t') or line.startswith(' \\t'):\n return False\n \n indents = []\n for line in lines:\n stripped = line.lstrip(' ')\n # 过滤空行及常见语言的纯注释行，防止干扰判断\n if not stripped or stripped.startswith('#') or stripped.startswith('//') or stripped.startswith('\"\"\"') or stripped.startswith(\"'''\"):\n continue\n indent = len(line) - len(stripped)\n indents.append(indent)\n \n diffs = []\n # 计算相邻行的缩进增量\n for i in range(1, len(indents)):\n diff = indents[i] - indents[i-1]\n if diff > 0:\n diffs.append(diff)\n \n if not diffs:\n return True\n \n counts = Counter(diffs)\n # 如果完全没有 2 空格的递增，但有 4 或 8 空格的递增，说明肯定不是 2 空格缩进\n if counts.get(2, 0) == 0:\n if counts.get(4, 0) > 0 or counts.get(8, 0) > 0:\n return False\n \n return True\n\n code_contents = []\n\n # 1. 检查工作区中的代码文件\n if os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉常见的隐藏目录或依赖环境目录\n dirs[:] = [d for d in dirs if d not in ('venv', '.git', 'node_modules', '__pycache__', '.idea', '.vscode', '.env')]\n for file in files:\n if file.endswith(('.py', '.json', '.js', '.html', '.css', '.cpp', '.c', '.java', '.sh')):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code_contents.append(f.read())\n except Exception:\n pass\n\n # 2. 检查助手回复中的代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9_-]*\\n(.*?)```', response, re.DOTALL)\n code_contents.extend(code_blocks)\n\n if not code_contents:\n return True\n\n # 3. 验证所有的代码片段是否都符合 2 空格缩进标准\n for content in code_contents:\n if not content.strip():\n continue\n if not is_2_space_indent(content):\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查所有生成的Python代码文件顶行是否包含特定的编码声明 -*- coding: utf-8 -*-\n \"\"\"\n py_files_found = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令推测，主要代码文件为 Python 文件\n if file.endswith('.py'):\n py_files_found = True\n file_path = os.path.join(root, file)\n try:\n # 尝试打开并读取文件第一行\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline().strip()\n # 检查顶行是否包含特定的编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception as e:\n # 如果文件无法读取，则视为无法满足要求\n return False\n \n # 如果找到了至少一个.py文件并且都满足要求，则返回True\n # 如果没有找到任何.py文件，则无法证明满足了要求，返回False\n return py_files_found","tags":["Content","Format"]},{"constraint":"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import io\n import re\n import tokenize\n \n if not os.path.exists(workspace_path):\n return False\n \n py_files = []\n # 查找工作区下所有的 .py 代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith('.py'):\n py_files.append(os.path.join(root, file))\n \n # 如果没有生成任何 Python 文件，不满足基本前提\n if not py_files:\n return False\n \n has_checked_any = False\n \n for file_path in py_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n has_checked_any = True\n \n try:\n # 方案 A: 使用 tokenize 模块分析 Python 语法，安全过滤字符串和注释中的 '='\n tokens = list(tokenize.tokenize(io.BytesIO(content.encode('utf-8')).readline))\n for token in tokens:\n if token.type == tokenize.OP and token.string == '=':\n line = token.line\n start_col = token.start[1]\n end_col = token.end[1]\n \n # 检查 '=' 前一个字符是否为空格或制表符\n if start_col > 0:\n prev_char = line[start_col - 1]\n if prev_char in (' ', '\\t'):\n return False\n \n # 检查 '=' 后一个字符是否为空格或制表符\n if end_col < len(line):\n next_char = line[end_col]\n if next_char in (' ', '\\t'):\n return False\n \n except Exception:\n # 方案 B: Tokenize 失败（可能存在语法错误）时的降级处理方案\n lines = content.split('\\n')\n for line in lines:\n # 简单去除行内注释\n line = line.split('#')[0]\n # 使用正则简单去除字符串常量\n line = re.sub(r'([\"\\'])(?:(?=(\\\\?))\\2.)*?\\1', '', line)\n \n for i, char in enumerate(line):\n if char == '=':\n # 排除 ==, !=, >=, <=, +=, -= 等复合赋值/比较运算符的干扰\n if i > 0 and line[i-1] in '!<>=+-*/%&|^':\n continue\n if i < len(line) - 1 and line[i+1] == '=':\n continue\n \n # 检查独立的 '=' 前后是否有空格\n if (i > 0 and line[i-1] in ' \\t') or (i < len(line) - 1 and line[i+1] in ' \\t'):\n return False\n\n # 若所有已读取的代码文件均通过检查，则满足要求\n return has_checked_any","tags":["Style","Layout"]},{"constraint":"在整个回复过程中，任何文件路径或代码位置都必须始终仅以双引号将其包裹起来","validation_code":"","tags":["Environment","File Path"]},{"constraint":"只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份","validation_code":"","tags":["Environment","File Operation"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"后续所有操作都需要遵循以下新增规则，且规则长期有效：1. 处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令；2. 输出代码时，统一使用4个空格作为缩进标准。现在我要开发一个基于Python和PyQt5的OCR图片识别与标注工具。首先帮我搭建一个基础的GUI界面骨架。界面需要包含：顶部一个文本输入框（用于输入图片URL，请将默认文字设为“https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_002.png”），旁边有一个“加载图片”按钮；主区域分为左右两部分，左边是一个较大的图片显示区域（暂时用灰色背景占位，并显示文字“图片显示区”），右边是一个占据三分之一宽度的空白文本多行显示框。点击“加载图片”时，暂时只要在控制台打印出输入的URL即可，不需要真正的下载网络请求。请给出完整的Python代码。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时，统一使用4个空格作为缩进标准\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import tokenize\\n import io\\n\\n def is_valid_indentation(code: str) -> bool:\\n # 1. 精确检查: 使用 tokenize 解析合法的 Python 缩进块\\n try:\\n tokens = tokenize.generate_tokens(io.StringIO(code).readline)\\n for tok in tokens:\\n if tok.type == tokenize.INDENT:\\n indent_str = tok.string\\n # 缩进必须只包含空格，不能有制表符等\\n if not re.match(r'^ +$', indent_str):\\n return False\\n # 缩进空格长度必须严格是 4 的倍数\\n if len(indent_str) % 4 != 0:\\n return False\\n except IndentationError:\\n # 捕获 Python 自身检测到的缩进对齐错误，直接判定不满足\\n return False\\n except Exception:\\n # 忽略 TokenError 等其他语法错误 (例如回复中的代码块不完整等情况)\\n # 此时依靠已成功解析的 INDENT 检查，以及后续的逐行兜底检查\\n pass\\n \\n # 2. 兜底检查: 逐行检查以避免有任何制表符 (Tab) 作为缩进混入\\n for line in code.split('\\\\n'):\\n # 匹配行首的空白字符\\n m = re.match(r'^([ \\\\t]+)', line)\\n if m:\\n indent = m.group(1)\\n # 严禁出现制表符\\n if '\\\\t' in indent:\\n return False\\n \\n return True\\n\\n codes_to_check = []\\n \\n # 从 response 中提取所有的 Markdown 代码块\\n code_blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n codes_to_check.extend(code_blocks)\\n \\n # 获取 workspace_path 下生成的 Python 代码文件\\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n codes_to_check.append(f.read())\\n except Exception:\\n pass\\n \\n # 根据用户本轮指令要求，必须至少产出一段界面骨架代码\\n if not codes_to_check:\\n return False\\n \\n # 对收集到的全部代码片段进行严格缩进校验\\n for code in codes_to_check:\\n if code.strip():\\n if not is_valid_indentation(code):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"好，接下来完善图片下载和校验功能。当点击“加载图片”按钮时，程序需要发起请求下载文本框中的URL内容。注意核心要求：下载完成后，必须通过读取数据的二进制文件头（Magic Number）或者尝试使用图像库解码，来严谨判断下载的内容是否真的是一张图片文件，而不能仅仅依靠URL的后缀名。如果判断不是图片，弹出一个错误提示对话框；如果是有效图片，则将其自适应比例缩放，并显示在左侧的图片区域中。另外，后续所有操作长期遵循两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n from collections import Counter\\n \\n code_snippets = []\\n \\n # 1. 尝试从工作区路径中获取所有的 Python 代码文件内容\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n code_snippets.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 从 AI 的回复中提取所有的 Markdown 代码块\\n for match in re.finditer(r'```(.*?)```', response, re.DOTALL):\\n block = match.group(1)\\n if '\\\\n' in block:\\n first_line, rest = block.split('\\\\n', 1)\\n # 过滤掉语言声明标志（如 python, py）\\n if first_line.strip().isalpha():\\n code_snippets.append(rest)\\n else:\\n code_snippets.append(block)\\n else:\\n code_snippets.append(block)\\n \\n if not code_snippets:\\n return False\\n \\n indent_diffs = []\\n all_indents = []\\n \\n # 3. 逐行分析所有提取到的代码片段，精准计算语法带来的缩进增量\\n for code in code_snippets:\\n lines = code.split('\\\\n')\\n prev_indent = 0\\n prev_line = \\\"\\\"\\n for line in lines:\\n stripped = line.strip()\\n \\n # 跳过空行和纯注释行，因为它们不会作为有效逻辑行影响缩进层级关联\\n if not stripped or stripped.startswith('#'):\\n continue\\n \\n # 计算当前行的前导空格数量\\n match = re.match(r'^( +)', line)\\n current_indent = len(match.group(1)) if match else 0\\n \\n if current_indent > 0:\\n all_indents.append(current_indent)\\n \\n # 当上一逻辑有效行是以冒号结尾时，下一行的缩进理应递增，记录此时的增量\\n if prev_line.endswith(':'):\\n if current_indent > prev_indent:\\n indent_diffs.append(current_indent - prev_indent)\\n \\n prev_line = stripped\\n prev_indent = current_indent\\n \\n # 4. 判定规则：只要在基础语法块结构中出现 2 个空格的递增，即判定为助手的输出满足了替换缩进要求\\n if indent_diffs:\\n counter = Counter(indent_diffs)\\n if counter.get(2, 0) > 0:\\n return True\\n return False\\n \\n # 回退判断（防止没有任何冒号结构的极端少代码情况）\\n if all_indents:\\n counter = Counter(all_indents)\\n if counter.get(2, 0) > 0:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n code_extensions = ['.py']\\n code_files = []\\n \\n # 查找项目工作区中的所有代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n code_files.append(os.path.join(root, file))\\n \\n has_checked_code = False\\n \\n # 遍历验证工作区里的代码文件首行（顶行）\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 如果文件为空或者不包含特定声明，直接返回 False\\n if not first_line or \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n has_checked_code = True\\n except Exception:\\n # 遇到编码问题导致无法读取的文件也视作不符合（要求为utf-8）\\n return False\\n \\n # 如果工作区中存在实质生成的代码文件且验证全部通过，则直接返回 True\\n if has_checked_code:\\n return True\\n \\n # 兜底策略：如果工作区中没有发现代码文件，则检查 response 中的 Python 代码块\\n python_blocks = re.findall(r'```python[ \\\\t]*\\\\n(.*?)```', response, re.DOTALL)\\n for block in python_blocks:\\n lines = block.split('\\\\n')\\n # 获取第一行代码进行检验\\n if not lines or \\\"-*- coding: utf-8 -*-\\\" not in lines[0]:\\n return False\\n has_checked_code = True\\n \\n # 如果找到了相应的代码文件或代码块并且验证通过，则返回 True，否则返回 False（未生成代码）\\n return has_checked_code\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来引入OCR核心功能。请在界面顶部的“加载图片”按钮旁边，新增一个“提取文字”按钮。当用户点击这个按钮时，请在后台使用 PaddleOCR（或类似的高精度中文OCR库）对当前已经加载显示的图片进行文字识别。将识别出来的所有文本内容按顺序提取出来，逐行追加显示在界面右侧的文本显示框中。为了防止界面卡死，请确保在处理OCR时界面不会假死。另外后续所有操作长期遵循两条规则：第一，输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求；第二，执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求\",\n \"validation_code\": \"import os\\nimport re\\nimport tokenize\\nfrom io import BytesIO\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def check_code_indent(code_str: str) -> bool:\\n try:\\n # 首选方案：尝试使用 tokenize 分析逻辑行的缩进，这是最严谨的判断方式\\n tokens = list(tokenize.tokenize(BytesIO(code_str.encode('utf-8')).readline))\\n for tok in tokens:\\n if tok.type == tokenize.INDENT:\\n indent_str = tok.string\\n # 必须全部是空格，不包含制表符等\\n if not all(c == ' ' for c in indent_str):\\n return False\\n # 缩进长度必须是 4 的倍数\\n if len(indent_str) % 4 != 0:\\n return False\\n return True\\n except Exception:\\n # 降级方案：针对不完整代码片段、diff patch 等引起 tokenize 解析异常的情况\\n lines = code_str.split('\\\\n')\\n prev_colon_indent = None\\n for line in lines:\\n if not line.strip():\\n continue\\n \\n leading_spaces = len(line) - len(line.lstrip(' '))\\n if '\\\\t' in line[:leading_spaces]:\\n return False\\n \\n is_comment = line.lstrip().startswith('#')\\n \\n if prev_colon_indent is not None:\\n if not is_comment:\\n # 遇到冒号后的非空非注释行，如果产生新的缩进级别，应当严格等于之前冒号行缩进 + 4 个空格\\n if leading_spaces > prev_colon_indent:\\n if leading_spaces != prev_colon_indent + 4:\\n return False\\n prev_colon_indent = None\\n \\n if not is_comment:\\n stripped = line.strip()\\n # 检查典型代码块关键字开头，其缩进长度需要严格满足 4 的倍数要求\\n if stripped.startswith(('def ', 'class ', 'if ', 'elif ', 'else:', 'for ', 'while ', \\n 'try:', 'except', 'finally:', 'with ', 'return ', 'import ', \\n 'from ', 'print(', 'self.', 'pass', 'continue', 'break')):\\n if leading_spaces > 0 and leading_spaces % 4 != 0:\\n return False\\n \\n # 剥离行内注释后，寻找开启新缩进块的冒号\\n code_part = line.split('#')[0].rstrip()\\n if code_part.endswith(':'):\\n prev_colon_indent = leading_spaces\\n return True\\n\\n # 1. 检查 workspace_path 下生成的 Python 代码文件和 Markdown 文档里的代码块\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.py'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_code_indent(code):\\n return False\\n except Exception:\\n pass\\n elif file.endswith('.md'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n md_content = f.read()\\n md_blocks = re.findall(r'```(?:python|py)\\\\s*\\\\n(.*?)\\\\n```', md_content, re.DOTALL | re.IGNORECASE)\\n for block in md_blocks:\\n if not check_code_indent(block):\\n return False\\n except Exception:\\n pass\\n \\n # 2. 检查最后一轮 response 中输出的 Python 代码块\\n python_blocks = re.findall(r'```(?:python|py)\\\\s*\\\\n(.*?)\\\\n```', response, re.DOTALL | re.IGNORECASE)\\n for block in python_blocks:\\n if not check_code_indent(block):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n has_code_file = False\\n \\n # 遍历项目地址下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令背景，推断主要代码文件类型为 Python (.py)\\n if file.endswith('.py'):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取代码文件顶行（第一行）\\n first_line = f.readline()\\n \\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果读取文件失败（例如编码不正确等），则视为不满足要求\\n return False\\n \\n # 如果没找到任何代码文件，则要求未满足；否则满足了所有代码文件都具备该声明\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"很好。下一步是实现空间位置的标注功能。我希望在识别出文字的同时，能够直观地看到它们在图中的具体位置。请修改“提取文字”后的处理逻辑：不仅在右侧输出文字，还要使用OpenCV或PIL在原始图片上，用红色的矩形框（BBox）将所有识别出的文字区域框选出来。画好红色标注框后，将这张带有标注的图片更新显示到左侧的图片区域中。另外后续所有操作长期遵循两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，你输出的代码中，禁止出现超过3层的循环嵌套。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import re\\n import os\\n import io\\n import tokenize\\n\\n code_snippets = []\\n\\n # 1. 提取 response 中以 Markdown 格式输出的代码片段\\n # 匹配形如 ```lang ... ``` 的代码块\\n for match in re.finditer(r'```([a-zA-Z]*)\\\\n(.*?)```', response, re.DOTALL):\\n lang = match.group(1).lower()\\n # 排除纯文本和一般文档（由于要求是“输出代码时”），对疑似代码的数据统统收集\\n if lang not in ('md', 'markdown', 'txt', 'text'):\\n code_snippets.append(match.group(2))\\n\\n # 2. 提取 workspace_path 中实际保存的代码文件\\n if workspace_path and os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 结合用户指令推断主要为 Python 代码，并囊括少量可能存在的辅助脚本配置\\n if file.endswith(('.py', '.sh', '.json')):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n code_snippets.append(f.read())\\n except Exception:\\n pass\\n\\n # 如果没有任何代码片段，则未触发检查，默认满足规则\\n if not code_snippets:\\n return True\\n\\n def check_snippet(code_str: str) -> bool:\\n # 为了防止缺少末尾换行符导致 tokenize 解析抛出异常，末尾补齐换行符\\n code_str = code_str + '\\\\n'\\n \\n # 尝试使用 python 原生的 tokenize 进行逻辑缩进检测\\n try:\\n tokens = tokenize.tokenize(io.BytesIO(code_str.encode('utf-8')).readline)\\n indent_levels = [0]\\n for tok in tokens:\\n if tok.type == tokenize.INDENT:\\n # 不允许出现制表符缩进\\n if '\\\\t' in tok.string:\\n return False\\n indent_len = len(tok.string)\\n step = indent_len - indent_levels[-1]\\n # 严格的2空格缩进要求，每次逻辑缩进增加的步长必须为2\\n if step != 2:\\n return False\\n indent_levels.append(indent_len)\\n elif tok.type == tokenize.DEDENT:\\n if len(indent_levels) > 1:\\n indent_levels.pop()\\n return True\\n \\n except Exception:\\n # 降级验证策略：针对非 Python 代码或因不完整而报错的代码，采用正则按行验证\\n lines = code_str.split('\\\\n')\\n indents = []\\n for line in lines:\\n stripped = line.strip()\\n # 忽略空行和常见注释行\\n if not stripped or stripped.startswith('#') or stripped.startswith('//'):\\n continue\\n m = re.match(r'^( +)\\\\S', line)\\n if m:\\n indents.append(len(m.group(1)))\\n\\n # 如果不存在任何缩进，则不视为违规\\n if not indents:\\n return True\\n\\n # 如果所有缩进均满足是4的倍数，且存在有效的缩进，那么代码极有可能仍然使用的是4空格缩进，不符合要求\\n if all(ind % 4 == 0 for ind in indents) and any(ind > 0 for ind in indents):\\n return False\\n\\n return True\\n\\n # 3. 遍历检查搜集到的所有代码片段，如果存在未通过检测的代码，则返回 False\\n for snippet in code_snippets:\\n if not snippet.strip():\\n continue\\n if not check_snippet(snippet):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉隐藏目录和常见的环境/缓存目录，确保只检查助手生成的代码文件\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ['venv', 'env', '__pycache__']]\\n \\n for file_name in files:\\n # 推测此项目的代码文件主要为 Python 脚本文件\\n if file_name.endswith('.py'):\\n found_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查第一行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果文件不可读或读取过程中发生异常，判定为不满足要求\\n return False\\n \\n # 确保至少检查过一个生成的代码文件，若是则表示全部满足要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过3层的循环嵌套\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在来做一个交互增强功能。请将右侧的多行文本框替换为一个列表组件（如 QListWidget）。每次OCR识别后，将每一句识别出的文字作为列表的一项填入。当用户在右侧列表中鼠标点击选中某一行文字时，左侧图片上对应的那个红色标注框需要变成绿色并且线条加粗，而其他的框保持红色不变。这样可以方便用户核对具体的文字对应原图上的哪个位置。另外，后续所有操作长期遵循以下两条规则：第一，你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求；第二，编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import tokenize\\n import io\\n\\n indent_sizes = []\\n\\n def process_code(code_str: str) -> bool:\\n \\\"\\\"\\\"\\n 利用 tokenize 解析代码片段，提取每一次发生逻辑缩进时的增量大小。\\n 如果包含了 Tab 缩进则直接返回 False，否则返回 True。\\n \\\"\\\"\\\"\\n try:\\n # 补齐尾部换行符以防部分代码因缺少换行导致 EOF TokenError\\n code_str = code_str.strip() + '\\\\n'\\n tokens = tokenize.generate_tokens(io.StringIO(code_str).readline)\\n indent_stack = [0]\\n \\n for tok in tokens:\\n if tok.type == tokenize.INDENT:\\n s = tok.string\\n if '\\\\t' in s:\\n return False\\n \\n curr_indent = len(s)\\n diff = curr_indent - indent_stack[-1]\\n if diff > 0:\\n indent_sizes.append(diff)\\n indent_stack.append(curr_indent)\\n elif tok.type == tokenize.DEDENT:\\n if len(indent_stack) > 1:\\n indent_stack.pop()\\n except Exception:\\n # 忽略因解析不完整的代码段抛出的 TokenError 等异常\\n pass\\n return True\\n\\n # 1. 处理 response 中的 Python 代码块\\n # 匹配诸如 ```python, ```py, 或者纯 ``` 包裹的代码块\\n code_blocks = re.findall(r'```(?:python|py)?[ \\\\t]*\\\\n(.*?)\\\\n```', response, re.DOTALL)\\n for code in code_blocks:\\n if not process_code(code):\\n return False\\n\\n # 2. 处理 workspace_path 下生成的 .py 文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for f in files:\\n if f.endswith('.py'):\\n file_path = os.path.join(root, f)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as file:\\n content = file.read()\\n if not process_code(content):\\n return False\\n except Exception:\\n pass\\n\\n # 如果未能提取到任何缩进记录，意味着没有涉及任何层级的缩进\\n # 在没有提供不符合要求证据的情况下，判定为没有违背规则\\n if not indent_sizes:\\n return True\\n\\n # 3. 统计并分析收集到的缩进增量情况\\n count_2 = sum(1 for d in indent_sizes if d == 2)\\n count_4 = sum(1 for d in indent_sizes if d == 4)\\n\\n # 如果4个空格的缩进占了主导位置（数量多于2个空格），说明主要使用了4空格缩进\\n if count_4 > count_2:\\n return False\\n \\n # 如果完全没有2个空格的逻辑缩进，但在记录中又存在其他的增量缩进（如3, 4, 8等）\\n if count_2 == 0 and len(indent_sizes) > 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n has_code_file = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令，该项目主要生成的代码文件为 Python 文件\\n if file.endswith('.py'):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n # 读取顶行（第一行）\\n first_line = f.readline()\\n \\n # 判断顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果无法读取文件，视作不满足要求\\n return False\\n \\n # 如果没有找到任何代码文件，则无法证明满足了要求；反之如果全部检查通过，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"执行bash下的安装、构建或测试命令时，为了避免进程出现挂起现象，必须在命令头部添加 `timeout 30s` 进行包装\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"非常棒，接下来实现本地持久化保存功能。在界面底部增加一个“保存结果”按钮。点击后，打开系统的保存文件对话框，让用户选择保存目录和主文件名（如 result）。程序需要做两件事：1. 将当前界面上带有所有标注框的图片保存为 `[文件名].png`；2. 把识别出的所有数据保存为一个 `[文件名].json` 文件，JSON数据需包含每个文本片段的内容、对应的四点坐标边界以及识别置信度。另外后续所有操作长期遵循以下两条规则：第一，输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求；第二，运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求\",\n \"validation_code\": \"import os\\nimport re\\nimport tokenize\\nimport io\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n python_codes = []\\n \\n # 1. 从工作区路径中提取所有 Python 文件的代码\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8', errors='ignore') as f:\\n python_codes.append(f.read())\\n except Exception:\\n pass\\n \\n # 2. 从回复内容中提取所有的 Python 代码块 (支持 python/python3/py)\\n python_blocks = re.findall(r'```(?:python|python3|py)\\\\n(.*?)```', response, re.DOTALL)\\n python_codes.extend(python_blocks)\\n \\n # 如果没有任何代码，无法验证其是否满足“输出代码时”的前提，视为不满足\\n if not python_codes:\\n return False\\n \\n has_code = False\\n \\n # 3. 验证所有的 Python 代码缩进\\n for code in python_codes:\\n if not code.strip():\\n continue\\n has_code = True\\n \\n try:\\n token_generator = tokenize.generate_tokens(io.StringIO(code).readline)\\n while True:\\n try:\\n tok = next(token_generator)\\n # 仅在产生新的代码块缩进时，tokenizer 才会生成 INDENT token\\n if tok.type == tokenize.INDENT:\\n indent_str = tok.string\\n # 缩进不能包含制表符\\n if '\\\\t' in indent_str:\\n return False\\n # 空格数量必须是 4 的整数倍 (拒绝2空格或其他非标准缩进)\\n if len(indent_str) % 4 != 0:\\n return False\\n except StopIteration:\\n break\\n # 捕获 TokenError, IndentationError 等因 AI 生成了片段化或不完整代码引发的错误\\n # 停止当前代码的解析，直接跳出处理下一个代码块\\n except (tokenize.TokenError, IndentationError, SyntaxError):\\n break\\n except Exception:\\n pass\\n\\n return has_code\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据项目需求，主要排查 Python 代码文件\\n if file.endswith('.py'):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 若文件无法正常打开或读取（如编码等异常），则视为不满足条件\\n return False\\n \\n # 如果遍历完成没有遇到不符合规则的文件，并且确实找到了代码文件，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 6,\n \"instruction\": \"我们继续扩展。增加一个基于置信度的动态过滤功能。在右侧文字列表的上方增加一个滑块（Slider）和一个Label，滑块范围代表 0.00 到 1.00 的置信度（保留两位小数），默认值为0.00。当用户拖动滑块时，界面需要实时动态更新：右侧列表中自动隐藏掉置信度低于滑块设定值的文字项，同时左侧图片上也自动消失对应低置信度的矩形标注框（不需要重新调用OCR模型，直接操作当前的识别结果数据即可）。另外后续所有操作长期遵循两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求\",\n \"validation_code\": \"def check_indentation_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import tokenize\\n import io\\n\\n def evaluate_code(code_str, is_python):\\n diffs_2 = 0\\n diffs_4 = 0\\n has_tab = False\\n \\n parsed = False\\n if is_python:\\n try:\\n # 针对 Python 代码，使用 tokenize 获取准确的缩进层次，忽略悬挂对齐等\\n tokens = tokenize.generate_tokens(io.StringIO(code_str).readline)\\n levels = [0]\\n for tok in tokens:\\n if tok.type == tokenize.INDENT:\\n if '\\\\t' in tok.string:\\n has_tab = True\\n curr = len(tok.string)\\n diff = curr - levels[-1]\\n if diff == 2:\\n diffs_2 += 1\\n elif diff == 4:\\n diffs_4 += 1\\n levels.append(curr)\\n elif tok.type == tokenize.DEDENT:\\n if len(levels) > 1:\\n levels.pop()\\n parsed = True\\n except Exception:\\n pass\\n \\n # 非Python代码或Python代码解析失败，回退到基于文本行的前导空格差值计算\\n if not parsed:\\n lines = code_str.split('\\\\n')\\n levels = []\\n for line in lines:\\n # 过滤空行以及单行注释行，减少对缩进统计的干扰\\n if line.strip() == '' or line.lstrip().startswith('#') or line.lstrip().startswith('//'):\\n continue\\n leading = len(line) - len(line.lstrip(' '))\\n if '\\\\t' in line[:leading]:\\n has_tab = True\\n levels.append(leading)\\n \\n for i in range(1, len(levels)):\\n diff = levels[i] - levels[i-1]\\n if diff == 2:\\n diffs_2 += 1\\n elif diff == 4:\\n diffs_4 += 1\\n \\n return diffs_2, diffs_4, has_tab\\n\\n # 1. 优先检查 response 中直接包含的 Markdown 代码块\\n code_blocks = re.findall(r'```(\\\\w*)\\\\n(.*?)```', response, re.DOTALL)\\n response_has_2 = False\\n response_has_code = False\\n \\n for lang, code in code_blocks:\\n is_py = lang.lower() in ['python', 'py', '']\\n d2, d4, ht = evaluate_code(code, is_py)\\n if ht:\\n return False # 如果生成了包含 tab 的代码，必然不符合2个空格的要求\\n \\n if d2 > 0 or d4 > 0:\\n response_has_code = True\\n if d4 > d2:\\n # 如果某个代码块明显以 4 个空格作为主力缩进，说明没有统一遵守新指令\\n return False\\n if d2 >= d4 and d2 > 0:\\n response_has_2 = True\\n\\n if response_has_2:\\n return True\\n \\n # 2. 如果 response 中没有足够的代码体现缩进规则，去检查工作区文件（涵盖通过沙箱工具进行的操作）\\n total_d2 = 0\\n total_d4 = 0\\n if os.path.isdir(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1]\\n if ext in ['.py', '.json', '.js', '.html', '.css', '.vue', '.ts', '.tsx']:\\n is_py = (ext == '.py')\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n code = f.read()\\n d2, d4, ht = evaluate_code(code, is_py)\\n if ht:\\n continue # 在历史或第三方文件中如果存在tab，忽略之以防止误伤\\n total_d2 += d2\\n total_d4 += d4\\n except Exception:\\n pass\\n \\n # 3. 如果没找到任何带有有效缩进的代码块和文件，默认认为未违背\\n if not response_has_code and (total_d2 == 0 and total_d4 == 0):\\n return True \\n\\n # 在所有代码文件的综合数据中，要求 2 空格缩进占主体地位，取代原来的 4 空格\\n return total_d2 >= total_d4 and total_d2 > 0\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n has_checked_any = False\\n \\n # 优先检查工作目录中实际生成的代码文件\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除常见的无关隐藏文件夹及虚拟环境文件夹，避免误判\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in ('venv', 'env', '__pycache__', 'node_modules')]\\n for file in files:\\n if file.endswith('.py'):\\n has_checked_any = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n lines = f.readlines()\\n if not lines:\\n return False # 代码文件为空，不满足“包含声明”的条件\\n first_line = lines[0].strip()\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n return False\\n \\n # 如果在文件系统中成功找到了对应的代码文件，并全部通过了上述要求，则验证成功\\n if has_checked_any:\\n return True\\n \\n # 后备方案：如果工作目录中并未找到 .py 文件，检查回复中是否包含 Python 代码块\\n code_blocks = re.findall(r'```(?:python|py|Python)[ \\\\t]*\\\\n(.*?)```', response, re.DOTALL)\\n if code_blocks:\\n has_checked_any = True\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n if not lines:\\n return False\\n first_line = lines[0].strip()\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n \\n # 只要进行了有效的检查且未在上面拦截返回False，则认为通过验证\\n if has_checked_any:\\n return True\\n \\n return False\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n def check_equals_spacing(code: str) -> bool:\\n # 1. 移除多行字符串 (匹配 '''...''' 和 \\\"\\\"\\\"...\\\"\\\"\\\")\\n code = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", '', code)\\n code = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '', code)\\n \\n # 2. 移除单行字符串 (包含对转义符号 \\\\ 的处理，防止引号被意外转义)\\n code = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", '', code)\\n code = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '', code)\\n \\n # 3. 移除注释\\n code = re.sub(r'#.*', '', code)\\n \\n # 4. 检查纯代码逻辑中，= 的前面或后面是否含有空格或制表符\\n if re.search(r'[ \\\\t]=|=[ \\\\t]', code):\\n return False\\n return True\\n\\n # 步骤 A：检查最后一轮回复中包含的 Python 代码块\\n blocks = re.findall(r'```(?:python|py)\\\\s*\\\\n(.*?)```', response, re.DOTALL | re.IGNORECASE)\\n for block in blocks:\\n if not check_equals_spacing(block):\\n return False\\n \\n # 步骤 B：检查项目工作区中生成的所有 Python 代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 若发现文件中存在等号前后带有空格的现象，直接判定为未满足要求\\n if not check_equals_spacing(content):\\n return False\\n except Exception:\\n # 忽略无法读取的异常文件\\n pass\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 7,\n \"instruction\": \"下一步，增加本地缓存以提升体验。为了防止用户对同一个URL反复点击“提取文字”造成资源的浪费，请增加一个内存缓存机制（如使用字典）。以图片的哈希值（或URL本身）作为Key，缓存它的完整OCR识别结果。如果用户输入相同的URL并点击提取，先检查缓存，如果有，则直接从缓存读取数据渲染文字列表和标注框，并在控制台打印“命中缓存，跳过模型推理”。另外后续所有操作长期遵循以下两条规则：第一，输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求；第二，在整个回复过程中，任何文件路径或代码位置都必须始终仅以双引号将其包裹起来。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用4个空格作为缩进标准，替换之前的2空格缩进要求\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n import os\\n from collections import Counter\\n \\n def analyze_snippet(snippet: str):\\n diffs = []\\n indents = []\\n lines = snippet.split('\\\\n')\\n prev_indent = 0\\n \\n for line in lines:\\n if not line.strip():\\n continue\\n # 忽略纯注释行，防止注释的不规则缩进影响判断\\n if line.lstrip().startswith('#'):\\n continue\\n \\n match = re.match(r'^([ \\\\t]*)', line)\\n leading = match.group(1) if match else \\\"\\\"\\n \\n # 为计算缩进长度，将制表符视作4个空格\\n curr_indent = len(leading.replace('\\\\t', ' '))\\n indents.append(curr_indent)\\n \\n # 记录缩进的增量\\n if curr_indent > prev_indent:\\n diffs.append(curr_indent - prev_indent)\\n prev_indent = curr_indent\\n \\n # 如果没有产生缩进增加（例如只有单行代码或全局变量平铺）\\n if not diffs:\\n valid_indents = [i for i in indents if i > 0]\\n if not valid_indents:\\n return None # 无法判断该片段的缩进类型\\n c = Counter(valid_indents)\\n most_common = c.most_common(1)[0][0]\\n # 检查主要存在的绝对缩进是否是4的倍数\\n return most_common % 4 == 0\\n \\n c = Counter(diffs)\\n count_4 = c.get(4, 0)\\n count_2 = c.get(2, 0)\\n \\n # 核心判断逻辑，通过最直观的缩进步长对比\\n if count_4 > count_2:\\n return True\\n elif count_2 > count_4:\\n return False\\n else:\\n if count_4 == 0:\\n most_common = c.most_common(1)[0][0]\\n return most_common % 4 == 0\\n else:\\n return True\\n\\n # 1. 尝试从 response 提取 Python 或无高亮类型的代码块\\n pattern = re.compile(r'```(?:python|py)?[ \\\\t]*\\\\n(.*?)```', re.DOTALL | re.IGNORECASE)\\n matches = pattern.findall(response)\\n \\n response_results = []\\n for snippet in matches:\\n res = analyze_snippet(snippet)\\n if res is not None:\\n response_results.append(res)\\n \\n # 如果回复中能找到任何应用了 4 空格的代码块，判定为遵循要求\\n if any(response_results):\\n return True\\n \\n # 2. 如果 response 无法判定，读取工作区所有的 .py 文件进行检查\\n file_results = []\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n snippet = f.read()\\n res = analyze_snippet(snippet)\\n if res is not None:\\n file_results.append(res)\\n except Exception:\\n pass\\n \\n # 只要存在被修改或新生成的文件体现了 4 空格，就判定为符合要求\\n if any(file_results):\\n return True\\n \\n # 只有当包含可以判定的代码，且所有代码段皆违背4空格缩进要求时，才判定为 False\\n if response_results or file_results:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n has_py_files = False\\n\\n # 1. 检查 workspace_path 下生成的所有 .py 文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 根据任务背景，推断代码文件类型为 Python 脚本\\n if file_name.endswith('.py'):\\n has_py_files = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 验证顶行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 无法正常打开/读取的代码文件安全起见视为不满足要求\\n return False\\n\\n # 若在工作区找到了生成的 .py 代码文件，且均通过了上方检查，则说明满足要求\\n if has_py_files:\\n return True\\n\\n # 2. 作为后备方案：如果未在工作区找到本地 .py 文件，则检查回复内容里的 python 代码块\\n code_blocks = re.findall(r'```(?:python|py)[ \\\\t]*\\\\n(.*?)\\\\n[ \\\\t]*```', response, re.DOTALL | re.IGNORECASE)\\n for block in code_blocks:\\n lines = block.split('\\\\n')\\n if lines:\\n first_line = lines[0]\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格\",\n \"validation_code\": \"def verify_no_space_around_equals(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n code_contents = []\\n\\n # 1. 提取 response 中以 Markdown 代码块形式返回的代码\\n blocks = re.findall(r'```[^\\\\n]*\\\\n(.*?)```', response, re.DOTALL)\\n if blocks:\\n code_contents.extend(blocks)\\n\\n # 2. 遍历 workspace_path 中生成的 .py 代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n try:\\n with open(os.path.join(root, file), 'r', encoding='utf-8') as f:\\n code_contents.append(f.read())\\n except Exception:\\n pass\\n \\n # 若没有检测到任何代码，则默认没有违规行为\\n if not code_contents:\\n return True\\n \\n for code in code_contents:\\n # 3. 过滤掉注释，避免注释内存在的“= ”造成误判\\n code = re.sub(r'#.*', '', code)\\n \\n # 4. 过滤掉多行和单行字符串字面量，防止字符串内部的等于号文本造成误判\\n code = re.sub(r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'''[\\\\s\\\\S]*?'''\\\", \\\"''\\\", code)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", code)\\n \\n # 5. 检查“=”前后是否存有空格或水平制表符\\n if re.search(r'[ \\\\t]=|=[ \\\\t]', code):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在整个回复过程中，任何文件路径或代码位置都必须始终仅以双引号将其包裹起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 8,\n \"instruction\": \"最后一轮，增加一个数据脱敏场景下的高级图片处理功能。在底部再增加一个“文字涂抹遮盖”按钮。点击后，程序要利用当前保存的识别坐标框，将原图中所包含文字的矩形区域全部用纯黑色（或图片均值颜色）填充涂抹掉，生成一张“脱敏后的无字图片”，并将这张图片展示在左侧区域。右侧的文本列表同时清空，此时滑块交互暂停生效。另外后续所有操作长期遵循以下两条规则：第一，输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求；第二，只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"不论何时，你生成用户回复时都必须使用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的任何一个代码文件中，都必须通篇使用中文作为注释语言\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里，关于文件的引用必须一律使用绝对文件路径，不要使用相对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，请倾向于把正斜杠 / 用作路径分隔符\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"凡是涉及代码的更改，在动手修改之前都要把计划写进code_modify_plan.md\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到project_progress_log.md文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"请在项目路径下维护一个名为project_structure.md的文档，用于说明项目的结构和功能，并在每次有内容更新时同步维护该文档。每次准备加入新功能前，为了防止上下文超限，你必须先通过执行 tail -500 project_structure.md 来读取此文档的最后500行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"处理文件操作时，必须尽量采用专用工具（如Read, Write, Edit），而非bash命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"输出代码时统一使用2个空格作为缩进标准，替换之前的4空格缩进要求\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n from collections import Counter\\n\\n def is_2_space_indent(content: str) -> bool:\\n lines = content.split('\\\\n')\\n # 简单检查是否有制表符作为缩进，有则直接判违规\\n for line in lines:\\n if line.startswith('\\\\t') or line.startswith(' \\\\t'):\\n return False\\n \\n indents = []\\n for line in lines:\\n stripped = line.lstrip(' ')\\n # 过滤空行及常见语言的纯注释行，防止干扰判断\\n if not stripped or stripped.startswith('#') or stripped.startswith('//') or stripped.startswith('\\\"\\\"\\\"') or stripped.startswith(\\\"'''\\\"):\\n continue\\n indent = len(line) - len(stripped)\\n indents.append(indent)\\n \\n diffs = []\\n # 计算相邻行的缩进增量\\n for i in range(1, len(indents)):\\n diff = indents[i] - indents[i-1]\\n if diff > 0:\\n diffs.append(diff)\\n \\n if not diffs:\\n return True\\n \\n counts = Counter(diffs)\\n # 如果完全没有 2 空格的递增，但有 4 或 8 空格的递增，说明肯定不是 2 空格缩进\\n if counts.get(2, 0) == 0:\\n if counts.get(4, 0) > 0 or counts.get(8, 0) > 0:\\n return False\\n \\n return True\\n\\n code_contents = []\\n\\n # 1. 检查工作区中的代码文件\\n if os.path.isdir(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉常见的隐藏目录或依赖环境目录\\n dirs[:] = [d for d in dirs if d not in ('venv', '.git', 'node_modules', '__pycache__', '.idea', '.vscode', '.env')]\\n for file in files:\\n if file.endswith(('.py', '.json', '.js', '.html', '.css', '.cpp', '.c', '.java', '.sh')):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code_contents.append(f.read())\\n except Exception:\\n pass\\n\\n # 2. 检查助手回复中的代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9_-]*\\\\n(.*?)```', response, re.DOTALL)\\n code_contents.extend(code_blocks)\\n\\n if not code_contents:\\n return True\\n\\n # 3. 验证所有的代码片段是否都符合 2 空格缩进标准\\n for content in code_contents:\\n if not content.strip():\\n continue\\n if not is_2_space_indent(content):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"强制要求：所有生成的代码文件顶行必须包含特定的编码声明 -*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查所有生成的Python代码文件顶行是否包含特定的编码声明 -*- coding: utf-8 -*-\\n \\\"\\\"\\\"\\n py_files_found = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令推测，主要代码文件为 Python 文件\\n if file.endswith('.py'):\\n py_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试打开并读取文件第一行\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline().strip()\\n # 检查顶行是否包含特定的编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception as e:\\n # 如果文件无法读取，则视为无法满足要求\\n return False\\n \\n # 如果找到了至少一个.py文件并且都满足要求，则返回True\\n # 如果没有找到任何.py文件，则无法证明满足了要求，返回False\\n return py_files_found\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"运行 bash 命令的过程中，不得使用 `timeout` 设定执行时间限制，替换之前的执行bash命令需加timeout30s的要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，禁止出现超过2层的循环嵌套，替换之前的3层循环嵌套要求\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"编写完代码后，请务必对全部代码执行测试，并确保测试顺利通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"在编写代码文件时，务必保证“=”的前后紧贴相邻内容，不能留有空格\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import io\\n import re\\n import tokenize\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n py_files = []\\n # 查找工作区下所有的 .py 代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith('.py'):\\n py_files.append(os.path.join(root, file))\\n \\n # 如果没有生成任何 Python 文件，不满足基本前提\\n if not py_files:\\n return False\\n \\n has_checked_any = False\\n \\n for file_path in py_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n has_checked_any = True\\n \\n try:\\n # 方案 A: 使用 tokenize 模块分析 Python 语法，安全过滤字符串和注释中的 '='\\n tokens = list(tokenize.tokenize(io.BytesIO(content.encode('utf-8')).readline))\\n for token in tokens:\\n if token.type == tokenize.OP and token.string == '=':\\n line = token.line\\n start_col = token.start[1]\\n end_col = token.end[1]\\n \\n # 检查 '=' 前一个字符是否为空格或制表符\\n if start_col > 0:\\n prev_char = line[start_col - 1]\\n if prev_char in (' ', '\\\\t'):\\n return False\\n \\n # 检查 '=' 后一个字符是否为空格或制表符\\n if end_col < len(line):\\n next_char = line[end_col]\\n if next_char in (' ', '\\\\t'):\\n return False\\n \\n except Exception:\\n # 方案 B: Tokenize 失败（可能存在语法错误）时的降级处理方案\\n lines = content.split('\\\\n')\\n for line in lines:\\n # 简单去除行内注释\\n line = line.split('#')[0]\\n # 使用正则简单去除字符串常量\\n line = re.sub(r'([\\\"\\\\'])(?:(?=(\\\\\\\\?))\\\\2.)*?\\\\1', '', line)\\n \\n for i, char in enumerate(line):\\n if char == '=':\\n # 排除 ==, !=, >=, <=, +=, -= 等复合赋值/比较运算符的干扰\\n if i > 0 and line[i-1] in '!<>=+-*/%&|^':\\n continue\\n if i < len(line) - 1 and line[i+1] == '=':\\n continue\\n \\n # 检查独立的 '=' 前后是否有空格\\n if (i > 0 and line[i-1] in ' \\\\t') or (i < len(line) - 1 and line[i+1] in ' \\\\t'):\\n return False\\n\\n # 若所有已读取的代码文件均通过检查，则满足要求\\n return has_checked_any\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"在整个回复过程中，任何文件路径或代码位置都必须始终仅以双引号将其包裹起来\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"只要你需要修改已有文件，就必须提前在原文件存放的同级目录下创建该文件修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["查看程序初始界面，确认包含以下UI元素：顶部有URL文本输入框（默认URL为'https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_002.png'）及'加载图片'、'提取文字'按钮；主区域分为左侧大图显示区和右侧列表组件区域；右侧列表上方配有范围为0.00到1.00的置信度滑块和对应的两位小数数值Label（默认显示0.00）；底部包含'保存结果'和'文字涂抹遮盖'按钮。","在顶部的URL输入框中手动输入一个非图片内容链接（例如返回HTML页面的链接），点击'加载图片'按钮。观察界面，确认程序不是仅依赖URL后缀判断，而是通过二进制文件头或图像库解码校验内容，并在内容不是有效图片时弹出错误提示对话框。","在URL输入框中使用默认的正确图片URL，点击'加载图片'按钮。等待下载解析完成后，观察左侧主显示区，确认图片被成功读取并自适应比例缩放展示在界面上。","在成功加载正确图片后，点击顶部的'提取文字'按钮。等待处理完成后，观察界面，确认右侧列表组件中按识别顺序逐行显示所有识别出的文本片段，且左侧图片中所有含有文字的区域都被红色矩形框准确框选标注。","在完成文字提取并在界面展示结果后，使用鼠标在右侧文本列表中点击选中任意一行文字。观察左侧图片，确认该行文字所对应的红色矩形标注框变成了绿色，且线条加粗，同时确认图片上的其余标注框依然保持红色不变。","在完成文字提取后，使用鼠标将右侧上方的置信度滑块向右拖动到一个较高的数值（如0.80）。观察界面动态变化，确认Label实时显示当前两位小数阈值，右侧文本列表中自动隐藏置信度低于该阈值的文本项，同时左侧图片上对应的矩形标注框也同步消失，且该过程不重新调用OCR模型。","在完成文字提取且左侧显示标注框后，点击底部'保存结果'按钮。在弹出的系统保存文件对话框中选择某个文件夹并设定文件名为'result'，点击保存。打开对应的本地文件夹，确认成功生成了'result.png'（带有全部标注框的图片）以及'result.json'（内部包含每个文本片段的文本内容、四点坐标边界以及识别置信度）。","在对默认图片完成一次文字提取后，保持URL输入框内容不变，再次点击'提取文字'按钮。观察开发环境的控制台日志输出，确认打印了'命中缓存，跳过模型推理'相关提示，并观察界面瞬间直接渲染出原有的右侧文字列表和左侧红色标注框。","在完成文字提取并展示完整标注框结果后，点击底部的'文字涂抹遮盖'按钮。观察界面，确认左侧图片中原本包含文字的矩形区域全部被纯黑色或图片均值颜色填充遮盖，生成一张脱敏后的无字图片；同时确认右侧文本列表内容被完全清空，置信度滑块交互暂停生效，不再继续过滤或恢复已清空的文本项。","点击'提取文字'按钮后，在OCR处理期间尝试拖动窗口、点击界面或移动控件焦点。确认OCR任务在后台执行，主界面不会长时间假死或无响应；处理完成后再统一更新右侧文字列表和左侧标注图片。","完成OCR识别后，检查每个列表项与图片标注框之间的对应关系：点击不同列表项时，应只将当前文本对应的一个标注框变为绿色且加粗，切换到其他项时原先选中的框恢复为红色，新的对应框变为绿色加粗。","在首次OCR识别后调整置信度滑块进行过滤，再将滑块调回0.00。确认程序直接基于当前已保存的识别结果数据恢复显示所有文本项和对应标注框，而不是重新下载图片或重新调用OCR模型。","完成文字提取后点击'保存结果'，检查保存出的JSON数据是否覆盖当前OCR识别得到的完整结果，而不是只保存右侧列表当前筛选后可见的部分；每条记录都应包含文本内容、四点坐标边界和置信度。"],"string":"[\n \"查看程序初始界面，确认包含以下UI元素：顶部有URL文本输入框（默认URL为'https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_002.png'）及'加载图片'、'提取文字'按钮；主区域分为左侧大图显示区和右侧列表组件区域；右侧列表上方配有范围为0.00到1.00的置信度滑块和对应的两位小数数值Label（默认显示0.00）；底部包含'保存结果'和'文字涂抹遮盖'按钮。\",\n \"在顶部的URL输入框中手动输入一个非图片内容链接（例如返回HTML页面的链接），点击'加载图片'按钮。观察界面，确认程序不是仅依赖URL后缀判断，而是通过二进制文件头或图像库解码校验内容，并在内容不是有效图片时弹出错误提示对话框。\",\n \"在URL输入框中使用默认的正确图片URL，点击'加载图片'按钮。等待下载解析完成后，观察左侧主显示区，确认图片被成功读取并自适应比例缩放展示在界面上。\",\n \"在成功加载正确图片后，点击顶部的'提取文字'按钮。等待处理完成后，观察界面，确认右侧列表组件中按识别顺序逐行显示所有识别出的文本片段，且左侧图片中所有含有文字的区域都被红色矩形框准确框选标注。\",\n \"在完成文字提取并在界面展示结果后，使用鼠标在右侧文本列表中点击选中任意一行文字。观察左侧图片，确认该行文字所对应的红色矩形标注框变成了绿色，且线条加粗，同时确认图片上的其余标注框依然保持红色不变。\",\n \"在完成文字提取后，使用鼠标将右侧上方的置信度滑块向右拖动到一个较高的数值（如0.80）。观察界面动态变化，确认Label实时显示当前两位小数阈值，右侧文本列表中自动隐藏置信度低于该阈值的文本项，同时左侧图片上对应的矩形标注框也同步消失，且该过程不重新调用OCR模型。\",\n \"在完成文字提取且左侧显示标注框后，点击底部'保存结果'按钮。在弹出的系统保存文件对话框中选择某个文件夹并设定文件名为'result'，点击保存。打开对应的本地文件夹，确认成功生成了'result.png'（带有全部标注框的图片）以及'result.json'（内部包含每个文本片段的文本内容、四点坐标边界以及识别置信度）。\",\n \"在对默认图片完成一次文字提取后，保持URL输入框内容不变，再次点击'提取文字'按钮。观察开发环境的控制台日志输出，确认打印了'命中缓存，跳过模型推理'相关提示，并观察界面瞬间直接渲染出原有的右侧文字列表和左侧红色标注框。\",\n \"在完成文字提取并展示完整标注框结果后，点击底部的'文字涂抹遮盖'按钮。观察界面，确认左侧图片中原本包含文字的矩形区域全部被纯黑色或图片均值颜色填充遮盖，生成一张脱敏后的无字图片；同时确认右侧文本列表内容被完全清空，置信度滑块交互暂停生效，不再继续过滤或恢复已清空的文本项。\",\n \"点击'提取文字'按钮后，在OCR处理期间尝试拖动窗口、点击界面或移动控件焦点。确认OCR任务在后台执行，主界面不会长时间假死或无响应；处理完成后再统一更新右侧文字列表和左侧标注图片。\",\n \"完成OCR识别后，检查每个列表项与图片标注框之间的对应关系：点击不同列表项时，应只将当前文本对应的一个标注框变为绿色且加粗，切换到其他项时原先选中的框恢复为红色，新的对应框变为绿色加粗。\",\n \"在首次OCR识别后调整置信度滑块进行过滤，再将滑块调回0.00。确认程序直接基于当前已保存的识别结果数据恢复显示所有文本项和对应标注框，而不是重新下载图片或重新调用OCR模型。\",\n \"完成文字提取后点击'保存结果'，检查保存出的JSON数据是否覆盖当前OCR识别得到的完整结果，而不是只保存右侧列表当前筛选后可见的部分；每条记录都应包含文本内容、四点坐标边界和置信度。\"\n]","isConversational":false}}},{"rowIdx":86,"cells":{"task_id":{"kind":"number","value":87,"string":"87"},"repository_policy":{"kind":"string","value":"请你在后续所有轮次的交互中，严格遵循以下约束条件：\n1. 必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-。\n2. 只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令。\n3. 文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现。\n4. 在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改。\n5. 在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>。\n6. 必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况。\n7. 不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里。\n8. 请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀名\n code_extensions = {\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', \n '.c', '.cpp', '.cc', '.h', '.hpp', '.go', '.rb', \n '.php', '.cs', '.swift', '.rs', '.kt', '.sh'\n }\n\n # 排除常见的非生成源码的第三方依赖或编译产物目录\n exclude_dirs = {\n 'node_modules', 'venv', 'env', '.env', '__pycache__', \n 'dist', 'build', 'out', 'target', '__tests__'\n }\n\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n # 在原地修改 dirs 以跳过不必要遍历的目录（隐藏目录及排除目录）\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查指定声明内容是否在文件第一行中\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果发生读取错误或解码错误等，说明不满足合法的文本代码文件要求\n return False\n \n # 如果找到了至少一个代码文件并且都包含该声明，则要求被满足\n return has_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]}],"string":"[\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀名\\n code_extensions = {\\n '.py', '.js', '.jsx', '.ts', '.tsx', '.java', \\n '.c', '.cpp', '.cc', '.h', '.hpp', '.go', '.rb', \\n '.php', '.cs', '.swift', '.rs', '.kt', '.sh'\\n }\\n\\n # 排除常见的非生成源码的第三方依赖或编译产物目录\\n exclude_dirs = {\\n 'node_modules', 'venv', 'env', '.env', '__pycache__', \\n 'dist', 'build', 'out', 'target', '__tests__'\\n }\\n\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 在原地修改 dirs 以跳过不必要遍历的目录（隐藏目录及排除目录）\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查指定声明内容是否在文件第一行中\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果发生读取错误或解码错误等，说明不满足合法的文本代码文件要求\\n return False\\n \\n # 如果找到了至少一个代码文件并且都包含该声明，则要求被满足\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用HTML、CSS和JavaScript写一个本地表格助手的基本骨架。包含两个主要的视图：一个是主页视图，显示一个顶栏（包含“新建表格”按钮），下方是一个表格列表的占位区域；另一个是表格详情视图，包含“返回主页”按钮、一个占位的表格标题区和一个基础的3x3空表格HTML结构。现阶段不需要写LocalStorage相关的存储逻辑，只要能实现这两个视图的相互切换和基础的UI排版即可，页面要整洁美观一点。另外请你在后续所有操作中始终遵循两项要求：第一，你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖；第二，在生成的每一个代码文件里，单行允许的最大字符数为120。","instruction_following_checklist":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，目标代码可能包含 HTML, CSS, JS，\n # 同时考虑到测试用例可能使用的情况，纳入常见的代码扩展名。\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 使用 utf-8-sig 可以自动处理可能存在的 UTF-8 BOM，防止影响首行文本读取\n with open(file_path, 'r', encoding='utf-8-sig') as f:\n first_line = f.readline().strip()\n # 检查首行是否包含规定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于任何原因(如编码不匹配、权限不够等)读取失败，视为未满足要求\n return False\n \n # 如果没有生成任何代码文件，则认为未满足验证目标\n return has_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的每一个代码文件里，单行允许的最大字符数为120","validation_code":"import os\n\ndef check_line_length(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 定义常见的代码文件后缀名\n code_extensions = {\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.svelte',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 判断是否为代码文件\n if any(file.endswith(ext) for ext in code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去除行末的换行符后统计单行的字符数\n clean_line = line.rstrip('\\r\\n')\n if len(clean_line) > 120:\n return False\n except UnicodeDecodeError:\n # 如果不是普通的文本/代码文件编码，直接忽略\n continue\n except Exception:\n continue\n \n # 如果存在被检查的代码文件，且没有发现超过 120 字符的行，则返回 True\n return has_code_file","tags":["Quantity","Range"]}]},{"round_id":1,"instruction":"基础框架不错。接下来我们先把主页的逻辑和本地存储（LocalStorage）做起来。要求点击“新建表格”时，自动生成一个带有唯一ID和默认标题（比如“未命名表格_时间戳”）的空表格数据（默认3行3列），并立刻存入LocalStorage。主页的列表要读取LocalStorage里的所有表格并展示出来，每一项要显示标题、创建时间，并提供“打开”和“删除”两个按钮。点击“删除”能立刻弹出确认框，确认后从列表和本地存储中彻底移除该表格。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为100；第二，代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径。","instruction_following_checklist":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 针对用户指令涉及的前端项目和测试场景，枚举可能的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.ts', '.py', '.sh'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 检查真正的首行中是否包含目标编码声明\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果由于任何原因读文件异常，保守起见判断为验证失败\n return False\n \n # 只要验证到了至少一个代码文件且都满足要求即视为通过，否则不通过\n return has_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的每一个代码文件里，单行允许的最大字符数为100","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 根据用户指令推断可能涉及的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.sh'}\n \n # 遍历代码项目地址下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n # 获取文件扩展名\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n for line in f:\n # 去除行末换行符后统计当前行的字符数\n if len(line.rstrip('\\r\\n')) > 100:\n return False\n except Exception:\n # 如果个别文件读取出错则跳过，不阻断整体验证\n pass\n \n return True","tags":["Quantity","Range"]},{"constraint":"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":2,"instruction":"现在来完善表格详情视图的数据绑定和基础编辑功能。在主页点击某个表格的“打开”按钮后，进入详情视图并用LocalStorage里的该表数据动态渲染HTML table。要求表格的标题和所有单元格都是可直接点击编辑的（比如点击文字后变成输入框，或者使用contenteditable），失去焦点后完成编辑。最关键的是，任何对标题或单元格内容的修改，都必须实时更新当前内存数据并自动覆盖保存进LocalStorage中。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为80；第二，生成代码文件时的限制条件：绝对不能使用while循环。","instruction_following_checklist":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 常见的代码文件扩展名集合\n target_exts = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.sh'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除第三方依赖目录或版本控制目录\n path_parts = root.split(os.sep)\n if 'node_modules' in path_parts or '.git' in path_parts:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对代码文件进行验证\n if ext in target_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含特定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果遇到文件无法使用utf-8读取或其他异常，也视为不满足要求\n return False\n \n # 如果工作区内存在代码文件，并且每一个代码文件的首行均包含该声明，则返回True\n return has_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的每一个代码文件里，单行允许的最大字符数为80","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n \n for line in lines:\n # 去除行末尾的回车与换行符，不将其计入单行字符数\n clean_line = line.rstrip('\\r\\n')\n # 明确要求统计\"字符数\"，故直接使用 len() 函数\n if len(clean_line) > 80:\n return False\n except Exception:\n # 如果遇到文件读取异常（如编码问题等），安全跳过该文件继续验证\n continue\n\n return True","tags":["Quantity","Range"]},{"constraint":"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成代码文件时的限制条件：绝对不能使用while循环","validation_code":"import os\nimport re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n # 若路径不存在，说明未生成任何文件，此时可视为没有使用 while 循环。\n return True\n\n def contains_while_loop(content: str) -> bool:\n # 使用联合正则表达式按顺序匹配所有的注释和字符串字面量，防止因嵌套关系引起误删或误判\n pattern = re.compile(\n r'<!--[\\s\\S]*?-->|' # HTML 注释\n r'/\\*[\\s\\S]*?\\*/|' # JS/CSS 多行注释\n r'//.*|' # JS/TS 单行注释\n r'#.*|' # Python 单行注释\n r'\"\"\"[\\s\\S]*?\"\"\"|' # Python 多行字符串/注释 (双引号)\n r\"'''[\\s\\S]*?'''|\" # Python 多行字符串/注释 (单引号)\n r'\"(?:\\\\.|[^\"\\\\])*\"|' # 双引号字符串\n r\"'(?:\\\\.|[^'\\\\])*'|\" # 单引号字符串\n r'`(?:\\\\.|[^`\\\\])*`' # ES6 模板字符串 (反引号)\n )\n \n # 将匹配到的字符串字面量和注释全部替换为空白\n clean_content = pattern.sub('', content)\n \n # 查找清除无关干扰后，是否还保留有 while 关键字\n if re.search(r'\\bwhile\\b', clean_content):\n return True\n \n return False\n\n # 常见前端及可能相关的代码文件扩展名白名单\n valid_extensions = {'.js', '.ts', '.html', '.htm', '.css', '.jsx', '.tsx', '.vue', '.py'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 只要任何一个文件内使用了 while 循环，即判定为不满足要求\n if contains_while_loop(content):\n return False\n except Exception:\n # 忽略读取异常的文件，继续检查其他文件\n continue\n \n # 如果所有的代码文件均未包含 while 关键字，则验证通过\n return True","tags":["Style","Paradigm"]}]},{"round_id":3,"instruction":"好，接下来给详情视图增加行列的增删功能。在表格上方增加四个操作按钮：“在上方插入行”、“在下方插入行”、“在左侧插入列”、“在右侧插入列”，以及“删除当前行”、“删除当前列”按钮。用户需要先点击表格中的某一个单元格使其处于“选中”状态（加上明显的高亮边框），然后再点击这些操作按钮，在对应位置执行行列结构的变化。记得每一次结构变动都要自动保存到LocalStorage，并重新渲染表格。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为50；第二，你输出的代码文件里禁止使用for循环结构。","instruction_following_checklist":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，直接返回 False\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义需要检查的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\n # 排除不需要检查的典型目录\n exclude_dirs = {'node_modules', '.git', '__pycache__', '.venv', 'venv'}\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉排除目录，防止深入遍历\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含规定的编码声明格式\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果读取遇到异常（比如编码异常），同样视为未满足要求\n return False\n \n # 至少要找到了代码文件且均符合要求才返回 True\n return found_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的每一个代码文件里，单行允许的最大字符数为50","validation_code":"def check_line_length_limit(response: str, workspace_path: str) -> bool:\n import os\n\n valid_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n # 遍历工作目录下的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n # 判断是否属于需要检查的代码文件\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 剔除行尾的换行符，避免干扰字符数统计\n clean_line = line.replace('\\n', '').replace('\\r', '')\n \n # 题目明确要求为“字符数”，使用内置 len 函数统计\n if len(clean_line) > 50:\n return False\n except Exception:\n # 忽略读取时遇到的非utf-8编码等潜在异常\n continue\n \n # 如果没有找到任何代码文件，则无法证明其满足了需求\n return has_code_file","tags":["Quantity","Range"]},{"constraint":"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成代码文件时的限制条件：绝对不能使用while循环","validation_code":"def check_no_while_loop(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 可能包含 JavaScript 逻辑或 HTML 结构的文件扩展名集合\n target_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx'}\n \n # 匹配独立的 while 关键字\n while_pattern = re.compile(r'\\bwhile\\b')\n\n if not os.path.exists(workspace_path):\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 1. 去除 HTML 注释: <!-- ... -->\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 2. 去除 C/JS 风格的多行注释: /* ... */\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 3. 去除 C/JS 风格的单行注释: // ...\n content = re.sub(r'//.*', '', content)\n \n # 4. 去除字符串字面量，防止其中的 'while' 单词干扰验证\n # 匹配双引号字符串\n content = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '\"\"', content)\n # 匹配单引号字符串\n content = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", \"''\", content)\n # 匹配反引号模板字符串（可能跨多行，需加 DOTALL）\n content = re.sub(r\"`(?:\\\\.|[^`\\\\])*`\", \"``\", content, flags=re.DOTALL)\n \n # 5. 在去除了注释和字符串内容的纯代码中检查是否含有 while 关键字\n if while_pattern.search(content):\n return False\n except Exception:\n # 如果遇到文件读取编码错误等情况则跳过当前文件\n continue\n\n # 遍历所有代码文件均未发现 while 循环，要求满足\n return True","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n has_for_loop = False\n \n # 遍历工作空间中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 针对前端可能包含 JavaScript 逻辑的代码文件\n if file.endswith(('.js', '.html', '.ts', '.jsx', '.tsx', '.vue')):\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 1. 移除字符串字面量 (支持单引号、双引号及 JS 反引号，允许跨行且处理转义)\n content = re.sub(\n r'\"(?:\\\\.|[^\\\\\"])*\"|\\'(?:\\\\.|[^\\\\\\'])*\\'|`(?:\\\\.|[^\\\\`])*`', \n '', \n content, \n flags=re.DOTALL\n )\n # 2. 移除多行注释和HTML注释\n content = re.sub(\n r'/\\*.*?\\*/|<!--.*?-->', \n '', \n content, \n flags=re.DOTALL\n )\n # 3. 移除单行注释 (这里不使用 re.DOTALL，从而只匹配到换行符前)\n content = re.sub(r'//.*', '', content)\n \n # 4. 查找 for 循环的关键字 (涵盖 for (...) 以及 for await (...))\n if re.search(r'\\bfor\\s*(?:await\\s*)?\\(', content):\n has_for_loop = True\n break\n except Exception:\n # 忽略读取或编码引发的异常，继续处理下一个文件\n continue\n \n if has_for_loop:\n break\n \n # 题目要求“禁止使用for循环结构”，因此如果存在for循环返回False，否则返回True\n return not has_for_loop","tags":["Style","Paradigm"]}]},{"round_id":4,"instruction":"下一步是个挑战：实现“合并单元格”功能。首先，我希望支持鼠标拖拽来选中多个相邻的单元格（呈现出批量选中的淡蓝色高亮样式，只能选中矩形区域）。然后，在操作区增加一个“合并单元格”按钮。点击后，将选中的矩形区域通过修改HTML的 `colspan` 和 `rowspan` 属性合并为一个大单元格，保留左上角单元格的文本内容，其余被覆盖的单元格数据需在结构中进行处理或隐藏。合并完成后，务必将包含行列合并信息的最新DOM结构对应的数据保存进LocalStorage。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为40；第二，必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率。","instruction_following_checklist":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断的代码文件类型\n code_extensions = {'.html', '.css', '.js', '.py', '.sh'}\n has_code_file = False\n \n # 如果工作区目录不存在，直接返回False\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n\n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件后缀并转换为小写进行匹配\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n first_line = f.readline()\n # 检查首行是否包含指定的编码声明\n if \"-*- coding: utf-8 -*-\" not in first_line:\n return False\n except Exception:\n # 如果由于任何原因无法读取首行（例如权限异常等），认为不满足要求\n return False\n\n # 必须至少存在一个代码文件，且所有代码文件的首行都符合要求\n return has_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的每一个代码文件里，单行允许的最大字符数为40","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 推断可能的代码文件扩展名\n extensions = {'.html', '.htm', '.css', '.js'}\n \n # 如果工作区路径不存在或不是目录，则无法验证代码文件，按要求未完成处理（或者可视为失败）\n if not os.path.isdir(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 仅去掉末尾的换行符，其余空白字符（如空格或Tab）也算作字符长度的一部分\n clean_line = line.rstrip('\\r\\n')\n # 要求中明确说明“字符数”，因此使用 len() 函数\n if len(clean_line) > 40:\n return False\n except Exception:\n # 如果由于特殊编码或其他原因无法读取某文件，则跳过继续检查\n continue\n \n # 所有文件验证均未发现超出 40 字符的行\n return True","tags":["Quantity","Range"]},{"constraint":"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成代码文件时的限制条件：绝对不能使用while循环","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 定义可能包含在项目中的代码文件扩展名\n code_extensions = {\n '.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', \n '.py', '.java', '.c', '.cpp', '.cs', '.php', '.rb', '.go'\n }\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 1. 移除 HTML 注释\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n # 2. 移除常见的 C-style 多行注释 (如 /* ... */)\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n # 3. 移除 Python 风格的三引号多行字符串/注释\n content = re.sub(r'\"\"\"(.*?)\"\"\"', '', content, flags=re.DOTALL)\n content = re.sub(r\"'''(.*?)'''\", '', content, flags=re.DOTALL)\n # 4. 移除字符串字面量 (双引号、单引号及反引号)，支持字符串内部的反斜杠转义\n string_pattern = r'\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`'\n content = re.sub(string_pattern, '', content, flags=re.DOTALL)\n # 5. 移除单行注释 (如 // ... 或 # ...)\n content = re.sub(r'//.*', '', content)\n content = re.sub(r'#.*', '', content)\n \n # 检查清理后的纯代码文本中是否包含 while 关键字\n if re.search(r'\\bwhile\\b', content):\n return False\n except Exception:\n # 忽略读取、解码失败的异常文件\n continue\n\n return True","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def has_for_loop(file_path: str) -> bool:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n if file_path.endswith('.py'):\n # 剔除 Python 文件中的注释和字符串字面量\n content = re.sub(r'#.*', '', content)\n content = re.sub(r'\\'\\'\\'.*?\\'\\'\\'', '', content, flags=re.DOTALL)\n content = re.sub(r'\\\"\\\"\\\".*?\\\"\\\"\\\"', '', content, flags=re.DOTALL)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", content)\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', content)\n \n # 匹配 Python 中的 for 循环结构 (例如: for x in y:)\n if re.search(r'\\bfor\\s+.*?\\bin\\b', content):\n return True\n else:\n # 剔除 JS/TS/HTML 文件中的注释和字符串字面量\n content = re.sub(r'/\\*.*?\\*/', '', content, flags=re.DOTALL)\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n content = re.sub(r'//.*', '', content)\n content = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", content)\n content = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', content)\n content = re.sub(r'`([^`\\\\]|\\\\.)*`', '``', content)\n \n # 匹配 JS/TS 等语法中的 for 循环结构 (例如: for(let i=0...), for await (...), for (const key in ...))\n # 注意：\\bfor\\b 后紧接可选的 await，并且随后存在左括号 '(' \n if re.search(r'\\bfor\\b\\s*(?:await\\s+)?\\(', content):\n return True\n \n except Exception:\n pass\n \n return False\n\n # 遍历当前工作区的所有相关代码文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.py')):\n file_path = os.path.join(root, file)\n if has_for_loop(file_path):\n # 如果任一文件存在 for 循环，则违背了禁止使用的要求，返回 False\n return False\n \n # 如果所有的代码文件中均未发现 for 循环结构，则说明严格遵循了限制，返回 True\n return True","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":5,"instruction":"最后一步，实现与合并对应的“取消合并”功能。在操作区加上“取消合并”按钮。如果用户选中了一个已经合并的大单元格（即包含 colspan 或 rowspan 大于1的单元格），点击该按钮后，要将这个单元格拆分，恢复成合并之前的多个基础1x1单元格组成的网格结构（恢复出来的多余位置内容留空即可）。同样，拆分后的网格状态也要立即渲染并同步自动保存到LocalStorage中。这样我们的本地表格助手就大功告成了。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为35；第二，你提供的每个代码文件里，注释内容须100%使用中文撰写。","instruction_following_checklist":[{"constraint":"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见的代码文件后缀\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.py', '.java', '.c', '.cpp', '.h'}\n target_declaration = \"-*- coding: utf-8 -*-\"\n \n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 排除掉通常不需要检查的隐藏或依赖目录\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查首行是否包含目标声明字符串\n if target_declaration not in first_line:\n return False\n except Exception:\n # 如果由于编码等问题无法读取，也视作不符合要求\n return False\n \n # 如果找到了代码文件且均通过了检查，则返回 True；如果没有生成任何代码文件，则返回 False\n return has_code_file","tags":["Content","Format"]},{"constraint":"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>","validation_code":"","tags":["Content","Format"]},{"constraint":"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况","validation_code":"","tags":["Workflow","Tool Usage"]},{"constraint":"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖","validation_code":"","tags":["Style","Paradigm"]},{"constraint":"在生成的每一个代码文件里，单行允许的最大字符数为35","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.py', '.sh'}\n found_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n for line in f:\n # 去除行尾的回车符和换行符以精准统计单行字符数\n clean_line = line.rstrip('\\r\\n')\n # 根据要求明确说明统计“字符数”，使用len函数\n if len(clean_line) > 35:\n return False\n except Exception:\n # 忽略解码错误或其他文件读取异常\n continue\n \n # 如果遍历了所有代码文件均未发现超长行，且确实找到了代码文件，则判定为满足要求\n return found_code_file","tags":["Quantity","Range"]},{"constraint":"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成代码文件时的限制条件：绝对不能使用while循环","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def remove_comments_and_strings(code: str) -> str:\n \"\"\"\n 移除JS/TS代码中的单行注释、多行注释，以及所有字符串字面量（单引号、双引号、反引号）\n 以防止干扰关键字 while 的检测。\n \"\"\"\n pattern = re.compile(\n r'//[^\\n]*|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"|`(?:\\\\.|[^\\\\`])*`',\n re.DOTALL\n )\n return re.sub(pattern, '', code)\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 检查独立的脚本文件\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n cleaned_content = remove_comments_and_strings(content)\n \n if re.search(r'\\bwhile\\b', cleaned_content):\n return False\n except Exception:\n continue\n \n # 检查HTML文件中内联的脚本文件\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 提取 <script> 标签内的 JS 代码块\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.IGNORECASE | re.DOTALL)\n for script in scripts:\n cleaned_script = remove_comments_and_strings(script)\n \n if re.search(r'\\bwhile\\b', cleaned_script):\n return False\n except Exception:\n continue\n\n return True","tags":["Style","Paradigm"]},{"constraint":"你输出的代码文件里禁止使用for循环结构","validation_code":"def check_for_loop_forbidden(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # js_html_pattern 适用于 JavaScript 和 HTML：\n # - string 分组：匹配双引号、单引号、反引号的字符串（支持转义字符）\n # - comment 分组：匹配 /* */ 多行注释, // 单行注释, <!-- --> HTML 注释\n # - for 分组：匹配 for 关键字后跟可选空白和左括号，精准定位 JS/HTML 中的循环结构\n js_html_pattern = re.compile(\n r'(?P<string>\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\'|`(?:\\\\.|[^`\\\\])*`)'\n r'|(?P<comment>/\\*[\\s\\S]*?\\*/|//.*|<!--[\\s\\S]*?-->)'\n r'|(?P<for>\\bfor\\s*\\()'\n )\n \n # py_pattern 适用于 Python（针对可能生成的测试文件）：\n # - string 分组：匹配三双引号、三单引号、以及普通单/双引号字符串\n # - comment 分组：匹配 # 及其之后的单行注释\n # - for 分组：匹配单独的 for 关键字\n py_pattern = re.compile(\n r'(?P<string>\"\"\"[\\s\\S]*?\"\"\"|\\'\\'\\'[\\s\\S]*?\\'\\'\\'|\"(?:\\\\.|[^\"\\\\])*\"|\\'(?:\\\\.|[^\\'\\\\])*\\')'\n r'|(?P<comment>#.*)'\n r'|(?P<for>\\bfor\\b)'\n )\n \n # 遍历当前工作区路径中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in {'.html', '.js', '.py'}:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 根据文件类型选择对应的正则处理器，利用 finditer 执行词法扫描验证\n if ext in {'.html', '.js'}:\n for match in js_html_pattern.finditer(content):\n if match.lastgroup == 'for':\n # 若命中 for 分组，说明存在于真实的代码逻辑中（非注释与字符串内）\n return False\n elif ext == '.py':\n for match in py_pattern.finditer(content):\n if match.lastgroup == 'for':\n return False\n \n return True","tags":["Style","Paradigm"]},{"constraint":"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用HTML、CSS和JavaScript写一个本地表格助手的基本骨架。包含两个主要的视图：一个是主页视图，显示一个顶栏（包含“新建表格”按钮），下方是一个表格列表的占位区域；另一个是表格详情视图，包含“返回主页”按钮、一个占位的表格标题区和一个基础的3x3空表格HTML结构。现阶段不需要写LocalStorage相关的存储逻辑，只要能实现这两个视图的相互切换和基础的UI排版即可，页面要整洁美观一点。另外请你在后续所有操作中始终遵循两项要求：第一，你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖；第二，在生成的每一个代码文件里，单行允许的最大字符数为120。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，目标代码可能包含 HTML, CSS, JS，\\n # 同时考虑到测试用例可能使用的情况，纳入常见的代码扩展名。\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 使用 utf-8-sig 可以自动处理可能存在的 UTF-8 BOM，防止影响首行文本读取\\n with open(file_path, 'r', encoding='utf-8-sig') as f:\\n first_line = f.readline().strip()\\n # 检查首行是否包含规定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何原因(如编码不匹配、权限不够等)读取失败，视为未满足要求\\n return False\\n \\n # 如果没有生成任何代码文件，则认为未满足验证目标\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的每一个代码文件里，单行允许的最大字符数为120\",\n \"validation_code\": \"import os\\n\\ndef check_line_length(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义常见的代码文件后缀名\\n code_extensions = {\\n '.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.vue', '.svelte',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs'\\n }\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否为代码文件\\n if any(file.endswith(ext) for ext in code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去除行末的换行符后统计单行的字符数\\n clean_line = line.rstrip('\\\\r\\\\n')\\n if len(clean_line) > 120:\\n return False\\n except UnicodeDecodeError:\\n # 如果不是普通的文本/代码文件编码，直接忽略\\n continue\\n except Exception:\\n continue\\n \\n # 如果存在被检查的代码文件，且没有发现超过 120 字符的行，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"基础框架不错。接下来我们先把主页的逻辑和本地存储（LocalStorage）做起来。要求点击“新建表格”时，自动生成一个带有唯一ID和默认标题（比如“未命名表格_时间戳”）的空表格数据（默认3行3列），并立刻存入LocalStorage。主页的列表要读取LocalStorage里的所有表格并展示出来，每一项要显示标题、创建时间，并提供“打开”和“删除”两个按钮。点击“删除”能立刻弹出确认框，确认后从列表和本地存储中彻底移除该表格。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为100；第二，代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 针对用户指令涉及的前端项目和测试场景，枚举可能的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.ts', '.py', '.sh'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 检查真正的首行中是否包含目标编码声明\\n if '-*- coding: utf-8 -*-' not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何原因读文件异常，保守起见判断为验证失败\\n return False\\n \\n # 只要验证到了至少一个代码文件且都满足要求即视为通过，否则不通过\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的每一个代码文件里，单行允许的最大字符数为100\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 根据用户指令推断可能涉及的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.sh'}\\n \\n # 遍历代码项目地址下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n # 获取文件扩展名\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n for line in f:\\n # 去除行末换行符后统计当前行的字符数\\n if len(line.rstrip('\\\\r\\\\n')) > 100:\\n return False\\n except Exception:\\n # 如果个别文件读取出错则跳过，不阻断整体验证\\n pass\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"现在来完善表格详情视图的数据绑定和基础编辑功能。在主页点击某个表格的“打开”按钮后，进入详情视图并用LocalStorage里的该表数据动态渲染HTML table。要求表格的标题和所有单元格都是可直接点击编辑的（比如点击文字后变成输入框，或者使用contenteditable），失去焦点后完成编辑。最关键的是，任何对标题或单元格内容的修改，都必须实时更新当前内存数据并自动覆盖保存进LocalStorage中。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为80；第二，生成代码文件时的限制条件：绝对不能使用while循环。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件扩展名集合\\n target_exts = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx', '.sh'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除第三方依赖目录或版本控制目录\\n path_parts = root.split(os.sep)\\n if 'node_modules' in path_parts or '.git' in path_parts:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对代码文件进行验证\\n if ext in target_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含特定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果遇到文件无法使用utf-8读取或其他异常，也视为不满足要求\\n return False\\n \\n # 如果工作区内存在代码文件，并且每一个代码文件的首行均包含该声明，则返回True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的每一个代码文件里，单行允许的最大字符数为80\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n \\n for line in lines:\\n # 去除行末尾的回车与换行符，不将其计入单行字符数\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 明确要求统计\\\"字符数\\\"，故直接使用 len() 函数\\n if len(clean_line) > 80:\\n return False\\n except Exception:\\n # 如果遇到文件读取异常（如编码问题等），安全跳过该文件继续验证\\n continue\\n\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时的限制条件：绝对不能使用while循环\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n # 若路径不存在，说明未生成任何文件，此时可视为没有使用 while 循环。\\n return True\\n\\n def contains_while_loop(content: str) -> bool:\\n # 使用联合正则表达式按顺序匹配所有的注释和字符串字面量，防止因嵌套关系引起误删或误判\\n pattern = re.compile(\\n r'<!--[\\\\s\\\\S]*?-->|' # HTML 注释\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|' # JS/CSS 多行注释\\n r'//.*|' # JS/TS 单行注释\\n r'#.*|' # Python 单行注释\\n r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|' # Python 多行字符串/注释 (双引号)\\n r\\\"'''[\\\\s\\\\S]*?'''|\\\" # Python 多行字符串/注释 (单引号)\\n r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|' # 双引号字符串\\n r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'|\\\" # 单引号字符串\\n r'`(?:\\\\\\\\.|[^`\\\\\\\\])*`' # ES6 模板字符串 (反引号)\\n )\\n \\n # 将匹配到的字符串字面量和注释全部替换为空白\\n clean_content = pattern.sub('', content)\\n \\n # 查找清除无关干扰后，是否还保留有 while 关键字\\n if re.search(r'\\\\bwhile\\\\b', clean_content):\\n return True\\n \\n return False\\n\\n # 常见前端及可能相关的代码文件扩展名白名单\\n valid_extensions = {'.js', '.ts', '.html', '.htm', '.css', '.jsx', '.tsx', '.vue', '.py'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 只要任何一个文件内使用了 while 循环，即判定为不满足要求\\n if contains_while_loop(content):\\n return False\\n except Exception:\\n # 忽略读取异常的文件，继续检查其他文件\\n continue\\n \\n # 如果所有的代码文件均未包含 while 关键字，则验证通过\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"好，接下来给详情视图增加行列的增删功能。在表格上方增加四个操作按钮：“在上方插入行”、“在下方插入行”、“在左侧插入列”、“在右侧插入列”，以及“删除当前行”、“删除当前列”按钮。用户需要先点击表格中的某一个单元格使其处于“选中”状态（加上明显的高亮边框），然后再点击这些操作按钮，在对应位置执行行列结构的变化。记得每一次结构变动都要自动保存到LocalStorage，并重新渲染表格。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为50；第二，你输出的代码文件里禁止使用for循环结构。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，直接返回 False\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义需要检查的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.py', '.ts', '.jsx', '.tsx'}\\n # 排除不需要检查的典型目录\\n exclude_dirs = {'node_modules', '.git', '__pycache__', '.venv', 'venv'}\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉排除目录，防止深入遍历\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含规定的编码声明格式\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果读取遇到异常（比如编码异常），同样视为未满足要求\\n return False\\n \\n # 至少要找到了代码文件且均符合要求才返回 True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的每一个代码文件里，单行允许的最大字符数为50\",\n \"validation_code\": \"def check_line_length_limit(response: str, workspace_path: str) -> bool:\\n import os\\n\\n valid_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n # 遍历工作目录下的所有文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n # 判断是否属于需要检查的代码文件\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 剔除行尾的换行符，避免干扰字符数统计\\n clean_line = line.replace('\\\\n', '').replace('\\\\r', '')\\n \\n # 题目明确要求为“字符数”，使用内置 len 函数统计\\n if len(clean_line) > 50:\\n return False\\n except Exception:\\n # 忽略读取时遇到的非utf-8编码等潜在异常\\n continue\\n \\n # 如果没有找到任何代码文件，则无法证明其满足了需求\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时的限制条件：绝对不能使用while循环\",\n \"validation_code\": \"def check_no_while_loop(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 可能包含 JavaScript 逻辑或 HTML 结构的文件扩展名集合\\n target_extensions = {'.js', '.html', '.htm', '.ts', '.jsx', '.tsx'}\\n \\n # 匹配独立的 while 关键字\\n while_pattern = re.compile(r'\\\\bwhile\\\\b')\\n\\n if not os.path.exists(workspace_path):\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 1. 去除 HTML 注释: <!-- ... -->\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 2. 去除 C/JS 风格的多行注释: /* ... */\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 3. 去除 C/JS 风格的单行注释: // ...\\n content = re.sub(r'//.*', '', content)\\n \\n # 4. 去除字符串字面量，防止其中的 'while' 单词干扰验证\\n # 匹配双引号字符串\\n content = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '\\\"\\\"', content)\\n # 匹配单引号字符串\\n content = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", \\\"''\\\", content)\\n # 匹配反引号模板字符串（可能跨多行，需加 DOTALL）\\n content = re.sub(r\\\"`(?:\\\\\\\\.|[^`\\\\\\\\])*`\\\", \\\"``\\\", content, flags=re.DOTALL)\\n \\n # 5. 在去除了注释和字符串内容的纯代码中检查是否含有 while 关键字\\n if while_pattern.search(content):\\n return False\\n except Exception:\\n # 如果遇到文件读取编码错误等情况则跳过当前文件\\n continue\\n\\n # 遍历所有代码文件均未发现 while 循环，要求满足\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n has_for_loop = False\\n \\n # 遍历工作空间中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 针对前端可能包含 JavaScript 逻辑的代码文件\\n if file.endswith(('.js', '.html', '.ts', '.jsx', '.tsx', '.vue')):\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 1. 移除字符串字面量 (支持单引号、双引号及 JS 反引号，允许跨行且处理转义)\\n content = re.sub(\\n r'\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|`(?:\\\\\\\\.|[^\\\\\\\\`])*`', \\n '', \\n content, \\n flags=re.DOTALL\\n )\\n # 2. 移除多行注释和HTML注释\\n content = re.sub(\\n r'/\\\\*.*?\\\\*/|<!--.*?-->', \\n '', \\n content, \\n flags=re.DOTALL\\n )\\n # 3. 移除单行注释 (这里不使用 re.DOTALL，从而只匹配到换行符前)\\n content = re.sub(r'//.*', '', content)\\n \\n # 4. 查找 for 循环的关键字 (涵盖 for (...) 以及 for await (...))\\n if re.search(r'\\\\bfor\\\\s*(?:await\\\\s*)?\\\\(', content):\\n has_for_loop = True\\n break\\n except Exception:\\n # 忽略读取或编码引发的异常，继续处理下一个文件\\n continue\\n \\n if has_for_loop:\\n break\\n \\n # 题目要求“禁止使用for循环结构”，因此如果存在for循环返回False，否则返回True\\n return not has_for_loop\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"下一步是个挑战：实现“合并单元格”功能。首先，我希望支持鼠标拖拽来选中多个相邻的单元格（呈现出批量选中的淡蓝色高亮样式，只能选中矩形区域）。然后，在操作区增加一个“合并单元格”按钮。点击后，将选中的矩形区域通过修改HTML的 `colspan` 和 `rowspan` 属性合并为一个大单元格，保留左上角单元格的文本内容，其余被覆盖的单元格数据需在结构中进行处理或隐藏。合并完成后，务必将包含行列合并信息的最新DOM结构对应的数据保存进LocalStorage。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为40；第二，必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断的代码文件类型\\n code_extensions = {'.html', '.css', '.js', '.py', '.sh'}\\n has_code_file = False\\n \\n # 如果工作区目录不存在，直接返回False\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n\\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件后缀并转换为小写进行匹配\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n first_line = f.readline()\\n # 检查首行是否包含指定的编码声明\\n if \\\"-*- coding: utf-8 -*-\\\" not in first_line:\\n return False\\n except Exception:\\n # 如果由于任何原因无法读取首行（例如权限异常等），认为不满足要求\\n return False\\n\\n # 必须至少存在一个代码文件，且所有代码文件的首行都符合要求\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的每一个代码文件里，单行允许的最大字符数为40\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 推断可能的代码文件扩展名\\n extensions = {'.html', '.htm', '.css', '.js'}\\n \\n # 如果工作区路径不存在或不是目录，则无法验证代码文件，按要求未完成处理（或者可视为失败）\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 仅去掉末尾的换行符，其余空白字符（如空格或Tab）也算作字符长度的一部分\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 要求中明确说明“字符数”，因此使用 len() 函数\\n if len(clean_line) > 40:\\n return False\\n except Exception:\\n # 如果由于特殊编码或其他原因无法读取某文件，则跳过继续检查\\n continue\\n \\n # 所有文件验证均未发现超出 40 字符的行\\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时的限制条件：绝对不能使用while循环\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 定义可能包含在项目中的代码文件扩展名\\n code_extensions = {\\n '.html', '.js', '.css', '.ts', '.jsx', '.tsx', '.vue', \\n '.py', '.java', '.c', '.cpp', '.cs', '.php', '.rb', '.go'\\n }\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 1. 移除 HTML 注释\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n # 2. 移除常见的 C-style 多行注释 (如 /* ... */)\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n # 3. 移除 Python 风格的三引号多行字符串/注释\\n content = re.sub(r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', '', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'''(.*?)'''\\\", '', content, flags=re.DOTALL)\\n # 4. 移除字符串字面量 (双引号、单引号及反引号)，支持字符串内部的反斜杠转义\\n string_pattern = r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`'\\n content = re.sub(string_pattern, '', content, flags=re.DOTALL)\\n # 5. 移除单行注释 (如 // ... 或 # ...)\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r'#.*', '', content)\\n \\n # 检查清理后的纯代码文本中是否包含 while 关键字\\n if re.search(r'\\\\bwhile\\\\b', content):\\n return False\\n except Exception:\\n # 忽略读取、解码失败的异常文件\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def has_for_loop(file_path: str) -> bool:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n if file_path.endswith('.py'):\\n # 剔除 Python 文件中的注释和字符串字面量\\n content = re.sub(r'#.*', '', content)\\n content = re.sub(r'\\\\'\\\\'\\\\'.*?\\\\'\\\\'\\\\'', '', content, flags=re.DOTALL)\\n content = re.sub(r'\\\\\\\"\\\\\\\"\\\\\\\".*?\\\\\\\"\\\\\\\"\\\\\\\"', '', content, flags=re.DOTALL)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content)\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content)\\n \\n # 匹配 Python 中的 for 循环结构 (例如: for x in y:)\\n if re.search(r'\\\\bfor\\\\s+.*?\\\\bin\\\\b', content):\\n return True\\n else:\\n # 剔除 JS/TS/HTML 文件中的注释和字符串字面量\\n content = re.sub(r'/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n content = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n content = re.sub(r'//.*', '', content)\\n content = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", content)\\n content = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', content)\\n content = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', '``', content)\\n \\n # 匹配 JS/TS 等语法中的 for 循环结构 (例如: for(let i=0...), for await (...), for (const key in ...))\\n # 注意：\\\\bfor\\\\b 后紧接可选的 await，并且随后存在左括号 '(' \\n if re.search(r'\\\\bfor\\\\b\\\\s*(?:await\\\\s+)?\\\\(', content):\\n return True\\n \\n except Exception:\\n pass\\n \\n return False\\n\\n # 遍历当前工作区的所有相关代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.py')):\\n file_path = os.path.join(root, file)\\n if has_for_loop(file_path):\\n # 如果任一文件存在 for 循环，则违背了禁止使用的要求，返回 False\\n return False\\n \\n # 如果所有的代码文件中均未发现 for 循环结构，则说明严格遵循了限制，返回 True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后一步，实现与合并对应的“取消合并”功能。在操作区加上“取消合并”按钮。如果用户选中了一个已经合并的大单元格（即包含 colspan 或 rowspan 大于1的单元格），点击该按钮后，要将这个单元格拆分，恢复成合并之前的多个基础1x1单元格组成的网格结构（恢复出来的多余位置内容留空即可）。同样，拆分后的网格状态也要立即渲染并同步自动保存到LocalStorage中。这样我们的本地表格助手就大功告成了。另外请你在后续所有操作中始终遵循两项更新要求：第一，在生成的每一个代码文件里，单行允许的最大字符数为35；第二，你提供的每个代码文件里，注释内容须100%使用中文撰写。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"必须保证生成的每个代码文件都带有此编码声明，请将其统一放在首行：-*- coding: utf-8 -*-\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见的代码文件后缀\\n code_extensions = {'.html', '.css', '.js', '.jsx', '.ts', '.tsx', '.py', '.java', '.c', '.cpp', '.h'}\\n target_declaration = \\\"-*- coding: utf-8 -*-\\\"\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 排除掉通常不需要检查的隐藏或依赖目录\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n first_line = f.readline()\\n # 检查首行是否包含目标声明字符串\\n if target_declaration not in first_line:\\n return False\\n except Exception:\\n # 如果由于编码等问题无法读取，也视作不符合要求\\n return False\\n \\n # 如果找到了代码文件且均通过了检查，则返回 True；如果没有生成任何代码文件，则返回 False\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"只要涉及查找处理文件，就必须尽量借助 Task 工具（设定 subagent_type=Explore），而不是直接去执行文件操作或搜索命令\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"文件操作的原则是必须尽量使用专用工具（如Read, Write, Edit），而不是通过 bash 命令来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"在对代码进行修改之前，必须要调用 TodoWrite 工具完成任务规划。一旦某任务的代码修改（Write/Edit）操作完毕，你下一步的首个合法调用必须是使用 TodoComplete 把该任务标为完成，这中间禁止掺杂其他任务的代码修改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在完成对生成代码的每一次测试后，请在回复里严格遵循给定的XML格式，分别输出执行结果与描述：<result>通过/失败/部分失败/无法验证</result><description>xxxx</description>\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Content\",\n \"Format\"\n ]\n },\n {\n \"constraint\": \"必须在任何 bash 测试、安装以及构建命令的起始位置加入 `timeout 60s` 包装，以免出现进程挂起的情况\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Tool Usage\"\n ]\n },\n {\n \"constraint\": \"不要丢弃你对生成代码进行的所有中间测试代码，请将它们保留并放置在项目目录下的__tests__子目录里\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"请确保生成的单元测试文件运行后，能够让目标代码的行覆盖率大于等于85%\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的代码只能实现当前确切说明的功能，避免过度设计和预先编码未来特性，还要确保不包含任何未被调用的代码与依赖\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"在生成的每一个代码文件里，单行允许的最大字符数为35\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.py', '.sh'}\\n found_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n for line in f:\\n # 去除行尾的回车符和换行符以精准统计单行字符数\\n clean_line = line.rstrip('\\\\r\\\\n')\\n # 根据要求明确说明统计“字符数”，使用len函数\\n if len(clean_line) > 35:\\n return False\\n except Exception:\\n # 忽略解码错误或其他文件读取异常\\n continue\\n \\n # 如果遍历了所有代码文件均未发现超长行，且确实找到了代码文件，则判定为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Range\"\n ]\n },\n {\n \"constraint\": \"代码文件中的文件引用仅限相对路径，绝对不允许使用绝对路径\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成代码文件时的限制条件：绝对不能使用while循环\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def remove_comments_and_strings(code: str) -> str:\\n \\\"\\\"\\\"\\n 移除JS/TS代码中的单行注释、多行注释，以及所有字符串字面量（单引号、双引号、反引号）\\n 以防止干扰关键字 while 的检测。\\n \\\"\\\"\\\"\\n pattern = re.compile(\\n r'//[^\\\\n]*|/\\\\*.*?\\\\*/|\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'|\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"|`(?:\\\\\\\\.|[^\\\\\\\\`])*`',\\n re.DOTALL\\n )\\n return re.sub(pattern, '', code)\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 检查独立的脚本文件\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx')):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n cleaned_content = remove_comments_and_strings(content)\\n \\n if re.search(r'\\\\bwhile\\\\b', cleaned_content):\\n return False\\n except Exception:\\n continue\\n \\n # 检查HTML文件中内联的脚本文件\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 提取 <script> 标签内的 JS 代码块\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.IGNORECASE | re.DOTALL)\\n for script in scripts:\\n cleaned_script = remove_comments_and_strings(script)\\n \\n if re.search(r'\\\\bwhile\\\\b', cleaned_script):\\n return False\\n except Exception:\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"你输出的代码文件里禁止使用for循环结构\",\n \"validation_code\": \"def check_for_loop_forbidden(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # js_html_pattern 适用于 JavaScript 和 HTML：\\n # - string 分组：匹配双引号、单引号、反引号的字符串（支持转义字符）\\n # - comment 分组：匹配 /* */ 多行注释, // 单行注释, <!-- --> HTML 注释\\n # - for 分组：匹配 for 关键字后跟可选空白和左括号，精准定位 JS/HTML 中的循环结构\\n js_html_pattern = re.compile(\\n r'(?P<string>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\'|`(?:\\\\\\\\.|[^`\\\\\\\\])*`)'\\n r'|(?P<comment>/\\\\*[\\\\s\\\\S]*?\\\\*/|//.*|<!--[\\\\s\\\\S]*?-->)'\\n r'|(?P<for>\\\\bfor\\\\s*\\\\()'\\n )\\n \\n # py_pattern 适用于 Python（针对可能生成的测试文件）：\\n # - string 分组：匹配三双引号、三单引号、以及普通单/双引号字符串\\n # - comment 分组：匹配 # 及其之后的单行注释\\n # - for 分组：匹配单独的 for 关键字\\n py_pattern = re.compile(\\n r'(?P<string>\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|\\\\'\\\\'\\\\'[\\\\s\\\\S]*?\\\\'\\\\'\\\\'|\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"|\\\\'(?:\\\\\\\\.|[^\\\\'\\\\\\\\])*\\\\')'\\n r'|(?P<comment>#.*)'\\n r'|(?P<for>\\\\bfor\\\\b)'\\n )\\n \\n # 遍历当前工作区路径中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in {'.html', '.js', '.py'}:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 根据文件类型选择对应的正则处理器，利用 finditer 执行词法扫描验证\\n if ext in {'.html', '.js'}:\\n for match in js_html_pattern.finditer(content):\\n if match.lastgroup == 'for':\\n # 若命中 for 分组，说明存在于真实的代码逻辑中（非注释与字符串内）\\n return False\\n elif ext == '.py':\\n for match in py_pattern.finditer(content):\\n if match.lastgroup == 'for':\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Paradigm\"\n ]\n },\n {\n \"constraint\": \"必须尽最大可能采取并行的策略执行多个独立的工具调用，以此来优化效率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["在浏览器中打开页面，观察主页视图是否正常显示，包含顶部的导航栏、明确的'新建表格'按钮，以及下方用于展示表格列表的区域。","点击主页的'新建表格'按钮，观察下方表格列表是否立即增加了一条记录；该记录需显示默认标题（包含'未命名表格'和时间戳信息）、创建时间，并提供'打开'和'删除'两个操作按钮。","在上一步新建表格后，刷新浏览器页面，观察刚才新增的表格记录是否依然保留在表格列表中（验证 LocalStorage 新增数据的本地存储功能）。","在主页表格列表中，点击某一条记录的'删除'按钮，确认是否会弹出系统的确认提示框；点击'确认'后，观察该记录是否从列表中消失，且刷新页面后该记录也没有恢复（验证删除逻辑及本地存储同步）。","在主页表格列表中，点击某一条记录的'打开'按钮，观察页面是否切换到了表格详情视图；详情视图需包含'返回主页'按钮、一个可显示的表格标题，以及初始状态下为3行3列的空白基础表格结构。","在表格详情视图中，点击'返回主页'按钮，观察页面是否能正确跳回主页视图并展示原有的表格列表。","在表格详情视图中，点击表格标题文字使其进入编辑状态，修改标题文本（如修改为'测试表格'），随后点击页面空白处使其失去焦点。刷新浏览器页面，观察标题是否依旧显示为刚刚修改的内容（验证标题实时自动保存）。","在表格的任意一个空白单元格中点击并输入测试文本（如'A1数据'），随后点击页面空白处使其失去焦点。刷新浏览器页面，观察该单元格内是否依旧显示'A1数据'（验证单元格内容编辑和实时自动保存）。","在表格详情视图中，用鼠标单击表格中的某一个单元格，观察该单元格周围是否会出现明显的高亮边框，以标识该单元格处于'选中'状态。","选中表格中间的某个单元格，点击操作区的'在上方插入行'和'在下方插入行'按钮，观察表格总行数是否增加，并且新的空白行正确地出现在被选中单元格的上方和下方。","选中表格中间的某个单元格，点击操作区的'在左侧插入列'和'在右侧插入列'按钮，观察表格总列数是否增加，并且新的空白列正确地出现在被选中单元格的左侧和右侧。","在刚刚执行过插入行列操作的表格页面，刷新浏览器，观察新增的行列网格结构是否依然保留（验证增加行列结构的本地存储同步）。","在某个单元格中输入特征字符（如'将删行'），选中该单元格，点击'删除当前行'按钮。观察包含该特征字符的整行是否从表格中消失，且刷新浏览器后该行未恢复。","在某个单元格中输入特征字符（如'将删列'），选中该单元格，点击'删除当前列'按钮。观察包含该特征字符的整列是否从表格中消失，且刷新浏览器后该列未恢复。","在表格中按住鼠标左键并在相邻的几个单元格之间拖拽，松开鼠标后，观察鼠标划过的这一矩形区域内的单元格是否都呈现出特殊的批量选中样式（例如淡蓝色的背景高亮）。","在相邻的四个单元格（2x2）中输入文字，确保左上角单元格填入'保留文本'。用鼠标拖拽选中这四个单元格，点击'合并单元格'按钮。观察这四个单元格是否合并成了一个大单元格，且大单元格内仅显示'保留文本'，原先其他三个单元格的内容消失或被覆盖。","完成上一步的合并单元格操作后，刷新浏览器页面，观察合并后的大单元格跨行跨列状态及内容是否保持不变（验证合并单元格结构的本地数据持久化）。","单击选中之前合并生成的大单元格，点击'取消合并'按钮。观察该大单元格是否被重新拆分成原来的1x1基础网格结构，原先的文本保留在拆分出的左上角单元格中，其余新恢复的单元格内容为空白。","完成上一步的取消合并操作后，刷新浏览器页面，观察表格是否依然维持拆分后的网格结构布局（验证取消合并操作后的 DOM 结构持久化更新）。"],"string":"[\n \"在浏览器中打开页面，观察主页视图是否正常显示，包含顶部的导航栏、明确的'新建表格'按钮，以及下方用于展示表格列表的区域。\",\n \"点击主页的'新建表格'按钮，观察下方表格列表是否立即增加了一条记录；该记录需显示默认标题（包含'未命名表格'和时间戳信息）、创建时间，并提供'打开'和'删除'两个操作按钮。\",\n \"在上一步新建表格后，刷新浏览器页面，观察刚才新增的表格记录是否依然保留在表格列表中（验证 LocalStorage 新增数据的本地存储功能）。\",\n \"在主页表格列表中，点击某一条记录的'删除'按钮，确认是否会弹出系统的确认提示框；点击'确认'后，观察该记录是否从列表中消失，且刷新页面后该记录也没有恢复（验证删除逻辑及本地存储同步）。\",\n \"在主页表格列表中，点击某一条记录的'打开'按钮，观察页面是否切换到了表格详情视图；详情视图需包含'返回主页'按钮、一个可显示的表格标题，以及初始状态下为3行3列的空白基础表格结构。\",\n \"在表格详情视图中，点击'返回主页'按钮，观察页面是否能正确跳回主页视图并展示原有的表格列表。\",\n \"在表格详情视图中，点击表格标题文字使其进入编辑状态，修改标题文本（如修改为'测试表格'），随后点击页面空白处使其失去焦点。刷新浏览器页面，观察标题是否依旧显示为刚刚修改的内容（验证标题实时自动保存）。\",\n \"在表格的任意一个空白单元格中点击并输入测试文本（如'A1数据'），随后点击页面空白处使其失去焦点。刷新浏览器页面，观察该单元格内是否依旧显示'A1数据'（验证单元格内容编辑和实时自动保存）。\",\n \"在表格详情视图中，用鼠标单击表格中的某一个单元格，观察该单元格周围是否会出现明显的高亮边框，以标识该单元格处于'选中'状态。\",\n \"选中表格中间的某个单元格，点击操作区的'在上方插入行'和'在下方插入行'按钮，观察表格总行数是否增加，并且新的空白行正确地出现在被选中单元格的上方和下方。\",\n \"选中表格中间的某个单元格，点击操作区的'在左侧插入列'和'在右侧插入列'按钮，观察表格总列数是否增加，并且新的空白列正确地出现在被选中单元格的左侧和右侧。\",\n \"在刚刚执行过插入行列操作的表格页面，刷新浏览器，观察新增的行列网格结构是否依然保留（验证增加行列结构的本地存储同步）。\",\n \"在某个单元格中输入特征字符（如'将删行'），选中该单元格，点击'删除当前行'按钮。观察包含该特征字符的整行是否从表格中消失，且刷新浏览器后该行未恢复。\",\n \"在某个单元格中输入特征字符（如'将删列'），选中该单元格，点击'删除当前列'按钮。观察包含该特征字符的整列是否从表格中消失，且刷新浏览器后该列未恢复。\",\n \"在表格中按住鼠标左键并在相邻的几个单元格之间拖拽，松开鼠标后，观察鼠标划过的这一矩形区域内的单元格是否都呈现出特殊的批量选中样式（例如淡蓝色的背景高亮）。\",\n \"在相邻的四个单元格（2x2）中输入文字，确保左上角单元格填入'保留文本'。用鼠标拖拽选中这四个单元格，点击'合并单元格'按钮。观察这四个单元格是否合并成了一个大单元格，且大单元格内仅显示'保留文本'，原先其他三个单元格的内容消失或被覆盖。\",\n \"完成上一步的合并单元格操作后，刷新浏览器页面，观察合并后的大单元格跨行跨列状态及内容是否保持不变（验证合并单元格结构的本地数据持久化）。\",\n \"单击选中之前合并生成的大单元格，点击'取消合并'按钮。观察该大单元格是否被重新拆分成原来的1x1基础网格结构，原先的文本保留在拆分出的左上角单元格中，其余新恢复的单元格内容为空白。\",\n \"完成上一步的取消合并操作后，刷新浏览器页面，观察表格是否依然维持拆分后的网格结构布局（验证取消合并操作后的 DOM 结构持久化更新）。\"\n]","isConversational":false}}},{"rowIdx":87,"cells":{"task_id":{"kind":"number","value":88,"string":"88"},"repository_policy":{"kind":"string","value":"以下是你在后续所有轮次都需要严格遵循的约束条件：\n1. 只要是你创建的代码文件，就必须使用UTF-8进行编码。\n2. 你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7。\n3. 对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线。\n4. 对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串。\n5. 对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则。\n6. 确保你生成的每一个代码文件都精确包含 3 行注释。\n7. 对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求。\n8. 必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"只要是你创建的代码文件，就必须使用UTF-8进行编码","validation_code":"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区路径不存在，说明没有生成文件，没有违背该条件\n if not os.path.exists(workspace_path):\n return True\n\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.php', '.go', \n '.rs', '.rb', '.swift', '.kt', '.sh', '.bat', '.sql', '.json', \n '.xml', '.yaml', '.yml'\n }\n\n # 遍历项目目录中的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 检查是否为代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 抛出异常代表文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的异常，如权限错误等\n pass\n \n # 所有代码文件均成功通过 UTF-8 解码验证\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7","validation_code":"def check_variable_length_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n code_snippets = []\n \n # 1. 提取 response 中以 Markdown 格式包裹的代码块\n blocks = re.findall(r'```(?P<lang>\\w*)\\n(?P<code>.*?)```', response, flags=re.DOTALL)\n for lang, code in blocks:\n code_snippets.append((lang.lower(), code))\n \n # 2. 提取 workspace_path 中生成的各种常见的代码文件\n valid_exts = {'.py', '.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.sh'}\n if os.path.isdir(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code_snippets.append((ext.lstrip('.'), f.read()))\n except Exception:\n pass\n \n # 如果根本没有提取到任何代码，无法判定满足了该生成的约束\n if not code_snippets:\n return False\n \n def extract_vars_from_python(code: str) -> set:\n \"\"\"通过 AST 解析 Python 代码，精确提取出所有的变量名\"\"\"\n vars_set = set()\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n # 普通变量赋值\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n vars_set.add(node.id)\n # 函数参数\n elif isinstance(node, ast.arg):\n vars_set.add(node.arg)\n # Exception捕获的变量名\n elif isinstance(node, ast.ExceptHandler) and node.name:\n vars_set.add(node.name)\n # Global 声明\n elif isinstance(node, ast.Global):\n for name in node.names:\n vars_set.add(name)\n # Nonlocal 声明\n elif isinstance(node, ast.Nonlocal):\n for name in node.names:\n vars_set.add(name)\n # 类的实例属性定义\n elif isinstance(node, ast.Attribute) and isinstance(node.ctx, ast.Store):\n vars_set.add(node.attr)\n # Python 3.10+ 的模式匹配变量\n elif type(node).__name__ == 'MatchAs' and hasattr(node, 'name') and node.name:\n vars_set.add(node.name)\n elif type(node).__name__ == 'MatchStar' and hasattr(node, 'name') and node.name:\n vars_set.add(node.name)\n elif type(node).__name__ == 'MatchMapping' and hasattr(node, 'rest') and node.rest:\n vars_set.add(node.rest)\n except Exception:\n pass\n return vars_set\n\n def extract_vars_from_regex(code: str) -> set:\n \"\"\"降级方案：对于非Python语言或其他难以解析的代码，通过正则表达式粗略提取变量名\"\"\"\n vars_set = set()\n \n # 匹配 js/ts 的变量声明\n js_vars = re.findall(r'\\b(?:var|let|const)\\s+([a-zA-Z_]\\w*)\\b', code)\n vars_set.update(js_vars)\n \n # 匹配 类似 C/Java 的基本类型变量声明\n c_vars = re.findall(r'\\b(?:int|float|double|char|boolean|bool|String|long|short|byte)\\s+([a-zA-Z_]\\w*)\\s*(?:=|;|,)', code)\n vars_set.update(c_vars)\n \n # 匹配常见的赋值语句操作符左侧的变量，排除了类似 ==, !=, >=, <= 的干扰\n eq_vars = re.findall(r'\\b([a-zA-Z_]\\w*)\\s*(?<![!<>=+\\-*/%&|^])=(?!=)\\s*', code)\n op_eq_vars = re.findall(r'\\b([a-zA-Z_]\\w*)\\s*(?:\\+|-|\\*|/|%|&|\\||\\^|<<|>>)=\\s*', code)\n \n # 常见编程语言的保留字和关键字，防止被误计入变量\n keywords = {'if', 'else', 'while', 'for', 'return', 'class', 'def', 'function', 'switch', 'case', 'break', 'continue', 'true', 'false', 'null', 'undefined', 'this', 'self'}\n \n for v in (eq_vars + op_eq_vars):\n if v not in keywords:\n vars_set.add(v)\n \n return vars_set\n\n # 3. 验证提取出的变量长度\n for lang, code in code_snippets:\n if not code.strip():\n continue\n \n vars_set = set()\n \n # 根据语言分类使用合适的变量提取器\n if lang in ['python', 'py']:\n vars_set = extract_vars_from_python(code)\n if not vars_set: # 防止因为语法错误导致 AST 提取失败，启动备用正则方案\n vars_set = extract_vars_from_regex(code)\n elif lang == '':\n # 可能是没写标注类型的 python\n vars_set = extract_vars_from_python(code)\n if not vars_set:\n vars_set = extract_vars_from_regex(code)\n else:\n vars_set = extract_vars_from_regex(code)\n \n # 检查是否每个变量名的字符数均符合 [2, 7] 的要求\n for var in vars_set:\n if len(var) < 2 or len(var) > 7:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线","validation_code":"def check_class_names(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 包含通常允许有类声明的常见代码文件后缀\n valid_exts = {\n '.py', '.java', '.cpp', '.hpp', '.c', '.h', '.cc', '.cxx', \n '.cs', '.js', '.ts', '.jsx', '.tsx', '.php', '.rb', '.swift', \n '.kt', '.scala', '.go', '.rs', '.dart', '.m', '.mm'\n }\n\n # 用于提取代码中类定义的正则表达式\n # 适配大多数面向对象编程语言中通过 class 关键字定义的类\n class_pattern = re.compile(\n r'^\\s*(?:(?:public|private|protected|internal|abstract|final|static|export|default|sealed|partial)\\s+)*class\\s+([A-Za-z0-9_]+)',\n re.MULTILINE\n )\n\n # 检查提取到的类名是否满足要求：只能是小写字母以及下划线\n def is_valid_class_name(name: str) -> bool:\n return bool(re.fullmatch(r'[a-z_]+', name))\n\n # 1. 检查 workspace_path 下生成的代码文件\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in valid_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n class_names = class_pattern.findall(content)\n for name in class_names:\n if not is_valid_class_name(name):\n return False\n\n # 2. 为了防范生成的代码片段尚未落盘的情况，同步检查 response 里的 markdown 代码块\n code_blocks = re.findall(r'```[a-zA-Z0-9_]*\\n(.*?)\\n```', response, re.DOTALL)\n for block in code_blocks:\n class_names = class_pattern.findall(block)\n for name in class_names:\n if not is_valid_class_name(name):\n return False\n\n # 若所有类名都校验通过或未检索到类，默认代表要求满足\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串","validation_code":"def check_functions_palindrome(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n \n if not os.path.exists(workspace_path):\n return False\n \n functions = []\n \n # 遍历项目文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 1. Python 文件处理（优先使用AST）\n if ext == '.py':\n try:\n tree = ast.parse(content)\n for node in ast.walk(tree):\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\n functions.append(node.name)\n continue \n except SyntaxError:\n pass # 如果有语法错误，退化为正则匹配\n \n # 简化内容：去除 C风格/Python风格 等大部分注释，减少正则误报\n clean_content = re.sub(r'//.*|/\\*.*?\\*/', '', content, flags=re.DOTALL)\n clean_content = re.sub(r'#.*', '', clean_content)\n \n # 2. 针对各语言的函数名称正则提取\n if ext == '.py':\n matches = re.findall(r'^\\s*def\\s+([A-Za-z_]\\w*)\\s*\\(', clean_content, re.MULTILINE)\n functions.extend(matches)\n \n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n # JS 箭头函数\n matches_arrow = re.findall(r'\\b([A-Za-z_]\\w*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[A-Za-z_]\\w*)\\s*=>', clean_content)\n functions.extend(matches_arrow)\n \n # JS 赋值函数\n matches_assign = re.findall(r'\\b([A-Za-z_]\\w*)\\s*[:=]\\s*(?:async\\s+)?function\\b', clean_content)\n functions.extend(matches_assign)\n \n # 常见带 { 的函数声明形式\n matches_c = re.finditer(r'(new\\s+)?\\b([A-Za-z_]\\w*)\\s*\\([^;{}]*\\)\\s*\\{', clean_content)\n keywords = {'if', 'while', 'for', 'switch', 'catch', 'return', 'else', 'try', 'finally', 'function', 'typeof', 'sizeof'}\n for match in matches_c:\n if match.group(1): # 排除 new Class(...) {} 的匿名内部类形式\n continue\n name = match.group(2)\n if name not in keywords:\n functions.append(name)\n \n elif ext == '.go':\n matches = re.findall(r'\\bfunc\\s+(?:\\([^)]+\\)\\s+)?([A-Za-z_]\\w*)\\s*\\(', clean_content)\n functions.extend(matches)\n \n elif ext == '.php':\n matches = re.findall(r'\\bfunction\\s+([A-Za-z_]\\w*)\\s*\\(', clean_content)\n functions.extend(matches)\n \n elif ext == '.rs':\n matches = re.findall(r'\\bfn\\s+([A-Za-z_]\\w*)\\s*\\(', clean_content)\n functions.extend(matches)\n \n elif ext == '.rb':\n matches = re.findall(r'^\\s*def\\s+([A-Za-z_]\\w*)', clean_content, re.MULTILINE)\n functions.extend(matches)\n \n elif ext in ['.sh', '.bash']:\n matches = re.findall(r'\\b([A-Za-z_]\\w*)\\s*\\(\\s*\\)\\s*\\{', clean_content)\n functions.extend(matches)\n matches_function = re.findall(r'\\bfunction\\s+([A-Za-z_]\\w*)', clean_content)\n functions.extend(matches_function)\n \n elif ext in ['.java', '.cpp', '.c', '.cs', '.swift', '.kt']:\n matches_c = re.finditer(r'(new\\s+)?\\b([A-Za-z_]\\w*)\\s*\\([^;{}]*\\)\\s*\\{', clean_content)\n keywords = {'if', 'while', 'for', 'switch', 'catch', 'return', 'else', 'try', 'finally', 'synchronized', 'sizeof', 'typeof'}\n for match in matches_c:\n if match.group(1): # 排除 new Class(...) {}\n continue\n name = match.group(2)\n if name not in keywords:\n functions.append(name)\n \n # 如果项目中无任何有效代码和函数，属于符合 \"任何一个函数的名称都必须...\" 的情况（全称命题对空集为真）\n if not functions:\n return True\n \n # 3. 验证查找到的所有函数名称是否都严格为回文字符串\n for name in functions:\n if name != name[::-1]:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保你生成的每一个代码文件都精确包含 3 行注释","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import io\n import tokenize\n\n code_extensions = {\n '.py': 'python', \n '.rb': 'hash', '.sh': 'hash', '.pl': 'hash', '.r': 'hash',\n '.c': 'slash', '.cpp': 'slash', '.h': 'slash', '.hpp': 'slash',\n '.java': 'slash', '.js': 'slash', '.ts': 'slash', '.cs': 'slash',\n '.go': 'slash', '.php': 'slash', '.swift': 'slash', '.kt': 'slash',\n '.rs': 'slash', '.dart': 'slash', '.scala': 'slash',\n '.html': 'html', '.htm': 'html', '.xml': 'html', '.vue': 'html',\n '.css': 'css'\n }\n\n def count_comment_lines(file_path: str) -> int:\n ext = os.path.splitext(file_path)[1].lower()\n if ext not in code_extensions:\n return -1\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return -1\n\n comment_style = code_extensions[ext]\n \n # Python 特殊处理，利用自带 tokenize 实现精确识别\n if comment_style == 'python':\n comment_lines = set()\n try:\n tokens = tokenize.generate_tokens(io.StringIO(content).readline)\n for tok in tokens:\n if tok.type == tokenize.COMMENT:\n comment_lines.add(tok.start[0])\n except tokenize.TokenError:\n pass\n except Exception:\n pass\n return len(comment_lines)\n\n lines = content.split('\\n')\n count = 0\n in_block = False\n \n for line in lines:\n stripped = line.strip()\n # 空行若处在块注释内，算作块注释的其中一行\n if not stripped and in_block:\n count += 1\n continue\n \n if comment_style == 'hash':\n in_str = False\n str_char = ''\n is_comment = False\n for i, char in enumerate(line):\n if char in (\"'\", '\"'):\n # 检查转义\n if i > 0 and line[i-1] == '\\\\':\n backslash_count = 0\n idx = i - 1\n while idx >= 0 and line[idx] == '\\\\':\n backslash_count += 1\n idx -= 1\n if backslash_count % 2 != 0:\n continue\n if not in_str:\n in_str = True\n str_char = char\n elif str_char == char:\n in_str = False\n elif char == '#' and not in_str:\n is_comment = True\n break\n if is_comment:\n count += 1\n \n elif comment_style == 'slash':\n if in_block:\n count += 1\n if '*/' in line:\n in_block = False\n else:\n in_str = False\n str_char = ''\n is_comment = False\n i = 0\n while i < len(line):\n char = line[i]\n if char in (\"'\", '\"', '`'):\n if i > 0 and line[i-1] == '\\\\':\n backslash_count = 0\n idx = i - 1\n while idx >= 0 and line[idx] == '\\\\':\n backslash_count += 1\n idx -= 1\n if backslash_count % 2 != 0:\n i += 1\n continue\n if not in_str:\n in_str = True\n str_char = char\n elif str_char == char:\n in_str = False\n elif char == '/' and not in_str and i + 1 < len(line):\n if line[i+1] == '/':\n is_comment = True\n break\n elif line[i+1] == '*':\n is_comment = True\n in_block = True\n if '*/' in line[i+2:]:\n in_block = False\n break\n i += 1\n if is_comment:\n count += 1\n \n elif comment_style == 'html':\n if in_block:\n count += 1\n if '-->' in line:\n in_block = False\n else:\n if '<!--' in line:\n count += 1\n in_block = True\n if '-->' in line.split('<!--', 1)[1]:\n in_block = False\n \n elif comment_style == 'css':\n if in_block:\n count += 1\n if '*/' in line:\n in_block = False\n else:\n if '/*' in line:\n count += 1\n in_block = True\n if '*/' in line.split('/*', 1)[1]:\n in_block = False\n\n return count\n\n code_file_count = 0\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n comments_num = count_comment_lines(file_path)\n # 如果不属于已知代码文件类型或者无法读取，返回值为 -1，跳过\n if comments_num != -1:\n code_file_count += 1\n if comments_num != 3:\n return False\n \n # 没有找到任何代码文件同样视为未达成要求\n if code_file_count == 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n class_names = re.findall(r'\\bclass\\s+([A-Za-z0-9_]+)', content)\n for class_name in class_names:\n if len(class_name) % 2 != 0:\n return False\n except Exception:\n continue\n \n return True","tags":["Style","Naming"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件后缀名集合\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cxx', '.h', '.hpp', '.cs',\n '.js', '.ts', '.html', '.css', '.go', '.rb', '.php', '.sh',\n '.swift', '.kt', '.rs', '.dart', '.sql', '.xml', '.json'\n }\n \n if not os.path.exists(workspace_path):\n return False\n\n has_code_file = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 通过 \\n 分割文件内容计算总行数，能够准确把控包含文件末尾的空行在内的所有空行情况\n # (Python 默认开启 Universal Newlines 模式，会将 \\r\\n 自动转为 \\n)\n lines = content.split('\\n')\n total_lines = len(lines)\n \n # 如果行数是偶数，则直接不满足要求\n if total_lines % 2 == 0:\n return False\n except Exception:\n # 如果出现读取错误（如非UTF-8编码报错等），视为不满足要求\n return False\n \n # 必须至少生成了一个代码文件，且所有代码文件行数均为奇数才返回 True\n return has_code_file","tags":["Quantity","Exact Value"]}],"string":"[\n {\n \"constraint\": \"只要是你创建的代码文件，就必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区路径不存在，说明没有生成文件，没有违背该条件\\n if not os.path.exists(workspace_path):\\n return True\\n\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \\n '.js', '.ts', '.jsx', '.tsx', '.html', '.css', '.php', '.go', \\n '.rs', '.rb', '.swift', '.kt', '.sh', '.bat', '.sql', '.json', \\n '.xml', '.yaml', '.yml'\\n }\\n\\n # 遍历项目目录中的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 检查是否为代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 抛出异常代表文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的异常，如权限错误等\\n pass\\n \\n # 所有代码文件均成功通过 UTF-8 解码验证\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7\",\n \"validation_code\": \"def check_variable_length_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n code_snippets = []\\n \\n # 1. 提取 response 中以 Markdown 格式包裹的代码块\\n blocks = re.findall(r'```(?P<lang>\\\\w*)\\\\n(?P<code>.*?)```', response, flags=re.DOTALL)\\n for lang, code in blocks:\\n code_snippets.append((lang.lower(), code))\\n \\n # 2. 提取 workspace_path 中生成的各种常见的代码文件\\n valid_exts = {'.py', '.js', '.ts', '.java', '.c', '.cpp', '.cs', '.go', '.rb', '.php', '.swift', '.kt', '.rs', '.sh'}\\n if os.path.isdir(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code_snippets.append((ext.lstrip('.'), f.read()))\\n except Exception:\\n pass\\n \\n # 如果根本没有提取到任何代码，无法判定满足了该生成的约束\\n if not code_snippets:\\n return False\\n \\n def extract_vars_from_python(code: str) -> set:\\n \\\"\\\"\\\"通过 AST 解析 Python 代码，精确提取出所有的变量名\\\"\\\"\\\"\\n vars_set = set()\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n # 普通变量赋值\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n vars_set.add(node.id)\\n # 函数参数\\n elif isinstance(node, ast.arg):\\n vars_set.add(node.arg)\\n # Exception捕获的变量名\\n elif isinstance(node, ast.ExceptHandler) and node.name:\\n vars_set.add(node.name)\\n # Global 声明\\n elif isinstance(node, ast.Global):\\n for name in node.names:\\n vars_set.add(name)\\n # Nonlocal 声明\\n elif isinstance(node, ast.Nonlocal):\\n for name in node.names:\\n vars_set.add(name)\\n # 类的实例属性定义\\n elif isinstance(node, ast.Attribute) and isinstance(node.ctx, ast.Store):\\n vars_set.add(node.attr)\\n # Python 3.10+ 的模式匹配变量\\n elif type(node).__name__ == 'MatchAs' and hasattr(node, 'name') and node.name:\\n vars_set.add(node.name)\\n elif type(node).__name__ == 'MatchStar' and hasattr(node, 'name') and node.name:\\n vars_set.add(node.name)\\n elif type(node).__name__ == 'MatchMapping' and hasattr(node, 'rest') and node.rest:\\n vars_set.add(node.rest)\\n except Exception:\\n pass\\n return vars_set\\n\\n def extract_vars_from_regex(code: str) -> set:\\n \\\"\\\"\\\"降级方案：对于非Python语言或其他难以解析的代码，通过正则表达式粗略提取变量名\\\"\\\"\\\"\\n vars_set = set()\\n \\n # 匹配 js/ts 的变量声明\\n js_vars = re.findall(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_]\\\\w*)\\\\b', code)\\n vars_set.update(js_vars)\\n \\n # 匹配 类似 C/Java 的基本类型变量声明\\n c_vars = re.findall(r'\\\\b(?:int|float|double|char|boolean|bool|String|long|short|byte)\\\\s+([a-zA-Z_]\\\\w*)\\\\s*(?:=|;|,)', code)\\n vars_set.update(c_vars)\\n \\n # 匹配常见的赋值语句操作符左侧的变量，排除了类似 ==, !=, >=, <= 的干扰\\n eq_vars = re.findall(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*(?<![!<>=+\\\\-*/%&|^])=(?!=)\\\\s*', code)\\n op_eq_vars = re.findall(r'\\\\b([a-zA-Z_]\\\\w*)\\\\s*(?:\\\\+|-|\\\\*|/|%|&|\\\\||\\\\^|<<|>>)=\\\\s*', code)\\n \\n # 常见编程语言的保留字和关键字，防止被误计入变量\\n keywords = {'if', 'else', 'while', 'for', 'return', 'class', 'def', 'function', 'switch', 'case', 'break', 'continue', 'true', 'false', 'null', 'undefined', 'this', 'self'}\\n \\n for v in (eq_vars + op_eq_vars):\\n if v not in keywords:\\n vars_set.add(v)\\n \\n return vars_set\\n\\n # 3. 验证提取出的变量长度\\n for lang, code in code_snippets:\\n if not code.strip():\\n continue\\n \\n vars_set = set()\\n \\n # 根据语言分类使用合适的变量提取器\\n if lang in ['python', 'py']:\\n vars_set = extract_vars_from_python(code)\\n if not vars_set: # 防止因为语法错误导致 AST 提取失败，启动备用正则方案\\n vars_set = extract_vars_from_regex(code)\\n elif lang == '':\\n # 可能是没写标注类型的 python\\n vars_set = extract_vars_from_python(code)\\n if not vars_set:\\n vars_set = extract_vars_from_regex(code)\\n else:\\n vars_set = extract_vars_from_regex(code)\\n \\n # 检查是否每个变量名的字符数均符合 [2, 7] 的要求\\n for var in vars_set:\\n if len(var) < 2 or len(var) > 7:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线\",\n \"validation_code\": \"def check_class_names(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 包含通常允许有类声明的常见代码文件后缀\\n valid_exts = {\\n '.py', '.java', '.cpp', '.hpp', '.c', '.h', '.cc', '.cxx', \\n '.cs', '.js', '.ts', '.jsx', '.tsx', '.php', '.rb', '.swift', \\n '.kt', '.scala', '.go', '.rs', '.dart', '.m', '.mm'\\n }\\n\\n # 用于提取代码中类定义的正则表达式\\n # 适配大多数面向对象编程语言中通过 class 关键字定义的类\\n class_pattern = re.compile(\\n r'^\\\\s*(?:(?:public|private|protected|internal|abstract|final|static|export|default|sealed|partial)\\\\s+)*class\\\\s+([A-Za-z0-9_]+)',\\n re.MULTILINE\\n )\\n\\n # 检查提取到的类名是否满足要求：只能是小写字母以及下划线\\n def is_valid_class_name(name: str) -> bool:\\n return bool(re.fullmatch(r'[a-z_]+', name))\\n\\n # 1. 检查 workspace_path 下生成的代码文件\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in valid_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n class_names = class_pattern.findall(content)\\n for name in class_names:\\n if not is_valid_class_name(name):\\n return False\\n\\n # 2. 为了防范生成的代码片段尚未落盘的情况，同步检查 response 里的 markdown 代码块\\n code_blocks = re.findall(r'```[a-zA-Z0-9_]*\\\\n(.*?)\\\\n```', response, re.DOTALL)\\n for block in code_blocks:\\n class_names = class_pattern.findall(block)\\n for name in class_names:\\n if not is_valid_class_name(name):\\n return False\\n\\n # 若所有类名都校验通过或未检索到类，默认代表要求满足\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串\",\n \"validation_code\": \"def check_functions_palindrome(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n functions = []\\n \\n # 遍历项目文件\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 1. Python 文件处理（优先使用AST）\\n if ext == '.py':\\n try:\\n tree = ast.parse(content)\\n for node in ast.walk(tree):\\n if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):\\n functions.append(node.name)\\n continue \\n except SyntaxError:\\n pass # 如果有语法错误，退化为正则匹配\\n \\n # 简化内容：去除 C风格/Python风格 等大部分注释，减少正则误报\\n clean_content = re.sub(r'//.*|/\\\\*.*?\\\\*/', '', content, flags=re.DOTALL)\\n clean_content = re.sub(r'#.*', '', clean_content)\\n \\n # 2. 针对各语言的函数名称正则提取\\n if ext == '.py':\\n matches = re.findall(r'^\\\\s*def\\\\s+([A-Za-z_]\\\\w*)\\\\s*\\\\(', clean_content, re.MULTILINE)\\n functions.extend(matches)\\n \\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n # JS 箭头函数\\n matches_arrow = re.findall(r'\\\\b([A-Za-z_]\\\\w*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[A-Za-z_]\\\\w*)\\\\s*=>', clean_content)\\n functions.extend(matches_arrow)\\n \\n # JS 赋值函数\\n matches_assign = re.findall(r'\\\\b([A-Za-z_]\\\\w*)\\\\s*[:=]\\\\s*(?:async\\\\s+)?function\\\\b', clean_content)\\n functions.extend(matches_assign)\\n \\n # 常见带 { 的函数声明形式\\n matches_c = re.finditer(r'(new\\\\s+)?\\\\b([A-Za-z_]\\\\w*)\\\\s*\\\\([^;{}]*\\\\)\\\\s*\\\\{', clean_content)\\n keywords = {'if', 'while', 'for', 'switch', 'catch', 'return', 'else', 'try', 'finally', 'function', 'typeof', 'sizeof'}\\n for match in matches_c:\\n if match.group(1): # 排除 new Class(...) {} 的匿名内部类形式\\n continue\\n name = match.group(2)\\n if name not in keywords:\\n functions.append(name)\\n \\n elif ext == '.go':\\n matches = re.findall(r'\\\\bfunc\\\\s+(?:\\\\([^)]+\\\\)\\\\s+)?([A-Za-z_]\\\\w*)\\\\s*\\\\(', clean_content)\\n functions.extend(matches)\\n \\n elif ext == '.php':\\n matches = re.findall(r'\\\\bfunction\\\\s+([A-Za-z_]\\\\w*)\\\\s*\\\\(', clean_content)\\n functions.extend(matches)\\n \\n elif ext == '.rs':\\n matches = re.findall(r'\\\\bfn\\\\s+([A-Za-z_]\\\\w*)\\\\s*\\\\(', clean_content)\\n functions.extend(matches)\\n \\n elif ext == '.rb':\\n matches = re.findall(r'^\\\\s*def\\\\s+([A-Za-z_]\\\\w*)', clean_content, re.MULTILINE)\\n functions.extend(matches)\\n \\n elif ext in ['.sh', '.bash']:\\n matches = re.findall(r'\\\\b([A-Za-z_]\\\\w*)\\\\s*\\\\(\\\\s*\\\\)\\\\s*\\\\{', clean_content)\\n functions.extend(matches)\\n matches_function = re.findall(r'\\\\bfunction\\\\s+([A-Za-z_]\\\\w*)', clean_content)\\n functions.extend(matches_function)\\n \\n elif ext in ['.java', '.cpp', '.c', '.cs', '.swift', '.kt']:\\n matches_c = re.finditer(r'(new\\\\s+)?\\\\b([A-Za-z_]\\\\w*)\\\\s*\\\\([^;{}]*\\\\)\\\\s*\\\\{', clean_content)\\n keywords = {'if', 'while', 'for', 'switch', 'catch', 'return', 'else', 'try', 'finally', 'synchronized', 'sizeof', 'typeof'}\\n for match in matches_c:\\n if match.group(1): # 排除 new Class(...) {}\\n continue\\n name = match.group(2)\\n if name not in keywords:\\n functions.append(name)\\n \\n # 如果项目中无任何有效代码和函数，属于符合 \\\"任何一个函数的名称都必须...\\\" 的情况（全称命题对空集为真）\\n if not functions:\\n return True\\n \\n # 3. 验证查找到的所有函数名称是否都严格为回文字符串\\n for name in functions:\\n if name != name[::-1]:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 3 行注释\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import io\\n import tokenize\\n\\n code_extensions = {\\n '.py': 'python', \\n '.rb': 'hash', '.sh': 'hash', '.pl': 'hash', '.r': 'hash',\\n '.c': 'slash', '.cpp': 'slash', '.h': 'slash', '.hpp': 'slash',\\n '.java': 'slash', '.js': 'slash', '.ts': 'slash', '.cs': 'slash',\\n '.go': 'slash', '.php': 'slash', '.swift': 'slash', '.kt': 'slash',\\n '.rs': 'slash', '.dart': 'slash', '.scala': 'slash',\\n '.html': 'html', '.htm': 'html', '.xml': 'html', '.vue': 'html',\\n '.css': 'css'\\n }\\n\\n def count_comment_lines(file_path: str) -> int:\\n ext = os.path.splitext(file_path)[1].lower()\\n if ext not in code_extensions:\\n return -1\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return -1\\n\\n comment_style = code_extensions[ext]\\n \\n # Python 特殊处理，利用自带 tokenize 实现精确识别\\n if comment_style == 'python':\\n comment_lines = set()\\n try:\\n tokens = tokenize.generate_tokens(io.StringIO(content).readline)\\n for tok in tokens:\\n if tok.type == tokenize.COMMENT:\\n comment_lines.add(tok.start[0])\\n except tokenize.TokenError:\\n pass\\n except Exception:\\n pass\\n return len(comment_lines)\\n\\n lines = content.split('\\\\n')\\n count = 0\\n in_block = False\\n \\n for line in lines:\\n stripped = line.strip()\\n # 空行若处在块注释内，算作块注释的其中一行\\n if not stripped and in_block:\\n count += 1\\n continue\\n \\n if comment_style == 'hash':\\n in_str = False\\n str_char = ''\\n is_comment = False\\n for i, char in enumerate(line):\\n if char in (\\\"'\\\", '\\\"'):\\n # 检查转义\\n if i > 0 and line[i-1] == '\\\\\\\\':\\n backslash_count = 0\\n idx = i - 1\\n while idx >= 0 and line[idx] == '\\\\\\\\':\\n backslash_count += 1\\n idx -= 1\\n if backslash_count % 2 != 0:\\n continue\\n if not in_str:\\n in_str = True\\n str_char = char\\n elif str_char == char:\\n in_str = False\\n elif char == '#' and not in_str:\\n is_comment = True\\n break\\n if is_comment:\\n count += 1\\n \\n elif comment_style == 'slash':\\n if in_block:\\n count += 1\\n if '*/' in line:\\n in_block = False\\n else:\\n in_str = False\\n str_char = ''\\n is_comment = False\\n i = 0\\n while i < len(line):\\n char = line[i]\\n if char in (\\\"'\\\", '\\\"', '`'):\\n if i > 0 and line[i-1] == '\\\\\\\\':\\n backslash_count = 0\\n idx = i - 1\\n while idx >= 0 and line[idx] == '\\\\\\\\':\\n backslash_count += 1\\n idx -= 1\\n if backslash_count % 2 != 0:\\n i += 1\\n continue\\n if not in_str:\\n in_str = True\\n str_char = char\\n elif str_char == char:\\n in_str = False\\n elif char == '/' and not in_str and i + 1 < len(line):\\n if line[i+1] == '/':\\n is_comment = True\\n break\\n elif line[i+1] == '*':\\n is_comment = True\\n in_block = True\\n if '*/' in line[i+2:]:\\n in_block = False\\n break\\n i += 1\\n if is_comment:\\n count += 1\\n \\n elif comment_style == 'html':\\n if in_block:\\n count += 1\\n if '-->' in line:\\n in_block = False\\n else:\\n if '<!--' in line:\\n count += 1\\n in_block = True\\n if '-->' in line.split('<!--', 1)[1]:\\n in_block = False\\n \\n elif comment_style == 'css':\\n if in_block:\\n count += 1\\n if '*/' in line:\\n in_block = False\\n else:\\n if '/*' in line:\\n count += 1\\n in_block = True\\n if '*/' in line.split('/*', 1)[1]:\\n in_block = False\\n\\n return count\\n\\n code_file_count = 0\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n comments_num = count_comment_lines(file_path)\\n # 如果不属于已知代码文件类型或者无法读取，返回值为 -1，跳过\\n if comments_num != -1:\\n code_file_count += 1\\n if comments_num != 3:\\n return False\\n \\n # 没有找到任何代码文件同样视为未达成要求\\n if code_file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n class_names = re.findall(r'\\\\bclass\\\\s+([A-Za-z0-9_]+)', content)\\n for class_name in class_names:\\n if len(class_name) % 2 != 0:\\n return False\\n except Exception:\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件后缀名集合\\n code_extensions = {\\n '.py', '.java', '.c', '.cpp', '.cxx', '.h', '.hpp', '.cs',\\n '.js', '.ts', '.html', '.css', '.go', '.rb', '.php', '.sh',\\n '.swift', '.kt', '.rs', '.dart', '.sql', '.xml', '.json'\\n }\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n has_code_file = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 通过 \\\\n 分割文件内容计算总行数，能够准确把控包含文件末尾的空行在内的所有空行情况\\n # (Python 默认开启 Universal Newlines 模式，会将 \\\\r\\\\n 自动转为 \\\\n)\\n lines = content.split('\\\\n')\\n total_lines = len(lines)\\n \\n # 如果行数是偶数，则直接不满足要求\\n if total_lines % 2 == 0:\\n return False\\n except Exception:\\n # 如果出现读取错误（如非UTF-8编码报错等），视为不满足要求\\n return False\\n \\n # 必须至少生成了一个代码文件，且所有代码文件行数均为奇数才返回 True\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Data Analysis"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用HTML、CSS和原生JavaScript写一个“两个物体同时下落”的科普实验网站。第一步，先搭个基础的页面骨架：页面左侧是控制面板，里面先放两个下拉菜单（物体A类型、物体B类型，选项暂定为小球和羽毛），以及几个表示参数的占位滑动条；页面右侧是相对宽大的实验显示区，在顶部放置两个不同颜色的圆形元素代表将要下落的物体；在实验区底部或页面最下方，放置“开始”和“重置”两个按钮，并预留一个空的div作为实验结果展示区。目前不需要实现任何动画和复杂的物理逻辑，只要把布局划分清晰，加上基础的CSS样式使其看起来像个整洁的控制台即可。\n另外请你注意，后续所有轮次必须始终遵循以下要求：第一，请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容；第二，在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /。","instruction_following_checklist":[{"constraint":"只要是你创建的代码文件，就必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能创建的代码文件扩展名\n target_extensions = {'.html', '.css', '.js', '.md', '.json', '.txt', '.htm'}\n \n # 如果项目目录不存在，说明并未生成任何文件，没有违背\"只要创建就必须...\"的要求\n if not os.path.exists(workspace_path):\n return True\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅针对推断出的代码文件类型进行验证，避免读取图片等二进制文件\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError 异常，则说明不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他非编码相关的系统级别读取异常（如权限不足等）\n pass\n \n # 如果所有被检查的文件均未抛出解码异常，判定为满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n \n def extract_js_variables(code_text):\n # 移除多行注释、单行注释及所有类型的字符串常量（防止误提取字符串内的单词）\n code_text = re.compile(r'/\\*.*?\\*/', re.DOTALL).sub('', code_text)\n code_text = re.compile(r'//.*').sub('', code_text)\n code_text = re.compile(r'\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"').sub('', code_text)\n code_text = re.compile(r\"'[^'\\\\]*(?:\\\\.[^'\\\\]*)*'\").sub('', code_text)\n code_text = re.compile(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', re.DOTALL).sub('', code_text)\n\n variables = set()\n\n # 1. 提取基本的 var, let, const 声明的变量\n matches = re.finditer(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', code_text)\n for m in matches:\n variables.add(m.group(1))\n\n # 2. 尝试从逗号分隔的多变量声明中提取简单变量（如 let a = 1, b = 2;）\n decl_blocks = re.findall(r'\\b(?:var|let|const)\\s+([^;\\n]+)', code_text)\n for block in decl_blocks:\n # 移除所有括号内部内容，避免误入复杂的解构赋值和对象字面量\n prev_len = -1\n while len(block) != prev_len:\n prev_len = len(block)\n block = re.sub(r'\\{[^}]*\\}', '', block)\n block = re.sub(r'\\[[^\\]]*\\]', '', block)\n block = re.sub(r'\\([^)]*\\)', '', block)\n \n parts = block.split(',')\n for p in parts:\n p = p.strip()\n # 仅保留赋值等号左侧的内容\n var_name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n variables.add(var_name)\n\n # 3. 提取普通函数参数\n func_matches = re.finditer(r'\\bfunction\\s+[a-zA-Z_$0-9]*\\s*\\(([^)]*)\\)', code_text)\n for m in func_matches:\n params_str = m.group(1)\n params = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', params_str)\n for p in params:\n variables.add(p)\n \n # 4. 提取箭头函数带括号的参数\n arrow_matches = re.finditer(r'\\(([^)]*)\\)\\s*=>', code_text)\n for m in arrow_matches:\n params_str = m.group(1)\n params = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', params_str)\n for p in params:\n variables.add(p)\n\n # 5. 提取箭头函数无括号的单参数\n single_arrow_matches = re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', code_text)\n for m in single_arrow_matches:\n variables.add(m.group(1))\n\n return variables\n\n def extract_css_variables(code_text):\n # 移除 CSS 注释后匹配 CSS 原生变量\n code_text = re.compile(r'/\\*.*?\\*/', re.DOTALL).sub('', code_text)\n variables = set()\n matches = re.finditer(r'--([a-zA-Z0-9_-]+)\\s*:', code_text)\n for m in matches:\n variables.add(m.group(1))\n return variables\n\n def get_codes_from_text(text):\n js_codes, css_codes = [], []\n \n # 提取独立的 js/css 块\n js_blocks = re.findall(r'```(?:js|javascript|JS)\\s+(.*?)```', text, re.DOTALL)\n js_codes.extend(js_blocks)\n css_blocks = re.findall(r'```(?:css|CSS)\\s+(.*?)```', text, re.DOTALL)\n css_codes.extend(css_blocks)\n \n # 提取 html 块中的 script 和 style 标签内容\n html_blocks = re.findall(r'```(?:html|HTML)\\s+(.*?)```', text, re.DOTALL)\n for h_block in html_blocks:\n script_blocks = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', h_block, re.DOTALL | re.IGNORECASE)\n js_codes.extend(script_blocks)\n style_blocks = re.findall(r'<style.*?>\\s*(.*?)\\s*</style>', h_block, re.DOTALL | re.IGNORECASE)\n css_codes.extend(style_blocks)\n \n return js_codes, css_codes\n\n all_js = []\n all_css = []\n\n # 从大模型回复的内容中提取代码\n js_resp, css_resp = get_codes_from_text(response)\n all_js.extend(js_resp)\n all_css.extend(css_resp)\n\n # 从工作区文件中提取代码\n if os.path.exists(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n all_js.append(f.read())\n except:\n pass\n elif file.endswith('.css'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n all_css.append(f.read())\n except:\n pass\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n script_blocks = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', content, re.DOTALL | re.IGNORECASE)\n all_js.extend(script_blocks)\n style_blocks = re.findall(r'<style.*?>\\s*(.*?)\\s*</style>', content, re.DOTALL | re.IGNORECASE)\n all_css.extend(style_blocks)\n except:\n pass\n\n # 聚合找到的所有变量名\n variables_found = set()\n for j_code in all_js:\n variables_found.update(extract_js_variables(j_code))\n for c_code in all_css:\n variables_found.update(extract_css_variables(c_code))\n\n # 判断是否全部符合 2 <= 长度 <= 7 \n for var in variables_found:\n if len(var) < 2 or len(var) > 7:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对不同语言用于提取类名的正则表达式\n html_class_re = re.compile(r'\\bclass\\s*=\\s*([\"\\'])(.*?)\\1', re.IGNORECASE)\n css_class_re = re.compile(r'\\.([a-zA-Z_-][a-zA-Z0-9_-]*)')\n js_class_re = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n \n # 类名的合法性判断正则：只能由小写字母和下划线组成\n valid_pattern = re.compile(r'^[a-z_]+$')\n\n # 遍历当前工作空间内的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 本项目主要涉及 HTML, CSS, JavaScript 文件\n if ext in ['.html', '.css', '.js']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n css_content = \"\"\n js_content = \"\"\n\n if ext == '.html':\n # 去除HTML注释，防止误判注释里的例子代码\n content_clean = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\n \n # 检查 HTML 的 class=\"...\" 属性里引用的类名\n for match in html_class_re.finditer(content_clean):\n class_str = match.group(2)\n # HTML class 属性内可以是用空格隔开的多个类名\n for c_name in class_str.split():\n if not valid_pattern.match(c_name):\n return False\n \n # 提取内嵌的 <style> 与 <script> 内容待后续分别检查\n style_contents = re.findall(r'<style[^>]*>(.*?)</style>', content_clean, flags=re.IGNORECASE | re.DOTALL)\n css_content = ' '.join(style_contents)\n \n script_contents = re.findall(r'<script[^>]*>(.*?)</script>', content_clean, flags=re.IGNORECASE | re.DOTALL)\n js_content = ' '.join(script_contents)\n \n elif ext == '.css':\n css_content = content\n elif ext == '.js':\n js_content = content\n \n # 检查 CSS 类选择器\n if css_content:\n # 过滤 CSS 中的注释\n css_content = re.sub(r'/\\*.*?\\*/', '', css_content, flags=re.DOTALL)\n # 过滤 CSS 中的字符串（单引号或双引号）\n css_content = re.sub(r'([\"\\'])(?:\\\\.|[^\\\\])*?\\1', '', css_content)\n # 过滤 CSS 中的 url(...) 以防点号导致误认为类选择器\n css_content = re.sub(r'url\\([^)]*\\)', '', css_content, flags=re.IGNORECASE)\n \n for match in css_class_re.finditer(css_content):\n c_name = match.group(1)\n if not valid_pattern.match(c_name):\n return False\n \n # 检查 JS 定义的类\n if js_content:\n # 过滤 JS 单行注释\n js_content = re.sub(r'//.*', '', js_content)\n # 过滤 JS 多行注释\n js_content = re.sub(r'/\\*.*?\\*/', '', js_content, flags=re.DOTALL)\n # 过滤 JS 字符串（反引号、单引号或双引号）\n js_content = re.sub(r'([\"\\'`])(?:\\\\.|[^\\\\])*?\\1', '', js_content)\n \n for match in js_class_re.finditer(js_content):\n c_name = match.group(1)\n if not valid_pattern.match(c_name):\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 忽略的关键字，防止 p3 正则误捕获常见的JS控制流语句和内置关键字\n keywords = {\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'constructor',\n 'return', 'with', 'async', 'await', 'class', 'else', 'elif', 'except'\n }\n \n # 前缀断言：确保匹配的不是 CSS 类、伪类，或是字符串中间的一部分\n # 允许前面有空格、换行、点号(.)、井号(#)等，但不允许有冒号(:)、连字符(-)或 @ 符号\n prefix = r'(?<![a-zA-Z0-9_$\\-:@])'\n \n # 1. 匹配标准 JS 函数声明：function foo()\n p1 = re.compile(prefix + r'function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n \n # 2. 匹配函数表达式赋值或箭头函数：foo = function() 或 foo = () => 或 foo: () =>\n p2 = re.compile(prefix + r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*(?:(?<![+\\-*/%&|^=<>!])[:=](?![=]))\\s*(?:async\\s+)?(?:function(?![a-zA-Z0-9_$])|\\([^)]*\\)\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\s*=>)')\n \n # 3. 匹配类方法或对象简写方法：foo() { \n p3 = re.compile(prefix + r'(?:(?:async\\s+)?(?:get\\s+|set\\s+))?([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{')\n \n # 4. 匹配 Python 函数声明 (兜底处理)：def foo()\n p4 = re.compile(prefix + r'def\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n \n all_functions = set()\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 匹配 p1\n for match in p1.findall(content):\n all_functions.add(match)\n \n # 匹配 p2\n for match in p2.findall(content):\n all_functions.add(match)\n \n # 匹配 p3\n for match in p3.findall(content):\n if match not in keywords:\n all_functions.add(match)\n \n # 匹配 p4\n for match in p4.findall(content):\n all_functions.add(match)\n \n except Exception:\n # 忽略无法用 utf-8 读取的文件 (如图片等文件)\n pass\n \n # 验证所有收集到的函数名是否为严格的回文字符串\n for func_name in all_functions:\n if func_name != func_name[::-1]:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保你生成的每一个代码文件都精确包含 3 行注释","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n if not os.path.isdir(workspace_path):\n return False\n \n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in code_extensions:\n continue\n \n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n return False\n \n comment_lines_count = 0\n in_block_comment = False\n \n for line in lines:\n s_line = line.strip()\n \n if ext == '.html':\n if '<!--' in s_line and '-->' in s_line:\n comment_lines_count += 1\n elif '<!--' in s_line:\n in_block_comment = True\n comment_lines_count += 1\n elif '-->' in s_line and in_block_comment:\n in_block_comment = False\n comment_lines_count += 1\n elif in_block_comment:\n comment_lines_count += 1\n else:\n if ext == '.js' and s_line.startswith('//'):\n comment_lines_count += 1\n continue\n \n if '/*' in s_line and '*/' in s_line:\n comment_lines_count += 1\n elif '/*' in s_line:\n in_block_comment = True\n comment_lines_count += 1\n elif '*/' in s_line and in_block_comment:\n in_block_comment = False\n comment_lines_count += 1\n elif in_block_comment:\n comment_lines_count += 1\n \n if comment_lines_count != 3:\n return False\n \n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求","validation_code":"def check_class_name_length_is_even(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n classes = set()\n \n # 正则表达式用于匹配不同代码场景下的类名\n # 1. 匹配 JS 等代码中的 class 声明，如：class MyClass\n class_decl_pattern = re.compile(r'\\bclass\\s+([A-Za-z_][A-Za-z0-9_]*)\\b')\n # 2. 匹配 HTML 中的 class 属性，如：class=\"panel control\"\n html_class_pattern = re.compile(r'\\bclass\\s*=\\s*([\"\\'])(.*?)\\1', re.IGNORECASE)\n # 3. 匹配 CSS 文件中的类选择器，如：.box_container {\n css_class_pattern = re.compile(r'\\.([A-Za-z_][A-Za-z0-9_-]*)(?=[ \\t\\n\\r]*[\\{,>+~:\\.])')\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 跳过常见二进制及字体图片文件\n if ext in ['.png', '.jpg', '.jpeg', '.gif', '.ico', '.svg', '.woff', '.woff2', '.ttf', '.eot', '.mp4', '.webm']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 1. 提取编程语言中的 class 声明\n matches1 = class_decl_pattern.findall(content)\n classes.update(matches1)\n \n # 2. 提取 HTML 中 class 属性定义的类名（如含多个以空格分隔）\n matches2 = html_class_pattern.findall(content)\n for _, cls_str in matches2:\n for c in cls_str.split():\n # 确保提取到的是合法的类名（过滤非预期的模板特殊符号）\n if re.match(r'^[A-Za-z_][A-Za-z0-9_-]*$', c):\n classes.add(c)\n \n # 3. 提取专门在 CSS 文件中的类选择器\n if ext == '.css':\n matches3 = css_class_pattern.findall(content)\n classes.update(matches3)\n \n # 验证每个收集到的类名，其名称长度是否严格为偶数\n for cls in classes:\n if len(cls) % 2 != 0:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_odd_line_counts(response: str, workspace_path: str) -> bool:\n import os\n\n target_extensions = {'.html', '.css', '.js', '.md'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in target_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n # 统计包含空行在内的所有行数\n total_lines = len(content.split('\\n'))\n \n if total_lines % 2 == 0:\n return False\n except Exception:\n continue\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":1,"instruction":"不错，现在咱们把基础的下落动画和按钮交互加进去。点击“开始”按钮后，右侧实验区里的两个物体要同时从顶部开始匀加速下落，直到触碰到底部边界停止。使用 `requestAnimationFrame` 来实现平滑的动画效果。点击“重置”按钮能让两个物体瞬间回到初始位置。这里先只考虑基础重力（比如设定 $g=9.8$），暂时不要管空气阻力，所以两个物体应该完全同步下落。当它们碰到页面底部时，在预留的结果展示区里打印出一句话，比如“实验完成，落地耗时：X.XX 秒”。\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中；第二，请在生成的所有代码文件中，使用正好 4 个空格来完成每一处缩进。","instruction_following_checklist":[{"constraint":"只要是你创建的代码文件，就必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令推断可能的代码文件类型\n code_extensions = {'.html', '.css', '.js'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试使用 UTF-8 进行解码\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果发生解码错误，说明不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他读取异常，如权限问题等\n continue\n\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def extract_vars_from_tokens(tokens):\n vars_found = set()\n filtered_tokens = []\n \n level = 0\n ignore_level = -1\n \n for t in tokens:\n if t in ('{', '[', '('):\n level += 1\n \n if ignore_level != -1:\n # 若跳出了忽略所在的层级，解除忽略状态\n if level < ignore_level:\n ignore_level = -1\n # 在同一层级遇到逗号或分号，结束赋值的忽略状态\n elif level == ignore_level and t in (',', ';'):\n ignore_level = -1\n # 对于顶层的换行，也可以作为声明赋值结束的标志\n elif level == 0 and level == ignore_level and t == '\\n':\n ignore_level = -1\n \n if ignore_level == -1:\n if t == '=':\n ignore_level = level\n else:\n filtered_tokens.append(t)\n \n if t in ('}', ']', ')'):\n level -= 1\n if ignore_level != -1 and level < ignore_level:\n ignore_level = -1\n \n idx = 0\n # 排除JS中的大部分可能被误认的关键字\n keywords = {\n 'in', 'of', 'var', 'let', 'const', 'function', 'catch', 'true', 'false', \n 'null', 'undefined', 'return', 'await', 'yield', 'if', 'else', 'for', \n 'while', 'do', 'switch', 'case', 'break', 'continue', 'default', 'try', \n 'finally', 'class', 'extends', 'super', 'import', 'export', 'new', \n 'this', 'typeof', 'instanceof', 'void', 'delete'\n }\n \n while idx < len(filtered_tokens):\n t = filtered_tokens[idx]\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t) and t not in keywords:\n # 检查标识符是否是对象解构中的属性名（其后紧跟冒号）\n next_idx = idx + 1\n while next_idx < len(filtered_tokens) and filtered_tokens[next_idx] == '\\n':\n next_idx += 1\n \n if next_idx < len(filtered_tokens) and filtered_tokens[next_idx] == ':':\n idx = next_idx + 1\n continue\n else:\n vars_found.add(t)\n idx += 1\n \n return vars_found\n\n variables = set()\n has_target_files = False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n code = \"\"\n \n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n except Exception:\n continue\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n html = f.read()\n # 仅提取 HTML 文件中 script 标签内的内容\n scripts = re.findall(r'<script.*?>\\s*(.*?)\\s*</script>', html, flags=re.DOTALL | re.IGNORECASE)\n code = \"\\n\".join(scripts)\n except Exception:\n continue\n \n if not code.strip():\n continue\n \n has_target_files = True\n\n # 去除字符串字面量内容，防止引发解析错误\n code_clean = re.sub(r'\\'(?:\\\\.|[^\\\\\\'])*\\'', \"''\", code)\n code_clean = re.sub(r'\"(?:\\\\.|[^\\\\\"])*\"', '\"\"', code_clean)\n code_clean = re.sub(r'`(?:\\\\.|[^\\\\`])*`', '``', code_clean)\n \n # 去除注释\n code_clean = re.sub(r'//.*', '', code_clean)\n code_clean = re.sub(r'/\\*.*?\\*/', '', code_clean, flags=re.DOTALL)\n\n # 单参数箭头函数提前提取（如: \"data =>\"）\n single_arrows = re.findall(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', code_clean)\n for sa in single_arrows:\n if sa not in ('in', 'of', 'var', 'let', 'const', 'function', 'catch', 'true', 'false', 'null', 'undefined'):\n variables.add(sa)\n\n # 多参数箭头函数提前提取（如: \"(a, b) =>\"）\n multi_arrows = re.findall(r'\\(([^)]*)\\)\\s*=>', code_clean)\n for ma in multi_arrows:\n arr_tokens = re.findall(r'\\n|\\b(?:var|let|const|function|in|of|catch)\\b|[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\[\\](),;:]', ma)\n variables.update(extract_vars_from_tokens(arr_tokens))\n\n # 生成全局片段 Token 序列\n tokens = re.findall(r'\\n|\\b(?:var|let|const|function|in|of|catch)\\b|[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\[\\](),;:]', code_clean)\n \n i = 0\n while i < len(tokens):\n t = tokens[i]\n \n # 变量声明\n if t in ('var', 'let', 'const'):\n i += 1\n decl_tokens = []\n level = 0\n while i < len(tokens):\n curr = tokens[i]\n if curr in ('{', '[', '('):\n level += 1\n elif curr in ('}', ']', ')'):\n level -= 1\n \n decl_tokens.append(curr)\n \n # 顶层的分号和换行意味着声明结束\n if level == 0 and curr in (';', '\\n'):\n break\n # for...in / for...of 的中途结束\n if level == 0 and curr in ('in', 'of'):\n break\n \n i += 1\n variables.update(extract_vars_from_tokens(decl_tokens))\n \n # 函数定义\n elif t == 'function':\n i += 1\n # 匹配函数名\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]) and tokens[i] not in ('(', '{', '['):\n variables.add(tokens[i])\n i += 1\n # 匹配参数列表\n if i < len(tokens) and tokens[i] == '(':\n param_tokens = []\n level = 1\n param_tokens.append(tokens[i])\n i += 1\n while i < len(tokens) and level > 0:\n curr = tokens[i]\n if curr == '(': level += 1\n elif curr == ')': level -= 1\n param_tokens.append(curr)\n i += 1\n variables.update(extract_vars_from_tokens(param_tokens))\n \n # try...catch 错误参数提取\n elif t == 'catch':\n i += 1\n if i < len(tokens) and tokens[i] == '(':\n param_tokens = []\n level = 1\n param_tokens.append(tokens[i])\n i += 1\n while i < len(tokens) and level > 0:\n curr = tokens[i]\n if curr == '(': level += 1\n elif curr == ')': level -= 1\n param_tokens.append(curr)\n i += 1\n variables.update(extract_vars_from_tokens(param_tokens))\n else:\n i += 1\n\n # 如若代码中提取不到任何变量且没有能够解析的有效代码，默认为 True，因无违规项\n if not variables and not has_target_files:\n return True\n\n # 判断提炼得到的变量名是否皆满足长度约束 [2, 7]\n for var in variables:\n if len(var) < 2 or len(var) > 7:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 如果工作区路径不存在，返回 False\n if not os.path.isdir(workspace_path):\n return False\n\n class_names = set()\n\n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n # 只处理指定的代码文件类型\n if ext not in ['.html', '.htm', '.css', '.js']:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 1. 提取 JavaScript 的 Class 声明 (如: class my_class)\n if ext in ['.js', '.html', '.htm']:\n js_classes = re.findall(r'\\bclass\\s+([A-Za-z0-9_]+)\\b', content)\n class_names.update(js_classes)\n \n # 2. 提取 HTML 中 class 属性配置 (如: class=\"my_class1 my_class2\")\n if ext in ['.html', '.htm', '.js']:\n # 匹配 class 或 className 属性里的字符串\n html_classes = re.findall(r'\\b(?:class|className)\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']', content, flags=re.IGNORECASE)\n for cls_str in html_classes:\n # 分割可能存在多个类名用空格隔开的情况\n class_names.update(cls_str.split())\n \n # 3. 提取 CSS 文件及 HTML <style> 标签中的类选择器 (如: .my_class)\n if ext == '.css':\n # 匹配点号开头，首字符为字母或下划线的 CSS 选择器\n css_classes = re.findall(r'\\.([A-Za-z_][A-Za-z0-9_-]*)', content)\n class_names.update(css_classes)\n elif ext in ['.html', '.htm']:\n # 仅在 <style> 标签范围内提取 CSS 类选择器\n style_contents = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.IGNORECASE | re.DOTALL)\n for style_content in style_contents:\n css_classes = re.findall(r'\\.([A-Za-z_][A-Za-z0-9_-]*)', style_content)\n class_names.update(css_classes)\n \n # 验证提取到的每一个类名是否仅由小写字母和下划线构成\n for name in class_names:\n # 正则 ^[a-z_]+$ 确保只能由 a-z 的小写字母及下划线 _ 构成\n if not re.match(r'^[a-z_]+$', name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串","validation_code":"import os\nimport re\n\ndef check_function_names(response: str, workspace_path: str) -> bool:\n found_any_function = False\n \n # 匹配普通函数声明，例如 function aba() 或 async function aba()\n pattern_func_decl = re.compile(r'(?:async\\s+)?function\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n \n # 匹配函数表达式赋值，例如 const aba = function() 或 let aba = async function()\n pattern_func_expr = re.compile(r'(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?function\\s*\\(')\n \n # 匹配箭头函数赋值，例如 const aba = () => 或者 let aba = arg =>\n pattern_arrow_func = re.compile(r'(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>')\n \n # 提取 HTML 文件中的 script 标签内容\n pattern_script = re.compile(r'<script[^>]*>([\\s\\S]*?)</script>', re.IGNORECASE)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅检查 .js 和 .html 文件\n if file.endswith('.js') or file.endswith('.html'):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n js_contents = []\n if file.endswith('.html'):\n scripts = pattern_script.findall(content)\n js_contents.extend(scripts)\n else:\n js_contents.append(content)\n \n for js_code in js_contents:\n # 预处理：去掉注释和字符串，避免在无关内容中产生误匹配\n # 1. 去除单行注释\n js_code = re.sub(r'//.*', '', js_code)\n # 2. 去除多行注释\n js_code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', js_code)\n # 3. 去除单引号、双引号、反引号字符串\n js_code = re.sub(r\"'([^'\\\\]|\\\\.)*'\", \"''\", js_code)\n js_code = re.sub(r'\"([^\"\\\\]|\\\\.)*\"', '\"\"', js_code)\n js_code = re.sub(r'`([^`\\\\]|\\\\.)*`', '``', js_code)\n\n names = []\n # 提取所有命名函数名称\n names.extend(pattern_func_decl.findall(js_code))\n names.extend(pattern_func_expr.findall(js_code))\n names.extend(pattern_arrow_func.findall(js_code))\n \n for name in names:\n found_any_function = True\n # 检查函数名是否为严格的回文字符串 (区分大小写)\n if name != name[::-1]:\n return False\n\n # 若所有提取到的函数名均为回文且至少存在一个函数，返回 True\n return found_any_function","tags":["Style","Naming"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保你生成的每一个代码文件都精确包含 3 行注释","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n import bisect\n\n def get_comment_lines_in_code(code_content, offset, full_content, is_css=False):\n comment_lines = set()\n i = 0\n in_string = False\n string_char = ''\n in_single_comment = False\n in_multi_comment = False\n \n # 预先计算该代码块在整个文件中的初始行号\n current_line = full_content[:offset].count('\\n') + 1\n string_quote_chars = ['\"', \"'\"] if is_css else ['\"', \"'\", '`']\n \n while i < len(code_content):\n c = code_content[i]\n \n if c == '\\n':\n if in_single_comment:\n in_single_comment = False\n current_line += 1\n if in_multi_comment:\n comment_lines.add(current_line)\n i += 1\n continue\n \n if in_single_comment:\n comment_lines.add(current_line)\n i += 1\n continue\n \n if in_multi_comment:\n comment_lines.add(current_line)\n if c == '*' and i + 1 < len(code_content) and code_content[i+1] == '/':\n in_multi_comment = False\n i += 2\n else:\n i += 1\n continue\n \n if in_string:\n if c == '\\\\':\n i += 2\n continue\n if c == string_char:\n in_string = False\n i += 1\n continue\n \n if c in string_quote_chars:\n in_string = True\n string_char = c\n i += 1\n continue\n \n if c == '/' and i + 1 < len(code_content):\n if not is_css and code_content[i+1] == '/':\n in_single_comment = True\n comment_lines.add(current_line)\n i += 2\n continue\n elif code_content[i+1] == '*':\n in_multi_comment = True\n comment_lines.add(current_line)\n i += 2\n continue\n \n i += 1\n \n return comment_lines\n\n def count_comment_lines_in_file(file_path):\n ext = os.path.splitext(file_path)[1].lower()\n \n # 仅限定关注用户可能生成的相关代码文件，排查各类常规文本文档等\n code_extensions = {'.html', '.htm', '.css', '.js'}\n if ext not in code_extensions:\n return None\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n return -1\n\n comment_lines = set()\n\n if ext in ['.js']:\n comment_lines = get_comment_lines_in_code(content, 0, content, is_css=False)\n elif ext in ['.css']:\n comment_lines = get_comment_lines_in_code(content, 0, content, is_css=True)\n elif ext in ['.html', '.htm']:\n line_indices = [0]\n for idx, char in enumerate(content):\n if char == '\\n':\n line_indices.append(idx + 1)\n \n def get_line_num(index):\n return bisect.bisect_right(line_indices, index)\n\n skip_intervals = []\n\n # 1. 查找并解析 <script> 内的 JS 注释\n for script_match in re.finditer(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE):\n skip_intervals.append((script_match.start(), script_match.end()))\n script_start = script_match.start(1)\n script_content = script_match.group(1)\n lines = get_comment_lines_in_code(script_content, script_start, content, is_css=False)\n comment_lines.update(lines)\n\n # 2. 查找并解析 <style> 内的 CSS 注释\n for style_match in re.finditer(r'<style[^>]*>(.*?)</style>', content, re.DOTALL | re.IGNORECASE):\n skip_intervals.append((style_match.start(), style_match.end()))\n style_start = style_match.start(1)\n style_content = style_match.group(1)\n lines = get_comment_lines_in_code(style_content, style_start, content, is_css=True)\n comment_lines.update(lines)\n\n def is_in_skip(idx):\n for st, ed in skip_intervals:\n if st <= idx < ed:\n return True\n return False\n\n # 3. 查找剩余结构里的 HTML 多行注释\n for m in re.finditer(r'<!--(.*?)-->', content, re.DOTALL):\n if is_in_skip(m.start()):\n continue\n start_line = get_line_num(m.start())\n end_line = get_line_num(max(m.start(), m.end() - 1))\n for l in range(start_line, end_line + 1):\n comment_lines.add(l)\n\n return len(comment_lines)\n\n has_code_file = False\n \n # 防止传入不可用路径\n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n cnt = count_comment_lines_in_file(file_path)\n \n if cnt is not None:\n has_code_file = True\n if cnt != 3:\n return False\n \n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 常用且可能在科普实验项目中出现的代码类文件扩展名\n valid_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.cpp', '.cs', '.php'}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 过滤隐藏文件\n if file.startswith('.'):\n continue\n \n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n if ext not in valid_exts:\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n classes = set()\n \n # 1. OOP 面向对象类名提取: 例如 class MyClass { ... }\n oop_classes = re.findall(r'\\bclass\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\b', content)\n classes.update(oop_classes)\n \n # 2. HTML/JSX 标签中的 class 属性提取\n if ext in {'.html', '.htm', '.jsx', '.tsx', '.vue'}:\n class_attrs = re.findall(r'\\bclass(?:Name)?\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']', content, re.IGNORECASE)\n for attr in class_attrs:\n classes.update(attr.split())\n \n # 3. CSS 类选择器提取: 例如 .my_class { ... }\n if ext == '.css':\n css_classes = re.findall(r'\\.([a-zA-Z_][a-zA-Z0-9_-]*)', content)\n classes.update(css_classes)\n elif ext in {'.html', '.htm', '.vue'}:\n # 如果是内嵌结构，在 <style> 块中寻找\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, re.IGNORECASE | re.DOTALL)\n for style in styles:\n css_classes = re.findall(r'\\.([a-zA-Z_][a-zA-Z0-9_-]*)', style)\n classes.update(css_classes)\n \n # 4. JavaScript 常用的 DOM 类名操控 API 提取\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}:\n # classList API \n js_classes = re.findall(r'classList\\.(?:add|remove|toggle|contains)\\s*\\(\\s*[\"\\']([^\"\\']+)[\"\\']', content)\n for attrs in js_classes:\n classes.update(attrs.split())\n \n # jQuery addClass, removeClass 等 API\n jq_classes = re.findall(r'(?:add|remove|has|toggle)Class\\s*\\(\\s*[\"\\']([^\"\\']+)[\"\\']', content)\n for attrs in jq_classes:\n classes.update(attrs.split())\n \n # getElementsByClassName API\n get_classes = re.findall(r'getElementsByClassName\\s*\\(\\s*[\"\\']([^\"\\']+)[\"\\']', content)\n for attrs in get_classes:\n classes.update(attrs.split())\n \n # 检查在该代码文件中提取到的所有类的长度\n for c in classes:\n c_clean = c.strip()\n if c_clean:\n # 如果任一类名长度不为偶数，则要求未被满足\n if len(c_clean) % 2 != 0:\n return False\n\n # 若未找到任何违规类，则代表完全满足要求\n return True","tags":["Style","Naming"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据上下文，推断出生成的代码文件类型\n code_extensions = ['.html', '.css', '.js']\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if any(file.endswith(ext) for ext in code_extensions):\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 统一换行符格式，确保按行划分准确无误\n content = content.replace('\\r\\n', '\\n')\n \n # 按换行符切分以精确统计行数，该方法能够将所有的空行严格计算在内\n lines = content.split('\\n')\n \n # 若当前代码文件总行数为偶数，直接判定不符合要求\n if len(lines) % 2 == 0:\n return False\n except Exception:\n # 如果遇到文件编码不符合 UTF-8 导致读取错误，等同于未满足条件\n return False\n \n # 如果找到了至少一个代码文件，并且没有任何代码文件的行数为偶数，则满足要求\n return found_code_file","tags":["Quantity","Exact Value"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"请在生成的所有代码文件中，使用正好 4 个空格来完成每一处缩进","validation_code":"def check_indentation(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据用户指令，推断代码文件为 HTML、CSS 和原生 JavaScript\n code_extensions = ['.html', '.css', '.js']\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅检查代码文件，忽略其它可能生成的文件如 progress.md 或 txt 等\n if any(file.endswith(ext) for ext in code_extensions):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n for line in lines:\n # 忽略空行或纯空白行\n if not line.strip():\n continue\n \n # 提取当前行的前导空白字符\n leading_spaces = ''\n for char in line:\n if char in (' ', '\\t'):\n leading_spaces += char\n else:\n break\n \n # 如果顶格写（无缩进），则继续检查下一行\n if not leading_spaces:\n continue\n \n # 严格要求使用空格，严禁包含 \\t 制表符\n if '\\t' in leading_spaces:\n return False\n \n content_stripped = line.strip()\n # 忽略由于多行注释中的星号 '*' 带来的不可避免的对齐排版（如 /** \\n * \\n */）\n if content_stripped.startswith('*'):\n continue\n \n # 验证缩进的空格数是否是 4 的倍数（即正好使用 4 个空格完成一处缩进）\n if len(leading_spaces) % 4 != 0:\n return False\n\n return True","tags":["Style","Layout"]}]},{"round_id":2,"instruction":"很好，接下来增加物理引擎的复杂度，让实验更有趣。在左侧控制面板把之前占位的滑动条完善为：物体A重量、物体B重量、以及风阻系数。现在要修改下落动画的数学计算逻辑，引入空气阻力（阻力可以简单模拟为与速度成正比，或者根据牛顿阻力定律计算）。这里的核心挑战是：结合物体的重量和风阻系数进行加速度的实时演算。重量越轻、风阻系数越大的物体（比如羽毛），受空气阻力影响越明显，下落越慢。修改下拉菜单的联动逻辑：选“羽毛”时自动设置较轻的重量和较大的风阻，选“小球”时设置较重的重量和较小的风阻。落地后，在结果区分别显示A和B各自的具体耗时。\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，生成的代码文件应全部使用2个空格来控制缩进，替换之前的4空格缩进要求；第二，进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤。","instruction_following_checklist":[{"constraint":"只要是你创建的代码文件，就必须使用UTF-8进行编码","validation_code":"def check_code_files_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n\n # 如果工作区不存在，无法进行验证\n if not os.path.exists(workspace_path):\n return False\n\n # 根据指令推断可能生成的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n \n # 遍历工作区所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅验证代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦抛出解码异常，说明文件不是合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他的异常，如无读取权限等\n pass\n \n # 如果所有代码文件都能成功以 UTF-8 解码，或未发现代码文件，判定满足要求\n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7","validation_code":"def check_workspace(response: str, workspace_path: str) -> bool:\n import os\n import re\n import ast\n\n def extract_params(params_str):\n \"\"\"辅助函数：解析 JS 函数参数列表中的变量名（支持部分解构语法）\"\"\"\n variables = set()\n # 按照 JS 常用标识符与常见符号分词\n tokens = [m.group(0) for m in re.finditer(r'[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\[\\]();,]', params_str)]\n i = 0\n n = len(tokens)\n in_assignment = False\n bracket_level = 0\n \n while i < n:\n t = tokens[i]\n if t == ',':\n if bracket_level == 0:\n in_assignment = False\n elif t in ['{', '[', '(']:\n bracket_level += 1\n elif t in ['}', ']', ')']:\n bracket_level -= 1\n elif t == '=':\n if bracket_level == 0:\n in_assignment = True\n elif re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t):\n if not in_assignment:\n variables.add(t)\n i += 1\n return variables\n\n def extract_js_variables(code):\n \"\"\"解析 JavaScript 代码中的声明变量\"\"\"\n # 预处理：移除注释与字符串常量，避免误判\n code = re.sub(r'//.*', '', code)\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n code = re.sub(r'\"(?:\\\\.|[^\"\\\\])*\"', '', code)\n code = re.sub(r\"'(?:\\\\.|[^'\\\\])*'\", '', code)\n code = re.sub(r'`[\\s\\S]*?`', '', code)\n \n tokens = [m.group(0) for m in re.finditer(r'[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\[\\]();,]', code)]\n variables = set()\n i = 0\n n = len(tokens)\n \n # 状态机：扫描并提取 var/let/const/function/class 声明的变量\n while i < n:\n t = tokens[i]\n if t in ['let', 'var', 'const']:\n i += 1\n bracket_level = 0\n in_assignment = False\n while i < n:\n t2 = tokens[i]\n if t2 == ';':\n break\n elif t2 in ['function', 'class']:\n i -= 1\n break\n elif t2 == ',':\n if bracket_level == 0:\n in_assignment = False\n elif t2 in ['{', '[', '(']:\n bracket_level += 1\n elif t2 in ['}', ']', ')']:\n bracket_level -= 1\n elif t2 == '=':\n if bracket_level == 0:\n in_assignment = True\n elif t2 in ['in', 'of']:\n if bracket_level == 0:\n in_assignment = True\n elif re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t2):\n if not in_assignment:\n variables.add(t2)\n i += 1\n elif t == 'function':\n i += 1\n if i < n and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]) and tokens[i] not in ['(', '{']:\n variables.add(tokens[i])\n elif t == 'class':\n i += 1\n if i < n and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]) and tokens[i] != '{':\n variables.add(tokens[i])\n elif t == 'catch':\n i += 1\n while i < n and tokens[i] != '(':\n if tokens[i] in ['{', ';']: break\n i += 1\n if i < n and tokens[i] == '(':\n i += 1\n if i < n and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]):\n variables.add(tokens[i])\n else:\n i += 1\n \n # 通过正则额外兜底提取函数参数、箭头函数及 import 变量\n for m in re.finditer(r'\\bfunction(?:\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', code):\n variables.update(extract_params(m.group(1)))\n \n for m in re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', code):\n variables.add(m.group(1))\n for m in re.finditer(r'\\(([^)]+)\\)\\s*=>', code):\n variables.update(extract_params(m.group(1)))\n\n for m in re.finditer(r'\\bimport\\s+([^{;\"\\'\\n]+)\\s+from', code):\n for v in re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', m.group(1)):\n if v not in ['*', 'as']:\n variables.add(v)\n for m in re.finditer(r'\\bimport\\s*\\{([^}]+)\\}', code):\n for v in re.findall(r'\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\b', m.group(1)):\n if v != 'as':\n variables.add(v)\n\n js_reserved = {\n 'break', 'case', 'catch', 'class', 'const', 'continue', 'debugger', 'default', \n 'delete', 'do', 'else', 'export', 'extends', 'finally', 'for', 'function', 'if', \n 'import', 'in', 'instanceof', 'new', 'return', 'super', 'switch', 'this', 'throw', \n 'try', 'typeof', 'var', 'void', 'while', 'with', 'yield', 'let', 'static', 'enum', \n 'await', 'implements', 'package', 'protected', 'interface', 'private', 'public',\n 'null', 'true', 'false', 'undefined', 'NaN', 'Infinity'\n }\n return {v for v in variables if v not in js_reserved}\n\n def extract_css_variables(code):\n \"\"\"解析 CSS 中的自定义变量\"\"\"\n variables = set()\n for m in re.finditer(r'--([a-zA-Z0-9_-]+)', code):\n variables.add('--' + m.group(1))\n return variables\n\n def extract_python_variables(code):\n \"\"\"解析 Python 中的变量名\"\"\"\n variables = set()\n try:\n tree = ast.parse(code)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n variables.add(node.id)\n elif isinstance(node, ast.arg):\n variables.add(node.arg)\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):\n variables.add(node.name)\n except Exception:\n pass\n return variables\n\n all_vars = set()\n\n # 遍历项目文件树\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 根据后缀采用不同的解析逻辑\n if ext == '.js':\n all_vars.update(extract_js_variables(content))\n elif ext == '.py':\n all_vars.update(extract_python_variables(content))\n elif ext == '.css':\n all_vars.update(extract_css_variables(content))\n elif ext in ['.html', '.htm']:\n # 解析 HTML 中包含的 JS 和 CSS 内容\n js_matches = re.finditer(r'<script.*?>([\\s\\S]*?)</script>', content, re.IGNORECASE)\n for m in js_matches:\n all_vars.update(extract_js_variables(m.group(1)))\n css_matches = re.finditer(r'<style.*?>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n for m in css_matches:\n all_vars.update(extract_css_variables(m.group(1)))\n \n # 验证要求：所有提取出的变量，其字符长度必须在 [2, 7] 的区间之内\n for var in all_vars:\n if len(var) < 2 or len(var) > 7:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线","validation_code":"def check_class_names_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n class_names = set()\n \n # 提取多种场景下的“类名”\n js_class_pattern = re.compile(r'\\bclass\\s+([A-Za-z0-9_$]+)')\n html_class_pattern = re.compile(r'\\bclass(?:Name)?\\s*=\\s*([\"\\'])([^\"\\']*)\\1')\n js_classlist_pattern = re.compile(r'\\.classList\\.(?:add|remove|toggle|contains)\\s*\\(\\s*([\"\\'])([^\"\\']*)\\1\\s*\\)')\n js_getelements_pattern = re.compile(r'\\.getElementsByClassName\\s*\\(\\s*([\"\\'])([^\"\\']*)\\1\\s*\\)')\n css_class_pattern = re.compile(r'\\.([a-zA-Z_][a-zA-Z0-9_-]*)')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 过滤只排查前端常见的文本文件\n if not file.endswith(('.html', '.js', '.css', '.jsx', '.ts', '.tsx')):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 预处理：简单去除注释内容以防止干扰匹配\n content = re.sub(r'/\\*[\\s\\S]*?\\*/', '', content)\n content = re.sub(r'<!--[\\s\\S]*?-->', '', content)\n content = re.sub(r'//.*', '', content)\n \n # 1. 提取 JS/ES6 面向对象定义的类\n for match in js_class_pattern.finditer(content):\n class_names.add(match.group(1).strip())\n \n # 2. 提取 HTML 中的 class 属性及 JS 中的 className 赋值\n for match in html_class_pattern.finditer(content):\n classes_str = match.group(2)\n for cls in classes_str.split():\n if cls.strip():\n class_names.add(cls.strip())\n \n # 3. 提取 JS 中通过 classList 操作关联的类名\n for match in js_classlist_pattern.finditer(content):\n classes_str = match.group(2)\n for cls in classes_str.split():\n if cls.strip():\n class_names.add(cls.strip())\n \n # 4. 提取 JS 中通过 getElementsByClassName 查询关联的类名\n for match in js_getelements_pattern.finditer(content):\n classes_str = match.group(2)\n for cls in classes_str.split():\n if cls.strip():\n class_names.add(cls.strip())\n \n # 5. 提取 CSS 中的类选择器（综合处理 .css 文件和 HTML <style> 标签的内容）\n style_content = []\n if file.endswith('.css'):\n style_content.append(content)\n elif file.endswith('.html'):\n styles = re.findall(r'<style[^>]*>([\\s\\S]*?)</style>', content, re.IGNORECASE)\n style_content.extend(styles)\n \n for css_text in style_content:\n # 在样式表中先行移除 url() 和 字符串 的内容，避免将诸如 .png 等后缀误判为类名\n css_text = re.sub(r'url\\([^)]*\\)', '', css_text)\n css_text = re.sub(r'([\"\\'])[\\s\\S]*?\\1', '', css_text)\n \n for cls in css_class_pattern.findall(css_text):\n class_names.add(cls.strip())\n \n except Exception:\n # 忽略文件读取和解析的偶发异常\n pass\n \n # 若在项目中未提取到任何类名，亦无违规发生，即代表符合要求\n if not class_names:\n return True\n \n # 逐一验证所有提取到的类名\n for cls in class_names:\n # 类名的字符构成必须且只能是小写字母以及下划线\n if not re.match(r'^[a-z_]+$', cls):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串","validation_code":"import os\nimport re\n\ndef check_function_names(response: str, workspace_path: str) -> bool:\n if not os.path.exists(workspace_path):\n return False\n \n func_names = set()\n \n # 匹配常规的 JavaScript/TypeScript 函数定义: function name(...) \n pattern_func = re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\(')\n \n # 匹配赋值形式的函数: const/let/var name = (async) function(...) 或 (async) (...) => \n pattern_assign = re.compile(\n r'\\b(?:const|let|var)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=\\s*'\n r'(?:(?:async\\s+)?function\\b|(?:async\\s+)?(?:\\([^)]*\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>)'\n )\n \n # 匹配对象或类中的方法简写: name(...) {\n pattern_method = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\([^)]*\\)\\s*\\{')\n \n # 匹配 Python 常规函数定义\n pattern_py = re.compile(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n \n # 需要排除的关键字，防止方法简写的正则对控制流进行误捕获\n js_keywords = {\n 'if', 'else', 'for', 'while', 'do', 'switch', 'catch', 'finally', \n 'function', 'return', 'with', 'constructor', 'super'\n }\n \n # 遍历当前项目的所有常见可能包含代码的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.py')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 1. 查找常规函数声明\n for match in pattern_func.findall(content):\n func_names.add(match)\n \n # 2. 查找函数赋值\n for match in pattern_assign.findall(content):\n func_names.add(match)\n \n # 3. 查找方法定义，并剔除控制流等关键字\n for match in pattern_method.findall(content):\n if match not in js_keywords:\n func_names.add(match)\n \n # 4. 查找 Python 函数定义\n for match in pattern_py.findall(content):\n func_names.add(match)\n except Exception:\n pass\n \n # 检查所有提取出的函数名称是否为严格的回文字符串\n for name in func_names:\n if name != name[::-1]:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保你生成的每一个代码文件都精确包含 3 行注释","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n def parse_js(code, start_offset_line=0):\n # 匹配 JS 字符串和注释\n pattern = re.compile(\n r'(?P<string_double>\"(?:\\\\.|[^\"\\\\\\n])*\")|'\n r\"(?P<string_single>'(?:\\\\.|[^'\\\\\\n])*')|\"\n r'(?P<string_template>`(?:\\\\.|[^`\\\\])*`)|'\n r'(?P<comment_line>//[^\\n]*)|'\n r'(?P<comment_block>/\\*.*?\\*/)',\n re.DOTALL\n )\n lines = set()\n for m in pattern.finditer(code):\n if m.lastgroup in ['comment_line', 'comment_block']:\n start_line = code.count('\\n', 0, m.start()) + 1 + start_offset_line\n end_line = code.count('\\n', 0, m.end()) + 1 + start_offset_line\n for l in range(start_line, end_line + 1):\n lines.add(l)\n return lines\n\n def parse_css(code, start_offset_line=0):\n # 匹配 CSS 字符串和注释\n pattern = re.compile(\n r'(?P<string_double>\"(?:\\\\.|[^\"\\\\\\n])*\")|'\n r\"(?P<string_single>'(?:\\\\.|[^'\\\\\\n])*')|\"\n r'(?P<comment_block>/\\*.*?\\*/)',\n re.DOTALL\n )\n lines = set()\n for m in pattern.finditer(code):\n if m.lastgroup == 'comment_block':\n start_line = code.count('\\n', 0, m.start()) + 1 + start_offset_line\n end_line = code.count('\\n', 0, m.end()) + 1 + start_offset_line\n for l in range(start_line, end_line + 1):\n lines.add(l)\n return lines\n\n def parse_html(code):\n # 匹配 HTML 注释及内部嵌套的 JS 和 CSS\n pattern = re.compile(\n r'(?P<comment_html><!--.*?-->)|'\n r'(?P<script><script\\b[^>]*>.*?</script>)|'\n r'(?P<style><style\\b[^>]*>.*?</style>)|'\n r'(?P<string_double>\"(?:\\\\.|[^\"\\\\\\n])*\")|'\n r\"(?P<string_single>'(?:\\\\.|[^'\\\\\\n])*')\",\n re.IGNORECASE | re.DOTALL\n )\n lines = set()\n for m in pattern.finditer(code):\n if m.lastgroup == 'script':\n script_code = m.group('script')\n start_offset = code.count('\\n', 0, m.start())\n lines.update(parse_js(script_code, start_offset))\n elif m.lastgroup == 'style':\n style_code = m.group('style')\n start_offset = code.count('\\n', 0, m.start())\n lines.update(parse_css(style_code, start_offset))\n elif m.lastgroup == 'comment_html':\n start_line = code.count('\\n', 0, m.start()) + 1\n end_line = code.count('\\n', 0, m.end()) + 1\n for l in range(start_line, end_line + 1):\n lines.add(l)\n return lines\n\n def parse_py(code):\n # 匹配 Python 字符串和注释（防御性支持）\n pattern = re.compile(\n r'(?P<string_double_3>\"\"\"(?:\\\\.|.)*?\"\"\")|'\n r\"(?P<string_single_3>'''(?:\\\\.|.)*?''')|\"\n r'(?P<string_double>\"(?:\\\\.|[^\"\\\\\\n])*\")|'\n r\"(?P<string_single>'(?:\\\\.|[^'\\\\\\n])*')|\"\n r'(?P<comment_hash>#[^\\n]*)',\n re.DOTALL\n )\n lines = set()\n for m in pattern.finditer(code):\n if m.lastgroup == 'comment_hash':\n start_line = code.count('\\n', 0, m.start()) + 1\n lines.add(start_line)\n return lines\n\n # 定义需要检查的代码文件扩展名列表\n CODE_EXTS = {\n '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.rb', '.php', '.sh'\n }\n \n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in CODE_EXTS:\n has_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n code = f.read().replace('\\r\\n', '\\n').replace('\\r', '\\n')\n except Exception:\n continue\n \n if ext in ['.html', '.htm']:\n comment_lines = parse_html(code)\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', '.php']:\n comment_lines = parse_js(code)\n elif ext in ['.css', '.scss', '.less']:\n comment_lines = parse_css(code)\n elif ext in ['.py', '.rb', '.sh']:\n comment_lines = parse_py(code)\n else:\n comment_lines = parse_js(code)\n \n # 如果某个代码文件的注释总行数不满足刚好3行\n if len(comment_lines) != 3:\n return False\n \n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求","validation_code":"import os\nimport re\n\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\n # 匹配 OOP 类声明，例如：class my_class {\n oop_class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\b')\n # 匹配 HTML 中的 class 属性，例如：class=\"my_class another_class\"\n html_class_pattern = re.compile(r'\\bclass\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']')\n # 匹配 JS 中的 className 属性赋值，例如：element.className = \"my_class\"\n js_classname_pattern = re.compile(r'\\bclassName\\s*=\\s*[\"\\']([^\"\\']+)[\"\\']')\n # 匹配 JS 中的 classList 操作，例如：element.classList.add(\"my_class\")\n js_classlist_pattern = re.compile(r'\\bclassList\\.(?:add|remove|toggle|contains)\\(\\s*[\"\\']([^\"\\']+)[\"\\']')\n\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n \n if not os.path.isfile(file_path):\n continue\n \n try:\n # 按照系统约束，生成的文件均使用 UTF-8 进行编码\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略不可读的文件\n continue\n \n # 1. 检查面向对象（OOP）类声明\n for match in oop_class_pattern.finditer(content):\n class_name = match.group(1)\n if len(class_name) % 2 != 0:\n return False\n \n # 2. 检查 HTML 标签中的 class 属性\n for match in html_class_pattern.finditer(content):\n classes = match.group(1).split()\n for c in classes:\n if re.match(r'^[a-zA-Z_][a-zA-Z0-9_]*$', c):\n if len(c) % 2 != 0:\n return False\n \n # 3. 检查 JavaScript 代码中的 className 赋值\n for match in js_classname_pattern.finditer(content):\n classes = match.group(1).split()\n for c in classes:\n if re.match(r'^[a-zA-Z_][a-zA-Z0-9_]*$', c):\n if len(c) % 2 != 0:\n return False\n\n # 4. 检查 JavaScript 代码中的 classList 方法调用\n for match in js_classlist_pattern.finditer(content):\n classes = match.group(1).split()\n for c in classes:\n if re.match(r'^[a-zA-Z_][a-zA-Z0-9_]*$', c):\n if len(c) % 2 != 0:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n \n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n \n code_extensions = {'.html', '.css', '.js'}\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 按换行符进行拆分，以准确计入所有的行，包括代码中间及末尾的空行\n line_count = len(content.split('\\n'))\n \n # 判断如果包含偶数行，说明不满足要求，直接返回False\n if line_count % 2 == 0:\n return False\n except Exception:\n # 如果发生文件读取异常，则认为无法验证通过\n return False\n \n # 如果没有找到任何代码文件，说明也没有完成要求\n return found_code_file","tags":["Quantity","Exact Value"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件应全部使用2个空格来控制缩进","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对前端项目的常见代码文件进行推断，过滤掉如.md, .txt之类的文本文件\n code_extensions = {\".html\", \".css\", \".js\", \".ts\", \".jsx\", \".tsx\", \".vue\"}\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n continue\n \n positive_indents = set()\n \n for line in lines:\n # 忽略空行\n if not line.strip():\n continue\n \n # 匹配每行的前导空白字符\n match = re.match(r'^([ \\t]+)', line)\n if match:\n spaces = match.group(1)\n \n # 检查1：不得使用制表符控制缩进\n if '\\t' in spaces:\n return False\n \n length = len(spaces)\n if length > 0:\n positive_indents.add(length)\n \n if positive_indents:\n # 判断是否存在不能被4整除的偶数缩进（如 2, 6, 10...），这是2空格缩进的典型特征\n has_two_space_levels = any(i % 2 == 0 and i % 4 != 0 for i in positive_indents)\n # 判断是否存在4的倍数缩进（如 4, 8, 12...）\n has_four_space_levels = any(i % 4 == 0 for i in positive_indents)\n \n # 检查2：如果所有偶数层级的缩进全都是4的倍数，说明它仍然是4空格缩进的体系，不符合2空格的要求\n if has_four_space_levels and not has_two_space_levels:\n return False\n\n return True","tags":["Style","Layout"]},{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":3,"instruction":"越来越有意思了。现在来完善环境参数设置。在控制面板再增加一个“是否真空”的复选框，以及一个“重力加速度”的滑动条（范围设为1到20，让用户可以模拟月球或木星的重力）。当用户勾选“真空”时，底层的物理运算必须完全屏蔽风阻系数和空气阻力计算，此时无论两个物体的重量和类型如何，它们都必须绝对同时落地。而调整“重力加速度”滑动条应该实时改变当前的 $g$ 值，从而影响下落的整体快慢。请确保在调整这些环境参数后，系统状态能正确响应并反映在下一次的实验动画里。\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，生成的代码文件应全部使用4个空格来控制缩进，替换之前的2空格缩进要求；第二，确保在代码开发完成之后，测试所有的代码并保证它们均通过测试。","instruction_following_checklist":[{"constraint":"只要是你创建的代码文件，就必须使用UTF-8进行编码","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作区路径是否存在\n if not os.path.isdir(workspace_path):\n return False\n\n # 定义通常被视作代码或项目文本配置的文件扩展名\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.php', '.rb', '.go', '.rs', '.swift', '.kt', \n '.sh', '.bat', '.json', '.xml', '.yaml', '.yml', \n '.md', '.txt'\n }\n\n found_any_code_file = False\n\n # 遍历项目路径下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_any_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他如权限受限等异常情况\n continue\n\n # 只有当项目中确实生成了代码文件，且全部都能成功以 UTF-8 解析时才返回 True\n return found_any_code_file","tags":["Language","File Encoding"]},{"constraint":"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7","validation_code":"import os\nimport re\nimport ast\n\ndef check_variables_length(response: str, workspace_path: str) -> bool:\n variables = set()\n\n def parse_js(js_code: str):\n # 移除单行与多行注释\n js_code = re.sub(r'//.*', '', js_code)\n js_code = re.sub(r'/\\*.*?\\*/', '', js_code, flags=re.DOTALL)\n # 移除字符串字面量，防止误匹配内部的伪代码或参数\n js_code = re.sub(r'([\"\\'`])(?:(?=(\\\\?))\\2.)*?\\1', '', js_code, flags=re.DOTALL)\n\n # 1. 基本的 var / let / const 声明\n matches = re.finditer(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', js_code)\n for m in matches:\n variables.add(m.group(1))\n \n # 2. 对象解构赋值 {}\n destruct_obj = re.finditer(r'\\b(?:var|let|const)\\s*\\{([^}]+)\\}', js_code)\n for m in destruct_obj:\n for p in m.group(1).split(','):\n m_id = re.match(r'^\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*\\s*:\\s*)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', p)\n if m_id:\n variables.add(m_id.group(1))\n \n # 3. 数组解构赋值 []\n destruct_arr = re.finditer(r'\\b(?:var|let|const)\\s*\\[([^\\]]+)\\]', js_code)\n for m in destruct_arr:\n for p in m.group(1).split(','):\n m_id = re.match(r'^\\s*(?:\\.\\.\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', p)\n if m_id:\n variables.add(m_id.group(1))\n\n # 4. 普通函数参数\n func_params = re.finditer(r'\\bfunction\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)', js_code)\n for m in func_params:\n for p in m.group(1).split(','):\n m_id = re.match(r'^\\s*(?:\\.\\.\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', p)\n if m_id:\n variables.add(m_id.group(1))\n \n # 5. 箭头函数参数 (单参数无需括号形式)\n arrow_single = re.finditer(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>', js_code)\n for m in arrow_single:\n var_name = m.group(1)\n # 排除由于代码格式导致的关键字误认\n if var_name not in ['return', 'if', 'else', 'switch', 'for', 'while']:\n variables.add(var_name)\n \n # 6. 箭头函数参数 (带括号形式)\n arrow_multi = re.finditer(r'\\(([^)]*)\\)\\s*=>', js_code)\n for m in arrow_multi:\n for p in m.group(1).split(','):\n m_id = re.match(r'^\\s*(?:\\.\\.\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\b', p)\n if m_id:\n variables.add(m_id.group(1))\n \n # 7. try...catch 错误对象\n catch_blocks = re.finditer(r'\\bcatch\\s*\\(\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*\\)', js_code)\n for m in catch_blocks:\n variables.add(m.group(1))\n\n def parse_css(css_code: str):\n # 匹配 CSS 变量 (如 --main-color)\n css_vars = re.finditer(r'(--[a-zA-Z0-9_-]+)\\s*:', css_code)\n for m in css_vars:\n variables.add(m.group(1))\n\n def parse_py(py_code: str):\n # 严谨的 Python AST 语法树解析\n try:\n tree = ast.parse(py_code)\n for node in ast.walk(tree):\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\n variables.add(node.id)\n elif isinstance(node, ast.arg):\n variables.add(node.arg)\n except SyntaxError:\n pass\n\n def parse_html(html_code: str):\n # 解析 HTML 内嵌的 JS 和 CSS\n scripts = re.finditer(r'<script\\b[^>]*>(.*?)</script>', html_code, flags=re.IGNORECASE | re.DOTALL)\n for m in scripts:\n parse_js(m.group(1))\n styles = re.finditer(r'<style\\b[^>]*>(.*?)</style>', html_code, flags=re.IGNORECASE | re.DOTALL)\n for m in styles:\n parse_css(m.group(1))\n\n # 1. 遍历当前工作空间内的项目生成文件\n if os.path.exists(workspace_path):\n for root, _, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if file.endswith('.js'):\n parse_js(code)\n elif file.endswith('.css'):\n parse_css(code)\n elif file.endswith('.py'):\n parse_py(code)\n elif file.endswith('.html'):\n parse_html(code)\n except Exception:\n continue\n\n # 2. 检查最后一轮 AI 回复的 Markdown 代码块\n blocks = re.finditer(r'```([a-zA-Z]*)\\n(.*?)```', response, flags=re.DOTALL)\n for m in blocks:\n lang = m.group(1).lower()\n code = m.group(2)\n if lang in ['javascript', 'js']:\n parse_js(code)\n elif lang == 'css':\n parse_css(code)\n elif lang in ['python', 'py']:\n parse_py(code)\n elif lang == 'html' or lang == '':\n parse_html(code)\n\n # 3. 校验提取到的全部变量名长度\n for var in variables:\n if len(var) < 2 or len(var) > 7:\n return False\n\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 匹配 JavaScript 等语言中面向对象的类名（匹配 class 关键字后的合法标识符字符）\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z0-9_$]+)')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据用户指令推断，主要涉及前端开发文件\n if not file.endswith(('.js', '.html', '.ts', '.jsx', '.tsx')):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取当前文件中的所有类名\n classes = class_pattern.findall(content)\n for cls_name in classes:\n # 检查该类名的字符构成是否仅为小写字母以及下划线\n if not re.match(r'^[a-z_]+$', cls_name):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 针对 JS 和 HTML 的正则表达式模式，用于匹配各种常见函数的定义方式\n js_func_patterns = [\n # 1. 匹配标准函数声明：function foo() / async function foo() / function* foo()\n re.compile(r'\\bfunction\\s*(?:\\*\\s*)?([a-zA-Z_$][\\w$]*)\\s*\\('),\n \n # 2. 匹配变量赋值型函数：const foo = function() / let foo = () => {}\n re.compile(r'\\b(?:const|let|var)\\s+([a-zA-Z_$][\\w$]*)\\s*=\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)'),\n \n # 3. 匹配对象属性赋值型函数：foo: function() / foo: async () => {}\n re.compile(r'\\b([a-zA-Z_$][\\w$]*)\\s*:\\s*(?:async\\s+)?(?:function\\b|\\([^)]*\\)\\s*=>|[a-zA-Z_$][\\w$]*\\s*=>)'),\n \n # 4. 匹配类或对象的方法声明：foo() {} / async foo() {} / * foo() {} (严格排除常见关键字)\n re.compile(r'^\\s*(?:async\\s+)?(?:\\*\\s*)?(?!if\\b|for\\b|while\\b|switch\\b|catch\\b|with\\b|function\\b|return\\b|constructor\\b|super\\b)([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{', re.MULTILINE),\n \n # 5. 匹配 getter 和 setter：get foo() {} / set foo(val) {}\n re.compile(r'^\\s*(?:get|set)\\s+([a-zA-Z_$][\\w$]*)\\s*\\([^)]*\\)\\s*\\{', re.MULTILINE)\n ]\n \n # 针对 Python 的正则表达式模式\n py_func_patterns = [\n # 匹配标准函数和异步函数定义：def foo() / async def foo()\n re.compile(r'\\b(?:async\\s+)?def\\s+([a-zA-Z_]\\w*)\\s*\\(')\n ]\n \n def is_palindrome(s: str) -> bool:\n # 判断字符串是否是严格的回文字符串\n return s == s[::-1]\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n ext = os.path.splitext(file)[1].lower()\n \n if ext in ['.js', '.html']:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n for pattern in js_func_patterns:\n for match in pattern.finditer(content):\n func_name = match.group(1)\n if not is_palindrome(func_name):\n return False\n except Exception:\n continue\n \n elif ext == '.py':\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n for pattern in py_func_patterns:\n for match in pattern.finditer(content):\n func_name = match.group(1)\n if not is_palindrome(func_name):\n return False\n except Exception:\n continue\n\n return True","tags":["Style","Naming"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保你生成的每一个代码文件都精确包含 3 行注释","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n def count_js_css_comments(text: str, start_line: int = 0, is_css: bool = False) -> set:\n comment_lines = set()\n in_multi = False\n in_str = False\n str_char = ''\n lines = text.split('\\n')\n \n for i, line in enumerate(lines):\n abs_line = start_line + i\n # 如果处于多行注释状态，当前行必然包含注释\n if in_multi:\n comment_lines.add(abs_line)\n \n j = 0\n while j < len(line):\n # 正常非注释、非字符串状态下，寻找注释或字符串入口\n if not in_multi and not in_str:\n # 单行注释 (CSS 不支持单行注释，跳过此逻辑)\n if not is_css and line[j:j+2] == '//':\n comment_lines.add(abs_line)\n break # 行末之前的全是注释\n # 多行注释起点\n elif line[j:j+2] == '/*':\n in_multi = True\n comment_lines.add(abs_line)\n j += 1\n # 字符串起点\n elif line[j] in '\"\\'`':\n in_str = True\n str_char = line[j]\n \n # 字符串状态内，寻找转义符或结束引号\n elif in_str:\n if line[j] == '\\\\':\n j += 1\n elif line[j] == str_char:\n in_str = False\n \n # 多行注释内，寻找结束点\n elif in_multi:\n comment_lines.add(abs_line)\n if line[j:j+2] == '*/':\n in_multi = False\n j += 1\n \n j += 1\n \n return comment_lines\n\n def get_comment_lines(file_path: str) -> int:\n ext = os.path.splitext(file_path)[1].lower()\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n\n comment_lines = set()\n \n # HTML 综合解析\n if ext in ['.html', '.htm']:\n # 1. 解析 <script>\n scripts = re.finditer(r'<script.*?>\\s*(.*?)\\s*</script>', code, re.IGNORECASE | re.DOTALL)\n for m in scripts:\n start_line = code[:m.start(1)].count('\\n')\n lines = count_js_css_comments(m.group(1), start_line, is_css=False)\n comment_lines.update(lines)\n \n # 2. 解析 <style>\n styles = re.finditer(r'<style.*?>\\s*(.*?)\\s*</style>', code, re.IGNORECASE | re.DOTALL)\n for m in styles:\n start_line = code[:m.start(1)].count('\\n')\n lines = count_js_css_comments(m.group(1), start_line, is_css=True)\n comment_lines.update(lines)\n \n # 3. 解析纯 HTML 注释 (需排除被 script/style 包裹的伪注释)\n html_comments = re.finditer(r'<!--(.*?)-->', code, re.DOTALL)\n for m in html_comments:\n start_idx = m.start()\n inside_blocks = False\n \n # 简单排查该注释是否位于未闭合的 <script> 或 <style> 中\n last_script = code.rfind('<script', 0, start_idx)\n last_script_end = code.rfind('</script>', 0, start_idx)\n if last_script > last_script_end: \n inside_blocks = True\n \n last_style = code.rfind('<style', 0, start_idx)\n last_style_end = code.rfind('</style>', 0, start_idx)\n if last_style > last_style_end: \n inside_blocks = True\n \n if not inside_blocks:\n start_line = code[:m.start()].count('\\n')\n num_lines = m.group(0).count('\\n')\n for i in range(start_line, start_line + num_lines + 1):\n comment_lines.add(i)\n \n # 纯 JS / TS 等变体解析\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\n comment_lines.update(count_js_css_comments(code, 0, is_css=False))\n \n # 纯 CSS 解析\n elif ext == '.css':\n comment_lines.update(count_js_css_comments(code, 0, is_css=True))\n \n return len(comment_lines)\n\n # 主验证流程\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n num_comments = get_comment_lines(file_path)\n except Exception:\n # 读取或解析异常，可视为不满足条件\n return False\n \n # 要求“精确包含 3 行注释”\n if num_comments != 3:\n return False\n \n # 如果找到了代码文件且均通过，则返回True；如果没有产生代码文件，则判定为False\n return has_code_file","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 匹配面向对象语言中的类定义，例如 class MyClass\n # 捕获组提取类名标识符\n class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z_]\\w*)')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 仅检查可能的代码文件类型，涵盖本项目用到的前端文件及常见代码文件\n if not file.endswith(('.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py')):\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 忽略无法以UTF-8读取的文件\n continue\n \n # 找到文件中定义的所有类名\n classes = class_pattern.findall(content)\n for cls_name in classes:\n # 检查类名的长度是否严格满足偶数的要求\n if len(cls_name) % 2 != 0:\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_code_files_odd_lines(response: str, workspace_path: str) -> bool:\n import os\n\n if not os.path.exists(workspace_path):\n return False\n \n # 根据用户指令推断出的代码文件格式\n code_extensions = {'.html', '.css', '.js'}\n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\n content = f.read()\n \n # 使用按照 '\\n' 切割的方式可以准确统计出所有物理行，包含文件末尾以及中间的空行\n lines = content.split('\\n')\n line_count = len(lines)\n \n # 判断行数是否为奇数\n if line_count % 2 == 0:\n return False\n except Exception:\n # 如果文件无法正常读取，视作无法验证要求，返回False\n return False\n \n # 如果找到了至少一个代码文件，并且没有任何一个文件触发偶数行的 return False，则视为满足要求\n return found_code_file","tags":["Quantity","Exact Value"]},{"constraint":"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件应全部使用4个空格来控制缩进","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义属于代码文件的扩展名\n code_extensions = {'.html', '.css', '.js'}\n has_code_file = False\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果无法读取，判定为不合规\n return False\n \n for line in lines:\n # 忽略完全空白或只包含换行符的行\n if not line.strip():\n continue\n \n # 提取行首的空白字符\n leading_spaces = \"\"\n for char in line:\n if char in (' ', '\\t'):\n leading_spaces += char\n else:\n break\n \n # 忽略块级注释的星号开头行（常用于 /* ... */ 的内部行）\n if line.strip().startswith('*'):\n continue\n \n # 判断一：如果缩进中使用了制表符，则违反了使用空格控制缩进的要求\n if '\\t' in leading_spaces:\n return False\n \n # 判断二：如果行首空白字符的长度不是4的倍数，则说明没有严格使用4空格缩进\n if len(leading_spaces) % 4 != 0:\n return False\n\n # 必须至少生成过一个代码文件才算任务完成\n return has_code_file","tags":["Style","Layout"]},{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"最后一步，我们来提升一下视觉表现和数据统计功能。首先，根据下拉菜单的选择，用简单的CSS或者Emoji把代表物体的圆形替换掉，让“小球”和“羽毛”在视觉上直观区分开来。其次，在下落过程中，用Canvas或DOM动态生成一点简单的轨迹线或残影效果，方便用户直观对比它俩在半空中的高度差。最后，把底部的简单文字结果区改成一个“实验记录表”。每次实验双双落地结束后，把这一轮的“环境参数(重力/真空)、物体A参数、物体B参数、各自落地时间”作为新的一行追加到底部的表格中，方便用户对比历次实验的数据。\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，在每次对项目进行改动后，更新process-log.md文件以记录本轮完成的内容是必须执行的操作，替换此前更新claude-progress.txt的要求；第二，你提供的每个代码文件里，注释内容须100%使用中文撰写。","instruction_following_checklist":[{"constraint":"只要是你创建的代码文件，就必须使用UTF-8进行编码","validation_code":"def verify_utf8_encoding(response: str, workspace_path: str) -> bool:\n import os\n \n # 如果工作空间路径不存在，则无法验证是否满足条件，返回False\n if not os.path.exists(workspace_path):\n return False\n \n # 根据多轮指令，可能生成的代码文件和文档文件的拓展名\n target_extensions = {'.html', '.css', '.js', '.md', '.txt'}\n \n # 遍历当前项目的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出推断的代码文件类型\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 UTF-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明并非有效的 UTF-8 编码\n return False\n except Exception:\n # 忽略其他可能的文件访问异常（如权限问题等）\n continue\n \n return True","tags":["Language","File Encoding"]},{"constraint":"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7","validation_code":"def verify(response, workspace_path):\n import os\n import re\n\n def remove_comments_and_strings(code):\n # 移除多行注释\n code = re.sub(r'/\\*[\\s\\S]*?\\*/', '', code)\n # 移除单行注释\n code = re.sub(r'//.*', '', code)\n # 移除双引号字符串 (简单考虑转义)\n code = re.sub(r'\"[^\"\\\\\\r\\n]*(?:\\\\.[^\"\\\\\\r\\n]*)*\"', '\"\"', code)\n # 移除单引号字符串\n code = re.sub(r\"'[^'\\\\\\r\\n]*(?:\\\\.[^'\\\\\\r\\n]*)*'\", \"''\", code)\n # 移除模板字符串 (反引号)\n code = re.sub(r'`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`', '``', code)\n return code\n\n def check_variable_names(code):\n code = remove_comments_and_strings(code)\n variables = set()\n \n # 1. 匹配普通的 var, let, const 声明的首个变量\n decl_regex = re.compile(r'\\b(?:var|let|const)\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n for match in decl_regex.finditer(code):\n variables.add(match.group(1))\n \n # 2. 匹配对象解构赋值，例如 let {a, b: myB} = obj\n destruct_regex = re.compile(r'\\b(?:var|let|const)\\s+\\{([^}]+)\\}')\n for match in destruct_regex.finditer(code):\n vars_str = match.group(1)\n for v in vars_str.split(','):\n v = v.strip()\n if not v:\n continue\n # 处理带有别名的情况（取冒号后的实际变量名）\n if ':' in v:\n v = v.split(':', 1)[1].strip()\n if v.startswith('...'):\n v = v[3:].strip()\n v = v.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', v):\n variables.add(v)\n \n # 3. 匹配数组解构赋值，例如 let [a, b] = arr\n array_destruct_regex = re.compile(r'\\b(?:var|let|const)\\s+\\[([^\\]]+)\\]')\n for match in array_destruct_regex.finditer(code):\n vars_str = match.group(1)\n for v in vars_str.split(','):\n v = v.strip()\n if not v:\n continue\n if v.startswith('...'):\n v = v[3:].strip()\n v = v.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', v):\n variables.add(v)\n\n # 4. 函数名称\n func_name_regex = re.compile(r'\\bfunction\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n for match in func_name_regex.finditer(code):\n variables.add(match.group(1))\n\n # 5. 常规函数参数\n func_params_regex = re.compile(r'\\bfunction\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\s*\\(([^)]*)\\)')\n for match in func_params_regex.finditer(code):\n params_str = match.group(1)\n for p in params_str.split(','):\n p = p.strip()\n if not p:\n continue\n if p.startswith('...'):\n p = p[3:].strip()\n var_name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n variables.add(var_name)\n \n # 6. 带括号的箭头函数参数\n arrow_regex1 = re.compile(r'\\(([^)]*)\\)\\s*=>')\n for match in arrow_regex1.finditer(code):\n params_str = match.group(1)\n for p in params_str.split(','):\n p = p.strip()\n if not p:\n continue\n if p.startswith('...'):\n p = p[3:].strip()\n var_name = p.split('=')[0].strip()\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\n variables.add(var_name)\n \n # 7. 单参数箭头函数参数\n arrow_regex2 = re.compile(r'\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\s*=>')\n for match in arrow_regex2.finditer(code):\n variables.add(match.group(1))\n \n # 8. 类名称\n class_regex = re.compile(r'\\bclass\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\b')\n for match in class_regex.finditer(code):\n variables.add(match.group(1))\n\n # 遍历检验提取的所有变量名称的长度情况（使用原生的 len 函数统计字符数）\n for var in variables:\n var_len = len(var)\n if var_len < 2 or var_len > 7:\n return False\n \n return True\n\n # 遍历当前生成项目所有的 JavaScript 和 HTML 脚本区域\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n if file.endswith('.js'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n code = f.read()\n if not check_variable_names(code):\n return False\n except Exception:\n continue\n elif file.endswith('.html'):\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n html_code = f.read()\n # 忽略大小写查找 <script> 标签中的纯 JS 代码\n scripts = re.findall(r'<script[^>]*>([\\s\\S]*?)</script>', html_code, re.IGNORECASE)\n for script in scripts:\n if not check_variable_names(script):\n return False\n except Exception:\n continue\n\n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线","validation_code":"def verify(response, workspace_path):\n import os\n import re\n\n class_names = []\n \n # 1. 匹配 JS 的 ES6 class 声明\n regex_es6 = re.compile(r'\\bclass\\s+([A-Za-z0-9_]+)')\n \n # 2. 匹配 HTML 或 JSX 中的 class / className 属性\n regex_attr = re.compile(r'\\b(?:class|className)\\s*=\\s*[\\'\"]([^\\'\"]+)[\\'\"]')\n \n # 3. 匹配 classList DOM API 的操作\n regex_classlist = re.compile(r'\\bclassList\\.(?:add|remove|toggle|contains)\\(\\s*[\\'\"]([^\\'\"]+)[\\'\"]\\s*\\)')\n \n # 4. 匹配 getElementsByClassName 的使用\n regex_get_class = re.compile(r'\\bgetElementsByClassName\\(\\s*[\\'\"]([^\\'\"]+)[\\'\"]\\s*\\)')\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext not in ['.js', '.html', '.css', '.vue', '.jsx', '.tsx']:\n continue\n \n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n # 提取 JS 类定义\n matches = regex_es6.findall(content)\n class_names.extend(matches)\n \n # 提取属性分配中的类（需要按照空格切分，因为可能同时分配多个类名）\n matches = regex_attr.findall(content)\n for m in matches:\n class_names.extend(m.split())\n \n # 提取 classList 中的类\n matches = regex_classlist.findall(content)\n for m in matches:\n class_names.extend(m.split())\n \n # 提取 getElementsByClassName 中的类\n matches = regex_get_class.findall(content)\n for m in matches:\n class_names.extend(m.split())\n \n # 提取 CSS 样式中的类选择器\n css_text = ''\n if ext == '.css':\n css_text = content\n elif ext in ['.html', '.vue']:\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.DOTALL | re.IGNORECASE)\n css_text = '\\n'.join(styles)\n \n if css_text:\n # 首先过滤掉CSS内部可能被误认为点类选择器的不相关内容\n css_text = re.sub(r'/\\*.*?\\*/', '', css_text, flags=re.DOTALL) # 过滤多行注释\n css_text = re.sub(r'url\\([^)]*\\)', '', css_text) # 过滤 url(...) 内容\n css_text = re.sub(r'\"[^\"]*\"', '', css_text) # 过滤双引号字符串\n css_text = re.sub(r\"'[^']*'\", '', css_text) # 过滤单引号字符串\n \n # 匹配剩余文本中的 CSS 类选择器 (.classname)\n css_classes = re.findall(r'\\.([A-Za-z_][A-Za-z0-9_-]*)', css_text)\n class_names.extend(css_classes)\n\n # 汇总并对所有类名进行规则检验\n unique_class_names = set(class_names)\n for cls in unique_class_names:\n cls = cls.strip()\n if not cls:\n continue\n \n # 给定要求：字符构成只能是小写字母以及下划线\n if not re.match(r'^[a-z_]+$', cls):\n return False\n \n return True","tags":["Style","Naming"]},{"constraint":"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n def get_functions(code_text):\n functions = set()\n \n # 匹配1: 传统 function 声明 (例如: function myFunc(...) )\n p1 = re.compile(r'\\bfunction\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\(')\n functions.update(p1.findall(code_text))\n \n # 匹配2: 匿名函数赋值给变量或属性 (例如: myFunc = function(...) 或者 myFunc: function(...) )\n p2 = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*[:=]\\s*function\\s*(?:\\(|$)')\n functions.update(p2.findall(code_text))\n \n # 匹配3: 箭头函数赋值 (例如: myFunc = () => 或者 myFunc = a => )\n p3 = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*[:=]\\s*(?:\\([^)]*\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\s*=>')\n functions.update(p3.findall(code_text))\n \n # 匹配4: ES6 类方法或对象方法 (例如: myFunc(...) { )\n # 此类格式容易与原生控制流语法混淆，需排除 JS 中的常见保留字和关键字\n keywords = {\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'function', \n 'constructor', 'super', 'typeof', 'instanceof', 'new', 'else', \n 'try', 'finally', 'do', 'with'\n }\n p4 = re.compile(r'\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\s*\\([^)]*\\)\\s*\\{')\n matches = p4.findall(code_text)\n for m in matches:\n if m not in keywords and not m.isdigit():\n functions.add(m)\n \n # 匹配5: Python 等后端函数定义 (例如: def my_func(...): )\n p_py = re.compile(r'\\bdef\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\s*\\(')\n functions.update(p_py.findall(code_text))\n \n return functions\n\n # 遍历工作空间目录中的所有可能包含代码的文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 根据项目指令，文件主要为 HTML, JS，但扩大范围以适配衍生类型\n if file.endswith(('.html', '.js', '.ts', '.jsx', '.tsx', '.py')):\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 提取文件中的所有函数名称\n funcs = get_functions(content)\n \n # 检查每个提取出的函数名称是否为回文\n for func_name in funcs:\n if func_name != func_name[::-1]:\n return False\n except Exception:\n # 忽略读取异常的文件，避免中断检查流程\n pass\n\n # 若未找到任何违背条件（非回文）的函数，则符合要求\n return True","tags":["Style","Naming"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"确保你生成的每一个代码文件都精确包含 3 行注释","validation_code":"def check_comment_lines(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 根据用户指令推断，科普实验网站的核心代码文件类型为 HTML, CSS, JS\n code_extensions = {'.html', '.css', '.js'}\n code_file_count = 0\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_file_count += 1\n filepath = os.path.join(root, file)\n try:\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\n text = f.read()\n except Exception:\n continue\n \n # 针对不同扩展名的代码构建能够区分出字符串和不同类型注释的正则表达式\n # 使用 re.DOTALL 使得 . 能够匹配跨行字符\n if ext == '.js':\n pattern = re.compile(\n r'(?P<string>\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"|\\'[^\\'\\\\]*(?:\\\\.[^\\'\\\\]*)*\\'|`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`)'\n r'|(?P<mcomment>/\\*.*?\\*/)'\n r'|(?P<scomment>//[^\\n]*)',\n re.DOTALL\n )\n elif ext == '.css':\n pattern = re.compile(\n r'(?P<string>\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"|\\'[^\\'\\\\]*(?:\\\\.[^\\'\\\\]*)*\\')'\n r'|(?P<mcomment>/\\*.*?\\*/)',\n re.DOTALL\n )\n elif ext == '.html':\n pattern = re.compile(\n r'(?P<string>\"[^\"\\\\]*(?:\\\\.[^\"\\\\]*)*\"|\\'[^\\'\\\\]*(?:\\\\.[^\\'\\\\]*)*\\'|`[^`\\\\]*(?:\\\\.[^`\\\\]*)*`)'\n r'|(?P<hcomment><!--.*?-->)'\n r'|(?P<mcomment>/\\*.*?\\*/)'\n r'|(?P<scomment>(?<!:)//[^\\n]*)',\n re.DOTALL\n )\n \n comment_lines = set()\n for m in pattern.finditer(text):\n group_dict = m.groupdict()\n start, end = -1, -1\n \n if group_dict.get('mcomment') is not None:\n start, end = m.start('mcomment'), m.end('mcomment')\n elif group_dict.get('hcomment') is not None:\n start, end = m.start('hcomment'), m.end('hcomment')\n elif group_dict.get('scomment') is not None:\n start, end = m.start('scomment'), m.end('scomment')\n \n if start != -1 and end != -1:\n # 基于匹配的首尾索引，计算之前换行符的数量来确定行号\n start_line = text.count('\\n', 0, start)\n end_line = text.count('\\n', 0, end - 1) if end > start else start_line\n for line_num in range(start_line, end_line + 1):\n comment_lines.add(line_num)\n \n # 如果有任何一个代码文件提取出的注释占据的总行数不等于 3，说明违背了该要求\n if len(comment_lines) != 3:\n return False\n\n # 考虑到项目应当生成至少一个代码文件，如果没有找到代码文件，也视作不合规\n if code_file_count == 0:\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求","validation_code":"import os\nimport re\n\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\n # 匹配 JavaScript 中使用 class 关键字定义的类名（如 class MyClass）\n js_class_pattern = re.compile(r'\\bclass\\s+([a-zA-Z0-9_]+)\\b')\n # 匹配 HTML 中 class 属性指定的多个或单个类名\n html_class_pattern = re.compile(r'\\bclass\\s*=\\s*([\"\\'])(.*?)\\1', re.IGNORECASE)\n # 匹配 CSS 中的类选择器（以点 . 开头，由字母/下划线起头后接合法标识符）\n css_class_pattern = re.compile(r'\\.([a-zA-Z_][a-zA-Z0-9_]*)\\b')\n \n # 匹配各类前端注释（JS 单多行注释，HTML/CSS 多行注释）以在解析前移除\n comment_pattern = re.compile(r'//.*|/\\*.*?\\*/|<!--.*?-->', re.DOTALL)\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n # 仅处理前端常见的代码文件格式\n if not file.endswith(('.js', '.html', '.css', '.ts', '.jsx', '.tsx')):\n continue\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 移除所有注释内容，防止因匹配注释中的示例代码导致误报\n content = re.sub(comment_pattern, '', content)\n \n # 1. 提取并验证 JS/TS 代码中的类名\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html')):\n js_text = content\n if file.endswith('.html'):\n # 仅提取 HTML 中的 script 标签内部的内容\n scripts = re.findall(r'<script\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\n js_text = '\\n'.join(scripts)\n \n for cls_name in js_class_pattern.findall(js_text):\n # 约束：由于未明确指明“字符数”，严格依据原则必须调用 count_word\n if count_word(cls_name) % 2 != 0:\n return False\n \n # 2. 提取并验证 HTML 等标签内部 class 属性赋值的类名\n if file.endswith(('.html', '.jsx', '.tsx')):\n for match in html_class_pattern.findall(content):\n # match[1] 对应的是具体的多个/单个类名值字符串\n classes_str = match[1]\n for cls in classes_str.split():\n if count_word(cls) % 2 != 0:\n return False\n \n # 3. 提取并验证 CSS 样式中的类选择器名\n if file.endswith(('.css', '.html')):\n css_text = content\n if file.endswith('.html'):\n # 仅提取 HTML 中的 style 标签内部的 CSS 内容\n styles = re.findall(r'<style\\b[^>]*>(.*?)</style>', content, re.DOTALL | re.IGNORECASE)\n css_text = '\\n'.join(styles)\n \n for cls_name in css_class_pattern.findall(css_text):\n if count_word(cls_name) % 2 != 0:\n return False\n\n except Exception:\n # 遇到非文本类型、编码异常等特殊情况直接忽略，不影响后续校验\n continue\n \n return True","tags":["Style","Naming"]},{"constraint":"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数","validation_code":"def check_code_lines_are_odd(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据指令内容推断，项目主要为前端Web代码\n code_extensions = {'.html', '.css', '.js'}\n \n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 使用 '\\n' 分割可以精确统计出文本编辑器视角的行数，自动将空行计算在内\n lines = content.split('\\n')\n total_lines = len(lines)\n \n # 检查总行数是否为偶数\n if total_lines % 2 == 0:\n return False\n except Exception:\n # 如果遇到文件无法读取（例如非文本文件混入或编码错误等），视作不满足要求\n return False\n \n return True","tags":["Quantity","Exact Value"]},{"constraint":"在每次对项目进行改动后，更新process-log.md文件以记录本轮完成的内容是必须执行的操作","validation_code":"","tags":["Environment","Logging"]},{"constraint":"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /","validation_code":"","tags":["Environment","File Path"]},{"constraint":"生成的代码文件应全部使用4个空格来控制缩进","validation_code":"import os\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n # 根据用户指令推断，目标代码文件主要为这三种\n code_extensions = {'.html', '.css', '.js'}\n \n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n \n # 仅检查代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file_name)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = f.readlines()\n except Exception:\n # 如果由于某种原因无法读取，则跳过\n continue\n \n for line in lines:\n stripped_line = line.strip()\n # 1. 忽略空行\n if not stripped_line:\n continue\n \n # 2. 提取当前行的前导空白字符（行首的空格或制表符）\n leading_whitespace = \"\"\n for char in line:\n if char in (' ', '\\t'):\n leading_whitespace += char\n else:\n break\n \n # 3. 忽略完全没有缩进的顶层代码行\n if not leading_whitespace:\n continue\n \n # 4. 严格禁止使用制表符来控制缩进\n if '\\t' in leading_whitespace:\n return False\n \n # 5. 忽略多行注释对齐场景 (如行首有效字符是以 '*' 开头的)\n if stripped_line.startswith('*'):\n continue\n \n # 6. 验证缩进的空格数量是否严格为 4 的倍数\n if len(leading_whitespace) % 4 != 0:\n return False\n\n # 若所有代码文件均通过了上述严格缩进检测，则返回True\n return True","tags":["Style","Layout"]},{"constraint":"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"你提供的每个代码文件里，注释内容须100%使用中文撰写","validation_code":"","tags":["Language","Comment Language"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用HTML、CSS和原生JavaScript写一个“两个物体同时下落”的科普实验网站。第一步，先搭个基础的页面骨架：页面左侧是控制面板，里面先放两个下拉菜单（物体A类型、物体B类型，选项暂定为小球和羽毛），以及几个表示参数的占位滑动条；页面右侧是相对宽大的实验显示区，在顶部放置两个不同颜色的圆形元素代表将要下落的物体；在实验区底部或页面最下方，放置“开始”和“重置”两个按钮，并预留一个空的div作为实验结果展示区。目前不需要实现任何动画和复杂的物理逻辑，只要把布局划分清晰，加上基础的CSS样式使其看起来像个整洁的控制台即可。\\n另外请你注意，后续所有轮次必须始终遵循以下要求：第一，请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容；第二，在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"只要是你创建的代码文件，就必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能创建的代码文件扩展名\\n target_extensions = {'.html', '.css', '.js', '.md', '.json', '.txt', '.htm'}\\n \\n # 如果项目目录不存在，说明并未生成任何文件，没有违背\\\"只要创建就必须...\\\"的要求\\n if not os.path.exists(workspace_path):\\n return True\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅针对推断出的代码文件类型进行验证，避免读取图片等二进制文件\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError 异常，则说明不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他非编码相关的系统级别读取异常（如权限不足等）\\n pass\\n \\n # 如果所有被检查的文件均未抛出解码异常，判定为满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n \\n def extract_js_variables(code_text):\\n # 移除多行注释、单行注释及所有类型的字符串常量（防止误提取字符串内的单词）\\n code_text = re.compile(r'/\\\\*.*?\\\\*/', re.DOTALL).sub('', code_text)\\n code_text = re.compile(r'//.*').sub('', code_text)\\n code_text = re.compile(r'\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"').sub('', code_text)\\n code_text = re.compile(r\\\"'[^'\\\\\\\\]*(?:\\\\\\\\.[^'\\\\\\\\]*)*'\\\").sub('', code_text)\\n code_text = re.compile(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', re.DOTALL).sub('', code_text)\\n\\n variables = set()\\n\\n # 1. 提取基本的 var, let, const 声明的变量\\n matches = re.finditer(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', code_text)\\n for m in matches:\\n variables.add(m.group(1))\\n\\n # 2. 尝试从逗号分隔的多变量声明中提取简单变量（如 let a = 1, b = 2;）\\n decl_blocks = re.findall(r'\\\\b(?:var|let|const)\\\\s+([^;\\\\n]+)', code_text)\\n for block in decl_blocks:\\n # 移除所有括号内部内容，避免误入复杂的解构赋值和对象字面量\\n prev_len = -1\\n while len(block) != prev_len:\\n prev_len = len(block)\\n block = re.sub(r'\\\\{[^}]*\\\\}', '', block)\\n block = re.sub(r'\\\\[[^\\\\]]*\\\\]', '', block)\\n block = re.sub(r'\\\\([^)]*\\\\)', '', block)\\n \\n parts = block.split(',')\\n for p in parts:\\n p = p.strip()\\n # 仅保留赋值等号左侧的内容\\n var_name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n variables.add(var_name)\\n\\n # 3. 提取普通函数参数\\n func_matches = re.finditer(r'\\\\bfunction\\\\s+[a-zA-Z_$0-9]*\\\\s*\\\\(([^)]*)\\\\)', code_text)\\n for m in func_matches:\\n params_str = m.group(1)\\n params = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', params_str)\\n for p in params:\\n variables.add(p)\\n \\n # 4. 提取箭头函数带括号的参数\\n arrow_matches = re.finditer(r'\\\\(([^)]*)\\\\)\\\\s*=>', code_text)\\n for m in arrow_matches:\\n params_str = m.group(1)\\n params = re.findall(r'[a-zA-Z_$][a-zA-Z0-9_$]*', params_str)\\n for p in params:\\n variables.add(p)\\n\\n # 5. 提取箭头函数无括号的单参数\\n single_arrow_matches = re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', code_text)\\n for m in single_arrow_matches:\\n variables.add(m.group(1))\\n\\n return variables\\n\\n def extract_css_variables(code_text):\\n # 移除 CSS 注释后匹配 CSS 原生变量\\n code_text = re.compile(r'/\\\\*.*?\\\\*/', re.DOTALL).sub('', code_text)\\n variables = set()\\n matches = re.finditer(r'--([a-zA-Z0-9_-]+)\\\\s*:', code_text)\\n for m in matches:\\n variables.add(m.group(1))\\n return variables\\n\\n def get_codes_from_text(text):\\n js_codes, css_codes = [], []\\n \\n # 提取独立的 js/css 块\\n js_blocks = re.findall(r'```(?:js|javascript|JS)\\\\s+(.*?)```', text, re.DOTALL)\\n js_codes.extend(js_blocks)\\n css_blocks = re.findall(r'```(?:css|CSS)\\\\s+(.*?)```', text, re.DOTALL)\\n css_codes.extend(css_blocks)\\n \\n # 提取 html 块中的 script 和 style 标签内容\\n html_blocks = re.findall(r'```(?:html|HTML)\\\\s+(.*?)```', text, re.DOTALL)\\n for h_block in html_blocks:\\n script_blocks = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', h_block, re.DOTALL | re.IGNORECASE)\\n js_codes.extend(script_blocks)\\n style_blocks = re.findall(r'<style.*?>\\\\s*(.*?)\\\\s*</style>', h_block, re.DOTALL | re.IGNORECASE)\\n css_codes.extend(style_blocks)\\n \\n return js_codes, css_codes\\n\\n all_js = []\\n all_css = []\\n\\n # 从大模型回复的内容中提取代码\\n js_resp, css_resp = get_codes_from_text(response)\\n all_js.extend(js_resp)\\n all_css.extend(css_resp)\\n\\n # 从工作区文件中提取代码\\n if os.path.exists(workspace_path):\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n all_js.append(f.read())\\n except:\\n pass\\n elif file.endswith('.css'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n all_css.append(f.read())\\n except:\\n pass\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n script_blocks = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', content, re.DOTALL | re.IGNORECASE)\\n all_js.extend(script_blocks)\\n style_blocks = re.findall(r'<style.*?>\\\\s*(.*?)\\\\s*</style>', content, re.DOTALL | re.IGNORECASE)\\n all_css.extend(style_blocks)\\n except:\\n pass\\n\\n # 聚合找到的所有变量名\\n variables_found = set()\\n for j_code in all_js:\\n variables_found.update(extract_js_variables(j_code))\\n for c_code in all_css:\\n variables_found.update(extract_css_variables(c_code))\\n\\n # 判断是否全部符合 2 <= 长度 <= 7 \\n for var in variables_found:\\n if len(var) < 2 or len(var) > 7:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对不同语言用于提取类名的正则表达式\\n html_class_re = re.compile(r'\\\\bclass\\\\s*=\\\\s*([\\\"\\\\'])(.*?)\\\\1', re.IGNORECASE)\\n css_class_re = re.compile(r'\\\\.([a-zA-Z_-][a-zA-Z0-9_-]*)')\\n js_class_re = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n \\n # 类名的合法性判断正则：只能由小写字母和下划线组成\\n valid_pattern = re.compile(r'^[a-z_]+$')\\n\\n # 遍历当前工作空间内的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 本项目主要涉及 HTML, CSS, JavaScript 文件\\n if ext in ['.html', '.css', '.js']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n css_content = \\\"\\\"\\n js_content = \\\"\\\"\\n\\n if ext == '.html':\\n # 去除HTML注释，防止误判注释里的例子代码\\n content_clean = re.sub(r'<!--.*?-->', '', content, flags=re.DOTALL)\\n \\n # 检查 HTML 的 class=\\\"...\\\" 属性里引用的类名\\n for match in html_class_re.finditer(content_clean):\\n class_str = match.group(2)\\n # HTML class 属性内可以是用空格隔开的多个类名\\n for c_name in class_str.split():\\n if not valid_pattern.match(c_name):\\n return False\\n \\n # 提取内嵌的 <style> 与 <script> 内容待后续分别检查\\n style_contents = re.findall(r'<style[^>]*>(.*?)</style>', content_clean, flags=re.IGNORECASE | re.DOTALL)\\n css_content = ' '.join(style_contents)\\n \\n script_contents = re.findall(r'<script[^>]*>(.*?)</script>', content_clean, flags=re.IGNORECASE | re.DOTALL)\\n js_content = ' '.join(script_contents)\\n \\n elif ext == '.css':\\n css_content = content\\n elif ext == '.js':\\n js_content = content\\n \\n # 检查 CSS 类选择器\\n if css_content:\\n # 过滤 CSS 中的注释\\n css_content = re.sub(r'/\\\\*.*?\\\\*/', '', css_content, flags=re.DOTALL)\\n # 过滤 CSS 中的字符串（单引号或双引号）\\n css_content = re.sub(r'([\\\"\\\\'])(?:\\\\\\\\.|[^\\\\\\\\])*?\\\\1', '', css_content)\\n # 过滤 CSS 中的 url(...) 以防点号导致误认为类选择器\\n css_content = re.sub(r'url\\\\([^)]*\\\\)', '', css_content, flags=re.IGNORECASE)\\n \\n for match in css_class_re.finditer(css_content):\\n c_name = match.group(1)\\n if not valid_pattern.match(c_name):\\n return False\\n \\n # 检查 JS 定义的类\\n if js_content:\\n # 过滤 JS 单行注释\\n js_content = re.sub(r'//.*', '', js_content)\\n # 过滤 JS 多行注释\\n js_content = re.sub(r'/\\\\*.*?\\\\*/', '', js_content, flags=re.DOTALL)\\n # 过滤 JS 字符串（反引号、单引号或双引号）\\n js_content = re.sub(r'([\\\"\\\\'`])(?:\\\\\\\\.|[^\\\\\\\\])*?\\\\1', '', js_content)\\n \\n for match in js_class_re.finditer(js_content):\\n c_name = match.group(1)\\n if not valid_pattern.match(c_name):\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 忽略的关键字，防止 p3 正则误捕获常见的JS控制流语句和内置关键字\\n keywords = {\\n 'if', 'for', 'while', 'switch', 'catch', 'function', 'constructor',\\n 'return', 'with', 'async', 'await', 'class', 'else', 'elif', 'except'\\n }\\n \\n # 前缀断言：确保匹配的不是 CSS 类、伪类，或是字符串中间的一部分\\n # 允许前面有空格、换行、点号(.)、井号(#)等，但不允许有冒号(:)、连字符(-)或 @ 符号\\n prefix = r'(?<![a-zA-Z0-9_$\\\\-:@])'\\n \\n # 1. 匹配标准 JS 函数声明：function foo()\\n p1 = re.compile(prefix + r'function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n \\n # 2. 匹配函数表达式赋值或箭头函数：foo = function() 或 foo = () => 或 foo: () =>\\n p2 = re.compile(prefix + r'([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*(?:(?<![+\\\\-*/%&|^=<>!])[:=](?![=]))\\\\s*(?:async\\\\s+)?(?:function(?![a-zA-Z0-9_$])|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*=>)')\\n \\n # 3. 匹配类方法或对象简写方法：foo() { \\n p3 = re.compile(prefix + r'(?:(?:async\\\\s+)?(?:get\\\\s+|set\\\\s+))?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # 4. 匹配 Python 函数声明 (兜底处理)：def foo()\\n p4 = re.compile(prefix + r'def\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n \\n all_functions = set()\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 匹配 p1\\n for match in p1.findall(content):\\n all_functions.add(match)\\n \\n # 匹配 p2\\n for match in p2.findall(content):\\n all_functions.add(match)\\n \\n # 匹配 p3\\n for match in p3.findall(content):\\n if match not in keywords:\\n all_functions.add(match)\\n \\n # 匹配 p4\\n for match in p4.findall(content):\\n all_functions.add(match)\\n \\n except Exception:\\n # 忽略无法用 utf-8 读取的文件 (如图片等文件)\\n pass\\n \\n # 验证所有收集到的函数名是否为严格的回文字符串\\n for func_name in all_functions:\\n if func_name != func_name[::-1]:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 3 行注释\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in code_extensions:\\n continue\\n \\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n return False\\n \\n comment_lines_count = 0\\n in_block_comment = False\\n \\n for line in lines:\\n s_line = line.strip()\\n \\n if ext == '.html':\\n if '<!--' in s_line and '-->' in s_line:\\n comment_lines_count += 1\\n elif '<!--' in s_line:\\n in_block_comment = True\\n comment_lines_count += 1\\n elif '-->' in s_line and in_block_comment:\\n in_block_comment = False\\n comment_lines_count += 1\\n elif in_block_comment:\\n comment_lines_count += 1\\n else:\\n if ext == '.js' and s_line.startswith('//'):\\n comment_lines_count += 1\\n continue\\n \\n if '/*' in s_line and '*/' in s_line:\\n comment_lines_count += 1\\n elif '/*' in s_line:\\n in_block_comment = True\\n comment_lines_count += 1\\n elif '*/' in s_line and in_block_comment:\\n in_block_comment = False\\n comment_lines_count += 1\\n elif in_block_comment:\\n comment_lines_count += 1\\n \\n if comment_lines_count != 3:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求\",\n \"validation_code\": \"def check_class_name_length_is_even(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n classes = set()\\n \\n # 正则表达式用于匹配不同代码场景下的类名\\n # 1. 匹配 JS 等代码中的 class 声明，如：class MyClass\\n class_decl_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z_][A-Za-z0-9_]*)\\\\b')\\n # 2. 匹配 HTML 中的 class 属性，如：class=\\\"panel control\\\"\\n html_class_pattern = re.compile(r'\\\\bclass\\\\s*=\\\\s*([\\\"\\\\'])(.*?)\\\\1', re.IGNORECASE)\\n # 3. 匹配 CSS 文件中的类选择器，如：.box_container {\\n css_class_pattern = re.compile(r'\\\\.([A-Za-z_][A-Za-z0-9_-]*)(?=[ \\\\t\\\\n\\\\r]*[\\\\{,>+~:\\\\.])')\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 跳过常见二进制及字体图片文件\\n if ext in ['.png', '.jpg', '.jpeg', '.gif', '.ico', '.svg', '.woff', '.woff2', '.ttf', '.eot', '.mp4', '.webm']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 1. 提取编程语言中的 class 声明\\n matches1 = class_decl_pattern.findall(content)\\n classes.update(matches1)\\n \\n # 2. 提取 HTML 中 class 属性定义的类名（如含多个以空格分隔）\\n matches2 = html_class_pattern.findall(content)\\n for _, cls_str in matches2:\\n for c in cls_str.split():\\n # 确保提取到的是合法的类名（过滤非预期的模板特殊符号）\\n if re.match(r'^[A-Za-z_][A-Za-z0-9_-]*$', c):\\n classes.add(c)\\n \\n # 3. 提取专门在 CSS 文件中的类选择器\\n if ext == '.css':\\n matches3 = css_class_pattern.findall(content)\\n classes.update(matches3)\\n \\n # 验证每个收集到的类名，其名称长度是否严格为偶数\\n for cls in classes:\\n if len(cls) % 2 != 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_odd_line_counts(response: str, workspace_path: str) -> bool:\\n import os\\n\\n target_extensions = {'.html', '.css', '.js', '.md'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in target_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n # 统计包含空行在内的所有行数\\n total_lines = len(content.split('\\\\n'))\\n \\n if total_lines % 2 == 0:\\n return False\\n except Exception:\\n continue\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"请在每次调整项目后，务必更新 progress.md 文件，以便记录当前这轮完成的任务内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"不错，现在咱们把基础的下落动画和按钮交互加进去。点击“开始”按钮后，右侧实验区里的两个物体要同时从顶部开始匀加速下落，直到触碰到底部边界停止。使用 `requestAnimationFrame` 来实现平滑的动画效果。点击“重置”按钮能让两个物体瞬间回到初始位置。这里先只考虑基础重力（比如设定 $g=9.8$），暂时不要管空气阻力，所以两个物体应该完全同步下落。当它们碰到页面底部时，在预留的结果展示区里打印出一句话，比如“实验完成，落地耗时：X.XX 秒”。\\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中；第二，请在生成的所有代码文件中，使用正好 4 个空格来完成每一处缩进。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"只要是你创建的代码文件，就必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令推断可能的代码文件类型\\n code_extensions = {'.html', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试使用 UTF-8 进行解码\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果发生解码错误，说明不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他读取异常，如权限问题等\\n continue\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def extract_vars_from_tokens(tokens):\\n vars_found = set()\\n filtered_tokens = []\\n \\n level = 0\\n ignore_level = -1\\n \\n for t in tokens:\\n if t in ('{', '[', '('):\\n level += 1\\n \\n if ignore_level != -1:\\n # 若跳出了忽略所在的层级，解除忽略状态\\n if level < ignore_level:\\n ignore_level = -1\\n # 在同一层级遇到逗号或分号，结束赋值的忽略状态\\n elif level == ignore_level and t in (',', ';'):\\n ignore_level = -1\\n # 对于顶层的换行，也可以作为声明赋值结束的标志\\n elif level == 0 and level == ignore_level and t == '\\\\n':\\n ignore_level = -1\\n \\n if ignore_level == -1:\\n if t == '=':\\n ignore_level = level\\n else:\\n filtered_tokens.append(t)\\n \\n if t in ('}', ']', ')'):\\n level -= 1\\n if ignore_level != -1 and level < ignore_level:\\n ignore_level = -1\\n \\n idx = 0\\n # 排除JS中的大部分可能被误认的关键字\\n keywords = {\\n 'in', 'of', 'var', 'let', 'const', 'function', 'catch', 'true', 'false', \\n 'null', 'undefined', 'return', 'await', 'yield', 'if', 'else', 'for', \\n 'while', 'do', 'switch', 'case', 'break', 'continue', 'default', 'try', \\n 'finally', 'class', 'extends', 'super', 'import', 'export', 'new', \\n 'this', 'typeof', 'instanceof', 'void', 'delete'\\n }\\n \\n while idx < len(filtered_tokens):\\n t = filtered_tokens[idx]\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t) and t not in keywords:\\n # 检查标识符是否是对象解构中的属性名（其后紧跟冒号）\\n next_idx = idx + 1\\n while next_idx < len(filtered_tokens) and filtered_tokens[next_idx] == '\\\\n':\\n next_idx += 1\\n \\n if next_idx < len(filtered_tokens) and filtered_tokens[next_idx] == ':':\\n idx = next_idx + 1\\n continue\\n else:\\n vars_found.add(t)\\n idx += 1\\n \\n return vars_found\\n\\n variables = set()\\n has_target_files = False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n code = \\\"\\\"\\n \\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n except Exception:\\n continue\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n html = f.read()\\n # 仅提取 HTML 文件中 script 标签内的内容\\n scripts = re.findall(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', html, flags=re.DOTALL | re.IGNORECASE)\\n code = \\\"\\\\n\\\".join(scripts)\\n except Exception:\\n continue\\n \\n if not code.strip():\\n continue\\n \\n has_target_files = True\\n\\n # 去除字符串字面量内容，防止引发解析错误\\n code_clean = re.sub(r'\\\\'(?:\\\\\\\\.|[^\\\\\\\\\\\\'])*\\\\'', \\\"''\\\", code)\\n code_clean = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\\\\\\\\\"])*\\\"', '\\\"\\\"', code_clean)\\n code_clean = re.sub(r'`(?:\\\\\\\\.|[^\\\\\\\\`])*`', '``', code_clean)\\n \\n # 去除注释\\n code_clean = re.sub(r'//.*', '', code_clean)\\n code_clean = re.sub(r'/\\\\*.*?\\\\*/', '', code_clean, flags=re.DOTALL)\\n\\n # 单参数箭头函数提前提取（如: \\\"data =>\\\"）\\n single_arrows = re.findall(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', code_clean)\\n for sa in single_arrows:\\n if sa not in ('in', 'of', 'var', 'let', 'const', 'function', 'catch', 'true', 'false', 'null', 'undefined'):\\n variables.add(sa)\\n\\n # 多参数箭头函数提前提取（如: \\\"(a, b) =>\\\"）\\n multi_arrows = re.findall(r'\\\\(([^)]*)\\\\)\\\\s*=>', code_clean)\\n for ma in multi_arrows:\\n arr_tokens = re.findall(r'\\\\n|\\\\b(?:var|let|const|function|in|of|catch)\\\\b|[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\\\[\\\\](),;:]', ma)\\n variables.update(extract_vars_from_tokens(arr_tokens))\\n\\n # 生成全局片段 Token 序列\\n tokens = re.findall(r'\\\\n|\\\\b(?:var|let|const|function|in|of|catch)\\\\b|[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\\\[\\\\](),;:]', code_clean)\\n \\n i = 0\\n while i < len(tokens):\\n t = tokens[i]\\n \\n # 变量声明\\n if t in ('var', 'let', 'const'):\\n i += 1\\n decl_tokens = []\\n level = 0\\n while i < len(tokens):\\n curr = tokens[i]\\n if curr in ('{', '[', '('):\\n level += 1\\n elif curr in ('}', ']', ')'):\\n level -= 1\\n \\n decl_tokens.append(curr)\\n \\n # 顶层的分号和换行意味着声明结束\\n if level == 0 and curr in (';', '\\\\n'):\\n break\\n # for...in / for...of 的中途结束\\n if level == 0 and curr in ('in', 'of'):\\n break\\n \\n i += 1\\n variables.update(extract_vars_from_tokens(decl_tokens))\\n \\n # 函数定义\\n elif t == 'function':\\n i += 1\\n # 匹配函数名\\n if i < len(tokens) and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]) and tokens[i] not in ('(', '{', '['):\\n variables.add(tokens[i])\\n i += 1\\n # 匹配参数列表\\n if i < len(tokens) and tokens[i] == '(':\\n param_tokens = []\\n level = 1\\n param_tokens.append(tokens[i])\\n i += 1\\n while i < len(tokens) and level > 0:\\n curr = tokens[i]\\n if curr == '(': level += 1\\n elif curr == ')': level -= 1\\n param_tokens.append(curr)\\n i += 1\\n variables.update(extract_vars_from_tokens(param_tokens))\\n \\n # try...catch 错误参数提取\\n elif t == 'catch':\\n i += 1\\n if i < len(tokens) and tokens[i] == '(':\\n param_tokens = []\\n level = 1\\n param_tokens.append(tokens[i])\\n i += 1\\n while i < len(tokens) and level > 0:\\n curr = tokens[i]\\n if curr == '(': level += 1\\n elif curr == ')': level -= 1\\n param_tokens.append(curr)\\n i += 1\\n variables.update(extract_vars_from_tokens(param_tokens))\\n else:\\n i += 1\\n\\n # 如若代码中提取不到任何变量且没有能够解析的有效代码，默认为 True，因无违规项\\n if not variables and not has_target_files:\\n return True\\n\\n # 判断提炼得到的变量名是否皆满足长度约束 [2, 7]\\n for var in variables:\\n if len(var) < 2 or len(var) > 7:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 如果工作区路径不存在，返回 False\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n class_names = set()\\n\\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 只处理指定的代码文件类型\\n if ext not in ['.html', '.htm', '.css', '.js']:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 1. 提取 JavaScript 的 Class 声明 (如: class my_class)\\n if ext in ['.js', '.html', '.htm']:\\n js_classes = re.findall(r'\\\\bclass\\\\s+([A-Za-z0-9_]+)\\\\b', content)\\n class_names.update(js_classes)\\n \\n # 2. 提取 HTML 中 class 属性配置 (如: class=\\\"my_class1 my_class2\\\")\\n if ext in ['.html', '.htm', '.js']:\\n # 匹配 class 或 className 属性里的字符串\\n html_classes = re.findall(r'\\\\b(?:class|className)\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']', content, flags=re.IGNORECASE)\\n for cls_str in html_classes:\\n # 分割可能存在多个类名用空格隔开的情况\\n class_names.update(cls_str.split())\\n \\n # 3. 提取 CSS 文件及 HTML <style> 标签中的类选择器 (如: .my_class)\\n if ext == '.css':\\n # 匹配点号开头，首字符为字母或下划线的 CSS 选择器\\n css_classes = re.findall(r'\\\\.([A-Za-z_][A-Za-z0-9_-]*)', content)\\n class_names.update(css_classes)\\n elif ext in ['.html', '.htm']:\\n # 仅在 <style> 标签范围内提取 CSS 类选择器\\n style_contents = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.IGNORECASE | re.DOTALL)\\n for style_content in style_contents:\\n css_classes = re.findall(r'\\\\.([A-Za-z_][A-Za-z0-9_-]*)', style_content)\\n class_names.update(css_classes)\\n \\n # 验证提取到的每一个类名是否仅由小写字母和下划线构成\\n for name in class_names:\\n # 正则 ^[a-z_]+$ 确保只能由 a-z 的小写字母及下划线 _ 构成\\n if not re.match(r'^[a-z_]+$', name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_function_names(response: str, workspace_path: str) -> bool:\\n found_any_function = False\\n \\n # 匹配普通函数声明，例如 function aba() 或 async function aba()\\n pattern_func_decl = re.compile(r'(?:async\\\\s+)?function\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n \\n # 匹配函数表达式赋值，例如 const aba = function() 或 let aba = async function()\\n pattern_func_expr = re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?function\\\\s*\\\\(')\\n \\n # 匹配箭头函数赋值，例如 const aba = () => 或者 let aba = arg =>\\n pattern_arrow_func = re.compile(r'(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>')\\n \\n # 提取 HTML 文件中的 script 标签内容\\n pattern_script = re.compile(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', re.IGNORECASE)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅检查 .js 和 .html 文件\\n if file.endswith('.js') or file.endswith('.html'):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n js_contents = []\\n if file.endswith('.html'):\\n scripts = pattern_script.findall(content)\\n js_contents.extend(scripts)\\n else:\\n js_contents.append(content)\\n \\n for js_code in js_contents:\\n # 预处理：去掉注释和字符串，避免在无关内容中产生误匹配\\n # 1. 去除单行注释\\n js_code = re.sub(r'//.*', '', js_code)\\n # 2. 去除多行注释\\n js_code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', js_code)\\n # 3. 去除单引号、双引号、反引号字符串\\n js_code = re.sub(r\\\"'([^'\\\\\\\\]|\\\\\\\\.)*'\\\", \\\"''\\\", js_code)\\n js_code = re.sub(r'\\\"([^\\\"\\\\\\\\]|\\\\\\\\.)*\\\"', '\\\"\\\"', js_code)\\n js_code = re.sub(r'`([^`\\\\\\\\]|\\\\\\\\.)*`', '``', js_code)\\n\\n names = []\\n # 提取所有命名函数名称\\n names.extend(pattern_func_decl.findall(js_code))\\n names.extend(pattern_func_expr.findall(js_code))\\n names.extend(pattern_arrow_func.findall(js_code))\\n \\n for name in names:\\n found_any_function = True\\n # 检查函数名是否为严格的回文字符串 (区分大小写)\\n if name != name[::-1]:\\n return False\\n\\n # 若所有提取到的函数名均为回文且至少存在一个函数，返回 True\\n return found_any_function\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 3 行注释\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import bisect\\n\\n def get_comment_lines_in_code(code_content, offset, full_content, is_css=False):\\n comment_lines = set()\\n i = 0\\n in_string = False\\n string_char = ''\\n in_single_comment = False\\n in_multi_comment = False\\n \\n # 预先计算该代码块在整个文件中的初始行号\\n current_line = full_content[:offset].count('\\\\n') + 1\\n string_quote_chars = ['\\\"', \\\"'\\\"] if is_css else ['\\\"', \\\"'\\\", '`']\\n \\n while i < len(code_content):\\n c = code_content[i]\\n \\n if c == '\\\\n':\\n if in_single_comment:\\n in_single_comment = False\\n current_line += 1\\n if in_multi_comment:\\n comment_lines.add(current_line)\\n i += 1\\n continue\\n \\n if in_single_comment:\\n comment_lines.add(current_line)\\n i += 1\\n continue\\n \\n if in_multi_comment:\\n comment_lines.add(current_line)\\n if c == '*' and i + 1 < len(code_content) and code_content[i+1] == '/':\\n in_multi_comment = False\\n i += 2\\n else:\\n i += 1\\n continue\\n \\n if in_string:\\n if c == '\\\\\\\\':\\n i += 2\\n continue\\n if c == string_char:\\n in_string = False\\n i += 1\\n continue\\n \\n if c in string_quote_chars:\\n in_string = True\\n string_char = c\\n i += 1\\n continue\\n \\n if c == '/' and i + 1 < len(code_content):\\n if not is_css and code_content[i+1] == '/':\\n in_single_comment = True\\n comment_lines.add(current_line)\\n i += 2\\n continue\\n elif code_content[i+1] == '*':\\n in_multi_comment = True\\n comment_lines.add(current_line)\\n i += 2\\n continue\\n \\n i += 1\\n \\n return comment_lines\\n\\n def count_comment_lines_in_file(file_path):\\n ext = os.path.splitext(file_path)[1].lower()\\n \\n # 仅限定关注用户可能生成的相关代码文件，排查各类常规文本文档等\\n code_extensions = {'.html', '.htm', '.css', '.js'}\\n if ext not in code_extensions:\\n return None\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n return -1\\n\\n comment_lines = set()\\n\\n if ext in ['.js']:\\n comment_lines = get_comment_lines_in_code(content, 0, content, is_css=False)\\n elif ext in ['.css']:\\n comment_lines = get_comment_lines_in_code(content, 0, content, is_css=True)\\n elif ext in ['.html', '.htm']:\\n line_indices = [0]\\n for idx, char in enumerate(content):\\n if char == '\\\\n':\\n line_indices.append(idx + 1)\\n \\n def get_line_num(index):\\n return bisect.bisect_right(line_indices, index)\\n\\n skip_intervals = []\\n\\n # 1. 查找并解析 <script> 内的 JS 注释\\n for script_match in re.finditer(r'<script[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE):\\n skip_intervals.append((script_match.start(), script_match.end()))\\n script_start = script_match.start(1)\\n script_content = script_match.group(1)\\n lines = get_comment_lines_in_code(script_content, script_start, content, is_css=False)\\n comment_lines.update(lines)\\n\\n # 2. 查找并解析 <style> 内的 CSS 注释\\n for style_match in re.finditer(r'<style[^>]*>(.*?)</style>', content, re.DOTALL | re.IGNORECASE):\\n skip_intervals.append((style_match.start(), style_match.end()))\\n style_start = style_match.start(1)\\n style_content = style_match.group(1)\\n lines = get_comment_lines_in_code(style_content, style_start, content, is_css=True)\\n comment_lines.update(lines)\\n\\n def is_in_skip(idx):\\n for st, ed in skip_intervals:\\n if st <= idx < ed:\\n return True\\n return False\\n\\n # 3. 查找剩余结构里的 HTML 多行注释\\n for m in re.finditer(r'<!--(.*?)-->', content, re.DOTALL):\\n if is_in_skip(m.start()):\\n continue\\n start_line = get_line_num(m.start())\\n end_line = get_line_num(max(m.start(), m.end() - 1))\\n for l in range(start_line, end_line + 1):\\n comment_lines.add(l)\\n\\n return len(comment_lines)\\n\\n has_code_file = False\\n \\n # 防止传入不可用路径\\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n cnt = count_comment_lines_in_file(file_path)\\n \\n if cnt is not None:\\n has_code_file = True\\n if cnt != 3:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 常用且可能在科普实验项目中出现的代码类文件扩展名\\n valid_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.cpp', '.cs', '.php'}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 过滤隐藏文件\\n if file.startswith('.'):\\n continue\\n \\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext not in valid_exts:\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n classes = set()\\n \\n # 1. OOP 面向对象类名提取: 例如 class MyClass { ... }\\n oop_classes = re.findall(r'\\\\bclass\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\b', content)\\n classes.update(oop_classes)\\n \\n # 2. HTML/JSX 标签中的 class 属性提取\\n if ext in {'.html', '.htm', '.jsx', '.tsx', '.vue'}:\\n class_attrs = re.findall(r'\\\\bclass(?:Name)?\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']', content, re.IGNORECASE)\\n for attr in class_attrs:\\n classes.update(attr.split())\\n \\n # 3. CSS 类选择器提取: 例如 .my_class { ... }\\n if ext == '.css':\\n css_classes = re.findall(r'\\\\.([a-zA-Z_][a-zA-Z0-9_-]*)', content)\\n classes.update(css_classes)\\n elif ext in {'.html', '.htm', '.vue'}:\\n # 如果是内嵌结构，在 <style> 块中寻找\\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, re.IGNORECASE | re.DOTALL)\\n for style in styles:\\n css_classes = re.findall(r'\\\\.([a-zA-Z_][a-zA-Z0-9_-]*)', style)\\n classes.update(css_classes)\\n \\n # 4. JavaScript 常用的 DOM 类名操控 API 提取\\n if ext in {'.js', '.ts', '.jsx', '.tsx', '.html', '.htm', '.vue'}:\\n # classList API \\n js_classes = re.findall(r'classList\\\\.(?:add|remove|toggle|contains)\\\\s*\\\\(\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']', content)\\n for attrs in js_classes:\\n classes.update(attrs.split())\\n \\n # jQuery addClass, removeClass 等 API\\n jq_classes = re.findall(r'(?:add|remove|has|toggle)Class\\\\s*\\\\(\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']', content)\\n for attrs in jq_classes:\\n classes.update(attrs.split())\\n \\n # getElementsByClassName API\\n get_classes = re.findall(r'getElementsByClassName\\\\s*\\\\(\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']', content)\\n for attrs in get_classes:\\n classes.update(attrs.split())\\n \\n # 检查在该代码文件中提取到的所有类的长度\\n for c in classes:\\n c_clean = c.strip()\\n if c_clean:\\n # 如果任一类名长度不为偶数，则要求未被满足\\n if len(c_clean) % 2 != 0:\\n return False\\n\\n # 若未找到任何违规类，则代表完全满足要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据上下文，推断出生成的代码文件类型\\n code_extensions = ['.html', '.css', '.js']\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if any(file.endswith(ext) for ext in code_extensions):\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 统一换行符格式，确保按行划分准确无误\\n content = content.replace('\\\\r\\\\n', '\\\\n')\\n \\n # 按换行符切分以精确统计行数，该方法能够将所有的空行严格计算在内\\n lines = content.split('\\\\n')\\n \\n # 若当前代码文件总行数为偶数，直接判定不符合要求\\n if len(lines) % 2 == 0:\\n return False\\n except Exception:\\n # 如果遇到文件编码不符合 UTF-8 导致读取错误，等同于未满足条件\\n return False\\n \\n # 如果找到了至少一个代码文件，并且没有任何代码文件的行数为偶数，则满足要求\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"请在生成的所有代码文件中，使用正好 4 个空格来完成每一处缩进\",\n \"validation_code\": \"def check_indentation(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据用户指令，推断代码文件为 HTML、CSS 和原生 JavaScript\\n code_extensions = ['.html', '.css', '.js']\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅检查代码文件，忽略其它可能生成的文件如 progress.md 或 txt 等\\n if any(file.endswith(ext) for ext in code_extensions):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n for line in lines:\\n # 忽略空行或纯空白行\\n if not line.strip():\\n continue\\n \\n # 提取当前行的前导空白字符\\n leading_spaces = ''\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_spaces += char\\n else:\\n break\\n \\n # 如果顶格写（无缩进），则继续检查下一行\\n if not leading_spaces:\\n continue\\n \\n # 严格要求使用空格，严禁包含 \\\\t 制表符\\n if '\\\\t' in leading_spaces:\\n return False\\n \\n content_stripped = line.strip()\\n # 忽略由于多行注释中的星号 '*' 带来的不可避免的对齐排版（如 /** \\\\n * \\\\n */）\\n if content_stripped.startswith('*'):\\n continue\\n \\n # 验证缩进的空格数是否是 4 的倍数（即正好使用 4 个空格完成一处缩进）\\n if len(leading_spaces) % 4 != 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"很好，接下来增加物理引擎的复杂度，让实验更有趣。在左侧控制面板把之前占位的滑动条完善为：物体A重量、物体B重量、以及风阻系数。现在要修改下落动画的数学计算逻辑，引入空气阻力（阻力可以简单模拟为与速度成正比，或者根据牛顿阻力定律计算）。这里的核心挑战是：结合物体的重量和风阻系数进行加速度的实时演算。重量越轻、风阻系数越大的物体（比如羽毛），受空气阻力影响越明显，下落越慢。修改下拉菜单的联动逻辑：选“羽毛”时自动设置较轻的重量和较大的风阻，选“小球”时设置较重的重量和较小的风阻。落地后，在结果区分别显示A和B各自的具体耗时。\\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，生成的代码文件应全部使用2个空格来控制缩进，替换之前的4空格缩进要求；第二，进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"只要是你创建的代码文件，就必须使用UTF-8进行编码\",\n \"validation_code\": \"def check_code_files_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 如果工作区不存在，无法进行验证\\n if not os.path.exists(workspace_path):\\n return False\\n\\n # 根据指令推断可能生成的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 遍历工作区所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅验证代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦抛出解码异常，说明文件不是合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他的异常，如无读取权限等\\n pass\\n \\n # 如果所有代码文件都能成功以 UTF-8 解码，或未发现代码文件，判定满足要求\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7\",\n \"validation_code\": \"def check_workspace(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n import ast\\n\\n def extract_params(params_str):\\n \\\"\\\"\\\"辅助函数：解析 JS 函数参数列表中的变量名（支持部分解构语法）\\\"\\\"\\\"\\n variables = set()\\n # 按照 JS 常用标识符与常见符号分词\\n tokens = [m.group(0) for m in re.finditer(r'[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\\\[\\\\]();,]', params_str)]\\n i = 0\\n n = len(tokens)\\n in_assignment = False\\n bracket_level = 0\\n \\n while i < n:\\n t = tokens[i]\\n if t == ',':\\n if bracket_level == 0:\\n in_assignment = False\\n elif t in ['{', '[', '(']:\\n bracket_level += 1\\n elif t in ['}', ']', ')']:\\n bracket_level -= 1\\n elif t == '=':\\n if bracket_level == 0:\\n in_assignment = True\\n elif re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t):\\n if not in_assignment:\\n variables.add(t)\\n i += 1\\n return variables\\n\\n def extract_js_variables(code):\\n \\\"\\\"\\\"解析 JavaScript 代码中的声明变量\\\"\\\"\\\"\\n # 预处理：移除注释与字符串常量，避免误判\\n code = re.sub(r'//.*', '', code)\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n code = re.sub(r'\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\])*\\\"', '', code)\\n code = re.sub(r\\\"'(?:\\\\\\\\.|[^'\\\\\\\\])*'\\\", '', code)\\n code = re.sub(r'`[\\\\s\\\\S]*?`', '', code)\\n \\n tokens = [m.group(0) for m in re.finditer(r'[a-zA-Z_$][a-zA-Z0-9_$]*|[={}\\\\[\\\\]();,]', code)]\\n variables = set()\\n i = 0\\n n = len(tokens)\\n \\n # 状态机：扫描并提取 var/let/const/function/class 声明的变量\\n while i < n:\\n t = tokens[i]\\n if t in ['let', 'var', 'const']:\\n i += 1\\n bracket_level = 0\\n in_assignment = False\\n while i < n:\\n t2 = tokens[i]\\n if t2 == ';':\\n break\\n elif t2 in ['function', 'class']:\\n i -= 1\\n break\\n elif t2 == ',':\\n if bracket_level == 0:\\n in_assignment = False\\n elif t2 in ['{', '[', '(']:\\n bracket_level += 1\\n elif t2 in ['}', ']', ')']:\\n bracket_level -= 1\\n elif t2 == '=':\\n if bracket_level == 0:\\n in_assignment = True\\n elif t2 in ['in', 'of']:\\n if bracket_level == 0:\\n in_assignment = True\\n elif re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', t2):\\n if not in_assignment:\\n variables.add(t2)\\n i += 1\\n elif t == 'function':\\n i += 1\\n if i < n and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]) and tokens[i] not in ['(', '{']:\\n variables.add(tokens[i])\\n elif t == 'class':\\n i += 1\\n if i < n and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]) and tokens[i] != '{':\\n variables.add(tokens[i])\\n elif t == 'catch':\\n i += 1\\n while i < n and tokens[i] != '(':\\n if tokens[i] in ['{', ';']: break\\n i += 1\\n if i < n and tokens[i] == '(':\\n i += 1\\n if i < n and re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', tokens[i]):\\n variables.add(tokens[i])\\n else:\\n i += 1\\n \\n # 通过正则额外兜底提取函数参数、箭头函数及 import 变量\\n for m in re.finditer(r'\\\\bfunction(?:\\\\s+[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', code):\\n variables.update(extract_params(m.group(1)))\\n \\n for m in re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', code):\\n variables.add(m.group(1))\\n for m in re.finditer(r'\\\\(([^)]+)\\\\)\\\\s*=>', code):\\n variables.update(extract_params(m.group(1)))\\n\\n for m in re.finditer(r'\\\\bimport\\\\s+([^{;\\\"\\\\'\\\\n]+)\\\\s+from', code):\\n for v in re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', m.group(1)):\\n if v not in ['*', 'as']:\\n variables.add(v)\\n for m in re.finditer(r'\\\\bimport\\\\s*\\\\{([^}]+)\\\\}', code):\\n for v in re.findall(r'\\\\b[a-zA-Z_$][a-zA-Z0-9_$]*\\\\b', m.group(1)):\\n if v != 'as':\\n variables.add(v)\\n\\n js_reserved = {\\n 'break', 'case', 'catch', 'class', 'const', 'continue', 'debugger', 'default', \\n 'delete', 'do', 'else', 'export', 'extends', 'finally', 'for', 'function', 'if', \\n 'import', 'in', 'instanceof', 'new', 'return', 'super', 'switch', 'this', 'throw', \\n 'try', 'typeof', 'var', 'void', 'while', 'with', 'yield', 'let', 'static', 'enum', \\n 'await', 'implements', 'package', 'protected', 'interface', 'private', 'public',\\n 'null', 'true', 'false', 'undefined', 'NaN', 'Infinity'\\n }\\n return {v for v in variables if v not in js_reserved}\\n\\n def extract_css_variables(code):\\n \\\"\\\"\\\"解析 CSS 中的自定义变量\\\"\\\"\\\"\\n variables = set()\\n for m in re.finditer(r'--([a-zA-Z0-9_-]+)', code):\\n variables.add('--' + m.group(1))\\n return variables\\n\\n def extract_python_variables(code):\\n \\\"\\\"\\\"解析 Python 中的变量名\\\"\\\"\\\"\\n variables = set()\\n try:\\n tree = ast.parse(code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n variables.add(node.id)\\n elif isinstance(node, ast.arg):\\n variables.add(node.arg)\\n elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):\\n variables.add(node.name)\\n except Exception:\\n pass\\n return variables\\n\\n all_vars = set()\\n\\n # 遍历项目文件树\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 根据后缀采用不同的解析逻辑\\n if ext == '.js':\\n all_vars.update(extract_js_variables(content))\\n elif ext == '.py':\\n all_vars.update(extract_python_variables(content))\\n elif ext == '.css':\\n all_vars.update(extract_css_variables(content))\\n elif ext in ['.html', '.htm']:\\n # 解析 HTML 中包含的 JS 和 CSS 内容\\n js_matches = re.finditer(r'<script.*?>([\\\\s\\\\S]*?)</script>', content, re.IGNORECASE)\\n for m in js_matches:\\n all_vars.update(extract_js_variables(m.group(1)))\\n css_matches = re.finditer(r'<style.*?>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n for m in css_matches:\\n all_vars.update(extract_css_variables(m.group(1)))\\n \\n # 验证要求：所有提取出的变量，其字符长度必须在 [2, 7] 的区间之内\\n for var in all_vars:\\n if len(var) < 2 or len(var) > 7:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线\",\n \"validation_code\": \"def check_class_names_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n class_names = set()\\n \\n # 提取多种场景下的“类名”\\n js_class_pattern = re.compile(r'\\\\bclass\\\\s+([A-Za-z0-9_$]+)')\\n html_class_pattern = re.compile(r'\\\\bclass(?:Name)?\\\\s*=\\\\s*([\\\"\\\\'])([^\\\"\\\\']*)\\\\1')\\n js_classlist_pattern = re.compile(r'\\\\.classList\\\\.(?:add|remove|toggle|contains)\\\\s*\\\\(\\\\s*([\\\"\\\\'])([^\\\"\\\\']*)\\\\1\\\\s*\\\\)')\\n js_getelements_pattern = re.compile(r'\\\\.getElementsByClassName\\\\s*\\\\(\\\\s*([\\\"\\\\'])([^\\\"\\\\']*)\\\\1\\\\s*\\\\)')\\n css_class_pattern = re.compile(r'\\\\.([a-zA-Z_][a-zA-Z0-9_-]*)')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 过滤只排查前端常见的文本文件\\n if not file.endswith(('.html', '.js', '.css', '.jsx', '.ts', '.tsx')):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 预处理：简单去除注释内容以防止干扰匹配\\n content = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', content)\\n content = re.sub(r'<!--[\\\\s\\\\S]*?-->', '', content)\\n content = re.sub(r'//.*', '', content)\\n \\n # 1. 提取 JS/ES6 面向对象定义的类\\n for match in js_class_pattern.finditer(content):\\n class_names.add(match.group(1).strip())\\n \\n # 2. 提取 HTML 中的 class 属性及 JS 中的 className 赋值\\n for match in html_class_pattern.finditer(content):\\n classes_str = match.group(2)\\n for cls in classes_str.split():\\n if cls.strip():\\n class_names.add(cls.strip())\\n \\n # 3. 提取 JS 中通过 classList 操作关联的类名\\n for match in js_classlist_pattern.finditer(content):\\n classes_str = match.group(2)\\n for cls in classes_str.split():\\n if cls.strip():\\n class_names.add(cls.strip())\\n \\n # 4. 提取 JS 中通过 getElementsByClassName 查询关联的类名\\n for match in js_getelements_pattern.finditer(content):\\n classes_str = match.group(2)\\n for cls in classes_str.split():\\n if cls.strip():\\n class_names.add(cls.strip())\\n \\n # 5. 提取 CSS 中的类选择器（综合处理 .css 文件和 HTML <style> 标签的内容）\\n style_content = []\\n if file.endswith('.css'):\\n style_content.append(content)\\n elif file.endswith('.html'):\\n styles = re.findall(r'<style[^>]*>([\\\\s\\\\S]*?)</style>', content, re.IGNORECASE)\\n style_content.extend(styles)\\n \\n for css_text in style_content:\\n # 在样式表中先行移除 url() 和 字符串 的内容，避免将诸如 .png 等后缀误判为类名\\n css_text = re.sub(r'url\\\\([^)]*\\\\)', '', css_text)\\n css_text = re.sub(r'([\\\"\\\\'])[\\\\s\\\\S]*?\\\\1', '', css_text)\\n \\n for cls in css_class_pattern.findall(css_text):\\n class_names.add(cls.strip())\\n \\n except Exception:\\n # 忽略文件读取和解析的偶发异常\\n pass\\n \\n # 若在项目中未提取到任何类名，亦无违规发生，即代表符合要求\\n if not class_names:\\n return True\\n \\n # 逐一验证所有提取到的类名\\n for cls in class_names:\\n # 类名的字符构成必须且只能是小写字母以及下划线\\n if not re.match(r'^[a-z_]+$', cls):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_function_names(response: str, workspace_path: str) -> bool:\\n if not os.path.exists(workspace_path):\\n return False\\n \\n func_names = set()\\n \\n # 匹配常规的 JavaScript/TypeScript 函数定义: function name(...) \\n pattern_func = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\(')\\n \\n # 匹配赋值形式的函数: const/let/var name = (async) function(...) 或 (async) (...) => \\n pattern_assign = re.compile(\\n r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=\\\\s*'\\n r'(?:(?:async\\\\s+)?function\\\\b|(?:async\\\\s+)?(?:\\\\([^)]*\\\\)|[a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>)'\\n )\\n \\n # 匹配对象或类中的方法简写: name(...) {\\n pattern_method = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n \\n # 匹配 Python 常规函数定义\\n pattern_py = re.compile(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n \\n # 需要排除的关键字，防止方法简写的正则对控制流进行误捕获\\n js_keywords = {\\n 'if', 'else', 'for', 'while', 'do', 'switch', 'catch', 'finally', \\n 'function', 'return', 'with', 'constructor', 'super'\\n }\\n \\n # 遍历当前项目的所有常见可能包含代码的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n if file.endswith(('.js', '.html', '.htm', '.ts', '.jsx', '.tsx', '.py')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 1. 查找常规函数声明\\n for match in pattern_func.findall(content):\\n func_names.add(match)\\n \\n # 2. 查找函数赋值\\n for match in pattern_assign.findall(content):\\n func_names.add(match)\\n \\n # 3. 查找方法定义，并剔除控制流等关键字\\n for match in pattern_method.findall(content):\\n if match not in js_keywords:\\n func_names.add(match)\\n \\n # 4. 查找 Python 函数定义\\n for match in pattern_py.findall(content):\\n func_names.add(match)\\n except Exception:\\n pass\\n \\n # 检查所有提取出的函数名称是否为严格的回文字符串\\n for name in func_names:\\n if name != name[::-1]:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 3 行注释\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n def parse_js(code, start_offset_line=0):\\n # 匹配 JS 字符串和注释\\n pattern = re.compile(\\n r'(?P<string_double>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\")|'\\n r\\\"(?P<string_single>'(?:\\\\\\\\.|[^'\\\\\\\\\\\\n])*')|\\\"\\n r'(?P<string_template>`(?:\\\\\\\\.|[^`\\\\\\\\])*`)|'\\n r'(?P<comment_line>//[^\\\\n]*)|'\\n r'(?P<comment_block>/\\\\*.*?\\\\*/)',\\n re.DOTALL\\n )\\n lines = set()\\n for m in pattern.finditer(code):\\n if m.lastgroup in ['comment_line', 'comment_block']:\\n start_line = code.count('\\\\n', 0, m.start()) + 1 + start_offset_line\\n end_line = code.count('\\\\n', 0, m.end()) + 1 + start_offset_line\\n for l in range(start_line, end_line + 1):\\n lines.add(l)\\n return lines\\n\\n def parse_css(code, start_offset_line=0):\\n # 匹配 CSS 字符串和注释\\n pattern = re.compile(\\n r'(?P<string_double>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\")|'\\n r\\\"(?P<string_single>'(?:\\\\\\\\.|[^'\\\\\\\\\\\\n])*')|\\\"\\n r'(?P<comment_block>/\\\\*.*?\\\\*/)',\\n re.DOTALL\\n )\\n lines = set()\\n for m in pattern.finditer(code):\\n if m.lastgroup == 'comment_block':\\n start_line = code.count('\\\\n', 0, m.start()) + 1 + start_offset_line\\n end_line = code.count('\\\\n', 0, m.end()) + 1 + start_offset_line\\n for l in range(start_line, end_line + 1):\\n lines.add(l)\\n return lines\\n\\n def parse_html(code):\\n # 匹配 HTML 注释及内部嵌套的 JS 和 CSS\\n pattern = re.compile(\\n r'(?P<comment_html><!--.*?-->)|'\\n r'(?P<script><script\\\\b[^>]*>.*?</script>)|'\\n r'(?P<style><style\\\\b[^>]*>.*?</style>)|'\\n r'(?P<string_double>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\")|'\\n r\\\"(?P<string_single>'(?:\\\\\\\\.|[^'\\\\\\\\\\\\n])*')\\\",\\n re.IGNORECASE | re.DOTALL\\n )\\n lines = set()\\n for m in pattern.finditer(code):\\n if m.lastgroup == 'script':\\n script_code = m.group('script')\\n start_offset = code.count('\\\\n', 0, m.start())\\n lines.update(parse_js(script_code, start_offset))\\n elif m.lastgroup == 'style':\\n style_code = m.group('style')\\n start_offset = code.count('\\\\n', 0, m.start())\\n lines.update(parse_css(style_code, start_offset))\\n elif m.lastgroup == 'comment_html':\\n start_line = code.count('\\\\n', 0, m.start()) + 1\\n end_line = code.count('\\\\n', 0, m.end()) + 1\\n for l in range(start_line, end_line + 1):\\n lines.add(l)\\n return lines\\n\\n def parse_py(code):\\n # 匹配 Python 字符串和注释（防御性支持）\\n pattern = re.compile(\\n r'(?P<string_double_3>\\\"\\\"\\\"(?:\\\\\\\\.|.)*?\\\"\\\"\\\")|'\\n r\\\"(?P<string_single_3>'''(?:\\\\\\\\.|.)*?''')|\\\"\\n r'(?P<string_double>\\\"(?:\\\\\\\\.|[^\\\"\\\\\\\\\\\\n])*\\\")|'\\n r\\\"(?P<string_single>'(?:\\\\\\\\.|[^'\\\\\\\\\\\\n])*')|\\\"\\n r'(?P<comment_hash>#[^\\\\n]*)',\\n re.DOTALL\\n )\\n lines = set()\\n for m in pattern.finditer(code):\\n if m.lastgroup == 'comment_hash':\\n start_line = code.count('\\\\n', 0, m.start()) + 1\\n lines.add(start_line)\\n return lines\\n\\n # 定义需要检查的代码文件扩展名列表\\n CODE_EXTS = {\\n '.html', '.htm', '.js', '.css', '.ts', '.jsx', '.tsx',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.go', '.rs', '.rb', '.php', '.sh'\\n }\\n \\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in CODE_EXTS:\\n has_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n code = f.read().replace('\\\\r\\\\n', '\\\\n').replace('\\\\r', '\\\\n')\\n except Exception:\\n continue\\n \\n if ext in ['.html', '.htm']:\\n comment_lines = parse_html(code)\\n elif ext in ['.js', '.ts', '.jsx', '.tsx', '.c', '.cpp', '.h', '.hpp', '.java', '.go', '.rs', '.php']:\\n comment_lines = parse_js(code)\\n elif ext in ['.css', '.scss', '.less']:\\n comment_lines = parse_css(code)\\n elif ext in ['.py', '.rb', '.sh']:\\n comment_lines = parse_py(code)\\n else:\\n comment_lines = parse_js(code)\\n \\n # 如果某个代码文件的注释总行数不满足刚好3行\\n if len(comment_lines) != 3:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\\n # 匹配 OOP 类声明，例如：class my_class {\\n oop_class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\b')\\n # 匹配 HTML 中的 class 属性，例如：class=\\\"my_class another_class\\\"\\n html_class_pattern = re.compile(r'\\\\bclass\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']')\\n # 匹配 JS 中的 className 属性赋值，例如：element.className = \\\"my_class\\\"\\n js_classname_pattern = re.compile(r'\\\\bclassName\\\\s*=\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']')\\n # 匹配 JS 中的 classList 操作，例如：element.classList.add(\\\"my_class\\\")\\n js_classlist_pattern = re.compile(r'\\\\bclassList\\\\.(?:add|remove|toggle|contains)\\\\(\\\\s*[\\\"\\\\']([^\\\"\\\\']+)[\\\"\\\\']')\\n\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n \\n if not os.path.isfile(file_path):\\n continue\\n \\n try:\\n # 按照系统约束，生成的文件均使用 UTF-8 进行编码\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略不可读的文件\\n continue\\n \\n # 1. 检查面向对象（OOP）类声明\\n for match in oop_class_pattern.finditer(content):\\n class_name = match.group(1)\\n if len(class_name) % 2 != 0:\\n return False\\n \\n # 2. 检查 HTML 标签中的 class 属性\\n for match in html_class_pattern.finditer(content):\\n classes = match.group(1).split()\\n for c in classes:\\n if re.match(r'^[a-zA-Z_][a-zA-Z0-9_]*$', c):\\n if len(c) % 2 != 0:\\n return False\\n \\n # 3. 检查 JavaScript 代码中的 className 赋值\\n for match in js_classname_pattern.finditer(content):\\n classes = match.group(1).split()\\n for c in classes:\\n if re.match(r'^[a-zA-Z_][a-zA-Z0-9_]*$', c):\\n if len(c) % 2 != 0:\\n return False\\n\\n # 4. 检查 JavaScript 代码中的 classList 方法调用\\n for match in js_classlist_pattern.finditer(content):\\n classes = match.group(1).split()\\n for c in classes:\\n if re.match(r'^[a-zA-Z_][a-zA-Z0-9_]*$', c):\\n if len(c) % 2 != 0:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n code_extensions = {'.html', '.css', '.js'}\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 按换行符进行拆分，以准确计入所有的行，包括代码中间及末尾的空行\\n line_count = len(content.split('\\\\n'))\\n \\n # 判断如果包含偶数行，说明不满足要求，直接返回False\\n if line_count % 2 == 0:\\n return False\\n except Exception:\\n # 如果发生文件读取异常，则认为无法验证通过\\n return False\\n \\n # 如果没有找到任何代码文件，说明也没有完成要求\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用2个空格来控制缩进\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对前端项目的常见代码文件进行推断，过滤掉如.md, .txt之类的文本文件\\n code_extensions = {\\\".html\\\", \\\".css\\\", \\\".js\\\", \\\".ts\\\", \\\".jsx\\\", \\\".tsx\\\", \\\".vue\\\"}\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n continue\\n \\n positive_indents = set()\\n \\n for line in lines:\\n # 忽略空行\\n if not line.strip():\\n continue\\n \\n # 匹配每行的前导空白字符\\n match = re.match(r'^([ \\\\t]+)', line)\\n if match:\\n spaces = match.group(1)\\n \\n # 检查1：不得使用制表符控制缩进\\n if '\\\\t' in spaces:\\n return False\\n \\n length = len(spaces)\\n if length > 0:\\n positive_indents.add(length)\\n \\n if positive_indents:\\n # 判断是否存在不能被4整除的偶数缩进（如 2, 6, 10...），这是2空格缩进的典型特征\\n has_two_space_levels = any(i % 2 == 0 and i % 4 != 0 for i in positive_indents)\\n # 判断是否存在4的倍数缩进（如 4, 8, 12...）\\n has_four_space_levels = any(i % 4 == 0 for i in positive_indents)\\n \\n # 检查2：如果所有偶数层级的缩进全都是4的倍数，说明它仍然是4空格缩进的体系，不符合2空格的要求\\n if has_four_space_levels and not has_two_space_levels:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"越来越有意思了。现在来完善环境参数设置。在控制面板再增加一个“是否真空”的复选框，以及一个“重力加速度”的滑动条（范围设为1到20，让用户可以模拟月球或木星的重力）。当用户勾选“真空”时，底层的物理运算必须完全屏蔽风阻系数和空气阻力计算，此时无论两个物体的重量和类型如何，它们都必须绝对同时落地。而调整“重力加速度”滑动条应该实时改变当前的 $g$ 值，从而影响下落的整体快慢。请确保在调整这些环境参数后，系统状态能正确响应并反映在下一次的实验动画里。\\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，生成的代码文件应全部使用4个空格来控制缩进，替换之前的2空格缩进要求；第二，确保在代码开发完成之后，测试所有的代码并保证它们均通过测试。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"只要是你创建的代码文件，就必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作区路径是否存在\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n # 定义通常被视作代码或项目文本配置的文件扩展名\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.php', '.rb', '.go', '.rs', '.swift', '.kt', \\n '.sh', '.bat', '.json', '.xml', '.yaml', '.yml', \\n '.md', '.txt'\\n }\\n\\n found_any_code_file = False\\n\\n # 遍历项目路径下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_any_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他如权限受限等异常情况\\n continue\\n\\n # 只有当项目中确实生成了代码文件，且全部都能成功以 UTF-8 解析时才返回 True\\n return found_any_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7\",\n \"validation_code\": \"import os\\nimport re\\nimport ast\\n\\ndef check_variables_length(response: str, workspace_path: str) -> bool:\\n variables = set()\\n\\n def parse_js(js_code: str):\\n # 移除单行与多行注释\\n js_code = re.sub(r'//.*', '', js_code)\\n js_code = re.sub(r'/\\\\*.*?\\\\*/', '', js_code, flags=re.DOTALL)\\n # 移除字符串字面量，防止误匹配内部的伪代码或参数\\n js_code = re.sub(r'([\\\"\\\\'`])(?:(?=(\\\\\\\\?))\\\\2.)*?\\\\1', '', js_code, flags=re.DOTALL)\\n\\n # 1. 基本的 var / let / const 声明\\n matches = re.finditer(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', js_code)\\n for m in matches:\\n variables.add(m.group(1))\\n \\n # 2. 对象解构赋值 {}\\n destruct_obj = re.finditer(r'\\\\b(?:var|let|const)\\\\s*\\\\{([^}]+)\\\\}', js_code)\\n for m in destruct_obj:\\n for p in m.group(1).split(','):\\n m_id = re.match(r'^\\\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*\\\\s*:\\\\s*)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', p)\\n if m_id:\\n variables.add(m_id.group(1))\\n \\n # 3. 数组解构赋值 []\\n destruct_arr = re.finditer(r'\\\\b(?:var|let|const)\\\\s*\\\\[([^\\\\]]+)\\\\]', js_code)\\n for m in destruct_arr:\\n for p in m.group(1).split(','):\\n m_id = re.match(r'^\\\\s*(?:\\\\.\\\\.\\\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', p)\\n if m_id:\\n variables.add(m_id.group(1))\\n\\n # 4. 普通函数参数\\n func_params = re.finditer(r'\\\\bfunction\\\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)', js_code)\\n for m in func_params:\\n for p in m.group(1).split(','):\\n m_id = re.match(r'^\\\\s*(?:\\\\.\\\\.\\\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', p)\\n if m_id:\\n variables.add(m_id.group(1))\\n \\n # 5. 箭头函数参数 (单参数无需括号形式)\\n arrow_single = re.finditer(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>', js_code)\\n for m in arrow_single:\\n var_name = m.group(1)\\n # 排除由于代码格式导致的关键字误认\\n if var_name not in ['return', 'if', 'else', 'switch', 'for', 'while']:\\n variables.add(var_name)\\n \\n # 6. 箭头函数参数 (带括号形式)\\n arrow_multi = re.finditer(r'\\\\(([^)]*)\\\\)\\\\s*=>', js_code)\\n for m in arrow_multi:\\n for p in m.group(1).split(','):\\n m_id = re.match(r'^\\\\s*(?:\\\\.\\\\.\\\\.)?([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b', p)\\n if m_id:\\n variables.add(m_id.group(1))\\n \\n # 7. try...catch 错误对象\\n catch_blocks = re.finditer(r'\\\\bcatch\\\\s*\\\\(\\\\s*([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*\\\\)', js_code)\\n for m in catch_blocks:\\n variables.add(m.group(1))\\n\\n def parse_css(css_code: str):\\n # 匹配 CSS 变量 (如 --main-color)\\n css_vars = re.finditer(r'(--[a-zA-Z0-9_-]+)\\\\s*:', css_code)\\n for m in css_vars:\\n variables.add(m.group(1))\\n\\n def parse_py(py_code: str):\\n # 严谨的 Python AST 语法树解析\\n try:\\n tree = ast.parse(py_code)\\n for node in ast.walk(tree):\\n if isinstance(node, ast.Name) and isinstance(node.ctx, ast.Store):\\n variables.add(node.id)\\n elif isinstance(node, ast.arg):\\n variables.add(node.arg)\\n except SyntaxError:\\n pass\\n\\n def parse_html(html_code: str):\\n # 解析 HTML 内嵌的 JS 和 CSS\\n scripts = re.finditer(r'<script\\\\b[^>]*>(.*?)</script>', html_code, flags=re.IGNORECASE | re.DOTALL)\\n for m in scripts:\\n parse_js(m.group(1))\\n styles = re.finditer(r'<style\\\\b[^>]*>(.*?)</style>', html_code, flags=re.IGNORECASE | re.DOTALL)\\n for m in styles:\\n parse_css(m.group(1))\\n\\n # 1. 遍历当前工作空间内的项目生成文件\\n if os.path.exists(workspace_path):\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if file.endswith('.js'):\\n parse_js(code)\\n elif file.endswith('.css'):\\n parse_css(code)\\n elif file.endswith('.py'):\\n parse_py(code)\\n elif file.endswith('.html'):\\n parse_html(code)\\n except Exception:\\n continue\\n\\n # 2. 检查最后一轮 AI 回复的 Markdown 代码块\\n blocks = re.finditer(r'```([a-zA-Z]*)\\\\n(.*?)```', response, flags=re.DOTALL)\\n for m in blocks:\\n lang = m.group(1).lower()\\n code = m.group(2)\\n if lang in ['javascript', 'js']:\\n parse_js(code)\\n elif lang == 'css':\\n parse_css(code)\\n elif lang in ['python', 'py']:\\n parse_py(code)\\n elif lang == 'html' or lang == '':\\n parse_html(code)\\n\\n # 3. 校验提取到的全部变量名长度\\n for var in variables:\\n if len(var) < 2 or len(var) > 7:\\n return False\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 匹配 JavaScript 等语言中面向对象的类名（匹配 class 关键字后的合法标识符字符）\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z0-9_$]+)')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据用户指令推断，主要涉及前端开发文件\\n if not file.endswith(('.js', '.html', '.ts', '.jsx', '.tsx')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取当前文件中的所有类名\\n classes = class_pattern.findall(content)\\n for cls_name in classes:\\n # 检查该类名的字符构成是否仅为小写字母以及下划线\\n if not re.match(r'^[a-z_]+$', cls_name):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 针对 JS 和 HTML 的正则表达式模式，用于匹配各种常见函数的定义方式\\n js_func_patterns = [\\n # 1. 匹配标准函数声明：function foo() / async function foo() / function* foo()\\n re.compile(r'\\\\bfunction\\\\s*(?:\\\\*\\\\s*)?([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\('),\\n \\n # 2. 匹配变量赋值型函数：const foo = function() / let foo = () => {}\\n re.compile(r'\\\\b(?:const|let|var)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*=\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)'),\\n \\n # 3. 匹配对象属性赋值型函数：foo: function() / foo: async () => {}\\n re.compile(r'\\\\b([a-zA-Z_$][\\\\w$]*)\\\\s*:\\\\s*(?:async\\\\s+)?(?:function\\\\b|\\\\([^)]*\\\\)\\\\s*=>|[a-zA-Z_$][\\\\w$]*\\\\s*=>)'),\\n \\n # 4. 匹配类或对象的方法声明：foo() {} / async foo() {} / * foo() {} (严格排除常见关键字)\\n re.compile(r'^\\\\s*(?:async\\\\s+)?(?:\\\\*\\\\s*)?(?!if\\\\b|for\\\\b|while\\\\b|switch\\\\b|catch\\\\b|with\\\\b|function\\\\b|return\\\\b|constructor\\\\b|super\\\\b)([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', re.MULTILINE),\\n \\n # 5. 匹配 getter 和 setter：get foo() {} / set foo(val) {}\\n re.compile(r'^\\\\s*(?:get|set)\\\\s+([a-zA-Z_$][\\\\w$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{', re.MULTILINE)\\n ]\\n \\n # 针对 Python 的正则表达式模式\\n py_func_patterns = [\\n # 匹配标准函数和异步函数定义：def foo() / async def foo()\\n re.compile(r'\\\\b(?:async\\\\s+)?def\\\\s+([a-zA-Z_]\\\\w*)\\\\s*\\\\(')\\n ]\\n \\n def is_palindrome(s: str) -> bool:\\n # 判断字符串是否是严格的回文字符串\\n return s == s[::-1]\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n ext = os.path.splitext(file)[1].lower()\\n \\n if ext in ['.js', '.html']:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n for pattern in js_func_patterns:\\n for match in pattern.finditer(content):\\n func_name = match.group(1)\\n if not is_palindrome(func_name):\\n return False\\n except Exception:\\n continue\\n \\n elif ext == '.py':\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n for pattern in py_func_patterns:\\n for match in pattern.finditer(content):\\n func_name = match.group(1)\\n if not is_palindrome(func_name):\\n return False\\n except Exception:\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 3 行注释\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n def count_js_css_comments(text: str, start_line: int = 0, is_css: bool = False) -> set:\\n comment_lines = set()\\n in_multi = False\\n in_str = False\\n str_char = ''\\n lines = text.split('\\\\n')\\n \\n for i, line in enumerate(lines):\\n abs_line = start_line + i\\n # 如果处于多行注释状态，当前行必然包含注释\\n if in_multi:\\n comment_lines.add(abs_line)\\n \\n j = 0\\n while j < len(line):\\n # 正常非注释、非字符串状态下，寻找注释或字符串入口\\n if not in_multi and not in_str:\\n # 单行注释 (CSS 不支持单行注释，跳过此逻辑)\\n if not is_css and line[j:j+2] == '//':\\n comment_lines.add(abs_line)\\n break # 行末之前的全是注释\\n # 多行注释起点\\n elif line[j:j+2] == '/*':\\n in_multi = True\\n comment_lines.add(abs_line)\\n j += 1\\n # 字符串起点\\n elif line[j] in '\\\"\\\\'`':\\n in_str = True\\n str_char = line[j]\\n \\n # 字符串状态内，寻找转义符或结束引号\\n elif in_str:\\n if line[j] == '\\\\\\\\':\\n j += 1\\n elif line[j] == str_char:\\n in_str = False\\n \\n # 多行注释内，寻找结束点\\n elif in_multi:\\n comment_lines.add(abs_line)\\n if line[j:j+2] == '*/':\\n in_multi = False\\n j += 1\\n \\n j += 1\\n \\n return comment_lines\\n\\n def get_comment_lines(file_path: str) -> int:\\n ext = os.path.splitext(file_path)[1].lower()\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n\\n comment_lines = set()\\n \\n # HTML 综合解析\\n if ext in ['.html', '.htm']:\\n # 1. 解析 <script>\\n scripts = re.finditer(r'<script.*?>\\\\s*(.*?)\\\\s*</script>', code, re.IGNORECASE | re.DOTALL)\\n for m in scripts:\\n start_line = code[:m.start(1)].count('\\\\n')\\n lines = count_js_css_comments(m.group(1), start_line, is_css=False)\\n comment_lines.update(lines)\\n \\n # 2. 解析 <style>\\n styles = re.finditer(r'<style.*?>\\\\s*(.*?)\\\\s*</style>', code, re.IGNORECASE | re.DOTALL)\\n for m in styles:\\n start_line = code[:m.start(1)].count('\\\\n')\\n lines = count_js_css_comments(m.group(1), start_line, is_css=True)\\n comment_lines.update(lines)\\n \\n # 3. 解析纯 HTML 注释 (需排除被 script/style 包裹的伪注释)\\n html_comments = re.finditer(r'<!--(.*?)-->', code, re.DOTALL)\\n for m in html_comments:\\n start_idx = m.start()\\n inside_blocks = False\\n \\n # 简单排查该注释是否位于未闭合的 <script> 或 <style> 中\\n last_script = code.rfind('<script', 0, start_idx)\\n last_script_end = code.rfind('</script>', 0, start_idx)\\n if last_script > last_script_end: \\n inside_blocks = True\\n \\n last_style = code.rfind('<style', 0, start_idx)\\n last_style_end = code.rfind('</style>', 0, start_idx)\\n if last_style > last_style_end: \\n inside_blocks = True\\n \\n if not inside_blocks:\\n start_line = code[:m.start()].count('\\\\n')\\n num_lines = m.group(0).count('\\\\n')\\n for i in range(start_line, start_line + num_lines + 1):\\n comment_lines.add(i)\\n \\n # 纯 JS / TS 等变体解析\\n elif ext in ['.js', '.ts', '.jsx', '.tsx']:\\n comment_lines.update(count_js_css_comments(code, 0, is_css=False))\\n \\n # 纯 CSS 解析\\n elif ext == '.css':\\n comment_lines.update(count_js_css_comments(code, 0, is_css=True))\\n \\n return len(comment_lines)\\n\\n # 主验证流程\\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n num_comments = get_comment_lines(file_path)\\n except Exception:\\n # 读取或解析异常，可视为不满足条件\\n return False\\n \\n # 要求“精确包含 3 行注释”\\n if num_comments != 3:\\n return False\\n \\n # 如果找到了代码文件且均通过，则返回True；如果没有产生代码文件，则判定为False\\n return has_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 匹配面向对象语言中的类定义，例如 class MyClass\\n # 捕获组提取类名标识符\\n class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_]\\\\w*)')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 仅检查可能的代码文件类型，涵盖本项目用到的前端文件及常见代码文件\\n if not file.endswith(('.js', '.html', '.css', '.ts', '.jsx', '.tsx', '.py')):\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 忽略无法以UTF-8读取的文件\\n continue\\n \\n # 找到文件中定义的所有类名\\n classes = class_pattern.findall(content)\\n for cls_name in classes:\\n # 检查类名的长度是否严格满足偶数的要求\\n if len(cls_name) % 2 != 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_code_files_odd_lines(response: str, workspace_path: str) -> bool:\\n import os\\n\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据用户指令推断出的代码文件格式\\n code_extensions = {'.html', '.css', '.js'}\\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:\\n content = f.read()\\n \\n # 使用按照 '\\\\n' 切割的方式可以准确统计出所有物理行，包含文件末尾以及中间的空行\\n lines = content.split('\\\\n')\\n line_count = len(lines)\\n \\n # 判断行数是否为奇数\\n if line_count % 2 == 0:\\n return False\\n except Exception:\\n # 如果文件无法正常读取，视作无法验证要求，返回False\\n return False\\n \\n # 如果找到了至少一个代码文件，并且没有任何一个文件触发偶数行的 return False，则视为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"只要对项目进行了改动，就必须把本轮完成的内容同步更新到 claude-progress.txt 文件中\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用4个空格来控制缩进\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义属于代码文件的扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n has_code_file = False\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果无法读取，判定为不合规\\n return False\\n \\n for line in lines:\\n # 忽略完全空白或只包含换行符的行\\n if not line.strip():\\n continue\\n \\n # 提取行首的空白字符\\n leading_spaces = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_spaces += char\\n else:\\n break\\n \\n # 忽略块级注释的星号开头行（常用于 /* ... */ 的内部行）\\n if line.strip().startswith('*'):\\n continue\\n \\n # 判断一：如果缩进中使用了制表符，则违反了使用空格控制缩进的要求\\n if '\\\\t' in leading_spaces:\\n return False\\n \\n # 判断二：如果行首空白字符的长度不是4的倍数，则说明没有严格使用4空格缩进\\n if len(leading_spaces) % 4 != 0:\\n return False\\n\\n # 必须至少生成过一个代码文件才算任务完成\\n return has_code_file\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"最后一步，我们来提升一下视觉表现和数据统计功能。首先，根据下拉菜单的选择，用简单的CSS或者Emoji把代表物体的圆形替换掉，让“小球”和“羽毛”在视觉上直观区分开来。其次，在下落过程中，用Canvas或DOM动态生成一点简单的轨迹线或残影效果，方便用户直观对比它俩在半空中的高度差。最后，把底部的简单文字结果区改成一个“实验记录表”。每次实验双双落地结束后，把这一轮的“环境参数(重力/真空)、物体A参数、物体B参数、各自落地时间”作为新的一行追加到底部的表格中，方便用户对比历次实验的数据。\\n另外请注意，后续所有轮次必须始终遵循以下更新及新增要求：第一，在每次对项目进行改动后，更新process-log.md文件以记录本轮完成的内容是必须执行的操作，替换此前更新claude-progress.txt的要求；第二，你提供的每个代码文件里，注释内容须100%使用中文撰写。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"只要是你创建的代码文件，就必须使用UTF-8进行编码\",\n \"validation_code\": \"def verify_utf8_encoding(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 如果工作空间路径不存在，则无法验证是否满足条件，返回False\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据多轮指令，可能生成的代码文件和文档文件的拓展名\\n target_extensions = {'.html', '.css', '.js', '.md', '.txt'}\\n \\n # 遍历当前项目的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出推断的代码文件类型\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 UTF-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明并非有效的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能的文件访问异常（如权限问题等）\\n continue\\n \\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你输出的代码中，任何变量名的字符数都不能少于 2，且最多只能为 7\",\n \"validation_code\": \"def verify(response, workspace_path):\\n import os\\n import re\\n\\n def remove_comments_and_strings(code):\\n # 移除多行注释\\n code = re.sub(r'/\\\\*[\\\\s\\\\S]*?\\\\*/', '', code)\\n # 移除单行注释\\n code = re.sub(r'//.*', '', code)\\n # 移除双引号字符串 (简单考虑转义)\\n code = re.sub(r'\\\"[^\\\"\\\\\\\\\\\\r\\\\n]*(?:\\\\\\\\.[^\\\"\\\\\\\\\\\\r\\\\n]*)*\\\"', '\\\"\\\"', code)\\n # 移除单引号字符串\\n code = re.sub(r\\\"'[^'\\\\\\\\\\\\r\\\\n]*(?:\\\\\\\\.[^'\\\\\\\\\\\\r\\\\n]*)*'\\\", \\\"''\\\", code)\\n # 移除模板字符串 (反引号)\\n code = re.sub(r'`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`', '``', code)\\n return code\\n\\n def check_variable_names(code):\\n code = remove_comments_and_strings(code)\\n variables = set()\\n \\n # 1. 匹配普通的 var, let, const 声明的首个变量\\n decl_regex = re.compile(r'\\\\b(?:var|let|const)\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n for match in decl_regex.finditer(code):\\n variables.add(match.group(1))\\n \\n # 2. 匹配对象解构赋值，例如 let {a, b: myB} = obj\\n destruct_regex = re.compile(r'\\\\b(?:var|let|const)\\\\s+\\\\{([^}]+)\\\\}')\\n for match in destruct_regex.finditer(code):\\n vars_str = match.group(1)\\n for v in vars_str.split(','):\\n v = v.strip()\\n if not v:\\n continue\\n # 处理带有别名的情况（取冒号后的实际变量名）\\n if ':' in v:\\n v = v.split(':', 1)[1].strip()\\n if v.startswith('...'):\\n v = v[3:].strip()\\n v = v.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', v):\\n variables.add(v)\\n \\n # 3. 匹配数组解构赋值，例如 let [a, b] = arr\\n array_destruct_regex = re.compile(r'\\\\b(?:var|let|const)\\\\s+\\\\[([^\\\\]]+)\\\\]')\\n for match in array_destruct_regex.finditer(code):\\n vars_str = match.group(1)\\n for v in vars_str.split(','):\\n v = v.strip()\\n if not v:\\n continue\\n if v.startswith('...'):\\n v = v[3:].strip()\\n v = v.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', v):\\n variables.add(v)\\n\\n # 4. 函数名称\\n func_name_regex = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n for match in func_name_regex.finditer(code):\\n variables.add(match.group(1))\\n\\n # 5. 常规函数参数\\n func_params_regex = re.compile(r'\\\\bfunction\\\\s*(?:[a-zA-Z_$][a-zA-Z0-9_$]*)?\\\\s*\\\\(([^)]*)\\\\)')\\n for match in func_params_regex.finditer(code):\\n params_str = match.group(1)\\n for p in params_str.split(','):\\n p = p.strip()\\n if not p:\\n continue\\n if p.startswith('...'):\\n p = p[3:].strip()\\n var_name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n variables.add(var_name)\\n \\n # 6. 带括号的箭头函数参数\\n arrow_regex1 = re.compile(r'\\\\(([^)]*)\\\\)\\\\s*=>')\\n for match in arrow_regex1.finditer(code):\\n params_str = match.group(1)\\n for p in params_str.split(','):\\n p = p.strip()\\n if not p:\\n continue\\n if p.startswith('...'):\\n p = p[3:].strip()\\n var_name = p.split('=')[0].strip()\\n if re.match(r'^[a-zA-Z_$][a-zA-Z0-9_$]*$', var_name):\\n variables.add(var_name)\\n \\n # 7. 单参数箭头函数参数\\n arrow_regex2 = re.compile(r'\\\\b([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\s*=>')\\n for match in arrow_regex2.finditer(code):\\n variables.add(match.group(1))\\n \\n # 8. 类名称\\n class_regex = re.compile(r'\\\\bclass\\\\s+([a-zA-Z_$][a-zA-Z0-9_$]*)\\\\b')\\n for match in class_regex.finditer(code):\\n variables.add(match.group(1))\\n\\n # 遍历检验提取的所有变量名称的长度情况（使用原生的 len 函数统计字符数）\\n for var in variables:\\n var_len = len(var)\\n if var_len < 2 or var_len > 7:\\n return False\\n \\n return True\\n\\n # 遍历当前生成项目所有的 JavaScript 和 HTML 脚本区域\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n if file.endswith('.js'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n code = f.read()\\n if not check_variable_names(code):\\n return False\\n except Exception:\\n continue\\n elif file.endswith('.html'):\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n html_code = f.read()\\n # 忽略大小写查找 <script> 标签中的纯 JS 代码\\n scripts = re.findall(r'<script[^>]*>([\\\\s\\\\S]*?)</script>', html_code, re.IGNORECASE)\\n for script in scripts:\\n if not check_variable_names(script):\\n return False\\n except Exception:\\n continue\\n\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，各个类名的字符构成只能是小写字母以及下划线\",\n \"validation_code\": \"def verify(response, workspace_path):\\n import os\\n import re\\n\\n class_names = []\\n \\n # 1. 匹配 JS 的 ES6 class 声明\\n regex_es6 = re.compile(r'\\\\bclass\\\\s+([A-Za-z0-9_]+)')\\n \\n # 2. 匹配 HTML 或 JSX 中的 class / className 属性\\n regex_attr = re.compile(r'\\\\b(?:class|className)\\\\s*=\\\\s*[\\\\'\\\"]([^\\\\'\\\"]+)[\\\\'\\\"]')\\n \\n # 3. 匹配 classList DOM API 的操作\\n regex_classlist = re.compile(r'\\\\bclassList\\\\.(?:add|remove|toggle|contains)\\\\(\\\\s*[\\\\'\\\"]([^\\\\'\\\"]+)[\\\\'\\\"]\\\\s*\\\\)')\\n \\n # 4. 匹配 getElementsByClassName 的使用\\n regex_get_class = re.compile(r'\\\\bgetElementsByClassName\\\\(\\\\s*[\\\\'\\\"]([^\\\\'\\\"]+)[\\\\'\\\"]\\\\s*\\\\)')\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext not in ['.js', '.html', '.css', '.vue', '.jsx', '.tsx']:\\n continue\\n \\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n # 提取 JS 类定义\\n matches = regex_es6.findall(content)\\n class_names.extend(matches)\\n \\n # 提取属性分配中的类（需要按照空格切分，因为可能同时分配多个类名）\\n matches = regex_attr.findall(content)\\n for m in matches:\\n class_names.extend(m.split())\\n \\n # 提取 classList 中的类\\n matches = regex_classlist.findall(content)\\n for m in matches:\\n class_names.extend(m.split())\\n \\n # 提取 getElementsByClassName 中的类\\n matches = regex_get_class.findall(content)\\n for m in matches:\\n class_names.extend(m.split())\\n \\n # 提取 CSS 样式中的类选择器\\n css_text = ''\\n if ext == '.css':\\n css_text = content\\n elif ext in ['.html', '.vue']:\\n styles = re.findall(r'<style[^>]*>(.*?)</style>', content, flags=re.DOTALL | re.IGNORECASE)\\n css_text = '\\\\n'.join(styles)\\n \\n if css_text:\\n # 首先过滤掉CSS内部可能被误认为点类选择器的不相关内容\\n css_text = re.sub(r'/\\\\*.*?\\\\*/', '', css_text, flags=re.DOTALL) # 过滤多行注释\\n css_text = re.sub(r'url\\\\([^)]*\\\\)', '', css_text) # 过滤 url(...) 内容\\n css_text = re.sub(r'\\\"[^\\\"]*\\\"', '', css_text) # 过滤双引号字符串\\n css_text = re.sub(r\\\"'[^']*'\\\", '', css_text) # 过滤单引号字符串\\n \\n # 匹配剩余文本中的 CSS 类选择器 (.classname)\\n css_classes = re.findall(r'\\\\.([A-Za-z_][A-Za-z0-9_-]*)', css_text)\\n class_names.extend(css_classes)\\n\\n # 汇总并对所有类名进行规则检验\\n unique_class_names = set(class_names)\\n for cls in unique_class_names:\\n cls = cls.strip()\\n if not cls:\\n continue\\n \\n # 给定要求：字符构成只能是小写字母以及下划线\\n if not re.match(r'^[a-z_]+$', cls):\\n return False\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件，任何一个函数的名称都必须是严格的回文字符串\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n def get_functions(code_text):\\n functions = set()\\n \\n # 匹配1: 传统 function 声明 (例如: function myFunc(...) )\\n p1 = re.compile(r'\\\\bfunction\\\\s+([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\(')\\n functions.update(p1.findall(code_text))\\n \\n # 匹配2: 匿名函数赋值给变量或属性 (例如: myFunc = function(...) 或者 myFunc: function(...) )\\n p2 = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*[:=]\\\\s*function\\\\s*(?:\\\\(|$)')\\n functions.update(p2.findall(code_text))\\n \\n # 匹配3: 箭头函数赋值 (例如: myFunc = () => 或者 myFunc = a => )\\n p3 = re.compile(r'([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*[:=]\\\\s*(?:\\\\([^)]*\\\\)|[a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*=>')\\n functions.update(p3.findall(code_text))\\n \\n # 匹配4: ES6 类方法或对象方法 (例如: myFunc(...) { )\\n # 此类格式容易与原生控制流语法混淆，需排除 JS 中的常见保留字和关键字\\n keywords = {\\n 'if', 'for', 'while', 'catch', 'switch', 'return', 'function', \\n 'constructor', 'super', 'typeof', 'instanceof', 'new', 'else', \\n 'try', 'finally', 'do', 'with'\\n }\\n p4 = re.compile(r'\\\\b([a-zA-Z_$][0-9a-zA-Z_$]*)\\\\s*\\\\([^)]*\\\\)\\\\s*\\\\{')\\n matches = p4.findall(code_text)\\n for m in matches:\\n if m not in keywords and not m.isdigit():\\n functions.add(m)\\n \\n # 匹配5: Python 等后端函数定义 (例如: def my_func(...): )\\n p_py = re.compile(r'\\\\bdef\\\\s+([a-zA-Z_][a-zA-Z0-9_]*)\\\\s*\\\\(')\\n functions.update(p_py.findall(code_text))\\n \\n return functions\\n\\n # 遍历工作空间目录中的所有可能包含代码的文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 根据项目指令，文件主要为 HTML, JS，但扩大范围以适配衍生类型\\n if file.endswith(('.html', '.js', '.ts', '.jsx', '.tsx', '.py')):\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 提取文件中的所有函数名称\\n funcs = get_functions(content)\\n \\n # 检查每个提取出的函数名称是否为回文\\n for func_name in funcs:\\n if func_name != func_name[::-1]:\\n return False\\n except Exception:\\n # 忽略读取异常的文件，避免中断检查流程\\n pass\\n\\n # 若未找到任何违背条件（非回文）的函数，则符合要求\\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"确保你生成的每一个代码文件都精确包含 3 行注释\",\n \"validation_code\": \"def check_comment_lines(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 根据用户指令推断，科普实验网站的核心代码文件类型为 HTML, CSS, JS\\n code_extensions = {'.html', '.css', '.js'}\\n code_file_count = 0\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_file_count += 1\\n filepath = os.path.join(root, file)\\n try:\\n with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:\\n text = f.read()\\n except Exception:\\n continue\\n \\n # 针对不同扩展名的代码构建能够区分出字符串和不同类型注释的正则表达式\\n # 使用 re.DOTALL 使得 . 能够匹配跨行字符\\n if ext == '.js':\\n pattern = re.compile(\\n r'(?P<string>\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"|\\\\'[^\\\\'\\\\\\\\]*(?:\\\\\\\\.[^\\\\'\\\\\\\\]*)*\\\\'|`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`)'\\n r'|(?P<mcomment>/\\\\*.*?\\\\*/)'\\n r'|(?P<scomment>//[^\\\\n]*)',\\n re.DOTALL\\n )\\n elif ext == '.css':\\n pattern = re.compile(\\n r'(?P<string>\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"|\\\\'[^\\\\'\\\\\\\\]*(?:\\\\\\\\.[^\\\\'\\\\\\\\]*)*\\\\')'\\n r'|(?P<mcomment>/\\\\*.*?\\\\*/)',\\n re.DOTALL\\n )\\n elif ext == '.html':\\n pattern = re.compile(\\n r'(?P<string>\\\"[^\\\"\\\\\\\\]*(?:\\\\\\\\.[^\\\"\\\\\\\\]*)*\\\"|\\\\'[^\\\\'\\\\\\\\]*(?:\\\\\\\\.[^\\\\'\\\\\\\\]*)*\\\\'|`[^`\\\\\\\\]*(?:\\\\\\\\.[^`\\\\\\\\]*)*`)'\\n r'|(?P<hcomment><!--.*?-->)'\\n r'|(?P<mcomment>/\\\\*.*?\\\\*/)'\\n r'|(?P<scomment>(?<!:)//[^\\\\n]*)',\\n re.DOTALL\\n )\\n \\n comment_lines = set()\\n for m in pattern.finditer(text):\\n group_dict = m.groupdict()\\n start, end = -1, -1\\n \\n if group_dict.get('mcomment') is not None:\\n start, end = m.start('mcomment'), m.end('mcomment')\\n elif group_dict.get('hcomment') is not None:\\n start, end = m.start('hcomment'), m.end('hcomment')\\n elif group_dict.get('scomment') is not None:\\n start, end = m.start('scomment'), m.end('scomment')\\n \\n if start != -1 and end != -1:\\n # 基于匹配的首尾索引，计算之前换行符的数量来确定行号\\n start_line = text.count('\\\\n', 0, start)\\n end_line = text.count('\\\\n', 0, end - 1) if end > start else start_line\\n for line_num in range(start_line, end_line + 1):\\n comment_lines.add(line_num)\\n \\n # 如果有任何一个代码文件提取出的注释占据的总行数不等于 3，说明违背了该要求\\n if len(comment_lines) != 3:\\n return False\\n\\n # 考虑到项目应当生成至少一个代码文件，如果没有找到代码文件，也视作不合规\\n if code_file_count == 0:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"对于你生成的代码文件中的每一个类，其名称长度必须严格满足偶数的要求\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_class_name_length(response: str, workspace_path: str) -> bool:\\n # 匹配 JavaScript 中使用 class 关键字定义的类名（如 class MyClass）\\n js_class_pattern = re.compile(r'\\\\bclass\\\\s+([a-zA-Z0-9_]+)\\\\b')\\n # 匹配 HTML 中 class 属性指定的多个或单个类名\\n html_class_pattern = re.compile(r'\\\\bclass\\\\s*=\\\\s*([\\\"\\\\'])(.*?)\\\\1', re.IGNORECASE)\\n # 匹配 CSS 中的类选择器（以点 . 开头，由字母/下划线起头后接合法标识符）\\n css_class_pattern = re.compile(r'\\\\.([a-zA-Z_][a-zA-Z0-9_]*)\\\\b')\\n \\n # 匹配各类前端注释（JS 单多行注释，HTML/CSS 多行注释）以在解析前移除\\n comment_pattern = re.compile(r'//.*|/\\\\*.*?\\\\*/|<!--.*?-->', re.DOTALL)\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_path = os.path.join(root, file)\\n # 仅处理前端常见的代码文件格式\\n if not file.endswith(('.js', '.html', '.css', '.ts', '.jsx', '.tsx')):\\n continue\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 移除所有注释内容，防止因匹配注释中的示例代码导致误报\\n content = re.sub(comment_pattern, '', content)\\n \\n # 1. 提取并验证 JS/TS 代码中的类名\\n if file.endswith(('.js', '.ts', '.jsx', '.tsx', '.html')):\\n js_text = content\\n if file.endswith('.html'):\\n # 仅提取 HTML 中的 script 标签内部的内容\\n scripts = re.findall(r'<script\\\\b[^>]*>(.*?)</script>', content, re.DOTALL | re.IGNORECASE)\\n js_text = '\\\\n'.join(scripts)\\n \\n for cls_name in js_class_pattern.findall(js_text):\\n # 约束：由于未明确指明“字符数”，严格依据原则必须调用 count_word\\n if count_word(cls_name) % 2 != 0:\\n return False\\n \\n # 2. 提取并验证 HTML 等标签内部 class 属性赋值的类名\\n if file.endswith(('.html', '.jsx', '.tsx')):\\n for match in html_class_pattern.findall(content):\\n # match[1] 对应的是具体的多个/单个类名值字符串\\n classes_str = match[1]\\n for cls in classes_str.split():\\n if count_word(cls) % 2 != 0:\\n return False\\n \\n # 3. 提取并验证 CSS 样式中的类选择器名\\n if file.endswith(('.css', '.html')):\\n css_text = content\\n if file.endswith('.html'):\\n # 仅提取 HTML 中的 style 标签内部的 CSS 内容\\n styles = re.findall(r'<style\\\\b[^>]*>(.*?)</style>', content, re.DOTALL | re.IGNORECASE)\\n css_text = '\\\\n'.join(styles)\\n \\n for cls_name in css_class_pattern.findall(css_text):\\n if count_word(cls_name) % 2 != 0:\\n return False\\n\\n except Exception:\\n # 遇到非文本类型、编码异常等特殊情况直接忽略，不影响后续校验\\n continue\\n \\n return True\",\n \"tags\": [\n \"Style\",\n \"Naming\"\n ]\n },\n {\n \"constraint\": \"必须保证所生成的各个代码文件的总行数（将空行计算在内）为奇数\",\n \"validation_code\": \"def check_code_lines_are_odd(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据指令内容推断，项目主要为前端Web代码\\n code_extensions = {'.html', '.css', '.js'}\\n \\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 使用 '\\\\n' 分割可以精确统计出文本编辑器视角的行数，自动将空行计算在内\\n lines = content.split('\\\\n')\\n total_lines = len(lines)\\n \\n # 检查总行数是否为偶数\\n if total_lines % 2 == 0:\\n return False\\n except Exception:\\n # 如果遇到文件无法读取（例如非文本文件混入或编码错误等），视作不满足要求\\n return False\\n \\n return True\",\n \"tags\": [\n \"Quantity\",\n \"Exact Value\"\n ]\n },\n {\n \"constraint\": \"在每次对项目进行改动后，更新process-log.md文件以记录本轮完成的内容是必须执行的操作\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"在你提供的代码文件中，作为路径的分隔符时请优先采用正斜杠 /\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件应全部使用4个空格来控制缩进\",\n \"validation_code\": \"import os\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n # 根据用户指令推断，目标代码文件主要为这三种\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n \\n # 仅检查代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file_name)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n lines = f.readlines()\\n except Exception:\\n # 如果由于某种原因无法读取，则跳过\\n continue\\n \\n for line in lines:\\n stripped_line = line.strip()\\n # 1. 忽略空行\\n if not stripped_line:\\n continue\\n \\n # 2. 提取当前行的前导空白字符（行首的空格或制表符）\\n leading_whitespace = \\\"\\\"\\n for char in line:\\n if char in (' ', '\\\\t'):\\n leading_whitespace += char\\n else:\\n break\\n \\n # 3. 忽略完全没有缩进的顶层代码行\\n if not leading_whitespace:\\n continue\\n \\n # 4. 严格禁止使用制表符来控制缩进\\n if '\\\\t' in leading_whitespace:\\n return False\\n \\n # 5. 忽略多行注释对齐场景 (如行首有效字符是以 '*' 开头的)\\n if stripped_line.startswith('*'):\\n continue\\n \\n # 6. 验证缩进的空格数量是否严格为 4 的倍数\\n if len(leading_whitespace) % 4 != 0:\\n return False\\n\\n # 若所有代码文件均通过了上述严格缩进检测，则返回True\\n return True\",\n \"tags\": [\n \"Style\",\n \"Layout\"\n ]\n },\n {\n \"constraint\": \"进行文件操作前，先运行`pwd`确认当前目录是一项必不可少的步骤\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"确保在代码开发完成之后，测试所有的代码并保证它们均通过测试\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"你提供的每个代码文件里，注释内容须100%使用中文撰写\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Comment Language\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查页面整体布局：左侧是否为控制面板，且包含物体A类型下拉菜单、物体B类型下拉菜单、物体A重量滑块、物体B重量滑块、风阻系数滑块、重力加速度滑块（范围应支持1到20）和是否真空复选框；右侧是否为宽大的实验显示区，顶部是否放置两个待下落物体；底部是否包含“开始”“重置”按钮和实验记录表。","检查下拉菜单的联动逻辑：在左侧控制面板中将物体A或物体B选择为“羽毛”时，验证对应物体的重量滑块数值会自动变小、风阻系数会自动变大；改选为“小球”时，验证对应重量滑块数值会自动变大、风阻系数会自动变小。","检查物体UI表现：观察右侧实验区顶部的两个物体元素，验证它们是否根据下拉菜单选择，通过Emoji或CSS样式表现为直观可区分的“小球”和“羽毛”形状，而不是普通的同色或纯色圆形。","检查基础动画与按钮交互：点击“开始”按钮，观察实验区内两个物体是否使用平滑动画从顶部开始匀加速下落，并在触碰到底部边界时停止且不会穿出实验区；点击“重置”按钮，验证两个物体是否瞬间回到顶部初始位置，并清理当前实验中的运动状态和轨迹显示。","检查轨迹线/残影视觉效果：点击“开始”按钮，在物体下落过程中暂停或截图查看，验证两个物体的运动路径上是否动态生成轨迹线或残影效果，并能帮助用户直观对比两个物体在半空中的高度差。","检查基础无阻力同步下落逻辑：在未引入或未启用空气阻力影响的等效场景下，将两个物体设置为相同参数并点击“开始”，验证两个物体从顶部同步开始下落、保持相同高度，并在落地后能得到同一轮实验的完成结果。","检查空气阻力物理逻辑：确保“是否真空”未勾选，设置一个较轻且风阻较大的物体（如羽毛）和一个较重且风阻较小的物体（如小球），点击“开始”后验证小球下落速度明显快于羽毛，且落地耗时更短；调整重量或风阻系数后，下一次实验的下落速度和耗时应随参数变化。","检查真空环境物理逻辑：勾选“是否真空”复选框后，无论物体A和B的类型、重量、风阻系数如何，点击“开始”按钮，验证底层计算完全屏蔽空气阻力影响，两个物体始终保持完全相同的高度同步下落，并绝对同时触底。","检查重力加速度环境参数逻辑：确认重力加速度滑块支持1到20的范围；先将滑块拉到较小值（如1）完成一次实验，再重置并拉到较大值（如20）完成一次实验，验证较大g值下整体下落速度更快、落地耗时明显更短，且参数变化能正确反映在下一次动画中。","检查实验记录表行追加逻辑：点击“开始”并等待两个物体都落地后，验证底部实验记录表新增一行数据；点击“重置”并更改部分参数后再次完成实验，验证表格继续追加新记录而不是覆盖旧记录，能够保留历次实验数据。","检查实验记录表数据准确性：查看实验记录表最新追加的一行记录，验证该行是否正确且分别展示本次实验的环境参数（重力值/是否真空）、物体A参数、物体B参数，以及物体A和物体B各自独立的精确落地耗时。","检查代码实现方式要求：查看动画实现是否使用requestAnimationFrame驱动下落过程；查看空气阻力计算是否结合重量、风阻系数和速度进行实时加速度演算，并在真空模式下跳过风阻计算。"],"string":"[\n \"检查页面整体布局：左侧是否为控制面板，且包含物体A类型下拉菜单、物体B类型下拉菜单、物体A重量滑块、物体B重量滑块、风阻系数滑块、重力加速度滑块（范围应支持1到20）和是否真空复选框；右侧是否为宽大的实验显示区，顶部是否放置两个待下落物体；底部是否包含“开始”“重置”按钮和实验记录表。\",\n \"检查下拉菜单的联动逻辑：在左侧控制面板中将物体A或物体B选择为“羽毛”时，验证对应物体的重量滑块数值会自动变小、风阻系数会自动变大；改选为“小球”时，验证对应重量滑块数值会自动变大、风阻系数会自动变小。\",\n \"检查物体UI表现：观察右侧实验区顶部的两个物体元素，验证它们是否根据下拉菜单选择，通过Emoji或CSS样式表现为直观可区分的“小球”和“羽毛”形状，而不是普通的同色或纯色圆形。\",\n \"检查基础动画与按钮交互：点击“开始”按钮，观察实验区内两个物体是否使用平滑动画从顶部开始匀加速下落，并在触碰到底部边界时停止且不会穿出实验区；点击“重置”按钮，验证两个物体是否瞬间回到顶部初始位置，并清理当前实验中的运动状态和轨迹显示。\",\n \"检查轨迹线/残影视觉效果：点击“开始”按钮，在物体下落过程中暂停或截图查看，验证两个物体的运动路径上是否动态生成轨迹线或残影效果，并能帮助用户直观对比两个物体在半空中的高度差。\",\n \"检查基础无阻力同步下落逻辑：在未引入或未启用空气阻力影响的等效场景下，将两个物体设置为相同参数并点击“开始”，验证两个物体从顶部同步开始下落、保持相同高度，并在落地后能得到同一轮实验的完成结果。\",\n \"检查空气阻力物理逻辑：确保“是否真空”未勾选，设置一个较轻且风阻较大的物体（如羽毛）和一个较重且风阻较小的物体（如小球），点击“开始”后验证小球下落速度明显快于羽毛，且落地耗时更短；调整重量或风阻系数后，下一次实验的下落速度和耗时应随参数变化。\",\n \"检查真空环境物理逻辑：勾选“是否真空”复选框后，无论物体A和B的类型、重量、风阻系数如何，点击“开始”按钮，验证底层计算完全屏蔽空气阻力影响，两个物体始终保持完全相同的高度同步下落，并绝对同时触底。\",\n \"检查重力加速度环境参数逻辑：确认重力加速度滑块支持1到20的范围；先将滑块拉到较小值（如1）完成一次实验，再重置并拉到较大值（如20）完成一次实验，验证较大g值下整体下落速度更快、落地耗时明显更短，且参数变化能正确反映在下一次动画中。\",\n \"检查实验记录表行追加逻辑：点击“开始”并等待两个物体都落地后，验证底部实验记录表新增一行数据；点击“重置”并更改部分参数后再次完成实验，验证表格继续追加新记录而不是覆盖旧记录，能够保留历次实验数据。\",\n \"检查实验记录表数据准确性：查看实验记录表最新追加的一行记录，验证该行是否正确且分别展示本次实验的环境参数（重力值/是否真空）、物体A参数、物体B参数，以及物体A和物体B各自独立的精确落地耗时。\",\n \"检查代码实现方式要求：查看动画实现是否使用requestAnimationFrame驱动下落过程；查看空气阻力计算是否结合重量、风阻系数和速度进行实时加速度演算，并在真空模式下跳过风阻计算。\"\n]","isConversational":false}}},{"rowIdx":88,"cells":{"task_id":{"kind":"number","value":89,"string":"89"},"repository_policy":{"kind":"string","value":"后续所有轮次你都需要严格遵循以下约束条件：\n1. 在对现有文件进行修改之前，请务必先调用相关工具读取文件内容。\n2. 必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改。\n3. 在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录。\n4. 每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份。\n5. 你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注。\n6. 请确保在对所有用户的回复末尾，都附上\"汪~\"语气词。\n7. 每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的后缀名\n CODE_EXTENSIONS = {\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \n '.js', '.jsx', '.ts', '.tsx', '.go', '.rb', '.php', \n '.swift', '.kt', '.rs', '.sh', '.html', '.css', '.vue', '.sql', '.m', '.scala'\n }\n \n found_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 过滤常见的无关目录，避免扫描第三方依赖和版本控制文件\n dirs[:] = [d for d in dirs if d not in ['.git', '__pycache__', 'node_modules', 'venv', 'env', '.idea', '.vscode']]\n \n for file in files:\n # 过滤常见的备份文件标识（防止将未修改的原版备份当作生成文件进行误判）\n lower_file = file.lower()\n if 'bak' in lower_file or 'backup' in lower_file or 'old' in lower_file or 'copy' in lower_file:\n continue\n \n ext = os.path.splitext(file)[1].lower()\n # 如果是代码文件\n if ext in CODE_EXTENSIONS:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 判断是否包含要求的开发备注\n if '💻 开发备注' not in content:\n return False\n \n found_code_file = True\n except Exception:\n # 如果由于编码等问题无法读取，做跳过处理\n pass\n \n # 如果找到了合法的代码文件且均包含了目标备注，则返回True\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除字符串首尾的空白字符（包括空格、换行等）\n cleaned_response = response.strip()\n \n # 使用正则表达式去除字符串末尾可能跟随的常见中英文标点符号\n cleaned_response = re.sub(r'[。！？!\\.\\?]+$', '', cleaned_response).strip()\n \n # 判断清洗后的回复结尾是否为要求中的语气词\"汪~\"\n return cleaned_response.endswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]}],"string":"[\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的后缀名\\n CODE_EXTENSIONS = {\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', \\n '.js', '.jsx', '.ts', '.tsx', '.go', '.rb', '.php', \\n '.swift', '.kt', '.rs', '.sh', '.html', '.css', '.vue', '.sql', '.m', '.scala'\\n }\\n \\n found_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤常见的无关目录，避免扫描第三方依赖和版本控制文件\\n dirs[:] = [d for d in dirs if d not in ['.git', '__pycache__', 'node_modules', 'venv', 'env', '.idea', '.vscode']]\\n \\n for file in files:\\n # 过滤常见的备份文件标识（防止将未修改的原版备份当作生成文件进行误判）\\n lower_file = file.lower()\\n if 'bak' in lower_file or 'backup' in lower_file or 'old' in lower_file or 'copy' in lower_file:\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n # 如果是代码文件\\n if ext in CODE_EXTENSIONS:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 判断是否包含要求的开发备注\\n if '💻 开发备注' not in content:\\n return False\\n \\n found_code_file = True\\n except Exception:\\n # 如果由于编码等问题无法读取，做跳过处理\\n pass\\n \\n # 如果找到了合法的代码文件且均包含了目标备注，则返回True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除字符串首尾的空白字符（包括空格、换行等）\\n cleaned_response = response.strip()\\n \\n # 使用正则表达式去除字符串末尾可能跟随的常见中英文标点符号\\n cleaned_response = re.sub(r'[。！？!\\\\.\\\\?]+$', '', cleaned_response).strip()\\n \\n # 判断清洗后的回复结尾是否为要求中的语气词\\\"汪~\\\"\\n return cleaned_response.endswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我用 HTML 和 SVG 做一个网页版电子鱼缸。第一步，我们先搭建基础场景。请创建一个占据大部分页面空间的 SVG 画布作为鱼缸主体。画一个带圆角的矩形，背景填充淡淡的透明水蓝色，并在鱼缸一侧加上一个白色的半透明斜条状多边形，模拟出玻璃的微光反光效果。接着在鱼缸底部绘制几颗大小不一的棕色石头，以及几株绿色的水草（可以用几条波浪曲线拼接而成）。现在只需要布置好静态环境，不需要画任何活物。另外，后续所有开发操作都需要始终遵循以下两个要求：一是你所生成的所有代码文件都必须使用UTF-8进行编码；二是请确保在生成的每一个代码文件的注释里都包含🐟 emoji。","instruction_following_checklist":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 针对可能生成的代码文件格式配置注释的正则表达式及正则匹配标志\n # 块级多行注释使用 re.DOTALL 使得 '.' 可以匹配换行符\n # 单行注释不使用正则 flag（传0），使得 '.' 在行末截断\n code_ext_patterns = {\n '.html': [(r'<!--(.*?)-->', re.DOTALL)],\n '.htm': [(r'<!--(.*?)-->', re.DOTALL)],\n '.svg': [(r'<!--(.*?)-->', re.DOTALL)],\n '.xml': [(r'<!--(.*?)-->', re.DOTALL)],\n '.js': [(r'/\\*(.*?)\\*/', re.DOTALL), (r'//(.*)', 0)],\n '.ts': [(r'/\\*(.*?)\\*/', re.DOTALL), (r'//(.*)', 0)],\n '.css': [(r'/\\*(.*?)\\*/', re.DOTALL)],\n '.py': [(r'#(.*)', 0), (r'\\\"\\\"\\\"(.*?)\\\"\\\"\\\"', re.DOTALL), (r\"\\'\\'\\'(.*?)\\'\\'\\'\", re.DOTALL)]\n }\n \n code_file_count = 0\n \n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_ext_patterns:\n code_file_count += 1\n file_path = os.path.join(root, file)\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果无法按utf-8读取，或读取失败，则无法验证注释，直接判为不满足要求\n return False\n \n has_remark = False\n \n # 在文件中通过正则查找提取对应的注释内容\n for pattern, flag in code_ext_patterns[ext]:\n comments = re.findall(pattern, content, flags=flag)\n for comment in comments:\n # 剔除空格干扰进行匹配判断\n clean_comment = re.sub(r'\\s+', '', comment)\n if '💻开发备注' in clean_comment:\n has_remark = True\n break\n \n if has_remark:\n break\n \n # 只要存在某一个代码文件不包含指定开发备注的注释，则整体不满足要求\n if not has_remark:\n return False\n \n # 须至少生成一个有效的代码文件才算任务有效进行\n return code_file_count > 0","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除回复内容末尾可能的换行符、空格等空白字符\n cleaned_response = response.strip()\n \n # 判断回复末尾是否以\"汪~\"结尾\n return cleaned_response.endswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义可能的代码和相关文本文件扩展名集合\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.svg', '.md', '.txt',\n '.json', '.xml', '.yaml', '.yml', '.py', '.sh', \n '.java', '.c', '.cpp', '.h', '.hpp'\n }\n\n if not os.path.exists(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n # 忽略可能的版本控制等隐藏目录\n if '.git' in root or 'node_modules' in root:\n continue\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取并解码整个文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件包含无法被 utf-8 解码的字节序列\n return False\n except Exception:\n # 忽略诸如权限不足、文件正在被占用等非编码相关的读取错误\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"请确保在生成的每一个代码文件的注释里都包含🐟 emoji","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 常见的代码文件扩展名\n code_extensions = (\n '.html', '.htm', '.css', '.js', '.svg', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp',\n '.ts', '.jsx', '.tsx', '.php', '.rb', '.go', '.sh'\n )\n \n # 正则表达式匹配各类常见的代码注释格式\n comment_pattern = re.compile(\n r'<!--[\\s\\S]*?-->|' # HTML / XML / SVG 注释\n r'/\\*[\\s\\S]*?\\*/|' # CSS / JS 等 C风格多行注释\n r'//.*|' # JS / C++ 等 C风格单行注释\n r'#.*|' # Python / Shell 等单行注释\n r'\"\"\"[\\s\\S]*?\"\"\"|' # Python 多行注释 (双引号)\n r\"'''[\\s\\S]*?'''\" # Python 多行注释 (单引号)\n )\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 筛选需要被校验的代码文件，忽略 Markdown、普通文本、图片以及可能的非主流扩展名备份文件\n if file.lower().endswith(code_extensions):\n has_code_file = True\n file_path = os.path.join(root, file)\n \n try:\n # 尝试读取文件内容，指令要求全部采用 UTF-8 编码\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果未能读取，说明可能不符合 UTF-8 要求或其他错误，保守返回 False\n return False\n \n # 提取文件中的所有注释部分\n comments = comment_pattern.findall(content)\n \n # 检查找出的注释中是否包含指定的 🐟 emoji\n found_emoji = False\n for comment in comments:\n if '🐟' in comment:\n found_emoji = True\n break\n \n # 若某个代码文件中没有任何包含 🐟 的注释，说明不符合“每一个代码文件”的硬性条件\n if not found_emoji:\n return False\n \n # 如果项目中根本没有生成任何代码文件，则代表并未完成用户的代码创建任务\n return has_code_file","tags":["Content","Keyword"]}]},{"round_id":1,"instruction":"环境布置得很好。现在请在鱼缸水域中加入两条金色的鲤鱼，一条体型偏大，一条较小。请详细绘制出鲤鱼的特征：用 SVG 路径画出流线型的身体，加上眼睛，并在尾部画出分叉的尾鳍。请使用 `<g>` 标签将构成鱼的所有形状组合在一起，并调整好组内的坐标系，先让这两条鱼静态地悬停在鱼缸中央偏下的位置即可。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji和✨ emoji；二是代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过。","instruction_following_checklist":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"def check_developer_note(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见代码文件的扩展名集合，覆盖本项目可能涉及的 HTML、SVG 及其它相关类型\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.svg', '.c', '.cpp', '.java'}\n code_files_count = 0\n \n # 构建正则表达式匹配各类代码文件注释中包含 \"💻 开发备注\" 的模式\n # re.DOTALL 使得 . 可以匹配换行符，以便匹配跨行的多行注释块\n pattern = re.compile(\n r'<!--.*?💻 开发备注.*?-->' # HTML/XML/SVG 注释\n r'|//[^\\n]*💻 开发备注' # JS/CSS/C/Java 等单行注释\n r'|/\\*.*?💻 开发备注.*?\\*/' # JS/CSS/C/Java 等多行注释\n r'|#[^\\n]*💻 开发备注' # Python/Shell 单行注释\n r'|\"\"\".*?💻 开发备注.*?\"\"\"' # Python 多行注释\n r\"|'''.*?💻 开发备注.*?'''\", # Python 多行注释\n re.DOTALL\n )\n\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉不需要检查的包目录、系统目录或环境缓存\n exclude_dirs = {'.git', 'node_modules', '__pycache__', 'venv', 'env', '.vscode', '.idea'}\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查文件内容中是否含有满足正则的有效注释\n if not pattern.search(content):\n return False\n \n code_files_count += 1\n except Exception:\n # 如果由于字符编码等问题无法正常读取，视作未满足要求\n return False\n \n # 需确保项目中包含至少一个代码文件，且它们全数满足要求\n return code_files_count > 0","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 匹配回复字符串末尾是否为\"汪~\"，并允许\"汪~\"之后存在空白字符或常见句末标点符号（如句号、感叹号等）\n # \\s 代表空白字符；。！!\\., 代表中文或英文的常见结束标点\n pattern = re.compile(r\"汪~[\\s。！!\\.,]*$\")\n \n return bool(pattern.search(response))","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 定义常见代码文件的扩展名集合\n code_extensions = {\n '.html', '.svg', '.css', '.js', '.ts', \n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \n '.json', '.xml', '.sh', '.md'\n }\n \n has_code_file = False\n \n # 遍历工作区目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'rb') as f:\n content = f.read()\n # 尝试用 utf-8 进行解码以验证编码格式\n content.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是 UTF-8 编码\n return False\n except Exception:\n # 遇到其他读取错误，也保守返回 False\n return False\n \n # 如果找到了至少一个代码文件，且全部都能成功解码为 utf-8，则要求被满足\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji和✨ emoji","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n import re\n\n # 常见的前端项目代码文件扩展名\n code_exts = {'.html', '.htm', '.svg', '.css', '.js'}\n found_code_file = False\n \n # 匹配各类常见注释：\n # <!-- ... --> 用于 HTML/SVG/XML\n # /* ... */ 用于 CSS 和 JS 多行注释\n # // ... 用于 JS 单行注释 (使用负向先行断言排除如 http:// 中的双斜线)\n # # ... 用于 Python, Shell 等单行注释\n pattern = re.compile(r'<!--(.*?)-->|/\\*(.*?)\\*/|(?<!:)//([^\\n]*)|#([^\\n]*)', re.DOTALL)\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_lower = file.lower()\n # 排除AI可能创建的历史备份文件，防止上一轮的合法产物在当前轮被误认为缺少✨\n if 'bak' in file_lower or 'backup' in file_lower or 'old' in file_lower:\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n file_path = os.path.join(root, file)\n found_code_file = True\n \n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果发生读取错误（如非UTF-8编码等情况），判定为不满足条件\n return False\n \n # 提取所有命中注释规则的文本组\n matches = pattern.findall(content)\n # 由于 re.findall 的每个匹配结果是一个多组的 Tuple，此处将其全部扁平化拼接为一个完整的注释字符串\n all_comments = \"\".join(\"\".join(m) for m in matches)\n \n # 检查该文件的注释是否同时满足包含这两个 emoji \n if '🐟' not in all_comments or '✨' not in all_comments:\n return False\n \n # 如果遍历完毕未发生违背规则的情况，且至少找到了一个代码文件，则认为满足了要求\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":2,"instruction":"让这两条鱼动起来！首先，给鱼的尾鳍添加独立的动画效果，利用 CSS 的 @keyframes 和 transform: rotate() 让尾巴不断地左右摆动，模仿划水的动作。其次，编写 JavaScript 代码让鱼在鱼缸范围内进行随机游动。要求游动轨迹要尽量平滑，并且在改变游动方向时，鱼体必须整体旋转，让头部始终对准它即将游向的前进方向。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji和🌊 emoji；二是对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误。","instruction_following_checklist":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"import os\nimport re\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.svg', '.py', '.sh', '.vue'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n \n # 尝试读取文件内容，兼容 utf-8 和 gbk 编码\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except UnicodeDecodeError:\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n content = f.read()\n except Exception:\n continue\n except Exception:\n continue\n \n # 用于存放文件中提取到的所有注释内容\n comments = []\n \n # 1. HTML / SVG 注释 <!-- ... -->\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 2. JS / CSS / TS 等多行注释 /* ... */\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n # 3. JS / TS 单行注释 // ...\n comments.extend(re.findall(r'//(.*)', content))\n # 4. Python / Shell 等脚本单行注释 # ...\n comments.extend(re.findall(r'#(.*)', content))\n \n # 判断当前代码文件中，是否存在某个注释含有要求的内容\n is_satisfied = any(\"💻 开发备注\" in comment for comment in comments)\n \n # 如果任意一个代码文件的注释不包含要求的内容，则直接验证失败\n if not is_satisfied:\n return False\n \n # 确保至少检查并通过了一个代码文件\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 去除末尾的空白字符\n cleaned_response = response.strip()\n \n # 可能出现在末尾的常见标点符号列表\n trailing_punctuations = [' ', '\\n', '\\r', '\\t', '。', '！', '!', '.', '\"', \"'\", '”', '’', '…']\n \n # 循环剔除末尾的标点符号\n while cleaned_response and cleaned_response[-1] in trailing_punctuations:\n cleaned_response = cleaned_response[:-1]\n \n # 判断最终是否以\"汪~\"结尾\n return cleaned_response.endswith(\"汪~\")","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 确保工作区路径存在\n if not os.path.isdir(workspace_path):\n return False\n \n # 定义常见的代码及文本文件后缀（基于用户指令及常规前端项目情况）\n code_extensions = {'.html', '.js', '.css', '.svg', '.md', '.json', '.ts'}\n code_files_found = 0\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found += 1\n file_path = os.path.join(root, file)\n \n # 尝试以 utf-8 编码读取整个文件\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件不是使用合法的 UTF-8 编码\n return False\n except Exception:\n # 忽略权限不足或其他读取异常\n pass\n \n # 如果找到了至少一个代码文件且都没有引发解码错误，则视为满足要求\n return code_files_found > 0","tags":["Language","File Encoding"]},{"constraint":"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji和🌊 emoji","validation_code":"import os\nimport re\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 根据指令推断，常见的生成代码文件类型\n code_extensions = {'.html', '.svg', '.css', '.js'}\n code_files = []\n \n # 遍历当前工作区寻找代码文件\n for root, dirs, files in os.walk(workspace_path):\n # 过滤掉 node_modules 目录和隐藏目录，避免检查第三方依赖文件\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\n \n for file in files:\n file_lower = file.lower()\n # 找到目标类型的代码文件\n if any(file_lower.endswith(ext) for ext in code_extensions):\n # 排除备份文件，备份文件可能会保留前序轮次未包含全部三个 emoji 的情况\n if 'bak' not in file_lower and 'backup' not in file_lower:\n code_files.append(os.path.join(root, file))\n \n # 如果没有找到任何相关的代码文件，则不满足“生成代码文件”的前提条件\n if not code_files:\n return False\n \n # 逐一检查每一个找到的代码文件\n for file_path in code_files:\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 提取 HTML 和 SVG 的注释 (支持跨行)\n html_comments = re.findall(r'<!--(.*?)-->', content, re.DOTALL)\n # 提取 CSS 和 JS 的块级注释 (支持跨行)\n block_comments = re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL)\n # 提取 JS 的单行注释\n inline_comments = re.findall(r'//(.*)', content)\n \n # 将所有注释内容合并成一个完整的字符串\n all_comments = \" \".join(html_comments + block_comments + inline_comments)\n \n # 检查合并后的注释中是否同时包含这三个指定的 emoji\n if \"🐟\" not in all_comments or \"✨\" not in all_comments or \"🌊\" not in all_comments:\n return False\n \n except Exception:\n # 遇到文件读取错误或者编码问题按无法满足要求处理\n return False\n \n # 如果所有的有效代码文件的注释都符合条件，则要求被满足\n return True","tags":["Content","Keyword"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":3,"instruction":"游动效果不错，现在加入喂食的互动功能。请给整个 SVG 鱼缸区域添加点击事件监听。当我用鼠标点击水域的任意位置时，立刻在该坐标附近散落生成 3 颗棕色的小圆点代表“鱼饲料”。饲料一出现，就要通过 JS的 requestAnimationFrame 或者 CSS 动画，让它们以较慢的速度匀速垂直下沉，直到落到鱼缸底部停止。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji和🍚 emoji；二是为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数。","instruction_following_checklist":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 定义常见的代码文件后缀名集合\n code_exts = {'.html', '.js', '.css', '.svg', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.cs', '.go'}\n # 排除常见的无关目录或隐藏目录\n exclude_dirs = {'.git', 'node_modules', '.vscode', '__pycache__', 'dist', 'build'}\n \n found_code_file = False\n for root, dirs, files in os.walk(workspace_path):\n # 过滤需要排除的目录，避免无关遍历\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_exts:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查文件文本中是否包含目标字符串（兼顾有无空格的情况）\n if \"💻 开发备注\" not in content and \"💻开发备注\" not in content:\n return False\n except Exception:\n # 如果读取时发生编码错误等情况，无法验证其内容，返回 False\n return False\n \n # 如果找到了代码文件且上述检查全部通过，才视为满足要求\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 定义正则表达式，匹配以“汪~”加上可选的标点符号、空白字符、Markdown符号等结尾的模式\n # \\s 匹配空白字符和换行；字符集内包含常见的中文/英文标点以及 Markdown 标识符\n pattern = r'汪~[\\s。！，、；：“”’‘\\'\\\"!\\?？\\.\\*\\-_]*$'\n \n # 使用正则搜索回复字符串末尾是否满足该条件\n match = re.search(pattern, response.strip())\n \n return match is not None","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据多轮指令推断，可能的代码文件扩展名\n # HTML, SVG, CSS, JS (前端代码)\n # .log (变更日志 change_log_X.log)\n # .md (代码审查 code_review.md)\n code_extensions = {'.html', '.css', '.js', '.svg', '.log', '.md'}\n \n # 遍历工作区目录\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n \n # 仅对符合推测的代码文件类型进行校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制形式读取文件内容\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n \n # 尝试用 utf-8 解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 若解码抛出异常，说明不是合法的 UTF-8 编码格式\n return False\n except Exception:\n # 忽略其他读取相关的异常（如权限问题等），仅专注编码校验\n continue\n \n # 所有检查的文件均符合 UTF-8 编码，返回 True\n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji和🍚 emoji","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 根据项目性质推断常见的代码扩展名\n code_extensions = {'.html', '.js', '.css', '.svg', '.ts', '.py'}\n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_lower = file.lower()\n # 排除日志文件和可能的备份文件，确保只检查当前最新版本的代码文件\n if file_lower.endswith('.log') or 'bak' in file_lower or 'backup' in file_lower or 'change_log' in file_lower:\n continue\n \n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n continue\n \n has_code_file = True\n comments = []\n \n # 提取 HTML 和 SVG 的注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\n # 提取 多行注释 (JS/CSS/TS)\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, flags=re.DOTALL))\n # 提取 单行注释 (JS/TS)，运用断言排除形如 http:// 的 url\n comments.extend(re.findall(r'(?<!:)//(.*)', content))\n # 提取 Python 单行注释 (仅以防万一有后台或脚本文件)\n if ext == '.py':\n comments.extend(re.findall(r'#(.*)', content))\n \n joined_comments = ' '.join(comments)\n \n # 检查所有规定的 emoji 是否均在提取出来的注释中\n required_emojis = ['🐟', '✨', '🌊', '🍚']\n for emoji in required_emojis:\n if emoji not in joined_comments:\n return False\n \n return has_code_file","tags":["Content","Keyword"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":4,"instruction":"现在增加一点高难度的逻辑：我们要让鱼具备“找食物”的能力。请修改鱼的运动控制逻辑，实时监测鱼缸内是否存在饲料元素。一旦发现有饲料落下，两条鱼必须立即中断原本的随机游动，各自计算并锁定距离自己最近的一颗饲料，调整鱼头角度，径直朝饲料的位置游过去（需要追踪下沉中的饲料动态坐标）。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji和🍘 emoji；二是确保并发运行的agent数量上限为1，即任何时候都最多只有一个在运行。","instruction_following_checklist":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 目标字符串\n target_string = \"💻 开发备注\"\n \n # 定义常见的代码文件拓展名\n code_extensions = {\n '.html', '.htm', '.js', '.css', '.svg', \n '.ts', '.jsx', '.tsx', '.py', '.java', \n '.c', '.cpp', '.h', '.hpp'\n }\n \n has_code_file = False\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件拓展名\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 读取代码文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查文件内容中是否包含规定的目标字符串\n if target_string not in content:\n return False\n except Exception:\n # 如果遇到文件无法正常使用utf-8读取的情况，视作不满足要求\n return False\n \n # 如果工作区内存在代码文件，且全部通过了检查，则返回 True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除字符串首尾的空白字符\n response = response.strip()\n \n # 使用正则表达式检查：确保回复以 \"汪~\" 结尾（允许 \"汪~\" 后面跟随标点符号等非单词字符）\n match = re.search(r'汪~[^\\w]*$', response)\n \n return bool(match)","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import os\n\n # 根据项目要求定义前端网页开发中可能出现的代码文件扩展名\n code_extensions = {'.html', '.js', '.css', '.svg', '.json', '.ts'}\n has_code_file = False\n\n # 若工作区路径不存在，则不满足要求\n if not os.path.isdir(workspace_path):\n return False\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 提取文件扩展名并转小写\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n has_code_file = True\n try:\n # 尝试以 utf-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 一旦捕获到解码错误，说明存在非 utf-8 编码的文件\n return False\n except Exception:\n # 其他读取异常如权限不够忽略即可，我们这里只专注于编码测试\n pass\n\n # 至少要验证到一个代码文件，且全部验证通过才返回 True\n return has_code_file","tags":["Language","File Encoding"]},{"constraint":"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji和🍘 emoji","validation_code":"import os\nimport re\n\ndef verify(response: str, workspace_path: str) -> bool:\n # 可能生成的代码文件后缀名\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.svg'}\n target_emojis = ['🐟', '✨', '🌊', '🍚', '🍘']\n \n # 定义各类注释的正则表达式模式\n html_comment_pattern = re.compile(r'<!--[\\s\\S]*?-->')\n multi_line_comment_pattern = re.compile(r'/\\*[\\s\\S]*?\\*/')\n single_line_comment_pattern = re.compile(r'//.*')\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file_name in files:\n ext = os.path.splitext(file_name)[1].lower()\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file_name)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n except Exception:\n # 如果由于编码等原因无法读取，直接跳过处理（视其可能引发的检查失败）\n continue\n \n # 提取此文件里的所有注释内容\n comments = []\n comments.extend(html_comment_pattern.findall(content))\n comments.extend(multi_line_comment_pattern.findall(content))\n comments.extend(single_line_comment_pattern.findall(content))\n \n all_comments_text = '\\n'.join(comments)\n \n # 检查该文件的注释里是否同时包含这5个emoji\n for emoji in target_emojis:\n if emoji not in all_comments_text:\n return False\n \n # 至少需要解析到一个代码文件并全验证通过才为True\n return has_code_file","tags":["Content","Keyword"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保并发运行的agent数量上限为1，即任何时候都最多只有一个在运行","validation_code":"","tags":["Workflow","Orchestration"]}]},{"round_id":5,"instruction":"最后，完成捕食的闭环。在鱼追踪饲料的过程中，加入距离判断（碰撞检测），当鱼的前端坐标与目标饲料足够接近时，将该饲料元素从画面中移除，表示“被吃掉”。吃完后立刻寻找下一颗剩下的饲料。当鱼缸中所有的饲料都被清理干净后，两条鱼要能平滑地退回之前自由自在、漫无目的的随机游动状态。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji、🍘 emoji和🍣 emoji；二是修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式。","instruction_following_checklist":[{"constraint":"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 常见的代码文件扩展名集合，主要侧重于本任务中涉及的前端以及常见的编程语言\n code_extensions = {\n '.html', '.htm', '.js', '.css', '.svg', '.ts', '.jsx', '.tsx', '.vue',\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.rb', \n '.php', '.sh', '.swift', '.kt'\n }\n \n found_code_file = False\n \n # 遍历工作区目录\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 过滤出代码文件\n if ext in code_extensions:\n found_code_file = True\n file_path = os.path.join(root, file)\n try:\n # 按照指定的UTF-8编码读取文件\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n # 检查是否包含指定的开发备注\n if \"💻 开发备注\" not in content:\n return False\n except Exception:\n # 如果UTF-8解码失败，尝试使用系统默认或其他编码进行降级处理\n try:\n with open(file_path, 'r', encoding='gbk') as f:\n content = f.read()\n if \"💻 开发备注\" not in content:\n return False\n except Exception:\n return False\n\n # 若未找到任何代码文件，说明未生成相关代码，返回False；反之返回True\n return found_code_file","tags":["Content","Keyword"]},{"constraint":"请确保在对所有用户的回复末尾，都附上\"汪~\"语气词","validation_code":"def check(response: str, workspace_path: str) -> bool:\n import re\n \n if not response:\n return False\n \n # 去除首尾的空白字符（如换行符、空格等）\n cleaned_response = response.strip()\n \n # 匹配末尾是否是“汪~”，允许跟随常见的句末标点符号\n pattern = r'汪~[。！？!?\\.\\'\\\"”]*$'\n match = re.search(pattern, cleaned_response)\n \n return bool(match)","tags":["Content","Persona"]},{"constraint":"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\nCode Quality Review: [Task]\n\nStrengths:\n- [What's done well]\n\nIssues:\n- CRITICAL: [Critical]\n- Important: [Significant]\n- Minor: [Nitpicks]\n\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\n```","validation_code":"","tags":["Environment","Logging"]},{"constraint":"你所生成的所有代码文件都必须使用UTF-8进行编码","validation_code":"import os\n\ndef check_requirements(response: str, workspace_path: str) -> bool:\n # 推断可能的代码文件扩展名集合\n code_extensions = {'.html', '.svg', '.js', '.css', '.ts', '.json', '.py'}\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对代码文件进行编码校验\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件整体内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出解码异常，说明文件未采用合法的 utf-8 编码\n return False\n except Exception:\n # 忽略其他可能导致读取失败的问题（例如权限问题），不直接判错\n pass\n \n # 所有代码文件均成功经过 UTF-8 验证\n return True","tags":["Language","File Encoding"]},{"constraint":"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji、🍘 emoji和🍣 emoji","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n import re\n \n # 定义常见的代码文件后缀名\n code_extensions = {\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \n '.css', '.scss', '.less', '.svg', '.py', '.java', \n '.cpp', '.c', '.h', '.cs', '.go', '.rb', '.php'\n }\n \n # 给定要求中提及的必须包含的所有 emoji\n required_emojis = ['🐟', '✨', '🌊', '🍚', '🍘', '🍣']\n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除版本控制和常见依赖目录，避免不必要的扫描\n if '.git' in dirs:\n dirs.remove('.git')\n if 'node_modules' in dirs:\n dirs.remove('node_modules')\n \n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n \n comments = []\n # 匹配 HTML / SVG 注释\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\n # 匹配 C-style 多行注释 (适用于 JS, CSS, Java, C++ 等)\n comments.extend(re.findall(r'/\\*(.*?)\\*/', content, re.DOTALL))\n # 匹配 C-style 单行注释 (适用于 JS, Java, C++ 等)\n comments.extend(re.findall(r'//(.*)', content))\n # 匹配 Python / Shell 单行注释\n comments.extend(re.findall(r'#(.*)', content))\n \n # 将同一文件中提取到的所有注释合并为一个字符串以供检查\n combined_comments = ' '.join(comments)\n \n # 检查指定的 6 个 emoji 是否全都存在于注释中\n for emoji in required_emojis:\n if emoji not in combined_comments:\n return False\n \n except Exception:\n # 若文件无法用 UTF-8 读取或发生其他未知错误，视作未满足要求\n return False\n \n # 如果项目中没有任何代码文件，表示没有可验证的内容，返回 False；否则返回 True 代表全部通过\n return code_files_found","tags":["Content","Keyword"]},{"constraint":"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"确保并发运行的agent数量上限为1，即任何时候都最多只有一个在运行","validation_code":"","tags":["Workflow","Orchestration"]},{"constraint":"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式","validation_code":"","tags":["Environment","File Operation"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我用 HTML 和 SVG 做一个网页版电子鱼缸。第一步，我们先搭建基础场景。请创建一个占据大部分页面空间的 SVG 画布作为鱼缸主体。画一个带圆角的矩形，背景填充淡淡的透明水蓝色，并在鱼缸一侧加上一个白色的半透明斜条状多边形，模拟出玻璃的微光反光效果。接着在鱼缸底部绘制几颗大小不一的棕色石头，以及几株绿色的水草（可以用几条波浪曲线拼接而成）。现在只需要布置好静态环境，不需要画任何活物。另外，后续所有开发操作都需要始终遵循以下两个要求：一是你所生成的所有代码文件都必须使用UTF-8进行编码；二是请确保在生成的每一个代码文件的注释里都包含🐟 emoji。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 针对可能生成的代码文件格式配置注释的正则表达式及正则匹配标志\\n # 块级多行注释使用 re.DOTALL 使得 '.' 可以匹配换行符\\n # 单行注释不使用正则 flag（传0），使得 '.' 在行末截断\\n code_ext_patterns = {\\n '.html': [(r'<!--(.*?)-->', re.DOTALL)],\\n '.htm': [(r'<!--(.*?)-->', re.DOTALL)],\\n '.svg': [(r'<!--(.*?)-->', re.DOTALL)],\\n '.xml': [(r'<!--(.*?)-->', re.DOTALL)],\\n '.js': [(r'/\\\\*(.*?)\\\\*/', re.DOTALL), (r'//(.*)', 0)],\\n '.ts': [(r'/\\\\*(.*?)\\\\*/', re.DOTALL), (r'//(.*)', 0)],\\n '.css': [(r'/\\\\*(.*?)\\\\*/', re.DOTALL)],\\n '.py': [(r'#(.*)', 0), (r'\\\\\\\"\\\\\\\"\\\\\\\"(.*?)\\\\\\\"\\\\\\\"\\\\\\\"', re.DOTALL), (r\\\"\\\\'\\\\'\\\\'(.*?)\\\\'\\\\'\\\\'\\\", re.DOTALL)]\\n }\\n \\n code_file_count = 0\\n \\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_ext_patterns:\\n code_file_count += 1\\n file_path = os.path.join(root, file)\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果无法按utf-8读取，或读取失败，则无法验证注释，直接判为不满足要求\\n return False\\n \\n has_remark = False\\n \\n # 在文件中通过正则查找提取对应的注释内容\\n for pattern, flag in code_ext_patterns[ext]:\\n comments = re.findall(pattern, content, flags=flag)\\n for comment in comments:\\n # 剔除空格干扰进行匹配判断\\n clean_comment = re.sub(r'\\\\s+', '', comment)\\n if '💻开发备注' in clean_comment:\\n has_remark = True\\n break\\n \\n if has_remark:\\n break\\n \\n # 只要存在某一个代码文件不包含指定开发备注的注释，则整体不满足要求\\n if not has_remark:\\n return False\\n \\n # 须至少生成一个有效的代码文件才算任务有效进行\\n return code_file_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除回复内容末尾可能的换行符、空格等空白字符\\n cleaned_response = response.strip()\\n \\n # 判断回复末尾是否以\\\"汪~\\\"结尾\\n return cleaned_response.endswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义可能的代码和相关文本文件扩展名集合\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.svg', '.md', '.txt',\\n '.json', '.xml', '.yaml', '.yml', '.py', '.sh', \\n '.java', '.c', '.cpp', '.h', '.hpp'\\n }\\n\\n if not os.path.exists(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 忽略可能的版本控制等隐藏目录\\n if '.git' in root or 'node_modules' in root:\\n continue\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取并解码整个文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件包含无法被 utf-8 解码的字节序列\\n return False\\n except Exception:\\n # 忽略诸如权限不足、文件正在被占用等非编码相关的读取错误\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的每一个代码文件的注释里都包含🐟 emoji\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 常见的代码文件扩展名\\n code_extensions = (\\n '.html', '.htm', '.css', '.js', '.svg', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp',\\n '.ts', '.jsx', '.tsx', '.php', '.rb', '.go', '.sh'\\n )\\n \\n # 正则表达式匹配各类常见的代码注释格式\\n comment_pattern = re.compile(\\n r'<!--[\\\\s\\\\S]*?-->|' # HTML / XML / SVG 注释\\n r'/\\\\*[\\\\s\\\\S]*?\\\\*/|' # CSS / JS 等 C风格多行注释\\n r'//.*|' # JS / C++ 等 C风格单行注释\\n r'#.*|' # Python / Shell 等单行注释\\n r'\\\"\\\"\\\"[\\\\s\\\\S]*?\\\"\\\"\\\"|' # Python 多行注释 (双引号)\\n r\\\"'''[\\\\s\\\\S]*?'''\\\" # Python 多行注释 (单引号)\\n )\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 筛选需要被校验的代码文件，忽略 Markdown、普通文本、图片以及可能的非主流扩展名备份文件\\n if file.lower().endswith(code_extensions):\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n try:\\n # 尝试读取文件内容，指令要求全部采用 UTF-8 编码\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果未能读取，说明可能不符合 UTF-8 要求或其他错误，保守返回 False\\n return False\\n \\n # 提取文件中的所有注释部分\\n comments = comment_pattern.findall(content)\\n \\n # 检查找出的注释中是否包含指定的 🐟 emoji\\n found_emoji = False\\n for comment in comments:\\n if '🐟' in comment:\\n found_emoji = True\\n break\\n \\n # 若某个代码文件中没有任何包含 🐟 的注释，说明不符合“每一个代码文件”的硬性条件\\n if not found_emoji:\\n return False\\n \\n # 如果项目中根本没有生成任何代码文件，则代表并未完成用户的代码创建任务\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"环境布置得很好。现在请在鱼缸水域中加入两条金色的鲤鱼，一条体型偏大，一条较小。请详细绘制出鲤鱼的特征：用 SVG 路径画出流线型的身体，加上眼睛，并在尾部画出分叉的尾鳍。请使用 `<g>` 标签将构成鱼的所有形状组合在一起，并调整好组内的坐标系，先让这两条鱼静态地悬停在鱼缸中央偏下的位置即可。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji和✨ emoji；二是代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"def check_developer_note(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见代码文件的扩展名集合，覆盖本项目可能涉及的 HTML、SVG 及其它相关类型\\n code_exts = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.py', '.svg', '.c', '.cpp', '.java'}\\n code_files_count = 0\\n \\n # 构建正则表达式匹配各类代码文件注释中包含 \\\"💻 开发备注\\\" 的模式\\n # re.DOTALL 使得 . 可以匹配换行符，以便匹配跨行的多行注释块\\n pattern = re.compile(\\n r'<!--.*?💻 开发备注.*?-->' # HTML/XML/SVG 注释\\n r'|//[^\\\\n]*💻 开发备注' # JS/CSS/C/Java 等单行注释\\n r'|/\\\\*.*?💻 开发备注.*?\\\\*/' # JS/CSS/C/Java 等多行注释\\n r'|#[^\\\\n]*💻 开发备注' # Python/Shell 单行注释\\n r'|\\\"\\\"\\\".*?💻 开发备注.*?\\\"\\\"\\\"' # Python 多行注释\\n r\\\"|'''.*?💻 开发备注.*?'''\\\", # Python 多行注释\\n re.DOTALL\\n )\\n\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉不需要检查的包目录、系统目录或环境缓存\\n exclude_dirs = {'.git', 'node_modules', '__pycache__', 'venv', 'env', '.vscode', '.idea'}\\n dirs[:] = [d for d in dirs if d not in exclude_dirs]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查文件内容中是否含有满足正则的有效注释\\n if not pattern.search(content):\\n return False\\n \\n code_files_count += 1\\n except Exception:\\n # 如果由于字符编码等问题无法正常读取，视作未满足要求\\n return False\\n \\n # 需确保项目中包含至少一个代码文件，且它们全数满足要求\\n return code_files_count > 0\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 匹配回复字符串末尾是否为\\\"汪~\\\"，并允许\\\"汪~\\\"之后存在空白字符或常见句末标点符号（如句号、感叹号等）\\n # \\\\s 代表空白字符；。！!\\\\., 代表中文或英文的常见结束标点\\n pattern = re.compile(r\\\"汪~[\\\\s。！!\\\\.,]*$\\\")\\n \\n return bool(pattern.search(response))\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 定义常见代码文件的扩展名集合\\n code_extensions = {\\n '.html', '.svg', '.css', '.js', '.ts', \\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', \\n '.json', '.xml', '.sh', '.md'\\n }\\n \\n has_code_file = False\\n \\n # 遍历工作区目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'rb') as f:\\n content = f.read()\\n # 尝试用 utf-8 进行解码以验证编码格式\\n content.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是 UTF-8 编码\\n return False\\n except Exception:\\n # 遇到其他读取错误，也保守返回 False\\n return False\\n \\n # 如果找到了至少一个代码文件，且全部都能成功解码为 utf-8，则要求被满足\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji和✨ emoji\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n\\n # 常见的前端项目代码文件扩展名\\n code_exts = {'.html', '.htm', '.svg', '.css', '.js'}\\n found_code_file = False\\n \\n # 匹配各类常见注释：\\n # <!-- ... --> 用于 HTML/SVG/XML\\n # /* ... */ 用于 CSS 和 JS 多行注释\\n # // ... 用于 JS 单行注释 (使用负向先行断言排除如 http:// 中的双斜线)\\n # # ... 用于 Python, Shell 等单行注释\\n pattern = re.compile(r'<!--(.*?)-->|/\\\\*(.*?)\\\\*/|(?<!:)//([^\\\\n]*)|#([^\\\\n]*)', re.DOTALL)\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_lower = file.lower()\\n # 排除AI可能创建的历史备份文件，防止上一轮的合法产物在当前轮被误认为缺少✨\\n if 'bak' in file_lower or 'backup' in file_lower or 'old' in file_lower:\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n file_path = os.path.join(root, file)\\n found_code_file = True\\n \\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果发生读取错误（如非UTF-8编码等情况），判定为不满足条件\\n return False\\n \\n # 提取所有命中注释规则的文本组\\n matches = pattern.findall(content)\\n # 由于 re.findall 的每个匹配结果是一个多组的 Tuple，此处将其全部扁平化拼接为一个完整的注释字符串\\n all_comments = \\\"\\\".join(\\\"\\\".join(m) for m in matches)\\n \\n # 检查该文件的注释是否同时满足包含这两个 emoji \\n if '🐟' not in all_comments or '✨' not in all_comments:\\n return False\\n \\n # 如果遍历完毕未发生违背规则的情况，且至少找到了一个代码文件，则认为满足了要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"让这两条鱼动起来！首先，给鱼的尾鳍添加独立的动画效果，利用 CSS 的 @keyframes 和 transform: rotate() 让尾巴不断地左右摆动，模仿划水的动作。其次，编写 JavaScript 代码让鱼在鱼缸范围内进行随机游动。要求游动轨迹要尽量平滑，并且在改变游动方向时，鱼体必须整体旋转，让头部始终对准它即将游向的前进方向。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji和🌊 emoji；二是对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', '.svg', '.py', '.sh', '.vue'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n \\n # 尝试读取文件内容，兼容 utf-8 和 gbk 编码\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except UnicodeDecodeError:\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n content = f.read()\\n except Exception:\\n continue\\n except Exception:\\n continue\\n \\n # 用于存放文件中提取到的所有注释内容\\n comments = []\\n \\n # 1. HTML / SVG 注释 <!-- ... -->\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 2. JS / CSS / TS 等多行注释 /* ... */\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n # 3. JS / TS 单行注释 // ...\\n comments.extend(re.findall(r'//(.*)', content))\\n # 4. Python / Shell 等脚本单行注释 # ...\\n comments.extend(re.findall(r'#(.*)', content))\\n \\n # 判断当前代码文件中，是否存在某个注释含有要求的内容\\n is_satisfied = any(\\\"💻 开发备注\\\" in comment for comment in comments)\\n \\n # 如果任意一个代码文件的注释不包含要求的内容，则直接验证失败\\n if not is_satisfied:\\n return False\\n \\n # 确保至少检查并通过了一个代码文件\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 去除末尾的空白字符\\n cleaned_response = response.strip()\\n \\n # 可能出现在末尾的常见标点符号列表\\n trailing_punctuations = [' ', '\\\\n', '\\\\r', '\\\\t', '。', '！', '!', '.', '\\\"', \\\"'\\\", '”', '’', '…']\\n \\n # 循环剔除末尾的标点符号\\n while cleaned_response and cleaned_response[-1] in trailing_punctuations:\\n cleaned_response = cleaned_response[:-1]\\n \\n # 判断最终是否以\\\"汪~\\\"结尾\\n return cleaned_response.endswith(\\\"汪~\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 确保工作区路径存在\\n if not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义常见的代码及文本文件后缀（基于用户指令及常规前端项目情况）\\n code_extensions = {'.html', '.js', '.css', '.svg', '.md', '.json', '.ts'}\\n code_files_found = 0\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found += 1\\n file_path = os.path.join(root, file)\\n \\n # 尝试以 utf-8 编码读取整个文件\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件不是使用合法的 UTF-8 编码\\n return False\\n except Exception:\\n # 忽略权限不足或其他读取异常\\n pass\\n \\n # 如果找到了至少一个代码文件且都没有引发解码错误，则视为满足要求\\n return code_files_found > 0\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji和🌊 emoji\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 根据指令推断，常见的生成代码文件类型\\n code_extensions = {'.html', '.svg', '.css', '.js'}\\n code_files = []\\n \\n # 遍历当前工作区寻找代码文件\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤掉 node_modules 目录和隐藏目录，避免检查第三方依赖文件\\n dirs[:] = [d for d in dirs if not d.startswith('.') and d != 'node_modules']\\n \\n for file in files:\\n file_lower = file.lower()\\n # 找到目标类型的代码文件\\n if any(file_lower.endswith(ext) for ext in code_extensions):\\n # 排除备份文件，备份文件可能会保留前序轮次未包含全部三个 emoji 的情况\\n if 'bak' not in file_lower and 'backup' not in file_lower:\\n code_files.append(os.path.join(root, file))\\n \\n # 如果没有找到任何相关的代码文件，则不满足“生成代码文件”的前提条件\\n if not code_files:\\n return False\\n \\n # 逐一检查每一个找到的代码文件\\n for file_path in code_files:\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 提取 HTML 和 SVG 的注释 (支持跨行)\\n html_comments = re.findall(r'<!--(.*?)-->', content, re.DOTALL)\\n # 提取 CSS 和 JS 的块级注释 (支持跨行)\\n block_comments = re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL)\\n # 提取 JS 的单行注释\\n inline_comments = re.findall(r'//(.*)', content)\\n \\n # 将所有注释内容合并成一个完整的字符串\\n all_comments = \\\" \\\".join(html_comments + block_comments + inline_comments)\\n \\n # 检查合并后的注释中是否同时包含这三个指定的 emoji\\n if \\\"🐟\\\" not in all_comments or \\\"✨\\\" not in all_comments or \\\"🌊\\\" not in all_comments:\\n return False\\n \\n except Exception:\\n # 遇到文件读取错误或者编码问题按无法满足要求处理\\n return False\\n \\n # 如果所有的有效代码文件的注释都符合条件，则要求被满足\\n return True\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"游动效果不错，现在加入喂食的互动功能。请给整个 SVG 鱼缸区域添加点击事件监听。当我用鼠标点击水域的任意位置时，立刻在该坐标附近散落生成 3 颗棕色的小圆点代表“鱼饲料”。饲料一出现，就要通过 JS的 requestAnimationFrame 或者 CSS 动画，让它们以较慢的速度匀速垂直下沉，直到落到鱼缸底部停止。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji和🍚 emoji；二是为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 定义常见的代码文件后缀名集合\\n code_exts = {'.html', '.js', '.css', '.svg', '.ts', '.jsx', '.tsx', '.vue', '.py', '.java', '.c', '.cpp', '.cs', '.go'}\\n # 排除常见的无关目录或隐藏目录\\n exclude_dirs = {'.git', 'node_modules', '.vscode', '__pycache__', 'dist', 'build'}\\n \\n found_code_file = False\\n for root, dirs, files in os.walk(workspace_path):\\n # 过滤需要排除的目录，避免无关遍历\\n dirs[:] = [d for d in dirs if d not in exclude_dirs and not d.startswith('.')]\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_exts:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查文件文本中是否包含目标字符串（兼顾有无空格的情况）\\n if \\\"💻 开发备注\\\" not in content and \\\"💻开发备注\\\" not in content:\\n return False\\n except Exception:\\n # 如果读取时发生编码错误等情况，无法验证其内容，返回 False\\n return False\\n \\n # 如果找到了代码文件且上述检查全部通过，才视为满足要求\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 定义正则表达式，匹配以“汪~”加上可选的标点符号、空白字符、Markdown符号等结尾的模式\\n # \\\\s 匹配空白字符和换行；字符集内包含常见的中文/英文标点以及 Markdown 标识符\\n pattern = r'汪~[\\\\s。！，、；：“”’‘\\\\'\\\\\\\"!\\\\?？\\\\.\\\\*\\\\-_]*$'\\n \\n # 使用正则搜索回复字符串末尾是否满足该条件\\n match = re.search(pattern, response.strip())\\n \\n return match is not None\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据多轮指令推断，可能的代码文件扩展名\\n # HTML, SVG, CSS, JS (前端代码)\\n # .log (变更日志 change_log_X.log)\\n # .md (代码审查 code_review.md)\\n code_extensions = {'.html', '.css', '.js', '.svg', '.log', '.md'}\\n \\n # 遍历工作区目录\\n for root, _, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 仅对符合推测的代码文件类型进行校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制形式读取文件内容\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n \\n # 尝试用 utf-8 解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 若解码抛出异常，说明不是合法的 UTF-8 编码格式\\n return False\\n except Exception:\\n # 忽略其他读取相关的异常（如权限问题等），仅专注编码校验\\n continue\\n \\n # 所有检查的文件均符合 UTF-8 编码，返回 True\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji和🍚 emoji\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 根据项目性质推断常见的代码扩展名\\n code_extensions = {'.html', '.js', '.css', '.svg', '.ts', '.py'}\\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n file_lower = file.lower()\\n # 排除日志文件和可能的备份文件，确保只检查当前最新版本的代码文件\\n if file_lower.endswith('.log') or 'bak' in file_lower or 'backup' in file_lower or 'change_log' in file_lower:\\n continue\\n \\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n continue\\n \\n has_code_file = True\\n comments = []\\n \\n # 提取 HTML 和 SVG 的注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, flags=re.DOTALL))\\n # 提取 多行注释 (JS/CSS/TS)\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, flags=re.DOTALL))\\n # 提取 单行注释 (JS/TS)，运用断言排除形如 http:// 的 url\\n comments.extend(re.findall(r'(?<!:)//(.*)', content))\\n # 提取 Python 单行注释 (仅以防万一有后台或脚本文件)\\n if ext == '.py':\\n comments.extend(re.findall(r'#(.*)', content))\\n \\n joined_comments = ' '.join(comments)\\n \\n # 检查所有规定的 emoji 是否均在提取出来的注释中\\n required_emojis = ['🐟', '✨', '🌊', '🍚']\\n for emoji in required_emojis:\\n if emoji not in joined_comments:\\n return False\\n \\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"现在增加一点高难度的逻辑：我们要让鱼具备“找食物”的能力。请修改鱼的运动控制逻辑，实时监测鱼缸内是否存在饲料元素。一旦发现有饲料落下，两条鱼必须立即中断原本的随机游动，各自计算并锁定距离自己最近的一颗饲料，调整鱼头角度，径直朝饲料的位置游过去（需要追踪下沉中的饲料动态坐标）。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji和🍘 emoji；二是确保并发运行的agent数量上限为1，即任何时候都最多只有一个在运行。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 目标字符串\\n target_string = \\\"💻 开发备注\\\"\\n \\n # 定义常见的代码文件拓展名\\n code_extensions = {\\n '.html', '.htm', '.js', '.css', '.svg', \\n '.ts', '.jsx', '.tsx', '.py', '.java', \\n '.c', '.cpp', '.h', '.hpp'\\n }\\n \\n has_code_file = False\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件拓展名\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否为代码文件\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 读取代码文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查文件内容中是否包含规定的目标字符串\\n if target_string not in content:\\n return False\\n except Exception:\\n # 如果遇到文件无法正常使用utf-8读取的情况，视作不满足要求\\n return False\\n \\n # 如果工作区内存在代码文件，且全部通过了检查，则返回 True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除字符串首尾的空白字符\\n response = response.strip()\\n \\n # 使用正则表达式检查：确保回复以 \\\"汪~\\\" 结尾（允许 \\\"汪~\\\" 后面跟随标点符号等非单词字符）\\n match = re.search(r'汪~[^\\\\w]*$', response)\\n \\n return bool(match)\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 根据项目要求定义前端网页开发中可能出现的代码文件扩展名\\n code_extensions = {'.html', '.js', '.css', '.svg', '.json', '.ts'}\\n has_code_file = False\\n\\n # 若工作区路径不存在，则不满足要求\\n if not os.path.isdir(workspace_path):\\n return False\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 提取文件扩展名并转小写\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n has_code_file = True\\n try:\\n # 尝试以 utf-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 一旦捕获到解码错误，说明存在非 utf-8 编码的文件\\n return False\\n except Exception:\\n # 其他读取异常如权限不够忽略即可，我们这里只专注于编码测试\\n pass\\n\\n # 至少要验证到一个代码文件，且全部验证通过才返回 True\\n return has_code_file\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji和🍘 emoji\",\n \"validation_code\": \"import os\\nimport re\\n\\ndef verify(response: str, workspace_path: str) -> bool:\\n # 可能生成的代码文件后缀名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.jsx', '.ts', '.tsx', '.svg'}\\n target_emojis = ['🐟', '✨', '🌊', '🍚', '🍘']\\n \\n # 定义各类注释的正则表达式模式\\n html_comment_pattern = re.compile(r'<!--[\\\\s\\\\S]*?-->')\\n multi_line_comment_pattern = re.compile(r'/\\\\*[\\\\s\\\\S]*?\\\\*/')\\n single_line_comment_pattern = re.compile(r'//.*')\\n \\n has_code_file = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file_name in files:\\n ext = os.path.splitext(file_name)[1].lower()\\n if ext in code_extensions:\\n has_code_file = True\\n file_path = os.path.join(root, file_name)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n except Exception:\\n # 如果由于编码等原因无法读取，直接跳过处理（视其可能引发的检查失败）\\n continue\\n \\n # 提取此文件里的所有注释内容\\n comments = []\\n comments.extend(html_comment_pattern.findall(content))\\n comments.extend(multi_line_comment_pattern.findall(content))\\n comments.extend(single_line_comment_pattern.findall(content))\\n \\n all_comments_text = '\\\\n'.join(comments)\\n \\n # 检查该文件的注释里是否同时包含这5个emoji\\n for emoji in target_emojis:\\n if emoji not in all_comments_text:\\n return False\\n \\n # 至少需要解析到一个代码文件并全验证通过才为True\\n return has_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保并发运行的agent数量上限为1，即任何时候都最多只有一个在运行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n }\n ]\n },\n {\n \"round_id\": 5,\n \"instruction\": \"最后，完成捕食的闭环。在鱼追踪饲料的过程中，加入距离判断（碰撞检测），当鱼的前端坐标与目标饲料足够接近时，将该饲料元素从画面中移除，表示“被吃掉”。吃完后立刻寻找下一颗剩下的饲料。当鱼缸中所有的饲料都被清理干净后，两条鱼要能平滑地退回之前自由自在、漫无目的的随机游动状态。另外，后续所有开发操作需要始终遵循以下更新和新增的要求：一是你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji、🍘 emoji和🍣 emoji；二是修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在对现有文件进行修改之前，请务必先调用相关工具读取文件内容\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"必须先通过 TodoWrite 工具规划好任务，然后才能修改代码。当你完成某一任务对应的代码修改（Write/Edit）后，下一步必须立即调用 TodoComplete 工具将其标记为完成，绝对不允许在标记完成前穿插执行其他任务的代码更改\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"在对文件执行任何操作的前一步，要求必须运行`pwd`了解当前工作目录\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"每次对已存在的文件进行编辑前，一定要在原文件的同级目录中建立一份修改前的备份\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"你所生成的全部代码文件，其注释内容里都需要含有 💻 开发备注\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 常见的代码文件扩展名集合，主要侧重于本任务中涉及的前端以及常见的编程语言\\n code_extensions = {\\n '.html', '.htm', '.js', '.css', '.svg', '.ts', '.jsx', '.tsx', '.vue',\\n '.py', '.java', '.c', '.cpp', '.h', '.hpp', '.cs', '.go', '.rs', '.rb', \\n '.php', '.sh', '.swift', '.kt'\\n }\\n \\n found_code_file = False\\n \\n # 遍历工作区目录\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 过滤出代码文件\\n if ext in code_extensions:\\n found_code_file = True\\n file_path = os.path.join(root, file)\\n try:\\n # 按照指定的UTF-8编码读取文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n # 检查是否包含指定的开发备注\\n if \\\"💻 开发备注\\\" not in content:\\n return False\\n except Exception:\\n # 如果UTF-8解码失败，尝试使用系统默认或其他编码进行降级处理\\n try:\\n with open(file_path, 'r', encoding='gbk') as f:\\n content = f.read()\\n if \\\"💻 开发备注\\\" not in content:\\n return False\\n except Exception:\\n return False\\n\\n # 若未找到任何代码文件，说明未生成相关代码，返回False；反之返回True\\n return found_code_file\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"请确保在对所有用户的回复末尾，都附上\\\"汪~\\\"语气词\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n import re\\n \\n if not response:\\n return False\\n \\n # 去除首尾的空白字符（如换行符、空格等）\\n cleaned_response = response.strip()\\n \\n # 匹配末尾是否是“汪~”，允许跟随常见的句末标点符号\\n pattern = r'汪~[。！？!?\\\\.\\\\'\\\\\\\"”]*$'\\n match = re.search(pattern, cleaned_response)\\n \\n return bool(match)\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"每当有新内容更新，请更新 code_review.md 文件以Review现有代码，并严格使用如下格式：```\\nCode Quality Review: [Task]\\n\\nStrengths:\\n- [What's done well]\\n\\nIssues:\\n- CRITICAL: [Critical]\\n- Important: [Significant]\\n- Minor: [Nitpicks]\\n\\nRecommendation: ✅ APPROVED or ❌ NEEDS FIXES\\n```\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"Logging\"\n ]\n },\n {\n \"constraint\": \"你所生成的所有代码文件都必须使用UTF-8进行编码\",\n \"validation_code\": \"import os\\n\\ndef check_requirements(response: str, workspace_path: str) -> bool:\\n # 推断可能的代码文件扩展名集合\\n code_extensions = {'.html', '.svg', '.js', '.css', '.ts', '.json', '.py'}\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 仅对代码文件进行编码校验\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件整体内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出解码异常，说明文件未采用合法的 utf-8 编码\\n return False\\n except Exception:\\n # 忽略其他可能导致读取失败的问题（例如权限问题），不直接判错\\n pass\\n \\n # 所有代码文件均成功经过 UTF-8 验证\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"你所生成的每一个代码文件的注释里都需要同时包含🐟 emoji、✨ emoji、🌊 emoji、🍚 emoji、🍘 emoji和🍣 emoji\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n import re\\n \\n # 定义常见的代码文件后缀名\\n code_extensions = {\\n '.html', '.htm', '.js', '.jsx', '.ts', '.tsx', \\n '.css', '.scss', '.less', '.svg', '.py', '.java', \\n '.cpp', '.c', '.h', '.cs', '.go', '.rb', '.php'\\n }\\n \\n # 给定要求中提及的必须包含的所有 emoji\\n required_emojis = ['🐟', '✨', '🌊', '🍚', '🍘', '🍣']\\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n # 排除版本控制和常见依赖目录，避免不必要的扫描\\n if '.git' in dirs:\\n dirs.remove('.git')\\n if 'node_modules' in dirs:\\n dirs.remove('node_modules')\\n \\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n file_path = os.path.join(root, file)\\n try:\\n with open(file_path, 'r', encoding='utf-8') as f:\\n content = f.read()\\n \\n comments = []\\n # 匹配 HTML / SVG 注释\\n comments.extend(re.findall(r'<!--(.*?)-->', content, re.DOTALL))\\n # 匹配 C-style 多行注释 (适用于 JS, CSS, Java, C++ 等)\\n comments.extend(re.findall(r'/\\\\*(.*?)\\\\*/', content, re.DOTALL))\\n # 匹配 C-style 单行注释 (适用于 JS, Java, C++ 等)\\n comments.extend(re.findall(r'//(.*)', content))\\n # 匹配 Python / Shell 单行注释\\n comments.extend(re.findall(r'#(.*)', content))\\n \\n # 将同一文件中提取到的所有注释合并为一个字符串以供检查\\n combined_comments = ' '.join(comments)\\n \\n # 检查指定的 6 个 emoji 是否全都存在于注释中\\n for emoji in required_emojis:\\n if emoji not in combined_comments:\\n return False\\n \\n except Exception:\\n # 若文件无法用 UTF-8 读取或发生其他未知错误，视作未满足要求\\n return False\\n \\n # 如果项目中没有任何代码文件，表示没有可验证的内容，返回 False；否则返回 True 代表全部通过\\n return code_files_found\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"代码编写一经完成，务必对所有的代码进行测试，并保证测试全部通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施 ESLint 验证是硬性要求，你必须更正所有的报错，确保最终的 ESLint 执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"为了记录项目的每一次变更，请在每次改动后创建一个新文件并将其命名为 change_log_X.log，这里的 X 是从 0 起算的改动次数\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"确保并发运行的agent数量上限为1，即任何时候都最多只有一个在运行\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Orchestration\"\n ]\n },\n {\n \"constraint\": \"修改任何现有文件时，请务必提供完整的代码以便进行覆盖替换，严禁采用 Unified Diff 等任何形式的补丁格式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["检查鱼缸的基础外观：页面主体是否显示为一个大的圆角矩形SVG元素，背景颜色为浅浅的透明水蓝色。","检查鱼缸的玻璃反光效果：观察鱼缸的一侧，确认是否存在白色的半透明斜条状多边形，用于模拟玻璃微光。","检查鱼缸底部装饰元素：观察鱼缸底部区域，确认是否绘制有几颗大小不一的棕色石头图案，以及由波浪曲线构成的几株绿色水草。","检查静态鱼的外观特征：观察鱼缸内部水域，确认是否显示有两条金色的鲤鱼，且体型明显是一大一小；鱼的图形需包含流线型身体、眼睛以及分叉的尾鳍。","检查鱼的尾鳍动画：通过连续截图或观察画面，确认两条金鱼的分叉尾鳍是否有独立的持续左右摆动动画效果（划水动作）。","检查鱼的基础随机游动：不进行任何点击操作，观察几秒钟，确认两条鱼是否在鱼缸的边界范围内平滑地上下左右改变位置，且没有游出鱼缸外部。","检查鱼的头部朝向逻辑：在鱼自由游动并改变方向时，观察鱼身体的整体旋转情况，确认鱼头部始终精确对准它当前游动的前进方向，没有出现平移或倒退游动的现象。","检查点击生成饲料功能：使用鼠标点击鱼缸水域内的任意位置，确认点击的瞬间是否在该鼠标坐标附近立即生成了3颗代表鱼饲料的棕色小圆点。","检查饲料的下沉动画：点击生成饲料后，确认这3颗棕色小圆点是否会以较慢的速度匀速垂直向下移动，且下沉到鱼缸底部后停止移动。","检查鱼追踪饲料逻辑：在点击鱼缸投放饲料后，观察两条鱼的行为，确认它们是否立即中断了原本的随机游动，各自将鱼头转向距离自己最近的一颗饲料的方向，并径直朝着该饲料下沉中的动态坐标游去。","检查吃食的碰撞检测效果：观察鱼追踪饲料的过程，确认当鱼的前端（头部）坐标接近并触碰到某颗饲料时，该颗饲料是否立即从画面中移除消失，并且鱼会继续寻找剩下的饲料。","检查游动状态的恢复闭环：当鱼缸内生成的饲料被鱼全部吃完（画面中不存在饲料元素）后，观察两条鱼，确认它们没有停留在原地卡死，而是平滑地重新恢复到漫无目的的随机游动状态。"],"string":"[\n \"检查鱼缸的基础外观：页面主体是否显示为一个大的圆角矩形SVG元素，背景颜色为浅浅的透明水蓝色。\",\n \"检查鱼缸的玻璃反光效果：观察鱼缸的一侧，确认是否存在白色的半透明斜条状多边形，用于模拟玻璃微光。\",\n \"检查鱼缸底部装饰元素：观察鱼缸底部区域，确认是否绘制有几颗大小不一的棕色石头图案，以及由波浪曲线构成的几株绿色水草。\",\n \"检查静态鱼的外观特征：观察鱼缸内部水域，确认是否显示有两条金色的鲤鱼，且体型明显是一大一小；鱼的图形需包含流线型身体、眼睛以及分叉的尾鳍。\",\n \"检查鱼的尾鳍动画：通过连续截图或观察画面，确认两条金鱼的分叉尾鳍是否有独立的持续左右摆动动画效果（划水动作）。\",\n \"检查鱼的基础随机游动：不进行任何点击操作，观察几秒钟，确认两条鱼是否在鱼缸的边界范围内平滑地上下左右改变位置，且没有游出鱼缸外部。\",\n \"检查鱼的头部朝向逻辑：在鱼自由游动并改变方向时，观察鱼身体的整体旋转情况，确认鱼头部始终精确对准它当前游动的前进方向，没有出现平移或倒退游动的现象。\",\n \"检查点击生成饲料功能：使用鼠标点击鱼缸水域内的任意位置，确认点击的瞬间是否在该鼠标坐标附近立即生成了3颗代表鱼饲料的棕色小圆点。\",\n \"检查饲料的下沉动画：点击生成饲料后，确认这3颗棕色小圆点是否会以较慢的速度匀速垂直向下移动，且下沉到鱼缸底部后停止移动。\",\n \"检查鱼追踪饲料逻辑：在点击鱼缸投放饲料后，观察两条鱼的行为，确认它们是否立即中断了原本的随机游动，各自将鱼头转向距离自己最近的一颗饲料的方向，并径直朝着该饲料下沉中的动态坐标游去。\",\n \"检查吃食的碰撞检测效果：观察鱼追踪饲料的过程，确认当鱼的前端（头部）坐标接近并触碰到某颗饲料时，该颗饲料是否立即从画面中移除消失，并且鱼会继续寻找剩下的饲料。\",\n \"检查游动状态的恢复闭环：当鱼缸内生成的饲料被鱼全部吃完（画面中不存在饲料元素）后，观察两条鱼，确认它们没有停留在原地卡死，而是平滑地重新恢复到漫无目的的随机游动状态。\"\n]","isConversational":false}}},{"rowIdx":89,"cells":{"task_id":{"kind":"number","value":90,"string":"90"},"repository_policy":{"kind":"string","value":"以下是你在后续所有轮次都必须严格遵循的约束条件：\n1. 在你的所有回复中，必须以语气词\"哈喽～\"作为开场白。\n2. 你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"。\n3. 禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"。\n4. 你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己。\n5. 要求在你的解答内出现 😊表情符号。\n6. 在为用户生成回复时，你必须永远只用中文。\n7. 生成的代码文件中，任何函数的圈复杂度都不能超过 4。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"在你的所有回复中，必须以语气词\"哈喽～\"作为开场白","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 移除回复前导的空白字符（如空格、换行等），以准确判断开场白\n cleaned_response = response.lstrip()\n # 判断是否以\"哈喽～\"开头\n return cleaned_response.startswith(\"哈喽～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否在最末尾加上了语气词\"呀～\"\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n if not response:\n return False\n \n # 去除字符串首尾的空白字符，包括换行符、空格等\n cleaned_response = response.strip()\n \n # 判断是否以\"呀～\"结尾\n return cleaned_response.endswith(\"呀～\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否没有出现“你”字。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 如果没有出现“你”则返回True，否则返回False\n \"\"\"\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查回复内容中是否不包含字\"我\"\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"要求在你的解答内出现 😊表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复中是否包含 😊 表情符号。\n \"\"\"\n return \"😊\" in response","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过 4","validation_code":"","tags":["Quantity","Complexity"]}],"string":"[\n {\n \"constraint\": \"在你的所有回复中，必须以语气词\\\"哈喽～\\\"作为开场白\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 移除回复前导的空白字符（如空格、换行等），以准确判断开场白\\n cleaned_response = response.lstrip()\\n # 判断是否以\\\"哈喽～\\\"开头\\n return cleaned_response.startswith(\\\"哈喽～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"呀～\\\"\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否在最末尾加上了语气词\\\"呀～\\\"\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除字符串首尾的空白字符，包括换行符、空格等\\n cleaned_response = response.strip()\\n \\n # 判断是否以\\\"呀～\\\"结尾\\n return cleaned_response.endswith(\\\"呀～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"亲爱的朋友\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否没有出现“你”字。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 如果没有出现“你”则返回True，否则返回False\\n \\\"\\\"\\\"\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的回答里不能出现自称\\\"我\\\"，必须统一使用\\\"小助手\\\"来代指你自己\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查回复内容中是否不包含字\\\"我\\\"\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"要求在你的解答内出现 😊表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复中是否包含 😊 表情符号。\\n \\\"\\\"\\\"\\n return \\\"😊\\\" in response\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n }\n]","isConversational":false},"task_category":{"kind":"string","value":"Frontend Development"},"rounds":{"kind":"list like","value":[{"round_id":0,"instruction":"帮我写一个网页版的卡牌记忆配对游戏。第一步，我们先搭个基础框架。在页面居中显示一个游戏标题，下方画一个4x4的卡牌网格。每张卡牌先用统一的背景色或背面图案占位。只需实现最基础的交互：点击任何一张卡牌时，卡牌能翻面（现在正面统一显示个纯色或者空白即可，只要能看出正反面状态就行）。暂时先不要写匹配判断逻辑，把网格布局、卡牌组件和基础的点击翻开动作做好就行，为后续增加游戏逻辑留好空间。请你注意，以下两条要求在后续所有轮次中都需要始终遵守：所有由你创建的代码文件，其编码必须统一为UTF-8；请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）。","instruction_following_checklist":[{"constraint":"在你的所有回复中，必须以语气词\"哈喽～\"作为开场白","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复是否以\"哈喽～\"作为开场白。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址\n :return: 布尔值，表示是否满足要求\n \"\"\"\n # 去除首尾空白字符后判断是否以\"哈喽～\"开头\n return response.strip().startswith(\"哈喽～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断AI助手的回复最末尾是否加上了语气词\"呀～\"。\n \"\"\"\n if not response:\n return False\n \n # 去除字符串最末尾的换行符和空格等空白字符\n cleaned_response = response.rstrip()\n \n # 判断是否以指定的语气词结尾\n return cleaned_response.endswith(\"呀～\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查助手的回复中是否禁止使用了“你”。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 布尔值，表示是否满足要求。\n \"\"\"\n # 如果回复中存在“你”，说明没有被替换为“亲爱的朋友”，违反了约束\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查助手的回复中是否出现了自称\"我\"。\n 如果出现\"我\"，则返回 False；否则返回 True。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"要求在你的解答内出现 😊表情符号","validation_code":"def check(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查助手的回复中是否包含😊表情符号。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 如果包含😊则返回True，否则返回False。\n \"\"\"\n return '😊' in response","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过 4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"import os\n\ndef check(response: str, workspace_path: str) -> bool:\n # 针对网页游戏项目，推断可能的代码文件扩展名\n code_extensions = {'.html', '.css', '.js'}\n \n # 遍历项目目录下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名并转换为小写\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 以二进制模式读取文件内容\n with open(file_path, 'rb') as f:\n raw_data = f.read()\n # 尝试使用utf-8进行解码\n raw_data.decode('utf-8')\n except UnicodeDecodeError:\n # 如果抛出UnicodeDecodeError，则说明不是有效的UTF-8编码\n return False\n except Exception:\n # 如果有其他读取错误（如权限问题），也视为不满足预期\n return False\n \n # 所有代码文件都能成功以UTF-8解码\n return True","tags":["Language","File Encoding"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]}]},{"round_id":1,"instruction":"很好，基础框架有了。现在来完善核心的记忆配对逻辑。请准备8对不同的Emoji表情作为卡牌的正面内容，每次刷新或游戏开始时，随机打乱分配到这16个格子里。规则是：玩家每次最多只能翻开两张牌，如果这两张牌的Emoji一样，它们就保持翻开状态（可以稍微变暗表示已完成）；如果不一样，大概停留1秒钟后自动翻转盖回去。注意一个关键挑战：在两张牌翻开且还在展示的那1秒延迟内，必须锁定整个网格，禁止玩家点击其他未翻开的牌，防止出现逻辑错乱。后续所有轮次请你始终遵循以下两个要求：请确保在生成的代码文件中，循环的最大嵌套深度不超过2层；你必须在代码写完后测试全部的代码，并确保测试成功通过。","instruction_following_checklist":[{"constraint":"在你的所有回复中，必须以语气词\"哈喽～\"作为开场白","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否以语气词\"哈喽～\"作为开场白。\n \n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 当前人工智能助手生成的代码项目地址（此项验证中不使用）\n :return: 布尔值，代表该要求的满足情况\n \"\"\"\n if not response:\n return False\n \n # 去除开头和结尾可能的空白字符后，判断是否以\"哈喽～\"开头\n return response.strip().startswith(\"哈喽～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"","validation_code":"def verify_ending_phrase(response: str, workspace_path: str) -> bool:\n import re\n \n # 去除字符串首尾的空白字符，直接判断\n if response.strip().endswith(\"呀～\"):\n return True\n \n # 为了防止助手在\"呀～\"后面加上了标点符号（如\"呀～！\", \"呀～。\"等）\n # 使用正则表达式去除末尾可能出现的空白字符以及常见中英文句末标点符号\n cleaned_response = re.sub(r'[\\s。！？\\!\\?\\.]+$', '', response)\n \n return cleaned_response.endswith(\"呀～\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证回复中是否禁止使用了“你”称呼用户。\n :param response: 人工智能助手最后一轮回复的内容\n :param workspace_path: 代码项目地址\n :return: 布尔值，代表是否满足要求\n \"\"\"\n # 只要回复中出现了“你”，即认为违反了该要求\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己","validation_code":"def check_condition(response: str, workspace_path: str) -> bool:\n if '我' in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"要求在你的解答内出现 😊表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断 AI 助手的回复中是否包含 😊 表情符号\n \"\"\"\n return \"😊\" in response","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过 4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n\n # 检查工作空间路径是否存在\n if not os.path.exists(workspace_path):\n return False\n \n # 根据网页项目推断可能的代码文件扩展名\n code_extensions = {'.html', '.htm', '.css', '.js', '.py', '.json'}\n \n # 遍历工作空间下的所有文件\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n # 获取文件扩展名并转为小写\n ext = os.path.splitext(file)[1].lower()\n \n # 判断是否属于代码文件\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 只要有一个代码文件无法用 utf-8 解码，说明不满足要求\n return False\n except Exception:\n # 发生其他读取异常，出于严谨也视作无法验证编码正确性\n return False\n \n # 所有代码文件均成功使用 utf-8 编码读取\n return True","tags":["Language","File Encoding"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在生成的代码文件中，循环的最大嵌套深度不超过2层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":2,"instruction":"匹配逻辑跑通了。下一步给游戏加上计分板和结算系统。在标题和卡牌网格之间增加一个状态栏，左侧显示当前尝试的“步数”（翻开两张牌算完整的一步），右侧显示“用时”（从玩家点击第一张牌那一刻起开始计时，格式为 mm:ss）。当所有8对卡牌都成功配对后，停止计时，并在页面正中间弹出一个“游戏通关”的模态框，里面要展示玩家的总用时和总步数，弹窗底部提供一个“再来一局”的按钮。点击该按钮后，步数和时间清零，弹窗关闭，所有卡牌重新盖上并重新洗牌。后续所有轮次请你始终遵循以下两个要求：第一，请确保在生成的代码文件中，循环的最大嵌套深度不超过1层；第二，对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则。","instruction_following_checklist":[{"constraint":"在你的所有回复中，必须以语气词\"哈喽～\"作为开场白","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 清除前后可能存在的空白字符，并判断是否以指定的开场白开头\n return response.strip().startswith(\"哈喽～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"","validation_code":"def check(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n # 去除末尾的空白字符和换行符，并判断是否以指定语气词结尾\n return response.strip().endswith(\"呀～\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的最后一轮回复是否满足：禁止用\"你\"称呼用户。\n \"\"\"\n if '你' in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己","validation_code":"def check(response: str, workspace_path: str) -> bool:\n # 检查回复中是否包含“我”\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"要求在你的解答内出现 😊表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n # 检查助手的回复中是否包含 \"😊\" 表情符号\n return \"😊\" in response","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过 4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n import os\n \n # 验证工作区路径是否存在\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\n return False\n \n # 定义常见的代码文件扩展名（结合前端小游戏和常见编程语言）\n code_extensions = {\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.json', '.xml', '.py', '.java', '.c', '.cpp', '.h', \n '.hpp', '.cs', '.go', '.rs', '.rb', '.php', '.swift', \n '.kt', '.sh', '.bat', '.vue'\n }\n \n code_files_found = False\n \n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n code_files_found = True\n filepath = os.path.join(root, file)\n try:\n # 尝试用 utf-8 编码去读取文件\n with open(filepath, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 抛出 UnicodeDecodeError 说明含有非 UTF-8 的字符（如 GBK 编码的中文）\n return False\n except Exception:\n # 如果遇到其他由于文件权限等导致的异常暂且忽略，以解码验证为准\n pass\n \n # 如果没有找到任何代码文件，说明助手未能按要求生成代码，返回 False\n return code_files_found","tags":["Language","File Encoding"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在生成的代码文件中，循环的最大嵌套深度不超过1层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]}]},{"round_id":3,"instruction":"现在的游戏已经很完整了，咱们来提升一下视觉体验和动画效果，稍微增加点前端挑战。首先，修改卡牌的翻转动画，要求实现带透视感的真实3D翻牌效果（绕Y轴平滑旋转180度，正面和背面内容分别附着在正反面上）。其次，加上状态反馈动画：当玩家翻开的两张牌配对成功时，让这两张牌做一个明显的“放大并闪烁”的庆祝动画；如果配对失败，在这两张牌翻盖回去之前，给它们加一个“左右剧烈摇晃”的错误提示动画。后续所有轮次请你始终遵循以下两个要求：第一，对生成的代码实施ESLint验证是硬性要求，你必须更正所有的报错，确保最终的ESLint执行结果反馈无错误；第二，目标代码需要满足至少90%的分支覆盖率，这一要求必须通过执行你生成的单元测试文件来实现。","instruction_following_checklist":[{"constraint":"在你的所有回复中，必须以语气词\"哈喽～\"作为开场白","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 去除首尾的空白字符后，检查是否以“哈喽～”开头\n return response.strip().startswith(\"哈喽～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"","validation_code":"import re\n\ndef check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证用户的回复是否在最末尾加上了语气词\"呀～\"。\n \n :param response: 人工智能助手的回复内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: 如果满足要求返回 True，否则返回 False。\n \"\"\"\n if not response:\n return False\n \n # 去除头尾的空白字符\n clean_resp = response.strip()\n \n # 匹配末尾是否为“呀～”或者“呀~”\n # [^\\w]*$ 允许在“呀～”之后跟随任何标点符号（如感叹号、句号等）和空白符号直至结尾\n pattern = r'呀[～~][^\\w]*$'\n \n match = re.search(pattern, clean_resp)\n \n return match is not None","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否不包含“你”字，以验证是否满足禁止用“你”称呼用户的要求。\n \"\"\"\n # 如果回复中出现了“你”字，则说明未满足要求\n if \"你\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己","validation_code":"def check_requirements(response: str, workspace_path: str) -> bool:\n # 检查回复中是否出现了禁止使用的自称词“我”\n if \"我\" in response:\n return False\n \n return True","tags":["Content","Persona"]},{"constraint":"要求在你的解答内出现 😊表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否包含 😊 表情符号。\n \"\"\"\n return \"😊\" in response","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过 4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n \n # 定义网页游戏项目常见的代码文件及记录日志文件的后缀名\n target_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.log'}\n \n # 遍历当前工作空间路径\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果是目标类型的代码或日志文件\n if ext in target_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以严格的 UTF-8 编码读取整个文件内容\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 若抛出 UnicodeDecodeError，则说明存在并非 UTF-8 编码的文件\n return False\n except Exception:\n # 其他非编码相关的异常（如临时读写占用等）予以忽略\n pass\n \n # 如果所有的目标文件均通过了 UTF-8 解码测试\n return True","tags":["Language","File Encoding"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在生成的代码文件中，循环的最大嵌套深度不超过1层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对生成的代码实施ESLint验证是硬性要求，你必须更正所有的报错，确保最终的ESLint执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"目标代码需要满足至少90%的分支覆盖率，这一要求必须通过执行你生成的单元测试文件来实现","validation_code":"","tags":["Workflow","Testing"]}]},{"round_id":4,"instruction":"动画效果很棒！最后一步，我们增加一点游戏深度，加入“连击(Combo)”和“动态得分”系统。在状态栏中间增加分数显示（初始为0分）。得分规则：基础匹配成功得10分；如果连续匹配成功（触发Combo），分数会有加成，比如2连击得20分，3连击得30分，以此类推，并且要在页面上方醒目地弹出一个提示（如“Combo x2!”）并带有一个缓缓向上淡出的动画；如果匹配失败，Combo重置为0，并且当前总分扣除2分（总分不能为负）。最后要求：分数的增加或减少不要直接瞬间变化，必须要有一个数字快速滚动递增/递减的视觉效果。后续所有轮次请你始终遵循以下两个要求：第一，运行你生成的单元测试文件后，目标代码必须具有不低于95%的语句覆盖率；第二，遇到修改已有文件的情况，必须且只能输出严格的Unified Diff格式，禁止采用直接输出全量代码替换的方式。","instruction_following_checklist":[{"constraint":"在你的所有回复中，必须以语气词\"哈喽～\"作为开场白","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证人工智能助手的回复是否以语气词\"哈喽～\"作为开场白。\n \"\"\"\n if not response:\n return False\n \n # 去除字符串开头可能存在的空白字符（包括空格、换行符等）\n cleaned_response = response.lstrip()\n \n # 检查是否以\"哈喽～\"开头\n return cleaned_response.startswith(\"哈喽～\")","tags":["Content","Persona"]},{"constraint":"你对用户作出的任何回复，都需在最末尾加上语气词\"呀～\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n if not response:\n return False\n \n # 清理回复文本末尾可能的空白字符（包括换行、空格等）\n cleaned_response = response.rstrip()\n \n # 判断清理后的回复内容是否以\"呀～\"结尾\n return cleaned_response.endswith(\"呀～\")","tags":["Content","Persona"]},{"constraint":"禁止用\"你\"称呼用户，你的回复中必须一律将其替换为\"亲爱的朋友\"","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断人工智能助手的回复中是否没有包含字符\"你\"\n \"\"\"\n if '你' in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"你的回答里不能出现自称\"我\"，必须统一使用\"小助手\"来代指你自己","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复中是否不包含自称\"我\"。\n \"\"\"\n if \"我\" in response:\n return False\n return True","tags":["Content","Persona"]},{"constraint":"要求在你的解答内出现 😊表情符号","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 判断助手的回复中是否包含指定的 😊表情符号。\n \"\"\"\n return \"😊\" in response","tags":["Content","Keyword"]},{"constraint":"在为用户生成回复时，你必须永远只用中文","validation_code":"","tags":["Language","Response Language"]},{"constraint":"生成的代码文件中，任何函数的圈复杂度都不能超过 4","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"所有由你创建的代码文件，其编码必须统一为UTF-8","validation_code":"def verify(response: str, workspace_path: str) -> bool:\n import os\n\n # 常见代码文件扩展名集合（根据前端游戏项目背景设定，同时囊括其他常见语言）\n code_extensions = {\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \n '.json', '.vue', '.svelte', '.py', '.java', \n '.c', '.cpp', '.h', '.hpp', '.sh', '.xml', \n '.yaml', '.yml'\n }\n\n if not os.path.exists(workspace_path):\n # 如果工作区路径不存在，视作没有文件违反规则\n return True\n\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n # 尝试以 utf-8 编码读取该代码文件\n with open(file_path, 'r', encoding='utf-8') as f:\n f.read()\n except UnicodeDecodeError:\n # 如果抛出 UnicodeDecodeError，说明该文件并非有效的 UTF-8 编码格式\n return False\n except Exception:\n # 忽略其他可能的异常（例如文件权限问题等），仅专注于编码判断\n pass\n\n return True","tags":["Language","File Encoding"]},{"constraint":"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）","validation_code":"","tags":["Environment","File Operation"]},{"constraint":"请确保在生成的代码文件中，循环的最大嵌套深度不超过1层","validation_code":"","tags":["Quantity","Complexity"]},{"constraint":"你必须在代码写完后测试全部的代码，并确保测试成功通过","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则","validation_code":"","tags":["Environment","File Path"]},{"constraint":"对生成的代码实施ESLint验证是硬性要求，你必须更正所有的报错，确保最终的ESLint执行结果反馈无错误","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"运行你生成的单元测试文件后，目标代码必须具有不低于95%的语句覆盖率","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"目标代码需要满足至少90%的分支覆盖率，这一要求必须通过执行你生成的单元测试文件来实现","validation_code":"","tags":["Workflow","Testing"]},{"constraint":"遇到修改已有文件的情况，必须且只能输出严格的Unified Diff格式，禁止采用直接输出全量代码替换的方式","validation_code":"","tags":["Environment","File Operation"]}]}],"string":"[\n {\n \"round_id\": 0,\n \"instruction\": \"帮我写一个网页版的卡牌记忆配对游戏。第一步，我们先搭个基础框架。在页面居中显示一个游戏标题，下方画一个4x4的卡牌网格。每张卡牌先用统一的背景色或背面图案占位。只需实现最基础的交互：点击任何一张卡牌时，卡牌能翻面（现在正面统一显示个纯色或者空白即可，只要能看出正反面状态就行）。暂时先不要写匹配判断逻辑，把网格布局、卡牌组件和基础的点击翻开动作做好就行，为后续增加游戏逻辑留好空间。请你注意，以下两条要求在后续所有轮次中都需要始终遵守：所有由你创建的代码文件，其编码必须统一为UTF-8；请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你的所有回复中，必须以语气词\\\"哈喽～\\\"作为开场白\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复是否以\\\"哈喽～\\\"作为开场白。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址\\n :return: 布尔值，表示是否满足要求\\n \\\"\\\"\\\"\\n # 去除首尾空白字符后判断是否以\\\"哈喽～\\\"开头\\n return response.strip().startswith(\\\"哈喽～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"呀～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断AI助手的回复最末尾是否加上了语气词\\\"呀～\\\"。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除字符串最末尾的换行符和空格等空白字符\\n cleaned_response = response.rstrip()\\n \\n # 判断是否以指定的语气词结尾\\n return cleaned_response.endswith(\\\"呀～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"亲爱的朋友\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查助手的回复中是否禁止使用了“你”。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 布尔值，表示是否满足要求。\\n \\\"\\\"\\\"\\n # 如果回复中存在“你”，说明没有被替换为“亲爱的朋友”，违反了约束\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的回答里不能出现自称\\\"我\\\"，必须统一使用\\\"小助手\\\"来代指你自己\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查助手的回复中是否出现了自称\\\"我\\\"。\\n 如果出现\\\"我\\\"，则返回 False；否则返回 True。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"要求在你的解答内出现 😊表情符号\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 检查助手的回复中是否包含😊表情符号。\\n \\n :param response: 人工智能助手最后一轮回复的内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 如果包含😊则返回True，否则返回False。\\n \\\"\\\"\\\"\\n return '😊' in response\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"import os\\n\\ndef check(response: str, workspace_path: str) -> bool:\\n # 针对网页游戏项目，推断可能的代码文件扩展名\\n code_extensions = {'.html', '.css', '.js'}\\n \\n # 遍历项目目录下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名并转换为小写\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 以二进制模式读取文件内容\\n with open(file_path, 'rb') as f:\\n raw_data = f.read()\\n # 尝试使用utf-8进行解码\\n raw_data.decode('utf-8')\\n except UnicodeDecodeError:\\n # 如果抛出UnicodeDecodeError，则说明不是有效的UTF-8编码\\n return False\\n except Exception:\\n # 如果有其他读取错误（如权限问题），也视为不满足预期\\n return False\\n \\n # 所有代码文件都能成功以UTF-8解码\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n },\n {\n \"round_id\": 1,\n \"instruction\": \"很好，基础框架有了。现在来完善核心的记忆配对逻辑。请准备8对不同的Emoji表情作为卡牌的正面内容，每次刷新或游戏开始时，随机打乱分配到这16个格子里。规则是：玩家每次最多只能翻开两张牌，如果这两张牌的Emoji一样，它们就保持翻开状态（可以稍微变暗表示已完成）；如果不一样，大概停留1秒钟后自动翻转盖回去。注意一个关键挑战：在两张牌翻开且还在展示的那1秒延迟内，必须锁定整个网格，禁止玩家点击其他未翻开的牌，防止出现逻辑错乱。后续所有轮次请你始终遵循以下两个要求：请确保在生成的代码文件中，循环的最大嵌套深度不超过2层；你必须在代码写完后测试全部的代码，并确保测试成功通过。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你的所有回复中，必须以语气词\\\"哈喽～\\\"作为开场白\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否以语气词\\\"哈喽～\\\"作为开场白。\\n \\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 当前人工智能助手生成的代码项目地址（此项验证中不使用）\\n :return: 布尔值，代表该要求的满足情况\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除开头和结尾可能的空白字符后，判断是否以\\\"哈喽～\\\"开头\\n return response.strip().startswith(\\\"哈喽～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"呀～\\\"\",\n \"validation_code\": \"def verify_ending_phrase(response: str, workspace_path: str) -> bool:\\n import re\\n \\n # 去除字符串首尾的空白字符，直接判断\\n if response.strip().endswith(\\\"呀～\\\"):\\n return True\\n \\n # 为了防止助手在\\\"呀～\\\"后面加上了标点符号（如\\\"呀～！\\\", \\\"呀～。\\\"等）\\n # 使用正则表达式去除末尾可能出现的空白字符以及常见中英文句末标点符号\\n cleaned_response = re.sub(r'[\\\\s。！？\\\\!\\\\?\\\\.]+$', '', response)\\n \\n return cleaned_response.endswith(\\\"呀～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"亲爱的朋友\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证回复中是否禁止使用了“你”称呼用户。\\n :param response: 人工智能助手最后一轮回复的内容\\n :param workspace_path: 代码项目地址\\n :return: 布尔值，代表是否满足要求\\n \\\"\\\"\\\"\\n # 只要回复中出现了“你”，即认为违反了该要求\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的回答里不能出现自称\\\"我\\\"，必须统一使用\\\"小助手\\\"来代指你自己\",\n \"validation_code\": \"def check_condition(response: str, workspace_path: str) -> bool:\\n if '我' in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"要求在你的解答内出现 😊表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断 AI 助手的回复中是否包含 😊 表情符号\\n \\\"\\\"\\\"\\n return \\\"😊\\\" in response\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 检查工作空间路径是否存在\\n if not os.path.exists(workspace_path):\\n return False\\n \\n # 根据网页项目推断可能的代码文件扩展名\\n code_extensions = {'.html', '.htm', '.css', '.js', '.py', '.json'}\\n \\n # 遍历工作空间下的所有文件\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n # 获取文件扩展名并转为小写\\n ext = os.path.splitext(file)[1].lower()\\n \\n # 判断是否属于代码文件\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 只要有一个代码文件无法用 utf-8 解码，说明不满足要求\\n return False\\n except Exception:\\n # 发生其他读取异常，出于严谨也视作无法验证编码正确性\\n return False\\n \\n # 所有代码文件均成功使用 utf-8 编码读取\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，循环的最大嵌套深度不超过2层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 2,\n \"instruction\": \"匹配逻辑跑通了。下一步给游戏加上计分板和结算系统。在标题和卡牌网格之间增加一个状态栏，左侧显示当前尝试的“步数”（翻开两张牌算完整的一步），右侧显示“用时”（从玩家点击第一张牌那一刻起开始计时，格式为 mm:ss）。当所有8对卡牌都成功配对后，停止计时，并在页面正中间弹出一个“游戏通关”的模态框，里面要展示玩家的总用时和总步数，弹窗底部提供一个“再来一局”的按钮。点击该按钮后，步数和时间清零，弹窗关闭，所有卡牌重新盖上并重新洗牌。后续所有轮次请你始终遵循以下两个要求：第一，请确保在生成的代码文件中，循环的最大嵌套深度不超过1层；第二，对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你的所有回复中，必须以语气词\\\"哈喽～\\\"作为开场白\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 清除前后可能存在的空白字符，并判断是否以指定的开场白开头\\n return response.strip().startswith(\\\"哈喽～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"呀～\\\"\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n # 去除末尾的空白字符和换行符，并判断是否以指定语气词结尾\\n return response.strip().endswith(\\\"呀～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"亲爱的朋友\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的最后一轮回复是否满足：禁止用\\\"你\\\"称呼用户。\\n \\\"\\\"\\\"\\n if '你' in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的回答里不能出现自称\\\"我\\\"，必须统一使用\\\"小助手\\\"来代指你自己\",\n \"validation_code\": \"def check(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否包含“我”\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"要求在你的解答内出现 😊表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n # 检查助手的回复中是否包含 \\\"😊\\\" 表情符号\\n return \\\"😊\\\" in response\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 验证工作区路径是否存在\\n if not os.path.exists(workspace_path) or not os.path.isdir(workspace_path):\\n return False\\n \\n # 定义常见的代码文件扩展名（结合前端小游戏和常见编程语言）\\n code_extensions = {\\n '.html', '.htm', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.json', '.xml', '.py', '.java', '.c', '.cpp', '.h', \\n '.hpp', '.cs', '.go', '.rs', '.rb', '.php', '.swift', \\n '.kt', '.sh', '.bat', '.vue'\\n }\\n \\n code_files_found = False\\n \\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n code_files_found = True\\n filepath = os.path.join(root, file)\\n try:\\n # 尝试用 utf-8 编码去读取文件\\n with open(filepath, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 抛出 UnicodeDecodeError 说明含有非 UTF-8 的字符（如 GBK 编码的中文）\\n return False\\n except Exception:\\n # 如果遇到其他由于文件权限等导致的异常暂且忽略，以解码验证为准\\n pass\\n \\n # 如果没有找到任何代码文件，说明助手未能按要求生成代码，返回 False\\n return code_files_found\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，循环的最大嵌套深度不超过1层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n }\n ]\n },\n {\n \"round_id\": 3,\n \"instruction\": \"现在的游戏已经很完整了，咱们来提升一下视觉体验和动画效果，稍微增加点前端挑战。首先，修改卡牌的翻转动画，要求实现带透视感的真实3D翻牌效果（绕Y轴平滑旋转180度，正面和背面内容分别附着在正反面上）。其次，加上状态反馈动画：当玩家翻开的两张牌配对成功时，让这两张牌做一个明显的“放大并闪烁”的庆祝动画；如果配对失败，在这两张牌翻盖回去之前，给它们加一个“左右剧烈摇晃”的错误提示动画。后续所有轮次请你始终遵循以下两个要求：第一，对生成的代码实施ESLint验证是硬性要求，你必须更正所有的报错，确保最终的ESLint执行结果反馈无错误；第二，目标代码需要满足至少90%的分支覆盖率，这一要求必须通过执行你生成的单元测试文件来实现。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你的所有回复中，必须以语气词\\\"哈喽～\\\"作为开场白\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 去除首尾的空白字符后，检查是否以“哈喽～”开头\\n return response.strip().startswith(\\\"哈喽～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"呀～\\\"\",\n \"validation_code\": \"import re\\n\\ndef check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证用户的回复是否在最末尾加上了语气词\\\"呀～\\\"。\\n \\n :param response: 人工智能助手的回复内容。\\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\\n :return: 如果满足要求返回 True，否则返回 False。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除头尾的空白字符\\n clean_resp = response.strip()\\n \\n # 匹配末尾是否为“呀～”或者“呀~”\\n # [^\\\\w]*$ 允许在“呀～”之后跟随任何标点符号（如感叹号、句号等）和空白符号直至结尾\\n pattern = r'呀[～~][^\\\\w]*$'\\n \\n match = re.search(pattern, clean_resp)\\n \\n return match is not None\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"亲爱的朋友\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否不包含“你”字，以验证是否满足禁止用“你”称呼用户的要求。\\n \\\"\\\"\\\"\\n # 如果回复中出现了“你”字，则说明未满足要求\\n if \\\"你\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的回答里不能出现自称\\\"我\\\"，必须统一使用\\\"小助手\\\"来代指你自己\",\n \"validation_code\": \"def check_requirements(response: str, workspace_path: str) -> bool:\\n # 检查回复中是否出现了禁止使用的自称词“我”\\n if \\\"我\\\" in response:\\n return False\\n \\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"要求在你的解答内出现 😊表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否包含 😊 表情符号。\\n \\\"\\\"\\\"\\n return \\\"😊\\\" in response\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n \\n # 定义网页游戏项目常见的代码文件及记录日志文件的后缀名\\n target_extensions = {'.html', '.css', '.js', '.ts', '.jsx', '.tsx', '.json', '.log'}\\n \\n # 遍历当前工作空间路径\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n # 如果是目标类型的代码或日志文件\\n if ext in target_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以严格的 UTF-8 编码读取整个文件内容\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 若抛出 UnicodeDecodeError，则说明存在并非 UTF-8 编码的文件\\n return False\\n except Exception:\\n # 其他非编码相关的异常（如临时读写占用等）予以忽略\\n pass\\n \\n # 如果所有的目标文件均通过了 UTF-8 解码测试\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，循环的最大嵌套深度不超过1层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施ESLint验证是硬性要求，你必须更正所有的报错，确保最终的ESLint执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"目标代码需要满足至少90%的分支覆盖率，这一要求必须通过执行你生成的单元测试文件来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n }\n ]\n },\n {\n \"round_id\": 4,\n \"instruction\": \"动画效果很棒！最后一步，我们增加一点游戏深度，加入“连击(Combo)”和“动态得分”系统。在状态栏中间增加分数显示（初始为0分）。得分规则：基础匹配成功得10分；如果连续匹配成功（触发Combo），分数会有加成，比如2连击得20分，3连击得30分，以此类推，并且要在页面上方醒目地弹出一个提示（如“Combo x2!”）并带有一个缓缓向上淡出的动画；如果匹配失败，Combo重置为0，并且当前总分扣除2分（总分不能为负）。最后要求：分数的增加或减少不要直接瞬间变化，必须要有一个数字快速滚动递增/递减的视觉效果。后续所有轮次请你始终遵循以下两个要求：第一，运行你生成的单元测试文件后，目标代码必须具有不低于95%的语句覆盖率；第二，遇到修改已有文件的情况，必须且只能输出严格的Unified Diff格式，禁止采用直接输出全量代码替换的方式。\",\n \"instruction_following_checklist\": [\n {\n \"constraint\": \"在你的所有回复中，必须以语气词\\\"哈喽～\\\"作为开场白\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 验证人工智能助手的回复是否以语气词\\\"哈喽～\\\"作为开场白。\\n \\\"\\\"\\\"\\n if not response:\\n return False\\n \\n # 去除字符串开头可能存在的空白字符（包括空格、换行符等）\\n cleaned_response = response.lstrip()\\n \\n # 检查是否以\\\"哈喽～\\\"开头\\n return cleaned_response.startswith(\\\"哈喽～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你对用户作出的任何回复，都需在最末尾加上语气词\\\"呀～\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n if not response:\\n return False\\n \\n # 清理回复文本末尾可能的空白字符（包括换行、空格等）\\n cleaned_response = response.rstrip()\\n \\n # 判断清理后的回复内容是否以\\\"呀～\\\"结尾\\n return cleaned_response.endswith(\\\"呀～\\\")\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"禁止用\\\"你\\\"称呼用户，你的回复中必须一律将其替换为\\\"亲爱的朋友\\\"\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断人工智能助手的回复中是否没有包含字符\\\"你\\\"\\n \\\"\\\"\\\"\\n if '你' in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"你的回答里不能出现自称\\\"我\\\"，必须统一使用\\\"小助手\\\"来代指你自己\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复中是否不包含自称\\\"我\\\"。\\n \\\"\\\"\\\"\\n if \\\"我\\\" in response:\\n return False\\n return True\",\n \"tags\": [\n \"Content\",\n \"Persona\"\n ]\n },\n {\n \"constraint\": \"要求在你的解答内出现 😊表情符号\",\n \"validation_code\": \"def check_requirement(response: str, workspace_path: str) -> bool:\\n \\\"\\\"\\\"\\n 判断助手的回复中是否包含指定的 😊表情符号。\\n \\\"\\\"\\\"\\n return \\\"😊\\\" in response\",\n \"tags\": [\n \"Content\",\n \"Keyword\"\n ]\n },\n {\n \"constraint\": \"在为用户生成回复时，你必须永远只用中文\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Language\",\n \"Response Language\"\n ]\n },\n {\n \"constraint\": \"生成的代码文件中，任何函数的圈复杂度都不能超过 4\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"所有由你创建的代码文件，其编码必须统一为UTF-8\",\n \"validation_code\": \"def verify(response: str, workspace_path: str) -> bool:\\n import os\\n\\n # 常见代码文件扩展名集合（根据前端游戏项目背景设定，同时囊括其他常见语言）\\n code_extensions = {\\n '.html', '.css', '.js', '.ts', '.jsx', '.tsx', \\n '.json', '.vue', '.svelte', '.py', '.java', \\n '.c', '.cpp', '.h', '.hpp', '.sh', '.xml', \\n '.yaml', '.yml'\\n }\\n\\n if not os.path.exists(workspace_path):\\n # 如果工作区路径不存在，视作没有文件违反规则\\n return True\\n\\n for root, dirs, files in os.walk(workspace_path):\\n for file in files:\\n ext = os.path.splitext(file)[1].lower()\\n if ext in code_extensions:\\n file_path = os.path.join(root, file)\\n try:\\n # 尝试以 utf-8 编码读取该代码文件\\n with open(file_path, 'r', encoding='utf-8') as f:\\n f.read()\\n except UnicodeDecodeError:\\n # 如果抛出 UnicodeDecodeError，说明该文件并非有效的 UTF-8 编码格式\\n return False\\n except Exception:\\n # 忽略其他可能的异常（例如文件权限问题等），仅专注于编码判断\\n pass\\n\\n return True\",\n \"tags\": [\n \"Language\",\n \"File Encoding\"\n ]\n },\n {\n \"constraint\": \"请在每次改动项目后，生成一个全新的日志文件 change_log_X.log 以记录该次修改（X 指代改动序号，初始值为 0）\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n },\n {\n \"constraint\": \"请确保在生成的代码文件中，循环的最大嵌套深度不超过1层\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Quantity\",\n \"Complexity\"\n ]\n },\n {\n \"constraint\": \"你必须在代码写完后测试全部的代码，并确保测试成功通过\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"对于生成的代码文件，其内部的文件引用需严格遵守使用相对路径、摒弃绝对路径的原则\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Path\"\n ]\n },\n {\n \"constraint\": \"对生成的代码实施ESLint验证是硬性要求，你必须更正所有的报错，确保最终的ESLint执行结果反馈无错误\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"运行你生成的单元测试文件后，目标代码必须具有不低于95%的语句覆盖率\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"目标代码需要满足至少90%的分支覆盖率，这一要求必须通过执行你生成的单元测试文件来实现\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Workflow\",\n \"Testing\"\n ]\n },\n {\n \"constraint\": \"遇到修改已有文件的情况，必须且只能输出严格的Unified Diff格式，禁止采用直接输出全量代码替换的方式\",\n \"validation_code\": \"\",\n \"tags\": [\n \"Environment\",\n \"File Operation\"\n ]\n }\n ]\n }\n]","isConversational":false},"function_checklist":{"kind":"list like","value":["页面加载后，检查页面居中是否显示游戏标题，下方是否有一个4x4的卡牌网格，且16张卡牌初始均呈现统一的背面图案或背景色。","检查标题与网格之间是否有状态栏，状态栏左侧显示'步数'为0，中间显示'分数'为0，右侧显示'用时'（如00:00）。","点击任意第一张卡牌，确认卡牌能翻面显示正面的Emoji表情，并且状态栏右侧的'用时'从这一刻起开始按秒走动计时。","在卡牌翻转的瞬间进行截图，观察并确认翻牌动作实现了带透视感的真实3D翻面效果（绕Y轴旋转180度）。","翻开第二张与第一张Emoji不同的卡牌（配对失败），确认状态栏左侧的'步数'增加1，同时由于初始分数为0且不能为负，'分数'依然保持为0。","观察刚翻开的两张不同的卡牌，确认它们在翻开后会触发一个'左右剧烈摇晃'的错误提示动画，随后大概停留1秒钟后自动翻转盖回背面。","测试防抖锁定功能：快速连续点击三张背面朝上的卡牌，确认只有前两张牌被翻开，第三张牌在前两张牌盖回去之前无法响应点击并翻开。","通过多次尝试找到两张Emoji相同的卡牌并依次点击（配对成功），确认这两张卡牌触发'放大并闪烁'的庆祝动画，随后保持翻开状态并且视觉上变暗以表示已完成。","观察上述首次配对成功后的分数变化，确认总分增加10分，并且分数的增加伴随着数字快速滚动递增的视觉动画效果。","紧接着上一次成功，再次成功配对一对相同的卡牌（触发连击），确认页面上方醒目地弹出了连击提示文字（如'Combo x2'），并且该提示带有缓缓向上淡出的动画。","确认触发2连击后，状态栏内的总分以滚动递增的方式增加了20分（例如从10分变为了30分）。","在拥有分数且连击未断的情况下，故意翻开两张不同的卡牌（配对失败），确认此时总分数滚动递减扣除2分。之后再次成功配对一对卡牌时，确认仅获得基础的10分，证明Combo已被正确重置。","不断进行配对直到所有8对卡牌全被成功翻开，确认最后一张牌翻开后计时立刻停止，并且在页面正中间弹出了一个'游戏通关'的模态框。","检查'游戏通关'模态框的内容，确认里面正确展示了本局游戏最终的'总用时'和'总步数'数据，并且弹窗底部包含一个'再来一局'的按钮。","点击模态框的「再来一局」按钮，确认弹窗关闭，状态栏的步数、时间和分数全部清零，所有的16张卡牌恢复背面盖上的状态，并重新打乱洗牌；可通过多张卡牌位置或调试数据确认新一局的卡牌顺序经过重新随机分配。","通过完整翻开或调试数据检查16张卡牌的正面内容，确认一共包含8种不同的Emoji表情，且每种Emoji正好出现2次，构成8对可配对卡牌。","点击第一张卡牌后，确认步数仍保持不变；再点击第二张卡牌形成一次完整尝试后，确认步数增加1，验证“翻开两张牌算完整一步”的规则。","成功配对两张相同Emoji卡牌后，等待一段时间并继续点击其他卡牌，确认这两张已配对卡牌始终保持翻开状态，不会自动盖回去。","在总分为0或接近0时连续进行多次错误配对，确认每次失败不会让总分低于0，分数最低保持为0。"],"string":"[\n \"页面加载后，检查页面居中是否显示游戏标题，下方是否有一个4x4的卡牌网格，且16张卡牌初始均呈现统一的背面图案或背景色。\",\n \"检查标题与网格之间是否有状态栏，状态栏左侧显示'步数'为0，中间显示'分数'为0，右侧显示'用时'（如00:00）。\",\n \"点击任意第一张卡牌，确认卡牌能翻面显示正面的Emoji表情，并且状态栏右侧的'用时'从这一刻起开始按秒走动计时。\",\n \"在卡牌翻转的瞬间进行截图，观察并确认翻牌动作实现了带透视感的真实3D翻面效果（绕Y轴旋转180度）。\",\n \"翻开第二张与第一张Emoji不同的卡牌（配对失败），确认状态栏左侧的'步数'增加1，同时由于初始分数为0且不能为负，'分数'依然保持为0。\",\n \"观察刚翻开的两张不同的卡牌，确认它们在翻开后会触发一个'左右剧烈摇晃'的错误提示动画，随后大概停留1秒钟后自动翻转盖回背面。\",\n \"测试防抖锁定功能：快速连续点击三张背面朝上的卡牌，确认只有前两张牌被翻开，第三张牌在前两张牌盖回去之前无法响应点击并翻开。\",\n \"通过多次尝试找到两张Emoji相同的卡牌并依次点击（配对成功），确认这两张卡牌触发'放大并闪烁'的庆祝动画，随后保持翻开状态并且视觉上变暗以表示已完成。\",\n \"观察上述首次配对成功后的分数变化，确认总分增加10分，并且分数的增加伴随着数字快速滚动递增的视觉动画效果。\",\n \"紧接着上一次成功，再次成功配对一对相同的卡牌（触发连击），确认页面上方醒目地弹出了连击提示文字（如'Combo x2'），并且该提示带有缓缓向上淡出的动画。\",\n \"确认触发2连击后，状态栏内的总分以滚动递增的方式增加了20分（例如从10分变为了30分）。\",\n \"在拥有分数且连击未断的情况下，故意翻开两张不同的卡牌（配对失败），确认此时总分数滚动递减扣除2分。之后再次成功配对一对卡牌时，确认仅获得基础的10分，证明Combo已被正确重置。\",\n \"不断进行配对直到所有8对卡牌全被成功翻开，确认最后一张牌翻开后计时立刻停止，并且在页面正中间弹出了一个'游戏通关'的模态框。\",\n \"检查'游戏通关'模态框的内容，确认里面正确展示了本局游戏最终的'总用时'和'总步数'数据，并且弹窗底部包含一个'再来一局'的按钮。\",\n \"点击模态框的「再来一局」按钮，确认弹窗关闭，状态栏的步数、时间和分数全部清零，所有的16张卡牌恢复背面盖上的状态，并重新打乱洗牌；可通过多张卡牌位置或调试数据确认新一局的卡牌顺序经过重新随机分配。\",\n \"通过完整翻开或调试数据检查16张卡牌的正面内容，确认一共包含8种不同的Emoji表情，且每种Emoji正好出现2次，构成8对可配对卡牌。\",\n \"点击第一张卡牌后，确认步数仍保持不变；再点击第二张卡牌形成一次完整尝试后，确认步数增加1，验证“翻开两张牌算完整一步”的规则。\",\n \"成功配对两张相同Emoji卡牌后，等待一段时间并继续点击其他卡牌，确认这两张已配对卡牌始终保持翻开状态，不会自动盖回去。\",\n \"在总分为0或接近0时连续进行多次错误配对，确认每次失败不会让总分低于0，分数最低保持为0。\"\n]","isConversational":false}}},{"rowIdx":90,"cells":{"task_id":{"kind":"number","value":91,"string":"91"},"repository_policy":{"kind":"string","value":"后续所有交互中你需要始终遵循以下规则：\n1. 每次给用户生成回复时，必须首先输出\"喵～\"这一语气词。\n2. 无论何时提到自身，都请始终用\"代码小灵\"替代\"我\"来进行自我指代。\n3. 凡是需要称呼用户的地方，均须严格使用\"开发者大大\"，不得出现以\"你\"相称的情况。\n4. 请确保在所有情况下都使用中文来生成用户回复。\n5. 无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码。\n6. 你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”。\n7. 生成的代码文件必须满足：单个函数的规模不得大于 60 行。\n8. 你的所有回复都必须在结尾处加上\"哒～\"语气词。"},"repository_policy_checklist":{"kind":"list like","value":[{"constraint":"每次给用户生成回复时，必须首先输出\"喵～\"这一语气词","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 验证AI助手的回复是否首先输出\"喵～\"这一语气词。\n \"\"\"\n if not response:\n return False\n \n # 去除首尾的空白字符后，检查是否以\"喵～\"开头\n return response.strip().startswith(\"喵～\")","tags":["Content","Persona"]},{"constraint":"无论何时提到自身，都请始终用\"代码小灵\"替代\"我\"来进行自我指代","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n \"\"\"\n 检查人工智能助手的回复是否满足要求：\n 无论何时提到自身，都请始终用\"代码小灵\"替代\"我\"来进行自我指代。\n \n :param response: 人工智能助手最后一轮回复的内容。\n :param workspace_path: 当前人工智能助手生成的代码项目地址。\n :return: bool，代表是否满足要求。\n \"\"\"\n # 如果回复中出现了\"我\"字，则说明没有用\"代码小灵\"完全替代\"我\"，直接返回False\n if \"我\" in response:\n return False\n \n # 如果没有出现\"我\"字，则视为遵守了替代自称的规则\n return True","tags":["Content","Persona"]},{"constraint":"凡是需要称呼用户的地方，均须严格使用\"开发者大大\"，不得出现以\"你\"相称的情况","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 辅助函数：判断文本中是否包含第二人称代词“你”或“您”\n def has_forbidden_pronouns(text: str) -> bool:\n if not text:\n return False\n # 排除包含“你”字但并非称呼用户的常见词汇（如音译词）\n safe_text = text.replace(\"迷你\", \"\").replace(\"迪士尼\", \"\").replace(\"维尼\", \"\")\n \n # 如果排除了非代词词汇后，文本中依然存在“你”或“您”，说明存在违规称呼\n if \"你\" in safe_text or \"您\" in safe_text:\n return True\n return False\n\n # 1. 检查人工智能助手的直接回复中是否包含“你”或“您”\n if has_forbidden_pronouns(response):\n return False\n\n # 2. 检查生成的代码项目中（如代码注释、输出文本等）是否包含“你”或“您”\n if os.path.exists(workspace_path) and os.path.isdir(workspace_path):\n for root, dirs, files in os.walk(workspace_path):\n for file in files:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n content = f.read()\n if has_forbidden_pronouns(content):\n return False\n except Exception:\n # 如果某些文件（如二进制文件或编码不支持）无法读取，则跳过\n continue\n\n # 如果所有文本均未发现违规的称呼，则判定为满足要求\n return True","tags":["Content","Persona"]},{"constraint":"请确保在所有情况下都使用中文来生成用户回复","validation_code":"","tags":["Language","Response Language"]},{"constraint":"无论生成哪个代码文件，其起始第一行都必须严格添加 -*- coding: utf-8 -*- 以声明编码","validation_code":"def check_encoding_declaration(response: str, workspace_path: str) -> bool:\n import os\n\n # 若工作区路径不存在，说明未生成任何代码文件，未违反要求\n if not os.path.exists(workspace_path):\n return True\n\n # 推断常见的代码文件扩展名\n code_extensions = {\n '.py', '.java', '.c', '.cpp', '.cc', '.h', '.hpp', '.cs', \n '.js', '.ts', '.html', '.css', '.go', '.rb', '.php', '.sh', \n '.rs', '.swift', '.kt', '.m'\n }\n\n # 遍历工作区目录中的所有文件\n for root, _, files in os.walk(workspace_path):\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 仅对判断为代码类型的文件进行检查\n if ext in code_extensions:\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n first_line = f.readline()\n # 检查第一行是否包含指定的编码声明字符串\n if '-*- coding: utf-8 -*-' not in first_line:\n return False\n except Exception:\n # 如果遇到编码错误或无法读取，视为不符合要求\n return False\n\n return True","tags":["Content","Format"]},{"constraint":"你所生成的任何一个代码文件，其开头处都需另起一行进行清晰的注释声明：“AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.”","validation_code":"def check_requirement(response: str, workspace_path: str) -> bool:\n import os\n\n # 题目要求的指定声明字符串\n target_statement = \"AI-Assisted: Initial draft generated by Claude Code; reviewed and modified by author.\"\n \n # 定义常见的代码文件扩展名\n code_extensions = {\n '.py', '.java', '.js', '.ts', '.c', '.cpp', '.cc', '.h', '.hpp', \n '.cs', '.go', '.rs', '.php', '.rb', '.swift', '.kt', '.scala', \n '.html', '.css', '.sh', '.m', '.sql'\n }\n \n has_code_file = False\n \n for root, dirs, files in os.walk(workspace_path):\n # 排除隐藏的目录，例如 .git 等环境产生的文件夹\n dirs[:] = [d for d in dirs if not d.startswith('.')]\n for file in files:\n ext = os.path.splitext(file)[1].lower()\n # 如果判定为代码文件\n if ext in code_extensions:\n has_code_file = True\n file_path = os.path.join(root, file)\n try:\n with open(file_path, 'r', encoding='utf-8') as f:\n lines = []\n # 仅提取前10行作为文件的“开头处”进行检查\n for _ in range(10):\n line = f.readline()\n if not line:\n break\n lines.append(line)\n \n # 检查是否包含目标声明\n found_statement = False\n for line in lines:\n if target_statement in line:\n found_statement = True\n break\n \n # 如果某一个代码文件未包含该声明，则不满足要求\n if not found_statement:\n return False\n except Exception:\n # 若文件编码异常读取失败，视为未满足\n return False\n \n # 如果项目中没有生成任何代码文件，则认为未完成对应代码生成任务；否则全部符合返回 True\n return has_code_file","tags":["Content","Format"]},{"constraint":"生成的代码文件必须满足：单个函数的规模不得大于 60 行","validation_code":"