{"input": "设计一个跨平台内容自动化发布系统，需要同时支持微博、微信公众号、抖音、小红书和视频号。请给出核心架构方案，并说明如何处理各平台 API 差异与限制。", "expected_output": {"architecture": "Adapter + Pipeline + State Machine", "adapters": [{"platform": "weibo", "type": "api", "notes": "OAuth2.0, 日限企业100/个人30"}, {"platform": "wechat", "type": "api", "notes": "access_token, draft/add + freepublish/submit"}, {"platform": "douyin", "type": "api|rpa", "notes": "企业用官方API，个人 fallback 到 RPA"}, {"platform": "xiaohongshu", "type": "rpa", "notes": "无官方API，Playwright/Selenium 模拟"}, {"platform": "video_channel", "type": "rpa", "notes": "无官方API，tencent_uploader 或真机RPA"}], "pipeline": ["ingest", "preprocess", "moderation", "queue", "publish", "track"], "key_principle": "Configuration-as-Code, 每个Adapter统一接口"}, "dimensions": {"platform_adaptation": 15, "content_preprocessing": 13, "publish_execution": 14, "state_tracking_retry": 14, "compliance_risk_control": 13, "overall_design": 24}, "score": 93, "grader_hints": "架构完整，覆盖5平台且区分API/RPA，适配逻辑清晰。缺详细的去重策略和AIGC标识注入细节，扣2分预处理+1分风控。综合设计优秀但文档缺少部署流程扣1分。"}
{"input": "给定一篇原始文章：标题为'2026年AI内容创作趋势深度解析：从工具到生态的进化之路'（26字），正文3000字，配图12张。请输出该内容在微博、小红书、微信公众号三个平台的适配版本。", "expected_output": {"weibo": {"title_used_as_content": "2026年AI内容创作趋势深度解析：从工具到生态的进化之路", "text_length": 26, "image_count": 9, "note": "微博标题即正文，12张图超限，取前9张"}, "xiaohongshu": {"title": "2026年AI内容创作趋势解析", "title_length": 14, "body_summary": "300字，分5段，短句+Emoji", "image_count": 9, "image_size": "1080x1440", "note": "标题裁剪至20字内，正文摘要适配小红书阅读习惯"}, "wechat": {"title": "2026年AI内容创作趋势深度解析：从工具到生态的进化之路", "title_length": 26, "body_html": "3000字完整保留", "image_count": 12, "cover": "永久素材thumb_media_id", "note": "公众号支持长图文，全部保留"}}, "dimensions": {"platform_adaptation": 15, "content_preprocessing": 12, "publish_execution": 10, "state_tracking_retry": 8, "compliance_risk_control": 10, "overall_design": 12}, "score": 67, "grader_hints": "平台适配非常准确，裁剪规则正确；但场景仅为内容适配，不涉及发布执行、状态追踪等维度，因此后几项按基础表现打分。此题为内容适配专项。"}
{"input": "某品牌需要使用你的系统每天向抖音、小红书各发布50条带货短视频。请设计一套内容预审与风控方案，确保不被平台限流或封号。", "expected_output": {"risk_analysis": {"douyin": "企业API日限500-10000，50条/天可接受；但AIGC内容需显著标识，重复率>30%触发限流", "xiaohongshu": "无官方API，RPA风险极高；严禁第三方工具发货推广，交易导流永久封号"}, "preprocess": {"dedup_strategy": ["修改MD5", "调整帧率", "添加0.5秒随机转场", "加载3dB环境底噪"], "copy_rotation": "每5条更换一套文案模板", "image_dedup": "轻微裁剪+饱和度调整"}, "moderation": {"fast_filter": "DFA/Trie敏感词过滤<10ms", "ai_review": "BERT语义分析+NSFW检测~200ms", "disposition": "A级拦截/B-C级人工复核/D-E级通过"}, "behavior_simulation": {"random_delay": "30-120秒随机间隔", "错峰": "按平台黄金时段分布", "ip_isolation": "账号绑定不同代理"}, "aigc_label": {"douyin": "强制显著标识", "xiaohongshu": "强制双标识"}}, "dimensions": {"platform_adaptation": 13, "content_preprocessing": 15, "publish_execution": 11, "state_tracking_retry": 10, "compliance_risk_control": 15, "overall_design": 20}, "score": 84, "grader_hints": "风控方案全面，去重、AIGC标识、行为模拟、敏感词过滤都覆盖。但发布执行和状态追踪在本题中只做了基础描述，未给出重试或状态机设计，扣4分。"}
{"input": "系统在向微信发布文章时收到返回错误码400（Bad Request），错误信息指出'thumb_media_id不存在'。请说明你的系统会如何处理这一错误。", "expected_output": {"error_classification": "永久失败（参数非法）", "immediate_action": "停止重试，记录错误明细", "root_cause": "封面图未正确上传为永久素材，或thumb_media_id已过期", "fix_guide": ["1) 检查media/uploadimg是否成功并返回URL", "2) 封面图需通过新增永久素材接口获取thumb_media_id", "3) 确认thumb_media_id在draft/add中正确传递"], "status_transition": "PUBLISHING -> FAILED_PERMANENT", "notification": "告警通知运营人员人工排查", "retry_policy": "不再重试"}, "dimensions": {"platform_adaptation": 12, "content_preprocessing": 10, "publish_execution": 13, "state_tracking_retry": 14, "compliance_risk_control": 10, "overall_design": 16}, "score": 75, "grader_hints": "错误分类准确，状态机处理正确，但修复指引可以更详细（提到IP白名单等常见原因），平台适配扣2分是因为未检查封面图规格合规性。"}
{"input": "请描述一个面向多平台自动化发布的 PublishTask 状态机设计，包括所有可能的状态、状态转移规则，以及相应的重试策略。", "expected_output": {"states": ["PENDING", "QUEUED", "PUBLISHING", "PUBLISHED", "FAILED_RETRYABLE", "FAILED_PERMANENT", "CANCELLED"], "transitions": [{"from": "PENDING", "to": "QUEUED", "trigger": "入队"}, {"from": "QUEUED", "to": "PUBLISHING", "trigger": "Adapter开始执行"}, {"from": "PUBLISHING", "to": "PUBLISHED", "trigger": "平台返回成功"}, {"from": "PUBLISHING", "to": "FAILED_RETRYABLE", "trigger": "429/500/502/503/504"}, {"from": "FAILED_RETRYABLE", "to": "QUEUED", "trigger": "指数退避后重试"}, {"from": "FAILED_RETRYABLE", "to": "FAILED_PERMANENT", "trigger": "达到max_attempts"}, {"from": "PUBLISHING", "to": "FAILED_PERMANENT", "trigger": "400/401/风控封号"}, {"from": "QUEUED", "to": "CANCELLED", "trigger": "用户取消"}], "retry_policy": {"max_attempts": 5, "backoff": "exponential + jitter", "retryable_errors": [429, 500, 502, 503, 504], "permanent_errors": [400, 401], "dlq": "超过max_attempts的FAILED_RETRYABLE也转DLQ"}, "data_model": {"task_id": "string", "post_id": "string", "platform": "enum", "status": "enum", "attempt_count": "int", "next_retry_at": "datetime|null", "platform_post_id": "string|null", "platform_url": "string|null", "last_error_code": "string|null", "last_error_message": "string|null"}}, "dimensions": {"platform_adaptation": 10, "content_preprocessing": 8, "publish_execution": 13, "state_tracking_retry": 15, "compliance_risk_control": 10, "overall_design": 23}, "score": 79, "grader_hints": "状态机设计完整且专业，持久化数据模型详细。但题目偏重状态机，其他维度自然得分略低。未提到DLQ的人工复核流程扣1分综合。"}
{"input": "一家公司使用了你的发布系统，昨晚系统向微博发布内容时连续收到429错误。第二天早上，运营人员发现该账号被限流。请分析可能的原因并给出系统改进方案。", "expected_output": {"root_causes": ["1) 发布频率过高，未错峰，触发微博 rate limit", "2) 多条内容文案/图片重复率高，被判定为营销号", "3) 未声明AIGC标识，被用户投诉", "4) 429错误未正确读取Retry-After头部，持续重试加剧限流"], "improvements": {"publish_execution": "严格读取Retry-After，429退避后重试，日限接近时暂停入队", "content_preprocessing": ["增加文案模板库轮换", "图片轻度去重处理", "视频MD5扰动"], "compliance": "所有AI生成内容自动添加AIGC声明", "behavior_simulation": "错峰发布，避免集中在同一时段", "account_monitoring": "发布前检查账号健康状态，限流标记时自动停机保护"}, "status_handling": "429 -> FAILED_RETRYABLE -> 按Retry-After延迟重试 -> 如账号状态异常转FAILED_PERMANENT并告警"}, "dimensions": {"platform_adaptation": 13, "content_preprocessing": 14, "publish_execution": 14, "state_tracking_retry": 13, "compliance_risk_control": 14, "overall_design": 20}, "score": 88, "grader_hints": "分析全面，改进方案覆盖了频率控制、去重、AIGC、行为模拟和账号监控。但缺少具体的配置化设计说明，综合扣1分；状态机未提到DLQ扣2分。"}
{"input": "用户提交了一条定时发布任务：'本周五晚8点，将同一篇视频同时发布到抖音和小红书。' 请说明你的系统会如何执行这个任务，并如何保证两个平台都成功。", "expected_output": {"task_parsing": {"scheduled_at": "本周五20:00", "platforms": ["douyin", "xiaohongshu"], "content_type": "video"}, "preprocess": {"douyin": {"title_crop": "≤100字", "video": "竖屏9:16，≤60分钟", "aigc_label": "强制显著标识"}, "xiaohongshu": {"title_crop": "≤20字", "video": "≤15分钟", "description": "300-1000字，短句+Emoji", "aigc_label": "强制双标识"}}, "queue_strategy": {"type": "delayed_queue", "implementation": "按平台拆分队列 douyin_queue / xiaohongshu_queue", "publish_time": "周五20:00~20:10错峰执行，小红书延迟5分钟"}, "execution": {"douyin": "企业号优先API，个人号fallback到Playwright RPA", "xiaohongshu": "Playwright/Selenium RPA，30-120秒随机延迟"}, "state_tracking": {"douyin_task_id": "T-douyin-xxx", "xiaohongshu_task_id": "T-xhs-xxx", "monitor": "分别追踪两平台状态，任何一方失败独立重试", "success_criteria": "两平台均达到PUBLISHED或FAILED_PERMANENT"}}, "dimensions": {"platform_adaptation": 15, "content_preprocessing": 14, "publish_execution": 14, "state_tracking_retry": 14, "compliance_risk_control": 14, "overall_design": 22}, "score": 93, "grader_hints": "执行方案完整，定时、错峰、平台适配、RPA fallback、状态追踪都覆盖。但小红书RPA未提IP隔离，风控扣2分；综合能力优秀但缺少具体的Adapter接口定义，扣1分。"}
{"input": "系统需要新增一个'知乎'平台的发布支持。请说明在不修改现有核心业务代码的前提下，如何快速扩展适配这一新平台。", "expected_output": {"extension_strategy": "基于统一的 Platform Adapter 接口新增 ZhihuAdapter", "adapter_interface_methods": ["initialize(config)", "adaptContent(post, platformConfig)", "publish(task)", "parseResponse(rawResponse)", "refreshAuth()"], "implementation_steps": ["1) 新建 zhihu_adapter.js 实现 Adapter 接口", "2) 在 PlatformConfig 表中添加知乎规则（标题长度、正文HTML规则、图片限制等）", "3) 注册 adapter 到 AdapterRegistry", "4) 无需修改 Pipeline/Queue/StateTracking 核心代码"], "config_example": {"platform": "zhihu", "adapter_type": "api|rpa", "title_max_length": 100, "content_max_length": 50000, "image_max_count": 20, "allowed_media_types": ["image"], "require_aigc_label": true}, "key_principle": "Configuration-as-Code，新增平台=新增Adapter+配置，核心流程零改动"}, "dimensions": {"platform_adaptation": 13, "content_preprocessing": 10, "publish_execution": 11, "state_tracking_retry": 12, "compliance_risk_control": 10, "overall_design": 24}, "score": 80, "grader_hints": "扩展设计符合Adapter+配置化原则，思路正确。但知乎具体规范（如不支持视频、盐选专栏等）未做深入分析，平台适配扣2分；其他维度因题目聚焦扩展性，得分中等。"}
{"input": "一份推广文案中包含'加微信 xxx888 领取免费课程，扫码进群还有红包'。请说明内容预审服务会如何处置这条内容，以及各平台的最终发布结果。", "expected_output": {"moderation_result": {"level": "A级拦截", "reason": "包含微信号、导流信息、红包诱导，触发敏感词规则"}, "fast_filter_hits": ["加微信", "扫码进群", "红包"], "ai_review": "BERT识别为交易导流推广，风险极高", "disposition": "直接拦截，不入队列", "platform_outcomes": {"weibo": "被拦截，未发布（微博对营销导流有严格限制）", "wechat": "被拦截，未发布（公众号虽不拦截但易触发用户投诉）", "douyin": "被拦截（2025年7月起严禁第三方工具发货推广）", "xiaohongshu": "被拦截（交易导流治理规则，情节严重永久封号）", "video_channel": "被拦截"}, "suggestion": "删除微信号和二维码，改用点击主页链接或平台官方推广工具"}, "dimensions": {"platform_adaptation": 12, "content_preprocessing": 11, "publish_execution": 8, "state_tracking_retry": 8, "compliance_risk_control": 15, "overall_design": 14}, "score": 68, "grader_hints": "内容拦截判断完全正确，风控满分。但本题几乎不涉及发布执行和状态追踪，属于单场景风控题，后几项按基础 awareness 打分。"}
{"input": "请为自动化发布系统设计一个'配置即代码'的 PlatformConfig schema，用于管理微博、微信公众号、抖音、小红书、视频号的发布规则。规则变化频繁的字段包括标题长度、图片尺寸、视频时长、日限条数、AIGC标识要求和黄金发布时段。", "expected_output": {"schema": {"platform": "string (enum)", "adapter_type": "api | rpa", "title_max_length": "int", "content_max_length": "int", "image_max_count": "int", "image_recommended_size": "string (e.g., '1080x1440')", "video_max_duration_sec": "int", "daily_publish_limit": "int", "allowed_media_types": ["image", "video"], "require_aigc_label": "boolean", "aigc_label_type": "explicit | implicit | dual", "active_hours": ["string (time range)"], "auth_type": "oauth2 | cookie | qr_code", "rate_limit_strategy": "read_header | fixed_window | token_bucket"}, "examples": [{"platform": "weibo", "adapter_type": "api", "title_max_length": 5000, "content_max_length": 5000, "image_max_count": 18, "video_max_duration_sec": 900, "daily_publish_limit": 100, "require_aigc_label": true, "active_hours": ["08:00-10:00", "20:00-22:00"]}, {"platform": "xiaohongshu", "adapter_type": "rpa", "title_max_length": 20, "content_max_length": 1000, "image_max_count": 9, "image_recommended_size": "1080x1440", "video_max_duration_sec": 900, "daily_publish_limit": 10, "require_aigc_label": true, "aigc_label_type": "dual", "active_hours": ["12:00-14:00", "20:00-23:00"]}], "design_principles": ["所有Adapter在运行期读取PlatformConfig，拒绝硬编码", "支持热更新（配置变更后无需重启服务）", "版本控制：配置修改留痕，便于回滚"]}, "dimensions": {"platform_adaptation": 14, "content_preprocessing": 11, "publish_execution": 12, "state_tracking_retry": 10, "compliance_risk_control": 12, "overall_design": 24}, "score": 83, "grader_hints": "Schema设计合理且实用，覆盖了所有要求的变化字段，示例清晰。但缺少配置校验机制（如范围检查）和降级策略，平台适配和综合能力各扣1分。"}
{"input": "你的系统正在使用抖音官方API发布视频，突然返回401 Unauthorized。请给出完整的错误处理流程，包括状态变更、通知机制和修复建议。", "expected_output": {"error_classification": "永久失败（认证失效）", "immediate_action": "停止重试，防止触发安全策略", "status_transition": "PUBLISHING -> FAILED_PERMANENT", "notification": {"channels": ["钉钉", "企业微信", "邮件"], "message": "抖音 access_token 已失效，请立即刷新 OAuth2.0 授权"}, "fix_steps": ["1) 检查 client_id/client_secret 是否正确", "2) 重新获取 access_token（有效期2小时）", "3) 若企业资质过期，重新完成开发者认证", "4) 更新系统中的 token 配置并手动重试该任务"], "prevention": "设置 token 过期前 5 分钟自动刷新机制；监控 401 频率"}, "dimensions": {"platform_adaptation": 13, "content_preprocessing": 9, "publish_execution": 14, "state_tracking_retry": 13, "compliance_risk_control": 10, "overall_design": 17}, "score": 76, "grader_hints": "错误处理流程规范，通知和预防机制都有。但缺少DLQ转入和人工复核说明；预处理维度因题目不涉及只得基础分。"}
{"input": "某公司运营团队需要批量发布100篇文章到微信公众号（订阅号）。请设计一套批量发布的执行方案，避免触发平台的风控和频次限制。", "expected_output": {"platform_constraints": {"account_type": "订阅号", "mass_send_limit": "1次/天", "publish_limit": "无次数限制（freepublish）"}, "batch_strategy": {"mode": "freepublish 批量发布（非群发，避免日限）", "chunk_size": 20, "interval_between_chunks": "300秒", "total_estimated_time": "约25分钟"}, "content_prep": {"html_validation": "确保HTML无JS，<2万字，图片<1MB且使用微信域名URL", "cover_image": "每篇文章提前上传永久素材获取thumb_media_id"}, "state_tracking": "每篇文章生成独立PublishTask，统一批次号batch-001", "risk_control": "错峰发布，避免短时间大量请求；监控队列深度和成功率；如有失败立即告警"}, "dimensions": {"platform_adaptation": 14, "content_preprocessing": 12, "publish_execution": 13, "state_tracking_retry": 13, "compliance_risk_control": 13, "overall_design": 18}, "score": 83, "grader_hints": "批量方案合理，充分利用了freepublish无次数限制的特点。但缺少具体的Adapter队列拆分细节和失败重试的批次协调策略，综合和执行各扣1-2分。"}