anbeime/skill

🎯 What is Skill Store?

Skill Store is a curated repository of AI‑agent “skills” – reusable, self‑contained descriptions of what an AI agent can do (e.g., generate PPTs, analyse papers, run e‑commerce pipelines, synthesize speech, etc.).

The project automatically pulls the latest 182 official skills from the community‑maintained awesome‑agent‑skills collection and adds 63 hand‑picked Chinese‑language skills created by the maintainers. All skills are stored as JSON/CSV files and can be browsed locally or via the live web UI at https://skill.vercel.app.

Core ideas

• Skill = a tiny, declarative plugin that tells an AI model how to perform a specific task and what inputs/outputs it expects.
• Central catalogue so developers don’t have to reinvent the wheel for common workflows (document handling, video creation, e‑commerce automation, etc.).
• Automatic sync – a GitHub Action runs every 24 h, crawling the upstream awesome‑agent‑skills repo and updating the local JSON files.
• Export & analysis – the data can be dumped to JSON or CSV for downstream tooling or research.

---

⚙️ How to get started

# 1. Clone the repo
git clone https://github.com/anbeime/skill.git
cd skill

# 2. Install Python dependencies
pip install -r requirements.txt

# 3. Run an update (once or as a daemon)
python main.py --once          # one‑off refresh
python main.py --daemon        # background 24‑h updater

# 4. Inspect the catalogue
cat data/skills.json          # official skills
cat data/local_skills.json    # locally authored skills

The main.py script also supports:

• --stats – prints a quick summary of counts per category.
• --export <file> – writes the catalogue to CSV.
• Verbose logging with -v.

---

📚 What’s inside?

Official skills (182)

Sourced from top‑tier teams (Anthropic, Vercel, Cloudflare, Hugging Face, Google Labs, etc.) and cover domains such as:

• Document processing (docx, pdf, xlsx, pptx)
• Code‑review and security tooling
• Model training & dataset handling
• Cloud platform best‑practice guides

Local skills (63)

Hand‑crafted, mostly Chinese‑language, grouped into 14 categories:

Category	Example skill	What it does
Content creation	content-creation-publisher	End‑to‑end article writing & publishing
Video creation	video-creation-suite	Full video production pipeline
E‑commerce	ecommerce-full-pipeline	Automated 1688 data scrape → product listing → marketing video → fulfillment
Document handling (built‑in)	pdf, docx, xlsx, pptx	Parse, edit, and generate the respective file types
AI‑assistant tools	agent‑team, multi‑agent‑meeting	Coordination primitives for multiple agents
…	…	…

Each skill is rated with stars (⭐) in the README, indicating community popularity or maintainers’ confidence.

---

🛠️ Tooling included

• Skill validator (tools/skill_validator.py) – checks that a SKILL.md file follows the specification, can output the skill as an XML prompt for LLM calls, and can read properties as JSON.
• Template (skills/_template/) – starter skeleton for contributors to add new skills.
• Export scripts – JSON ↔ CSV conversion for analytics.

---

📊 Data format snapshot

{
  "skills": [
    {
      "name": "Anthropic/docx",
      "description": "Process Word documents",
      "link": "https://github.com/Anthropic/docx",
      "category": "Document",
      "source": "awesome-agent-skills",
      "crawled_at": "2026-02-02T17:07:33"
    }
  ],
  "total": 182,
  "updated_at": "2026-02-02T17:07:33"
}

Local skills follow a similar schema but include extra metadata such as core_skills, sub_skills, and API requirement breakdowns.

---

🚀 Who might use this?

• AI‑agent developers who need a ready‑made library of prompts and wrappers for common tasks.
• Product teams building internal assistants (e.g., a sales‑assistant or a document‑review bot) and want to plug in existing skills.
• Researchers studying the ecosystem of LLM‑agent capabilities; the JSON dump provides a tidy dataset.
• Educators & hobbyists looking for concrete examples of how to structure an LLM‑driven skill.

---

🤝 Contributing

1. Fork the repo.
2. Add a new skill under skills/ using the provided template.
3. Run the validator (python tools/skill_validator.py validate <skill>).
4. Open a PR – the maintainers will review and merge.

You can also submit a skill via the issue template linked in the README.

---

📌 Quick take‑away

Skill Store is not a model or a training framework; it is a catalogue of reusable, declarative AI‑agent capabilities that can be imported into any LLM‑orchestrated system. Its value lies in the automated syncing of a large public skill set and the addition of a focused Chinese‑language collection, all packaged with validation tools and export options for downstream use.