The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The install_as_skill.sh script performs unverified package installations of yt-dlp, pysrt, and python-dotenv from PyPI. While these are common libraries, they are downloaded from an untrusted source at install time.
COMMAND_EXECUTION (LOW): The skill makes extensive use of subprocess.run to call ffmpeg and yt-dlp. The risk of command injection is significantly mitigated by the use of argument lists instead of shell strings, and by the implementation of a temporary directory strategy in scripts/burn_subtitles.py to handle potentially malicious file paths.
PROMPT_INJECTION (LOW): Vulnerable to Indirect Prompt Injection (Category 8).
Ingestion points: scripts/download_video.py fetches untrusted subtitle data from YouTube.
Boundary markers: Absent. Prompts in scripts/translate_subtitles.py and scripts/generate_summary.py do not use delimiters or instructions to ignore embedded commands in the subtitles.
Capability inventory: The skill can execute system commands via FFmpeg and yt-dlp and write files to the local system.
Sanitization: No sanitization or filtering is performed on the subtitle text before it is interpolated into LLM prompts.

youtube-clipper