菜单
Use when starting work on a specific technology (typescript, react, python, vite, nextjs, etc), when user asks "find a skill for X", "search skills.sh", "best practices for X", or when you need procedural knowledge for a task. Searches the open agent skills ecosystem at skills.sh.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | find-skill-sh |
| description | Use when starting work on a specific technology (typescript, react, python, vite, nextjs, etc), when user asks "find a skill for X", "search skills.sh", "best practices for X", or when you need procedural knowledge for a task. Searches the open agent skills ecosystem at skills.sh. |
Search the open agent skills ecosystem for procedural knowledge.
Check local cache before searching online:
.claude/plugins/kuro/skills/_skills-sh-data/
Available cached skills:
vite - Vite build tool (read _skills-sh-data/vite/INDEX.md)vitest - Vitest testing framework (read _skills-sh-data/vitest/INDEX.md)vercel-react-best-practices - React best practices (read _skills-sh-data/vercel-react-best-practices/AGENTS.md)If the technology matches a cached skill, read the INDEX.md directly - no network needed.
Use WebSearch with site filter:
site:skills.sh {technology}
Examples:
site:skills.sh typescript - TypeScript skillssite:skills.sh react testing - React testing skillssite:skills.sh python fastapi - Python FastAPI skillssite:skills.sh nextjs - Next.js skillsShow user the top 3-5 results in a table:
| # | Skill | Source | Installs |
|---|---|---|---|
| 1 | typescript-advanced-types | wshobson/agents | 2.5K |
| 2 | typescript-best-practices | 0xbigboss/claude-code | 177 |
Let user select which skill to apply, or auto-select top result if only one is highly relevant.
Use WebFetch on the skill page URL:
https://skills.sh/{owner}/{repo}/{skill-name}
Example:
WebFetch("https://skills.sh/wshobson/agents/typescript-advanced-types", "Extract the complete skill content including best practices, patterns, and guidelines")
Extract and apply the skill's:
| Technology | Action |
|---|---|
| Vite | Read local: _skills-sh-data/vite/INDEX.md |
| Vitest | Read local: _skills-sh-data/vitest/INDEX.md |
| React | Read local: _skills-sh-data/vercel-react-best-practices/AGENTS.md OR search site:skills.sh react |
| TypeScript | Search: site:skills.sh typescript |
| Next.js | Search: site:skills.sh nextjs |
| Python | Search: site:skills.sh python |
| Testing | Search: site:skills.sh testing |
| API Design | Search: site:skills.sh api |
https://skills.sh/?q={query}https://skills.sh/{owner}/{repo}/{skill-name}https://skills.sh/ (37K+ skills)User: "I'm working on TypeScript, find relevant skills"
Agent workflow:
site:skills.sh typescripthttps://skills.sh/wshobson/agents/typescript-advanced-typesvercel-labs/agent-skills and anthropics/skills are high qualityUse when the user wants to toggle question mode — making Claude ask clarifying questions before making edits. Triggers on 'question_toggle', 'toggle questions', 'ask me before editing', 'stop asking questions', or when the user wants to control whether Claude asks for confirmation before changes.
Use when the user wants multiple polymathic agents to analyze a problem, review code, debate an idea, or provide diverse expert perspectives. Triggers on 'consult polymaths', 'ask the polymaths', 'get polymath opinions', 'polymath review', 'run it by the polymaths', 'what would the polymaths say', or any request for multi-perspective expert analysis. Also trigger when the user says 'consult-polymaths' or wants to 'task the polymaths'. This is different from /consult (which queries external LLM APIs) — this skill spawns internal polymathic thinking agents.
Use when the user wants to run comprehensive end-to-end testing of a web application. Triggers on 'run e2e tests', 'test this app end to end', 'test all user journeys', 'do a full test pass'. Launches parallel research sub-agents then uses agent-browser for testing.
Use when creating a structured implementation plan with parallel task decomposition for team execution. Triggers on 'k-plan', 'create a plan', 'plan this task', 'decompose this work', or when the user needs a detailed plan document with task breakdown, dependencies, and team assignment before implementation.
Use when starting an autonomous task execution loop with structured git commits tied to task IDs. Triggers on 'k-start-worker-loop', 'start worker loop', 'autonomous task loop', 'work through tasks automatically', or when the user wants to iterate through a task list with automatic verification and git commits per task.
Use when the user needs browser automation specifically using playwright-cli — navigating pages, filling forms, managing tabs, taking screenshots, recording traces, or mocking network requests. Triggers on 'use playwright', 'playwright-cli', 'open browser with playwright'.