Skip to content
Get packs

AI Red Teaming

Plan and run structured adversarial tests against AI apps to find jailbreaks and data leaks

Source Download skill
Repository
sickn33/agentic-awesome-skills
Stars
49,023
Licence
MIT
Files
1 file

What it does

When installed, the agent gains a structured playbook for adversarially testing AI-powered applications: designing attack suites for jailbreaks, prompt injection, data exfiltration of system prompts, unsafe tool use and token-abuse denial of service, then logging results in a risk register. It also includes a categorized library of example attack prompts and a Python harness template for sending them to a model. The skill strongly gates any active testing behind a written-authorization check, defaulting to read-only defensive guidance otherwise.

Great for

  • Security or AI engineers preparing a pre-launch review of an LLM-powered feature
  • Teams that need documented adversarial testing evidence for a compliance audit
  • Developers responding to a reported prompt injection or jailbreak incident

Skip it if

  • Testing systems you do not own or have written permission to assess, since probing others' AI services can be illegal
  • Production systems directly, because the harness sends live prompts that may trigger destructive tool calls or costs
  • Anyone expecting a turnkey scanner: the script is a template and its pass/fail judgement is not validated

Try asking

  • “Build a jailbreak and prompt-injection test suite for our customer support chatbot before launch”
  • “Help me set up a risk register to track severity and exploitability of findings from our LLM red team”
  • “Review our RAG assistant for system prompt leakage using the data exfiltration test category”
  • “Draft an authorized test plan and confirmation checklist for assessing our coding agent's tool abuse risks”
Step-by-step guide: how to use AI Red TeamingInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    ai-red-teaming.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip ai-red-teaming.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip ai-red-teaming.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Always state the exact target and written scope before asking the agent to run anything; the skill requires this confirmation step.
  • Use a synthetic sandbox with no real secrets, destructive tools or unrelated accounts, as the skill itself recommends.
  • Treat the script's attack_successful output as a hint only and write an independent rubric per test case before using results as a gate.

What's inside

  • SKILL.md

Redistributed unmodified from sickn33/agentic-awesome-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.