multi-reference-image

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored free text is accepted as the user-provided --prompt and sent directly to the AI Hive generation endpoint via cmd_image() → client.generate_image() (and similarly --system/--prompt in cmd_chat()), so the runtime ingests arbitrary text from whomever runs the skill.