zchary1106/figurecraft/figurecraft-charts

FigureCraft Charts

Own quantitative chart tasks from data inspection through publication export. Never generate a factual chart with an image-generation model.

Workflow

1. Read the source data and identify fields, types, units, ordering, missing

values, uncertainty meaning, audience, and final physical size.

1. Choose the chart by communicative intent:

• discrete magnitude comparison: bar;
• ordered progression: line;
• paired relationship: scatter;
• distribution: histogram or box;
• matrix pattern: heatmap.

1. Build a validated FigureSpec using a chart.* kind.
2. Locate the sibling figurecraft skill and run its

scripts/figure.py validate, then render.

1. Inspect values, axes, units, legend/direct labels, clipping, and output size.
2. Deliver SVG plus source and manifest; add PNG/PDF when requested.

Publication rules

• Bars start at zero unless a disclosed scientific exception is necessary.
• Uncertainty requires kind, level, method, and explicit fields.
• Use a muted baseline and stronger primary method when comparison is intended.
• Use direct line labels for small series sets.
• Use color plus marker, line style, position, or text.
• Avoid rainbow maps, 3D bars, unreported smoothing, and unnecessary dual axes.
• Design at the requested final width; do not fix crowding only by shrinking text.
• Select the core output medium and declare provenance. Synthetic demonstration

values must be visibly distinguished from measured results.

Quality gate

Data mismatch, misleading axes, missing units, undefined uncertainty, clipped labels, or invalid output are blocking failures.