skill-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| SKILL_MCP_ROOTS | No | Comma/colon-separated roots (overrides defaults) | |
| SKILL_MCP_STATE_DIR | No | Binding state dir (default ~/.config/skill-mcp) | ~/.config/skill-mcp |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_skillsA | Catalog local SKILL.md skills from configured roots; include shadowed duplicates with why. |
| suggest_skillsA | Lexically rank skills for a prompt. Hard cap 3; default token budget 4000 (CJK-aware). Returns reasons and dropped (cap/budget). |
| bind_skillsC | Persist lean binding (name/description/path only). Max 3. Pass clear:true or none to clear. Optional scope: session (in-memory), project (.skill-mcp/binding.json), or global (~/.config/skill-mcp/binding.json). Default global. Priority: session > project > global. |
| get_bindingA | Return current lean skill binding. Default is the resolved binding (session > project > global). Pass scope to inspect one layer. Includes the host contract (only these N skills). |
| whyA | Explain why the current skills are bound. |
| estimate_tokensB | Estimate tokens for one/many/all skills (mode: body|description). |
| doctorA | Dry-run only: check skill roots exist, counts, shadowed, and binding path. Never mutates. |
| archive_idleA | List idle and shadowed user-tier skills. Default dry-run reports what would move. Pass apply:true to archive into a recoverable path. Never silent-deletes. |
| read_skillB | Read full SKILL.md body by explicit skill name (opt-in; other tools stay lean). |
| rescan_skillsA | Rescan skill roots into the in-memory catalog without restarting the MCP process. list_skills and suggest_skills then reflect new/changed/removed skills. |
| write_host_contractA | Write the current binding contract into Claude CLAUDE.md and Cursor .cursor/rules in one shot. The files list only bound skills and forbid others. |
| native_skills_bypassA | Optional: write Claude skillOverrides name-only so the host listing does not dump every skill description. enabled:true writes; enabled:false restores the previous settings in one step. Risk: routing fight with Claude Code native skills. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| binding-contract | Host/model may use only the currently bound skills |