The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements functionality to parse Ableton Live Set (.als) files, which are Gzip-compressed XML documents. This introduces a surface for indirect prompt injection where malicious instructions could be embedded within track names, clip metadata, or project comments to influence the agent's behavior during project audits.
Ingestion points: The inspect_als.py script in gemini.md reads and extracts data from user-provided .als files.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard natural language commands found within the project XML.
Capability inventory: The skill provides capabilities for Python scripting, OSC network communication, and modifying application configuration files (Options.txt).
Sanitization: The project parser uses standard XML extraction without filtering or sanitizing strings for instructional content.
[COMMAND_EXECUTION]: The documentation includes instructions for launching the Ableton Live executable via the command line with project path arguments and for monitoring application logs using system commands like tail or Get-Content. While these are standard operational procedures, they represent a direct interface with the host shell.

ableton-live