Prompt Fuzzer

Prompt Fuzzer, from Prompt Security, tests a system prompt for injection weaknesses. You give it your system prompt, it runs a library of attacks against it (instruction override, prompt leakage, role play, encoding tricks) with a model of your choice, and it scores how well the prompt held. It's interactive and quick, which makes it a good tool for iterating on prompt hardening.

Narrow scope; it's about your prompt, not the model's overall safety.