Skill-to-MCP

You can turn Claude Skills into MCP server resources that are accessible through the Model Context Protocol. This lets you discover, inspect, and read skill content alongside related files from a centralized MCP endpoint, enabling seamless integration with LLM applications and biomedical workflows.

How to use

This MCP server exposes three core tools that you will use through your MCP client: get_available_skills, get_skill_details, and get_skill_related_file. get_available_skills lists all discovered skills with metadata, get_skill_details returns the full SKILL.md content plus the skill’s file listing, and get_skill_related_file reads a specific file inside a skill’s directory with safeguards to prevent accessing outside the skill folders.

How to install

Prerequisites: you need Python 3.11+ installed on your system. You also need a running MCP client capable of consuming MCP servers.

Install the Skill-to-MCP package and run it with a skills directory. You have two common approaches: running directly with uvx or installing via Python’s package manager and starting the server.

Configuration

Choose a directory that contains your skills. Each skill must live in its own subdirectory with a SKILL.md file containing YAML frontmatter and any supporting scripts, references, or examples.

Configure your MCP client to point to the Skill-to-MCP server using one of the provided command-and-arguments patterns.

Two representative configuration examples are shown below. They both use uvx to launch the Skill-to-MCP server and pass the path to your skills directory.

{
  "mcpServers": {
    "skill-to-mcp": {
      "command": "uvx",
      "args": ["skill_to_mcp", "--skills-dir", "/path/to/your/skills"],
      "env": {
        "UV_PYTHON": "3.12"
      }
    }
  }
}

{
  "mcpServers": {
    "skill-to-mcp": {
      "command": "uvx",
      "args": ["skill_to_mcp"],
      "env": {
        "UV_PYTHON": "3.12",
        "SKILLS_DIR": "/path/to/your/skills"
      }
    }
  }
}

Security and access

The server validates file paths to ensure you can only access files within a skill’s directory. This helps prevent directory traversal attacks when reading skill content or related files.

Creating skills and structure

Place your skills under the configured skills directory. Each skill must have its own subdirectory containing a SKILL.md with YAML frontmatter that defines name and description, followed by the skill content and related resources.

Usage with the three core tools

get_available_skills returns a list of available skills, including their names, descriptions, and paths. get_skill_details returns the SKILL.md content and a list of files for a given skill. get_skill_related_file reads a specific file within a skill’s directory while enforcing safety checks.

Troubleshooting

If skills are not discovered, verify that the skills directory exists and contains subdirectories with SKILL.md files. Ensure the MCP client is configured to point to the Skill-to-MCP server and that the environment variables (such as UV_PYTHON) match your runtime expectations.

Notes

You can run multiple Skill-to-MCP instances with different skills collections by providing separate skills-dir values in separate mcpServers configurations.

Available tools

get_available_skills

Lists all available skills with metadata including name, description, and path.

get_skill_details

Returns the full SKILL.md content and a listing of files for a specific skill.

get_skill_related_file

Reads a file inside a skill directory with safeguards to prevent directory traversal.