The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Python scripts located at ~/agentic-music-producer-os/scripts/score_cli.py and lyria_cli.py to perform score validation, compilation, and audio generation. These are local operations essential to the skill's primary function.
[INDIRECT_PROMPT_INJECTION]: The skill operates by having the agent write a compact music DSL to a text file (score.txt), which is then processed by command-line tools. This pipeline creates a surface where user-provided input, if not correctly handled by the agent, could attempt to influence the script's execution.
Ingestion points: The agent writes to ~/agentic-music-producer-os/catalog/<slug>/score.txt based on user requests.
Boundary markers: None identified in the prompt template for the DSL content.
Capability inventory: Shell execution of Python scripts via the score_cli.py and lyria_cli.py tools.
Sanitization: Not explicitly defined; the skill relies on the validation script (score_cli.py validate) to check the generated content.
[SAFE]: The skill includes explicit instructions for secure credential management, directing the user/agent to load the GEMINI_API_KEY from a registry like Infisical rather than pasting it directly into the instructions or scripts.

score-composer