visual-style

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

• Third-party content exposure detected (medium risk: 0.30). In “Extract” mode, the workflow ingests user-supplied website content by receiving a URL and then fetching/parsing the page to generate visual-style.md (including freeform text that can contain indirect prompt-injection payloads).

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

• Potentially malicious external URL detected (high risk: 0.80). The extractor explicitly instructs the agent to "Fetch the page — Use web fetch to get the HTML/CSS" and then builds style_prompt_full from the fetched site (example URL shown: https://stripe.com), so external website content is fetched at runtime and directly controls generated prompts.