produce-video

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

• Third-party content exposure detected (high risk: 0.75). The skill ingests outsider-authored free text via the working transcript produced by invoking /transcribe-video on the user-provided recording; that transcript is derived from the recording’s spoken content (i.e., not authored by the operating user) and is then fed into the agent’s LLM context for P2 flagging and P4 overlay-trigger planning.

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

• Potentially malicious external URL detected (high risk: 0.90). The overlay templates include a runtime-loaded remote script (https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js) which the agent-launchable preview/browser will fetch and execute to run animations (i.e., remote code fetched at runtime and relied on by the overlays).