adversarial-review

adversarial-review is a skill for Claude Code from DerekYRC/agent-thinking-skills. It costs 144 tokens per session (1,868 once invoked), scanned A, original, Apache-2.0.

A security and stability review workflow that examines code from an attacker's point of view. It looks for bugs involving boundaries, security, performance, data consistency, errors, and concurrent operations.

In plain words
What is it for?
Use it to audit selected files or recent branch changes at a quick, standard, or deep review level and receive ranked findings.
Why use it?
It helps reveal risks that ordinary code review may miss, without changing the code.

Skill for Claude Code

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

Part of the first-principles plugin — 2 skills shipped together

Good fit Use it to audit selected files or recent branch changes at a quick, standard, or deep review level and receive ranked findings.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/derekyrc/agent-thinking-skills/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 DerekYRC/agent-thinking-skills --skill adversarial-review
Clone the repo
git clone --depth 1 https://github.com/DerekYRC/agent-thinking-skills

Made for: Claude Code.

Or install first-principles, the plugin that ships this one along with the rest of its 2 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/derekyrc/agent-thinking-skills/adversarial-review/github.svg)](https://agentmods.dev/skills/derekyrc/agent-thinking-skills/adversarial-review)
Your own site
<a href="https://agentmods.dev/skills/derekyrc/agent-thinking-skills/adversarial-review"><img src="https://agentmods.dev/badge/skills/derekyrc/agent-thinking-skills/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/derekyrc/agent-thinking-skills/adversarial-review"><img src="https://agentmods.dev/badge/skills/derekyrc/agent-thinking-skills/adversarial-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 144 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,868 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 $0.00144 $0.01868
Opus 5 $0.00072 $0.00934
Sonnet 5 $0.00029 $0.00374
Haiku 4.5 $0.00014 $0.00187

Measured 10d ago against content hash 8ad57879dc33, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, 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 · 216 lines

How it starts

The opening of the file, as written. The whole thing — 216 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Review

You are an attacker. Your task is to find and report vulnerabilities and risks in code. You do NOT fix them — you report them.

Determine Review Level

Parse user phrasing to select the review level:

User says Level Agent count
"quick review" / "quick adversarial review" Quick 2-3 agents
"adversarial review" / "adversarial audit" (default) Standard 6 agents
"deep adversarial review" / "thorough adversarial review" Deep 12-18 agents

Review Flow

1. Determine Scope

Use the user-specified scope if provided (files, directories, modules). If no scope given, review the current branch's recent changes (git diff main...HEAD or equivalent). Tell the user what scope you chose.

2. Launch Review Agents (in parallel)

Launch agents according to the selected tier. Each agent gets the attack prompt for its assigned dimension(s). All agents run in parallel.

Important: Use the Workflow tool or Agent tool to spawn agents concurrently. Do NOT review sequentially — parallel execution is the core value of adversarial review.

3. Collect and Rank Findings

After all agents return, merge, deduplicate, and rank findings:

  • Critical: causes data loss, security breach, system crash, or unrecoverable failure
  • Medium: causes incorrect behavior, degraded service, or recoverable failure under specific conditions
  • Low: edge case with minimal impact, or code smell that could become a problem later
  • Info: observations worth noting but not actionable now

Every Critical and Medium finding MUST include:

  • Exact file location with line number
  • Concrete reproduction steps
  • Impact description
  • One-line suggested fix direction

4. Output Report

Use the exact report template below. Do not skip sections.

6 Review Dimensions and Agent Prompts

Dimension 1: Boundary Conditions

Agent prompt: "You are a boundary condition tester. Scan the code for: null/undefined handling gaps, oversized inputs, special characters in strings, time anomalies (future dates, past dates, epoch edge cases), negative values where only positives are expected, integer overflow, type confusion between similar types. For each finding, provide: exact location, reproduction steps, and impact. Be specific — 'might crash' is not enough, explain exactly what input and what happens."

Read the full file on GitHub · 216 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. 10d ago First seen · 216 lines · 144 tokens per session scan A 8ad57879dc33

Subscribe to this mod's changes

adversarial-review is a skill published in the GitHub repository DerekYRC/agent-thinking-skills (2 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 144 tokens to every session and 1,868 once invoked, about $0.0007 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.