Skill-Cortex (Lite) / Skill-Cortex (精简版)
一个第三方 MCP 服务器:让所有 IDE 获得 Claude Code Skills 的能力。
A third-party MCP server: Enable all IDEs to access Claude Code Skills capabilities.
Features / 特性
- Skill Discovery & Indexing / 技能发现与索引: Automatically scan and index SKILL.md files from multiple sources / 自动扫描和索引来自多个来源的 SKILL.md 文件
- Flexible Configuration / 灵活配置: Support for custom skill roots and cache paths / 支持自定义技能根目录和缓存路径
- Tag Management / 标签管理: Built-in tag validation and management system / 内置标签验证和管理系统
- Skill Tree Navigation / 技能树导航: Hierarchical browsing of skills by category / 按类别分层浏览技能
- Search Functionality / 搜索功能: Full-text search across all indexed skills / 对所有索引技能进行全文搜索
- Skill Details / 技能详情: Detailed information retrieval for each skill / 每个技能的详细信息检索
- Enhanced Import Tools / 增强导入工具: Advanced import script with configuration files, progress tracking, and robust error handling / 具有配置文件、进度跟踪和健壮错误处理的高级导入脚本
- No Bundled Skills in Repo / 仓库不内置 Skills: This repository does not ship skills by default; put your skills under
~/.claude/skillsor./.skills/ 本仓库默认不内置 skills,请将 skills 放到~/.claude/skills或./.skills - Context Compression / 上下文压缩: Token-friendly skill retrieval with section-based extraction (summary, instructions, examples, full) / 支持按 section 提取技能内容,节省 token 消耗
Prerequisites / 先决条件
- Python 3.11 or higher / Python 3.11 或更高版本
- pip package manager / pip 包管理器
Key Technologies / 关键技术
- MCP (Model Context Protocol): For server integration / 用于服务器集成
- Python Standard Library: File system operations and caching / 文件系统操作和缓存
Project Structure / 项目结构
skill_cortex_server/
├── skill_cortex/ # Main package / 主包
│ ├── __init__.py # Package initialization / 包初始化
│ ├── server.py # MCP server implementation / MCP 服务器实现
│ ├── models.py # Data models / 数据模型
│ ├── config.py # Configuration management / 配置管理
│ ├── scanner.py # Skill file scanner / 技能文件扫描器
│ ├── index_store.py # Index storage and caching / 索引存储和缓存
│ ├── tags_registry.py # Tag management system / 标签管理系统
│ └── frontmatter.py # Frontmatter parsing / 前置元数据解析
├── config/ # Configuration files / 配置文件
│ └── examples/ # Example configuration files / 示例配置文件
│ ├── skills-config.yaml # YAML configuration example / YAML配置示例
│ └── skills-config.json # JSON configuration example / JSON配置示例
├── .kiro/ # Kiro specs and development files / Kiro规范和开发文件
│ └── specs/ # Feature specifications / 功能规范
├── .skill_cortex_cache/ # Cache directory / 缓存目录
│ └── index.json # Skill index cache / 技能索引缓存
├── .skill_cortex_sources/ # Imported skills source / 导入的技能源
├── .skills/ # Local skills directory / 本地技能目录
│ └── imported/ # Imported skills / 导入的技能
├── import_skills.py # Enhanced skill import script / 增强的技能导入脚本
├── pyproject.toml # Project configuration / 项目配置
├── README.md # This file / 本文件
└── tags.md # Allowed tags list / 允许的标签列表Quick Start / 快速开始
Installation / 安装
# Install in editable mode / 以可编辑模式安装
pip install -e .
# Or install from PyPI (when available) / 或从 PyPI 安装(当可用时)
pip install skill-cortex-liteRun the Server / 运行服务器
# Run directly / 直接运行
skill-cortex
# Or use Python module / 或使用 Python 模块
python -m skill_cortex.serverMCP Tools / MCP 工具
1. list_skill_tree / 列出技能树
Browse skills in a hierarchical tree structure / 以分层树结构浏览技能
Parameters / 参数:
path(optional): Starting path in the skill tree / 技能树中的起始路径
Example / 示例:
{
"path": "coding"
}Response / 响应: Returns a hierarchical tree of skills with their metadata / 返回包含元数据的分层技能树
2. search_skills / 搜索技能
Search for skills by query text / 通过查询文本搜索技能
Parameters / 参数:
query(optional): Search query string / 搜索查询字符串tags(optional): Array of tags to filter by / 用于筛选的标签数组
Example / 示例:
{
"query": "database",
"tags": ["coding", "data"]
}Response / 响应: Returns matching skills with their details / 返回匹配的技能及其详细信息
3. get_skill_details / 获取技能详情
Get detailed information about a specific skill with context compression / 获取特定技能的详细信息,支持上下文压缩
Parameters / 参数:
skill_id(required): Unique identifier of the skill / 技能的唯一标识符section(optional): What to return / 返回内容类型"summary"(default): Frontmatter + description snapshot (token-friendly) / 仅元数据和描述摘要(节省token)"instructions": Main instructions section only / 仅主要指令部分"examples": Code examples section only / 仅代码示例部分"full": Complete SKILL.md content / 完整内容
max_lines(optional): Limit output to N lines / 限制输出行数
Example / 示例:
{
"skill_id": "coding/database/query",
"section": "summary"
}{
"skill_id": "coding/database/query",
"section": "instructions",
"max_lines": 50
}Response / 响应: Returns skill information based on section parameter / 根据 section 参数返回技能信息
4. update_tags / 更新标签
Manage tags inside skills frontmatter / 管理 skills 的 frontmatter tags
Parameters / 参数:
mode(required): "list" or "apply" / 操作模式:"list" 或 "apply"updates(optional): required when mode is "apply" / 当 mode 为 "apply" 时需要提供
Example / 示例:
{
"mode": "list"
}{
"mode": "apply",
"updates": [
{"skill_id": "...", "tags": ["python", "mcp"]}
]
}Response / 响应: Returns the current tags list or operation result / 返回当前标签列表或操作结果
5. create_new_skill / 创建新技能
Create a new skill with automatic validation and indexing / 创建新技能并自动验证和索引
Parameters / 参数:
path(required): Skill path (e.g., "coding/python-helper" or "simple-skill") / 技能路径description(required): Skill description (1-1024 characters) / 技能描述(1-1024字符)tags(optional): Array of tags / 标签数组instructions(optional): Custom instructions (if not provided, a template will be generated) / 自定义指令(如未提供则生成模板)license(optional): License information / 许可证信息metadata(optional): Metadata dict (e.g., {"author": "example", "version": "1.0"}) / 元数据字典create_scripts_dir(optional): Whether to create scripts/ directory / 是否创建 scripts/ 目录create_references_dir(optional): Whether to create references/ directory / 是否创建 references/ 目录create_assets_dir(optional): Whether to create assets/ directory / 是否创建 assets/ 目录
Example / 示例:
{
"path": "coding/python-helper",
"description": "A helper skill for Python development tasks",
"tags": ["python", "coding", "development"],
"create_scripts_dir": true
}{
"path": "simple-skill",
"description": "A simple skill example",
"tags": ["example"],
"instructions": "## Instructions\n\nDetailed instructions here...",
"metadata": {"author": "example-user", "version": "1.0"}
}Response / 响应: Returns creation result with skill path and details / 返回创建结果,包含技能路径和详细信息
Notes / 注意事项:
- Skills are created in the
.skills/directory / 技能创建在.skills/目录中 - Skill names must be lowercase with hyphens only / 技能名称必须是小写字母和连字符
- Supports category paths (e.g., "coding/python-helper") / 支持分类路径
- Automatically validates name format and updates index / 自动验证名称格式并更新索引
6. delete_existing_skill / 删除现有技能
Delete a user-created skill (requires confirmation) / 删除用户创建的技能(需要确认)
Parameters / 参数:
skill_id(required): Unique identifier of the skill to delete / 要删除的技能的唯一标识符confirm(required): Must be True to actually delete (False shows preview only) / 必须为 True 才能实际删除(False 仅显示预览)
Example / 示例:
{
"skill_id": ".skills:coding/python-helper/SKILL.md",
"confirm": false
}{
"skill_id": ".skills:simple-skill/SKILL.md",
"confirm": true
}Response / 响应: Returns deletion result or preview information / 返回删除结果或预览信息
Safety Features / 安全特性:
- Only allows deletion of skills in
.skills/directory / 仅允许删除.skills/目录中的技能 - Cannot delete imported skills (
.skills/imported/) / 无法删除导入的技能 - Cannot delete source repository skills (
.skill_cortex_sources/) / 无法删除源仓库技能 - Requires explicit
confirm=trueto execute deletion / 需要显式confirm=true才能执行删除 - Automatically updates index after deletion / 删除后自动更新索引
Configuration / 配置
Environment Variables / 环境变量
-
SKILL_CORTEX_ROOTS: Comma-separated list of skill root directories / 逗号分隔的技能根目录列表- Default:
~/.claude/skills,./.skills/ 默认值 - Example:
/path/to/skills1,/path/to/skills2/ 示例
- Default:
-
SKILL_CORTEX_CACHE_PATH: Path to the index cache file / 索引缓存文件路径- Default:
./.skill_cortex_cache/index.json/ 默认值 - Example:
/custom/path/cache.json/ 示例
- Default:
-
SKILL_CORTEX_TAGS_PATH: Path to the allowed tags file / 允许的标签文件路径- Default:
./tags.md/ 默认值 - Example:
/custom/path/tags.md/ 示例
- Default:
Example Configuration / 配置示例
{
"mcpServers": {
"skill-cortex-lite": {
"command": "skill-cortex",
"args": [],
"env": {
"SKILL_CORTEX_ROOTS": "/Users/username/skills1,/Users/username/skills2",
"SKILL_CORTEX_CACHE_PATH": "/Users/username/.skill_cortex_cache/index.json",
"SKILL_CORTEX_TAGS_PATH": "/Users/username/tags.md"
}
}
}
}Importing Skills / 导入技能
The project includes an enhanced import script with advanced features for importing skills from public repositories / 项目包含一个增强的导入脚本,具有从公共仓库导入技能的高级功能
Enhanced Import Features / 增强导入功能
🔧 Configuration File Support / 配置文件支持
- YAML/JSON Configuration: Use configuration files to customize repository lists without modifying code / 使用配置文件自定义仓库列表,无需修改代码
- Auto-discovery: Automatically finds
skills-config.yaml,skills-config.yml, orskills-config.jsonin current directory / 自动查找当前目录中的配置文件 - Custom config path: Use
--configoption to specify custom configuration file / 使用--config选项指定自定义配置文件
📊 Progress Display / 进度显示
- Real-time progress: Shows current repository being processed with step-by-step feedback / 显示当前处理的仓库和逐步反馈
- Skill counting: Displays number of skills found in each repository / 显示每个仓库中找到的技能数量
- Comprehensive summary: Final report with statistics, timing, and success/failure counts / 包含统计、时间和成功/失败计数的最终报告
🛡️ Robust Error Handling / 健壮错误处理
- Continue on error: Single repository failure doesn't stop the entire import process / 单个仓库失败不会停止整个导入过程
- Detailed error reporting: Clear error messages with specific failure reasons / 清晰的错误消息和具体失败原因
- Error categorization: Different handling for network, file system, and Git errors / 对网络、文件系统和Git错误的不同处理
🔍 Enhanced Preview Mode / 增强预览模式
- Detailed dry-run: Shows repository URLs, skill paths, and counts before actual import / 在实际导入前显示仓库URL、技能路径和计数
- Clear indicators: Clearly shows when running in preview mode with no actual changes / 清楚显示预览模式,不进行实际更改
Built-in Skills / 内置技能
This repository does not bundle skills by default. Use the import script (optional) to fetch skills from public repositories / 本仓库默认不内置 skills,可使用导入脚本(可选)从公共仓库拉取:
1. agentskills/agentskills
A comprehensive collection of AI agent skills and capabilities / 一个全面的 AI 代理技能和能力集合
- Features / 特性: Provides a wide range of skills for AI agents, including coding, data analysis, and problem-solving / 为 AI 代理提供广泛的技能,包括编码、数据分析和问题解决
- Skill Types / 技能类型: Coding, debugging, testing, optimization, and more / 编码、调试、测试、优化等
- Usage / 用途: Enhances AI agent capabilities with practical, reusable skills / 通过实用的、可重用的技能增强 AI 代理能力
2. anthropics/skills
Official skills repository from Anthropic / Anthropic 的官方技能仓库
- Features / 特性: Officially maintained skills optimized for Claude AI / 官方维护的针对 Claude AI 优化的技能
- Skill Types / 技能类型: Documentation, code generation, system administration, and m
…