figpolisher

FigPolisher

A Claude Code plugin that polishes matplotlib plotting scripts into paper-ready scientific figures while preserving scientific semantics.

Installation

# From the Claude Code CLI
/plugin install /path/to/FigPolisher

Or add to your Claude Code plugins manually by pointing to this directory.

Usage

/figpolisher path/to/plot.py
/figpolisher path/to/plot.py paper

FigPolisher will:

1. Audit your script (plot type, semantic mapping, style issues)
2. Apply publication-quality styling (typography, colors, layout, legends)
3. Write a non-destructive polished copy (<name>.figpolished.py)
4. Render a preview and write a quality report

What it does

• Semantic color mapping: Sequential palettes for ordered variables (e.g., redshift), linestyle distinction for truth vs model
• Typography: Proper font hierarchy for paper figures
• Layout: Journal-ready dimensions (3.35in single-col, 6.9in double-col)
• Axes: Clean spines, inward ticks, minor ticks
• Legends: Non-obstructive placement, no frame

Configuration

Works out of the box with built-in defaults. Drop a .figpolisher.yaml in your project root to customize. See references/default-config.yaml for all available options.

Output

File	Location
Polished script	<stem>.figpolished.py
Preview image	.figpolisher/previews/<stem>.png
Report	.figpolisher/reports/<stem>_report.md
Semantic map	.figpolisher/semantics/<stem>_semantics.yaml

Architecture

Uses a subagent pipeline for performance:

• Phase 1 (parallel): Script auditor + Config reader
• Phase 2 (sequential): Unified polish
• Phase 3 (parallel): Quality reviewer + Preview renderer
• Phase 4: Report generation