visual-style

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

• Third-party content exposure detected (high risk: 1.00). This skill's extract workflow (see references/extractors/from-website.md, from-video.md, and from-pdf.md) explicitly fetches and analyzes arbitrary website URLs, videos, and PDFs and then synthesizes a style_prompt_full that is used verbatim by connectors (e.g., HeyGen), so untrusted third‑party content can directly influence prompts and downstream tool actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's "Extract from Website" workflow explicitly fetches a user-provided website URL at runtime (e.g., https://stripe.com) and synthesizes a style_prompt_full from that fetched HTML/CSS which is then injected verbatim into generation prompts, so remote site content can directly control the agent's prompts.