patient-question-content

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

• Third-party content exposure detected (high risk: 0.85). At runtime the workflow calls UnifAPI to fetch public web/search results (e.g., geo/serp, seo/keywords/*, news/search) and to open Reddit comment text via reddit/posts/{id}/comments, which is outsider-authored free-form prose that can be ingested into the agent’s LLM context.