adversarial-review

adversarial-review is a skill for Claude Code from monkey1wizard/Golem-Agents-Legion. It costs 0 tokens per session (725 once invoked), scanned A, original, MIT.

A review method for plans, code changes, documents, decisions, and claims. It first states the strongest version of the proposal, then tests it against independently gathered evidence and returns an APPROVE, REVISE, or REJECT verdict.

In plain words
What is it for?
Use it to challenge a proposed plan, diff, document, decision, or claim and record evidence for the resulting verdict.
Why use it?
It makes reviews less dependent on the author's assumptions and prevents unsupported claims from being treated as established facts. It also treats missing or unrun checks as unresolved rather than successful.

Skill for Claude Code

Written for Claude Code: background in frontmatter. Also seen: mentions subagents.

Good fit Use it to challenge a proposed plan, diff, document, decision, or claim and record evidence for the resulting verdict.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/monkey1wizard/golem-agents-legion/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 monkey1wizard/Golem-Agents-Legion --skill adversarial-review
Clone the repo
git clone --depth 1 https://github.com/monkey1wizard/Golem-Agents-Legion

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-review

README.md
[![agentmods](https://agentmods.dev/badge/skills/monkey1wizard/golem-agents-legion/adversarial-review/github.svg)](https://agentmods.dev/skills/monkey1wizard/golem-agents-legion/adversarial-review)
Your own site
<a href="https://agentmods.dev/skills/monkey1wizard/golem-agents-legion/adversarial-review"><img src="https://agentmods.dev/badge/skills/monkey1wizard/golem-agents-legion/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/monkey1wizard/golem-agents-legion/adversarial-review"><img src="https://agentmods.dev/badge/skills/monkey1wizard/golem-agents-legion/adversarial-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 725 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 $0.00000 $0.00725
Opus 5 $0.00000 $0.00362
Sonnet 5 $0.00000 $0.00145
Haiku 4.5 $0.00000 $0.00072

Measured 5d ago against content hash da10c186ef6b, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, 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 5d 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/gal-core/skills/adversarial-review/SKILL.md · 57 lines

How it starts

The opening of the file, as written. The whole thing — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Review

Single-source method for adversarial review across GAL roles. This skill is target-agnostic: apply it to plans, diffs, docs, decisions, claims, and review write-back.

Forked Invocation

Direct invocation (/adversarial-review <target>) runs in an independent subagent with no access to the invoking conversation. That isolation is the point: a review inherited from the authoring context is not an independent judgment.

Review target: $ARGUMENTS

  • Resolve any referenced path against the repository and read it yourself. Nothing about the target is already known to you.
  • Gather evidence with your own tools before forming a verdict. A finding with no repo evidence, no observed behavior, and no named receipt is refuted by rule 2, including your own.
  • Return one verdict block — APPROVE / REVISE / REJECT — with the evidence cited by file path and line.
  • With no target given, return REVISE: no review target provided.

Roles that apply this method from their own agent context (architect, analyst, designer, tester, steward) read the sections below and ignore this one.

Core Method

  1. Steel-man the proposal first. Restate the strongest honest version of the proposal before attacking it. Do not argue against a weaker straw-man.
  2. Refute by default under doubt. If a claim is under-evidenced, treat it as refuted until evidence arrives. Missing evidence is not a soft pass.
  3. Run the minimalism gate by pointer, not restatement. Apply conventions/minimalism.md before accepting any new mechanism, abstraction, file, or dependency. Do not restate the ladder here.
  4. Enforce evidence discipline. Anchor findings in concrete repo evidence, observable behavior, deterministic receipts, or an explicitly named absence of evidence. Do not invent certainty.
  5. Use explicit verdict vocabulary.
    • APPROVE = proceed
    • REVISE = fixable gaps; stop and loop back
    • REJECT = wrong direction; rethink from first principles
  6. Treat REVISE as a jidoka stop-line. Do not continue downstream while a blocking review gap remains open.
  7. NotRun is never pass. Missing execution, missing receipt, timeout, disconnected-partial, or unobserved behavior cannot be upgraded by optimistic prose.

Read the full file on GitHub · 57 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. 5d ago First seen · 57 lines · 0 tokens per session scan A da10c186ef6b

Subscribe to this mod's changes

adversarial-review is a skill published in the GitHub repository monkey1wizard/Golem-Agents-Legion (15 stars, last pushed yesterday), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 725 tokens. 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 2026-09-04.