The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via user-controlled file paths, which could lead to unauthorized command execution if the AI agent does not sanitize the input.
Ingestion points: The skill accepts an audio file path (<input>) and an optional stems folder path (<stem>) from the user in SKILL.md.
Boundary markers: There are no boundary markers or instructions to quote the input variables (e.g., using "$INPUT") to prevent shell command injection.
Capability inventory: The skill has access to the Bash tool, which it uses to run ffmpeg, helpers/loudness.py, and helpers/stem_analyze.py. These capabilities provide a direct vector for shell-based attacks.
Sanitization: There is no mention of sanitizing, escaping, or validating the file paths before they are interpolated into the Bash command strings.
[COMMAND_EXECUTION]: The skill performs shell command execution using the Bash tool to process audio files based on user-supplied parameters.
Evidence: Instructions in SKILL.md explicitly direct the agent to run helpers/loudness.py <input> and helpers/stem_analyze.py <stem>.
Evidence: The skill uses references/ffmpeg-mastering-chains.md to construct and execute ffmpeg filter chains on user-provided input files.

audio-engineer