adversarial / default

Adversarial code review stress-tests an action's plan and draft by running independent critic passes to identify failure modes, audience misinterpretations, and bash safety issues. It returns a ship or hold decision with a risk-calibrated findings trail to block unsafe deployments.

To run an adversarial review on a pull request, invoke the skill against the PR. It executes parallel critic passes for failure-mode, audience, second-order, and bash-invocation analysis, then posts PR issue comments with a risk-calibrated findings trail and a ship or hold decision.

Adversarial review applies to fallback reviews for general playbooks and ad-hoc PR workflows. It evaluates blockers across the plan, draft, and recorded execution trail to determine if an action should ship or be held for rework.

The best way to assess bash invocation risks is using an adversarial review that includes a dedicated bash-invocation analysis critic pass. This identifies potential safety issues in recorded commands and adds them to a durable audit trail for governance.

Yes, review thoroughness scales for high-risk actions by adding domain-specific reviewers if configured for the playbook. This risk-calibrated depth ensures critical changes receive more rigorous failure-mode and second-order consequence analysis.

You should hold a pull request when adversarial review identifies blockers across the plan, draft, or execution trail. It returns a hold decision with a rework brief detailing the failure modes and risks found, preventing unsafe outputs from shipping.