owid-data

OWID Skills

Agent skills for working with Our World in Data. Teach your AI coding agent — Claude Code, OpenAI Codex, Gemini CLI, Cursor, GitHub Copilot, and others — to search our charts, download the data behind them, and analyze it correctly.

Our World in Data publishes thousands of charts and datasets on global problems: poverty, health, energy, climate, education, and more. These skills give agents the knowledge to use that data well — the right APIs, the right query parameters, and the caveats that matter (country harmonization, citations, metadata).

> Status: early and experimental. Interfaces may change. Feedback and issues are welcome!

What you can do

Once installed, you can ask your agent things like:

• "Get life expectancy data for the US and UK since 1950 and plot the trend."
• "Find OWID charts about renewable energy adoption."
• "Download CO₂ emissions per country and compute per-capita values using OWID population data."
• "Make a scatter plot of child mortality against GDP per capita."

The skills trigger automatically when relevant — you don't need to invoke them by name.

Skills

Skills follow the open Agent Skills format (SKILL.md), so they work with any agent that supports the standard.

| Skill | What it does | Requires | |---|---|---| | search-charts | Search OWID's published charts by keyword | curl, jq | | fetch-chart-data | Download the data and metadata behind any chart | curl, jq | | joining-data | Join OWID data with external sources (per-capita metrics, scatter plots vs GDP, …) | duckdb | | owid-catalog | Python-native access to the full OWID catalog (charts, tables, indicators) via the owid-catalog library | uv (or pip) |

The HTTP-based skills are lightweight and language-agnostic. owid-catalog is the richer option when Python is available — it returns metadata-aware DataFrames and covers the full data catalog beyond published charts.

Installation

Claude Code

Install as a plugin from the marketplace:

/plugin marketplace add owid/skills
/plugin install owid@owid-skills

Other agents (Codex, Gemini CLI, Cursor, Copilot, …)

These are standard Agent Skills, read as-is by Codex, Gemini CLI, Cursor, GitHub Copilot, and many other tools — no Claude-specific setup required. The skills CLI detects which of your installed agents support skills (75+ supported) and copies them into each one's directory:

npx skills add owid/skills            # into the current project
npx skills add owid/skills --global   # user-level, across all your projects

Add --agent '*' to install to every supported agent, or --list to preview the skills first.

Manual

The skills are plain Agent Skills folders, so you can also copy or symlink them into whatever directory your agent reads. Clone the repo:

git clone https://github.com/owid/skills owid-skills

Then put owid-skills/skills/* where your agent looks for skills:

• Per project — ./.agents/skills/ (the shared convention read by Codex, Cursor, OpenCode, …)
• Per user — your agent's own skills directory, e.g. ~/.codex/skills/, ~/.gemini/skills/, or ~/.claude/skills/

Prerequisites

The skills use a few common command-line tools: curl, jq, duckdb, and uv. Install them with your package manager (e.g. brew install jq duckdb uv), or on macOS run:

curl -sSL https://raw.githubusercontent.com/owid/skills/main/install-prerequisites-macos.sh | bash

Using the data

Data published by Our World in Data is open: it is available under the Creative Commons BY license, and it builds on the work of the original data providers. The skills instruct agents to surface proper citations — please keep them when you publish results.

Development

Want to add or improve a skill? See AGENTS.md for repo conventions (skill layout, registration, testing). In short:

• Each skill is a directory under skills/ with a SKILL.md (frontmatter name must match the directory name).
• Register new skills in .claude-plugin/marketplace.json.
• Test with claude plugin validate . and by loading the plugin directly: claude --debug --plugin-dir .
• Plugins are versionless on purpose: every commit to main is a release.

This repository is limited to skills that rely on public OWID endpoints and common CLI tools; skills that require OWID-internal infrastructure or credentials are out of scope.

License

Apache-2.0