Keep enough evidence to repeat the result
For this exercise, keep the baseline and edited plots in separate directories:
| File | Purpose |
|---|---|
runs/baseline/summary.png |
The original plot, including overlapping bars and clipped labels |
runs/with-fix/summary.png |
The plot rendered from your edited source |
runs/with-fix/summary.alt.txt |
The separately written and reviewed alternative text, no more than 150 words |
Use the commands in your language path. Give each attempt a new output path. Keep generated plots under ignored runs/ and source edits in Git.
Record the commands and results
Record the source file, output paths, rendering command, behavior tests, figure checker, and actual outcomes. Add visual observations: separated bars, readable labels, and unchanged counts and order.
Include the alt-text path, contrast, group identification, and grayscale review. Write and review alt text separately; plotting commands create only the PNG.
Render and inspect again after the final source edit. If a check cannot run, record the environment problem and leave its result unverified.
Share only reviewed evidence
Give the agent source, approved before/after images, and test results. This workshop uses invented totals. For a study, review code and images before sharing; keep study data and detailed logs in the approved analysis environment. See lab data guidance.