AI Skill
This page does not apply to the environment you selected. It covers: AI Skill.
The Datama skill teaches an AI assistant to build interactive Datama visualizations. You ask a question about your data in plain language; the assistant picks the module, builds the settings, and gives you a standalone HTML page.
Your data stays in the page: it is rendered in your browser, and rows are not sent to Datama.
What the skill does
Section titled “What the skill does”The skill handles two modules, and picks one per page:
| Module | When | For example |
|---|---|---|
| Compare | You want to know why a KPI changed between two sides: two periods, two segments, A vs B. | Why did revenue drop between Q1 and Q2? |
| Explore | You want to break down a metric by one or two dimensions, with no comparison needed. | Break down revenue by country. |
When a question mixes both, or is unclear, the assistant uses Compare.
The skill works with any assistant that supports the open Agent Skills format: Claude, ChatGPT, Cursor, GitHub Copilot, VS Code, Gemini CLI… See the Agent Skills client list.
Download the skill
Section titled “Download the skill”- In your Datama account, open Web app integration (or My account › Extensions), and select AI Skill.
- Choose Include license in the skill package:
- on (default): the pages run in paid mode. The zip is tied to your subscription: do not share it;
- off: the pages run in demo mode, to try the skill.
- Click Download skill. You get
datama.zip.
Download it again after a change of subscription.
Install the skill
Section titled “Install the skill”- In a web assistant: open the Skills section of its settings,
choose Upload a skill, and select
datama.zip. The datama skill appears in the list, where you can turn it on or off. - In an IDE or a desktop agent: unzip
datama.zip, and copy thedatamafolder into the skills folder of the tool, for example.agents/skills/datama/at the root of your project.
Menu names and folders vary by tool: check the documentation of your assistant.
Use the skill
Section titled “Use the skill”Mention datama in your prompt, with a slash command such as /datama or
simply in your question, and give the data:
- an uploaded CSV or Excel file;
- data pasted in the conversation;
- a connector the assistant can use (Google Sheets, BigQuery…);
- or ask the assistant to generate a test dataset.
Above about 20,000 rows, the assistant aggregates the data by the dimensions you need first.
Examples:
- Compare Q1 vs Q2 revenue and explain the change.
- Decompose the drop in conversion rate between mobile and desktop.
- Plot conversion by device and channel.
The assistant reads the data, separates dimensions and metrics, builds the
market equation, and delivers an .html page
that opens in any browser.
Adjust the page
Section titled “Adjust the page”Keep talking to the assistant: it changes the settings of the page instead of rebuilding it.
- Look: a palette (modern, ocean, sunset, datama, excel…), your own colors, or a style preset (standard, modern, presentation…). Ask to reverse the colors when a decrease is good news, for costs or churn.
- Compare chart and comment: place the comment below, on the right, on the left or hide it; change its size; add a title and a subtitle; hide the values on the bars; show the gap between two bars, in % or in value.
Get the results as text
Section titled “Get the results as text”With a licensed skill, the assistant can read back what Datama computed (the drivers table and the comment) and quote it in its answer or a report. This works in Claude, once the page has been opened, and in Claude Code or Cursor. Other assistants deliver the page only. The assistant never computes the figures itself.
Download the page
Section titled “Download the page”With a licensed skill, Download as HTML in the header of the page saves it as a file.