adversarial-reviewer

adversarial-reviewer is an agent for Claude Code from recomposesh/recompose. It costs 46 tokens per session (221 once invoked), scanned A, original, MIT.

A code-review agent that looks for correctness, security, and design problems that automated checks may miss. It reviews a change without editing it and returns findings ordered by severity.

In plain words
What is it for?
Use it before a feature commit to inspect a diff against its purpose and acceptance criteria, identify likely breakages, and flag issues that need a human decision.
Why use it?
Passing linters and tests does not prove that a change is safe or correct in real use. Concrete failure scenarios make hidden risks easier to assess.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter.

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.

agentmods
npx agentmods add agents/recomposesh/recompose/adversarial-reviewer
Clone the repo
git clone --depth 1 https://github.com/recomposesh/recompose

Made for: Claude Code.

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.

agentmods badge for adversarial-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/recomposesh/recompose/adversarial-reviewer.svg)](https://agentmods.dev/agents/recomposesh/recompose/adversarial-reviewer)
Your own site
<a href="https://agentmods.dev/agents/recomposesh/recompose/adversarial-reviewer"><img src="https://agentmods.dev/badge/agents/recomposesh/recompose/adversarial-reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 46 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 221 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00046 $0.00221
Opus 5 $0.00023 $0.00111
Sonnet 5 $0.00009 $0.00044
Haiku 4.5 $0.00005 $0.00022

Measured 6d ago against content hash f10cbaeb3aa8, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

adversarial-reviewer 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 6d 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.

.claude/agents/adversarial-reviewer.md · 16 lines

What it actually says

You attack the diff to find what passes the linters and still breaks. You read the change and its context, and you change nothing.

Expect the diff, the feature intent, and the acceptance criteria. Project memory carries the standing failure patterns worth a second look. Read them before you start.

Return findings ranked most severe first. For each finding, give the location, the defect, and a concrete failure scenario that shows the break. When the diff holds up, say so in one line rather than pad the list.

The review runs as a pair. Two instances review the same change, and the dispatch overrides one seat to the most capable model for deliberate model diversity. Report your findings to the caller, and flag anything that needs a human call.

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.

  1. 6d ago First seen · 16 lines · 46 tokens per session scan A f10cbaeb3aa8

Subscribe to this mod's changes

adversarial-reviewer is an agent published in the GitHub repository recomposesh/recompose (27 stars, last pushed 9d ago), licensed MIT. It adds 46 tokens to every session and 221 once invoked, about $0.0002 per session on Opus 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.