Menu
Use when scanning this repository skill catalog and routing work to the correct skill group.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
| description | Use when scanning this repository skill catalog and routing work to the correct skill group. |
| name | skills-index |
AI-Fleet 技能群速查 | 29 Groups, 668 skills | 单一事实源:
configs/skill-groups.jsonCross-CLI: Claude Code / Codex / Gemini 均应在任务启动时读取本文件
架构 (2026-03-31 重构):
/ 命令只显示 Skill Group（组合技能），不显示个体 skill_lib/ 目录（项目级 + 全局级），文件名为 SPEC.md_lib/<skill-name>/SPEC.md加载方式:
~/.codex/instructions.md 引用，或项目根目录 SKILL.md；Codex 原生 skills 在 ~/.codex/skills/ai-tools-launcher.sh 的 ensure_agent_docs() 自动复制到项目目录单一事实源:
configs/skill-groups.json| 最多同时激活 3 个组 | base 组始终活跃
| Group ID | Name | Core Skills | Trigger Keywords | Est. Tokens |
|---|---|---|---|---|
| base | Base Essentials | planning-with-files, git-commit, verification-before-completion, code-review | (always active) | 2800 |
| frontend-dev | Frontend Development | frontend-design, react-best-practices, frontend-testing | frontend, React, dashboard, component | 3500 |
| design-system | Design System | design-taste-frontend, product-design, stitch-design-pipeline | design, redesign, brand, prototype, Stitch | 3000 |
| doc-styling | Document Styling | html-style-router | report, guide, briefing, analysis | 2000 |
| product | Product Management | brainstorming, ralph-wiggum, compliance-docs, stitch-design-pipeline, cognitive-skeleton | brainstorm, PRD, product, feature, prototype, mental model, 思维模型 | 4500 |
| architecture | Architecture & Engineering | rpi-research -> rpi-plan -> rpi-implement (sequenced) | architecture, new feature, ADR | 3500 |
| research | Deep Research | deep-research, notebooklm, agent-reach, x-reader | research, SOTA, read article, fetch URL | 2000 |
| data | Data & Analytics | bigdata-core, bigdata-viz | data, ML, analytics, visualization | 3000 |
| devops | DevOps & Quality | git-commit, code-review, github-pr-creation, verification-before-completion | git, PR, review, deploy | 3000 |
| media | Media Production | video-constrict, remotion-best-practices | video, image, media, download | 2500 |
| skill-ops-pipeline | Agent & Skill Operations | skill-creator, creating-skills, skill-group-loop | skill creator, eval skill, optimize skill, migrate agent | 3000 |
| orchestration | Heavy Orchestration (Conductor) | conductor-go, conductor-orchestrator, conductor-evaluators, conductor-board | /go, conductor, board meeting, full orchestration | 4000 |
| device-automation | Device Automation | android-device-automation, ios-device-automation, desktop-computer-automation | android, ios, mobile, device, desktop automation | 2500 |
| browser-scraping | Browser & Scraping | browser-automation, chrome-bridge-automation, scrapling-crawler | scrape, crawl, Cloudflare, headless, Electron | 2500 |
| knowledge | Knowledge Management | obsidian-markdown, obsidian-bases | obsidian, vault, PDF, canvas, mind map | 2000 |
| automation | Workflow Automation | n8n-automation, ghost-os-automation | n8n, ghost os, 小红书, macOS automation | 2500 |
| market-strategy | Market Strategy & GTM | competitor-analysis, competitive-battlecard, gtm-strategy, cognitive-skeleton | GTM, market research, pricing, SWOT, growth, 芒格, framework | 3000 |
| frontend-polish | Frontend Polish | design-taste-frontend, redesign-existing-projects, jh3y-hover-cards | polish, 打磨, UI improvement, redesign | 2500 |
Activation: auto-detect (task keyword match) | explicit ("use frontend-dev group") | task-type (SOP trigger)
Groups with execution.mode: "loop" in configs/skill-groups.json run their skills iteratively:
| Group | Loop Skills | Gate | Max Iter |
|---|---|---|---|
| frontend-dev | frontend-design -> react-best-practices -> frontend-testing | npx tsc --noEmit | 3 |
| architecture | rpi-research -> rpi-plan -> rpi-implement | ai check --json --no-tests | 3 |
| product | brainstorming -> ralph-wiggum -> compliance-docs | completion_promise (PRD_COMPLETE) | 2 |
| skill-ops-pipeline | skill-creator -> creating-skills -> skill-group-loop | registry_validator --groups --strict | 3 |
| orchestration | conductor-go -> conductor-orchestrator -> conductor-evaluators | completion_promise (TRACK_COMPLETE) | 5 |
Trigger: "group loop", "skill loop", "PDCA loop", "iterative improvement"
Orchestrator: skill-group-loop skill (.claude/skills/skill-group-loop/SKILL.md)
| 场景 | 推荐模式 | 理由 |
|---|---|---|
| 单文件修改/小 bugfix | base (无组切换) | 开销最低 |
| 中等功能 (1-3 文件) | architecture (RPI) | 人工可控的三阶段 |
| 大型功能 (5+ 文件, 多依赖) | orchestration (Conductor) | 自动 DAG 并行 + Board 审核 + 知识学习 |
| 多 Agent 并行 (独立任务) | Superset Desktop + 任意组 | 进程隔离 + worktree，每个 Agent 内部自选组 |
安装到用户主目录的 Claude Skills（通过 npm exec -- skills add <user/repo> -a claude-code -g），不在 SKILL-MANIFEST.json 整合审计范围内，但参与 runtime 路由：
| Skill | Source | Trigger | Routing Note |
|---|---|---|---|
| kami | tw93/kami v1.0.0 (2026-04-20, MIT) | resume / 简历 / letter / 信件 / portfolio / 作品集 / slides / 演示稿 / 一页纸 / 白皮书 + 羊皮纸衬线美学 | 补 html-style-router 正交轴：1 风格 × 6 模板。模板优先请求 → kami；风格优先请求 → html-style-router。详见 outputs/reports/research-learning-swarm/2026-04-21-kami-learning-digest.md |
| cc-design | ZeroZ-lab/cc-design (2026-04-21) | brand clone (Apple/Stripe/Vercel/Notion/Linear/Tesla/Airbnb/OpenAI…) / landing page / React prototype / UI mockup / slide deck / tweaks variant / Playwright design-verify | 补三个正交轴：(1) 消费科技+开发者工具+汽车+fintech 的品牌克隆（html-style-router 只覆盖 editorial/consulting 12 风格，无 consumer tech）(2) React+Babel inline setup + Playwright 设计验证协议 (3) iOS/Android/macOS/browser frame 模板。68+ 品牌 via getdesign.md WebFetch on-demand（非 bundle）。比 frontend-design/impeccable-design 更具体（后两者仅哲学指引，无执行协议） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| cognitive-skeleton | layers/L3-intelligence/skills | 决策/策略/风险/增长/问题分析，211 模型格栅（111 Munger + 100 PM），9 场景路由 |
| planning-with-files | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 大型任务，需跨会话断点续传 |
| workflow-router | layers/L3-intelligence/skills | 任务场景路由，识别 scenario/scope/mode |
| engineering-protocol | layers/L3-intelligence/skills | 最小 diff + 拓扑映射协议注入 |
| ralph-loop | layers/L3-intelligence/skills, codex-skills | 多轮自纠错直到完成承诺 |
| ralph-wiggum | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 需求 -> 规划 -> 迭代闭环（AUDIENCE_JTBD） |
| brainstorming | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 创建功能/组件/修改行为前的需求探索 |
| continue | .claude/skills/ | 跨会话恢复上下文 |
| prompt-engineering-100 | layers/L3-intelligence/skills | 需求分析 -> 可执行提示词生成 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| writing-plans | skills/shared/, ~/.codex/skills/ | 编写实现计划 |
| executing-plans | skills/shared/, ~/.codex/skills/ | 执行实现计划 |
| subagent-driven-development | skills/shared/, ~/.claude/skills/, ~/.codex/skills/ | 并行 SubAgent 实现 |
| dispatching-parallel-agents | skills/shared/, ~/.claude/skills/, ~/.codex/skills/ | 2+ 独立任务并行分发 |
| test-driven-development | layers/L3-intelligence/skills/skills/ | TDD 实现 |
| systematic-debugging | layers/L3-intelligence/skills/skills/ | Bug/测试失败/异常排查 |
| vibe-debug | layers/L3-intelligence/skills/skills/ | 自动化 TDD 调试流水线 (reproduce -> test -> fix-loop -> verify) |
| using-git-worktrees | skills/shared/, ~/.codex/skills/ | 隔离 feature 工作 |
| finishing-a-development-branch | skills/shared/, ~/.codex/skills/ | 完成后合并/PR/清理 |
| using-superpowers | skills/shared/, ~/.codex/skills/ | 会话初始化，发现可用 skills |
| Skill | 目录 | 触发场景 |
|---|---|---|
| git-commit | skills/shared/, ~/.claude/skills/, ~/.codex/skills/ | 提交变更（Conventional Commits） |
| commit-helper | .claude/skills/ | 分析 diff 生成 commit message |
| github-pr-creation | skills/shared/, ~/.codex/skills/ | 创建 PR |
| github-pr-merge | skills/shared/, ~/.codex/skills/ | 合并 PR |
| github-pr-review | skills/shared/, ~/.codex/skills/ | 处理 PR review 反馈 |
| release-skills | .claude/skills/ | 通用发布工作流（自动检测版本文件/changelog，9 步发布，7 语言 changelog，baoyu-skills） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| code-review | .claude/skills/ | 代码质量/安全/可维护性审查 |
| requesting-code-review | skills/shared/, ~/.codex/skills/ | 完成任务后请求审查 |
| receiving-code-review | skills/shared/, ~/.codex/skills/ | 收到审查反馈时 |
| verification-before-completion | skills/shared/, ~/.codex/skills/ | 宣称完成前验证 |
| creating-skills | skills/shared/, ~/.claude/skills/, ~/.codex/skills/ | 创建新 skill |
| writing-skills | skills/shared/, ~/.codex/skills/ | 编写/编辑 skill |
| bug-hunt | ~/.claude/skills/ | 对抗式 3-Agent Bug 狩猎（adversarial bug hunting） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| frontend-design | .claude/skills/ | 生产级前端界面设计 |
| frontend-testing | .claude/skills/ | 前端性能/控制台/响应式/E2E 验证 |
| react-best-practices | .claude/skills/ | React/Next.js 性能优化 |
| product-design | .claude/skills/ | 画布/原型/品牌资产 |
| tanstack-start-dashboard | .claude/skills/ | TanStack Start + shadcn/ui 后台脚手架 |
| diceui | .claude/skills/ | shadcn/ui 业务组件库 |
| jh3y-hover-cards | .claude/skills/ | CodePen hover blur 效果封装 |
| valq-powerbi-planning | .claude/skills/ | ValQ 风格规划组件/图表 |
| ui-skills | codex-skills | UI 约束规则安装 |
| web-interface-guidelines | codex-skills | Vercel Web 界面规范安装 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| bigdata-core | .claude/skills/ | 大规模数据处理（Polars/Dask/Vaex） |
| bigdata-ml | .claude/skills/ | ML 训练/部署（sklearn/PyTorch/Transformers） |
| bigdata-viz | .claude/skills/ | 数据可视化（Matplotlib/Seaborn/Plotly） |
| excel-analysis | .claude/skills/ | Excel/CSV 分析与处理 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| algo-core | .claude/skills/ | 图算法/符号数学/优化/仿真/概率编程 |
| algo-dl | .claude/skills/ | 深度学习（PyTorch/Transformers/GNN/RL） |
| karpathy-autoresearch | .claude/skills/ | 自动迭代实验框架（Karpathy 方法论，ML 训练/调参/代码优化） |
| research-paper-writing | ~/.claude/skills/ | ML/CV/NLP 学术论文写作（详细技能集） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| brainstorm-ideas-new | .claude/skills/ | 新产品创意发散（多视角） |
| brainstorm-ideas-existing | .claude/skills/ | 已有产品创意发散 |
| identify-assumptions-new | .claude/skills/ | 新产品假设识别（8 类风险） |
| identify-assumptions-existing | .claude/skills/ | 已有产品假设识别 |
| prioritize-assumptions | .claude/skills/ | 假设优先级（Impact x Risk 矩阵） |
| brainstorm-experiments-new | .claude/skills/ | 新产品实验设计（pretotype） |
| brainstorm-experiments-existing | .claude/skills/ | 已有产品实验设计 |
| opportunity-solution-tree | .claude/skills/ | OST 机会方案树（Teresa Torres） |
| interview-script | .claude/skills/ | 用户访谈脚本（JTBD 探询） |
| summarize-interview | .claude/skills/ | 访谈摘要结构化 |
| analyze-feature-requests | .claude/skills/ | 需求分析与优先级排序 |
| metrics-dashboard | .claude/skills/ | 指标看板设计 |
| prioritize-features | .claude/skills/ | 功能优先级（9 种框架） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| product-vision | .claude/skills/ | 产品愿景构建 |
| product-strategy | .claude/skills/ | 产品战略（9 section） |
| value-proposition | .claude/skills/ | 价值主张设计（6 部分 JTBD） |
| business-model | .claude/skills/ | 商业模式画布（BMC 9 模块） |
| lean-canvas | .claude/skills/ | 精益画布 |
| startup-canvas | .claude/skills/ | 创业画布（Strategy + Lean 组合） |
| swot-analysis | .claude/skills/ | SWOT 分析 |
| pestle-analysis | .claude/skills/ | PESTLE 宏观环境分析 |
| porters-five-forces | .claude/skills/ | 波特五力分析 |
| ansoff-matrix | .claude/skills/ | 安索夫矩阵（增长战略） |
| monetization-strategy | .claude/skills/ | 变现策略设计 |
| pricing-strategy | .claude/skills/ | 定价策略分析 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| create-prd | .claude/skills/ | PRD 撰写（8 section） |
| brainstorm-okrs | .claude/skills/ | OKR 构思 |
| outcome-roadmap | .claude/skills/ | 结果导向路线图 |
| sprint-plan | .claude/skills/ | Sprint 规划 |
| user-stories | .claude/skills/ | 用户故事（3C） |
| job-stories | .claude/skills/ | Job Stories（JTBD 格式） |
| wwas | .claude/skills/ | Why-What-Acceptance 格式 backlog |
| test-scenarios | .claude/skills/ | 测试场景生成 |
| stakeholder-map | .claude/skills/ | 利益相关方地图（Power x Interest） |
| pre-mortem | .claude/skills/ | 事前验尸分析 |
| retro | .claude/skills/ | Sprint 回顾 |
| release-notes | .claude/skills/ | 发布说明生成 |
| summarize-meeting | .claude/skills/ | 会议纪要结构化 |
| dummy-dataset | .claude/skills/ | 生成测试数据集 |
| prioritization-frameworks | .claude/skills/ | 9 种优先级框架参考 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| user-personas | .claude/skills/ | 用户画像（3 personas + JTBD） |
| user-segmentation | .claude/skills/ | 用户分群 |
| market-segments | .claude/skills/ | 市场细分 |
| market-sizing | .claude/skills/ | 市场规模（TAM/SAM/SOM） |
| customer-journey-map | .claude/skills/ | 用户旅程地图 |
| competitor-analysis | .claude/skills/ | 竞品分析 |
| sentiment-analysis | .claude/skills/ | 情感分析 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| sql-queries | .claude/skills/ | 自然语言生成 SQL |
| cohort-analysis | .claude/skills/ | 留存/cohort 分析 |
| ab-test-analysis | .claude/skills/ | A/B 测试统计显著性分析 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| beachhead-segment | .claude/skills/ | 登陆市场选择 |
| ideal-customer-profile | .claude/skills/ | ICP 定义 |
| gtm-strategy | .claude/skills/ | GTM 策略规划 |
| gtm-motions | .claude/skills/ | GTM 动作类型（7 种） |
| growth-loops | .claude/skills/ | 增长飞轮设计 |
| competitive-battlecard | .claude/skills/ | 竞争对手 Battlecard |
| Skill | 目录 | 触发场景 |
|---|---|---|
| marketing-ideas | .claude/skills/ | 营销创意（低成本） |
| positioning-ideas | .claude/skills/ | 定位差异化 |
| value-prop-statements | .claude/skills/ | 价值主张文案 |
| product-name | .claude/skills/ | 产品命名 |
| north-star-metric | .claude/skills/ | 北极星指标定义 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| review-resume | .claude/skills/ | PM 简历审查 |
| draft-nda | .claude/skills/ | NDA 起草 |
| privacy-policy | .claude/skills/ | 隐私政策起草 |
| grammar-check | .claude/skills/ | 语法/逻辑/文风检查 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| pdf-processing | .claude/skills/ | PDF 提取/填表/合并/转换 |
| pdf-layout-analysis | .claude/skills/ | PDF 版面分析（Docker 服务） |
| compliance-docs | .claude/skills/ | 合规文档/政策映射/审计报告 |
| codegen-doc | ~/.claude/skills/ | 代码生成文档（Code-to-Documentation） |
| md-report-summary | ~/.claude/skills/ | Markdown 报告摘要生成 |
| paper-write | ~/.claude/skills/ | 学术论文写作 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| zread-introspection | layers/L3-intelligence/skills/skills/ | Zread MCP 源码级自省，允许 OpenClaw 检索自身框架源码 |
| obsidian-markdown | .claude/skills/ | Obsidian 风格 Markdown（wikilinks/callouts） |
| obsidian-bases | .claude/skills/ | Obsidian Bases（数据库视图） |
| json-canvas | .claude/skills/ | JSON Canvas 文件（mind map/flowchart） |
| notebooklm | .claude/skills/, layers/L3-intelligence/skills | Google NotebookLM 查询 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| hyperx-trade | layers/L3-intelligence/skills/skills/ | Hyperliquid 钱包分析、清算热图与自动化跟单 (Copy-Trading) |
| Skill | 目录 | 触发场景 |
|---|---|---|
| video (registry) | layers/L3-intelligence/skills | 一句话生成视频 |
| video-constrict | .claude/skills/ | 视频压缩到目标大小 |
| remotion-best-practices | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | Remotion 视频创作 + 短视频制作工作流 |
| nanobanana-image-gen | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | Gemini Flash/Pro 图片生成/编辑（MCP） |
| downloadhd | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 54+ 平台视频/图片/音频下载（浏览器自动化） |
| surge | ~/bin/surge | Go TUI 多线程下载管理器 (v0.6.10, 32 并发连接, 多镜像, 守护进程架构) |
| baoyu-translate | ~/.claude/skills/ | 三模式翻译 (快速/标准/精修, 长文分块, 术语表, 7.1K stars) |
| makeownsrt | ~/.claude/skills/ | MKV 字幕提取 + 翻译 |
| wechat-article-writer | ~/.claude/skills/ | 微信公众号文章生成 |
| excalidraw-diagram-skill | ~/.claude/skills/ | Excalidraw 图表生成 |
| drawio-diagram | ~/.claude/skills/ | Draw.io 图表生成 |
| codegen-diagram | ~/.claude/skills/ | 代码生成图表（Code-to-Diagram） |
| pptgen-drawio | ~/.claude/skills/ | PPT 生成（含 Draw.io 图表） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| better-icons | .claude/skills/ | 200K+ 图标搜索（150+ collections: Lucide/Heroicons/Material Design 等，npx better-icons search/get） |
| canvas-design | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 画布级视觉艺术创作（海报/设计稿） |
| design-taste-frontend | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 反 AI Purple 前端设计（可调参数/反模式目录/硬件加速） |
| redesign-existing-projects | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 审计既有 UI 并升级为高端设计（不破坏功能） |
| stitch-design-pipeline | .claude/skills/ | Google Stitch AI 原型设计流水线：PRD -> 多方案生成 -> AI 比稿 -> 选定 -> 前端代码 |
| full-output-enforcement | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 强制完整输出（禁止 TODO/占位符/截断） |
| lobe-icons | .claude/skills/ | LobeHub AI/LLM 模型图标 |
| svglogo-icons | .claude/skills/ | 国内矢量 LOGO 图标库 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| uxc-universal-cli | layers/L3-intelligence/skills/skills/ | OpenAPI/GraphQL/MCP/gRPC 通用 API 渐进式发现与确定性调用 |
| n8n-automation | .claude/skills/ | n8n 工作流/Code 节点/表达式 |
| osint-framework | .claude/skills/ | OSINT 资源目录（合规模式） |
| agent-browser-electron | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | Electron 桌面应用 CDP 自动化 |
| agent-browser-slack | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | Slack 桌面端 CDP 自动化 |
| xiaohongshu-automation | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 小红书 CDP 自动发帖/搜索/评论/分析 |
| superset (desktop) | /Applications/Superset.app | 并行 AI Agent 管理器 (git worktree, Claude Code/Codex/OpenCode, Electron TUI) |
| claude-conductor (CLI) | npm global | Claude Code 文档框架 (CONDUCTOR.md/ARCHITECTURE.md/BUILD.md 自动生成) |
| cli-relay | bin/cli-relay, layers/L3-intelligence/skills/skills/ | 本地 AI CLI 代理服务器 (localhost:8317, OAuth+APIKey, 多 Provider 负载均衡/故障转移) |
| claude-to-im | ~/.claude/skills/ | Claude Code 远程 IM 桥接 (Telegram/Discord/飞书, 流式预览, 权限控制) |
| agentic-seo | ~/.claude/skills/ | SEO 分析（12 子技能 + 6 专家 Agent） |
| apify-competitor-intelligence | ~/.claude/skills/ | 竞品分析（via Apify） |
| apify-trend-analysis | ~/.claude/skills/ | 趋势追踪（Google/IG/FB/YT/TikTok, via Apify） |
| apify-ultimate-scraper | ~/.claude/skills/ | 通用 AI 爬虫（via Apify） |
| trends24 | ~/.claude/skills/ | Twitter/X 热搜趋势（Trends24） |
| x-tweet-fetcher | ~/.claude/skills/ | Twitter/X 推文抓取（无需 API Key，v1.6.3） |
| grok-search | ~/.claude/skills/ | xAI Grok 实时 Web + X/Twitter 搜索（server-side tools, 结构化 JSON + citations） |
| grok-twitter-search | ~/.claude/skills/ | xAI Grok Twitter 专用智能搜索 |
| tavily-search | ~/.claude/skills/ | Tavily API 搜索（AI 增强结果，深度搜索，域名过滤） |
| multi-search-engine | ~/.claude/skills/ | 17 引擎集成搜索（百度/微信/Google/DuckDuckGo/WolframAlpha，无需 API Key） |
| web-multi-search | ~/.claude/skills/ | 多引擎并行 Web 搜索 |
| wechat-article-search | ~/.claude/skills/ | 微信公众号文章全文 + 图片提取 |
| bestblogs-daily-digest | ~/.claude/skills/ | BestBlogs.dev 每日早报（7 API 并行，AI Top 10 筛选，3 种输出格式） |
| qiaomu-mondo-poster-design | ~/.claude/skills/ | 一句话生成大师级海报/封面设计（Mondo 风格，需 Nanobanana Pro API） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| wechat-decrypt | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | 微信 v4.0 数据库解密/消息搜索/实时监控（Windows） |
| vphone-cli | skills/shared/, ~/.claude/skills/, ~/.codex/skills/, ~/.gemini/skills/ | Apple Silicon 虚拟 iPhone（iOS 26 安全研究） |
| Skill | 目录 | 触发场景 |
|---|---|---|
| agent-fetch | .claude/skills/ | L4 HTTP 结构化数据提取（Readability/JSON-LD/Next.js/WordPress API/text-density，CSS selector，proxy） |
| scrapling-crawler | .claude/skills/ | 反爬虫/Cloudflare bypass 爬取（Scrapling v0.4.1，StealthyFetcher + Camoufox） |
| ghost-os-automation | .claude/skills/ | L3 macOS 桌面自动化（AX Tree 29 MCP tools，任意 macOS 应用交互，10-20x token 省） |
| playwright-interactive | ~/.claude/skills/ | Playwright 持久会话 UI 调试 (OpenAI curated, js_repl, Electron 支持) |
| playwright-codex | ~/.claude/skills/ | Playwright 浏览器自动化 (导航/表单/截图/数据提取) |
| chrome-devtools-mcp | MCP (user-level) | Google 官方 Chrome DevTools MCP -- 浏览器调试/性能分析/无障碍审计/网络监控 (CDP, Puppeteer) |
| screenshot-codex | ~/.claude/skills/ | 桌面/窗口/区域截图 (OpenAI curated) |
| security-best-practices | ~/.claude/skills/ | 安全最佳实践审查 (OpenAI curated, 语言/框架特定) |
| Skill | 目录 | 触发场景 |
|---|---|---|
| strategic-compact | ~/.claude/skills/ | 策略性 compact (hook 计数 50 次工具调用后建议, 逻辑边界压缩) |
| continuous-learning-v2 | ~/.claude/skills/ | 持续学习 v2.1 (项目隔离 instinct, 置信度评分, 自动进化为 skill) |
| search-first | ~/.claude/skills/ | 先搜索后编码 (调用 researcher agent 查找现有工具/库/模式) |
| verification-loop | ~/.claude/skills/ | 验证循环 (编辑后自动检查, 构建/lint/测试闭环) |
| enterprise-agent-ops | ~/.claude/skills/ | 企业级 Agent 运维 (可观测性/安全边界/生命周期管理) |
| skill-prompt-convert | ~/.claude/skills/ | Skill 转 Prompt 转换器 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| ecc-hooks | layers/L3-intelligence/skills/skills/ | ECC 全生命周期钩子引擎 (Pre/PostToolUse 拦截假死/自动格式化/防失忆) |
| postmortem (registry) | layers/L3-intelligence/skills | 事故尸检 + 回归触发器 |
| reflection (registry) | layers/L3-intelligence/skills | 阶段性反思与沉淀 |
| agent-eval-system (registry) | layers/L3-intelligence/skills | Agent 可审计评估框架 |
| Skill | 目录 | 触发场景 |
|---|---|---|
| self-improving-agent | layers/L3-intelligence/skills/skills/ | 自我进化/记忆增强 (捕获偏好与避免重复踩坑) |
| agent-teams-swarm (registry) | layers/L3-intelligence/skills | Agent Teams 蜂群编排 |
| multi-agent-orchestration (registry) | layers/L3-intelligence/skills/skills/ | 多智能体双脑协作规划 (PM2并发执行/Symphony/Harness) |
| conductor-orchestrator | layers/L3-intelligence/skills/skills/ | Multi-agent orchestration v3 (/go goal-driven, parallel workers, Board of Directors, 42 skills + 22 commands) |
| skill-group-loop | .claude/skills/ | 组合技能循环引擎 (Multi-Skill PDCA, 组内技能轮流执行直到质量门禁通过) |
| Skill | 目录 | 触发场景 |
|---|---|---|
| rpi-research | .claude/skills/ | Phase 1: 可行性研究（GO/NO-GO） |
| rpi-plan | .claude/skills/ | Phase 2: 实现规划（PM + UX + 工程） |
| rpi-implement | .claude/skills/ | Phase 3: 分阶段实现 + 验证门禁 |
| Capsule ID | 类型 | Tags |
|---|---|---|
| cli-supervisor | openclaw-skill | ops, tmux, agents |
| dna-inheritance | openclaw-skill | dna, skills, inheritance |
| openclaw-best-practices | openclaw-skill | openclaw, security, memory, model-routing |
| agentic-engineering-ceo-mode | openclaw-skill | agentic-engineering, closed-loop, guardrails |
# 按关键词搜索 skill
ai dna search "<keyword>"
# 查看注册表
cat layers/L3-intelligence/skills/skills/skills-registry.json | python3 -c "
import json, sys
for k, v in json.load(sys.stdin)['skills'].items():
if '<keyword>' in str(v): print(f'{k}: {v[\"name\"]}')"
# 列出所有 Claude Code skills
ls .claude/skills/
# 列出所有跨 Agent skills
ls ~/.gemini/skills/
| 文件 | 用途 |
|---|---|
| CLAUDE.md | 工作准则（单一事实源） |
| GLOSSARY.md | 术语表与定义库（快速查词） |
| TOOLBOX.md | 工具/接口/服务速查 |
| AGENTS.md | 角色协作规范 |
Maurice | maurice_wen@proton.me