judicial-first-impression-en

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

• Third-party content exposure detected (medium risk: 0.30). The workflow ingests outsider-authored free text directly as the “submission” in its required runtime steps (“Cold read. Read the whole text once…” and subsequent reads/quotations), so an outsider can submit the content that the LLM evaluates.