code-apps-design-guide

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

• Third-party content exposure detected (high risk: 0.85). The skill ingests outsider-authored free text when the user provides a public CSS/website/PDF link: it fetches the external page via fetch_webpage or reads the uploaded document in Step 2, then uses that extracted prose/content to generate design-guide.md and tokens that are passed into /create-code-app (LLM context).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches user-provided external assets at runtime (e.g., a CSS stylesheet or website link such as https://cdn.example.com/brand.css) via fetch_webpage and uses their contents to generate design tokens and the design guide, so remote content will directly influence the agent's outputs.