The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/analyze_recording.py executes ffmpeg and ffprobe using subprocess.run(). These calls use list-based arguments without shell=True, which is a secure practice to prevent command injection.
[EXTERNAL_DOWNLOADS]: The SKILL.md file provides instructions to install standard Python dependencies (numpy, soundfile, scipy, pyloudnorm, pesq, pystoi, librosa) and system utilities (ffmpeg, sox) from official registries and package managers.
[DATA_EXPOSURE]: The skill processes audio recordings from user-specified directories (e.g., Blackbox recording paths). It performs analysis locally and uses temporary directories for intermediate processing, with no evidence of data exfiltration or network operations.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates a standard data processing surface by reading file paths and metadata from a user-provided directory.
Ingestion points: Directory path input via command line; metadata.json and audio files in scripts/analyze_recording.py.
Boundary markers: Not applicable to binary audio data.
Capability inventory: Subprocess execution of media tools; file system read/write in temporary directories.
Sanitization: Uses pathlib for file operations and list-based subprocess calls, effectively sanitizing input against shell injection.

audio-quality-check