plan-review

Parallel code review is orchestrated by coordinating three independent reviewers in an Agent Teams leader-delegate setup to assess logic, quality, and performance simultaneously, followed by a cross-check phase to validate findings before compilation.

An Agent Teams leader-delegate setup works by assigning three delegate reviewers to evaluate code independently, then the leader collects and cross-checks their findings to compile a final consolidated review report, ensuring faster and more reliable results.

Yes, you can use Git branches to define the review scope. The review process identifies active tasks from Git branches and supporting spec files to automatically determine what code changes require evaluation.

The final code review report is saved as the authoritative artifact in `.spec/{slug}/review.md`, preserving traceability and enabling downstream QA by storing the consolidated leader report within the project spec.

Prerequisites include an active project with supporting spec files like `.spec/{slug}/files.md` when available, and adherence to the guidelines and prerequisites described in the plan-review workflow before initiating the review.

You should avoid parallel agent-based code review if your project lacks active Git branches for task discovery or does not support the Agent Teams leader-delegate architecture required to coordinate the three reviewers.