The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is designed to fetch audio assets from legitimate, well-known services such as Freesound, Mixkit, and Pixabay for use in soundtrack production.
The audit.py script uses urllib.request to download remote assets for the purpose of verifying that local files match the original sources (integrity checking).
[COMMAND_EXECUTION]: Multiple scripts (analyze.py, mix.py, master.py, mux.mjs) execute system commands via subprocess or child_process to perform audio processing and metadata extraction.
These calls target established utilities like ffmpeg and ffprobe using list-based arguments, which minimizes command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data, such as manifest files (audio/assets/manifests/*.json) and cue lists (out/cues.json), which influence the agent's automated design decisions.
Ingestion points: Ingests JSON manifests and cue files from the local project directory and remote URL references for asset verification.
Boundary markers: The instructions use detailed technical specifications and placeholders to delimit data, though explicit 'ignore embedded instructions' markers are absent.
Capability inventory: Uses subprocess.run (ffmpeg/ffprobe), urllib.request, and file system write operations.
Sanitization: Relies on standard Python JSON parsing and strictly defined data structures for audio parameters.

real-sound