litreview

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow includes ingesting “sources” (titles/abstracts/full-text notes) from outsider-authored literature and then rendering those notes into markdown/JSON outputs that are typically fed into the agent’s LLM context for synthesis; this is an outsider free-text ingestion path (public web/literature content → LLM context via the synthesis artifacts).