Curated cheat sheets for Claude, Cursor, and the rest

What it does This repository is a curated directory of over 1,400 agent skills—structured context files that teach AI coding assistants how to work with specific APIs, frameworks, and workflows. It aggregates official skill definitions published by engineering teams at Anthropic, Google, Stripe, Cloudflare, and dozens of other companies, alongside community submissions, and catalogs them for use across Claude Code, Cursor, GitHub Copilot, Gemini CLI, and similar tools. In practice, it functions as a cross-vendor card catalog where each entry tells an LLM how to behave like a domain expert.

The interesting bit The maintainers draw a hard line between hand-picked, team-tested skills and what they dismiss as “AI-slop” bulk-generated repositories. Because the collection is format-agnostic, a skill published by Stripe or Vercel can be adapted across multiple agent platforms instead of remaining locked to a single vendor’s ecosystem.

Key highlights

• Over 1,400 skills indexed, with a live badge tracking the current count.
• Official contributions from major teams including Anthropic, Google Labs, Vercel, Stripe, Cloudflare, Netlify, Sentry, Figma, and Hugging Face.
• Cross-platform compatibility spanning Claude Code, Codex, Gemini CLI, Cursor, GitHub Copilot, OpenCode, and Windsurf.
• Companion browsing site at officialskills.sh with per-skill documentation links.
• Explicitly curated by hand rather than scraped or auto-generated.

Caveats

• The repository is essentially a heavily formatted README and index; the actual skill files live in upstream repositories or on the external officialskills.sh platform.
• The README embeds prominent banner advertising for the maintainer’s own VoltAgent framework and includes sponsor slots, so the editorial curation and commercial promotion are intertwined.

Verdict Bookmark this if you want vendor-approved agent context without chasing scattered GitHub organizations. Look elsewhere if you need a standalone runtime or framework—this is a curated index, not an executable product.