adversarial-review

adversarial-review is a skill for Claude Code from InjayTseng/graph-engineering-on-research. It costs 37 tokens per session (289 once invoked), scanned A, original, MIT.

A structured review process in which separate reviewers try to disprove the conclusions of an important document.

In plain words
What is it for?
It is for stress-testing reports, plans, analyses, and other written conclusions before they are trusted or acted on.
Why use it?
It helps find weak evidence, hidden assumptions, and errors in documents that have already been revised and seem finished.

Skill for Claude Code

Written for Claude Code: disable-model-invocation in frontmatter. Also seen: mentions subagents.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the graph-engineering plugin — 7 skills shipped together

Good fit It is for stress-testing reports, plans, analyses, and other written conclusions before they are trusted or acted on.

Compare 6 skills from other repositories ↓
Install

Getting it into your agent

This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.

Claude Code
/plugin marketplace add InjayTseng/graph-engineering-on-research
Claude Code
/plugin install graph-engineering

Made for: Claude Code.

Or install graph-engineering, the plugin that ships this one along with the rest of its 7 skills.

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/injaytseng/graph-engineering-on-research/adversarial-review/github.svg)](https://agentmods.dev/skills/injaytseng/graph-engineering-on-research/adversarial-review)
Your own site
<a href="https://agentmods.dev/skills/injaytseng/graph-engineering-on-research/adversarial-review"><img src="https://agentmods.dev/badge/skills/injaytseng/graph-engineering-on-research/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/injaytseng/graph-engineering-on-research/adversarial-review"><img src="https://agentmods.dev/badge/skills/injaytseng/graph-engineering-on-research/adversarial-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 37 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 289 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 $0.00037 $0.00289
Opus 5 $0.00018 $0.00144
Sonnet 5 $0.00007 $0.00058
Haiku 4.5 $0.00004 $0.00029

Measured 10d ago against content hash aaddc231b884, 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 10d 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.

skills/adversarial-review/SKILL.md · 18 lines

What it actually says

Read ${CLAUDE_PLUGIN_ROOT}/prompts/03-adversarial-review.md in full — this template has no single copy block; the whole file is the workflow: Prep, then attacker dispatch, then integration rules, then outputs. The document under review is the user's input below.

Document under review: $ARGUMENTS

Harness rules:

  • If the input above is empty, ask the user which document or set of frozen conclusions to attack — do not pick one yourself.
  • If the input above is a file path, read that file; it is the document under review.
  • If the input above (or the document itself) is in Chinese, use ${CLAUDE_PLUGIN_ROOT}/prompts/03-adversarial-review.zh-TW.md instead and work in Traditional Chinese throughout.
  • You do the Prep and Integration layers yourself in this context — the template marks them as the layers that cannot be outsourced. Each attacker MUST be a separate subagent with fresh context that sees only its own de-identified slice, dispatched in parallel. Never simulate attackers sequentially in this context.
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. 10d ago First seen · 18 lines · 37 tokens per session scan A aaddc231b884

Subscribe to this mod's changes

adversarial-review is a skill published in the GitHub repository InjayTseng/graph-engineering-on-research (11 stars, last pushed 9d ago), licensed MIT. It adds 37 tokens to every session and 289 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 2026-08-31.

Related

Other skills, from other repositories

code-review

A code-review checklist for examining code without rewriting it. It checks correctness, security, readability and performance, then labels findings by seriousness.

radif-ru/ai-multi-agent-system · 0 tokens

pr-checks

Run comprehensive PR checks including reviewing CodeRabbit comments, ensuring PR description quality, running pre-commit hooks, tests, and validation. Use on an existing PR to address review feedback.

massgen/MassGen · 40 tokens

asa-evidence-grounding-auditor

Use when auditing whether a skill decomposition is evidence-grounded, over-inference-safe, and publishable.

3yesore/LetUenforskills · 30 tokens

simplify

Review recent code changes for reuse, quality, and efficiency. Use when the user asks for a code-quality pass, requests "/simplify", or asks to audit recent changes for duplication or technical debt.

qinghaoliqaq/AutoCoder · 43 tokens

continual-learning

Nightly refinement of an existing per-repo review-style prompt using this reviewer's own finding outcomes. Read confirmed (resolved-by-commit / thumbs-up) and dismissed (thumbs-down) findings, promote the bug patterns the team actually fixes, demote the false-positive patterns, reconcile against the current prompt…

langchain-ai/open-swe · 89 tokens

pandic-office

Convert Markdown to PDF (or DOCX/EPUB/HTML) using the pandoc CLI. Use when asked to produce a PDF report, brief, summary, or any document where the input is Markdown and the output should be a polished, paginated file.

Team-Commonly/commonly · 59 tokens