What problem does it solve? Reviewing large pull requests or entire codebases with a single agent causes context degradation and missed defects. This Skill splits the work into orthogonal functional clusters, audits each with an isolated subagent, and synthesizes findings into a prioritized blast-radius scorecard. ## Core Features & Use Cases - Automated Cluster Discovery: A deterministic Python engine partitions git diffs or whole repositories into domain-based clusters with associated test targets and volume metrics. - Parallel Subagent Audits: Each cluster is reviewed by an isolated subagent that runs its test suite first and reports only empirically grounded findings with exact file:line citations. - Cross-Boundary Contract Verification: The orchestrator validates exported symbols, signatures, and shared schemas across cluster boundaries, then produces a P0/P1/P2 scorecard with blast-radius assessment. - Use Case: Before merging a 2,000-line feature branch touching physics, IO, and transforms modules, run the audit to get per-cluster verdicts, a contract matrix, and a merge action plan. ## Quick Start Ask the agent to run a codebase audit on the current branch diff against main and produce a prioritized findings scorecard.
