image-gen-cloudflare

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md explicitly instructs downloading arbitrary user-provided URLs for reference images ("If the user gives a URL, download it first: curl -L -o /tmp/ref_image.jpg """), and those externally fetched images are ingested by the img2img workflow (scripts/img2img.py) where their content can directly influence model behavior, creating a clear vector for untrusted third-party content to affect actions.