agentleFS
jpoindexterjpoindexter/tufte-skills/skills/tufte/SKILL.md

tufte

Skill1 starsChanged 4 months agoMITView on GitHub

Use when invoked by name (/tufte in Claude Code, $tufte in Codex, or 'tufte' in any agent), or when creating, reviewing, or auditing any data display — chart, graph, table, dashboard, map, timeline, sparkline, diagram, annotated image, slide deck, meeting doc, or evidence presentation — and it's unclear which tufte-* skill applies. Router for the 38-skill Edward Tufte pack.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every sibling skill is installed beside this one, in the same skills directory: ../<slug>/SKILL.md relative to this file (e.g.

    line 13
  2. Very specific:

    Single-file fallback: if the siblings are not installed, read CANON.md in this folder — the complete pack condensed into one document.

    line 17
  3. Distinctive:

    Treat any arguments after the invocation as the task.

    line 21

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Tufte Router

Front door to the 38-skill Edward Tufte pack. Route the task to the smallest set of sibling skills, load them, apply them to the work at hand. Do not return a catalog or a description of the pack — use it. (Tufte's own trilogy framing, VE p.10: VDQI = pictures of numbers, EI = pictures of nouns, VE = pictures of verbs.)

Loading sibling skills — works in any agent

Every sibling skill is installed beside this one, in the same skills directory: ../<slug>/SKILL.md relative to this file (e.g. ~/.claude/skills/tufte-sparklines/SKILL.md, ~/.codex/skills/tufte-sparklines/SKILL.md).

  • Claude Code: invoke the sibling by slug with the Skill tool.
  • Codex CLI, Copilot CLI, Gemini CLI, opencode, or any other agent: read the sibling's SKILL.md with your file-read tool, then follow it. Reading the file IS loading the skill — no special tool required.
  • Single-file fallback: if the siblings are not installed, read CANON.md in this folder — the complete pack condensed into one document. Use it whenever loading multiple siblings is impractical in your runtime.

Route

  1. Treat any arguments after the invocation as the task. No arguments → use the active task.
  2. Select 1–4 skills from the routing table. Load them, apply them, and name which you used in one line.
  3. About to ship a display? Run the Master Audit below and load the matching skill for any point that fails.

Routing table

Task signalLoad
New chart or graphic from scratch; choosing graphic vs table vs sentencetufte-graphical-excellence
Axis scaling, truncated baseline, "is this misleading", pictogram sizingtufte-graphical-integrity
Decluttering, "too busy", gridlines, fills, framestufte-data-ink-ratio + tufte-chartjunk
Visual noise, figure/ground, layer weights, boxed labelstufte-layering-and-separation
Many series or conditions to compare; faceting; trellistufte-small-multiples
"How much data can this hold", sparse-vs-dense callstufte-data-density
Inline trends, KPI rows, word-sized graphics, table cells with historytufte-sparklines
Choosing or reviewing colortufte-color-in-information-design + tufte-redundant-encoding
Geographic or spatial datatufte-data-maps + tufte-micro-macro-readings
Schedules, timetables, movement over timetufte-space-time-graphics
Causal claims, A/B results, "did X cause Y", pre/post chartstufte-causal-reasoning-in-graphics + tufte-evidence-corruption
Slide decks, report vs presentation, bullet hierarchiestufte-cognitive-style + tufte-analytical-design
Diagrams with arrows or connectors; architecture diagramstufte-link-differentiation
Annotating images, screenshots, photos as evidencetufte-mapped-pictures
Before/after, side-by-side, overlay comparisonstufte-parallelism
Multi-image argument, poster, explainer compositiontufte-visual-confections
Labels, legends, table typography, linebreaks in datatufte-typography-for-data
Multivariate data on a flat screen; >2 variablestufte-escaping-flatland + tufte-multifunctioning-elements
Reports/docs where figures, tables, and text are segregated; "see Figure 3"tufte-integration-of-modes
Annotating evidence, commentary layers, explaining a display in placetufte-annotation-as-thinking
UI copy, wayfinding, help text, labels at the point of actiontufte-instructions-at-point-of-need
Step-by-step narrative, instructional sequence, motion in stillstufte-narrative-and-sequence
Evidence quality, cherry-picking, model-searching, integrity of analysistufte-evidence-corruption + tufte-visual-thinking
Aesthetic polish, proportion, line weight, friendlinesstufte-aesthetics-and-technique
Sentence vs table vs graphic; dense reference tables, supertablestufte-tables-and-text-tables
An org keeps shipping bad graphics; who makes them; review gatestufte-graphical-competence
Suspected deceptive display; instructions showing hidden mechanicstufte-disinformation-design
Chrome, frames, mounts, packaging accreting around finished worktufte-presentation-of-artifacts
Backlogs, nav menus, taxonomies, tag systems — what order and whytufte-lists
Design reviews, decision meetings, pre-reads, study-hall protocoltufte-meetings-and-documents
Auditing an analysis or dataset for errors, bias, provenance, fraudtufte-data-forensics
Reference indexes, back matter, asset registries, changelogstufte-visual-index
Dashboard, KPI card grid, metric tiles, status boardtufte-sparklines + tufte-tables-and-text-tables + tufte-data-density
Small-screen or mobile display; chart on a phone; how small can this gotufte-data-density + tufte-sparklines
A specific chart type is in question — pie, donut, funnel, gauge, stacked area, dual axis, 3-Dtufte-graphical-integrity + tufte-graphical-excellence
Long table, many rows, big list, long log — must read at a glance and up closetufte-micro-macro-readings + tufte-tables-and-text-tables
Heatmap, density plot, choropleth, session or attention maptufte-micro-macro-readings + tufte-color-in-information-design
Accessibility, color-blind safety, grayscale survival, contrast of encodingstufte-redundant-encoding + tufte-color-in-information-design
PRD, spec, RFC, technical doc, or report carrying numbers and figurestufte-integration-of-modes + tufte-meetings-and-documents
Vendor demo, sales deck, procurement or tool evaluation, benchmark claimtufte-disinformation-design + tufte-data-forensics
Portfolio, case study, app-store screenshots, showcasing finished worktufte-presentation-of-artifacts + tufte-visual-confections
Glossary, definitions, terminology page, taxonomy of termstufte-lists + tufte-tables-and-text-tables
Incident timeline, project timeline, Gantt, sequence of eventstufte-space-time-graphics + tufte-narrative-and-sequence
Aspect ratio, proportion of a plot, slopes look wrong, chart is too tall or too widetufte-aesthetics-and-technique + tufte-sparklines
"Compared with what?" — missing baseline, control, or reference classtufte-causal-reasoning-in-graphics + tufte-parallelism
How strong should a contrast be — smallest effective difference, contrast sizingtufte-layering-and-separation
"Our audience won't understand this"; dumbing a display down for stakeholderstufte-graphical-competence + tufte-cognitive-style
Accepting an inherited format, template, style sheet, or default chart typetufte-visual-thinking
Interaction hides data — tooltips, hovers, drill-downs, collapsed panelstufte-annotation-as-thinking + tufte-micro-macro-readings

Precedence when several rows fire. Pick the narrowest matching row, then add at most one broader row. Specific collisions:

  • Analysis-quality tasks: tufte-data-forensics when auditing a dataset or method; tufte-evidence-corruption when auditing a presentation of results; tufte-causal-reasoning-in-graphics when a causal claim is at stake. Load one, not all three.
  • Decks: tufte-meetings-and-documents when the question is the meeting (format, pre-read, time budget); tufte-cognitive-style when the question is the deck (density, bullets, headlines). tufte-analytical-design only when defending a design against a standard.
  • Annotation: tufte-mapped-pictures when the base is a photograph or scientific image needing scales and grids; tufte-annotation-as-thinking when the base is a chart, diagram, or screen needing commentary.

Master Audit — run before shipping any data display

Each point names the skill to load for depth on a failure.

  1. Format — does this deserve a graphic at all (vs sentence/table)? Right form chosen? → tufte-graphical-excellence
  2. Truth — Lie Factor ≈ 1.0; visual dims ≤ data dims; no truncated baseline or spare dimension → tufte-graphical-integrity
  3. Ink — two-pass erasure done; grid gray or gone; no fills/3-D/shadows → tufte-data-ink-ratio
  4. Chartjunk — no moiré fills, dark grids, or ducks → tufte-chartjunk
  5. Density — data matrix as large as the question warrants; shrinkable by half? → tufte-data-density
  6. Multivariate — variables shown vs variables the phenomenon has → tufte-escaping-flatland
  7. Comparison — "compared with what?" answered on the display; everything compared in one eyespan, like against like, causal variable on the axis → tufte-causal-reasoning-in-graphics (the question), tufte-small-multiples (the eyespan)
  8. Layering — weight ∝ importance; 1+1=3 effects hunted; nothing boxed → tufte-layering-and-separation
  9. Color — role named (label/measure/represent/decorate); no red/green solo → tufte-color-in-information-design
  10. Type — direct labels, no legend; horizontal; tabular figures → tufte-typography-for-data
  11. Documentation — author, sources, scales, units, date on the display itself → tufte-analytical-design
  12. Integrity — data traceable; no cherry-picked baselines, bins, or models; headline = actual finding → tufte-evidence-corruption

Guardrails

  • Load 1–4 siblings, not the whole pack; CANON.md covers the everything-at-once case.
  • Routing must end in application to the current task, not a summary of what the skills say.
  • Chart selection and encoding mechanics (which chart type, perceptual channel ranking) live outside this pack — pair with a data-visualization skill if installed.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.