adversarial-review

To run an adversarial code review, trigger the review workflow on your diffs. Three specialized sub-agents simultaneously analyze the changes from security, quality, and test perspectives, then a synthesizer merges their findings to surface issues.

Multi-perspective adversarial code review coordinates independent reviewers to analyze the same diff simultaneously. Use it for security-sensitive changes, refactors, and feature additions across codebases when multiple stakeholders require independent perspectives.

The synthesizer merges findings from the security, quality, and test reviewers using a defined consensus workflow. Disagreements between perspectives are treated as signal rather than noise to catch issues no single reviewer could find.

Yes, you can use this workflow for refactors across codebases. The review process collects full file context and diffs, applying specialized parallel sub-agents to ensure security, quality, and test coverage for complex changes.

No, the adversarial code review workflow operates without external dependencies. It implements parallel sub-agents internally to analyze diffs and synthesize findings using its own defined consensus workflow.

A three-way adversarial review provides deep, narrow focus from specialized security, quality, and test reviewers simultaneously. This approach catches issues no single reviewer ever could by treating disagreements between perspectives as signal.