adversarial-review

adversarial-review is a skill for Claude Code from overflowy/herdr-adversarial-review. It costs 92 tokens per session (1,013 once invoked), scanned A, original, MIT.

A process for having a separate reviewer challenge code changes or a development plan from an adversarial perspective. It launches a reviewer, waits for its findings, and checks those findings against the intended outcome.

In plain words
What is it for?
Use it for red-team reviews, second opinions, finding weaknesses in a code diff, or testing whether a proposed plan achieves its goal.
Why use it?
It helps expose overlooked bugs, weak assumptions, and gaps between what the changes were meant to do and what they actually do.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the adversarial-review plugin — 1 skill shipped together

Good fit Use it for red-team reviews, second opinions, finding weaknesses in a code diff, or testing whether a proposed plan achieves its goal.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/overflowy/herdr-adversarial-review/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.

Any agent
npx skills add overflowy/herdr-adversarial-review --skill adversarial-review
Clone the repo
git clone --depth 1 https://github.com/overflowy/herdr-adversarial-review

Made for: Claude Code.

Or install adversarial-review, the plugin that ships this one along with the rest of its 1 skill.

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/skills/overflowy/herdr-adversarial-review/adversarial-review/github.svg)](https://agentmods.dev/skills/overflowy/herdr-adversarial-review/adversarial-review)
Your own site
<a href="https://agentmods.dev/skills/overflowy/herdr-adversarial-review/adversarial-review"><img src="https://agentmods.dev/badge/skills/overflowy/herdr-adversarial-review/adversarial-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.

agentmods 80×15 button for adversarial-review

Your own site · 80×15
<a href="https://agentmods.dev/skills/overflowy/herdr-adversarial-review/adversarial-review"><img src="https://agentmods.dev/badge/skills/overflowy/herdr-adversarial-review/adversarial-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 92 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,013 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 $0.00092 $0.01013
Opus 5 $0.00046 $0.00507
Sonnet 5 $0.00018 $0.00203
Haiku 4.5 $0.00009 $0.00101

Measured 11d ago against content hash 1e93f99d13d7, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, 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 11d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/reviewer.sh), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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-review/SKILL.md · 108 lines

How it starts

The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Review

scripts/reviewer.sh drives the whole reviewer lifecycle - it builds the prompt, splits a herdr pane, launches safecodex, submits, waits through permission prompts, and verifies the output file. Your job is the two things it can't do: state the intent and verify the findings.

<skill-dir> below is this skill's directory (shown when the skill loads) - substitute its absolute path.

The reviewer runs interactively in the split pane, so the user can watch it, approve prompts, or steer it.

1. State the intent, then start

Determine what to review (default: the working diff) and - critically - the intent: what the author is trying to achieve. The reviewer challenges whether the work achieves the intent well, not whether the intent is correct. If you can't infer the intent, ask.

Write it to a file (one or two sentences, always in English), show it to the user, and start:

cat > /tmp/intent.md <<'EOF'
<what the author is trying to achieve>
EOF
<skill-dir>/scripts/reviewer.sh start --intent-file /tmp/intent.md --diff

Target selection:

  • --diff - working diff (git diff HEAD) plus untracked files. The default.
  • --file PATH - review a plan, doc, or specific file. Repeatable.
  • Both - e.g. a diff reviewed against the plan it implements.
  • --timeout MS - overall budget (default 900000).

Never paste diffs or file contents yourself; the flags collect them.

Run start with the Bash tool's run_in_background: true. A review can outlast the foreground Bash timeout cap (which would kill the wait loop mid-review), and backgrounding keeps the session free while the reviewer works - pick up the output when the background task completes. ask is quick enough to run in the foreground.

On success it prints REVIEW_FILE=.... On failure it exits non-zero with a diagnostic and a tail of the reviewer pane - relay that and stop; the pane stays open for inspection.

If the reviewer hits a permission prompt, the script notifies the user, prints a BLOCKED note, and keeps waiting - tell the user to approve it in the pane.

Read the full file on GitHub · 108 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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. 11d ago First seen · 108 lines · 92 tokens per session scan A 1e93f99d13d7

Subscribe to this mod's changes

adversarial-review is a skill published in the GitHub repository overflowy/herdr-adversarial-review (7 stars, last pushed 1mo ago), licensed MIT. It adds 92 tokens to every session and 1,013 once invoked, about $0.0005 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.