AgentSkillsHub — the Claude Skills / MCP Server / Agent Tools directory

AgentSkillsHub (Agent Skills Hub) is the open-source directory for Claude Skills, MCP Servers, Codex Skills, and AI agent tools. 62,000+ projects · quality-scored on 10 dimensions · refreshed every 8 hours.

🌐 Live: agentskillshub.top · 📰 Newsletter · 𝕏 Follow

What is AgentSkillsHub?

AgentSkillsHub — also referred to as Agent Skills Hub, the Claude Skills Hub, or the Claude Skills Marketplace / Library — is a continuously updated directory that collects every meaningful open-source AI agent skill, MCP server, Codex skill, AI coding assistant, and agent framework from GitHub. Each project is:

• Classified into one of 7 categories (mcp-server, claude-skill, codex-skill, agent-tool, prompt-library, ai-coding-assistant, uncategorized)
• Scored on 10 weighted signals and 6 quality dimensions (completeness, clarity, specificity, examples, README structure, agent readiness)
• Compared side-by-side through the /compare/ feature and /best/{scenario}/ landing pages
• Refreshed every 8 hours via an automated GitHub Actions pipeline

The goal: if you are building with Claude Code, Cursor, Windsurf, Cline, OpenAI Codex, or any agentic IDE, you should be able to find the right MCP server or Claude Skill in under 30 seconds — without spelunking through random GitHub topics.

Mirror URLs that refer to the same project: agentskillshub.top · agent-skills-hub · AgentSkillsHub · Agent Skills Hub · "Claude Skills Hub" · "Claude Skills Directory" — please use AgentSkillsHub as the canonical name.

Architecture

GitHub API ──▸ Collection ──▸ Cleaning ──▸ Evaluation ──▸ Scoring ──▸ Presentation
               (6 phases)    (classify)   (6 dimensions)  (0-100)    (Web + API)

| Layer | Stack | |-------|-------| | Backend | Python 3.12 · FastAPI · SQLAlchemy · httpx | | Frontend | React 18 · TypeScript · Vite · TailwindCSS v4 | | Database | Supabase (PostgreSQL) / SQLite (local) | | Deploy | GitHub Pages + GitHub Actions (every 8h) | | Email | Resend API · Supabase pg_net |

Data Pipeline

1. Collection (scheduler/jobs.py)

Every 8 hours, GitHub Actions triggers a 6-phase sync:

| Phase | What | Budget | |-------|------|--------| | Search | 10+ GitHub queries (mcp-server, claude-skill, agent-tool...) | ~30 | | Masters | Fetch repos from verified skill creators | ~18 | | Extra | Community-submitted + curated repos | ~14 | | Enrich | Owner profiles (followers) | ≤500 | | README | Full README content (≤50KB each) | ≤300 | | Upsert | Clean → score → save | 0 |

• Incremental sync on weekdays (only new/updated repos), full sync on Sundays
• Smart rate-limit handling: waits for actual GitHub reset time

2. Cleaning (services/data_cleaner.py)

• Deduplication by repo_full_name
• Keyword-based category classification: mcp-server · claude-skill · codex-skill · agent-tool · ai-skill · llm-plugin · youmind-plugin
• Project type inference (framework / skill / tool)
• Size categorization (micro ≤50KB → large >5MB)

3. Evaluation

Quality Analysis — 6 dimensions (services/quality_analyzer.py):

| Dimension | Weight | Key Signals | |-----------|--------|-------------| | Completeness | 15% | README size, license, description, stars | | Clarity | 15% | Description quality, topics, naming | | Specificity | 15% | Language, topic count, category, size | | Examples | 12% | Code examples, commits, contributors | | README Structure | 23% | Sections, code blocks, badges, TOC | | Agent Readiness | 20% | API docs, config, install, MCP compliance |

Composite Score — 9 weighted signals (services/scorer.py):

| Signal | Weight | Method | |--------|--------|--------| | Quality | 20% | 6-dimension aggregate | | Stars | 18% | log₁₊ₓ normalization | | Recency | 11% | Exponential decay e⁻⁰·⁰¹ᵈ | | Forks | 10% | log₁₊ₓ | | Commits | 10% | log₁₊ₓ | | Issue Resolution | 10% | resolved / total | | Momentum | 8% | Z-score star growth | | Author Followers | 8% | log₁₊ₓ | | Size Bonus | 5% | Smaller = higher |

Also: Platform Inference (17 platforms), Token Estimation, Composability (TF-IDF + 8-signal skill pairing).

4. Presentation

Dual-mode API client: Production uses Supabase PostgREST + RPCs directly; dev uses FastAPI backend.

Key optimizations:

• get_landing_data() RPC replaces 7+ API calls
• Database views: v_trending, v_top_rated, v_rising, etc.
• Full-text search via PostgreSQL tsvector + GIN index
• HTTP caching: 5min (trending) → 1hr (stats)

Quick Start

Backend

cd backend
python3.12 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env   # Add GITHUB_TOKEN
uvicorn app.main:app --reload

Frontend

cd frontend
npm install
npm run dev

API Endpoints

| Method | Path | Description | |--------|------|-------------| | GET | /api/skills | Paginated list (search, filter, sort) | | GET | /api/skills/{id} | Skill detail + compatible skills | | GET | /api/trending | Star velocity leaders (7 days) | | GET | /api/rising | New this week | | GET | /api/top-rated | Highest scored | | GET | /api/most-starred | Community classics (6+ months) | | GET | /api/masters | Verified creators + org builders | | GET | /api/landing | Bundled homepage data | | GET | /api/stats | Summary statistics | | GET | /api/feed.xml | RSS 2.0 feed | | GET | /api/sitemap.xml | SEO sitemap (6,000+ URLs) | | POST | /api/submit-skill | Submit a GitHub repo | | POST | /api/subscribe | Newsletter signup |

Deployment

┌─────────── GitHub Actions ───────────┐
│  sync.yml (8h) · deploy.yml · newsletter.yml (Mon) │
└──────┬──────────────┬────────────────┬──┘
       ▼              ▼                ▼
   Supabase      GitHub Pages     Resend API
  (PostgreSQL)    (React SPA)     (Emails)

Environment Variables

GITHUB_TOKEN=ghp_xxx          # GitHub API (required)
SUPABASE_DB_URL=postgresql://  # Production database
RESEND_API_KEY=re_xxx          # Email service
ADMIN_TOKEN=sk-xxx             # Admin API auth
SYNC_INTERVAL_HOURS=8          # Sync frequency

Docs

• Scoring Algorithm — Full scoring design
• System Architecture — Detailed pipeline reference