bmad-code-review

Adversarial code review surfaces defects by running parallel review layers like Blind Hunter and Edge Case Hunter concurrently against a diff. This structured triage approach categorizes findings into actionable groups to guide remediation and ensure alignment with intent.

To run a code review, instruct the AI by saying "run code review" or "review this code" and provide your diff when prompted. The Skill then executes parallel review layers and returns structured triage findings with clear locations for remediation.

Yes, you can validate code changes against a spec by providing optional spec or context documents. The Acceptance Auditor layer specifically checks your diff to ensure acceptance criteria are met and the implementation aligns with the intended design.

The best way to triage code review findings is using a structured triage framework that categorizes issues from parallel review layers into actionable groups. This provides clear locations in the diff for targeted remediation and improves overall spec alignment.

Yes, the code review method works without a formal specification document because spec and context documents are optional. It will still adversarially review your code changes using parallel layers to surface defects, though spec validation will be limited.

Parallel review layers like Blind Hunter, Edge Case Hunter, and Acceptance Auditor operate concurrently to surface different types of issues that standard diff analysis might miss. This adversarial approach ensures comprehensive defect detection and structured triage.