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.
npx agentmods add instructions/decoy-run/decoy-redteam/agents-mdgit clone --depth 1 https://github.com/decoy-run/decoy-redteamWrote 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.
[](https://agentmods.dev/instructions/decoy-run/decoy-redteam/agents-md)<a href="https://agentmods.dev/instructions/decoy-run/decoy-redteam/agents-md"><img src="https://agentmods.dev/badge/instructions/decoy-run/decoy-redteam/agents-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01312 | $0.01312 |
| Opus 5 | $0.00656 | $0.00656 |
| Sonnet 5 | $0.00262 | $0.00262 |
| Haiku 4.5 | $0.00131 | $0.00131 |
Grade A, and why
decoy-redteam AGENTS.md 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 4d 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.
How it starts
The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
decoy-redteam — Agent Reference
Autonomous red team tool for MCP servers. Zero dependencies. Node.js >= 18.
What This Tool Does
Connects to MCP servers, sends adversarial payloads to their tools, and reports exploitable vulnerabilities. Three layers of attack sophistication: deterministic (free), AI-adaptive (Guard Pro), autonomous cross-server (Guard Pro).
Architecture
decoy-redteam/
├── index.mjs — public API re-exports
├── bin/cli.mjs — CLI entry point
├── lib/
│ ├── attacks.mjs — 53 attack descriptors (pure data, no side effects)
│ ├── engine.mjs — discover → probe → plan → execute → build stories
│ ├── transport.mjs — MCP stdio connection (persistent, pooled)
│ ├── coverage.mjs — honest coverage % calculation
│ └── report.mjs — JSON + SARIF output
├── test/
│ ├── fixtures/mock-server.mjs — deliberately vulnerable MCP server
│ ├── attacks.test.mjs
│ ├── engine.test.mjs
│ ├── coverage.test.mjs
│ └── cli.test.mjs
└── package.json — zero dependencies, ES modules
Attack Categories
| Category | Count | What it tests |
|---|---|---|
| input-injection | ~16 | SQL, command, path traversal, SSRF, template injection |
| prompt-injection | ~10 | Direct override, role hijack, indirect via output, encoding, multi-turn |
| credential-exposure | ~8 | .env, cloud creds, SSH keys, shell history, error oracle |
| protocol-attacks | ~7 | Malformed JSON-RPC, capability escalation, replay, method injection |
| schema-boundary | ~7 | Type coercion, null bytes, overflow, extra props, NoSQL operators |
| privilege-escalation | ~5 | Scope escape, undeclared access, dotfiles, argument smuggling |
Total: 54 (run grep -c "^ id:" lib/attacks.mjs to confirm).
Safety Model
- Default: dry-run (shows plan, executes nothing)
--live: requires interactive confirmation, read-only + protocol attacks only--live --full: includes potentially-destructive attacks, extra warning- No
--yesflag. Deliberate friction. DECOY_REDTEAM_CONFIRM=yesenv var for CI only
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.
- 4d ago First seen · 123 lines · 1,312 tokens per session scan A 0607b3842093
decoy-redteam AGENTS.md is an instructions file published in the GitHub repository decoy-run/decoy-redteam (2 stars, last pushed 2mo ago), licensed MIT. It adds 1,312 tokens to every session, about $0.0066 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.
Other instructions, from other repositories
red-run CLAUDE.md
Claude Code instructions for blacklanternsecurity/red-run, covering red-run, engagement workflow, token budget, architecture and orchestrator variants.
aigis CLAUDE.md
Claude Code instructions for killertcell428/aigis, covering aigis — project guidelines, zenn記事管理（zenn cli + github連携）, ディレクトリ構成, 記事の作成 and プレビュー.
tooltrust-scanner CLAUDE.md
Claude Code instructions for AgentSafe-AI/tooltrust-scanner: Use the /browse skill from gstack for all web browsing. Never use mcpclaude-in-chrome tools.
evidence-to-skill CLAUDE.md
Instructions for Sanexxxx777/evidence-to-skill, covering project guidance, boundaries and required verification.
promptsonar AGENTS.md
Instructions for meghal86/promptsonar, covering promptsonar agent instructions, operating principles, repo layout, common commands and change workflow.
benchjack CLAUDE.md
Instructions for benchjack/benchjack, covering claude.md, project overview, tech stack, repository layout and common commands.