The Agent Skills Directory

[COMMAND_EXECUTION]: The test suite (scripts/test_lyria.py) uses os.system and subprocess.run to verify the functionality of the CLI tool. These calls are restricted to checking for the presence of the uv tool and executing internal scripts with --help flags. These are standard development practices with no external input flowing into the commands.
[EXTERNAL_DOWNLOADS]: The skill uses official Google Generative AI libraries (google-genai) to interact with the Lyria models. All network communication is directed to well-known Google API endpoints for the intended purpose of music generation.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied text and images to generate music, the output is binary audio data (MP3/WAV). This significantly limits the attack surface for secondary instruction execution as the generated content cannot trigger subsequent shell commands or logic within the agent environment.

lyria