eagle3-triage

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

• Third-party content exposure detected (high risk: 0.70). The required workflow reads experiment log files from experiments/<exp_id>/ (via find ... | tail -200), and those logs are outsider-authored free text (e.g., produced by other runs/users or external systems) that can contain arbitrary error messages/strings, which are then ingested into the agent’s LLM context for diagnosis.