Prompt And Review
# Prompt and Review Reference
## Good defaults
- Wide two-column canvas unless the paper needs a single-column figure.
- White or very light background.
- Low-saturation palette.
- Sparse labels.
- Clear reading path.
- One focal region for the paper's contribution.
- Captions own full figure titles.
- Use panel labels only when the figure has real subpanels.
Avoid generic rows of equal boxes. A strong paper figure usually has hierarchy: one main claim, supporting evidence, and a controlled amount of context.
## Visual directions
### Editorial schematic
Use for method or data-construction figures.
- Make the contribution object visually dominant.
- Let source evidence feed into a larger target artifact.
- Keep literal text short.
- Rebuild all labels later.
### Diagnostic dashboard
Use for error analysis, robustness, or ablation figures.
- Treat charts as evidence, not decoration.
- Use shared legends and consistent ordering.
- Place analytical contrast near the data.
- Rebuild all chart geometry from source data.
### Split-panel training analysis
Use for learning curves and checkpoint comparisons.
- Pair the panels around one question.
- Use a shared checkpoint marker when relevant.
- Use line style and markers for grayscale readability.
## Negative constraints
Start with these and add task-specific ones:when to use it
Community prompt sourced from the open-source GitHub repo Evan-Joseph/agent-workflow-skills (Apache-2.0). A "Prompt And Review" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.
tags
roleplaycommunitygeneral
source
Evan-Joseph/agent-workflow-skills · Apache-2.0