adversarial review gate — a dedicated, read-only skeptic dispatched after an impl wave, before merge-to-main / bead-close, whose only job is to disprove "done." Hunts stub bodies, mock-the-unit tests, unverified acceptance criteria, unreproducible runtime claims, composition gaps. Distinct from /check (which interrogates the plan, pre-impl).

Dispatch a read-only adversarial review gate post-implementation to disprove completion claims. It hunts stub bodies, mocked tests, and composition gaps, routing verdicts to SHIP, FIX-FIRST, or REJECT before merging to main.

Evidence-based pre-merge validation is an adversarial review process that requires file:line locations to support findings. It targets unverified acceptance criteria and unreproducible runtime claims, ensuring test reports accurately reflect implementation scope.

Run an adversarial review gate after your implementation wave to scrutinize test integrity. The read-only skeptic identifies mocked unit tests and missing error paths, appending file:line findings and a verdict to impl bead notes before merge.

No, an adversarial quality gate requires read-only execution. It performs code scrutiny to disprove false completion without modifying files, routing a SHIP, FIX-FIRST, or REJECT verdict based strictly on evidence-based findings.

You need an adversarial review gate after implementation, right before merge-to-main. It is distinct from pre-implementation checks, specifically targeting composition gaps and unreproducible runtime claims in delivered code rather than interrogating the plan.

The gate cannot invent issues and only validates post-implementation, pre-merge surfaces. It requires real implementation to scrutinize and relies on file:line evidence, meaning it cannot catch quality failures in code paths lacking explicit acceptance criteria.