Convert a user task into an execution-ready implementation plan with assumptions, risks, files likely to change, validation steps, rollback notes, and approval readiness. Use during Plan Mode.
Keep code changes bound to an approved plan, avoid unrelated refactors, respect off-limits files, and summarize changed files. Use during Execute Mode.
Review implementation against an approved plan and return PASS, PARTIAL PASS, or FAIL with missing requirements, unexpected changes, risks, and test/build status. Use during Validator Mode.
Inspect project structure, frameworks, package managers, entry points, configs, and architecture before implementation planning. Use when planning requires understanding an unfamiliar codebase.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Generate safe git status, diff summaries, changed file summaries, and commit message suggestions without committing or pushing. Use after execution or validation.
Find and respect project rules such as AGENTS.md, CLAUDE.md, README.md, .cursor/rules, .factory/rules, .pi, and .env.example before planning or execution.