What is adversarial-code-review?

A code-review process in which separate agents write or inspect changes, challenge the findings, and filter feedback. A pull request is a proposed code change submitted for review before it is merged.

It helps review pull requests before merging, run automated reviews in CI, and critique implementation plans or designs.

It reduces low-value review comments and focuses attention on issues that are both important and likely to be real.

adversarial-code-review — Skill by mlopscommunity

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent

npx skills add mlopscommunity/Coding-Agents-Conference-skills --skill adversarial-code-review

Clone the repo

git clone --depth 1 https://github.com/mlopscommunity/Coding-Agents-Conference-skills

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/mlopscommunity/coding-agents-conference-skills/adversarial-code-review/github.svg)](https://agentmods.dev/skills/mlopscommunity/coding-agents-conference-skills/adversarial-code-review)

Your own site

<a href="https://agentmods.dev/skills/mlopscommunity/coding-agents-conference-skills/adversarial-code-review"><img src="https://agentmods.dev/badge/skills/mlopscommunity/coding-agents-conference-skills/adversarial-code-review/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/mlopscommunity/coding-agents-conference-skills/adversarial-code-review"><img src="https://agentmods.dev/badge/skills/mlopscommunity/coding-agents-conference-skills/adversarial-code-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 47 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 2,248 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00047	$0.02248
Opus 5.5	$0.00019	$0.00899
Sonnet 5.5	$0.00009	$0.00450
Haiku 4.5	$0.00005	$0.00225

Measured 2mo ago against content hash a9539a162637, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

adversarial-code-review scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/adversarial-code-review/SKILL.md · 185 lines

How it starts

The opening of the file, as written. The whole thing — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Code Review

Overview

A multi-agent review pattern where one agent builds (or authors), a second agent critiques the code, and a third agent critiques the review itself. This layered adversarial approach filters out low-value nitpicks and surfaces only high-confidence, high-priority issues that deserve human attention.

Core principle: Fewer, higher-quality review comments build trust. Filter ruthlessly to high confidence + high priority only. Target roughly two comments per PR.

Dependency: Claude Code CLI with --append-system-prompt support. Optionally, a different model for the review pass than the one used for writing.

When to Use

Reviewing pull requests before merge, especially when review quality matters more than speed
As a CI-integrated automated reviewer that developers actually read instead of ignore
When existing automated reviews produce too much noise and developers have stopped trusting them
During versioned critique cycles where a plan or design needs iterative refinement

When NOT to Use

Trivial PRs (typo fixes, dependency bumps, single-line config changes)
When you need instant feedback during live pairing sessions (too slow for interactive use)
As a replacement for human review on security-critical or compliance-gated changes

Common Mistakes

Mistake	Why it's wrong
Surfacing every finding to the developer	Noise kills trust. Developers stop reading reviews that cry wolf. Filter to ~2 high-priority, high-confidence comments per PR.
Using the same model for writing and reviewing	The model is biased toward its own patterns. Use a different model for review than the one that wrote the code — it catches different classes of issues.
Skipping the meta-reviewer (third agent)	Without a check on the reviewer, you get false positives and nitpicks dressed up as critical findings. The meta-reviewer filters the reviewer's output.
Running adversarial review without priming the critic	A neutral prompt produces polite, hedging reviews. Tell the reviewer the code likely contains bugs to prime it for genuine criticality.
Treating all review comments as equal priority	Without confidence and priority scoring, developers cannot triage. Every comment must carry explicit confidence (high/medium/low) and priority (high/medium/low).

Read the full file on GitHub · 185 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 185 lines · 47 tokens per session scan A a9539a162637

Subscribe to this mod's changes

adversarial-code-review is a skill published in the GitHub repository mlopscommunity/Coding-Agents-Conference-skills (39 stars, last pushed 7mo ago), licensed Apache-2.0. It adds 47 tokens to every session and 2,248 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Code review

adversarial-code-review next to 6 skills from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
adversarial-code-reviewmlopscommunity/Coding-Agents-Conference-skills	Skill	39	7mo ago	Apache-2.0	A	47	Claude Code, Codex
autoreviewopenclaw/clawhub	Skill	9.6k	today	MIT	A	17	Claude Code, Codex
omh-code-reviewrlaope/oh-my-hermes	Skill	5.5k	today	MIT	A	49	Claude Code, Codex
agent-teams-simplify-and-hardenLeoYeAI/openclaw-master-skills	Skill	2.2k	6d ago	MIT	A	115	Claude Code, Codex
revdiff-planumputun/revdiff	Skill	953	today	MIT	A	84	Claude Code
re0-mergeLilMGenius/paperthin	Skill	1.4k	yesterday	MIT	A	51	Claude Code, Codex
explore-unknownsdzhng/skills	Skill	1.5k	today	MIT	A	162	Claude Code, Codex