The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external commands using subprocess.run to manage video processing tasks. Specifically, it calls ffmpeg for merging and transcoding video scenes in merge_scenes.py and stream_render.py. These calls are implemented using list-based arguments rather than shell strings, which follows security best practices to prevent command injection.
[EXTERNAL_DOWNLOADS]: The prepare_env.py script facilitates the installation of necessary Python dependencies (opencv-python, numpy, av, Pillow) from standard package registries. This is a common setup procedure for skills requiring specific computational environments.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves parsing user-provided SRT subtitle files and JSON annotation configurations. This represents a potential attack surface where a malicious input file could attempt to influence the agent's behavior during the planning or review stages. However, the current implementation focuses on structural data processing, and no exploitation patterns were observed.
[SAFE]: No patterns of data exfiltration, credential theft, obfuscation, or persistence were found. The use of a local HTML preview tool (preview.html) relies on the browser's File System Access API, which requires explicit user permission for file operations and is a standard approach for offline web-based editors.

srt-whiteboard-animation