grill-skill

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In grill-skill.eval.yaml, the runtime path prompt → ask/answer → generate tasks/spec ingests outsider-authored free text only from the user during the interactive interview (the harness feeds it as prompt text), which could contain indirect prompt-injection, so it’s a medium risk.