The Agent Skills Directory

[COMMAND_EXECUTION]: The Python utility scripts/gd_audio.py utilizes subprocess.run to interface with ffmpeg and ffprobe. These calls are used for legitimate technical tasks such as measuring loudness (EBU R128), extracting metadata, and generating spectrograms. The commands are constructed using list-based arguments without shell invocation, mitigating common command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process AI-generated audio assets from external sources like Scenario or ElevenLabs.
Ingestion points: External MP3/WAV files are read by gd_audio.py via ffprobe for technical auditing.
Boundary markers: The workflow suggests presenting analysis results to the agent, but does not define explicit delimiters for this data in the prompt interpolation.
Capability inventory: The skill can execute local commands (ffmpeg), write to the project filesystem, and modify Godot .import files.
Sanitization: Metadata is parsed through structured JSON output from ffprobe and validated using standard library JSON and regex tools, reducing the risk of malicious payload execution through corrupted media metadata.
[SAFE]: The GDScript components for bus management (audio_buses.gd) and scene auditing (audio_audit.gd) utilize standard Godot APIs (AudioServer, ProjectSettings, ResourceSaver) to perform their functions. The installation process strictly copies local script files to the project's addons directory, following standard engine conventions for extension deployment.

godot-audio