The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module to execute external binaries for audio processing. Specifically, m21kit/render.py, m21kit/verify.py, and scripts/audio_to_score.py call fluidsynth, ffmpeg, and ffprobe to perform MIDI-to-audio rendering, stem mixing, and metadata extraction. These calls are constructed as argument lists, which is a secure practice to prevent shell injection.
[EXTERNAL_DOWNLOADS]: The m21kit/render.py module contains a helper function to download a General MIDI soundfont from GitHub if it is not present on the system. Furthermore, scripts/fetch_score.py provides a centralized interface for downloading music scores from several online repositories including GitHub, Stanford University's KernScores, and TheSession.org.
[INDIRECT_PROMPT_INJECTION]: The skill processes music data from external sources, which represents a potential attack surface for indirect prompt injection if the ingested data contains malicious instructions.
Ingestion points: scripts/fetch_score.py (fetching remote scores) and scripts/audio_to_score.py (transcribing external audio).
Boundary markers: No explicit boundary markers or 'ignore' instructions are applied to the musical data during parsing.
Capability inventory: The skill can execute shell commands for audio rendering and write files to disk.
Sanitization: The skill relies on the structure-enforcing parsers of the music21 library to process incoming musical data.

programmatic-music