Skillz

Skillz is an MCP server designed to make Claude-style skills accessible as callable tools for a variety of agent frameworks, including Codex, Copilot, and Cursor. It addresses the need for integrating modular, reusable skills into agent-based workflows by discovering, exposing, and executing these skills for any MCP-compatible client.

skill archive. md file containing YAML front matter that describes the skill, and may include additional resources such as scripts, datasets, or example files. Skillz exposes the authored instructions and resources for each discovered skill and can execute any bundled helper scripts. skill files. For compatibility with Claude Code, a flat directory layout is required, but Skillz also supports more complex organizational schemes.

Skillz can be run directly or within a Docker container for isolation, with configuration options available to specify the skills directory and command-line arguments. It provides several command-line options, including the ability to list all discovered skills, choose the transport protocol (stdio, HTTP, or SSE), set the host and port for networked transports, and enable verbose logging. The tool is available via GitHub and PyPI, and a Docker image is also provided. Users are cautioned that Skillz is an experimental proof-of-concept and may be unsafe; running skills in sandboxes or containers is recommended due to the potential risk of executing untrusted code.

Skillz is positioned as an infrastructure solution for developers and teams working with agentic systems who require a method to modularly integrate and manage skills across different agent platforms.