The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to clone the official Microsoft VibeVoice repository from GitHub (github.com/microsoft/VibeVoice.git) and download models from Microsoft's Hugging Face collection. These are trusted sources.
[COMMAND_EXECUTION]: The skill provides standard installation and inference commands, including pip install, git clone, and running Python demo scripts from the cloned repository. These are necessary for the skill's functionality and use official, standard paths.
[REMOTE_CODE_EXECUTION]: The skill includes a shell script for downloading experimental voices (bash demo/download_experimental_voices.sh). This script originates from the official Microsoft repository and is intended for legitimate model data retrieval.
[DATA_EXPOSURE]: No sensitive file paths or credentials are used. The use of local WebSocket servers (localhost:7870) for real-time demos is standard practice for local testing of speech-to-text applications.

vibe-voice