Getting Started

What is SkillRepo?

SkillRepo is a registry for AI agent skills built on the open agent-skills standard. It provides a single source of truth where teams can publish, version, and manage the skills that power their AI coding assistants.

For developers, the skillrepo CLI keeps your local library in sync with the registry. A single npx skillrepo init validates your access key, detects your dev environment, and pulls your library to disk. After that, skillrepo update (or the optional Claude Code SessionStart hook installed during init) keeps the on-disk copy current. Your agent reads those files directly during a session — no per-activation network round-trip. External agents that don’t have filesystem access can read the same library through SkillRepo’s remote endpoint; see the MCP Reference for that path.

Creating an Account

Head to Sign up and create an account with your email or sign in with GitHub. The Publisher tier is free and includes unlimited published skills and 1 access key — plenty for an individual publisher getting started.

Team includes a 14-day trial; cards aren’t charged until the trial ends. Volume, contract, or self-hosted needs go through sales — see the pricing page.

Generating an Access Key

Joining a team (recommended)

If your team admin invites you via the Team page, the invitation email walks you through everything — click the accept link, sign in, and the email body explains how to run npx skillrepo init in your project. The CLI mints your personal access key on first run, writes it to ~/.claude/skillrepo/config.json (chmod 0600 on POSIX) and to .env.local, and pulls the team's library to disk. No manual key copy-and-paste.

Joining a team via the CLI

If a teammate has invited you to a team, you can let the CLI accept the invitation for you — even before clicking the email's accept link. Run npx skillrepo init in any project; the browser flow lists any pending invitations for your email alongside the accounts you already belong to. Picking the invitation accepts it and mints a key for that team in a single step. See the init command in the CLI Reference for the full flow.

Manual fallback

If you signed up directly (Publisher tier) or want to mint a key by hand, open the Connect page. Click Create key — the new sk_live_* value is shown once. Copy it and paste it into npx skillrepo init when prompted, or set SKILLREPO_ACCESS_KEY in your shell profile.

SkillRepo enforces one active access key per user. You’ll see a Rotate key CTA on Connect once you have a key — rotating revokes the existing key and mints a new one in a single transaction. See Access Keys in a Team for the full flow.

Connecting Your Dev Environment

The skillrepo CLI is the only setup you need. One command detects your dev environment, pulls your library to disk, and (in Claude Code) wires a SessionStart hook so future sessions auto-sync.

npx skillrepo init

# Pull the latest versions of every skill in your library
skillrepo update

# Add a skill to your library and pull it locally
skillrepo add @alice/pdf-tools

# See what's in your library
skillrepo list

# Search the public registry
skillrepo search "kubernetes"

# Fetch a single skill without modifying your library
skillrepo get @alice/pdf-tools

# Remove a skill from your library
skillrepo remove @alice/pdf-tools

Browsing and Discovering Skills

Browse the public catalog at Skills Catalog. Filter by category, sort by popularity, and view skill details before adding a skill to your library with skillrepo add @owner/name.

Understanding Skill Structure

Every skill is defined by a SKILL.md file that combines YAML frontmatter with markdown instructions. Here is a minimal example:

---
name: code-review
description: >
  Review code for bugs, security issues,
  and best practices.
version: 1.0.0
license: MIT
compatibility:
  - cursor
  - claude-code
  - jetbrains-ai
metadata:
  category: development
  tags:
    - code-quality
    - review
allowed-tools:
  - Read
  - Grep
  - Glob
---

# Code Review

You are an expert code reviewer. When activated,
review the provided code for:

1. **Bugs** - logic errors, off-by-one, null refs
2. **Security** - injection, auth issues, secrets
3. **Best practices** - naming, structure, DRY

Provide clear, actionable feedback with code
examples for each finding.