aio11y

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

• Third-party content exposure detected (high risk: 0.75). Runtime LLM context is populated with production “generations” (request/response/tool calls) from user sessions, which can include outsider-authored free text (e.g., user-provided messages or web/tool content) that the evaluators/LLM judges ingest when scoring; this is not limited to user-authored content.