adversarial-review

adversarial-review is a command for Claude Code from haunchen/gemini-plugin-cc. It costs 11 tokens per session (772 once invoked), scanned A, original, MIT.

A command that asks Gemini, through the Antigravity CLI, to challenge the decisions in a design or code change. A devil's advocate review looks for weaknesses and suggests alternatives rather than simply approving the work.

In plain words
What is it for?
It helps review a file, a group of files, or the current Git changes from a critical opposing perspective.
Why use it?
It exposes assumptions and trade-offs that the original author may have missed before implementation.

Command for Claude Code

Written for Claude Code: allowed-tools in frontmatter.

Part of the gemini plugin — 5 commands shipped together

Good fit It helps review a file, a group of files, or the current…

Compare 6 commands from other repositories ↓
Install with agentmods
npx agentmods add commands/haunchen/gemini-plugin-cc/adversarial-review
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.

Clone the repo
git clone --depth 1 https://github.com/haunchen/gemini-plugin-cc

Made for: Claude Code.

Or install gemini, the plugin that ships this one along with the rest of its 5 commands.

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-review

README.md
[![agentmods](https://agentmods.dev/badge/commands/haunchen/gemini-plugin-cc/adversarial-review.svg)](https://agentmods.dev/commands/haunchen/gemini-plugin-cc/adversarial-review)
Your own site
<a href="https://agentmods.dev/commands/haunchen/gemini-plugin-cc/adversarial-review"><img src="https://agentmods.dev/badge/commands/haunchen/gemini-plugin-cc/adversarial-review.svg" alt="Measured on agentmods" height="20"></a>
Per session 11 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 772 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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00011 $0.00772
Opus 5 $0.00005 $0.00386
Sonnet 5 $0.00002 $0.00154
Haiku 4.5 $0.00001 $0.00077

Measured 6d ago against content hash 6709399ccb04, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

adversarial-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 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.

plugins/gemini/commands/adversarial-review.md · 60 lines

How it starts

The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Get a devil's advocate review using Gemini via the Antigravity CLI (agy). Instead of finding bugs, this challenges your design decisions and proposes alternatives.

Step 1: Parse --model parameter

Check if $ARGUMENTS contains --model <value>:

  • If yes: extract the value as MODEL, remove --model <value> from $ARGUMENTS
  • If no: set MODEL = flash

Map the alias to an agy model slug:

  • flashgemini-3.6-flash-high (the default)
  • progemini-3.1-pro-high — an older generation than 3.6 flash; available for explicit opt-in, not recommended
  • Anything else is passed through unchanged (run agy models to list available slugs).

Step 2: Determine input

If $ARGUMENTS (after --model removal) is provided:

  • If it contains glob characters (* or ?), use the Glob tool to expand it, then Read each matched file
  • Otherwise, Read the single file directly
  • Concatenate all file contents as REVIEW_INPUT

If $ARGUMENTS is empty:

  • Run: git diff HEAD 2>/dev/null
  • If the command fails (e.g., no commits yet) or the diff is empty, also try: git diff --cached
  • If still empty, tell the user: "No changes found. Provide a file path or make some changes first."
  • Store the diff output as REVIEW_INPUT

Step 3: Call agy

The system prompt lives in the gemini-adversarial-review agent, installed by /gemini:setup. The agent's tools whitelist keeps the run read-only — there is no separate policy file.

Run the following bash command, passing REVIEW_INPUT via stdin:

output=$(printf "%s" "$REVIEW_INPUT" | agy --agent gemini-adversarial-review --model $MODEL --print-timeout 5m 2>&1)
echo "$output"

Note: We pipe input via stdin instead of -p to handle large diffs and special characters safely.

If the output says no output produced and names a denied permission (read_file), the agent tried to open a file and agy auto-denied it: a headless run cannot show a permission prompt, and the denial throws away the whole review. The install is fine — re-run, and note that whether the agent reaches for a file depends on what it finds in the diff, so this does not reproduce every time.

Read the full file on GitHub · 60 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.

  1. 6d ago First seen · 60 lines · 0 tokens per session scan A 6709399ccb04

Subscribe to this mod's changes

adversarial-review is a command published in the GitHub repository haunchen/gemini-plugin-cc (1 stars, last pushed 23d ago), licensed MIT. It adds 11 tokens to every session and 772 once invoked, about $0.0001 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 2026-08-31.