Token导航 LogoToken导航TokenDH.com
研究检索操作浏览器github未标认证来源可访问clear审计提醒

skill-finder技能查找器

Agent Skill

skill-finder 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

474

周安装

19

GitHub Stars

公开资料未说明

下载量

154
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

3

许可证

MIT

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:skill-finder(技能查找器)
来源仓库:https://github.com/dangogit/skill-finder
仓库路径:skills/skill-finder
安装命令:
npx skills add https://github.com/dangogit/skill-finder --skill skill-finder
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/dangogit/skill-finder --skill skill-finder

简介

技能查找器用于智能发现和安装skills.sh上的技能,提供基于自然语言的技能搜索功能。

  • 适用于需要快速找到合适技能来完成任务的开发者。
  • 能自动检测宿主环境(Claude Code、Cursor等),解析用户意图并匹配相关技能。
  • 安装前请确认权限范围和维护状态,注意是否会触发联网、命令执行或文件读写操作。
  • skill-finder 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Skill Finder

The smartest way to discover and install skills from skills.sh.

Trigger Phrases

  • /skill-finder
  • /skill-finder [description]
  • "find a skill for..."
  • "is there a skill for..."
  • "what skill should I use for..."

Process Overview

User Request
     │
     ▼
┌─────────────────────────┐
│ 0. Check Installed      │ ← Don't recommend what user already has
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 1. Detect Agent         │ ← Auto-detect Claude Code, Cursor, etc.
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 2. Parse Intent         │ ← Extract keywords, map to search terms
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 3. Targeted Search      │ ← Search with specific keywords + categories
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 4. Context Check        │ ← Only if needed for disambiguation
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 5. Rank & Compare       │ ← Quality signals + comparison mode
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 6. Recommend            │ ← Present with quality badges
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 7. Install              │ ← Global or project, auto-detect agent
└─────────────────────────┘
     │
     ▼
┌─────────────────────────┐
│ 8. Related Skills       │ ← Suggest complementary skills
└─────────────────────────┘

Step 0: Check Already Installed Skills

Before searching, check what's already installed:

# Check common skill locations
ls -la .claude/skills/ 2>/dev/null
ls -la .cursor/rules/ 2>/dev/null
ls -la .agents/skills/ 2>/dev/null
ls -la ~/.claude/skills/ 2>/dev/null  # Global

If the skill user needs is already installed:

You already have **{skill-name}** installed at `{path}`.

→ Want me to invoke it now, or find alternatives?

If outdated version is installed:

You have **{skill-name}** installed, but there may be updates.

→ Want me to check for updates, or use the current version?

Step 1: Detect Agent

Auto-detect which AI agent the user is using:

Check ForAgentInstall Flag
.claude/ directory or CLAUDE.mdClaude Code-a claude-code
.cursor/ directoryCursor-a cursor
.github/copilot/ directoryGitHub Copilot-a copilot
.cline/ directoryCline-a cline
.codex/ directoryCodex-a codex
.continue/ directoryContinue-a continue

Store detected agent for installation step.

If multiple detected or none detected:

  • Default to claude-code in Claude Code context
  • Or ask user which agent they're using

Step 2: Parse Intent & Extract Keywords

From user request, extract:

  • Primary topic (e.g., "payments", "authentication", "testing")
  • Specific technologies mentioned (e.g., "Stripe", "Supabase", "Playwright")
  • Action type (e.g., "add", "build", "improve", "debug")

Map to search keywords:

User SaysSearch Keywords
payments, billing, checkoutstripe, payment, billing, checkout, subscription
auth, login, sign inauth, authentication, login, oauth, jwt, session, clerk, better-auth
tests, testing, TDDtest, testing, playwright, jest, vitest, e2e, tdd, qa
database, DB, datapostgres, supabase, prisma, mongodb, sql, database, migration
AI, LLM, RAG, embeddingsrag, llm, embedding, ai-sdk, mcp, vector, openai, anthropic
frontend, UI, Reactreact, nextjs, vue, svelte, frontend, tailwind, ui, component
backend, API, serverfastapi, express, nestjs, backend, api, server, node
DevOps, CI/CD, deploydocker, k8s, kubernetes, github-actions, terraform, deploy, ci
email, notificationsemail, resend, notification, messaging, smtp
docs, PDF, Excelpdf, docx, xlsx, pptx, document, report
video, animationremotion, video, animation, threejs, canvas, webgl
scraping, automationbrowser, scrape, automation, crawl, puppeteer, playwright
marketing, SEO, growthseo, marketing, cro, copywriting, analytics, conversion
mobile, React Native, iOSreact-native, ios, android, swiftui, mobile, flutter, expo
securitysecurity, auth, encryption, vulnerability, audit, secrets

Step 3: Targeted Search

IMPORTANT: Use the CLI search command first to get proper skill references:

npx skills search {keyword} --limit 10

This returns skills in the correct format: author/repo@skill-name

Understanding the format:

  • author/repo@skill-name - Full skill reference from search results
  • author/repo - What you pass to npx skills add (installs all skills in that repo)

For WebFetch fallback (if CLI unavailable):

WebFetch: https://skills.sh/?search={primary_keyword}
Prompt: List matching skills with:
        - Full reference (author/repo@skill-name format)
        - Install count
        - Description
        Focus on skills matching: {specific_technologies_mentioned}

If user request is vague (just "/skill-finder"):

  • Ask: "What are you trying to accomplish?"
  • Wait for response before searching

Step 4: Context Check (Only If Needed)

Skip context gathering if:

  • User specified exact technology (e.g., "Stripe", "Playwright")
  • Search returned one clear winner
  • User request is unambiguous

Gather context if:

  • Multiple similar skills found (need stack to disambiguate)
  • User request is generic (e.g., "database best practices")
  • Need to verify compatibility

Quick context check:

FileWhat to Extract
package.jsonFramework (react, next, vue), key deps, has supabase/stripe/etc
pyproject.tomlPython framework (fastapi, django, flask)
go.modGo project confirmation
Cargo.tomlRust project confirmation
.env.exampleIntegrated services (Stripe, Supabase, etc.)

Summarize as: Stack: {framework}, {language}, {key_dependencies}


Step 5: Rank with Quality Signals

Quality indicators to factor into ranking:

SignalHow to IdentifyWeight
Official/VerifiedAuthor is stripe/, anthropics/, vercel/, supabase/, resend/High
High Installs1000+ installsMedium
Stack MatchSkill matches user's detected tech stackHigh
Specific MatchSkill name contains user's exact queryHigh

Official/Verified Authors (show ⭐ badge):

  • stripe/ - Stripe official
  • anthropics/ - Anthropic official
  • vercel/ or vercel-labs/ - Vercel official
  • supabase/ - Supabase official
  • resend/ - Resend official
  • remotion-dev/ - Remotion official
  • better-auth/ - Better Auth official
  • callstackincubator/ - Callstack (React Native experts)
  • cloudai-x/ - Three.js skills
  • trailofbits/ - Security experts

Step 5b: Comparison Mode

When multiple skills are similarly relevant, show comparison:

## Comparing {Category} Skills

| Feature | {skill-1} | {skill-2} | {skill-3} |
|---------|-----------|-----------|-----------|
| Installs | {count} | {count} | {count} |
| Official | ⭐ Yes | No | No |
| Your Stack | ✅ Match | ✅ Match | ❌ Different |
| Focus | {focus} | {focus} | {focus} |

**Recommendation:** {skill-name} ({reason})

→ Want details on any of these, or should I install the recommendation?

Trigger comparison mode when:

  • Top 2-3 skills have similar relevance scores
  • User asks "what's the difference between..."
  • Skills serve same purpose but different approaches

Safety Checks

Before recommending any skill, verify:

1. Author Verification

✅ TRUSTED - Official/Verified authors (install without warning):
   stripe/, anthropics/, vercel/, vercel-labs/, supabase/, resend/,
   remotion-dev/, better-auth/, callstackincubator/, trailofbits/

⚠️ COMMUNITY - Other authors (show warning before install):
   "This skill is from a community author. Review before installing."

2. Typosquatting Check

Before recommending, verify skill name doesn't look like a typo of a popular skill:

  • stripe-best-pracitces → suspicious (typo of stripe-best-practices)
  • react-best-prctices → suspicious
  • supabase-postgers → suspicious

If suspicious: Don't recommend, or warn user explicitly.

3. Install Confirmation

Always show before installing:

**About to install:** {skill-name}
**Author:** {author}
**Source:** https://github.com/{author}/{repo}

⚠️ Installing skills runs code on your machine.
   Review the source if you don't trust the author.

→ Proceed with installation? (y/n)

4. Link to Source

Always provide GitHub link so user can review:

📂 **View source:** https://github.com/{author}/{repo}/blob/main/skills/{skill}/SKILL.md

Step 6: Present Recommendations

Format with quality badges:

## Recommended Skills

Based on your {stack} project and goal to {userIntent}:

1. **{skill-name}** by {author} ({installs} installs) ⭐ Official
   {Why this matches - be specific about the fit}

2. **{skill-name}** ({installs} installs)
   {Why this is relevant, any trade-offs vs #1}

3. **{skill-name}** ({installs} installs)
   {When you'd choose this instead}

→ Want me to install any of these?

Quality badges:

  • ⭐ = Official/Verified author
  • 🔥 = Trending (high installs)
  • ✅ = Perfect stack match

Step 7: Installation

CRITICAL: The CLI uses author/repo format, NOT author/skill-name!

From search results like jezweb/claude-skills@tanstack-query:

  • jezweb/claude-skills = the repo (use this for install)
  • tanstack-query = specific skill name in that repo

⚠️ IMPORTANT: Installing a repo installs ALL skills in it!

Some repos contain many skills:

  • jezweb/claude-skills has 90+ skills (probably too many)
  • vercel-labs/next-skills has 3 skills (focused)

Prefer focused repos with fewer skills relevant to user's need.

If a large repo is the only option:

  1. Warn the user: "This repo contains {N} skills. Install anyway?"
  2. Suggest browsing: https://skills.sh/{author}/{repo} to see all skills
  3. After install, suggest removing unneeded skills

Ask about installation scope:

Install **{skill-name}** from `{author}/{repo}` to:
1. **This project** (./agents/skills/) - Recommended
2. **Globally** (all your projects)

→ Which do you prefer? (default: this project)

⚠️ Note: This will install {N} skills from this repo.

Installation commands:

# Project-level (default) - uses author/repo, NOT skill name
npx skills add {author}/{repo} -a {detected_agent} -y

# Global installation
npx skills add {author}/{repo} -a {detected_agent} -y -g

After successful installation:

✓ Installed skills from **{author}/{repo}** to {install_path}

Agent: {detected_agent}
Scope: {project|global}
Skills installed: {list of skill names}

Step 8: Suggest Related Skills

After installing, suggest complementary skills:

Installed SkillRelated Skills to Suggest
stripe-*webhook-patterns, subscription-management, billing-automation
*-auth, authenticationsession-management, secrets-management, security-*
supabase-*row-level-security, database-migration, sql-optimization
playwright-*, testingtest-driven-development, e2e-testing-patterns, qa-test-planner
react-*react-testing, react-performance, react-state-management
nextjs-*vercel-deployment, nextjs-app-router-patterns
fastapi-*async-python-patterns, python-testing-patterns
docker-*k8s-*, deployment-pipeline-design
ai-sdk, rag-*embedding-strategies, mcp-builder, llm-evaluation
remotion-*video processing skills, animation skills

Format:

**You might also need:**
- **{related-skill-1}** - {why relevant to what they just installed}
- **{related-skill-2}** - {why relevant}

→ Want me to install any of these too?

Step 9: Check for Updates (Optional)

When user has skills installed, offer update check:

npx skills check

If updates available:

📦 **Updates Available**

| Skill | Current | Latest | Changes |
|-------|---------|--------|---------|
| stripe-integration | 1.0.0 | 1.2.0 | New webhook patterns |
| react-best-practices | 2.1.0 | 2.3.0 | React 19 support |

→ Want me to update these?

Update command:

npx skills update

Error Handling

No Matches Found

I couldn't find skills that directly match "{user_request}".

**Suggestions:**
- Try different keywords (e.g., "auth" instead of "login system")
- Be more specific (e.g., "Supabase auth" instead of "auth")
- Browse skills.sh directly: https://skills.sh/

**Popular categories you might explore:**
- Authentication: `better-auth`, `clerk-auth`, `supabase-auth`
- Testing: `playwright-skill`, `webapp-testing`
- Payments: `stripe-best-practices`

What would you like to try?

skills.sh Unreachable

I'm having trouble reaching skills.sh right now.

**Options:**
1. Try again in a moment
2. Browse directly: https://skills.sh/
3. Tell me what you need and I'll help without a skill

What would you like to do?

Installation Failed

Installation failed: {error_message}

**Try manually:**

npx skills add {author}/{skill-name}


**Or browse the skill:** [https://skills.sh/{author}/{skill-name}](https://skills.sh/%7Bauthor%7D/%7Bskill-name%7D)

**Common fixes:**

- Run `npm cache clean --force` and retry
- Check your internet connection
- Try installing globally with `-g` flag

Skill Already Installed

**{skill-name}** is already installed at `{path}`.

**Options:**
1. Invoke it now for your current task
2. Check for updates
3. Find alternative skills
4. Reinstall (overwrite)

What would you like to do?

Examples

Fast Path (Clear Intent)

User: /skill-finder Stripe payments

Agent: Let me search for Stripe skills...
       [Runs: npx skills search stripe --limit 10]

       ## Recommended Skills

       Based on your goal to add Stripe payments:

       1. **stripe-best-practices** from `stripe/ai-skills` (354 installs) ⭐ Official
          Official Stripe patterns - checkout, webhooks, subscriptions
          Repo contains 2 skills (focused)

       2. **stripe-integration** from `wshobson/agents` (349 installs)
          End-to-end integration patterns
          ⚠️ Repo contains 15 skills

       → Want me to install any of these?

User: Install #1

Agent: Install from `stripe/ai-skills` to:
       1. This project (./agents/skills/) - Recommended
       2. Globally (all your projects)

       This repo contains 2 skills - both Stripe-related.

User: 1

Agent: [Runs: npx skills add stripe/ai-skills -a claude-code -y]

       ✓ Installed 2 skills from **stripe/ai-skills** to .agents/skills/

       Skills installed:
       - stripe-best-practices
       - stripe-webhooks

       → Want me to apply these to your current task?

With Comparison Mode

User: /skill-finder authentication

Agent: I found several auth skills. Here's a comparison:

       ## Comparing Authentication Skills

       | Feature | better-auth | clerk-auth | supabase-auth |
       |---------|-------------|------------|---------------|
       | Installs | 3,373 | 129 | 142 |
       | Official | ⭐ Yes | No | No |
       | Self-hosted | ✅ | ❌ Cloud | ✅ |
       | Social login | ✅ | ✅ | ✅ |
       | Your Stack | ✅ Next.js | ✅ Next.js | ⭐ You use Supabase |

       **Recommendation:** supabase-auth (best fit - you already use Supabase)

       → Want me to install supabase-auth, or details on another?

Already Installed

User: /skill-finder React testing

Agent: You already have **webapp-testing** installed at `.claude/skills/webapp-testing/`.

       **Options:**
       1. Invoke it now for your React testing needs
       2. Find additional/alternative testing skills
       3. Check for updates

       What would you like to do?

User: 2

Agent: ## Additional Testing Skills

       Since you have webapp-testing, here are complementary skills:

       1. **react-testing** (412 installs)
          React-specific testing patterns (RTL, hooks testing)

       2. **test-driven-development** (1,289 installs) ⭐ Popular
          TDD methodology and workflow

       → Want me to install any of these?

Update Check

User: /skill-finder --check-updates

Agent: 📦 **Checking your installed skills...**

       | Skill | Status |
       |-------|--------|
       | stripe-best-practices | ✅ Up to date |
       | webapp-testing | 🔄 Update available (1.2.0 → 1.3.0) |
       | react-best-practices | ✅ Up to date |

       → Want me to update webapp-testing?

Performance Tips

  1. Be specific — "Stripe checkout" finds better matches than "payments"
  2. Mention your stack — "React Native auth" narrows results
  3. One topic at a time — Split "auth and payments" into two searches
  4. Check installed first — Avoid duplicates with /skill-finder --installed
  5. Use official skills — Look for the ⭐ badge for verified quality

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

04

需要参考平台分布和安装热度时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Claude Code

28.91%
按下载量换算45

OpenCode

21.76%
按下载量换算34

Cursor

18.56%
按下载量换算29

Codex

12.08%
按下载量换算19

Antigravity

8.16%
按下载量换算13

Gemini CLI

3.53%
按下载量换算5

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。

来源信息

继续浏览同类 Skills