What problem does it solve? Codebase audits often produce vague advice that dies in the conversation where it was found. This Skill surveys a repository as a senior advisor, vets every finding against the actual code, and produces prioritized, fully self-contained implementation plans that a different model or agent with zero prior context can execute, test, and maintain. ## Core Features & Use Cases - Read-only codebase audit: Surveys correctness, security, performance, test coverage, tech debt, dependencies, DX, docs, and product direction without ever modifying source code. - Handoff plan generation: Writes plan files under plans/ with inlined context, verification commands, scope boundaries, STOP conditions, and machine-checkable done criteria. - Execution and follow-through: Dispatches an isolated-worktree executor subagent per plan, reviews its diff like a tech lead, reconciles stale plans, and optionally publishes plans as GitHub issues via --issues. - Use Case: Point it at a legacy service and ask for a security-focused audit; it returns a vetted findings table, then writes ordered plans (e.g. rotate a committed credential, add authorization checks) that a cheaper executor model can carry out independently. ## Quick Start Ask the agent to audit this repository with the improve skill and turn the top findings into executable plans.
