Bilibili and Video Site Downloader (yt-dlp)

Problem

Engineers working on automation, content archiving, or offline playback often need to retrieve a single video file from Bilibili or another supported site. Passing a web link to a script is not enough, because videos are usually split into audio and video streams, and quality selection, login state, and multi-part logic all affect the final artifact. This skill wraps yt-dlp downloading, format selection, and ffmpeg merging into a clear workflow, reducing the cost of assembling manual parameters.

How It Works

The skill runs downloads through scripts/download_video.py, with these core steps:
- Confirm the save directory: missing directories are created automatically, and the final file is named after the video title.
- Select quality: it defaults to the best available option, or supports 1080, 720, 480, and audio.
- Download and merge: the script installs yt-dlp automatically, selects formats, and uses ffmpeg to merge the result into .mp4.
- Verify the result: after completion, it checks that an MP4 file named after the title exists in the target directory and returns the full path.

b23.tv short links and Bilibili links with ?t= parameters can be passed directly, without stripping the playback position. When VIP access or higher bitrates are needed, a cookies file path can be provided to unlock better quality.

Boundaries

By default, the skill downloads only the current video; downloading an entire multi-part set requires changing the noplaylist setting. Anime, member-only, strongly authenticated, or copyright-restricted content may fail. The environment requires Python 3.10+ and ffmpeg; if ffmpeg is missing, the merge step will fail. If a download cannot be completed, the limitation should be reported honestly instead of claiming the file was saved.