agentleFS
Roxy-ljyRoxy-ljy/bilibili-video-summary-skill/SKILL.md

bilibili-video-summary

Skill1 starsChanged 6 months agoMITView on GitHub

Create local Markdown summaries for Bilibili, b23.tv, and YouTube videos by extracting metadata and subtitles/transcripts first, then having Codex write the summary and detailed notes. Use when the user asks to summarize, outline, extract key points from, or prepare notes for Bilibili/B-site/BV videos, b23.tv short links, or YouTube videos, especially when they want a local .md document.

  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. kiakun-collab/kiakun-skills★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

The file

Show its rules and text

We split this skill into the rules it gives an agent

19 rules. About 0 are common, 13 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈13

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user gives a Bilibili, b23.tv, or YouTube video URL and asks for a summary, notes, outline, timeline, key points, …

    line 8
  2. Distinctive:

    Run the helper script to extract metadata and subtitles/transcripts to JSON.

    line 12
  3. Distinctive:

    Read the JSON output.

    line 13
  4. Very specific:

    If items[0].has_transcript is true, use items[0].transcript to write the requested summary yourself.

    line 14
  5. Very specific:

    If items[0].has_transcript is false for a Bilibili video, use the browser subtitle fallback before audio transcription:

    line 15
  6. Distinctive:

    Start playback if needed.

    line 17
  7. Very specific:

    If the API, yt-dlp, and browser subtitle fallback all fail, use audio transcription with --transcribe-audio.

    line 23
  8. Distinctive:

    If audio transcription also fails, state that the video has no accessible subtitles and audio transcription failed.

    line 25
  9. Distinctive:

    Prefer Python directly because this skill is used on Windows as well as Unix-like systems:

    line 29
  10. Distinctive:

    When subtitles are unavailable and a real transcript is still needed, run the audio transcription fallback:

    line 47
  11. Very specific:

    Use --asr-model small for better Chinese transcription when speed and model download size are acceptable.

    line 53
  12. Very specific:

    Use this only when the JSON output says has_transcript: false and the URL is a Bilibili page.

    line 57
  13. Distinctive:

    Verify subtitles appear over the video.

    line 67
  14. Distinctive:

    Prefer DOM text extraction if the subtitle text is present in the page.

    line 69
  15. Distinctive:

    If DOM extraction does not expose the subtitle, use browser screenshots at short intervals and OCR only the subtitle region if an OCR tool …

    line 70
  16. Distinctive:

    Stop when the video ends or enough transcript has been collected for the requested summary.

    line 72
  17. Distinctive:

    If the subtitle menu has no Chinese or AI subtitle track, stop and report that no accessible subtitle is available.

    line 75
  18. Distinctive:

    Use this only after subtitle extraction and browser subtitle fallback have failed, or when the user explicitly asks to transcribe audio.

    line 79
  19. Very specific:

    Write in Chinese by default.

    line 125

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Bilibili Video Summary

Use this skill when the user gives a Bilibili, b23.tv, or YouTube video URL and asks for a summary, notes, outline, timeline, key points, or a local Markdown document. By default, create a local Markdown file unless the user explicitly asks for chat-only output.

Workflow

  1. Run the helper script to extract metadata and subtitles/transcripts to JSON.
  2. Read the JSON output.
  3. If items[0].has_transcript is true, use items[0].transcript to write the requested summary yourself.
  4. If items[0].has_transcript is false for a Bilibili video, use the browser subtitle fallback before audio transcription:
    • Open the video page in the browser.
    • Start playback if needed.
    • Open the player subtitle menu, usually labelled Subtitle, CC, or Chinese Zi Mu.
    • Select a Chinese subtitle track when available, especially Chinese AI or Zhongwen AI.
    • Confirm that the player shows a message equivalent to "subtitles switched to Chinese".
    • Extract visible subtitle text from the page while the video plays, deduplicate repeated lines, and use the collected text as the transcript.
    • Mark the transcript source as browser subtitle fallback in the local Markdown file.
  5. If the API, yt-dlp, and browser subtitle fallback all fail, use audio transcription with --transcribe-audio.
  6. Save a local .md file by default. Include source URL, title, channel, duration, transcript source, summary, and detailed notes.
  7. If audio transcription also fails, state that the video has no accessible subtitles and audio transcription failed. Do not pretend to have watched the video.

Commands

Prefer Python directly because this skill is used on Windows as well as Unix-like systems:

python scripts\summarize.py --url "https://www.bilibili.com/video/BV..." --include-transcript --output "$env:TEMP\video_summary.json"

For b23.tv links:

python scripts\summarize.py --url "https://b23.tv/..." --include-transcript --output "$env:TEMP\video_summary.json"

For YouTube:

python scripts\summarize.py --url "https://www.youtube.com/watch?v=..." --include-transcript --output "$env:TEMP\video_summary.json"

When subtitles are unavailable and a real transcript is still needed, run the audio transcription fallback:

python scripts\summarize.py --url "https://www.bilibili.com/video/BV..." --include-transcript --transcribe-audio --asr-model base --output "$env:TEMP\video_summary.json"

Use --asr-model small for better Chinese transcription when speed and model download size are acceptable.

Browser Subtitle Fallback

Use this only when the JSON output says has_transcript: false and the URL is a Bilibili page.

The goal is to recover subtitles that Bilibili exposes only in the player UI. Do not use this path when API subtitles are already available.

Suggested procedure:

  1. Open the Bilibili URL in the browser automation tool.
  2. Click play and wait until the player controls are visible.
  3. Open the subtitle menu. It may appear as Subtitle, CC, or Chinese UI text for subtitles.
  4. Select Chinese or Chinese AI subtitles if available. The UI may show entries similar to Chinese, Chinese AI, Zhongwen, or Zhongwen AI.
  5. Verify subtitles appear over the video.
  6. Collect subtitle text while playing:
    • Prefer DOM text extraction if the subtitle text is present in the page.
    • If DOM extraction does not expose the subtitle, use browser screenshots at short intervals and OCR only the subtitle region if an OCR tool is available.
    • Deduplicate consecutive repeated captions.
    • Stop when the video ends or enough transcript has been collected for the requested summary.
  7. In the Markdown file, include Transcript source: browser subtitle fallback.

If the subtitle menu has no Chinese or AI subtitle track, stop and report that no accessible subtitle is available.

Audio Transcription Fallback

Use this only after subtitle extraction and browser subtitle fallback have failed, or when the user explicitly asks to transcribe audio.

The helper script downloads the best available audio with yt-dlp and transcribes it locally with faster-whisper.

Recommended command:

python scripts\summarize.py --url "<video-url>" --include-transcript --transcribe-audio --asr-model base --output "$env:TEMP\video_summary.json"

Notes:

  • The first ASR run may download the Whisper model and can be slow.
  • base is faster and smaller. small usually gives better Chinese results but is slower.
  • The default language hint is zh. Use --asr-language "" only when the video is multilingual and automatic language detection is preferred.
  • Temporary audio is deleted by default. Use --audio-output-dir "<dir>" only when the audio file should be kept for debugging.
  • In the Markdown file, include Transcript source: audio_asr.

Dependencies

Install once if needed:

python -m pip install -r requirements.txt

The script needs yt-dlp on PATH. The Python package install normally provides the yt-dlp command.

Audio transcription also needs faster-whisper; it is included in requirements.txt.

JSON Output

The output JSON has:

  • items[0].title
  • items[0].url
  • items[0].source
  • items[0].channel
  • items[0].duration
  • items[0].has_transcript
  • items[0].transcript when --include-transcript is used and subtitles are available
  • items[0].metadata
  • items[0].metadata.transcript_source, such as subtitle_api, yt_dlp_subtitle, browser subtitle fallback, audio_asr, or none

Markdown Output Style

Write in Chinese by default. In the generated Markdown, do not use English section titles or field labels. Use Chinese headings and field labels throughout, and convert raw source keys like subtitle_api or audio_asr into human-readable Chinese.

Structure the output with local-language section titles for video info, summary, detailed content, and reusable points.

In the info block, use local-language labels for source, link, BV ID, title, channel, duration, publish date, transcript source, and note.

For technical videos, explicitly extract:

  • Problem statement
  • Inputs and outputs
  • Key method or mechanism
  • Metrics or evaluation criteria
  • Baselines or comparison points
  • Open questions to verify from the transcript

Limits

Bilibili videos without public subtitles can still be summarized through audio transcription, but ASR quality depends on audio clarity, language, background music, and model size. If subtitles and ASR both fail, ask the user for a subtitle file or transcript.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.