message-batches-api

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

• Third-party content exposure detected (high risk: 0.70). The required workflow builds batch requests by embedding the runtime variable document into the LLM messages content ("Classify this document:\n\n{document}"), so if documents contains outsider-authored free text, that text is fed into the agent’s LLM context via the batch API.