The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the shell command python3 -m music_analyzer tonality "<audio_file_path>" where the input is a user-controlled variable. This pattern is susceptible to command injection if the agent does not effectively sanitize the audio file path.
[INDIRECT_PROMPT_INJECTION]: The skill processes external audio files, which serves as an ingestion point for untrusted data that could contain malicious instructions in metadata or file contents. * Ingestion points: The <audio_file_path> parameter in SKILL.md. * Boundary markers: The file path is enclosed in double quotes within the shell execution string, but no additional instructions are provided to the model to ignore potential prompt injection markers within the file's data. * Capability inventory: The skill invokes a Python module via the shell, providing a mechanism for command execution as defined in SKILL.md. * Sanitization: Although a validation step for the file path is mentioned, no specific implementation details or safety regex are provided to ensure the input is benign.

music-tonality