evaluate

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow (e.g., Opik error-analysis/trace reading and evaluating RAG) ingests LLM trace contents—including user inputs and retrieved documents—at the point of “read each trace” and “capture complete traces,” where those trace texts can include outsider-authored free text from the user’s data being evaluated.