adversarial-review

adversarial-review is a skill for Claude Code from XRenSiu/claude-code-forge. It costs 115 tokens per session (5,342 once invoked), scanned A, original, MIT.

A team-based review process that checks completed code and requirements from several viewpoints, including specification, code quality, security, and red-team attack testing. A final decision combines the findings.

In plain words
What is it for?
Use it before releasing a major feature or when reviewing security-sensitive or high-risk code. It produces a combined decision such as approve, approve with changes, or request changes.
Why use it?
It reduces the chance that one reviewer misses defects, security weaknesses, or cases where the implementation does not meet the requirements.

Skill for Claude Code

Written for Claude Code: disable-model-invocation in frontmatter.

Part of the forge-teams plugin — 7 skills, 20 agents shipped together

Good fit Use it before releasing a major feature or when reviewing security-sensitive or high-risk code. It produces a combined decision such as approve, approve with changes, or request changes.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/xrensiu/claude-code-forge/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 XRenSiu/claude-code-forge --skill adversarial-review
Clone the repo
git clone --depth 1 https://github.com/XRenSiu/claude-code-forge

Made for: Claude Code.

Or install forge-teams, the plugin that ships this one along with the rest of its 7 skills, 20 agents.

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/xrensiu/claude-code-forge/adversarial-review.svg)](https://agentmods.dev/skills/xrensiu/claude-code-forge/adversarial-review)
Your own site
<a href="https://agentmods.dev/skills/xrensiu/claude-code-forge/adversarial-review"><img src="https://agentmods.dev/badge/skills/xrensiu/claude-code-forge/adversarial-review.svg" alt="Measured on agentmods" height="20"></a>
Per session 115 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 5,342 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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 $0.00115 $0.05342
Opus 5 $0.00057 $0.02671
Sonnet 5 $0.00023 $0.01068
Haiku 4.5 $0.00012 $0.00534

Measured 8d ago against content hash e3ac603c8a58, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

adversarial-review scanned grade A with 1 finding 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 8d 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.

Runs shell commandslowCapability

Expected in a hook, worth knowing in a rule or an instructions file.

grep -rn "exec\|spawn\|execSync" --include="*.ts" src/
plugins/forge-teams/skills/adversarial-review/SKILL.md · 608 lines

How it starts

The opening of the file, as written. The whole thing — 608 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Review + Red Team

用多个审查员的并行审查 + 红队攻击替代单人审查的有限视角。

实测数据:单人代码审查的缺陷检出率约 25-40%。对抗式审查通过 4 角色并行(规格 + 代码 + 安全 + 红队)+ 交叉验证 + 综合裁决,将缺陷检出率提升到 70%+,尤其是安全漏洞的检出率提升显著。

Announce at start: "I'm using the adversarial-review skill to create an agent team with parallel reviewers and a red team attacker."

前置条件:

  1. 实现代码已完成(Phase 4 产出)
  2. PRD 文档可用(用于规格合规审查)
  3. 需要启用 Agent Teams 实验性功能。 在 settings.json 中添加: "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }

When to Use

digraph {
    rankdir=TB;
    start [label="实现完成\n需要审查" shape=oval];
    q1 [label="安全敏感?" shape=diamond];
    q2 [label="高风险\n功能?" shape=diamond];
    q3 [label="需要多维度\n交叉验证?" shape=diamond];
    adversarial [label="adversarial-review\n(本 Skill)" shape=box style=filled fillcolor=lightgreen];
    simple [label="简单代码审查" shape=box];

    start -> q1;
    q1 -> adversarial [label="是"];
    q1 -> q2 [label="否"];
    q2 -> q3 [label="是"];
    q2 -> simple [label="否"];
    q3 -> adversarial [label="是"];
    q3 -> simple [label="否"];
}

The 3-Phase Protocol

┌──────────────────────────────────────────────────────────────────────┐
│                    ADVERSARIAL REVIEW + RED TEAM                     │
├──────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  Phase A: PARALLEL REVIEW      4 个审查员并行审查                    │
│           ↓                    spec + code + security + red team     │
│                                各自独立产出审查报告                   │
│                                                                      │
│  Phase B: CROSS EXAMINATION    审查员交叉检验                        │
│           ↓                    共享发现 → 相互补充 → 挑战分歧        │
│                                                                      │
│  Phase C: SYNTHESIS            review-synthesizer 综合裁决           │
│                                去重 → 排序 → 交叉验证 → 裁决        │
│                                                                      │
│  [FEEDBACK LOOP]:              如果 REQUEST CHANGES                  │
│           ↓                    → Phase 6 (debug/fix)                 │
│           → Re-review (scope-limited)                                │
│                                                                      │
└──────────────────────────────────────────────────────────────────────┘

Read the full file on GitHub · 608 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. 8d ago First seen · 608 lines · 115 tokens per session scan A e3ac603c8a58

Subscribe to this mod's changes

adversarial-review is a skill published in the GitHub repository XRenSiu/claude-code-forge (2 stars, last pushed today), licensed MIT. It adds 115 tokens to every session and 5,342 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.