csv-data-visualizer

SKILL.md

CSV Data Visualizer

Overview

This skill enables comprehensive data visualization and analysis for CSV files. It provides three main capabilities: (1) creating individual interactive visualizations using Plotly, (2) automatic data profiling with statistical summaries, and (3) generating multi-plot dashboards. The skill is optimized for exploratory data analysis, statistical reporting, and creating presentation-ready visualizations.

When to Use This Skill

Invoke this skill when users request:

"Visualize this CSV data"

"Create a histogram/scatter plot/box plot from this data"

"Show me the distribution of [column]"

"Generate a dashboard for this dataset"

"Profile this CSV file" or "Analyze this data"

"Create a correlation heatmap"

"Show trends over time"

"Compare [variable] across [categories]"

Core Capabilities
1. Individual Visualizations

Create specific chart types for detailed analysis using the visualize_csv.py script.

Available Chart Types:

Statistical Plots:

# Histogram - distribution of numeric data
python3 scripts/visualize_csv.py data.csv --histogram column_name --bins 30

# Box plot - show quartiles and outliers
python3 scripts/visualize_csv.py data.csv --boxplot column_name

# Box plot grouped by category
python3 scripts/visualize_csv.py data.csv --boxplot salary --group-by department

# Violin plot - distribution with probability density
python3 scripts/visualize_csv.py data.csv --violin column_name --group-by category

Relationship Analysis:

# Scatter plot with automatic trend line
python3 scripts/visualize_csv.py data.csv --scatter height weight

# Scatter plot with color and size encoding
python3 scripts/visualize_csv.py data.csv --scatter x y --color category --size value

# Correlation heatmap for all numeric columns
python3 scripts/visualize_csv.py data.csv --correlation

Time Series:

# Line chart for single variable
python3 scripts/visualize_csv.py data.csv --line date sales

# Multiple variables on same chart
python3 scripts/visualize_csv.py data.csv --line date "sales,revenue,profit"

Categorical Data:

# Bar chart (counts categories automatically)
python3 scripts/visualize_csv.py data.csv --bar category

# Pie chart for composition
python3 scripts/visualize_csv.py data.csv --pie region

Output Formats:
Specify output file with desired format extension:

# Interactive HTML (default)
python3 scripts/visualize_csv.py data.csv --histogram age -o output.html

# Static image formats
python3 scripts/visualize_csv.py data.csv --scatter x y -o plot.png
python3 scripts/visualize_csv.py data.csv --correlation -o heatmap.pdf
python3 scripts/visualize_csv.py data.csv --bar category -o chart.svg

2. Automatic Data Profiling

Generate comprehensive data quality and statistical reports using the data_profile.py script.

Text Report (default):

python3 scripts/data_profile.py data.csv

HTML Report:

python3 scripts/data_profile.py data.csv -f html -o report.html

JSON Report:

python3 scripts/data_profile.py data.csv -f json -o profile.json

What the Profiler Provides:

File information (size, dimensions)

Dataset overview (shape, memory usage, duplicates)

Column-by-column analysis (types, missing data, unique values)

Missing data patterns and completeness

Statistical summary for numeric columns (mean, std, quartiles, skewness, kurtosis)

Categorical column analysis (frequency counts, most/least common values)

Data quality checks (high missing data, duplicate rows, constant columns, high cardinality)

When to Use Profiling:
Always recommend running data profiling BEFORE creating visualizations when:

User is unfamiliar with the dataset

Data quality is unknown

Need to identify appropriate visualization types

Exploring a new dataset for the first time

3. Multi-Plot Dashboards

Create comprehensive dashboards with multiple visualizations using the create_dashboard.py script.

Automatic Dashboard:
Analyzes data types and automatically creates appropriate visualizations:

python3 scripts/create_dashboard.py data.csv

Custom output location:

python3 scripts/create_dashboard.py data.csv -o my_dashboard.html

Control number of plots:

python3 scripts/create_dashboard.py data.csv --max-plots 9

Custom Dashboard from Config:
Create a JSON configuration file specifying exact plots:

python3 scripts/create_dashboard.py data.csv --config config.json

Dashboard Config Format:

{
  "title": "Sales Analysis Dashboard",
  "plots": [
    {"type": "histogram", "column": "revenue"},
    {"type": "box", "column": "revenue", "group_by": "region"},
    {"type": "scatter", "column": "advertising", "group_by": "revenue"},
    {"type": "bar", "column": "product_category"},
    {"type": "correlation"}
  ]
}

Dashboard Plot Types:

histogram: Distribution of numeric column

box: Box plot, optionally grouped by category

scatter: Relationship between two numeric columns

bar: Count of categorical values

correlation: Heatmap of numeric correlations

Workflow Decision Tree

Use this decision tree to determine the appropriate approach:

User provides CSV file
│
├─ "Profile this data" / "Analyze this data" / Unfamiliar dataset
│  └─> Run data_profile.py first
│     Then offer visualization options based on findings
│
├─ "Create dashboard" / "Overview of the data" / Multiple visualizations needed
│  ├─ User knows exact plots wanted
│  │  └─> Create JSON config → run create_dashboard.py with config
│  └─ User wants automatic dashboard
│     └─> Run create_dashboard.py (auto mode)
│
└─ Specific visualization requested ("histogram", "scatter plot", etc.)
   └─> Use visualize_csv.py with appropriate flag

Best Practices

Starting Analysis

Always profile first for unfamiliar datasets: python3 scripts/data_profile.py data.csv

Review the profiling output to understand:

Column data types and ranges

Missing data patterns

Data quality issues

Statistical distributions

Choosing Visualizations

Consult references/visualization_guide.md for detailed guidance. Quick reference:

Distribution: Histogram, box plot, violin plot

Relationship: Scatter plot, correlation heatmap

Time series: Line chart

Categories: Bar chart (preferred) or pie chart (use sparingly)

Comparison: Box plot grouped by category

Creating Dashboards

Automatic dashboard: Good for initial exploration

Custom dashboard: Better for presentations or specific analysis goals

Limit plots: Keep to 6-9 plots maximum for readability

Logical grouping: Group related visualizations together

Output Considerations

HTML: Best for interactive exploration (zoom, pan, hover tooltips)

PNG/PDF: Best for reports and presentations

SVG: Best for publications requiring vector graphics

Dependencies

The scripts require these Python packages:

pip install pandas plotly numpy

For static image export (PNG, PDF, SVG), also install:

pip install kaleido

Example Workflows

Exploratory Data Analysis

# 1. Profile the data
python3 scripts/data_profile.py sales_data.csv -f html -o profile.html

# 2. Create automatic dashboard
python3 scripts/create_dashboard.py sales_data.csv -o dashboard.html

# 3. Dive deeper with specific plots
python3 scripts/visualize_csv.py sales_data.csv --scatter price sales --color region
python3 scripts/visualize_csv.py sales_data.csv --boxplot revenue --group-by product

Report Generation

# Create specific visualizations for report
python3 scripts/visualize_csv.py data.csv --histogram age -o fig1_distribution.png
python3 scripts/visualize_csv.py data.csv --scatter income age -o fig2_correlation.png
python3 scripts/visualize_csv.py data.csv --bar category -o fig3_categories.png

# Generate data summary
python3 scripts/data_profile.py data.csv -f html -o data_summary.html

Interactive Dashboard

# Create custom dashboard for presentation
# 1. First, create config.json with desired plots
# 2. Generate dashboard
python3 scripts/create_dashboard.py data.csv --config config.json -o presentation_dashboard.html

Troubleshooting

"Column not found" errors:

Run data profiling to see exact column names

CSV columns are case-sensitive

Check for leading/trailing spaces in column names

Empty or incorrect visualizations:

Verify data types (numeric vs categorical)

Check for missing data in plotted columns

Ensure sufficient non-null values exist

Script execution errors:

Verify dependencies are installed: pip list | grep plotly

Check Python version: Python 3.6+ required

For image export issues, install kaleido: pip install kaleido

Resources

scripts/

visualize_csv.py: Main visualization script with all chart types

data_profile.py: Automatic data profiling and quality analysis

create_dashboard.py: Multi-plot dashboard generator

references/

visualization_guide.md: Comprehensive guide for choosing appropriate chart types, best practices, and common patterns