Data Visualization Discipline (Judgment Layer)

What it does

This skill does not teach chart implementation — it encodes the judgment calls ("should it be drawn this way?") across five stages, each with a hand-off check:

1. Define intent — every chart needs one sentence: "the reader should conclude ___". If you can't write it, delete the chart. Exactly one north-star metric; process metrics get demoted.
2. Validate the data — mean vs. median by purpose, counting "unknown" in denominators, handling an unfinished final period (don't plot it / separate line / comparable measure), and locking a single as-of date, period definition and data source for the whole deliverable.
3. Choose the form — stacked = part-to-whole only, dual axes banned for mixed units, ≥6 series → small multiples, sankey node/self-flow/color rules, table column pruning, plus how to visualize non-numeric (qualitative) content.
4. Encode — length/position must be strictly linear (no additive base), color encodes category or threshold only, alert red reserved for negatives, same entity = same color/order across all charts, a second channel beyond red–green, and causality rules (same place, same beat, same source) for quantitative animation.
5. Pre-delivery gate (9 items) — all nine must be answered against an actually rendered artifact, and each demands observable evidence: a text-hidden screenshot you can still read relations from, a color count, measured bar-length ratios.

Deeper material lives in references/ (chart selection & statistics, visual form selection for qualitative content, graphical perception science: Cleveland & McGill, Bertin, Tufte, Few).

Who it's for

• Analysts and PMs who repeatedly ship weekly/monthly reports to execs or HQ
• Developers building charts in HTML reports, React/Vue dashboards, matplotlib/plotly/ECharts/D3
• Anyone puzzled why a "clean-looking" chart keeps getting rejected

Examples

• At kickoff: "Build a weekly lead report page" → the skill first extracts audience × form (static conclusion chart vs. interactive tool), the north-star metric, per-chart conclusion sentences, and a comparison list.
• Reviewing existing charts: "What's wrong with these four charts?" → inconsistent entity colors/ordering, invisible zeros in stacked bars, mixed-unit dual axes, chart/table duplication.
• Right before handoff: walk the 9 gates, producing a label-hidden screenshot, measured length ratios, and numerator/denominator (unknown N, handling: excluded/plotted) lines.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/daymade/claude-code-skills/HEAD/data-visualization-discipline/SKILL.md

What I want: (describe your task here)

Install the skill I found on Claude Skill Mart.
Copy the data-visualization-discipline folder from the GitHub repo daymade/claude-code-skills into my ~/.claude/skills/data-visualization-discipline/.
When it's done, tell me in one line what this skill can do.