The Agent Skills Directory

[COMMAND_EXECUTION]: The burn_subtitles_to_video function in scripts/video_editor.py contains a fallback mechanism that utilizes subprocess.run(cmd_str, shell=True). The command string is constructed by directly interpolating the video_file and output_file variables, which are sourced from user-supplied command-line arguments. An attacker can provide a filename containing shell metacharacters (such as ;, &, or |) to break out of the intended command and execute arbitrary system commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of filenames and JSON-formatted subtitle text.
Ingestion points: User-provided arguments for --videos, --video, --audio, and --subtitles in scripts/video_editor.py.
Boundary markers: None are present to distinguish between data and instructions.
Capability inventory: The script has the ability to execute shell commands via subprocess.run to invoke ffmpeg and ffprobe.
Sanitization: There is no validation or escaping of input filenames before they are passed to the shell or interpolation into the ffmpeg command structure.

byted-util-video-editor