bmad-code-review

Adversarial code review on a pull request is performed by running parallel review layers—Blind Hunter, Edge Case Hunter, and Acceptance Auditor—on the diff to surface issues. You provide the code diff and trigger the review process.

To find edge cases in a branch diff, use a multi-layer code review approach that includes an Edge Case Hunter. It analyzes the changes alongside integration concerns to surface potential boundary failures before merging.

Yes, you can check code changes against acceptance criteria using an Acceptance Auditor layer. It evaluates the provided diff and optional specifications to ensure the implementation aligns with the intended requirements.

Code review findings from a commit range are triaged by normalizing and deduplicating the results. They are then classified into categories like patch, defer, or reject to present actionable decisions.

Yes, this code review works without a specification by focusing on the provided diff. However, supplying optional specs and context enhances the Acceptance Auditor layer's ability to verify alignment with intent.

Adversarial review should be used when evaluating risky changes, security-sensitive patches, or complex feature rollouts. It reduces risk by orchestrating parallel review layers to surface hidden issues before merging.