# SkillRepo

> SkillRepo is the open distribution and governance layer for AI agent skills. One graded library lands where every agent already reads — Claude Code, Cursor, Windsurf, VS Code + Copilot, Gemini CLI, Codex CLI, and Cline — and stays current, so a team runs the versions it approved instead of a different copy on every laptop. Pin the approved version per repo, see which repos have drifted, and recall a bad version across your organization. A solo developer starts on one library across every agent, always current; a team lands everyone on the same approved set.

SkillRepo implements the [agentskills.io](https://agentskills.io/specification) specification. Skills are reusable instruction bundles that AI agents load on demand; SkillRepo delivers them where each agent reads and governs which version runs where.

## What SkillRepo does

SkillRepo owns two category terms — Skill Distribution and Skill Governance — told through three value pillars:

- Distribute — one library reaches every agent and stays current.
- Govern — approved-version pins, per-repo scoping, drift visibility, organization-wide recall.
- Trust — a two-layer A–F safety grade, provenance, and verified GitHub identity on what you pull; your own skills private by default.

Grades are signals, not certification — review a skill before you run it. 36.8% of public skills contain a security flaw (Snyk ToxicSkills, February 2026: https://snyk.io/blog/toxicskills-malicious-ai-agent-skills-clawhub/).

## Core
- [Skills catalog](https://skillrepo.dev/skills): Browse all published skills
- [Leaderboard](https://skillrepo.dev/leaderboard): Top-rated skills and publishers
- [Documentation](https://skillrepo.dev/docs): Publishing, getting started, MCP reference
- [Connect an MCP client](https://skillrepo.dev/docs/mcp-clients): Connect Claude Desktop (Cowork), ChatGPT, or any MCP client to your library via a custom connector
- [Pricing](https://skillrepo.dev/pricing): Plans and limits
- [About](https://skillrepo.dev/about): Mission and context

## Solutions
- [For Teams](https://skillrepo.dev/for-teams): One governed library — the versions you approve, in every repo — with any drift in plain sight.
- [For Publishers](https://skillrepo.dev/for-publishers): Publish once. Get found, attributed, distributed.
- [For Agents](https://skillrepo.dev/for-agents): Pull the right skills at runtime, in every dev environment SkillRepo supports.
- [Skillsets](https://skillrepo.dev/docs/skillsets): Scope a team library to what each repo needs via skillrepo.json — at the versions you approve, with per-repo compliance you can see.

## Compare
- [SkillRepo vs git](https://skillrepo.dev/vs/git): Git is where your skills live; SkillRepo keeps your whole team on the current, approved set in every agent, and shows which version each repo runs.
- [SkillRepo vs skills.sh](https://skillrepo.dev/vs/skills-sh): skills.sh finds and installs a skill; SkillRepo keeps a team on one approved version of it, current in every dev environment.

## Specification
- [agentskills.io specification](https://agentskills.io/specification): The standard SkillRepo implements

## Full catalog
- [llms-full.txt](https://skillrepo.dev/llms-full.txt): All public skills with descriptions, capabilities, and URLs — designed for LLM ingestion

## For agents
- [AGENTS.md](https://skillrepo.dev/AGENTS.md): Short guide for coding agents working with SkillRepo
- [Platform MCP endpoint](https://skillrepo.dev/api/mcp): Account-scoped MCP server (auth: `sk_live_*` access keys)
- [Catalog MCP endpoint](https://skillrepo.dev/api/v1/catalog-mcp): Public discovery MCP server (auth: `ck_live_*` catalog keys via hello@skillrepo.dev)
- [Catalog REST API](https://skillrepo.dev/api/v1/catalog/): Read-only HTTP surface for discovering published skills
- [MCP discovery](https://skillrepo.dev/.well-known/mcp.json): Machine-readable listing of MCP endpoints (provisional schema)