publication-figure-forge

name

publication-figure-forge

description

Creates and audits publication-ready scientific figures in R. Use for ggplot2, corrplot, journal figure standards, TIFF/EPS/PDF export, COSTAR prompts, scatterplots, boxplots, bar charts, line plots, heatmaps, correlation matrices, correlograms, corrplot significance plots, volcano plots, Kaplan-Meier curves, forest plots, multi-panel figures, clinical figures, omics figures, ecology/environment figures, and reviewer-proof visualization checks.

Publication Figure Forge

Use this skill when a research project needs publication-ready figures or reviewer-proof figure code, especially in R/ggplot2.

Core Principle

A figure should be designed for submission before it is exported. Do not treat font, units, legend position, resolution, and color as afterthoughts.

Separate:

• data meaning;
• statistical summary;
• figure type;
• journal requirements;
• accessibility;
• export format;
• reviewer failure points.

Intake

Identify:

• figure purpose;
• data frame name and glimpse() output if available;
• x, y, grouping, facet, label, time, survival, or effect-size variables;
• units for every axis;
• target journal or field;
• single-column or double-column size;
• color needs: grayscale, color-blind safe, or journal palette;
• output format: TIFF, EPS, PDF, PNG for draft only;
• whether the result needs clinical, omics, ecology, economics, or general science conventions.

Load:

• references/journal-figure-standards.md for submission defaults and review checks.
• references/figure-type-playbook.md for choosing plot type and expected elements.
• references/correlation-plot-playbook.md for corrplot-based correlation matrices, significance labels, clustering, p-value handling, and nine reusable correlation-plot variants.
• templates/costar-figure-prompt.md when the user wants a prompt for another AI assistant.
• templates/correlation-plot-audit.md when the figure is a correlation matrix or correlogram.
• templates/figure-review-checklist.md before finalizing.
• scripts/publication_theme.R when generating R code.
• scripts/corrplot_correlation_template.R when the user has a tab-delimited matrix and wants reproducible corrplot PDFs.

Workflow

1. Confirm the figure claim and variables.
2. Choose the figure type.
3. Specify axis labels with units.
4. Choose an accessible palette and shape/linetype backup.
5. Use theme_publication() from scripts/publication_theme.R.
6. Place legends inside the panel when space allows.
7. Export with save_figure() at 300 DPI or higher.
8. Check the figure against reviewer failure points.
9. If combining panels, tag panels with A/B/C and keep sizes consistent.

For correlation matrices, also:

• confirm whether rows are variables or samples before transposing;
• choose Pearson, Spearman, or Kendall based on the variable scale and distribution;
• compute p-values only when the sample size and assumptions are defensible;
• use multiple-testing correction when many correlations are tested;
• state that correlation is association, not causation.

Output Modes

R Figure Code

Packages:
Data assumptions:
Plot code:
Export code:
Reviewer checks:

COSTAR Prompt

Use templates/costar-figure-prompt.md and fill in:

• Context;
• Objective;
• Style;
• Tone;
• Audience;
• Response format.

Figure Audit

Verdict:
Figure type:
Axis units:
Font:
Palette:
Legend:
Export:
Common reviewer issue:
Fix:

Correlation Plot Plan

Input orientation:
Correlation method:
Missing-value rule:
P-value method:
Multiple-testing correction:
Plot variant:
Ordering / clustering:
Palette:
Significance display:
Output size and format:
Interpretation limit:

Guardrails

• Do not export final submission figures with the IDE export button.
• Do not omit units from axes.
• Do not use red/green as the only contrast.
• Do not use low-resolution PNG for journal submission unless the journal explicitly allows it.
• Do not label exploratory associations as causal.
• Do not add significance stars without naming the statistical test and correction.
• Do not hide raw data in group-comparison plots when sample size allows jitter/points.
• Do not overcrowd volcano or heatmap labels; label rules must be explicit.
• Do not transpose a correlation matrix input without checking whether rows are features or samples.
• Do not show unadjusted significance stars across many correlations without explaining multiplicity.