anbeime/skill: A Self-Updating Store of 245 Agent Skills With a Built-In Validator

Finding a single agent skill is easy. Keeping a local library current — knowing what is out there, what changed overnight, and whether a SKILL.md file even parses — is the part that does not scale. anbeime/skill, a 7k-star repository, attacks that operational side: it maintains a synced, classified, exportable index of agent skills plus a validator you can run before you install anything.

Why This Skill Matters

The store has two layers, and the README's stats block reports the inventory precisely: 182 official skills crawled from VoltAgent's awesome-agent-skills repository, plus 63 local skills (27 core, 30 sub-skills, and 6 system built-ins), for 245 total. The official side aggregates skills from named teams — the README lists Anthropic, Vercel, Cloudflare, Trail of Bits, Hugging Face, Google Labs, Stripe, Supabase, Expo, Sentry, and Better Auth, with per-team counts. The local side is a set of curated Chinese-language skills for verticals including content publishing, video creation, e-commerce marketing, presentations, voice synthesis, and document analysis.

Two features make it more than a list. A GitHub Action re-crawls the upstream awesome list every 24 hours, so the index tracks new arrivals without manual work. And a bundled validator checks each skill's SKILL.md against the Agent Skills specification before you waste an agent session on a malformed one.

Installation

The store is a Python project you clone and run locally:

git clone https://github.com/anbeime/skill.git
cd skill
pip install -r requirements.txt

Real Workflow: Sync the Index, Then Audit a Skill Before Installing It

Pull the latest snapshot once, or run the daemon to keep it current:

python main.py --once

python main.py --daemon

The crawl writes two data files you can read directly: data/skills.json for official skills and data/local_skills.json for the local library. Prefer browsing? The README points at an online store at skill.vercel.app backed by the same data.

Before installing a skill into your agent, run the validator on it:

python tools/skill_validator.py validate skills/agent-team

The same tool reads a skill's properties as JSON and can generate an XML prompt for AI invocation:

python tools/skill_validator.py read-properties skills/agent-team --json

When you want the data for your own tooling, export the index:

python main.py --export skills.csv

Tips

• python main.py --stats prints the current data statistics — a quick way to see how much the last crawl picked up.
• The README documents JSON and CSV export as first-class features, so you can pipe the index into scripts rather than reading HTML.
• Backup zip packages cover the local skills; the README notes the six system built-ins (pptx, xlsx, pdf, docx, find-skill, skill-creator) ship without zip packages.
• The README includes docs on the Agent Skills specification and skill-writing best practices, plus a "create a skill in 30 seconds" quickstart — useful even if you never run the crawler.

When Not to Use This

The store manages an index and a local library; its README documents no one-command install into an agent runtime. You still copy or link a skill folder into your agent's skills directory yourself. And if you need exactly one well-known skill, the upstream repository is the primary source — the store earns its keep when you want bulk, freshness tracking, and pre-install validation.

See the leaderboard for more skills.