nano-banana-2-multi-reference-image

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

• Third-party content exposure detected (medium risk: 0.30). Externally authored free text enters the runtime via the CLI --prompt argument, which is packaged into the API request body (messages content for chat or prompt for image generation) in scripts/imagegen.py and sent to POST /generation/image (and similarly POST /chat/text), without any filtering or item-selection gating.