05-red-team

05-red-team is a cursor rule for coding agents from AG-Dmitry/composer-enhanced. It costs 26 tokens per session (566 once invoked), scanned A, original, MIT.

A code-review rule that deliberately tries to break a new implementation before it is accepted.

In plain words
What is it for?
Use it after coding to compare the change with its requirements, test edge cases mentally or in code, and verify suspected failure points.
Why use it?
It exposes requirement gaps, logic mistakes, missing-value failures, asynchronous bugs, and regressions that a quick review may miss.

Cursor rule

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 rules/ag-dmitry/composer-enhanced/05-red-team
Clone the repo
git clone --depth 1 https://github.com/AG-Dmitry/composer-enhanced

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 05-red-team

README.md
[![agentmods](https://agentmods.dev/badge/rules/ag-dmitry/composer-enhanced/05-red-team.svg)](https://agentmods.dev/rules/ag-dmitry/composer-enhanced/05-red-team)
Your own site
<a href="https://agentmods.dev/rules/ag-dmitry/composer-enhanced/05-red-team"><img src="https://agentmods.dev/badge/rules/ag-dmitry/composer-enhanced/05-red-team.svg" alt="Measured on agentmods" height="20"></a>
Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 566 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 $0.00026 $0.00566
Opus 5 $0.00013 $0.00283
Sonnet 5 $0.00005 $0.00113
Haiku 4.5 $0.00003 $0.00057

Measured 5d ago against content hash 93178ab93aaf, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

05-red-team 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.

agent-protocols/05-red-team.mdc · 59 lines

How it starts

The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Phase 5 — Red-team (Self-critique)

Persona: Critic. This is the phase that most distinguishes a thinking model from a guessing one. Switch hats completely: your job is to prove the Implementer wrong, not to confirm it. Assume there is at least one bug and go find it.

Mindset

  • "Looks fine" is not an allowed conclusion. You must produce a concrete list of suspected failure points, then confirm or refute each one by reading the code.
  • If you cannot find anything wrong, you have not looked hard enough — re-read the diff against the success criteria and edge cases one more time.

Attack checklist

Work through templates/red-team-checklist.md. At minimum, hunt for:

  • Requirement drift — does the code actually satisfy each phase-1 success criterion? Re-read them literally.
  • Logic errors — off-by-one, inverted conditions, wrong operator, wrong variable.
  • Null / undefined / empty — every input that could be missing.
  • Async & lifecycle — unawaited promises, races, missing cleanup (subscriptions, timers, listeners), error rejection paths.
  • Error handling — what happens on failure? Is it swallowed? Mislabeled?
  • Regressions — what existing behavior or test could this break? Check call sites found in phase 2.
  • Boundaries & contracts — types, validation at external boundaries, return shapes other code depends on.
  • Hidden incompleteness — stubs, TODOs, hardcoded values presented as done.
  • Project-specific gates — see project-adapter.mdc for your project's security, convention, and framework-specific checks.

Output

Record in the Red-team findings section of the task-memory file:

  • Each suspected issue, with file/line.
  • Verdict: confirmed bug / false alarm.
  • For confirmed bugs: the fix, or a deliberate, justified acceptance.

A confirmed bug sends you to 07-reflect-retry.mdc to fix it (this counts as part of the work, not a retry iteration unless the gates in phase 6 had already run).

Read the full file on GitHub · 59 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 · 59 lines · 26 tokens per session scan A 93178ab93aaf

Subscribe to this mod's changes

05-red-team is a cursor rule published in the GitHub repository AG-Dmitry/composer-enhanced (3 stars, last pushed 2mo ago), licensed MIT. It adds 26 tokens to every session and 566 once invoked, about $0.0001 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.