The Agent Skills Directory

[COMMAND_EXECUTION]: The srt_utils.py script executes various system utilities (such as nvidia-smi, wmic, sysctl, and fc-list) via subprocess.run to gather hardware metadata regarding GPU availability, memory, and installed fonts to optimize the video processing pipeline.
[DATA_EXFILTRATION]: The skill instructs the agent to use yt-dlp --cookies-from-browser chrome, which involves accessing sensitive browser data (cookies) to facilitate authentication with YouTube's servers for age-gated or private content.
[DYNAMIC_EXECUTION]: The srt_utils.py script employs importlib.import_module to dynamically load the unidecode library for title transliteration, which is a form of runtime code loading.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading and translating external SRT files, which introduces a surface for indirect prompt injection.
Ingestion points: The parse_srt function in srt_utils.py reads user-provided or generated subtitle files.
Boundary markers: While SKILL.md provides detailed translation constraints, it lacks explicit safety delimiters or 'ignore' warnings for the subtitle content itself during the AI translation step.
Capability inventory: The skill has the ability to execute shell commands (ffmpeg, yt-dlp, whisper) and perform file system writes.
Sanitization: There is no evidence of content sanitization or filtering applied to the subtitle text before it is processed by the language model.

yt2bb