incident-response

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/incident_triage.py, the required runtime workflow ingests a security event JSON from --input (file) or stdin and then LLM-digests/derives triage decisions by flattening and keyword-scanning raw_payload/event fields via _flatten_to_string, classify_incident, check_false_positives, and check_sev_escalation_triggers, so an outsider who can submit those event payloads can inject free text that is processed without selecting a specific trusted item first.