paper-craft-skills

English | 中文

Turn academic papers into polished method figures, visual slide decks, and in-depth articles — zero config, one command.

From arxiv link to publication-ready visuals, AIGC slide decks, and deep-dive articles.
Drop a paper, pick a style, get output that looks like a human expert made it.

---

How to install

Copy this into Codex or Claude Code:

Please install zsyggg/paper-craft-skills for me.
GitHub: https://github.com/zsyggg/paper-craft-skills

That's it. The agent handles clone, symlink, and registration. No API keys. No accounts. If you prefer a terminal:

npx skills add zsyggg/paper-craft-skills

Works with: Codex · Claude Code · Cursor · Windsurf

---

What's inside

---

paper-deck — Visual slide decks that don't look like templates

paper-deck turns a paper, article, or technical note into a designed slide deck. It first builds a deck brief and slide-by-slide outline, then writes reproducible visual prompts, generates polished 16:9 slide images, and merges them into .pptx and .pdf.

It is built for iteration: every page has its own prompt, so you can ask for precise changes like “make slide 5 more journal-like”, “replace slide 8 with a real benchmark chart”, or “keep the layout but switch the cover to liquid glass”.

_{Four compact style presets for the same paper topic}

It also supports real source visuals. When a PDF contains strong figures, tables, plots, or screenshots, the skill plans which slides should use them, where they should be cropped, and how they should be framed. Real visuals can be mounted into clean academic panels, evidence blocks, or glass-style layouts instead of being hallucinated from scratch.

/paper-deck https://arxiv.org/abs/1706.03762
/paper-deck /path/to/paper.pdf --style journal-minimal --slides 12
/paper-deck notes.md --style liquid-glass

---

paper-comic — How it works

/paper-comic https://arxiv.org/abs/1706.03762

Reads the paper, then recommends:

  I suggest 6 figures:
  1. Cover: one-line contribution + visual anchor
  2. Transformer architecture overview
  3. Self-attention mechanism
  4. Multi-head attention detail
  5. Encoder / Decoder Block
  6. Key results

  Or generate only 1 overview figure, or expand to 8 detailed mechanism figures.
  Language? [Chinese / English]  Style? [sketchnote / paper-figure]  Scope and count?

Example outputs

paper-figure — clean, publication-grade

sketchnote — bright, warm, approachable

Full walkthrough: examples/paper-illustrated/attention-is-all-you-need

---

paper-analyzer — Deep articles that read like a human expert wrote them

Not a paper translator — a re-interpreter. It reads the full paper, searches GitHub for open-source implementations, cross-references code with the paper, and writes in your chosen style.

Three writing styles

Features

/paper-analyzer https://arxiv.org/abs/1706.03762     # arxiv link
/paper-analyzer /path/to/paper.pdf                     # local PDF
/paper-analyzer                                         # then paste text

---

License

MIT

---

Star History