Stop guessing whether your LLM resists prompt injection - fire the payloads and read the scorecard

There are two honest questions to ask about prompt injection, and they are not the same question. The first is where could it happen? - which lines of code fold untrusted text into a prompt. The secon