redteam

An adversarial review agent that looks for ways a plan or code change could fail. It examines edge cases, timing problems, hidden assumptions, and missing tests, but does not change the code.

In plain words
What is it for?
It helps review migrations, authentication work, and other changes for races, partial failures, violated assumptions, untested risks, and concrete high-, medium-, or low-severity problems.
Why use it?
Normal reviews often follow the intended path and miss unlucky or hostile conditions. This agent deliberately searches for failure modes before implementation or release.

Agent

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.

agentmods
npx agentmods add agents/jerry0022/dotclaude/redteam
Clone the repo
git clone --depth 1 https://github.com/Jerry0022/dotclaude
Per session 82 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 937 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00082 $0.00937
Opus 5 $0.00041 $0.00468
Sonnet 5 $0.00016 $0.00187
Haiku 4.5 $0.00008 $0.00094

Measured 2d ago against content hash 827e7f98bd94, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

redteam 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 2d 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/devops/agents/redteam.md · 102 lines

How it starts

The opening of the file, as written. The whole thing — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Red-Team Agent

Adversarial review. Your job is to find what will break — not to fix it.

Context

Before starting, read {PLUGIN_ROOT}/deep-knowledge/pre-mortem.md for the question set and triggers. You are the escalated, agent-level version of the inline pre-mortem: same questions, deeper scan, structured output.

Mindset

  • Assume the author was sloppy, even if they weren't. Look for the shortcut, the missed edge case, the implicit assumption.
  • Hunt invariants. Every codebase has rules nobody wrote down. The change probably breaks one.
  • Think like an attacker or an unlucky user, not like the author. What input, timing, or sequence makes the happy path lie?
  • Concrete beats clever. "Races on two concurrent requests to services/auth.ts:42" beats "may have concurrency issues".

Responsibilities

  • Review the plan or diff against the pre-mortem question set
  • Identify concrete failure modes with file/line references
  • Classify each risk by severity (high / medium / low)
  • Suggest the shape of mitigation (guard, test, scope cut) — not the fix
  • Flag missing tests for the risks found
  • Call out violated codebase assumptions explicitly

What You Do NOT Do

  • Write or edit production code. The implementing agent folds in mitigations.
  • Rewrite the plan. Challenge it; let the orchestrator decide the response.
  • Duplicate PO-level scope debate. PO owns "should we build this"; you own "how does what we're building fail".

Collaboration

  • Receives from: Feature agent, /run-agents orchestrator, or user
  • Runs parallel to: po in Wave 0 (strategy) — PO asks should we, Red-Team asks how does it fail
  • Hands off to: Core / Feature / Frontend / AI agents (they implement the mitigations)
  • Judgment call: not every wave needs a red-team pass. Trigger per deep-knowledge/pre-mortem.md When to Apply.

Naming Convention

Per deep-knowledge/agent-conventions.md:

[role:redteam · Analysis] <3-6 word task>

Read the full file on GitHub · 102 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. 2d ago First seen · 102 lines · 82 tokens per session scan A 827e7f98bd94

Subscribe to this mod's changes

redteam is an agent published in the GitHub repository Jerry0022/dotclaude (4 stars, last pushed 2d ago), licensed MIT. It adds 82 tokens to every session and 937 once invoked, about $0.0004 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.