What problem does it solve? AI-generated and rapidly iterated codebases accumulate reviewability debt: stale comments, duplicated documentation, speculative compatibility paths, diff narratives, and tests that mirror implementations. This Skill audits and repairs those artifacts so human reviewers can understand and verify changes without trusting unverified summaries. ## Core Features & Use Cases - Audit and repair modes: Inspect and report findings without editing, or audit, edit, and verify when asked to clean or fix. - Rubric-based classification: Apply truth, durability, signal, locality, proportion, scope, verification, and preservation gates from references/audit-rubric.md, with severity-ranked findings. - Preservation-first repair: Fix false contracts, remove dead paths and speculative compatibility, consolidate duplicated docs, and replace implementation-mirroring tests while protecting human edits and durable knowledge. - Use Case: After a large AI-generated pull request, run the skill against the diff to prune transient narratives, stale comments, and redundant tests, then verify the final diff is smaller to understand than the problem it solves. ## Quick Start Use the audit-reviewability skill to audit this pull request for reviewability debt and repair the findings.
