Guide · Agent skill
How to use the Data Analyst Expert agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Data Analyst Expert skill is a set of instructions that tells an AI agent to work on data the way a careful analyst would: check data quality first, clean it, explore it, chart it and report findings in plain language. It is useful for analysts and students working with Python or SQL, and for non-specialists who need a summary of what a dataset shows. This guide explains what the skill changes, how to install it, how to trigger it and where its limits are.
Data Analyst ExpertRightNow-AI/openfang · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the data-analyst ZIP from this page.
- 2Unzip the folder. Unzip it so SKILL.md is inside a data-analyst folder.
- 3Place the skill. Move the folder to ~/.claude/skills/ for personal use, or to a project's .claude/skills/ folder.
- 4Or upload in the apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 5Make sure Python libraries are available. Confirm pandas, numpy, matplotlib and seaborn are installed, since the instructions rely on them and the skill does not bundle them.
- 6Ask a data question. Describe the dataset and the question in plain words, and attach or point to the data.
- 7Review the output. Check the cleaning steps, charts and statistics before using the conclusions.
What an agent skill is and what this one changes
An agent skill is a folder containing a SKILL.md file. The file holds instructions that the agent reads when a task matches the skill's description. This skill is named data-analyst and is described as a data analysis expert for statistics, visualization, pandas, and exploration.
Once installed, the agent starts data work with exploratory analysis. It inspects the dataset for missing values, duplicates and odd formats, cleans it, looks at distributions and correlations, and produces labelled charts. It also applies statistical checks such as choosing an appropriate test, reporting effect sizes and confidence intervals, and warning against causal claims drawn from correlations.
- The skill is instructions and best practices only.
- It does not add new tools by itself. Analysis relies on Python with pandas, numpy, matplotlib and seaborn, and on SQL access if SQL queries are needed. The skill references these libraries but does not bundle them.
Install the Data Analyst skill
Download the ZIP file from this page. Unzip it so that SKILL.md sits inside a data-analyst folder. Then put that folder in one of two places: your personal skills folder, or a project-specific folder.
You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal install: place the data-analyst folder in ~/.claude/skills/
- Project install: place the data-analyst folder in the .claude/skills/ folder of your project
- Claude apps: Settings → Capabilities → Skills, then upload the skill
~/.claude/skills/data-analyst/SKILL.mdTrigger the skill with natural requests
You do not need a special command. Describe the data task in plain words and mention the dataset. The agent decides whether the skill applies based on its description, which covers statistics, visualization, pandas and exploration.
Requests that fit the skill include asking for a first look at a file, asking for a chart of a trend, asking whether two groups differ, or asking what a dataset shows in plain language. Give the agent the data itself or the file path and say what question you want answered.
- "Look at sales.csv and tell me what's wrong with it before we analyse anything."
- "Clean this customer table and document each step."
- "Is the difference in average order value between the two regions real? Report effect size and confidence intervals."
- "Make a line chart of monthly signups with labelled axes and a title."
- "Summarise in plain English what this survey data shows."
What the skill asks the agent to do
The source instructions set a consistent workflow. Data quality comes first, followed by exploration, cleaning, charting and statistics. The agent is told to choose visualizations by data type, for example bar charts for categories, line charts for time series, scatter plots for correlations and histograms for distributions.
On charts, the instructions ask for a title, labelled axes and units, restrained colour that highlights the key insight, and no 3D or cluttered pie charts. On statistics, the instructions ask for central tendency and spread reported together, and for assumptions such as normality to be checked before parametric tests.
- Inspect with df.shape, df.dtypes, df.head(), df.describe() and df.isnull().sum()
- Clean with deliberate missing-value handling, pd.to_datetime date parsing, df.duplicated() checks and documented steps
- Chart with titles, labelled axes, annotated thresholds and high-DPI export for reports
- Test with t-tests, chi-square or Mann-Whitney, always alongside effect sizes and confidence intervals
Check the safety notes before use
This skill runs no scripts and uses no network access. It is a text file of instructions. Anything the agent executes comes from the tools and libraries already available in your environment, not from the skill itself.
The skill gives no privacy or data-handling guidance. If your data is sensitive or regulated, review the agent's code and outputs before relying on them, and decide yourself what may be shared with the agent.
- Runs scripts: no
- Uses network: no
- No privacy or data-handling rules are included
Tips for better results
Give the agent context about the data: what each column means, where it came from and what decision the analysis supports. The skill asks for cleaning choices to depend on context, so context changes the result. Ask for the cleaning steps to be written down so the analysis can be repeated.
Keep the skill's own warnings in mind when reading results. Correlation does not prove causation, small samples are unreliable, and aggregation level can reverse a trend (Simpson's paradox).
- State the question before the data, so the agent knows which tests matter.
- Ask for a list of cleaning steps as part of the output.
- Ask for effect sizes and confidence intervals alongside p-values.
- Ask the agent to flag any causal language it is tempted to use.
What people use it for
- First-pass quality check on a new CSV, covering missing values, duplicates and inconsistent date formats.
- Exploring the distribution and correlations of numeric fields before building a model.
- Comparing two groups with an appropriate test, reported with effect size and confidence intervals.
- Producing a labelled chart with a title, axes and units for a team report.
- Writing a plain-language summary of a dataset for a non-specialist reader.
Frequently asked questions
How do I install the data-analyst agent skill?
Download the ZIP from this page and unzip it so SKILL.md sits in a data-analyst folder. Place that folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
What do I need to run this skill?
The instructions use Python with pandas, numpy, matplotlib and seaborn, which are referenced but not bundled, so you need them installed. SQL access to a database is needed only if you want SQL queries. The skill itself runs no scripts and uses no network.
Is the data-analyst skill free?
The source material for this page does not state pricing or licence terms beyond the listed licence, which is Apache-2.0. Check the repository for the licence text before reuse.
Does the skill add new tools to the agent?
No. It is a set of instructions and best practices in SKILL.md. The analysis depends on the Python libraries and SQL access already available in your environment.
Can I use this skill with sensitive or regulated data?
The skill gives no privacy or data-handling guidance, so it is not suited to running analysis on sensitive or regulated data without human review. Review the code and outputs and decide what data may be shared.
How is this different from just asking an AI to analyse data?
The skill sets a consistent workflow: data quality checks first, documented cleaning, chart conventions and statistical checks such as effect sizes and assumption testing. Without it, the agent may skip these steps unless you ask for them.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are an analyst, student or team that wants a structured, careful workflow and consistent cleaning and charting conventions when working with Python or SQL. Pick something else if you need privacy guidance for sensitive data, or if you expect the agent to deliver causal or high-stakes conclusions, which the skill itself warns against.
Download the skill