stitch-uview-plus-components

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime path explicitly downloads Stitch-authored HTML from htmlCode.downloadUrl (via scripts/fetch-stitch.sh) and then uses that retrieved HTML/screenshot as the source to generate the LLM’s output context, which is outsider free text from a third-party design system.

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

• Potentially malicious external URL detected (high risk: 0.80). This skill explicitly downloads external design HTML at runtime via scripts/fetch-stitch.sh using the provided htmlCode.downloadUrl (e.g., a stitch.withgoogle.com URL) after calling Stitch MCP, and that fetched HTML is used as the direct input to drive the agent's conversion output, so the htmlCode.downloadUrl (hosted on stitch.withgoogle.com) is a runtime external dependency that can directly control prompts.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)