| name | video-toolkit |
| description | Intelligent video processor for downloading media and extracting transcripts from YouTube and 1000+ supported sites. Automatically handles format selection, subtitle extraction, and post-processing. |
| allowed-tools | Read Write Glob Grep Bash(yt-dlp:*) Bash(ffmpeg:*) Bash(which:*) Bash(python3:*) |
Video Toolkit
This skill provides intelligent capabilities to download video/audio and extract transcripts using yt-dlp.
Usage Guidelines
Goal-Oriented Workflow: DO NOT ask for confirmation unless the user's intent is ambiguous.
- If user says "Download": Download the best available quality video.
- If user says "Transcribe" or "Summarize": Extract subtitles (auto-generated if needed) and convert to clean text.
- If user says "Audio only": Extract mp3/m4a.
Smart Defaults
- Language: Match the video's recognized language. Fallback to English (
en).
- Cookies: Use specific browser cookies (e.g.,
--cookies-from-browser chrome) ONLY if the download fails with "Sign in" or "403 Forbidden" errors.
- Format:
mp4 for video, srt -> txt for transcripts.
Workflows
1. Transcript & Text Extraction
Best for extracting content for reading, summarizing, or RAG.
-
Extract Subtitles:
Downloads subtitles (manual or auto-generated) without downloading the video.
yt-dlp --write-sub --write-auto-sub --sub-lang "en,zh-Hans,.*" --convert-subs srt --skip-download -o "%(title)s.%(ext)s" "[URL]"
(Note: Adjust --sub-lang priority based on the video's likely language if known)
-
Convert to Text:
Run the cleaning script to produce a readable .txt file:
python3 scripts/clean_transcript.py "[FILE].srt"
2. Media Download
Best for saving files locally.
3. Advanced & Troubleshooting
- Permission Errors: If you encounter 403 or Sign-in errors, retry with cookies:
yt-dlp --cookies-from-browser chrome ... (or firefox/safari/edge depending on user OS).
- Specific Formats: If user requests specific resolution (e.g., 1080p) or format, consult the yt-dlp Full Manual.
- Unsupported Sites: Check the Supported Sites List if the URL seems obscure.
References