Compose
Build structured attack instructions for AI red teaming. Select techniques, configure your approach, and generate a meta-prompt: an instruction set you feed to a separate LLM to produce adversarial prompts against your target.
New to these terms? Start with What is AI Red Teaming? or check the Glossary. For production workflows, see Our Tools.
Responsible use. This tool generates meta-prompts for authorized AI red teaming and safety testing. The output is an instruction set, not an attack. Use it to test systems you have permission to evaluate. Read the responsible use guidelines.