skills

You can expose SKILL.md files as callable MCP tools in Docker by running the Skillz MCP server. It reads frontmatter from local directories or zip archives and registers each skill as an invocable tool for MCP clients.

Yes, you can load Claude-style skills from zip archives in an MCP workflow. The Skillz runtime supports loading skills organized in either root directories or zip archives, parsing the SKILL.md frontmatter to expose them as callable tools.

Yes, the Skillz runtime is required to invoke optional helper scripts through an MCP client. The runtime integrates with the MCP client, loads skills from the root directory, and supports executing optional bundled scripts, references, and assets.

The best way to automate multi-skill workflows across different projects is using a Skillz MCP server in Docker. It exposes multiple SKILL.md definitions through a single API, allowing you to orchestrate and invoke various skills within CI/CD pipelines or automated tasks.

The Skillz MCP server discovers skills defined by SKILL.md frontmatter by scanning a mounted skills root directory. It parses the frontmatter metadata and optional linked resources, then registers each skill as a distinct callable tool for the MCP client.

When embedding callable skills in CI/CD pipelines with Docker, you must organize skills into folders or zip archives and mount the root directory correctly. The Skillz runtime requires this specific directory structure to successfully load and expose the skills.