The Agent Skills Directory

[COMMAND_EXECUTION]: The dependency management script scripts/_install-deps utilizes sudo apt-get install to install system-level audio packages. This requires the user to provide administrative privileges for the skill to function on Linux systems.
[COMMAND_EXECUTION]: Core functionality relies on the execution of multiple external binaries and shell scripts for audio playback, recording, and server management, including piper, aplay, sox, and listen-server.
[EXTERNAL_DOWNLOADS]: The skill automatically fetches external resources from well-known services: voice models are downloaded from Hugging Face via curl, and Python dependencies are installed from PyPI using pip at both bootstrap and runtime.
[PROMPT_INJECTION]: The skill's duplex mode creates an indirect prompt injection surface by processing live audio into agent input.
Ingestion points: Transcription of microphone input via scripts/listen.
Boundary markers: Absent from the instructions provided to the agent regarding the handling of transcribed text.
Capability inventory: Subprocess execution (audio tools, installers), file writing, and network operations (downloads).
Sanitization: Not present; the skill only normalizes text for identifying stop phrases.

voice-mode