critical-review

Automated adversarial review dispatches parallel reviewer agents to stress-test lifecycle artifacts, synthesizing their findings into a structured critique. This process surfaces hidden risks in plans and specs before project commitments are finalized.

Parallel adversarial review is a technique that dispatches multiple reviewer agents simultaneously to challenge an artifact from distinct angles. It derives tailored challenges based on artifact contents and project context, consolidating findings into a synthesized narrative.

Yes, the critical-review process operates within Claude Code lifecycle workflows to adversarially review research artifacts. It loads project context to tailor challenges and outputs structured results including per-angle envelopes and a synthesized Objections narrative.

Synthesizing parallel critique findings is best handled by consolidating per-angle envelopes into a structured narrative. The process generates an actionable Objections narrative and optionally persists B-class findings as residue data for morning reports.

Artifact integrity during critical reviews is maintained through fallback and verification mechanisms. The process includes atomic SHA checks and drift detection to ensure the reviewed artifact remains unaltered and verified throughout the adversarial cycle.

Adversarial reviews should be run on specs and plans immediately before making project commitments. Running parallel reviewer agents at this stage stress-tests the artifacts to surface hidden risks and generate actionable objections prior to finalization.