imagenCN

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md Step 1–4, the user-provided text prompt is refined and then passed directly at runtime into scripts/generate_image.py (args.prompt) and further into the image-generation REST calls (e.g., Ark/Hunyuan/Zhipu/DashScope) without selecting trusted/first-party text sources first.