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 workflow to fetch guidelines from a public URL at runtime (“WebFetch to retrieve the latest rules”), and that fetched free-form text is then used to drive the review—an outsider-authored web content path that can be ingested into the agent’s LLM context.

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

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