What is agent-red-team?

A method for authorised security testing of your own Cognigy AI Agent by trying targeted attacks against its rules and safeguards.

Use it to define the test scope, inspect the agent’s configuration, create relevant probes, verify suspected failures, score results, and produce a findings report.

It reveals where the agent may ignore its intended boundaries and provides repeatable evidence for reviewing the findings.

agent-red-team — Skill for Claude Code by Cognigy/cognigy-plugin

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 Cognigy/cognigy-plugin --skill agent-red-team

Clone the repo

git clone --depth 1 https://github.com/Cognigy/cognigy-plugin

Made for: Claude Code.

Or install cognigy, the plugin that ships this one along with the rest of its 16 skills, 2 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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/cognigy/cognigy-plugin/agent-red-team/github.svg)](https://agentmods.dev/skills/cognigy/cognigy-plugin/agent-red-team)

Your own site

<a href="https://agentmods.dev/skills/cognigy/cognigy-plugin/agent-red-team"><img src="https://agentmods.dev/badge/skills/cognigy/cognigy-plugin/agent-red-team/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/cognigy/cognigy-plugin/agent-red-team"><img src="https://agentmods.dev/badge/skills/cognigy/cognigy-plugin/agent-red-team.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 64 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 4,917 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. Third-party audits

NVIDIA SkillSpector warn 7 Sept 2026

SkillSpector: 1 finding, up to high

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

high YARA Match · line 3
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Fix: Remove offensive tool references and exploit code. Legitimate agent skills should not contain penetration testing tools, exploit frameworks, or reconnaissance utilities.

How audits are shown

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.

Model	Per session	Once invoked
Fable 5.1	$0.00064	$0.04917
Opus 5.5	$0.00026	$0.01967
Sonnet 5.5	$0.00013	$0.00983
Haiku 4.5	$0.00006	$0.00492

Measured 1mo ago against content hash 9f4846cce7f4, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

agent-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 1mo 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.

plugin/skills/agent-red-team/SKILL.md · 350 lines

How it starts

The opening of the file, as written. The whole thing — 350 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Red-Teaming a Cognigy AI Agent

Adversarial testing of an LLM-backed AI Agent: read its real configuration, derive the policy it is supposed to enforce, attack those specific boundaries, and report what broke with reproducible evidence.

Scope. Only agents in the caller's own Cognigy organisation, reached with the caller's own API key. This is authorised testing of the user's own system.

This skill teaches a method, not a probe list. Probes are generated per target from that target's config. A fixed script of canned jailbreaks produces false confidence — see Probe discipline.

Match effort to stakes. The discipline below — benign controls, reproduce-before-confirm, never blaming the agent for a transport artifact — is cheap and always worth it; skipping it is how false findings happen. Only the reporting apparatus scales. Use quick depth for a casual check and standard/thorough when the result must be trusted by someone who didn't run it, or compared against a later run.

Phase 0 — Scope contract

Establish this before any probe, state it back to the user, and get confirmation. Defaults are the conservative option.

Setting	Options	Default
Target	`aiAgentId`	— (must be told)
Depth	`quick` / `standard` / `thorough`	`standard`
Environment mutation	`none` / `scratch-store`	`none`
Fixes	`report-only` / `propose-and-apply`	`report-only`

Environment mutation gates the retrieval-injection technique only. scratch-store permits creating a throwaway knowledge store and temporarily repointing the agent's knowledge tool at it.
Fixes gates Phase 4 — changes to the target's own state. report-only means the run never calls update_ai_agent, and never delete_resource on a resource that already existed. It does not block the mandatory teardown of resources the run itself created (e.g. deleting a scratch knowledge store under a scratch-store contract): cleanup is guaranteed regardless of the fixes setting — see Mutation protocol.
Depth sets probe budget: quick ≈ 15, standard ≈ 35–45, thorough ≈ 80+. Only thorough includes repeat runs to measure non-determinism.

Read the full file on GitHub · 350 lines

Files

What ships with it

2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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.

1mo ago Changed · +5 lines 9f4846cce7f4
2mo ago First seen · 345 lines · 64 tokens per session scan A 38cebf6e4e44

Subscribe to this mod's changes

agent-red-team is a skill published in the GitHub repository Cognigy/cognigy-plugin (18 stars, last pushed yesterday), licensed MIT. It adds 64 tokens to every session and 4,917 once invoked, about $0.0003 per session on Opus 5.5. 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.