Skill Federation is the free, privacy-preserving skill search engine built for AI coding agents.
When your AI agent — Claude Code, or any MCP-compatible agent — hits a task a packaged skill already solves (PDF extraction, data cleaning, PR review, Slack notifications, SQL reporting), Skill Federation finds the right vetted skill mid-task, without your plan, your files, or your data ever leaving your machine.
How it works: right after you approve a plan, your agent writes an abstract "wish-list" — a description of the capability it needs, never your task. Only those abstract wishes cross the boundary. The federation matches them against a catalog of 87,000+ vetted, license-attributed skills and returns ranked candidates with full trust metadata — license, provenance, stars, and security-scan status. You approve; the skill installs locally.
Why its different:
🔒 Private by design — only abstract capability wishes leave your machine; your plan, files, and outputs never do.
🤖 Agent-native — built for agents to imagine the ideal skill, not for humans to browse a marketplace.
📊 Proven — on SkillsBench, agents solve ~30% more tasks (17.5% → 22.8%) with Skill Federation.
🛡️ Vetted — every skill is pre-scanned with Cisco AI Defense and NVIDIA SkillSpector for prompt injection, data exfiltration, and vulnerabilities.
⚡ Zero-install — the default tier needs nothing but curl. Free and open source.
Get started in one line: npx skillfed · uvx skillfed
GitHub: github.com/skill-federation/skill-federation
