video-edit

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 executes a public model via torch.hub.load('snakers4/silero-vad') in scripts/jump_cut_vad_singlepass.py (and invokes npx/npm in scripts/insert_3d_transition.py/video_effects), and those third‑party artifacts are executed and produce VAD/rendering outputs that directly determine FFmpeg editing steps, so untrusted remote content can materially influence tool actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill uses torch.hub.load('snakers4/silero-vad') at runtime (which fetches and executes code from https://github.com/snakers4/silero-vad) to perform Silero VAD and is a required dependency, so this remote repository fetch executes external code during runtime.