OpenAgentSkill

OpenAgentSkill is a skill registry and recommendation layer for AI agents. It is built so an agent can find, compare, and install a reusable skill automatically, with one selected skill, alternatives, policy decisions, audit notes, and an install plan returned before execution.

The registry maps a described task to a ranked skill using signals such as workflow fit, maintenance, stars, freshness, metadata, repo health, quality score, permission hints, risks, and readiness notes. The page also describes a trust profile for each candidate skill and an install review step that shows agent-safe next steps before execution. OpenAgentSkill says it can help with web scraping, coding agents, RAG and knowledge work, and browser automation, and it is presented as a way to turn scattered GitHub projects into ranked, auditable, install-ready capabilities.

Both humans and agents can use it. People can browse the index, while agents can call the registry through resolve, recommendation, and skill endpoints. The page names Codex, Claude Code, Cursor, MCP-compatible agents, and custom runtimes or internal runners as environments that can use it. Example API paths shown on the page include /api/agent/resolve, /api/agent/skills/crawl4ai, and /api/skills/crawl4ai/install?format=text.

OpenAgentSkill states that its registry includes 20,505 indexed skills, 860K+ downloads, and 104 agent surfaces.