video-edit

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

• Third-party content exposure detected (low risk: 0.10). The workflow downloads and transcribes arbitrary YouTube videos whose transcripts can contain adversarial text ingested by the LLM (Claude) during chapter generation.

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

• Potentially malicious external URL detected (high risk: 0.70). The script uses PyTorch Hub to fetch the Silero VAD model via torch.hub.load with 'snakers4/silero-vad', which represents an external third-party operational dependency from an unofficial repository at runtime without cryptographic verification, falling into the high-risk provenance tier.