Faster Whisper Gpu
Security checks across malware telemetry and agentic risk
Overview
This appears to be a coherent local audio transcription skill, with the main considerations being expected ML package installation, model downloads, and user-directed output file writes.
Before installing, be comfortable with downloading the listed Python ML packages and pretrained models. Use the tool on audio files you choose, and save outputs to non-sensitive paths to avoid accidental overwrites.
VirusTotal
64/64 vendors flagged this skill as clean.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Installing or first running the skill may download packages and models from external sources, even though transcription itself is described as local.
The skill depends on external Python packages and pretrained model artifacts. This is expected for Faster Whisper, but users should be aware that installation and first use may fetch third-party code or model files.
pip install faster-whisper torch ... Models are automatically downloaded on first use to `~/.cache/huggingface/hub/`.
Install from trusted package indexes, review dependency versions if needed, and understand that models will be cached locally.
A transcript, subtitle, or JSON output could overwrite an existing file if the same output path is chosen.
The transcription output can be written to a user-specified file. This is normal for a transcription tool, but users should choose output paths carefully to avoid overwriting files.
Path(output_file).write_text(text, encoding="utf-8")
Use explicit output filenames in safe locations and avoid pointing the output option at important existing files.