What is adversarial-review CLAUDE.md?

Instructions for an adversarial code-review plugin used with Claude Code. They explain how to send work to another AI model for critique and compare that review with the host agent’s analysis.

Reviewing code, plans, and prompts across different AI systems. They help choose a reviewer, handle provider failures, and identify which findings were independently confirmed.

They provide a defined review process instead of relying only on the agent that produced the work. They also describe fallback reviewers and warnings when independent review is unavailable.

adversarial-review CLAUDE.md — Instructions file for Claude Code and Codex by robertoecf: Instructions for an adversarial code-review plugin

Install

Getting it into your agent

There is no command for this one: it runs only inside a plugin, and the catalogue could not identify which plugin ships it. The source is linked below.

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/instructions/robertoecf/adversarial-review/claude-md.svg)](https://agentmods.dev/instructions/robertoecf/adversarial-review/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/robertoecf/adversarial-review/claude-md"><img src="https://agentmods.dev/badge/instructions/robertoecf/adversarial-review/claude-md.svg" alt="Measured on agentmods" height="20"></a>

Per session 1,599 This file is loaded in full into every session.

When invoked 1,599 The same file — it is already loaded in full.

Security scan B 1 finding. 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.

Model	Per session	Once invoked
Fable 5.1	$0.01599	$0.01599
Opus 5	$0.00800	$0.00800
Sonnet 5	$0.00320	$0.00320
Haiku 4.5	$0.00160	$0.00160

Measured 6d ago against content hash e70ff493b964, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade B, and why

adversarial-review CLAUDE.md scanned grade B with 1 finding 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

- **`forced_login_method = "chatgpt"`** must be in `~/.codex/config.toml` for

CLAUDE.md · 116 lines

How it starts

The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md - Claude Code directives

Plugin: adversarial-review v0.9.4

Cross-host adversarial review for coding workflows. Detects which agent host the SKILL.md is running under and routes review to another model family. Claude code reviews use direct-xAI Grok 4.5 first, while Claude plan reviews use Codex Luna first. Codex-host routing is author-aware: Codex-authored artifacts use the Pi Grok chain first, while Grok-authored artifacts use Codex Luna first. Grok and Pi hosts use Codex first. Each route has explicit secondary providers, then a quality-first Antigravity Claude, Gemini, and GPT ladder, then degraded host-self with explicit warning.

v0.9.4 adds author-aware Codex routing through ADVERSARIAL_REVIEW_AUTHOR and changes the Antigravity default to Gemini 3.7 Flash High. It also accepts Antigravity's tabular model catalog and falls through on empty provider output. The previous release replaced the OpenCode Go fallback with DeepSeek V4 Flash xhigh and kept the Luna reviewer at max effort. v0.9.1 originally made Gemini 3.6 Flash High the first Antigravity candidate. Routing policy 2026-07-23 makes Grok 4.5, xAI's frontier model, the sole direct-xAI model at high effort. v0.9.0 routes direct Moonshot review to Kimi K3 xhigh and adds quota-aware Antigravity failover across providers. v0.8.2 pinned every Codex reviewer invocation to GPT-5.6 Luna with read-only sandboxing. v0.8.1 added a mandatory simplicity counterfactual to plan and code reviews: hold required behavior and safety fixed, then challenge YAGNI, false seams, duplicate ownership, missed in-repo reuse, and unnecessary machinery. Existing canonical Effect implementations count as reuse evidence; Effect and lower LOC are never goals by themselves.

Architecture

Single source of truth: skills/<name>/SKILL.md. Same file for both hosts.
Cross-host routing lives in lib/call-external.sh. Skills NEVER call codex exec, claude -p, grok -p, or pi -p directly. They pipe prompts into the lib script.
Host detection in lib/detect-host.sh (override -> env -> PPID walk).
Anti-recursion: ADVERSARIAL_REVIEW_DEPTH env counter incremented at each cross-agent hop; refuse on ≥ 1.
Main session does the synthesis. No haiku Agent courier subagent. The agent reading the SKILL.md runs lib/call-external.sh, runs its own independent analysis, cross-validates, returns unified output.
Degraded mode when all externals fail: stdout banner ⚠️ DEGRADED MODE, exit 2 from call-external.sh, surfaced at the top of skill output.

Read the full file on GitHub · 116 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.

6d ago First seen · 116 lines · 1,599 tokens per session scan B e70ff493b964

Subscribe to this mod's changes

adversarial-review CLAUDE.md is an instructions file published in the GitHub repository robertoecf/adversarial-review (8 stars, last pushed 23d ago), licensed MIT. It adds 1,599 tokens to every session, about $0.0080 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Code review adversarial-review claude-code claude-code-plugin claude-code-plugins cross-model

river-review GEMINI.md

Gemini CLI instructions for s977043/river-review, covering gemini instructions (river-review), 必須チェック, gemini-specific tips, prompt assembly and development tips.

s977043/river-review · 318 tokens

AI-Prompt-Guide-MCP CLAUDE.md

Claude Code instructions for Blakeem/AI-Prompt-Guide-MCP, a project described as: Deprecated. MCP server and Claude Code plugin with task driven agent orchestration workflows, guides, markdown knowledge graphs, with intelligent context injection.

Blakeem/AI-Prompt-Guide-MCP · 5 tokens

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 6,785 tokens

adversarial-review CLAUDE.md