The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of scripts/run.py with arguments derived from user input (style-box-text and lyrics.txt). If the agent executes these via a shell without rigorous escaping or sanitization, it creates a potential command injection vulnerability.
[DYNAMIC_EXECUTION]: The scripts/run.py script utilizes os.execv() to re-execute itself using a Python interpreter from a specific virtual environment. This process replacement technique is a form of dynamic execution that can be abused if local file paths are manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data from the command line and external files (lyrics) to construct prompts for music models or to pass as arguments to the execution script.
[INDIRECT_PROMPT_INJECTION]: Ingestion points: User-provided topic/brief and external lyrics files via the --lyrics flag.
[INDIRECT_PROMPT_INJECTION]: Boundary markers: None identified for security purposes; the skill uses bracketed music tags which do not serve as safety delimiters.
[INDIRECT_PROMPT_INJECTION]: Capability inventory: Uses the Bash tool to execute python3 scripts/run.py which interacts with various remote music model APIs.
[INDIRECT_PROMPT_INJECTION]: Sanitization: The skill instructions do not specify explicit sanitization or validation of user-provided strings before they are passed as command-line arguments to the execution script.

music-prompt