Agent skill - Statistics Estonia

Today I published my first agent skill: using Statistics Estonia databases. Often I have some simple question and I know data exists, but I can't be bothered to figure out the clunky UIs and directory trees and subtle variants of tables. This is now automated for me.

It's built for Claude Code, but it obviously works with all other agents following the Agent Skills spec. You can find it at taivop/skill-statistics-estonia and the easiest quickstart should be:

git clone https://github.com/taivop/skill-statistics-estonia .claude/skills/statistics-estonia

Example

At one point I wanted to visualize the pay gap beyond a single number. Here's the relevant prompt:

Make a chart of cumulative salary distribution of men vs women, latest data available

The actual report is much longer. But the main chart:

You'll find more examples here.

Observations

Creating the skill initially itself was relatively simple: I fed an agent the API documentation of Statistics Estonia. The level-up came from adding metadata to each table so agents using this skill can understand the nuances and context of a table; this is done through a bit of HTML-scraping and URL-matching, which is a bit brittle but probably won't break soon.

I'm a bit confused about how to best package and distribute a skill. At first I tried to put everything (including the human readable README.md) into the Github repo, but then the agent got confused by the readme, so I had to remove it. Potential users also benefit from examples, but I also don't want to have these visible to the agent; for now I just instruct the agent to not look at the examples directory, but this seems a bit inelegant.