Pack No. 28 · Business

Data Analyst

From a raw file to numbers you can defend.

Who it's forAnalysts, operators and founders who answer questions from spreadsheets, CSV exports and SQL. You bring the data and the question. No warehouse connection is required: paste results or upload a file. Most cards are plain instructions that also load on claude.ai. On a terminal agent, charts need Python and Excel files need LibreOffice.
Install the "Data Analyst" skill collection from starterdeck.dev. It is a set of agent skills, each fetched from its author's GitHub repository at a pinned commit with the open-source skills CLI.

Check first that Node.js 22.20 or newer is installed (node --version). If it is not, stop and tell me.

Then run these commands exactly as written, one after another. Do not change them and do not add others. If one fails, stop and show me the error.

npx skills add anthropics/knowledge-work-plugins@0e0fcacddcc892eb92a1b7873ff9d68159c47571 --skill explore-data analyze sql-queries statistical-analysis data-visualization build-dashboard validate-data -g -y
npx skills add anthropics/skills@e4bb5d960d87126ff3b3a8af1a104b7db67560ba --skill xlsx -g -y
npx skills add anthropics/financial-services@3e32c2287dcb637c184c01cc0ad82d6af7e9bb0f --skill audit-xls -g -y

When they have finished, list the skills that were installed.

Paste it into Claude Code, Codex, Cursor, Gemini CLI or any other agent with a terminal. It fetches 9 skills from 3 author repos, each pinned to the version we reviewed, with the commands on the Terminal tab. Nothing runs until you send the message.

What's inside

9 skills. Each card links to its author's repository at the exact version we reviewed.

  1. Explore data 1/9 Data profiling Ready Profiles a table or file before you analyse it: size, column types, empty cells, value ranges, duplicates and suspicious values, then suggests what to look at next. @anthropics Apache-2.0

    Explore data

    Look before you count. It profiles the file first: size, column types, empty cells, ranges and duplicates, so the surprises come before the analysis and not after.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac
  2. Analyze data 2/9 Data analysis Ready Answers a data question at the depth it needs, from one number to a full report, and checks row counts, empty cells and totals before it presents a result. @anthropics Apache-2.0

    Analyze data

    Answers the question at the depth it needs, from one number to a full report, and checks row counts and totals before it shows you a result.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac
  3. SQL queries 3/9 SQL Ready Writes SQL in your database's syntax, with notes for PostgreSQL, Snowflake, BigQuery, Redshift and Databricks and patterns for window functions, cohorts, funnels and duplicates. @anthropics Apache-2.0

    SQL queries

    Most data questions end in a query. This keeps the agent in the right dialect and brings patterns for cohorts, funnels and duplicates. The analyze card calls on it by name.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac
  4. Statistical analysis 4/9 Statistics Ready Guides the statistics in an analysis: mean or median, percentiles, trends, outliers and significance tests, and flags claims the data cannot carry, like correlation read as cause. @anthropics Apache-2.0

    Statistical analysis

    Stops the agent from reporting a bare average or calling a small difference a finding. It brings medians, percentiles, significance checks and the caveats a careful analyst adds.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac
  5. Excel spreadsheets 5/9 Excel files Needs app Creates, edits and reads Excel files with live formulas instead of pasted values, recalculates them with LibreOffice and holds the file back while formula errors remain. @anthropics Apache-2.0

    Excel spreadsheets

    The file your colleagues will open. It builds workbooks with live formulas instead of pasted values and recalculates them before you get them back.

    Needs an app or CLI Community · Anthropic Apache-2.0
    Source anthropics/skills at e4bb5d9
  6. Spreadsheet audit 6/9 Formula audit Ready Audits a selection, a sheet or a whole financial model for formula errors, hidden hardcodes, broken links and statements that do not tie, and reports before it changes anything. @anthropics Apache-2.0

    Spreadsheet audit

    For the spreadsheet you inherited. It hunts formula errors, hidden hardcodes and broken links across a sheet or a whole model.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/financial-services at 3e32c22
  7. Data visualization 7/9 Charts Needs app Picks the chart type that fits the data, writes the Python code with matplotlib, seaborn or plotly, and applies rules for colour, titles, honest axes and accessibility. @anthropics Apache-2.0

    Data visualization

    Chooses the chart for the question, writes the Python that draws it and keeps the axes honest. The analyze card calls on it by name.

    Needs an app or CLI Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac
  8. Build a dashboard 8/9 Dashboards Ready Builds a dashboard as one HTML file that opens in a browser: headline numbers, Chart.js charts, filters and a sortable table, with your data embedded in the file. @anthropics Apache-2.0

    Build a dashboard

    Turns a CSV or a query result into one HTML file with filters and charts that you can send to a colleague, with no BI tool and no server.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac
  9. Validate an analysis 9/9 Analysis QA Ready Reviews an analysis before you share it: method, calculations, charts and conclusions. Checks it for common pitfalls and rates it ready, ready with caveats or in need of revision. @anthropics Apache-2.0

    Validate an analysis

    The second pass before the numbers leave your desk. It looks for inflated joins, partial periods and averages of averages, then says what to fix and which caveats to state.

    Works right away Community · Anthropic Apache-2.0
    Source anthropics/knowledge-work-plugins at 0e0fcac

Good to know

What you bring

  • Python 3 with openpyxl, pandas and markitdown
  • LibreOffice, used to recalculate formulas
  • Python 3 with pandas, matplotlib and seaborn (plotly for interactive charts)

Pick one, not both

  • Whole plugin: seven of these cards come from Anthropic's data plugin. In Claude Code or Cowork you can install that plugin instead, which also brings slash commands and warehouse connectors. Use the plugin or the single cards, not both.
  • SQL dialects: sql-queries covers PostgreSQL, Snowflake, BigQuery, Redshift and Databricks. It has no notes for MySQL, SQL Server, SQLite or DuckDB.
  • Dashboards: build-dashboard loads its chart library from a CDN, so the file needs a connection to draw its charts.
  • Product metrics: for a weekly or monthly review against targets, see metrics-review in the Product Manager pack.
  • Outside sources: for questions your own data cannot answer, the Researcher pack adds a research workflow and source checks. It shares explore-data and analyze with this pack.
Review slipPack No. 28
  • Every skill file read in full
  • Pinned to a reviewed version
  • Author and licence on every card
  • Requirements and costs disclosed
  • No sponsored skills inside
Last reviewed
Nov 2026
Reviewed