Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
ng avatar

Adversarial Review

  • 11 repo stars
  • Updated July 6, 2026
  • ng/adversarial-review

adversarial-review is a Claude Code plugin marketplace entry that runs adversarial multi-model code review with progressive cost-gating so developers can catch and optionally fix consensus issues on pull requests.

About

adversarial-review is a Claude Code marketplace plugin that performs adversarial, multi-model code review on pull requests with mechanical checks first, then Optimizer and Skeptic agents that challenge findings before auto-fixing consensus issues. Use it when you want deep review with cost gating, structured reports, and optional CI automation via the documented GitHub Action. It supports review-only mode with --no-fix when you only need findings.

  • Mechanical lint, typecheck, build, and test before paid agent review
  • Optimizer and Skeptic agents with progressive cost-gating
  • Auto-fix consensus Critical/Major issues with bounded verification
  • GitHub Action ng/adversarial-review@v1 for PR workflows

Adversarial Review by the numbers

  • Data as of Jul 13, 2026 (Skillselion catalog sync)
/plugin marketplace add ng/adversarial-review

Add your badge

Show developers this marketplace is listed on Skillselion. Paste this into your README.

Listed on Skillselion
repo stars11
Last updatedJuly 6, 2026
Repositoryng/adversarial-review

How do you get rigorous, multi-model code review on PRs without running expensive agents on trivial diffs?

Runs adversarial multi-model PR reviews with mechanical gates, Optimizer and Skeptic agents, optional auto-fix, verification loops, and GitHub Action integration.

Who is it for?

Teams using Claude Code or GitHub Actions who want structured adversarial review with optional auto-fix on pull requests.

Skip if: Changes that only need a quick human skim without mechanical checks, agent spend, or PR comment noise.

What you get

A structured review report, PR or MR comments, and optionally applied consensus fixes after verification iterations.

Recommended Marketplaces

FAQ

What agents perform the review?

The Optimizer and The Skeptic review independently and challenge each other per the README pipeline.

Can I disable auto-fix?

Yes; use --no-fix flags documented for /Adversarial Review:run and the GitHub Action mode input.

Why avoid synchronize triggers in CI?

The README warns reviews are slow and expensive, recommending labeled re-runs instead of every push.

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.