digipop-coloring-book-creation

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

• Third-party content exposure detected (high risk: 0.80). The skill's Phase 1 "Research" workflow explicitly instructs the agent to search and download images from open/public sites (e.g., Unsplash, Pexels, Pixabay, OpenClipart, Wikimedia Commons, DeviantArt) and to have a visual model "SEE the candidates" (saving thumbnails/full-resolution images and urls), so untrusted, user-generated third-party content is ingested and directly drives selection and processing decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill calls external fal.ai endpoints at runtime (e.g., fal-ai/recraft/vectorize, fal-ai/star-vector, fal-ai/image2svg, fal-ai/image-preprocessors/lineart) to perform vectorization/line-art processing — these are required runtime dependencies that execute remote code, so they present a measurable risk.