Skills_Curator

Use when the user mentions a skill or plugin by name, asks "should I install X?", asks "what skills fit this project?", asks to audit installed skills, asks whether a skill is safe to use, or wants to customize an external skill for their stack. Evaluates skills against the user's actual project (pros / cons / conflicts / ADOPT-PARTIAL-SKIP verdict), recommends what fits, scans for security red flags before adoption, persists every decision to a registry the user keeps in Project Knowledge, and forks stack-mismatched skills into project-tailored versions. Decide once, never re-decide.

Use when the user mentions a skill/plugin by name, asks "should I install X", asks for skill recommendations, wants a security check on a skill, asks about duplicates or conflicts, wants to create a new skill, or wants to migrate skills across agent platforms (Cursor, Codex, Gemini, etc.).

The intelligence layer for Claude skills — without Python. Maintains a trust-rated catalog (curated entries + live GitHub topic search via curl), identifies what fits your project (stack, deps, CLAUDE.md), recommends with pros/cons and per-project customization advice, persists every decision, and migrates across 55 supported agent platforms. Same model as the full version, no engine — the agent does the work via Bash, Read, Glob, Grep, Write. Use when the user mentions a skill, asks "should I install X", asks to evaluate / recommend / audit / check a skill, asks "what skills fit this project", asks for a list of supported platforms, or wants to migrate skills to another agent. Choose this over the full version when you don't have Python, want zero dependencies, or prefer transparent agent reasoning.