Adversarial Review

Automated critique-fix loop using fresh-eyes reviewer subagents. Each round, a new subagent reviews the code cold — no knowledge of why decisions were made, no attachment to the implementation. The main agent fixes Critical and Suggestion findings, then a fresh reviewer checks again. Loop stops when only Nitpicks remain or after 3 rounds.

When to Use

• After implementation, before committing — especially for features and refactors
• When changes touch critical paths, security-sensitive code, or complex logic
• When you want a review that's harder than what you'd give yourself
• Position in pipeline: implement → /simplify → /adversarial-review → /finalize → commit

When NOT to Use