adversarial-review

An adversarial review hunts for hidden regressions and unsafe assumptions by applying a skeptical analysis pass to your uncommitted workspace, base branch, or explicit commit SHA before high-stakes merges or deployments.

Yes, you can focus the review by forwarding a specific threat model via a prompt argument, steering the skeptical analysis toward targeted concerns like security, rollback safety, billing, or data loss to generate a risk-led report.

Checking release risk for migrations involves running an adversarial review against a specific commit SHA or branch to proactively surface hidden regressions and unsafe assumptions before deploying high-stakes engineering changes.

Yes, the workflow supports reviewing a specified base branch by passing the correct review target flags, analyzing either the current uncommitted workspace by default or an explicit commit SHA.

To generate a risk-led report for pre-deploy checks, run a skeptical Codex review against your targeted engineering changes while forwarding a focused threat model via the prompt argument to evaluate unsafe assumptions and regressions.