plan-change

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/plan_runtime.py, the stateless run reads outsider-provided free text from request_file via request_bytes = request.read_bytes() and then parses/validates it in detect_request_source() (including handoff receipt/body parsing), so the LLM ingests this text at seal time.