The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/midictl.py executes external MIDI utilities such as sendmidi, amidi, aconnect, and aseqdump to interact with system MIDI ports.
Execution is performed using subprocess.run and subprocess.call with list-based arguments and without the shell=True flag, which is a secure practice to prevent shell injection.
[EXTERNAL_DOWNLOADS]: The skill's documentation recommends installing sendmidi and receivemidi via Homebrew (brew install gbevin/tools/sendmidi). These are established, legitimate open-source tools within the music technology community.
[DATA_EXPOSURE]: The skill parses and authors Standard MIDI Files (.mid) on the local filesystem. It contains no network-enabled code or patterns suggesting data exfiltration.
[PROMPT_INJECTION]: The skill incorporates $ARGUMENTS into its context to process user commands. This is standard functionality for this type of agent skill and does not bypass existing safety protocols.

media-midi