web-design-guidelines

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md instructs the runtime workflow to fetch guidelines from a public URL (outsider-authored web content) and then “check against all rules in the fetched guidelines,” which implies the fetched free-text is ingested into the agent/LLM context for rule application.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs fetching runtime guidelines from https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md via WebFetch and uses that fetched content to determine rules and output format, so the external URL directly controls agent behavior at runtime.