Chart to editable data
Skill StaryMoon/chart-to-editable-data-skill/skills/chart-to-editable-data
Recover editable CSV series, an SVG reconstruction, and a pixel-level verification overlay from line-chart screenshots or raster figures. Use when Codex needs to digitize a published chart, rebuild a plot for analysis, compare curves, or recover approximate values with explicit calibration and uncertainty.From its SKILL.md
npx -y skills add StaryMoon/chart-to-editable-data-skill --skill chart-to-editable-dataAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.7 KB, 324 tokens by cl100k_base, as published. Nobody here has run it
Chart to Editable Data
Digitize visible line series without pretending that raster recovery is exact. Always preserve the source image, calibration values, detected pixels, and an overlay that lets the user verify the extraction.
Workflow
- Inspect the image and identify the inner plot rectangle in pixels:
left,top,right,bottom. - Read the numeric axis endpoints. Do not infer logarithmic axes as linear.
- Choose one or more visible line colors. Use
autoonly for clean charts. - Run:
python3 scripts/digitize_chart.py chart.png \
--output output/chart-data \
--plot-box 84,42,920,610 \
--x-range 0,100 \
--y-range 20,40 \
--series-color '#2f6fed,#e45756'
- Inspect
verification-overlay.png. If detections include text, markers from another series, or grid lines, tighten the color threshold or crop the plot box. - Deliver
data.csvtogether withreconstructed.svgandmanifest.json. State that values are digitized estimates.
Quality rules
- Never omit the verification overlay.
- Do not digitize bar charts, heatmaps, or log axes with the line-chart path.
- Keep missing x positions missing; do not invent long interpolated spans.
- Report axis calibration and pixel bounds with the result.
- Prefer an author-provided table when one exists.
Read references/calibration.md for coordinate rules and common failure modes.
What ships with it: 3 files
8.4 KB alongside SKILL.md, 1 of them executable
agents/
- openai.yaml245 B
references/
- calibration.md781 B
scripts/
- digitize_chart.pyruns7.4 KB