Turning a Safety Policy into Test Cases

"A Prompt Injection Vulnerability occurs when user prompts alter the LLM's behavior or output in unintended ways."

That sentence opens the entry for prompt injection, the first risk in the OWASP Top 10 for LLM Applications. OWASP is a community that publishes security guidance, and this list names ten security risks for LLM features. The sentence is accurate, and it describes a real risk. As written, however, nobody can grade a single output of the job ticket assistant against the words "unintended ways".