plot

name	plot
description	Generate publication-quality plots from data files or DataFrames
tags	["visualization","plotting","figures","data-analysis"]

Plot Skill - Data Visualization Agent

Generate publication-quality plots from CSV/TSV data or pandas DataFrames using natural language descriptions.

Usage

Basic Syntax

/plot <data_file> "<plot_description>"

Examples

Bar chart:

/plot results.csv "bar chart of citations by year"

Scatter plot with trend:

/plot feature_importance.csv "scatter plot of SHAP values vs feature importance with regression line"

Heatmap:

/plot correlation_matrix.csv "heatmap showing correlation between features"

Box plot:

/plot distributions.tsv "box plot comparing ecosystem diversity across biomes"

Line plot:

/plot timeseries.csv "line plot of citation counts over time by category"

Multiple subplots:

/plot metrics.csv "create 2x2 subplot grid: top-left accuracy, top-right precision, bottom-left recall, bottom-right F1 score"

Advanced Options

Specify columns explicitly

/plot data.csv "bar chart" --x ecosystem --y feature_count --hue category

Customize appearance

/plot data.csv "scatter plot" --title "Feature Analysis" --figsize 10,8 --dpi 300

Save in multiple formats

/plot data.csv "line plot" --output figure.png --formats png,pdf,svg

Use existing DataFrame (from context)

# In a Python cell or script
df = pd.read_csv('data.csv')
# Then use /plot with --dataframe flag

Supported Plot Types

Type	Description	Use Case
bar	Bar chart	Categorical comparisons
barh	Horizontal bar chart	Long category names
scatter	Scatter plot	Relationships between variables
line	Line plot	Time series, trends
box	Box plot	Distribution comparisons
violin	Violin plot	Detailed distributions
heatmap	Heat map	Correlation matrices
histogram	Histogram	Single variable distribution
kde	Kernel density	Smooth distributions
pair	Pair plot	Multi-variable relationships
count	Count plot	Frequency counts
strip	Strip plot	Individual data points
swarm	Swarm plot	Categorical scatter

Style Presets

/plot data.csv "scatter plot" --style nature  # Nature journal style
/plot data.csv "bar chart" --style science   # Science journal style
/plot data.csv "line plot" --style minimal   # Minimal clean style
/plot data.csv "heatmap" --style dark        # Dark theme

Output Options

Default: Saves to figures/generated/ with timestamp Custom path: --output my_figure.png Multiple formats: --formats png,pdf,svg DPI: --dpi 300 (default) or --dpi 600 (high-res)

Examples with Real Data

Example 1: Citation Gap Analysis Results

/plot manuscript/citation_gaps/gap_analysis_targeted.json \
  "scatter plot of citation count vs priority score, colored by gap type"

Example 2: Repository Statistics

/plot data_tables/file_statistics.tsv \
  "horizontal bar chart showing top 20 files by size"

Example 3: Feature Importance

/plot feature_importance.csv \
  "bar chart of top 30 features sorted by SHAP value with error bars"

Example 4: Confusion Matrix

/plot confusion_matrix.csv \
  "heatmap with annotations showing prediction accuracy per ecosystem"

Implementation Details

Tools Available

• Python execution for data loading and processing
• matplotlib for plotting
• seaborn for statistical visualizations
• pandas for data manipulation
• Read/Write tools for file I/O

Workflow

1. Parse user's plot description using LLM understanding
2. Load data from specified file or DataFrame
3. Determine appropriate plot type and parameters
4. Generate plot with publication-quality settings
5. Save in requested format(s)
6. Return figure path and preview

Quality Standards

• 300 DPI default (publication quality)
• Vector formats (PDF, SVG) for scalability
• Seaborn styling for professional appearance
• Proper axis labels and legends
• Tight layout to avoid clipping

Error Handling

Invalid data file:

❌ Error: File not found: data.csv
   Available files in current directory: [list]

Missing columns:

❌ Error: Column 'citations' not found in data
   Available columns: [feature, importance, p_value]

Ambiguous plot request:

⚠️  Multiple interpretations possible:
   1. Scatter plot of A vs B
   2. Line plot of A over B
   Which would you like? (1/2)

Tips for Best Results

1. Be specific: "scatter plot of X vs Y" is better than "plot the data"
2. Specify aesthetics: Include "with trend line" or "colored by category"
3. Use clear column names: The skill will attempt to match fuzzy column names
4. Check data first: Use head data.csv to verify column names
5. Iterate: Start simple, then add customization

Related Skills

• /rrwrite-extract-figures-tables - Extract figures from repository
• /rrwrite-draft-section - Use plots in manuscript sections

Technical Notes

Dependencies: matplotlib, seaborn, pandas, numpy Performance: Handles datasets up to ~1M rows efficiently Memory: Loads entire dataset into memory (use sampling for huge files)