What problem does it solve? Review findings from external reviewers, CI audits, or PR comments are often stale, overstated, or wrong, and acting on them blindly wastes effort or introduces regressions. This Skill forces every finding to be treated as a claim and classified with concrete evidence before any code is changed. ## Core Features & Use Cases - Evidence-Based Classification: Labels each finding as CONFIRMED, PARTIALLY_VALID, DISPROVED, UNVERIFIED, PRE_EXISTING, or OUT_OF_SCOPE based on source, diff, runtime, or test evidence. - Scoped Review Workflow: Establishes the reviewed scope via branch, base, PR number, and git diff before validating any claim. - Behavior-Level Validation: Prefers running the smallest safe command and checking callers over trusting source-string matches or stale line numbers. - Use Case: Given a bundle of 12 requested-changes comments on a pull request, classify each one with cited evidence, fix only the confirmed items, and document why the rest were rejected. ## Quick Start Validate these review findings against the current branch and classify each one with evidence before making any changes.