detect-text-watermark

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/detect_text_watermark.py, outsider-authored free text reaches the runtime via the --text argument or by reading an arbitrary user-provided file path (analyse(args.text, ...) / stream = core.read_text_stream(path) then passing text into detector_covert_channel), so the workflow directly ingests attacker-supplied text without selecting a specific trusted item first.