Scientific Figure QA

Use this skill when a user asks whether a figure is ready for a manuscript,
conference paper, journal submission, poster, rebuttal, or camera-ready
package.

Checks

• Color accessibility: color vision deficiency, grayscale distinguishability,
  contrast on white backgrounds, and redundant marker/line encodings.
• Palette provenance: confirm canonical, journal-inspired non-official,
  external GPL, or Notion-derived experimental status.
• Typography and layout: final-size font legibility, axis labels with units,
  legend placement, panel labels, consistent spacing, and annotation clarity.
• Statistical visual integrity: uncertainty/error bars, significance markers,
  colorbar labels, centered diverging maps only for data with a real midpoint.
• Export readiness: vector output for line art, 300+ dpi raster output,
  transparent-background risks, and journal/venue file-format constraints.

Palette QA

For palette-specific checks, use the palette CLI rather than inventing color
metadata:

python scripts/palette.py get okabe_ito --format json
python scripts/palette.py recommend --figure-type umap --paper-venue Nature --style-source all --explain

Do not promote Notion-derived candidates to reviewed status during QA unless a
separate manual review explicitly records the review evidence.