{"skill":{"slug":"feedback-loop","displayName":"Feedback Loop","summary":"Feedback Loop - Collect, analyze, and act on user feedback for continuous agent improvement","description":"---\nname: feedback-loop\nversion: 1.0.0\ndescription: Feedback Loop - Collect, analyze, and act on user feedback for continuous agent improvement\nauthor: OpenClaw\nlicense: MIT\ntags:\n  - feedback\n  - analytics\n  - improvement\n  - tracking\n  - agent-optimization\ncli:\n  command: feedback-loop\n  entry: bin/cli.js\ndependencies: []\n---\n\n# Feedback Loop Skill\n\n反馈循环器 - 一个完整的反馈收集、分析和改进跟踪系统，用于 OpenClaw 代理的持续优化。\n\n## 功能概述\n\n### 核心功能\n\n1. **反馈收集（Feedback Collection）**\n   - 显式反馈：用户主动评分、评论\n   - 隐式反馈：从交互模式自动检测（完成、重试、放弃等）\n   - 自动检测：基于会话模式智能识别反馈信号\n\n2. **反馈分析（Feedback Analysis）**\n   - 聚类分析：按类别、情感、评分分组\n   - 趋势分析：时间序列趋势检测\n   - 情感分析：正/负/中性情感分布\n   - 模式检测：识别 recurring issues、行为模式等\n\n3. **改进建议生成（Improvement Suggestions）**\n   - 基于分析结果自动生成可执行的改进建议\n   - 优先级排序（high/medium/low）\n   - 包含具体 action items 和预期影响\n\n4. **效果跟踪（Effect Tracking）**\n   - 跟踪建议实施进度\n   - 测量实施前后的影响\n   - 生成综合报告\n\n## 安装\n\n```bash\n# 通过 ClawHub 安装（推荐）\nclawhub install feedback-loop\n\n# 或手动安装\ncd ~/.openclaw/workspace/skills/feedback-loop\nnpm install\nnpm link\n```\n\n## 使用方法\n\n### CLI 命令\n\n#### 1. 提供反馈（provide）\n\n**显式反馈：**\n```bash\nfeedback-loop provide --type explicit --rating 5 --comment \"Excellent response!\" --category accuracy\n```\n\n**隐式反馈：**\n```bash\nfeedback-loop provide --type implicit --signal completion --sessionId sess123\nfeedback-loop provide --type implicit --signal retry --metrics '{\"retryCount\": 3}'\n```\n\n**参数说明：**\n- `--type`: feedback 类型（explicit 或 implicit）\n- `--rating`: 评分（1-5, thumbs_up, thumbs_down）\n- `--comment`: 可选评论\n- `--category`: 反馈类别（accuracy, speed, helpfulness 等）\n- `--signal`: 隐式信号类型（completion, retry, abandon, correction 等）\n- `--sessionId`: 会话标识符\n- `--metrics`: JSON 格式的性能指标\n- `--context`: JSON 格式的上下文信息\n\n#### 2. 分析反馈（analyze）\n\n```bash\n# 分析最近一周的反馈\nfeedback-loop analyze --range week\n\n# 只分析显式反馈\nfeedback-loop analyze --explicit-only\n\n# 分析最近一个月的数据，输出 JSON\nfeedback-loop analyze --range month --output json\n```\n\n**参数说明：**\n- `--range`: 时间范围（day, week, month, all）\n- `--explicit-only`: 仅分析显式反馈\n- `--output`: 输出格式（json, pretty）\n\n#### 3. 生成建议（suggest）\n\n```bash\n# 生成最多 5 条建议\nfeedback-loop suggest --max 5\n\n# 专注于特定类别\nfeedback-loop suggest --focus quality --max 10\n```\n\n**参数说明：**\n- `--max`: 最大建议数量\n- `--focus`: 专注的类别\n- `--output`: 输出格式\n\n#### 4. 跟踪进度（track）\n\n```bash\n# 跟踪建议实施进度\nfeedback-loop track fb_123456 --phase implementation --status in_progress\n\n# 标记为已完成\nfeedback-loop track fb_123456 --phase deployed --status completed --notes \"Successfully implemented\"\n```\n\n**参数说明：**\n- `--phase`: 实施阶段（planning, implementation, testing, deployed）\n- `--status`: 状态（in_progress, completed, blocked）\n- `--notes`: 附加说明\n- `--metrics`: JSON 格式的进度指标\n\n#### 5. 查看统计（stats）\n\n```bash\nfeedback-loop stats\nfeedback-loop stats --output json\n```\n\n#### 6. 列出数据（list）\n\n```bash\n# 列出反馈\nfeedback-loop list feedback --limit 10\nfeedback-loop list feedback --type explicit\n\n# 列出建议\nfeedback-loop list suggestions --status pending\nfeedback-loop list suggestions --category quality\n\n# 列出跟踪记录\nfeedback-loop list tracking --phase implementation\n```\n\n#### 7. 生成报告（report）\n\n```bash\nfeedback-loop report\nfeedback-loop report --output json\n```\n\n#### 8. 导出数据（export）\n\n```bash\nfeedback-loop export --format json --output data.json\nfeedback-loop export --format csv --output data.csv\n```\n\n### 编程接口\n\n```javascript\nconst FeedbackLoop = require('./src/index');\n\nconst fl = new FeedbackLoop();\n\n// 提供反馈\nfl.provide({\n  type: 'explicit',\n  rating: 5,\n  comment: 'Great!',\n  category: 'helpfulness'\n});\n\n// 分析\nconst analysis = fl.analyze({ timeRange: 'week' });\n\n// 生成建议\nconst suggestions = fl.suggest({ maxSuggestions: 5 });\n\n// 跟踪\nfl.track(suggestionId, {\n  phase: 'implementation',\n  status: 'in_progress'\n});\n\n// 获取统计\nconst stats = fl.getStats();\n\n// 获取报告\nconst report = fl.getReport();\n```\n\n## 触发方式\n\n### 主动收集\n- 在会话结束时自动请求评分\n- 定期生成分析报告\n- 检测到低满意度时触发改进流程\n\n### 自动检测\n- 高重试率 → 推断用户遇到困难\n- 快速完成 → 推断用户满意\n- 早期放弃 → 推断响应不符合期望\n- 多次追问 → 推断高参与度\n\n## 数据结构\n\n### 反馈记录\n```json\n{\n  \"id\": \"fb_1234567890_abc123\",\n  \"timestamp\": \"2024-01-15T10:30:00.000Z\",\n  \"type\": \"explicit\",\n  \"sessionId\": \"sess_123\",\n  \"rating\": 5,\n  \"comment\": \"Excellent response!\",\n  \"category\": \"accuracy\",\n  \"metadata\": {},\n  \"source\": \"cli\"\n}\n```\n\n### 隐式反馈\n```json\n{\n  \"id\": \"fb_1234567890_def456\",\n  \"timestamp\": \"2024-01-15T10:30:00.000Z\",\n  \"type\": \"implicit\",\n  \"sessionId\": \"sess_123\",\n  \"signal\": \"completion\",\n  \"metrics\": { \"responseTime\": 2500 },\n  \"context\": { \"autoDetected\": true },\n  \"inferredSentiment\": \"positive\"\n}\n```\n\n### 改进建议\n```json\n{\n  \"id\": \"fb_1234567890_ghi789\",\n  \"timestamp\": \"2024-01-15T10:30:00.000Z\",\n  \"title\": \"Address recurring issues in accuracy\",\n  \"description\": \"5 negative feedback items identified...\",\n  \"category\": \"accuracy\",\n  \"priority\": \"high\",\n  \"actionItems\": [...],\n  \"expectedImpact\": \"Reduce negative feedback...\",\n  \"status\": \"pending\"\n}\n```\n\n## 最佳实践\n\n1. **定期分析**：每周运行一次分析，及时发现趋势\n2. **快速响应**：对高优先级建议立即采取行动\n3. **持续跟踪**：记录每个建议的实施进度\n4. **衡量影响**：实施后对比前后数据\n5. **闭环管理**：确保每个反馈都有对应的改进行动\n\n## 文件结构\n\n```\nfeedback-loop/\n├── SKILL.md           # 技能文档\n├── package.json       # 项目配置\n├── bin/\n│   └── cli.js         # CLI 入口\n├── src/\n│   ├── index.js       # 主入口\n│   ├── storage.js     # 数据存储\n│   ├── collector.js   # 反馈收集\n│   ├── analyzer.js    # 反馈分析\n│   ├── suggester.js   # 建议生成\n│   └── tracker.js     # 效果跟踪\n├── data/              # 数据目录（自动生成）\n│   ├── feedback.json\n│   ├── analysis.json\n│   ├── suggestions.json\n│   └── tracking.json\n└── test/\n    └── run.js         # 测试脚本\n```\n\n## 注意事项\n\n- 数据存储在 `data/` 目录下，定期备份重要数据\n- 建议设置定期清理策略，避免数据文件过大\n- 敏感反馈数据应注意隐私保护\n\n## 版本历史\n\n- **1.0.0** - 初始版本\n  - 完整的反馈收集功能\n  - 多维度分析能力\n  - 智能建议生成\n  - 效果跟踪系统\n","topics":["Feedback","Improvement","Agent Optimization","Analytics","Tracking"],"tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":677,"installsAllTime":25,"installsCurrent":5,"stars":0,"versions":1},"createdAt":1773286398968,"updatedAt":1780693864172},"latestVersion":{"version":"1.0.0","createdAt":1773286398968,"changelog":"Initial release of Feedback Loop skill.\n\n- Collect explicit and implicit user feedback with CLI commands and programming interface.\n- Analyze feedback with clustering, trend, sentiment, and pattern detection.\n- Automatically generate prioritized improvement suggestions with actionable items.\n- Track progress and effects of implemented suggestions.\n- Generate statistics and reports; export data in JSON/CSV.\n- Organize all feedback, suggestions, and tracking data for ongoing agent optimization.","license":"MIT-0"},"metadata":{"setup":[],"os":null,"systems":null},"owner":{"handle":"harrylabsj","userId":"s17a8m9q4jybb46cv60h4fxard83hmsn","displayName":"haidong","image":"https://avatars.githubusercontent.com/u/144880725?v=4"},"moderation":null}