design-md

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow fetches outsider-authored free text/HTML from htmlCode.downloadUrl (via web_fetch/read_url_content) and parses it into LLM-readable prose, which can include arbitrary content from the target Stitch project’s screen code.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches screen assets at runtime (e.g., the htmlCode.downloadUrl and screenshot.downloadUrl retrieved via web_fetch / read_url_content after calling [prefix]:get_screen) and injects/parses that remote HTML/image content to drive prompt generation and the DESIGN.md synthesis, so these runtime download URLs directly influence agent outputs.