codex-review

An adversarial review for architecture decisions evaluates non-diff artifacts to catch blindspots and ensure unbiased analysis. You can run a one-shot Codex review on your spec or decision document to get categorized findings before implementation.

Adversarial plan review is a high-reasoning analysis of non-diff artifacts like debugging hypotheses and architecture decisions. You need it before starting implementation of a complex plan to catch blindspots and validate soundness.

Yes, Codex provides one-shot adversarial reviews for debugging hypotheses. It analyzes your non-diff artifacts and returns categorized findings marked as BLOCKING, IMPORTANT, or ADVISORY, plus AGREE notes for balanced feedback.

Yes, you need the Codex CLI installed with appropriate permissions to perform adversarial architecture decision reviews. The Skill relies on Codex to execute the high-reasoning analysis on your planning and decision documents.

The best way to validate a complex software architecture plan is to get a second opinion through an adversarial review. This one-shot analysis categorizes potential issues as BLOCKING, IMPORTANT, or ADVISORY to ensure comprehensive and unbiased analysis.

Adversarial plan review categorizes findings into BLOCKING, IMPORTANT, and ADVISORY levels to prioritize issues, while also providing AGREE notes to balance the critique and highlight sound elements of your architecture decisions.