Exploratory Data Analysis Skill

A systematic first pass over any dataset: structure, quality, distributions, relationships, and what to do next.

Try in Runcell
Made by
Runcell team
Category
Data Analysis
Format
SKILL.md
Works with
Claude Code, Codex, Cursor, Runcell (Jupyter)

Showcase

Illustrative output. The real run happens in your notebook.

DataFrame
ColumnTypeMissingFlag
ageint0.4%
incomefloat18.2%skewed, outliers
signup_datedate0%after target date
Column profile with the flags that matter.
Findings
1. signup_date is after the target event in 6% of rows (leakage)
2. income missing not at random (higher for churned users)
3. 212 duplicate customer_id rows
Next: drop signup_date, impute income by segment, dedupe
Three to five findings, ranked by impact.

Description

Exploratory data analysis in Python is a checklist that experienced analysts carry in their heads: shape and types, missingness, duplicates, distributions, outliers, correlations, leakage risks. This skill writes the checklist down so an agent runs it the same way every time and reports what it found.

It produces a short readable summary, not a forty-page auto-report. The plots and tables stay in notebook cells so you can drill into anything that looks wrong.

  • Profiles columns: types, cardinality, missing values, duplicates
  • Plots distributions and flags skew, outliers and impossible values
  • Computes correlations and highlights redundant or leaking features
  • Checks target balance and time ordering for modeling datasets
  • Writes a plain-language summary with the three to five findings that matter
  • Suggests cleaning and feature steps as a follow-up list
Editor's Recommendation

Recommended by

Runcell team

Why we built this skill

Auto-EDA reports are long and nobody reads them. We wanted the checklist an experienced analyst runs in their head, and a short list of what actually matters. The leakage and time-ordering checks are the ones that have saved models from looking better than they were.

What you get

  • A data quality summary
  • Distribution and correlation plots in cells
  • A findings list ranked by impact
  • A suggested next-steps checklist

Use it when

  • The first hour with a new dataset
  • Checking a dataset before training a model
  • Reviewing a colleague's data before you trust their results

FAQ

What is an EDA skill for Claude Code?

It is a skill file that tells the agent how to explore a dataset systematically with pandas, matplotlib and seaborn: what to check, in what order, and how to summarize the findings. It turns "look at the data" into a repeatable procedure.

How is it different from ydata-profiling or similar auto-EDA tools?

Those tools generate an exhaustive report. The skill generates judgment: it runs the checks, then tells you the few things that matter and why. It can call a profiling library as one of its steps.

Does it work on large datasets?

It samples when needed and says so. For files that do not fit in memory it switches to chunked reads or Polars, if available in your environment.

What comes after EDA?

Usually cleaning, then either statistics or modeling. The findings list feeds the Excel data analysis, statistical test selection and dataviz skills directly.

Related Skills

View all
Data Analysis

Excel Data Analysis Skill

Read, clean and analyze spreadsheets with pandas. Reproducible answers, not a rewritten sheet.

Runcell teamFree
Statistics

Statistical Test Selection Skill

From question and data to the right test, with assumptions checked and results reported correctly.

Runcell teamFree
Publication-style dose-response figureVisualization

DataViz Skill

Chart choice, consistent styling and journal-ready export for matplotlib, seaborn and plotly.

Runcell teamFree

Information

Format
SKILL.md (Agent Skills)
Works with
Claude Code, Codex, Cursor, Runcell (Jupyter)
Runs in
A live Jupyter kernel in Runcell, or your terminal agent
Maintained by
Runcell team
Last updated
Sep 20, 2026
Price
Free

Run this skill in your notebook

Runcell is a Jupyter-native AI agent that executes every step in a live kernel. Results are cells you can inspect and rerun.

Try in Runcell