Skip to content
Conduit

skill

data-visualization

无论是设计初学者还是科研从业者，有做数据可视化图表需求的都可以用这个skill.

albieduan-coder/data-visualizationcommunitylistedunreviewed
Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/albieduan-coder/data-visualization itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 22, 2026

viz-planner

A Claude Skill that turns a raw dataset into a confirmed visualization plan, then working, self-tested plotting code — instead of jumping straight from "here's my data" to "here's a bar chart."

一个用于数据可视化的 Claude Skill：拿到数据后先诊断、定方案、和你确认，再动手写绘图代码， 而不是一上来就丢一张默认样式的图表给你。

Why / 为什么做这个

Most bad charts don't come from bad code — they come from picking the wrong chart type, or quietly hiding a gap in the data behind a confident-looking visualization. This skill enforces a short workflow that catches both problems before any plotting code gets written.

大多数难看/误导性的图表问题不在代码，而在图表类型选错了，或者用一张"看起来很自信"的图 掩盖了数据本身的缺口。这个 skill 强制走一遍简短的工作流，在写代码之前先把这两类问题挡住。

Workflow / 工作流

  1. Diagnose 诊断数据 — grain, completeness, freshness, embedded caveats, what's simply not in the data (never fabricate a missing number).
  2. Choose a track 选轨道 — academic/publication (pairs with academic-figure skill for Nature/Science-style rules) vs. expressive/creative (dashboards, decks).
  3. Propose & confirm 出方案并确认 — concrete chart plan, panel count driven by how many distinct stories the data supports (not a fixed 2×2 template), and an explicit choice of where the code runs: hand off a script for you to run locally (cheaper), or render directly in-sandbox (more convenient, costs more tokens on the self-test/read-image-back cycle).
  4. Build 构建 — one self-contained Python + matplotlib script per figure, self-tested by actually rendering and reviewing the image before calling it done.
  5. Verify & explain 验证并讲解 — re-check the final render, then explain what it shows in plain language, naming any data limitations that shape the conclusions.

Structure / 目录结构

viz-planner/
├── SKILL.md                       # the skill itself — read this first
├── references/
│   └── chart-vocabulary.md        # chart types organized by job-to-be-done
└── scripts/
    ├── geo_basemap.py              # dependency-free GeoJSON basemap / LAEA projection
    └── layout_helpers.py           # squarify treemap layout, Bezier alluvial ribbons

Usage / 用法

Drop this folder into your Claude Skills directory (Claude Code, Claude Agent SDK, or Cowork), or install it as a saved skill. It triggers automatically whenever you hand Claude a dataset and ask for a chart, dashboard, or "more visualization angles" — you don't need to name it explicitly.

把整个文件夹放进 Claude 的 skills 目录（Claude Code / Agent SDK / Cowork 均可），或作为 已保存的 skill 加载。之后只要你丢一份数据集给 Claude 并要求可视化，就会自动触发，不需要 显式点名。

License / 许可

MIT许可

↑↓ navigate openesc close