web-design-guidelines

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches external guidelines from https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md in Step 2.1 and navigates to user-provided URLs to retrieve DOM/snapshots in Step 3, and the fetched guideline content is described as containing rules and output-format instructions that the agent must read and apply, so untrusted third-party content can materially influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly calls WebFetch at runtime to retrieve and use the guidelines from https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md, and states that the fetched content contains the rules and output-format instructions that drive the review, so this remote file directly controls agent behavior.