health

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

• Third-party content exposure detected (high risk: 0.75). The skill’s “Extended mode: Memory Quality Analysis” calls mem0 APIs like get_memories/get_event_status at runtime and then ingests the returned memory text (e.g., previews and content used for duplicate/stale/contradiction/orphan checks) into the agent’s LLM context, where those memory records are outsider-authored user/session content not chosen by the operating user for inclusion in the current prompt.