What problem does it solve? Single-pass code reviews tend to under-focus when one reviewer juggles multiple concerns at once, so security holes, hollow tests, and logic bugs slip through. This Skill runs three independent, narrowly-scoped reviewers in parallel and reconciles their verdicts into one actionable call. ## Core Features & Use Cases - Parallel multi-lens review: Launches three read-only subagents simultaneously — security, test-quality, and design-correctness — each blind to the others so findings stay independent. - Structured JSON findings: Every lens returns a verdict (pass/concerns/fail) plus severity-tagged findings with file:line references and one-line recommendations. - Verdict reconciliation: Combines the three verdicts into an overall GO, CAUTION, or STOP call, re-reading the source code itself when lenses disagree rather than defaulting to the loudest finding. - Use Case: Before merging a change to auth.py, run a review swarm to get a security audit, a check that the tests actually assert meaningful behavior, and a logic review — all in the wall-clock time of a single review pass. ## Quick Start Run a review swarm on src/auth.py to get security, test-quality, and design findings with an overall GO, CAUTION, or STOP verdict.
