Skillhub MCP

🚀 Skillhub MCP

Skillhub MCP bridges the gap between Claude-style skills and MCP clients, enabling any MCP client to call the same skills. It offers cross - client reuse, flexible packaging, and more.

🔗 Links

• PyPI: https://pypi.org/project/skillhub-mcp/
• PyPI v1.0.1: https://pypi.org/project/skillhub-mcp/1.0.1/
• Skills directory: http://skills.214140846.net/

mcp-name: io.github.214140846/skillhub-mcp

You may encounter difficulties when using Claude-style skills (SKILL.md):

• Your client uses MCP, not Claude Skills.
• Your team uses multiple agents (Cursor, Copilot, Codex, etc.), making it difficult to reuse skills across tools.
• You need a more flexible way to organize and distribute skills (nested folders, zip packaging).

Skillhub MCP solves these problems by converting Claude-style skills into MCP tools, allowing any MCP client to use the same skills.

⚠️ Important Note

This is an experimental project. Skills may contain scripts or resources. Treat them as untrusted and run them in sandboxes or containers if possible.

❓ Is this an MCP server or an MCP client?

This project is an MCP server.

• Skillhub MCP (this repo): Runs as a server process and exposes tools and resources to clients.
• MCP clients: Editors or agents such as Cursor, Claude Code, Codex, etc. They start or connect to MCP servers.

✨ Features

• Cross - client reuse: Install once and use from any MCP client.
• Flexible packaging: Support nested directories, .zip and .skill archives.
• Skill resources: Expose scripts, datasets, and examples as MCP resources (files that the client can read).
• Resource fallback: Provide a fetch_resource tool for clients without native MCP resource support.
• Multiple transports: Support stdio (default), http, and sse.

🚀 Quick Start

Default skills root: ~/.skillhub-mcp

uvx (recommended)

{
  "skillhub-mcp": {
    "command": "uvx",
    "args": ["skillhub-mcp@latest"]
  }
}

Use a custom skills root:

{
  "skillhub-mcp": {
    "command": "uvx",
    "args": ["skillhub-mcp@latest", "/path/to/skills"]
  }
}

📦 Installation in Popular Editors (MCP Clients)

Below are minimal working examples for mainstream “vibe coding” editors.

Cursor

Cursor supports configuring MCP servers via mcp.json. Add the following to your global ~/.cursor/mcp.json or project .cursor/mcp.json, then restart Cursor.

{
  "mcpServers": {
    "skillhub-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": ["skillhub-mcp@latest", "/path/to/skills"]
    }
  }
}

Claude Code

Option A: Configure via Claude Code CLI (recommended for quick setup):

claude mcp add --transport stdio skillhub-mcp -- uvx skillhub-mcp@latest /path/to/skills

Option B: Project - scoped configuration via .mcp.json at your project root. You may need to explicitly allow project MCP servers in .claude/settings.json.

./.mcp.json

{
  "mcpServers": {
    "skillhub-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": ["skillhub-mcp@latest", "/path/to/skills"]
    }
  }
}

./.claude/settings.json (approve only this server)

{
  "enabledMcpjsonServers": ["skillhub-mcp"]
}

Codex (OpenAI)

Option A: Use the Codex CLI to add a stdio MCP server:

codex mcp add skillhub-mcp -- uvx skillhub-mcp@latest /path/to/skills

Option B: Edit ~/.codex/config.toml:

[mcp_servers.skillhub-mcp]
command = "uvx"
args = ["skillhub-mcp@latest", "/path/to/skills"]

💻 Skill Format

Skillhub MCP discovers skills under the root directory (default ~/.skillhub-mcp). Each skill can be:

• A directory containing SKILL.md.
• A .zip or .skill archive containing SKILL.md (at the archive root or inside a single top - level folder).

All other files become downloadable MCP resources for your agent to read. Note: Skillhub MCP does not execute scripts; the client decides whether and how to run them.

Example layout:

~/.skillhub-mcp/
├── summarize-docs/
│   ├── SKILL.md
│   ├── summarize.py
│   └── prompts/example.txt
├── translate.zip
├── analyzer.skill
└── web-search/
    └── SKILL.md

Archive rules:

translate.zip
├── SKILL.md
└── helpers/
    └── translate.js

data-cleaner.zip
└── data-cleaner/
    ├── SKILL.md
    └── clean.py

🔧 Directory Structure: Skillhub MCP vs Claude Code

Claude Code expects a flat skills directory (each immediate subdirectory is one skill). Skillhub MCP is more flexible:

• Nested directories are supported.
• .zip / .skill packaged skills are supported.

If you need Claude Code compatibility, keep the flat layout.

📚 CLI Reference

skillhub-mcp [skills_root] [options]

⚠️ Safety Notes

• Skills are not just prompts; they can include scripts and arbitrary files.
• Skillhub MCP does not run scripts, but your client might. It is recommended to run in a sandbox or container.

🌐 Language

• English: README.md
• Chinese: README.zh-CN.md

👨‍💻 About the Author

I focus on AI SaaS going global, covering the full journey from idea validation and vibe coding to product development, infrastructure, SEO, backlinks, and growth experiments.

Everything shared here comes from real projects, real traffic, and real revenue attempts.

• Feishu Knowledge Base: Thor’s AI Going - Global Content Planning A structured knowledge base documenting hands - on experience in AI product overseas expansion, including demand discovery, execution strategies, and common pitfalls.
• Blog: Thor - AI Blog Long - form notes and case studies on building, launching, and iterating AI products in public.
• Open - source Project (High Star): Smart Campus System
  • GitHub: https://github.com/214140846/TOGO_School_Miniprograme
  • Gitee: https://gitee.com/zengyunengineer/TOGO_School_Miniprograme
• Social: Jike Sharing real - time thoughts on indie hacking, AI tools, and product growth.
• Product:
  • AI Video Generation Platform: Sora 2 Sora 2 ai An online platform for AI - powered video generation, focused on practical use cases and real user workflows.
  • AI Video & Image Generation: AI Video & Image Collection Model pages:
    • Grok Video
    • Sora 2
    • Veo 3.1
    • Veo 3
    • Veo 2
    • Kling 2.6
    • Wan 2.5
    • Seedance
    • Nano Banana 2
    • Nano Banana Pro A curated collection of AI video and image generation tools, experiments, and capability tracking.
• AI Video & Image Collection: https://www.notion.so/2e7600937cb3808c818efe79141f7ee6