User

Skillet User Guide

Skillet is an MCP-native skill discovery toolkit. This skill covers using skillet as a consumer: finding and using skills at runtime via MCP prompts.

Adding Skillet

MCP server (recommended) -- add to .mcp.json or ~/.claude/settings.json:

{
  "mcpServers": {
    "skillet": {
      "command": "skillet",
      "args": ["serve"]
    }
  }
}

Install the binary:

cargo install skillet-mcp

Docker (no install needed):

{
  "mcpServers": {
    "skillet": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "ghcr.io/joshrotenberg/skillet:latest"]
    }
  }
}

Searching for Skills

CLI:

skillet search "rust development"
skillet search "*" --category development
skillet search "*" --tag pytest
skillet search "*" --owner joshrotenberg

MCP tools (when running as server):

• search_skills(query) -- full-text BM25 search over skill metadata and content
• list_categories() -- browse available categories with counts
• list_skills_by_owner(owner) -- list all skills by a publisher
• info_skill(owner_name) -- detailed information about a specific skill

Using Skills

Skills are served as MCP prompts. An agent connects to skillet, searches for a relevant skill, and uses it as a prompt for the current session. No files written, no restart needed.

1. search_skills("rust development")
2. Use the skill prompt for joshrotenberg/rust-dev
3. Follow the skill's instructions

Configuration

Skillet is configured via ~/.config/skillet/config.toml and CLI flags.

Adding custom repos:

skillet serve --remote https://github.com/org/skills.git

Config file (~/.config/skillet/config.toml):

[repos]
remote = ["https://github.com/org/skills.git"]

[server]
discover_local = false

To skip the official repo, use --no-official-repo.

User Preferences

If you prefer a specific workflow, tell your agent:

• "always inline" -- only use skills via MCP prompts, never write files
• If unclear, default to inline use via prompts