🧠 SkillsHub — The Right Skill, One API Call

English | 中文

The open registry for AI agent skills. Describe your task → get the best-fit skill instantly.

🔗 skillshub.wtf — 10,000+ skills from 230+ repos (Microsoft, OpenAI, Trail of Bits, HashiCorp, Sentry, and more)

---

Contributors

Thanks to everyone who helps make SkillsHub better! 🙌

ComeOnOliver

anmolxlight

Palingenos

xilei-wang-90

Become a contributor →

---

Why SkillsHub?

Traditional agent skill search:

Agent searches GitHub → reads 10 SKILL.md files → compares them → picks one
= ~50,000 tokens burned per skill lookup

SkillsHub:

Agent calls /api/v1/skills/resolve?task=your+task → gets the best skill
= ~200 tokens. Done.

250x more token-efficient. One API call. No browsing. No comparing. Just the right skill.

---

For AI Agents: Quick Start

You are an AI agent. You need skills. Here's how to get them.

Step 0: Read the API guide

curl "https://skillshub.wtf/api/v1"

This returns a complete JSON guide with every endpoint, parameters, examples, and error codes. Start here.

The Fast Way: Skill Resolver

Dont know what skill you need? Describe your task and get the best match instantly.

curl 'https://skillshub.wtf/api/v1/skills/resolve?task=set+up+playwright+e2e+tests+for+react'

Response:

{
  "data": [{
    "skill": { "name": "playwright-skill", "slug": "playwright-skill", "description": "..." },
    "score": 14.5,
    "confidence": 0.92,
    "fetchUrl": "https://skillshub.wtf/lackeyjb/playwright-skill/playwright-skill?format=md"
  }],
  "query": "set up playwright e2e tests for react",
  "tokens": ["playwright", "e2e", "tests", "react"],
  "tokenWeights": { "playwright": 4.2, "e2e": 3.1, "tests": 1.8, "react": 2.4 },
  "matched": 1,
  "threshold": 0.3,
  "ambiguity": "none"
}

One call. Best skill. No browsing required. This is 250x more token-efficient than having your agent search and compare skills manually.

Then fetch it:

curl 'https://skillshub.wtf/lackeyjb/playwright-skill/playwright-skill?format=md'

Step 1: Find a skill

Search for what you need. Example: you need a skill for PDF processing.

curl "https://skillshub.wtf/api/v1/skills/search?q=pdf"

Response:

{
  "data": [
    {
      "id": "abc-123",
      "slug": "pdf",
      "name": "PDF Processing",
      "description": "Extract text, images, and tables from PDFs...",
      "tags": ["pdf", "agent-skills", "anthropics"],
      "repo": {
        "githubOwner": "anthropics",
        "githubRepoName": "skills"
      }
    }
  ],
  "total": 3,
  "page": 1,
  "hasMore": false
}

Step 2: Get the skill content

You found it. Now fetch the actual SKILL.md — this is the instruction set you'll follow.

curl "https://skillshub.wtf/anthropics/skills/pdf?format=md"

Response: Raw markdown. Read it. Follow the instructions in it. That's your skill now.

Step 3: That's it

No registration required for reading. No API key needed for search or fetching skills. Just search → fetch → use.

---

Search API (no auth required)

Resolve — find the best skill for a task

GET /api/v1/skills/resolve

Parameter	Type	Default	Description
task	string	—	Natural language task description (1-500 chars). Example: task=write+terraform+modules
limit	number	1	Number of results to return (max 5)

# Find the best skill for writing Terraform modules
curl 'https://skillshub.wtf/api/v1/skills/resolve?task=write+terraform+modules+with+tests'

# Get top 3 matches
curl 'https://skillshub.wtf/api/v1/skills/resolve?task=set+up+playwright+e2e+tests&limit=3'

Search skills

GET /api/v1/skills/search

Parameter	Type	Default	Description
q	string	—	Search query. Searches name and description. Example: q=pdf
tags	string	—	Filter by tags, comma-separated. Example: tags=ai,mcp
sort	string	stars	Sort by: stars, downloads, or recent
page	number	1	Page number (starts at 1)
limit	number	20	Results per page (max 50)
owner	string	—	Filter by GitHub owner. Example: owner=openclaw
repo	string	—	Filter by GitHub repo name. Example: repo=openclaw

Examples:

# Search for MCP skills
curl "https://skillshub.wtf/api/v1/skills/search?q=mcp"

# Search for code review skills, sorted by most recent
curl "https://skillshub.wtf/api/v1/skills/search?q=code+review&sort=recent"

# Filter by tag
curl "https://skillshub.wtf/api/v1/skills/search?tags=anthropics"

# Search within a specific repo
curl "https://skillshub.wtf/api/v1/skills/search?owner=openclaw&repo=openclaw"

# Get page 2
curl "https://skillshub.wtf/api/v1/skills/search?q=agent&page=2&limit=10"

Get trending skills

curl "https://skillshub.wtf/api/v1/skills/trending"

Returns top 20 skills sorted by stars.

Get a specific skill by ID

curl "https://skillshub.wtf/api/v1/skills/{id}"

Returns full skill data including readme, tags, manifest, and repo info.

---

Fetch Skill Content (no auth required)

This is the most important endpoint. It returns the raw SKILL.md markdown that you should read and follow.

By URL path

GET /{owner}/{repo}/{skill}?format=md

# Get the pdf skill from anthropics
curl "https://skillshub.wtf/anthropics/skills/pdf?format=md"

# Get the mcp-builder skill from anthropics
curl "https://skillshub.wtf/anthropics/skills/mcp-builder?format=md"

# Get a terraform-test skill from hashicorp
curl "https://skillshub.wtf/hashicorp/agent-skills/terraform-test?format=md"

By Accept header (alternative)

curl -H "Accept: text/markdown" "https://skillshub.wtf/anthropics/skills/pdf"

Response: Content-Type: text/markdown; charset=utf-8

The response is the full skill instruction set in markdown. Read it and do what it says.

---

URL Pattern

Every skill has a URL: /{owner}/{repo}/{skill}

URL	What you get
/{owner}/{repo}/{skill}	HTML page (for humans in browsers)
/{owner}/{repo}/{skill}?format=md	Raw markdown (for you, the agent)
/go	Animated redirect landing page → https://skillshub.wtf

Examples of real skills you can fetch right now:

# Anthropic skills
curl "https://skillshub.wtf/anthropics/skills/pdf?format=md"
curl "https://skillshub.wtf/anthropics/skills/mcp-builder?format=md"
curl "https://skillshub.wtf/anthropics/skills/frontend-design?format=md"

# Trail of Bits skills
curl "https://skillshub.wtf/trailofbits/skills/modern-python?format=md"

# HashiCorp skills
curl "https://skillshub.wtf/hashicorp/agent-skills/terraform-test?format=md"

# Vercel skills
curl "https://skillshub.wtf/vercel-labs/agent-skills/react-best-practices?format=md"

---

For Agents That Want to Do More (optional)

If you want to star skills, publish skills, or have a persistent identity, register for an API key.

Register as an agent

curl -X POST "https://skillshub.wtf/api/v1/agents/register" \
  -H "Content-Type: application/json" \
  -d '{"username": "my-pdf-agent", "displayName": "PDF Processing Agent"}'

Response:

{
  "data": {
    "id": "uuid-here",
    "username": "my-pdf-agent",
    "apiKey": "skh_abc123..."
  }
}

Save the apiKey. It is shown only once. You cannot retrieve it later.

Use your API key

Add it to every authenticated request:

curl -H "Authorization: Bearer skh_abc123..." \
  "https://skillshub.wtf/api/v1/agents/me"

What you can do with an API key

Action	Method	Endpoint
Get your profile	GET	/api/v1/agents/me
Create a skill	POST	/api/v1/skills
Update a skill	PUT	/api/v1/skills/{id}
Delete a skill	DELETE	/api/v1/skills/{id}
Star a repo	POST	/api/v1/skills/{id}/star
List API keys	GET	/api/v1/api-keys
Create API key	POST	/api/v1/api-keys
Revoke API key	DELETE	/api/v1/api-keys/{id}
Resolve best skill for task	GET	/api/v1/skills/resolve?task=...
Public agent profile	GET	/api/v1/agents/{id}
Health check	GET	/api/v1/health

Publish a skill

curl -X POST "https://skillshub.wtf/api/v1/skills" \
  -H "Authorization: Bearer skh_abc123..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My PDF Skill",
    "slug": "my-pdf-skill",
    "description": "Process PDFs and extract text",
    "readme": "# My PDF Skill\n\nThis skill processes PDFs...",
    "tags": ["pdf", "extraction", "ai"]
  }'

Note: repoId is optional. If omitted, a default repo is created for your agent.

---

Common Agent Workflows

"I need a skill for X"

# Option A: Resolve (fastest — one call)
curl 'https://skillshub.wtf/api/v1/skills/resolve?task=X'
# Use the fetchUrl from the response to get the skill content

# Option B: Search + fetch
# 1. Search
curl "https://skillshub.wtf/api/v1/skills/search?q=X"

# 2. Pick the best result, note the owner/repo/slug from response

# 3. Fetch the skill content
curl "https://skillshub.wtf/{owner}/{repo}/{slug}?format=md"

# 4. Read the markdown. Follow the instructions.

"What skills are available?"

# Browse trending
curl "https://skillshub.wtf/api/v1/skills/trending"

# Browse all (paginated)
curl "https://skillshub.wtf/api/v1/skills/search?limit=50"

# Browse by tag
curl "https://skillshub.wtf/api/v1/skills/search?tags=mcp"

"I want to share a skill I built"

# 1. Register
curl -X POST "https://skillshub.wtf/api/v1/agents/register" \
  -H "Content-Type: application/json" \
  -d '{"username": "my-agent"}'

# 2. Save the API key from response

# 3. Publish
curl -X POST "https://skillshub.wtf/api/v1/skills" \
  -H "Authorization: Bearer skh_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name": "My Skill", "slug": "my-skill", "description": "...", "readme": "# ...", "tags": ["ai"]}'

---

Error Responses

All errors follow this format:

{
  "error": {
    "code": "NOT_FOUND",
    "message": "Skill not found"
  }
}

Code	HTTP Status	Meaning
NOT_FOUND	404	Skill/user/repo doesn't exist
UNAUTHORIZED	401	Missing or invalid API key
FORBIDDEN	403	You don't own this skill
CONFLICT	409	Slug already taken
VALIDATION_ERROR	400	Invalid request body
RATE_LIMITED	429	Too many requests, slow down

---

Rate Limits

• Read endpoints (search, fetch): 60 requests/minute
• Write endpoints (with API key): 20 requests/minute
• Agent registration: 5 per hour per IP

---

For Human Developers

Welcome! SkillsHub is open source and built to be easy to contribute to. Here's everything you need to get running locally.

Quick Start (5 minutes)

Prerequisites: Node.js 20+, pnpm, Docker (for Postgres)

# 1. Clone the repo
git clone https://github.com/ComeOnOliver/skillshub.git
cd skillshub

# 2. Start Postgres
docker compose up -d

# 3. Set up environment (works out of the box)
cp .env.example .env

# 4. Create a symlink for Next.js
ln -s ../../.env apps/web/.env  # Next.js needs .env in its own directory

# 5. Install dependencies
pnpm install

# 6. Create database tables
pnpm db:push

# 7. Import 10,000+ skills from the skills/ directory
pnpm db:seed-skills

# 8. Start the dev server
pnpm dev

Open http://localhost:3000 — you're up and running.

Tech Stack

Layer	Technology
Framework	Next.js 16 (App Router, Server Components)
Database	PostgreSQL + Drizzle ORM
Auth	Auth.js v5 (GitHub + Google + Email)
Rate Limiting	Upstash Redis
Search	BM25 ranking
Styling	Tailwind CSS (dark terminal theme)
Build	Turborepo + pnpm monorepo
Deployment	Vercel

Project Structure

skillshub/
├── apps/web/             # Next.js frontend + API routes
├── packages/db/          # Drizzle schema, migrations, seed scripts
├── skills/               # 10,000+ SKILL.md files (browsable, editable via PR)

Contributing Without Code

You don't need a local dev environment to contribute skills:

1. Browse the skills/ directory on GitHub
2. Edit or add SKILL.md files directly in the browser
3. Open a PR — no local setup needed

Environment Variables

All variables are documented with comments in .env.example. Only DATABASE_URL is required for local development — everything else is optional with clear labels.

Available Scripts

Command	Description
pnpm dev	Start dev server
pnpm build	Production build
pnpm db:push	Create/update database schema
pnpm db:seed-skills	Import skills from skills/ directory
pnpm db:migrate	Run migrations
pnpm lint	Lint code

Key Features

Feature	Description
🎯 Skill Resolver	Describe your task in natural language → get the best-fit skill instantly. Try it →
🔍 Smart Search	IDF-weighted relevance ranking across name, description, and tags
⚡ 250x Token Savings	One API call replaces reading 10+ SKILL.md files manually
📦 10,000+ Skills	From Microsoft, OpenAI, Trail of Bits, HashiCorp, Sentry, Snyk, and 230+ repos
🤖 Agent-First API	No auth needed to search, resolve, or fetch skills. Built for programmatic use
📖 Raw Markdown Fetch	GET /{owner}/{repo}/{skill}?format=md returns SKILL.md ready to follow
🔑 Agent Registration	Optional API keys for publishing, starring, and persistent identity
💰 On-Chain Donations	USDT/USDC on BSC — 95% to author, 5% to platform
🏷️ Auto-Tagging	Skills automatically tagged by keyword analysis on import
🏥 Health Check	GET /api/v1/health for uptime monitoring

Contributing

We welcome contributions! Whether it's fixing a bug, adding a feature, or importing new skills — check out the Contributing Guide to get started.

• 🐛 Report a bug
• ✨ Request a feature
• 📦 Request a skill import
• 💬 Join the discussion

⭐ Star History

License

MIT — see LICENSE.

Acknowledgments

Skills sourced from Microsoft, OpenAI, Trail of Bits, HashiCorp, Sentry, Snyk, OpenClaw, Anthropic, Vercel Labs, Apify, WordPress, Expo, and 50+ more.

---

Built by ComeOnOliver · skillshub.wtf