The Agent Skills Directory

[SAFE]: The skill provides standard media processing functionalities including audio analysis, format conversion, video clipping, and thumbnail generation. All operations are performed using well-known and reputable libraries such as MoviePy, Librosa, Pydub, and Pillow.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted media files (audio/video) and associated metadata (SRT/JSON), it does not use the content of these files to construct or influence the agent's system prompts. The data is processed strictly as binary media or structured layout data, minimizing the risk of indirect prompt injection.
[COMMAND_EXECUTION]: The scripts utilize the argparse module to handle user-supplied file paths and parameters. The interaction with system binaries like FFmpeg is abstracted through the ffmpeg-python and moviepy libraries, which handle command construction safely without exposing raw shell access to untrusted input.

media-toolkit