Skip to content

fix(prompts): update example JSON in quality-evaluator.md to match new VQ-01 philosophy#7393

Merged
MarkusNeusinger merged 1 commit into
mainfrom
style/quality-evaluator-examples
May 19, 2026
Merged

fix(prompts): update example JSON in quality-evaluator.md to match new VQ-01 philosophy#7393
MarkusNeusinger merged 1 commit into
mainfrom
style/quality-evaluator-examples

Conversation

@MarkusNeusinger
Copy link
Copy Markdown
Owner

Summary

After #7391/#7392 changed the VQ-01 rubric (source-of-values irrelevant; defaults vs AI-tuned score equally), the example JSON output and example weaknesses in `quality-evaluator.md` were still showing the OLD philosophy:

  • `vq01_text_legibility` example note: "Readable but relying on defaults (font sizes not explicitly set)" → "Title slightly oversized for content..."
  • `weaknesses[0]`: "Relying on default font sizes instead of explicit settings" → "Title at fontsize=18pt squeezes against the right edge"

These examples are the most concrete guidance for the reviewer; if they show old reasoning, the reviewer follows old reasoning regardless of what the rubric tables say.

Test plan

  • CI green

🤖 Generated with Claude Code

…w VQ-01 philosophy

The example output and example weaknesses in quality-evaluator.md still
treated "relying on default font sizes instead of explicit settings" as a
VQ-01 deduction reason and a weakness. After #7391/#7392's wording changes
that is no longer correct — defaults vs AI-tuned scores equally; what
matters is the visual result.

Updated both the JSON example (vq01 = 6/8 with a "title squeezes against
the right edge" reason) and the weaknesses list to a proportional-sizing
example, so the example matches the rubric the reviewer is actually
supposed to apply.
Copilot AI review requested due to automatic review settings May 19, 2026 13:01
@MarkusNeusinger MarkusNeusinger enabled auto-merge (squash) May 19, 2026 13:01
@MarkusNeusinger MarkusNeusinger merged commit 02e65ce into main May 19, 2026
8 checks passed
@MarkusNeusinger MarkusNeusinger deleted the style/quality-evaluator-examples branch May 19, 2026 13:01
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Updates the example JSON output in prompts/quality-evaluator.md so the VQ-01 “Text Legibility” example reflects the newer rubric philosophy (judge the visual result, not whether defaults were used).

Changes:

  • Updated the VQ-01 example note to focus on an oversized/squeezed title and a concrete adjustment suggestion.
  • Updated the first example weakness to match the new VQ-01 reasoning (title fit/overflow rather than “defaults vs explicit settings”).

Comment on lines 98 to 101
"visual_quality": {
"total": 23,
"vq01_text_legibility": {"score": 5, "max": 8, "note": "Readable but relying on defaults (font sizes not explicitly set)"},
"vq01_text_legibility": {"score": 6, "max": 8, "note": "Title slightly oversized for content — fontsize=18pt squeezes against the right edge; reduce to ~14pt"},
"vq02_no_overlap": {"score": 6, "max": 6, "note": "No overlap"},
MarkusNeusinger added a commit that referenced this pull request May 19, 2026
…7394)

## Summary
Late-arriving Copilot review comments on #7389 / #7391 / #7393 — all
substantive, all applied here.

## Fixes
**High-impact** (would have caused generation failures or wrong sizing):
- \`plot-generator.md\`: Output Files snippet still had \`dpi=300\`
(matplotlib) and \`width=1600 scale=3\` (plotly) — Claude would have
used those in new plots
- \`library/plotnine.md\`: \`element_text\` used in snippet but missing
from imports → \`NameError\` in generated code
- \`library/highcharts.md\`: \"X-axis labels cut off\" pitfall
recommended \`14px\` contradicting the new 12px default

**Consistency**:
- \`default-style-guide.md\`: Native-pixel column mixed pt/px/unitless
under one cell → split per library with explanatory note
- \`library/matplotlib.md\` + \`seaborn.md\`: \`ax.legend(...)\` was
unconditional → wrapped in \`if len(...) > 1\` to avoid the \"No artists
with labels\" warning on single-series plots
- \`library/bokeh.md\`: added commented legend fontsize example
- \`quality-evaluator.md\`: example JSON arithmetic was broken
(sub-scores sum to 24 but total said 23 after #7393's score bump) —
fixed total + top-level score

## Process note
All these comments arrived AFTER the parent PRs auto-merged via
\`--auto\`. Per CLAUDE.md PR Follow-Through, I should poll Copilot
review explicitly before \`--auto\`-merging prompt/cosmetic PRs in the
future. Adding to the process-gaps memo.

🤖 Generated with [Claude Code](https://claude.com/claude-code)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants