Dataset Explorer
Profile datasets to understand shape, quality, distributions, and key patterns
Someone hands you a CSV with 50 columns and no documentation. Before you can analyze anything, you need to understand what you're looking at — column types, missing values, distributions, outliers. That first-look step takes longer than the analysis.
Who it's for: data analysts receiving new datasets, data scientists starting exploratory analysis, BI teams onboarding new data sources, researchers working with unfamiliar datasets, anyone who opens a CSV and thinks 'where do I even start?'
Example
"Profile this 500K-row customer dataset" → Column classification, distribution summaries, missing value map, outlier detection, relationship discovery between fields, data quality flags, and recommended follow-up analyses
CLAUDE.md Template
New here? 3-minute setup guide → | Already set up? Copy the template below.
npx ccplaybooks add dataset-explorer# Dataset Explorer
## Your Role
You are a data profiling assistant. You help users understand new datasets by analyzing their shape, quality, distributions, and key patterns. You produce comprehensive data profiles and recommend follow-up analyses.
## Workflow
### 1. Access the Data
Read the provided file (CSV, Excel, Parquet, JSON) or connect to a database table. Load into a working dataset and infer column types.
### 2. Understand StructureGet new playbooks like this one
One email a week with new Claude Code workflows. Free, like everything here.
No spam. Unsubscribe anytime.
What This Does
Profiles any dataset -- CSV, Excel, Parquet, or JSON -- to help you understand its shape, quality, and patterns before diving into analysis. The assistant classifies columns, calculates distributions, flags data quality issues, discovers relationships, and recommends follow-up analyses. Think of it as a comprehensive "first look" at any new data.
Quick Start
Step 1: Download the Template
Click Download above to get the CLAUDE.md file.
Step 2: Set Up Your Project
Create a project folder and place the template inside:
mkdir -p ~/Projects/data-exploration
mv ~/Downloads/CLAUDE.md ~/Projects/data-exploration/
Add your data files (CSV, Excel, Parquet, JSON) to the same folder.
Step 3: Start Working
cd ~/Projects/data-exploration
claude
Say: "Profile this dataset and tell me what's in it"
What Gets Analyzed
The profiling covers every column in your dataset:
- All columns -- Null rates, distinct counts, most/least common values
- Numeric columns -- Min, max, mean, median, standard deviation, percentiles (p1 through p99), zero and negative counts
- String columns -- Length statistics, empty strings, pattern analysis, case consistency
- Date columns -- Date ranges, gaps in time series, distribution by month/week
- Boolean columns -- True/false/null counts and ratios
Quality Assessment
The assistant flags potential data quality issues automatically:
- High null rates -- Columns with over 5% nulls get a warning; over 20% gets an alert
- Suspicious values -- Placeholder values like "N/A", "TBD", "test", or
999999 - Cardinality surprises -- Columns that should be high-cardinality but have few distinct values (or vice versa)
- Format inconsistencies -- Mixed case, trailing whitespace, or "USA" vs "US" vs "United States"
- Business rule violations -- Negative quantities, end dates before start dates, percentages over 100%
Tips
- For very large files (100M+ rows), the profiling will use sampling by default. Mention if you need exact counts.
- This is ideal as a first step before any analysis -- get the lay of the land before writing specific queries.
- The quality flags are heuristic -- not every flag is a real problem, but each is worth a quick look.
- Ask for relationship discovery to find foreign key candidates and column hierarchies (e.g., country > state > city).
Example Prompts
"Profile this dataset and tell me what's in it"
"What's the data quality like in sales_data.csv? Flag any issues."
"Show me the distributions for all numeric columns in this file"
"What columns could I use to join this table with the users table?"
"Explore this dataset and recommend 5 analyses I should run next"