eachlabs-image-generation

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

• Third-party content exposure detected (high risk: 0.70). This skill accepts and uses external, user-supplied images/URLs as model inputs (e.g., parameters like image_url in wan-v2-6, image_prompt_urls in Bria variants, image_reference_url in Photon, and various "image" fields in Photomaker/other models) which are fetched from public/untrusted sources and read/interpreted by the prediction workflow, exposing the agent to untrusted third-party content.