What problem does it solve? Code changes often land with hidden regressions, security gaps, or scope drift from the user stories they were meant to implement. This Skill performs a structured, high-confidence review of a change's diff before it ships, filtering out noise so only real, change-introduced issues surface. ## Core Features & Use Cases - Multi-lens diff review: Judges the diff across correctness, error handling, security, test coverage, scope fidelity against acceptance criteria, and repo conventions. - Confidence-scored findings: Scores every candidate finding 0-100 and discards anything below 80, eliminating pre-existing issues, linter-catchable problems, and nitpicks. - Severity-ranked verdict: Tags surviving findings as blocking, important, or minor, fixes clear-cut defects, and only finishes the stage when no blocking or important finding remains. - Use Case: Before merging a branch implementing a user story, run the review to verify the diff satisfies the story's acceptance criteria, introduces no security holes, and keeps existing tests valid. ## Quick Start Ask your AI agent to review the current change's diff using the defprod-change-review skill and report any blocking or important findings.
