stitch-design-md

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md에 따르면 런타임에 Stitch 프로젝트의 screen 메타데이터 및 htmlCode.downloadUrl에서 HTML 코드를 다운로드/파싱하는 흐름이 있으며, 이 HTML/시각 자산에 포함된 외부 사용자가 제공한 텍스트(디자인 자산)가 LLM 입력으로 직접 읽힐 수 있습니다.

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

• Potentially malicious external URL detected (high risk: 1.00). This skill explicitly instructs the agent to fetch project HTML and screenshots at runtime via fields like htmlCode.downloadUrl and screenshot.downloadUrl (used with web_fetch/read_url_content), which will be injected into the model's context and directly control the agent's analysis and generated prompts.