crisis-detection-intervention-ai

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

• Third-party content exposure detected (medium risk: 0.30). The runtime LLM-ingestion surface is user-supplied crisis text passed into the detector (CLI crisis_detector.ts uses args.join(' ') → detector.detect(text) and then prints/uses it), so outsider-authored free text directly reaches the required workflow without any trusted-source gating.