What problem does it solve? Deep pull request reviews often miss critical bugs or drown reviewers in unvalidated false positives. This Skill structures PR review as a multi-phase pipeline that maximizes recall through parallel exploration while enforcing independent validation before any finding is reported. ## Core Features & Use Cases - Parallel Explorer Lanes: Six base lanes (correctness, security, dependencies, docs/intent, tests, performance) plus triggered micro-lanes for risk areas like git safety, schema drift, and config ratchets. - Independent Validation Pipeline: Reviewer subagents confirm, disprove, or classify every candidate with file:line evidence, followed by critic challenge for HIGH/CRITICAL findings. - Obligation Extraction & Metrics: Reconstructs what the PR must deliver from checkboxes, issues, and commits, then records review quality metrics and knowledge writeback. - Use Case: When reviewing a large feature branch touching authentication and schema code, run this workflow to get grouped, evidence-backed findings with a merge recommendation (APPROVE, REQUEST_CHANGES, or BLOCK) instead of ad-hoc comments. ## Quick Start Ask the AI to run a swarm PR review on the current branch diff against main and report validated findings with a merge recommendation.
