How to use the Data Analyst Workbook agent skill

What the data-analyst skill makes the agent do

Agent skills are folders of instructions that an AI agent loads when a task matches them. This one, named data-analyst, tells the agent to work through a dataset in a fixed order rather than jumping straight to an answer.

First, the agent reads the data once and checks what each field means, how many rows there are, and whether there are blanks, duplicates or inconsistent units. It asks you about anything ambiguous before continuing. Second, it builds an Excel workbook: Sheet 1 holds the cleaned detail with units and date formats unified and a note of what was excluded, and Sheet 2 holds the summary with grouped statistics, period-on-period and year-on-year comparisons and Top-N lists, all written as formulas rather than typed results. Key metrics get charts: line charts for trends and bar or pie charts for composition. Third, the agent gives 3-5 conclusions in its reply, each with an observation, a supporting figure and a likely explanation or recommendation.

• Sheet 1: cleaned detail with a note of excluded rows
• Sheet 2: grouped statistics, period comparisons and Top-N, all formula-driven
• Charts: line for trends, bar or pie for composition
• Reply: 3-5 conclusions, and 'insufficient data' where the data cannot support a claim

Install the data-analyst skill

Download the skill ZIP from this page and unzip it. The skill is a single file, SKILL.md, inside its folder. You can install it in one of two places depending on how you use your agent.

For a personal setup, unzip the folder into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. The agent loads skills from these locations, so the folder should contain SKILL.md directly.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside the project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Check what you need before you start

The skill needs a spreadsheet-capable environment. Its instructions refer to 'the spreadsheet tools available here' and do not name a specific library, so the agent must have a way to create Excel files in the environment you are using. If it cannot create a workbook, the skill cannot deliver its main output.

The skill does not run scripts and does not use the network, according to its files. The only file in the package is SKILL.md.

• A spreadsheet-capable environment where the agent can create an Excel workbook
• Your data either attached to the chat or pasted in as a table

How to trigger the skill with natural requests

You do not need special syntax. Describe the data and what you want, using plain words. The skill is matched to requests about analysing tables, cleaning data, building workbooks or charting results. The skill's own quick prompts give a good pattern to copy.

Attach your file or paste the table, then ask for either conclusions, a workbook, or both. If you want charts, say so. If you already know which fields matter, name them, because the agent will ask about anything unclear before building the workbook.

• 'Analyse this table and give me the key conclusions'
• 'Turn this data into an analysis workbook with charts'
• 'Clean this sales export, group it by region and show the top 10 products'
• 'Compare this month with last month and chart the trend'

Read the output and check the conclusions

Open the workbook and check Sheet 1 first. The note of what was excluded tells you which rows were removed and why, so confirm those decisions match your intent. Then check Sheet 2. Because the summary uses formulas, you can click into cells to see how each figure was calculated and change the source data to see results update.

Read the conclusions as hypotheses. Each one should give a figure and a likely explanation or recommendation. Where the data is too thin, the agent should say 'insufficient data' instead of guessing, and you should treat that as a useful answer.

• Confirm the exclusion note on Sheet 1
• Click into Sheet 2 formulas to verify calculations
• Treat conclusions as starting points for your own judgement

Safety notes and limits

The skill is built around a single table and spreadsheet output. It is not designed for large datasets or production pipelines, and it is not meant to make repeatable automated reports. Do not rely on its summary alone for decisions with legal, medical or financial consequences, because the conclusions are inferred from the data and not validated by a specialist.

The skill runs no scripts and uses no network access, so the data stays with the agent and environment you are already using. Review the workbook yourself before sharing it.

• Not designed for large datasets or production pipelines
• Do not base legal, medical or financial decisions on the summary alone
• Runs no scripts and uses no network, per its files