motion-graphics

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

• Third-party content exposure detected (medium risk: 0.65). Outsider free text can enter the LLM context via the search-driven categories (webpage/news/tweet/asset-fusion): Step 2 uses media-use to fetch/capture outsider-authored page/article/tweet text, and Step 3’s Director Part 2 prompt includes the resolved assets/index.md/capture-derived content for the subagent to design around.

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

• Potentially malicious external URL detected (high risk: 0.90). The bake-basemap.mjs helper is executed at runtime for the maps basemap lane and injects/executes remote JS and data (e.g. https://cdn.jsdelivr.net/npm/maplibre-gl@5.24.0/dist/maplibre-gl.js and it fetches https://cdn.jsdelivr.net/npm/world-atlas@2.0.2/countries-110m.json and external tile URLs like https://server.arcgisonline.com/...), so untrusted remote content is downloaded and executed as a required part of the bake step.