ultraqa

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow (“RUN ADVERSARIAL DYNAMIC E2E SCENARIOS” and “RUN BASELINE VERIFICATION”/“--interactive”) explicitly generates and executes prompt-injection/hostile-input scenarios, meaning outsider-authored free text from the user’s goal/interaction is consumed at runtime to drive the adversarial testing.