verify

To validate implementation claims against repository truth, run an adversarial evaluation pass that checks concrete claims, diffs, and implementations against actual code and required specs. This surfaces hidden gaps and produces a findings-first report indicating blocked checks.

Auditing a concrete plan for missing specs and test sync requires an adversarial review against repository truth to identify scope blockers. The process yields concrete findings on plan safety, updates plan files with verdicts, and flags when next slices are unsafe to execute.

Yes, you can review implementation diffs before merging by applying adversarial attack-style evaluation against repository truth and required specs. This prevents hidden gaps by rigorously validating correctness and traceability before code integration.

Findings-first plan review works by adversarially evaluating concrete targets against repository truth and required tests, surfacing concrete issues with scope and blockers upfront. It guarantees a report focused on identified flaws and clearly indicates any blocked checks.

You should update plan files with verification verdicts when the evaluated target is an explicit plan and next execution slices are not safe. This records the adversarial findings directly into the plan, ensuring downstream work is blocked until issues are resolved.

Yes, evidence-based claims validation works without external dependencies, operating natively on repository truth, required specs, and bounded tests. It requires no additional environment setup to perform adversarial evaluations and produce findings-first reports.