stitch

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly fetches and ingests content from the third‑party Stitch service — scripts/stitch.mjs calls screen.getHtml()/screen.getImage() and downloadUrlToFile (which uses fetch) and SKILL.md documents the download-html/download-image and "converting Stitch output into React/Vue" workflow, so HTML/images returned by the external Stitch API (potentially user-generated/untrusted) are read and can influence subsequent tool actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill auto-installs and imports the external package spec "@google/stitch-sdk@0.0.3" at runtime via npm (fetched from the npm registry, e.g. https://registry.npmjs.org), which downloads and executes remote code the skill depends on.