fde-opportunity-map

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

• Third-party content exposure detected (medium risk: 0.30). In eval/run_harness.py, the harness passes the outsider-authored free-text case prompt (case.json derived from held-out-cases.json) into the required runtime runner as input for the enabled/disabled FDE opportunity-map workflow, exposing the agent to user-authored text.