Claude data analysis skills (data plugin)
The data plugin is a set of analyst skills Anthropic published for Claude Cowork and Claude Code. It writes SQL, explores a table, charts the result, builds a dashboard and reviews the analysis before it goes to anyone.
claude plugin install data@knowledge-work-pluginsAnthropic's knowledge-work-plugins repository bundles skills by job function, and the data plugin is the one for analysis. Its README calls it a data analyst plugin built mainly for Claude Cowork, Anthropic's desktop app for agent work, that also runs in Claude Code. It is open source under Apache 2.0. This page covers what is inside, how to install it and where it needs your help. For background on skills, read how Claude Skills work.
What the skills do
The plugin's skills folder holds these, each callable as a slash command or picked up from your request:
- analyze: answers a data question, from a single metric to a full report for stakeholders.
- explore-data: profiles a new table or file, with null rates, distributions, duplicates and odd values.
- write-query: writes SQL for your dialect, including Snowflake, BigQuery and Postgres.
- create-viz: turns results into a chart with Python, static or interactive.
- build-dashboard: builds one self-contained HTML file with KPI cards, charts, filters and tables.
- validate-data: checks methodology, calculations and bias before you present.
- statistical-analysis, sql-queries and data-visualization: background knowledge the others draw on.
- data-context-extractor: helps you write a company-specific version that records how your team defines its metrics.
With or without a database
Connect a data warehouse through an MCP server (the README names Snowflake, Databricks and BigQuery) and Claude can read schemas and run queries itself. Without one, paste query results or upload CSV and Excel files, or let Claude write SQL for you to run and then analyze what you paste back.
When to use it
- A quick answer from a sales or product export without opening a BI tool.
- A shareable dashboard for a weekly review.
- A second check on your own analysis before a meeting.
If the deliverable is a workbook, the Excel skill is the better fit. Our data prompts and business prompts work in any chat without installing anything.
How to install it
Claude Cowork: install plugins from claude.com/plugins.
Claude Code:
claude plugin marketplace add anthropics/knowledge-work-plugins
claude plugin install data@knowledge-work-plugins
Skills then fire when relevant, and commands appear with the plugin prefix, such as /data:write-query.
Example request
/data:analyze Here is [file].csv with [what each row is]. What drove the change in [metric] between [period 1] and [period 2]? Show the query or calculation for each number, then run validate-data on your own answer.
Limits
- The README calls these generic starting points. They get more useful once you add your own metric definitions and tools.
- Direct queries need an MCP connection to your warehouse, with the access rights that implies. Give it a read-only account.
- Results from pasted files are only as good as the export. Profile new data with explore-data before trusting totals.
- The plugin is written for Cowork and Claude Code. Anthropic does not document it for Claude.ai chat or the API.
Sources
- data plugin README — anthropics/knowledge-work-plugins, GitHub, accessed October 2026
- knowledge-work-plugins README — GitHub, accessed October 2026
- data plugin skills folder — GitHub, accessed October 2026