What it does

CSV Data Analyzer reads a sample of your CSV (or the full file if small) and builds a structural understanding — column names, types, null counts, unique value distributions, and numeric ranges. It then answers questions in plain English: "what are the top 5 product categories by revenue this quarter?" gets you both the pandas code to compute it and the human-readable answer.

For larger datasets, it generates efficient SQL or pandas queries that you run locally — it never sends your raw data anywhere. It also detects common data quality issues: duplicate rows, date columns stored as strings, currency values with mixed formats, and columns that should be categorical. Every insight comes with the code so you can reproduce or extend the analysis yourself.

How to install

bash

npx skills add user/csv-analyzer

The command fetches the skill definition, validates the schema, and registers it with your Claude Code environment. No global npm packages are installed. The skill is stored in your project's .claude/skills/ directory.

How to use

Paste your CSV headers and a sample of rows along with your question:

prompt

Analyze this data: [paste CSV headers + sample rows]
Question: [your question]

Claude reads the structure and generates both the code and the answer. A typical output for "what are the top 5 categories by revenue?" looks like this:

python

# Top 5 product categories by revenue
top_cats = (
    df.groupby('category')['revenue']
      .sum()
      .sort_values(ascending=False)
      .head(5)
      .reset_index()
)

# Plain-English summary:
# Electronics leads with $842K (34% of total), followed by
# Apparel ($510K), Home & Garden ($298K), Sports ($241K),
# and Beauty ($187K). Electronics and Apparel together
# account for 55% of all revenue this quarter.

You can follow up with additional questions like "show me month-over-month growth for Electronics" or "are there any anomalies in the revenue column?"

Configuration

Steer the output with inline instructions when you need a different format or focus:

prompt

# SQL instead of pandas
Analyze this data: [...] Generate a SQL query instead of pandas

# Anomaly detection
Analyze this data: [...] Find all anomalies in the revenue column

# Non-technical summary
Analyze this data: [...] Summarize this data for a non-technical audience

Tip: For large CSVs, paste just the header row and 20–30 representative rows. The skill builds a full structural model from a sample and generates queries you run against the complete file — keeping your data private and local.

Related skills

Pair CSV Data Analyzer with Data Visualizer to turn your tabular insights into charts. Use JSON Transformer when your source data arrives in JSON and needs flattening before analysis, and Dashboard Builder to package the key metrics into a live dashboard component.