Skill Loader MCP Server
Enables discovering, fetching, validating, and converting Claude skills from the skills.sh marketplace and GitHub repositories.
README
Skill Loader MCP Server
An MCP (Model Context Protocol) server for discovering, fetching, validating, and converting Claude skills from the skills.sh marketplace and GitHub repositories.
What's New in v1.1.0
- skills.sh API integration: Search uses the
/api/search?q=<query>endpoint — no authentication needed - Authenticated endpoints:
list_skillsandget_leaderboarduse the/api/v1/skillsendpoint (requiresSKILLS_SH_API_KEY) - Smarter Power conversion:
convert_to_powernow generatesmcp.jsonwhen a skill has dependencies or tool references, and creates asteering/directory when a skill has 3+ complex sections - MCP SDK upgrade: Updated to
@modelcontextprotocol/sdkv1.29
Features
- Discover Skills — Browse and search the skills.sh marketplace via its public search API
- Fetch Skills — Download raw skill content directly from GitHub
- Security Validation — Scan skills for dangerous commands, suspicious patterns, and code injection
- Format Conversion — Convert skills to Kiro steering files or power format (with optional
mcp.jsonandsteering/directory) - Complete Workflow — Import skills end-to-end with a single command (fetch + validate + convert)
Installation
Global Installation
npm install -g @goldzulu/skill-loader-mcp-server
Local Installation
npm install @goldzulu/skill-loader-mcp-server
Configuration
Environment Variables
| Variable | Required | Description |
|---|---|---|
SKILLS_SH_API_KEY |
For list_skills and get_leaderboard only |
API key for the skills.sh authenticated /api/v1/skills endpoint. Request from Vercel if needed. Not required for search_skills. |
With Kiro
Add to your mcp.json:
{
"mcpServers": {
"skill-loader": {
"command": "npx",
"args": ["-y", "@goldzulu/skill-loader-mcp-server"],
"env": {
"SKILLS_SH_API_KEY": "your-api-key-here"
},
"description": "Skill Loader MCP Server for managing Claude skills"
}
}
}
With Claude Desktop
Add to your Claude Desktop configuration:
{
"mcpServers": {
"skill-loader": {
"command": "skill-loader-mcp-server",
"env": {
"SKILLS_SH_API_KEY": "your-api-key-here"
}
}
}
}
The env block is optional — only needed if you want to use list_skills or get_leaderboard.
Standalone
skill-loader-mcp-server
Available Tools
1. search_skills
Search for skills by keyword using the skills.sh search API. No authentication required.
Parameters:
query(required): Search querylimit(optional): Max results (default: 20, max: 50)
Example:
{
"tool": "search_skills",
"arguments": { "query": "pdf", "limit": 5 }
}
2. list_skills
List all available skills from skills.sh with pagination. Requires SKILLS_SH_API_KEY.
Parameters:
page(optional): Page number (default: 1)pageSize(optional): Results per page (default: 50, max: 100)
Example:
{
"tool": "list_skills",
"arguments": { "page": 1, "pageSize": 10 }
}
3. get_leaderboard
Get trending or top-installed skills. Requires SKILLS_SH_API_KEY.
Parameters:
timeframe(optional):'all'or'24h'(default:'all')limit(optional): Max results (default: 20, max: 50)
Example:
{
"tool": "get_leaderboard",
"arguments": { "timeframe": "24h", "limit": 10 }
}
4. fetch_skill
Fetch raw skill content from GitHub.
Parameters:
identifier(required): Skill name orowner/repoformat
Example:
{
"tool": "fetch_skill",
"arguments": { "identifier": "anthropics/pdf-extractor" }
}
5. validate_skill
Validate skill content for security issues.
Parameters:
content(required): Skill content to validateurl(optional): Source URL for verification
Example:
{
"tool": "validate_skill",
"arguments": {
"content": "---\nname: Test\n---\n\n# Test",
"url": "https://example.com/skill.md"
}
}
6. convert_to_steering
Convert skill to Kiro steering file format.
Parameters:
content(required): Skill contentsourceUrl(optional): Original source URL
Example:
{
"tool": "convert_to_steering",
"arguments": {
"content": "---\nname: Test\ndescription: A test skill\n---\n\n# Test",
"sourceUrl": "https://example.com/skill.md"
}
}
7. convert_to_power
Convert skill to Kiro power format. Generates mcp.json when the skill has dependencies or tools, and a steering/ directory when the skill has 3+ complex sections.
Parameters:
content(required): Skill contentsourceUrl(optional): Original source URL
Example:
{
"tool": "convert_to_power",
"arguments": {
"content": "---\nname: Test\ndescription: A test skill\n---\n\n# Test",
"sourceUrl": "https://example.com/skill.md"
}
}
8. import_skill
Complete import workflow (fetch + validate + convert).
Parameters:
identifier(required): Skill identifieroutputFormat(required):'steering'or'power'skipValidation(optional): Skip security validation (default: false)
Example:
{
"tool": "import_skill",
"arguments": {
"identifier": "anthropics/pdf-extractor",
"outputFormat": "steering"
}
}
Security
The server includes security validation that scans for:
- Dangerous commands (
rm -rf,sudo,eval,exec) - Suspicious file operations (
/etc/,/usr/,/bin/) - Code injection patterns (
${...},$(...)) - Untrusted sources (non-GitHub URLs)
Skills that fail security validation will be blocked from import unless explicitly skipped.
Caching
The server caches skills.sh search results in memory for 1 hour to reduce API calls and improve performance. The cache is automatically refreshed when expired.
Error Handling
All tools return errors in a consistent JSON format with descriptive messages. Common error types:
- Network errors: Connection issues, timeouts, HTTP errors
- Validation errors: Security issues, format problems
- Parsing errors: YAML syntax errors, invalid markdown
- Resolution errors: Skill not found, ambiguous identifiers
Development
Build
npm run build
Test
npm test
Watch Mode
npm run test:watch
Requirements
- Node.js 18 or higher
- npm or yarn
License
MIT
Contributing
Contributions are welcome! Please see CONTRIBUTING.md for guidelines.
Support
For issues and questions, please open an issue on GitHub.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。