design-loop

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly downloads and ingests HTML/screenshots from a third-party generation service (Google Stitch) via screen.getHtml()/download URLs (references/stitch-sdk.md and SKILL.md) and also mentions extracting content from existing public sites for the design system, and that downloaded/ingested HTML is post-processed and used to update the site and the next .design/next-prompt.md baton—meaning untrusted third-party content is read and can influence subsequent agent actions.