Summary

The searches and the reads sit in front of the single call that writes to your machine, which is the right order for this.

skillsmp_search is the keyword path and skillsmp_ai_search the intent path, and both stop at a listing; skillsmp_list_repo_skills and skillsmp_get_skill_content then let you read the actual SKILL.md before anything lands. Only skillsmp_install_skill touches your setup, and it is explicit about where — the agent list is named rather than inferred, and global decides user-level against project-level. Search results carry star counts and GitHub URLs, so checking provenance does not need a separate tool.

What it is

A SkillsMP client with 5 tools: keyword and semantic search over the marketplace, reads of a skill's file content and of a repository's skill list, and installation into a named set of coding agents.

What you get

skillsmp_search finds skills by keyword with paging and sorting by stars or recency, returning name, description, author, star count and GitHub URL.
skillsmp_ai_search takes a natural-language description of what you are trying to accomplish, for when keywords are not enough.
skillsmp_get_skill_content reads a SKILL.md from a GitHub repository by owner, repo, path and branch, and returns a directory listing when that path holds no skill file.
skillsmp_list_repo_skills lists every skill in a repository from an owner/repo shorthand, without installing anything.
skillsmp_install_skill installs named skills from a repository into target agents — claude-code, cursor, codex, opencode, antigravity, github-copilot and roo — with a global flag choosing a user-level install over a project-level one.

Requirements

A SkillsMP API key, supplied as SKILLSMP_API_KEY.

Setup effort

One command plus a key — claude mcp add skillsmp -- npx -y skillsmp-mcp-server --env SKILLSMP_API_KEY=your_api_key, then supply credentials