adversarial-review

Name: adversarial-review
Availability: InStock
Author: Compozy

Spawn opposite-model reviewers to critique code and produce verdicts.

2.5k|134|Updated Mar 28, 2026

Search Tags:#review #code-review #adversarial #workflow-orchestration #ai-assisted #opposite-model #lens-based

Ready to run

npx skills add https://github.com/compozy/compozy --skill adversarial-review-compozy

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: adversarial-review
Source: https://github.com/compozy/compozy/tree/main/.agents/skills/adversarial-review
Command: npx skills add https://github.com/compozy/compozy --skill adversarial-review-compozy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Adversarial code review orchestrates cross-model critique by spawning reviewers on the opposite model to challenge work and surface critical issues without altering the source.

Core Features & Use Cases

Spawn reviewers on the opposite model (Claude vs Codex) to attack work from distinct lenses grounded in brain principles.
Generate a structured verdict and per-lens findings, mapping each critique to a specific principle and providing concrete remediation.
Automate the review orchestration after large diffs, planning phases, or cook sessions to improve review coverage and reduce blind spots.

Quick Start

Trigger an adversarial review after a substantial change to obtain a cross-model verdict and actionable findings.

Frequently Asked Questions about adversarial-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I run an adversarial code review to catch blind spots after a large diff?▼

Adversarial code review orchestrates cross-model critique by spawning reviewers on the opposite model to attack work from distinct lenses. Trigger it after large diffs or planning sessions to obtain a synthesized verdict and per-lens findings without altering the source.

What is cross-model AI code critique and how does it surface failure modes?▼

Cross-model AI code critique spawns reviewers on the opposite model, such as Claude versus Codex, to challenge work. It surfaces diverse failure modes by grounding each critique in a specific brain principle and providing concrete remediation.

When should I use an opposite-model review instead of a standard code review?▼

Use an opposite-model review when standard reviews miss edge cases after large diffs, planning phases, or cook sessions. It enforces cross-model CLI execution with standardized prompts to reduce blind spots and improve review coverage.

Can I use this adversarial review workflow with both Claude and Codex models?▼

Yes, the adversarial review workflow explicitly supports spawning reviewers on the opposite model, pitting Claude versus Codex. It requires opposite-model CLI execution to generate structured per-lens findings and a synthesized verdict.

What format do the outputs take when running an automated adversarial code review?▼

Automated adversarial code review enforces outputs including a synthesized verdict and per-lens findings. Each critique maps to a specific principle and provides concrete remediation steps for actionable review coverage.

Does adversarial code review modify my source code during the critique process?▼

Adversarial code review does not alter the source. It orchestrates cross-model critique strictly to surface critical issues and generate a structured verdict, leaving your original codebase untouched.