What problem does it solve? Diff-only code reviews miss broken caller contracts, duplicated bugs in similar implementations, and design inconsistencies because they never read the code surrounding the change. This Skill forces a structured deep review that reads callers, similar implementations, related tests, and downstream consumers before judging a pull request. ## Core Features & Use Cases - Four-lens evaluation: Judges specification compliance, code correctness (bugs, security, performance), code quality (naming, complexity, duplication, abstraction), and overall design consistency as separate dimensions. - Mandatory context gathering: Uses grep-based exploration to enumerate callers, find similar implementations needing the same fix, and trace how changed values are handled downstream. - Fabrication prevention and verdicts: Requires every Important-level finding to cite actually-read files and lines, and outputs a merge verdict (mergeable / needs fixes / not mergeable) with severity-rated findings. - Use Case: Before merging a multi-file refactor, ask for a deep review; the Skill greps all call sites of changed functions, checks whether sibling modules need the same fix, and reports contract mismatches with file:line evidence. ## Quick Start Ask the AI to deeply review the current branch diff against main, including surrounding code, callers, and design consistency.
