create-ugc-style-video

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md Phase 1/2 requires runtime web fetching of outsider-authored brand/product pages (“WebFetch the brand site… top product candidates with hero images”), and that fetched free-form text is then used to assemble prompts that are fed into the agent/LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly WebFetches the user-provided brand site at runtime (e.g., https://beautybyearth.com) and uses scraped product pages/images to assemble prompts and image_refs that directly control model inputs, so the external site content can directly control agent prompts.