AI Blocks › Keep your code secure
AI system attack tester
Point it at an AI app. Get prompt injection, jailbreak and tool abuse risks found, ranked and mapped to known attacks.
AI system attack tester is a free AI skill for the job "Keep your code secure". Give the instructions to an AI assistant that supports the required files and tools, then review its output.
Free · No account · MIT licence · Made by Alireza Rezvani · Level: advanced · Thinking: high
Made by Alireza Rezvani. See the original on GitHub; we copied it as it is, with its licence.
What it does
A ranked list of the ways someone could abuse your AI app, each tied to a known attack type. Test it before real users or attackers do. A scanner script runs prompts against your system, looking for prompt injection patterns and jailbreak openings. It also scores the risk that private training data can be extracted, the risk that bad data was slipped into training, and the damage a tool-using agent could do if tricked into misusing its tools. Findings are mapped to the MITRE ATLAS list of attacks on AI. The skill also suggests guardrail designs. Testing deeper than black box needs written permission. It needs Python.
You give it
- Test prompts, or a description of your AI app
- What the app can access and which tools it can call
- Written permission for deeper testing (when needed)
You get
- Injection and jailbreak findings
- Risk scores for data leaks and poisoning
- Tool misuse risks
- Guardrail fixes ranked by severity
Use it when
- You are about to launch a chatbot or agent that touches customer data.
- Your AI can send emails or run tools and you worry it could be tricked.
- A security review asks how you handle prompt injection.
- You want a repeatable set of attack prompts for each release.
How to use it
- Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
- Ask for the job in your own words. Your AI opens the skill when it fits.
- It works best when your AI can run code.