develop

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

• Third-party content exposure detected (high risk: 0.80). Phase 8 (Document → Research context) explicitly instructs the agent to "Use hashi-reddit" to find related community discussions, which are public, user-generated posts that the agent would ingest and interpret and thus could enable indirect prompt injection.