SkillNet

Open Infrastructure for Creating, Evaluating, and Connecting AI Agent Skills

Search 400,000+ community skills · One-line install · Auto-create from repos / docs / logs
5-dimension quality scoring · Semantic relationship graph

⭐ If you like our project, please give us a star on GitHub for the latest updates!

SkillNet is an open-source platform that treats AI agent skills as first-class, shareable packages — like npm for AI capabilities. It provides end-to-end tooling to search, install, create, evaluate, and organize skills, so agents can learn from the community and continuously grow.

• We conducted experiments on ALFWorld, WebShop, and ScienceWorld. Reproduction scripts are provided here.

📢 News

• 🤖 [2026-03-26] JiuwenClaw × SkillNet Integrated! — JiuwenClaw now natively integrates SkillNet as its built-in skill marketplace. One-click search and install curated skills. View Guide →

• 🔌 [2026-03-12] SkillNet MCP Server Released! — We've launched the Model Context Protocol (MCP) integration (maintained by CycleChain, special thanks for this great contribution!). Learn more →

• 📄 [2026-03-04] SkillNet Technical Report Released! — We've published the comprehensive SkillNet Technical Report, covering the system architecture, automated creation pipeline, multi-dimensional evaluation methodology, and the released open-source toolkits. View Report →

• 🦞 [2026-02-23] OpenClaw Integration Released! — SkillNet is now available as a built-in skill for OpenClaw. One command to install, zero config to use. The agent automatically searches, downloads, creates, evaluates, and analyzes skills on your behalf. Get started →

✨ Key Features

📖 Table of Contents

• Quick Start
• REST API
• Python SDK
• CLI Reference
• Configuration
• Example: Scientific Discovery
• OpenClaw Integration
• Model Context Protocol (MCP)
• Contributing
• Citation

🚀 Quick Start

pip install skillnet-ai

Or using uv:

uv venv
source .venv/bin/activate
uv pip install skillnet-ai

from skillnet_ai import SkillNetClient

client = SkillNetClient()  # No API key needed for search & download

# Search for skills
results = client.search(q="pdf", limit=5)
print(results[0].skill_name, results[0].stars)

# Install a skill
client.download(url=results[0].skill_url, target_dir="./my_skills")

🌐 SkillNet Web — Search, download individual skills, and explore curated skill collections through the SkillNet website.

🤖 OpenClaw + SkillNet — See SkillNet in action with OpenClaw. The agent automatically searches, creates, evaluates, and analyzes skills on your behalf. Learn more →

🌐 REST API

The SkillNet search API is free, public, and requires no authentication.

# Keyword search
curl "http://api-skillnet.openkg.cn/v1/search?q=pdf&sort_by=stars&limit=5"

# Semantic search
curl "http://api-skillnet.openkg.cn/v1/search?q=reading%20charts&mode=vector&threshold=0.8"

{
  "data": [
    {
      "skill_name": "pdf-extractor-v1",
      "skill_description": "Extracts text and tables from PDF documents.",
      "author": "openkg-team",
      "stars": 128,
      "skill_url": "https://...",
      "category": "Productivity"
    }
  ],
  "meta": { "query": "pdf", "mode": "keyword", "total": 1, "limit": 10 },
  "success": true
}

🐍 Python SDK

Initialize

from skillnet_ai import SkillNetClient

client = SkillNetClient(
    api_key="sk-...",         # Required for create / evaluate / analyze
    # base_url="...",         # Optional: custom LLM endpoint
    # github_token="ghp-..." # Optional: for private repos
)

Search

# Keyword search
results = client.search(q="pdf", limit=10, min_stars=5, sort_by="stars")

# Semantic search
results = client.search(q="analyze financial PDF reports", mode="vector", threshold=0.85)

if results:
    print(f"{results[0].skill_name} ⭐{results[0].stars}")

Install

local_path = client.download(
    url="https://github.com/anthropics/skills/tree/main/skills/skill-creator",
    target_dir="./my_skills"
)

Create

Convert diverse sources into structured skill packages with a single call:

# From conversation logs / execution traces
client.create(trajectory_content="User: rename .jpg to .png\nAgent: Done.", output_dir="./skills")

# From GitHub repository
client.create(github_url="https://github.com/zjunlp/DeepKE", output_dir="./skills")

# From office documents (PDF / PPT / Word)
client.create(office_file="./guide.pdf", output_dir="./skills")

# From natural language prompt
client.create(prompt="A skill for web scraping article titles", output_dir="./skills")

Evaluate

Score any skill across 5 quality dimensions. Accepts local paths or GitHub URLs.

result = client.evaluate(
    target="https://github.com/anthropics/skills/tree/main/skills/algorithmic-art"
)
# Returns: { "safety": {"level": "Good", "reason": "..."}, "completeness": {...}, ... }

Analyze Relationships

Map the connections between skills in a local directory — outputs similar_to, belong_to, compose_with, and depend_on edges.

relationships = client.analyze(skills_dir="./my_skills")

for rel in relationships:
    print(f"{rel['source']} --[{rel['type']}]--> {rel['target']}")
# PDF_Parser --[compose_with]--> Text_Summarizer

💻 CLI Reference

The CLI ships with pip install skillnet-ai and offers the same features with rich terminal output.

Use skillnet <command> --help for full options.

Search

skillnet search "pdf"
skillnet search "analyze financial reports" --mode vector --threshold 0.85
skillnet search "visualization" --category "Development" --sort-by stars --limit 10

Install

skillnet download https://github.com/anthropics/skills/tree/main/skills/algorithmic-art
skillnet download <url> -d ./my_agent/skills
skillnet download <private_url> --token <your_github_token>

# Use a mirror for faster downloads in restricted networks
skillnet download <url> --mirror https://ghfast.top/

Create

# From trajectory file
skillnet create ./logs/trajectory.txt -d ./generated_skills

# From GitHub repo
skillnet create --github https://github.com/owner/repo

# From office document (PDF, PPT, Word)
skillnet create --office ./docs/guide.pdf

# From prompt
skillnet create --prompt "A skill for extracting tables from images"

Evaluate

skillnet evaluate https://github.com/anthropics/skills/tree/main/skills/algorithmic-art
skillnet evaluate ./my_skills/web_search
skillnet evaluate ./my_skills/tool --category "Development" --model gpt-4o

Analyze

skillnet analyze ./my_agent_skills
skillnet analyze ./my_agent_skills --no-save   # print only, don't write file
skillnet analyze ./my_agent_skills --model gpt-4o

⚙️ Configuration

Environment Variables

search and download (public repos) work without any credentials.

Recommended mirror: https://ghfast.top/ — set GITHUB_MIRROR or pass --mirror to speed up downloads in restricted networks.

Linux / macOS:

export API_KEY="sk-..."
export BASE_URL="https://..."  # optional

Windows PowerShell:

$env:API_KEY = "sk-..."
$env:BASE_URL = "https://..."  # optional

🔬 Example: Scientific Discovery

A complete end-to-end demo showing how an AI Agent uses SkillNet to autonomously plan and execute a complex scientific workflow — from raw scRNA-seq data to a cancer target validation report.

👉 Try the Interactive Demo (Website → Scenarios → Science)  |  📓 View Notebook

🤖 OpenClaw Integration

SkillNet integrates with OpenClaw as a built-in, lazy-loaded skill. Once installed, your agent automatically:

• Searches existing skills before starting complex tasks
• Creates new skills from repos, documents, or completed work
• Evaluates & analyzes your local library for quality and inter-skill relationships

Community skills guide execution → successful outcomes become new skills → periodic analysis keeps the library clean.

npm i -g clawhub
clawhub install skillnet --workdir ~/.openclaw/workspace
openclaw gateway restart

Install the skillnet skill from ClawHub.

{
  "skills": {
    "entries": {
      "skillnet": {
        "enabled": true,
        "apiKey": "sk-REPLACE_ME",
        "env": {
          "BASE_URL": "https://api.openai.com/v1",
          "GITHUB_TOKEN": "ghp_REPLACE_ME"
        }
      }
    }
  }
}

Search SkillNet for a "docker" skill and summarize the top result.

Create a skill from this GitHub repo: https://github.com/owner/repo (then evaluate it).

🔌 Model Context Protocol (MCP) Integration

The SkillNet MCP Server (maintained by CycleChain) is a high-performance bridge that enables AI agents (such as Claude Desktop, Cursor, Antigravity and Windsurf) to interact with the SkillNet ecosystem using the Model Context Protocol.

It empowers agents to autonomously search, download, create, and evaluate 300,000+ specialized skills directly within your IDE or desktop environment.

git clone https://github.com/CycleChain/skillnet-mcp
cd skillnet-mcp
npm install && npm run build

docker pull fmdogancan/skillnet-mcp:latest

{
  "mcpServers": {
    "skillnet": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "fmdogancan/skillnet-mcp:latest"],
      "env": {
        "API_KEY": "your_api_key_here"
      }
    }
  }
}

docker build -t skillnet-mcp-local .

{
  "mcpServers": {
    "skillnet": {
      "command": "node",
      "args": ["/absolute/path/to/skillnet-mcp/build/index.js"],
      "env": {
        "API_KEY": "your_api_key_here"
      }
    }
  }
}

🤝 Contributing

Contributions of all kinds are welcome! Whether it's fixing a typo, adding a feature, or sharing a new skill — every contribution counts.

1. Fork the repository
2. Create a feature branch (git checkout -b feat/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feat/amazing-feature)
5. Open a Pull Request

📤 Contribute skills (Website → Contribute → Submit via URL / Upload Local Skill / Batch Upload Skills)

You can also open an Issue to report bugs or suggest features.

📚 Citation

If you find this work useful, please kindly ⭐ the repo and cite our paper!

@misc{liang2026skillnetcreateevaluateconnect,
      title={SkillNet: Create, Evaluate, and Connect AI Skills}, 
      author={Yuan Liang and Ruobin Zhong and Haoming Xu and Chen Jiang and Yi Zhong and Runnan Fang and Jia-Chen Gu and Shumin Deng and Yunzhi Yao and Mengru Wang and Shuofei Qiao and Xin Xu and Tongtong Wu and Kun Wang and Yang Liu and Zhen Bi and Jungang Lou and Yuchen Eleanor Jiang and Hangcheng Zhu and Gang Yu and Haiwen Hong and Longtao Huang and Hui Xue and Chenxi Wang and Yijun Wang and Zifei Shan and Xi Chen and Zhaopeng Tu and Feiyu Xiong and Xin Xie and Peng Zhang and Zhengke Gui and Lei Liang and Jun Zhou and Chiyu Wu and Jin Shang and Yu Gong and Junyu Lin and Changliang Xu and Hongjie Deng and Wen Zhang and Keyan Ding and Qiang Zhang and Fei Huang and Ningyu Zhang and Jeff Z. Pan and Guilin Qi and Haofen Wang and Huajun Chen},
      year={2026},
      eprint={2603.04448},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2603.04448}, 
}