aref-vc/tufte-claude-skill

★ 102HTMLAudience · developerComplexity · 1/5Setup · easy

Mindmap

mindmap
  root((tufte-claude-skill))
    What it does
      Generates charts
      Tufte principles
      Claude Code skill
    Ten principles
      Data ink ratio
      Small multiples
      Graphical integrity
    Input
      Natural language
      Make me a chart
    Setup
      Clone to skills dir
      Auto loads

mindmap root((tufte-claude-skill)) What it does Generates charts Tufte principles Claude Code skill Ten principles Data ink ratio Small multiples Graphical integrity Input Natural language Make me a chart Setup Clone to skills dir Auto loads

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Ask Claude Code to make a chart and receive a Tufte-compliant visualization with correct data-ink ratio and no chartjunk

USE CASE 2

Convert raw data tables into publication-quality small-multiple charts for comparing multiple series on a shared scale

USE CASE 3

Validate whether an existing chart would pass Tufte's graphical integrity test, including lie factor and bar-chart zero-baseline checks

Tech stack

HTMLClaude Code

Getting it running

Difficulty · easy Time to first run · 5min

Requires Claude Code installed. Clone repo to ~/.claude/skills/tufte and relaunch Claude Code.

In plain English

This repository is a skill plugin for Claude Code, the AI coding assistant tool from Anthropic. Once installed, it intercepts natural-language requests such as 'make me a chart' and generates charts that follow the data visualization principles of Edward Tufte, an American statistician known for rigorous work on how to present quantitative information clearly and honestly. The skill is distilled from three of Tufte's foundational books. 'The Visual Display of Quantitative Information' covers how to depict numbers with statistical integrity. 'Envisioning Information' covers how to organize information in space. 'Visual Explanations' covers how to show motion, causality, and narrative in visuals. Together these yield ten principles the skill applies when generating charts. The principles include: maximizing the fraction of ink on the chart that represents actual data versus decorative or structural overhead, removing chart elements that do not add information, such as frame boxes, minor gridlines, and redundant axis ticks, avoiding encoding a single quantity in multiple redundant visual formats at the same time, ensuring bar charts start at zero to preserve graphical integrity, and creating small-multiple layouts when a dataset has multiple cross-sections that should be compared on a consistent scale. A well-designed chart should communicate one clear shape from across the room and reveal individual data points when examined closely. Installation takes one command: clone the repository into a specific subfolder inside your Claude Code skills directory. The skill loads automatically the next time Claude Code starts. Typing /tufte or asking Claude to 'make a chart of anything' confirms it is working. No additional configuration is needed. The repository is written primarily as documentation and configuration for the skill. No license is stated in the README.

Copy-paste prompts

Prompt 1

Using the tufte-claude-skill, make me a bar chart showing monthly revenue for the past 12 months. Remove all non-essential chart elements and maximize the data-ink ratio.

Prompt 2

I have a dataset with 5 different customer segments over time. Use the tufte-claude-skill to create small multiples showing each segment on a shared scale.

Prompt 3

Using the tufte-claude-skill, take this CSV data and generate a chart that would pass Tufte's graphical integrity test. Explain which of the ten principles apply here.

Prompt 4

I want to understand data-ink ratio. Using the tufte-claude-skill, generate two versions of the same chart: one with all Tufte principles applied and one without, so I can compare them.

Open on GitHub → Explain another repo

← aref-vc on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.