motion-graphics

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The codebase includes high-risk abuse patterns: optional automatic upload of local images to an external API (embedding an API key in requests), silent/automatic skill update instructions, and unsandboxed headless browser runs that load remote CDN/map tile endpoints — all of which could be abused for data exfiltration, credential leakage, or supply‑chain/remote‑code injection.

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

• Third-party content exposure detected (medium risk: 0.30). In phases/source/guide.md the workflow conditionally resolves asset_needs for search-driven categories by performing RWA-style search (news/web/tweet/image) and then freezes the selected content into assets/ for later LLM-driven planning/design.

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

• Potentially malicious external URL detected (high risk: 0.90). The bake helper and templates load and execute remote JavaScript at runtime (e.g. maplibre + topojson in the puppeteer-baked PAGE and GSAP in the sample template), e.g. https://cdn.jsdelivr.net/npm/maplibre-gl@5.24.0/dist/maplibre-gl.js and https://cdn.jsdelivr.net/npm/topojson-client@3.1.0/dist/topojson-client.min.js and https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js are fetched and executed by the headless browser during skill runtime.