The Agent Skills Directory

[SAFE]: The skill implements a custom security module, _pathguard.py, which ensures that all file output operations are restricted to the workspace root or system temporary directories. This effectively mitigates path traversal and unauthorized file write risks.
[COMMAND_EXECUTION]: The skill utilizes system utilities ffmpeg and afplay for audio manipulation and playback. These are invoked using safe subprocess.run patterns with list-based arguments, and all file paths passed to these commands are first validated by the path guard.
[EXTERNAL_DOWNLOADS]: Network activity is restricted to the official ElevenLabs API endpoints. External dependencies like the requests library and the ffmpeg binary are standard tools for the skill's purpose and are correctly documented in the setup instructions.
[SAFE]: Voice cloning functionality includes a dedicated path resolver, _resolve_sample_path, which prevents the skill from reading arbitrary files by restricting audio sample access to a specific, configurable samples directory.

elevenlabs