wdkns-skills
Four skills for Codex: YouTube and Bilibili videos into LaTeX PDF notes, SRT refinement and tensor formula diagrams.
- 825 ★ on the repository
- Last commit
- GPL-3.0
Extended access
Why this level
- The skills download videos, run `ffmpeg`, `whisper` and `xelatex` and write files into the working folder.
- For high-quality Bilibili, `yt-dlp` reads your browser cookies.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/wdkns-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills download videos, run `ffmpeg`, `whisper` and `xelatex` and write files into the working folder.
- For high-quality Bilibili, `yt-dlp` reads your browser cookies.
This is third-party code. Review the repository files before installing.
What it does
wdkns-skills bundles four skills for lectures and technical material. `youtube-render-pdf` turns a YouTube lecture or talk into structured Chinese LaTeX notes with frames, formulas and code snippets and compiles them to PDF. The `bilibili-render-pdf` version is tuned for Bilibili quirks: many videos have no subtitles, some quality levels need a login, and videos split into parts. `subtitle-refine` corrects Chinese SRT subtitles after automatic recognition, keeping timing and meaning, and checks the result with the `check_clean_srt.py` script. `tensor-formula-viz` turns a formula or tensor code into a three-zone diagram: definition, computation in coloured blocks and an explanation of dimensions. The diagrams suit slides, for example for walking through attention and tensor parallelism.
Who it is for. AI students and researchers who learn from video lectures and want detailed PDF notes.
Good fit when
- A YouTube lecture runs two hours and you need PDF notes with frames and formulas.
- A Bilibili video has no subtitles, so speech recognition has to run.
- A course transcript has recognition errors and the timing must stay untouched.
- The attention formula must appear on a slide with the dimensions of every tensor aligned.
- You want a separate agent to check the first draft of the notes for missed details.
Not a fit when
- You need a three-paragraph summary: the skill builds a detailed document with sections.
- The agent cannot read images, so it cannot pick key frames.
- The machine has no TeX Live with Chinese support or ImageMagick and you do not want to install them.
Example request
`$youtube-render-pdf` with a lecture link: prepare detailed notes with frames and formulas and deliver the finished PDF.Limitations
The notes come out in Chinese. You need `yt-dlp`, `ffmpeg`, `xelatex` (TeX Live with CTeX) and `magick`; Bilibili also needs `whisper`. The author requires an agent with good vision, otherwise frame selection does not work. In Codex sub-agents start only when the request asks for them directly. High-quality Bilibili needs browser cookies, for example through `yt-dlp --cookies-from-browser chrome`. The GPL-3.0 license sets conditions on distributing derivative works. The last commit was made in October 2026.
How to disable. Delete the skill folders from `~/.codex/skills`. The finished `.tex` and PDF files stay in working directories and are removed separately.
FAQ
How do I install wdkns-skills in Codex?
Create the directory with `mkdir -p ~/.codex/skills`, then copy the folders you need: `cp -R skills/youtube-render-pdf ~/.codex/skills/` or `cp -R skills/bilibili-render-pdf ~/.codex/skills/`. After that call the skill by name along with a video link.
How does the Bilibili version differ from YouTube?
Subtitles come in three tiers: embedded, Whisper recognition, frame-only mode. High quality needs a cookie login, for multi-part videos the skill asks which parts to take, and calls to like and subscribe are dropped.
How do I speed up a large set of notes?
Ask the agent to launch sub-agents: the author gives a layout with one agent for the outline, five for sections, one for figures and one for terminology consistency.
How do I check that the notes are complete?
After the first version ask a separate reviewer agent to compare the `.tex` with the original subtitles and look for omissions. It only reports findings and changes nothing, so the loop can repeat until the notes are complete.
What diagrams does tensor-formula-viz draw?
Coloured blocks for masks and sparsity, stripes for sharding and tensor parallelism, stacks of slices for four-dimensional tensors and arrows for AllReduce, transpose and concat. The skill traces `matmul`, `einsum`, `reshape` and `permute` in code.
Discussion
No questions or reviews yet.
Related
SkillsTurns a lecture recording into slide-illustrated notes and an HTML page where video, transcript and summary stay in sync.
A CLI that turns PDFs, videos, web pages, audio and repos into study packs with quizzes and flashcards, or SKILL.md
16 ready skills for Cursor, Claude Code and Codex: theses, PDF and Office, draw.io diagrams, project documentation
docx-skill-4-cn-paper: a Word skill for Chinese papers
docx-skill-4-cn-paper
The docx-editor-cn skill builds Word files to Chinese university rules: margins, fonts, native formulas and three-line tables
Eight skills for finished papers and theses in LaTeX and Typst: compile, references, language polish, cover letters and a reviewer-style audit
A set of three research skills: paper writing, academic Word and PowerPoint documents, scientific computing and figures
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| wdkns-skills (this card) | no | 825 | ||
| YouTube Transcript Server | no | not required | 603 | gives an agent a YouTube video transcript by link; wdkns-skills builds a finished LaTeX PDF from the video and its frames |
| bilibili-cli | no | 1,143 | a CLI for videos, subtitles and search on Bilibili; here Bilibili is the source for a PDF study guide |
README in short
The README lists the four skills and describes the video pair in more detail: shared writing rules, the frame selection strategy, the LaTeX template and the Bilibili differences. A table of external dependencies and tips for launching sub-agents follow.