skills.rest
rodrigobranas
rodrigobranasCOMMUNITY
Explore all skills by this author β†’
Website𝕏@rodrigobranasGitHub

adversarial-review

Spawn opposing-model reviewers to critique code diffs and synthesize verdicts.

1|1|Updated Apr 18, 2026
Search Tags:#code-review#cross-model#ai-safety#adversarial-review#review-lenses#brain-principles
One-click install
npx skills add https://github.com/rodrigobranas/formacao_ia_arq_t1_4 --skill adversarial-review-rodrigobranas
Or copy as Structured Prompt for Agentβ–Ό
Please help me install this Agent Skill.
Skill: adversarial-review
Source: https://github.com/rodrigobranas/formacao_ia_arq_t1_4/tree/main/mastra-rag/.agents/skills/adversarial-review
Command: npx skills add https://github.com/rodrigobranas/formacao_ia_arq_t1_4 --skill adversarial-review-rodrigobranas

SYSTEM DOCUMENTATION & REQUIREMENTS

πŸ’‘ This Skill includes references (resource) components.

What problem does it solve?

Adversarial code review uses the opposite AI model to critique work, surfacing blind spots and quality gaps before shipping. It automatically spawns 1–3 reviewers from the opposing model to challenge the submission with distinct lenses and returns a synthesized verdict rather than making changes directly.

Core Features & Use Cases

  • Spawn reviewers on the opposite model (Claude spawns Codex, Codex spawns Claude) to critique diffs from multiple lenses.
  • Provide a structured verdict with prioritized findings (high/medium/low) and explicit evidence citations.
  • Enforce hard constraints: reviewers do not modify code; the output is actionable feedback for the author.
  • Triggers: "adversarial review" after large diffs, planning sessions, or major changes.

Quick Start

Trigger an adversarial review after a large diff to spawn 1–3 reviewers on the opposite model and synthesize a final verdict.

Frequently Asked Questions about adversarial-review

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

FAQPage Schema
What is adversarial code review and how does cross-model critique work?β–Ό

Adversarial code review uses the opposite AI model to critique work, spawning 1–3 reviewers from the opposing model to challenge diffs with distinct lenses. It synthesizes a final verdict with categorized findings rather than modifying code directly.

How do I run an adversarial review on a large diff?β–Ό

Trigger an adversarial review after a large diff, planning session, or major change. The system automatically spawns reviewers on the opposite AI model, such as Claude spawning Codex or Codex spawning Claude, to apply multiple review lenses.

Can I use cross-model reviewers for planning sessions instead of code changes?β–Ό

Yes, adversarial review applies to planning sessions or major changes, not just large diffs. The opposing model reviewers challenge the submission with distinct lenses to surface blind spots and quality gaps before shipping.

Does the adversarial review process modify my source code directly?β–Ό

No, reviewers operate under hard constraints and do not modify code. The output is actionable feedback for the author, providing a structured verdict with prioritized high, medium, and low findings and explicit evidence citations.

What is the best way to surface blind spots in AI-assisted code changes?β–Ό

Using cross-model adversarial review is a strong approach to surface blind spots, as it forces the opposite AI model to critique the work. It provides a synthesized verdict with explicit references to guiding brain principles and categorized findings.

Related Skills

Explore More Software Engineering Skills→
tense-itense-i
1

adversarial-review

Orchestrate adversarial code reviews with opposing-model reviewers and synthesized verdicts.

Community
Advanced
alexanderopalexanderop

reviewing-notes

Audit markdown notes for broken wiki-links, missing summaries, tags, and orphans.

Community
Intermediate
mikewolfdmikewolfd

semi-formal-code-review

Analyze code changes to produce evidence-backed reviews with file:line citations.

Community
Advanced
compozycompozy
2.5k

adversarial-review

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

Official
Advanced