← ClaudeAtlas

loadout-curatorlisted

Decide which agent skills, MCP servers, and tools should be active for the current repository, and keep that set current. Use when the user asks what skills they should install, why an agent is missing context about their stack, whether their setup is up to date, or when starting work in an unfamiliar codebase.
VirajMishra1/loadout · ★ 4 · AI & Automation · score 77
Install: claude install-skill VirajMishra1/loadout
# Loadout Curator Match the active skill set to the repository in front of you. Too few skills and the agent lacks context; too many and every prompt carries weight it does not need. This skill wraps the `loadout` CLI so the catalog, provenance, and safety findings stay current rather than being copied into this file. ## Prerequisite ```bash loadout --version ``` If that fails, tell the user to install it (`npm install --global loadout-ai`) and stop; do not guess at catalog contents. ## Starting work in a repository Read what the codebase actually is before suggesting anything: ```bash loadout recommend --project . ``` This prints detected signals (languages, frameworks, test runners) and rule-based suggestions with a confidence label. The signals are the useful part — they tell you what the repository is without you reading every file. Report the detected stack first, then the suggestions worth acting on. Skip low-confidence suggestions unless the user asks for the full list. `recommend` reads the project and the catalog. It does not prove quality, and it says so; repeat that framing rather than presenting suggestions as verified. ## Proposing an active set ```bash loadout optimize --project . ``` This proposes which already-downloaded library skills should be active here. Use it when the user has a Maximum-mode library, or after `recommend` shows the project needs something different from what is on. Present the proposal as a diff in plain words — what turns