bmad-code-review

Adversarial code review automates multi-layer analysis on diffs to surface edge cases and ensure specification compliance. It uses parallel reviewer roles to identify coverage gaps and produces triaged findings for quick action.

Triage code review findings into structured categories like patches, decisions, defers, and dismissals. A unified findings format provides location references, priority levels, and recommended actions to streamline the review process.

Specification alignment during code review is ensured by loading optional spec and project documentation as context. This diff-driven context awareness informs the review, reduces ambiguity, and verifies compliance with requirements.

Yes, you can run parallel reviewer roles such as Blind Hunter, Edge Case Hunter, and Acceptance Auditor on code diffs. These roles operate simultaneously to identify gaps in coverage and specification conformance.

Step-file driven workflows enforce reproducible code reviews by maintaining clear progression and generating auditable artifacts. This deterministic process loads diffs, specs, and context documents across multiple review layers.

Automated triage for software development workflows relies heavily on the quality of provided diffs, specs, and context documents. Limitations arise when optional project documentation is missing, increasing ambiguity during the automated review.