{
  "markdown": "# Skillhub MCP\n\n<p align=\"center\">\n  <img src=\"./assets/logo.png\" alt=\"Skillhub MCP logo\" width=\"160\" />\n</p>\n\n[![PyPI version](https://img.shields.io/pypi/v/skillhub-mcp.svg)](https://pypi.org/project/skillhub-mcp/)\n[![PyPI downloads](https://img.shields.io/pypi/dm/skillhub-mcp.svg)](https://pypi.org/project/skillhub-mcp/)\n\n## Links\n\n- PyPI: https://pypi.org/project/skillhub-mcp/\n- PyPI v1.0.1: https://pypi.org/project/skillhub-mcp/1.0.1/\n- Skills directory: http://skills.214140846.net/\n\nmcp-name: io.github.214140846/skillhub-mcp\n\nYou already have Claude-style skills (`SKILL.md`), but in practice you often hit a wall:\n\n- your client speaks MCP, not Claude Skills\n- your team uses multiple agents (Cursor, Copilot, Codex, etc.), so skills are painful to reuse across tools\n- you want a more flexible way to organize and ship skills (nested folders, zip packaging)\n\n**Skillhub MCP** bridges that gap: it turns Claude-style skills into MCP tools, so any MCP client can call the same skills.\n\n> ⚠️ Experimental. Skills may contain scripts/resources. Treat them as untrusted and run with sandboxes/containers when possible.\n\n## Is this an MCP server or an MCP client?\n\nThis project is an **MCP server**.\n\n- **Skillhub MCP (this repo)**: runs as a server process and exposes tools/resources to clients.\n- **MCP clients**: editors/agents like Cursor, Claude Code, Codex, etc. They start or connect to MCP servers.\n\n## What You Get\n\n- Cross-client reuse: install once, use from any MCP client\n- Flexible packaging: nested directories, `.zip` and `.skill` archives\n- Skill resources: expose scripts/datasets/examples as MCP resources (files the client can read)\n- Resource fallback: a `fetch_resource` tool for clients without native MCP resource support\n- Multiple transports: `stdio` (default), `http`, `sse`\n\n## Quick Start\n\nDefault skills root: `~/.skillhub-mcp`\n\n### uvx (recommended)\n\n```json\n{\n  \"skillhub-mcp\": {\n    \"command\": \"uvx\",\n    \"args\": [\"skillhub-mcp@latest\"]\n  }\n}\n```\n\nUse a custom skills root:\n\n```json\n{\n  \"skillhub-mcp\": {\n    \"command\": \"uvx\",\n    \"args\": [\"skillhub-mcp@latest\", \"/path/to/skills\"]\n  }\n}\n```\n\n## Install in Popular Editors (MCP Clients)\n\nBelow are minimal working examples for mainstream “vibe coding” editors.\n\n### Cursor\n\nCursor supports configuring MCP servers via `mcp.json`. Add the following to your\nglobal `~/.cursor/mcp.json` or project `.cursor/mcp.json`, then restart Cursor.\n\n```json\n{\n  \"mcpServers\": {\n    \"skillhub-mcp\": {\n      \"type\": \"stdio\",\n      \"command\": \"uvx\",\n      \"args\": [\"skillhub-mcp@latest\", \"/path/to/skills\"]\n    }\n  }\n}\n```\n\n### Claude Code\n\nOption A: configure via Claude Code CLI (recommended for quick setup):\n\n```bash\nclaude mcp add --transport stdio skillhub-mcp -- uvx skillhub-mcp@latest /path/to/skills\n```\n\nOption B: project-scoped configuration via `.mcp.json` at your project root. You\nmay need to explicitly allow project MCP servers in `.claude/settings.json`.\n\n`./.mcp.json`\n\n```json\n{\n  \"mcpServers\": {\n    \"skillhub-mcp\": {\n      \"type\": \"stdio\",\n      \"command\": \"uvx\",\n      \"args\": [\"skillhub-mcp@latest\", \"/path/to/skills\"]\n    }\n  }\n}\n```\n\n`./.claude/settings.json` (approve only this server)\n\n```json\n{\n  \"enabledMcpjsonServers\": [\"skillhub-mcp\"]\n}\n```\n\n### Codex (OpenAI)\n\nOption A: use the Codex CLI to add a stdio MCP server:\n\n```bash\ncodex mcp add skillhub-mcp -- uvx skillhub-mcp@latest /path/to/skills\n```\n\nOption B: edit `~/.codex/config.toml`:\n\n```toml\n[mcp_servers.skillhub-mcp]\ncommand = \"uvx\"\nargs = [\"skillhub-mcp@latest\", \"/path/to/skills\"]\n```\n\n## Skill Format\n\nSkillhub MCP discovers skills under the root directory (default `~/.skillhub-mcp`).\nEach skill can be:\n\n- a directory containing `SKILL.md`\n- a `.zip` or `.skill` archive containing `SKILL.md` (at the archive root or\n  inside a single top-level folder)\n\nAll other files become downloadable MCP resources for your agent to read. Note:\nSkillhub MCP does not execute scripts; the client decides whether/how to run them.\n\nExample layout:\n\n```text\n~/.skillhub-mcp/\n├── summarize-docs/\n│   ├── SKILL.md\n│   ├── summarize.py\n│   └── prompts/example.txt\n├── translate.zip\n├── analyzer.skill\n└── web-search/\n    └── SKILL.md\n```\n\nArchive rules:\n\n```text\ntranslate.zip\n├── SKILL.md\n└── helpers/\n    └── translate.js\n```\n\n```text\ndata-cleaner.zip\n└── data-cleaner/\n    ├── SKILL.md\n    └── clean.py\n```\n\n## Directory Structure: Skillhub MCP vs Claude Code\n\nClaude Code expects a flat skills directory (each immediate subdirectory is one\nskill). Skillhub MCP is more permissive:\n\n- nested directories are discovered\n- `.zip` / `.skill` packaged skills are supported\n\nIf you need Claude Code compatibility, keep the flat layout.\n\n## CLI Reference\n\n`skillhub-mcp [skills_root] [options]`\n\n| Flag / Option | Description |\n| --- | --- |\n| positional `skills_root` | Optional skills directory (defaults to `~/.skillhub-mcp`). |\n| `--transport {stdio,http,sse}` | Transport (default `stdio`). |\n| `--host HOST` | Bind address for HTTP/SSE transports. |\n| `--port PORT` | Port for HTTP/SSE transports. |\n| `--path PATH` | URL path for HTTP transport. |\n| `--list-skills` | List discovered skills and exit. |\n| `--verbose` | Emit debug logging. |\n| `--log` | Mirror verbose logs to `/tmp/skillhub-mcp.log`. |\n\n## Safety Notes\n\n- Skills are not \"just prompts\": they can include scripts and arbitrary files.\n- Skillhub MCP does not run scripts, but your client might. Prefer running in a sandbox/container.\n\n## Language\n\n- English: `README.md`\n- 中文: `README.zh-CN.md`\n\n\n## About the Author\n\nI 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**.\n\nEverything shared here comes from real projects, real traffic, and real revenue attempts.\n\n- **Feishu Knowledge Base**:\n    \n    [Thor’s AI Going-Global Content Planning](https://my.feishu.cn/wiki/space/7271588985498140676?ccm_open_type=lark_wiki_spaceLink&open_tab_from=wiki_home)\n    \n    A structured knowledge base documenting hands-on experience in AI product overseas expansion, including demand discovery, execution strategies, and common pitfalls.\n    \n- **Blog**:\n    \n    [Thor-AI Blog](https://www.notion.so/Thor-AI-2eaf0388ab4680d0a98bedc8d290e1be?pvs=21)\n    \n    Long-form notes and case studies on building, launching, and iterating AI products in public.\n    \n- **Open-source Project (High Star)**:\n    \n    **Smart Campus System**\n    \n    - GitHub: https://github.com/214140846/TOGO_School_Miniprograme\n    - Gitee: https://gitee.com/zengyunengineer/TOGO_School_Miniprograme\n- **Social**:\n    \n    [Jike](https://web.okjike.com/u/159D450D-2193-4739-8825-AA8EBEC2E9B4)\n    \n    Sharing real-time thoughts on indie hacking, AI tools, and product growth.\n    \n- **Product**:\n    - **AI Video Generation Platform**:\n        \n        [Sora 2](https://sora2.cloud/)\n        [Sora 2 ai](https://sora2.cloud/home)\n        \n        An online platform for AI-powered video generation, focused on practical use cases and real user workflows.\n        \n    - **AI Video & Image Generation**:\n        \n        [AI Video & Image Collection](https://ricebowl.ai/)\n\n      Model pages:\n      - [Grok Video](https://ricebowl.ai/m/grok-video)\n      - [Sora 2](https://ricebowl.ai/m/sora/sora-2)\n      - [Veo 3.1](https://ricebowl.ai/m/veo/veo-3-1)\n      - [Veo 3](https://ricebowl.ai/m/veo/veo-3)\n      - [Veo 2](https://ricebowl.ai/m/veo/veo-2)\n      - [Kling 2.6](https://ricebowl.ai/m/kling-2-6)\n      - [Wan 2.5](https://ricebowl.ai/m/wan/wan-2-5)\n      - [Seedance](https://ricebowl.ai/m/seedance)\n      - [Nano Banana 2](https://ricebowl.ai/m/nano-banana-2)\n      - [Nano Banana Pro](https://ricebowl.ai/m/nano-banana-pro)\n        \n        \n        A curated collection of AI video and image generation tools, experiments, and capability tracking.\n        \n- **AI Video & Image Collection**:\n    \n    [https://www.notion.so/2e7600937cb3808c818efe79141f7ee6](https://www.notion.so/2e7600937cb3808c818efe79141f7ee6?pvs=21)\n",
  "bytes": 8056,
  "sha": "052c32e329539d5621c442ebbc15bc6e22a0023a67d191ffa80a37449273558f",
  "repo_slug": "214140846/skillhub-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_214140846_skillhub_mcp_96df8d26/readme"
}