Give your AI agents the skill of visualizing data the way Edward Tufte intended.

Based on Edward Tufte's The Visual Display of Quantitative Information

Every chart your agent produces gets scored against Tufte's principles — lie factor measured, chartjunk stripped, redundant ink removed, labels moved inline, axes replaced with range-frames, monetary values inflation-adjusted, and the result rendered as a clean SVG. Not as a suggestion. As a workflow.

Before	After

Same data, both times. No decorative bars, no arbitrary colors. Cleveland dot plots that let the actual signal — and the uncertainty — speak for itself.

Individual (Claude Code)

1. Open Claude Desktop → switch to the Cowork tab
2. Go to Plugins
3. Add this repo as a plugin source
4. Upload the ZIP file

Organization (admin)

1. Go to Organization Settings → Plugins
2. Select GitHub as the source
3. Paste this repo URL
4. Set market preference: Required / Available / Default

Requires a paid Claude plan (Pro, Max, Team, or Enterprise) with Cowork enabled.

Three skills, one VDQI-sourced reference. The router picks between the two action skills; the principles file (mirrored into both skills) is the source-grounded encoding of Tufte's specific techniques — nine criteria with numeric anchors, ten chart genres with construction recipes, a chartjunk taxonomy, and 13+14 named anti-pattern/exemplar catalogues — all cited to VDQI by page.

Ships with helper scripts and the tufte-css typography bundle (MIT, vendored under skills/render-tufte-chart/assets/tufte-css/):

• assess-graphical-excellence/scripts/deflate.py — inflation adjustment for monetary time series. Requires real CPI values; errors on a missing year rather than guessing.
• render-tufte-chart/scripts/render_line_svg.py — Tufte-style time-series line chart (VDQI C10).
• render-tufte-chart/scripts/small_multiples.py — grid of identical mini-charts sharing one scale (VDQI C5).
• render-tufte-chart/scripts/quartile_plot.py — Tufte's stripped-down box plot (VDQI C1, pp.124–125).
• render-tufte-chart/scripts/range_frame.py — scatterplot with axis lines spanning only the data range (VDQI C2, pp.130–132); pass --marginal-dash for the dot-dash plot (C3, p.133).
• render-tufte-chart/scripts/wrap_html.py — wraps any SVG in a Tufte-styled HTML page using the vendored ET Book typography; copies the stylesheet and fonts next to the output so it opens in any browser with no network.

Start with the orchestrator — it detects your intent and routes you:

/orchestrate-tufte-vdqi

Or invoke an action skill directly:

/assess-graphical-excellence
/render-tufte-chart

assess-graphical-excellence   → nine-criteria scores, lie factor, prioritised fixes (B1–B7)

assess-graphical-excellence   → diagnose and emit remedies
render-tufte-chart            → rebuild honoring those remedies

render-tufte-chart            → produces an SVG that bakes in B1–B7 by construction
assess-graphical-excellence   → (optional) confirm the result scores well

deflate.py (B7)               → convert to real <base-year> dollars first
render-tufte-chart            → plot the real-terms series with a labelled axis

render-tufte-chart            → produces chart.svg
wrap_html.py                  → wraps it in a tufte-css page (ET Book typography)
                                → outputs chart.html + ./tufte-assets/ siblings

small_multiples.py            → grid of identical mini-charts, shared scales
                                ("inevitably comparative, deftly multivariate" — VDQI p.170)

quartile_plot.py              → Tufte's stripped-down box plot (VDQI pp.124–125)
                                erased box; offset IQR; median tick

range_frame.py                → axis lines span only data min..max (VDQI pp.130–132)
range_frame.py --marginal-dash → adds dot-dash marginals (VDQI p.133)

Concepts in this project were inspired by The Visual Display of Quantitative Information by Edward Tufte. No text has been reproduced.