# skills-hub.ai > The open registry for AI coding skills. Reusable SKILL.md files from Anthropic, Google, Microsoft, Cloudflare, Vercel, Stripe, Supabase, and 100+ other sources. Install in one command. Works with Claude Code, Cursor, OpenAI Codex CLI, Windsurf, GitHub Copilot, Cline, Gemini CLI, and any MCP-compatible tool. ## What is skills-hub.ai? skills-hub.ai is a free, open-source registry of AI coding skills. A "skill" is a SKILL.md file — a Markdown document with YAML frontmatter (name, description, version, category, platforms) and natural-language instructions. The Agent Skills standard is an open format adopted by 26+ AI coding tools. Skills make AI assistants dramatically better at specific tasks by giving them reusable, versioned instructions instead of re-explaining your code review process every session. Use the CLI to install: `npx @skills-hub-ai/cli install `. Use the MCP server to expose all installed skills in any MCP-compatible client: `npx @skills-hub-ai/mcp`. ## Who is it for? Developers and engineering teams using AI coding assistants. Anyone who's tired of re-explaining their project conventions, deploy process, code review checklist, or testing patterns to an AI every session. Anyone publishing reusable AI workflows for others to install. ## Key facts - Published skills across 23 categories - 100+ official upstream sources (Anthropic, Google, Microsoft, Vercel, Stripe, Cloudflare, Supabase, Sentry, Expo, Hugging Face, OpenAI, MongoDB, Redis, Elastic, Prisma, HashiCorp, Netlify, Deno, Render, and more) - Daily sync from upstream GitHub repos — never stale - Free and open-source. No paywalls, no premium tiers, no rate limits. - Open Agent Skills standard (SKILL.md) — portable across tools - MIT licensed CLI, MCP server, shared types, and skill parser ## How skills work A SKILL.md file: ```markdown --- name: code-review description: Thorough code review — checks correctness, security, performance, tests. version: 1.0.0 category: review platforms: - CLAUDE_CODE - CURSOR - CODEX_CLI --- When the user asks you to review code, follow these steps... ``` The CLI downloads SKILL.md to the right directory for your AI tool. The MCP server exposes installed skills as MCP prompts so they appear as slash commands in any compatible client. ## Supported AI tools - Claude Code (Anthropic) - Claude Desktop (Anthropic) - Cursor (Anysphere) - OpenAI Codex CLI - Windsurf (Codeium) - GitHub Copilot - Cline (formerly Claude Dev) - Continue.dev - Gemini CLI (Google) - Any MCP-compatible client ## Categories (23) Engineering: build, test, qa, review, deploy, docs, security, ux, analysis, productivity, integration, ops, combo, meta. Non-engineering: marketing, product, creative, data, business, healthcare, accessibility, game-dev, research. ## Core pages - Homepage: https://skills-hub.ai — landing, top skills, FAQ - Browse: https://skills-hub.ai/browse — filter skills by category, platform, source, tag, SDLC phase, quality score, security grade - Skill detail: https://skills-hub.ai/skills/ — instructions, install command, ratings, version history, related skills - Sources: https://skills-hub.ai/sources — 100+ official upstream sources - Source detail: https://skills-hub.ai/sources/ — all skills from one company/repo - Categories: https://skills-hub.ai/categories — 23 categories - Bundles: https://skills-hub.ai/bundles — curated multi-skill collections - Best-of: https://skills-hub.ai/best — programmatic "best X" pages - Leaderboard: https://skills-hub.ai/leaderboard — most-installed skills 24h/7d/30d/all-time - Tags: https://skills-hub.ai/tags — free-form tags - Blog: https://skills-hub.ai/blog — articles on SKILL.md, agent skills, AI infra - Authors: https://skills-hub.ai/authors — contributor profiles ## Platform landing pages - https://skills-hub.ai/claude-skills — Claude Code skills overview - https://skills-hub.ai/cursor-rules — Cursor rules library (.cursorrules and .mdc) - https://skills-hub.ai/codex-skills — OpenAI Codex CLI skills - https://skills-hub.ai/mcp-skills — skills via Model Context Protocol - https://skills-hub.ai/skill-md — the SKILL.md format ## Docs - https://skills-hub.ai/docs/getting-started — install your first skill in 30 seconds - https://skills-hub.ai/docs/cli — full CLI reference - https://skills-hub.ai/docs/integrations — wire up Claude Code, Cursor, Windsurf, Copilot, Cline, MCP - https://skills-hub.ai/docs/workflows — chain skills into pipelines ## CLI ```bash npx @skills-hub-ai/cli init # guided setup npx @skills-hub-ai/cli install # install one skill npx @skills-hub-ai/cli install --target cursor # for Cursor npx @skills-hub-ai/cli install # install everything in .skills.json npx @skills-hub-ai/cli search "code review" # search catalog npx @skills-hub-ai/cli publish ./SKILL.md # publish your own npx @skills-hub-ai/cli suggest # recommend skills for your project npx @skills-hub-ai/cli login # GitHub, Google, or email ``` ## MCP Server ```bash # Local stdio MCP server (Claude Code) claude mcp add skills-hub -- npx @skills-hub-ai/mcp # Remote MCP endpoint (any client) https://api.skills-hub.ai/mcp ``` The MCP server exposes search, skill-detail, source-listing, and category-browsing tools, plus every installed skill as an MCP prompt. ## Public REST API Base: https://api.skills-hub.ai/api/v1/ - `GET /skills` — list/search skills (cursor pagination) - `GET /skills/` — single skill detail with instructions - `GET /skills/count` — total published skill count (cached 1h) - `GET /skills?source=` — filter by upstream source - `GET /external-sources` — list all 100+ sources with counts - `GET /external-sources/` — source detail - `GET /categories` — list categories - `GET /tags` — list tags No auth required for read endpoints. Rate limit: 100 req/min unauthenticated. ## Comparison: skills vs Cursor rules vs system prompts - **Skills (SKILL.md)**: portable across tools, load on demand via slash command or matcher, versioned, installable from a registry. Best for recurring jobs (code review, deploy, refactor). - **Cursor rules (.cursorrules / .mdc)**: Cursor-specific, load on every conversation, scope to file globs. Best for ambient project context (stack, conventions, forbidden patterns). - **System prompts**: tool-specific, set in app settings, opaque to teammates. Best for personal preferences. Most teams use both rules and skills: rules for "always-on" context, skills for "on-demand" jobs. ## Publishing Three ways to publish: 1. **CLI**: write SKILL.md → `npx @skills-hub-ai/cli login` → `npx @skills-hub-ai/cli publish` 2. **Web**: https://skills-hub.ai/publish — describe what you want and AI generates name, description, category, tags, and instructions 3. **GitHub auto-sync**: register your repo as an external source and skills publish automatically on push Every public skill goes through an automated AI quality review (correctness, completeness, security) before appearing in the catalog. You get a creator dashboard showing installs, ratings, and referrer data. ## Organizations Teams can create organizations with GitHub auth. Org-scoped skills are private to members. Organizations get a private registry, member management, install analytics, and audit logs. ## Open standards - SKILL.md (Agent Skills standard) — open format, no vendor lock-in - MCP (Model Context Protocol) — Anthropic's open standard for AI tool connectivity - All registries indexable via the JSON Feed at https://skills-hub.ai/feed.json - Atom feed at https://skills-hub.ai/feed.xml ## Pricing Completely free. No premium tiers, no usage limits, no paywalls. Forever. ## License - CLI: MIT (`@skills-hub-ai/cli`) - MCP server: MIT (`@skills-hub-ai/mcp`) - Shared types: MIT (`@skills-hub-ai/shared`) - Skill parser: MIT (`@skills-hub-ai/skill-parser`) - Individual skills: license set by their author (most are MIT, Apache 2.0, or CC0) ## Repository - Registry (community-contributed SKILL.md files): https://github.com/tinh2/skills-hub-registry - npm org: https://www.npmjs.com/org/skills-hub-ai ## Contact - Web: https://skills-hub.ai - Support: https://skills-hub.ai/about - Privacy: https://skills-hub.ai/privacy - Terms: https://skills-hub.ai/terms