Claude Typeset — Design skill for Claude Code
A Claude Code skill: set Markdown as a real page.
How to install Claude Typeset
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open nzeribe/claude-typeset and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Typeset does
A Claude Code skill: set Markdown as a real page. Fifteen typographic rules from Bringhurst's Elements of Typographic Style, plus the tools that make them enforceable — measure the finished PDF, number it, verify it before you send it.
Alternatives in Design
- Planning With Files — Claude Code skill implementing Manus-style persistent markdown planning — the workflow pattern behind the $2B 17.4k ★
- Polish UI — Polish UI elements by auditing SCSS and TSX markup against the Anytype design system, then fixing discrepancie 8.7k ★
- Vox Director — Turn one topic into a finished Vox-style paper-collage explainer/ad video — automated end to end on Atlas Clou 2.1k ★
README
claude-typeset
A [Claude Code](https://docs.anthropic.com/en/docs/claude-code) skill that makes your model set documents like a typographer instead of a word processor.
Point it at a Markdown file and you get a real page: a self-contained HTML sheet and a properly **measured** PDF, plus the three instruments that keep it honest — one that measures the finished page, one that numbers it, one that verifies it before it goes anywhere.
The reason it works is not the stylesheet. It is that the model is carrying a **doctrine** — fifteen rules taken from Bringhurst's *The Elements of Typographic Style* and paid for in real, specific failures. Each rule says what it cost. That is in [DOCTRINE.md](./DOCTRINE.md), and it is the part worth reading even if you never install anything.
Why your PDFs look cheap
Almost certainly not the font. **The measure.**
Bringhurst's first rule of the page is 45–75 characters per line, 66 ideal. Past about 75 the eye loses the line on the return sweep and the page stops being text and becomes a grey slab. A default 30pt print margin on A4 leaves a 189mm column — about **97 characters**. That is what "it looks cheap" usually is, and no amount of font-swapping will fix it.
The second rule is the one that costs people a day: **you cannot compute the measure, only measure it.** The same nominal size of the same named font on the same column gave 68 characters through one renderer and 89 through another — the browser had embedded the face as Type 3 and set it narrower. So the pipeline ends by reading the finished PDF back and counting.
Install
Claude Code skills live in `~/.claude/skills//`:
git clone https://github.com/nzeribe/claude-typeset.git ~/.claude/skills/typeset
pip install pymupdf
bash ~/.claude/skills/typeset/install.sh
You also need a Chromium-family browser (Chrome, Chromium, Brave or Edge) installed — it is the typesetter, for the reason in rule 7. The installer checks for all three and tells you w
Related Skills
Quiz Generator
Claude Code skill for writing exam-style multiple-choice questions. Rules for balanced-length options and dist
Promptimizer
Promptimizer for Claude Fable 5.1: a Claude skill that rewrites a rough ask into a finished Fable prompt using
Mining Brief Agent
MCP-based mining daily brief agent: 3 MCP servers (mining news / NI 43-101 mineral-resource PDF extraction / c
Faostat Analytical Brief
Produce a FAOSTAT-style multi-page analytical brief (PDF + xlsx appendix)
Brochure
Make a brochure or company book (booklet, tri-fold, bi-fold), guided step by step (logo, colours from your log
Lovable Skill Wireframe
The /wireframe skill for Lovable: maps your app's real pages, copy, layout and interactive elements into a pla
Related Agents
Mj Typographer
Typographer with a style-book mind — typeface selection, pairing, hierarchy, spacing systems, and typographic
Style Extractor
Extracts a custom style+formatting profile from example memos and/or written rules. Produces prose-style.md (a
Fullstack QA Debugger
QA and debugging agent for fullstack applications. Use to validate API connections, test error states, measure