bmad-code-review

Adversarial code review on a pull request is automated by launching parallel reviewer roles to surface different problem classes. Findings are then deduplicated and triaged into decision_needed, patch, defer, or dismiss buckets for engineers.

Triage in automated diff analysis is the process of normalizing and deduplicating code review findings. It classifies issues into decision_needed, patch, defer, or dismiss buckets so engineering teams can prioritize and fix acceptance and logical problems faster.

To review uncommitted changes against a spec file, the Skill constructs a diff from your local workspace and audits it with spec awareness. It runs an Acceptance Auditor role to ensure your code changes align with the defined sprint context and story requirements.

Yes, triaged code review findings can be synced to sprint tracking artifacts. The workflow integration writes categorized findings back to story files, appends deferred items, and updates story status to keep engineering backlogs current.

Automated code review works with commit ranges, local branches, staged changes, and provided diffs. It constructs the diff from these inputs to run parallel adversarial reviews and audit the changes against your spec files and sprint context.