wjs-dubbing-video
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the Face Landmarker model from Google's official MediaPipe repository (storage.googleapis.com). This is a trusted source required for the visual speaker diarization feature.\n- [COMMAND_EXECUTION]: The script dub.py uses the subprocess module to invoke ffprobe and ffmpeg for audio analysis and video muxing. Arguments are passed as a list rather than a single string, which follows security best practices to prevent shell injection.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided SRT files to generate speech. While the data is not specifically sanitized for prompt injection, the risk is minimal as the content is consumed by a Text-to-Speech (TTS) engine rather than being executed as code or instructions for the agent itself.\n
- Ingestion points: Reads local subtitle files (.srt) in scripts/dub.py and scripts/visual_diarize.py.\n
- Boundary markers: No explicit delimiters or ignore-instructions are used for the processed SRT text.\n
- Capability inventory: Invokes ffmpeg/ffprobe subprocesses and makes network requests to the Volcano TTS API.\n
- Sanitization: Subtitle content is joined and cleaned of specific punctuation but is not filtered for adversarial prompt patterns.
Audit Metadata