Agent skills

Use AI coding agents like Cursor, Claude Code, Codex, and Antigravity to build and maintain your Lightdash semantic layer.

AI coding agents can help you build and maintain your Lightdash semantic layer more efficiently. By installing Lightdash skills, you enable these agents to understand your semantic layer structure and use the latest Lightdash features correctly.

What are skills?

Skills are context files that teach AI coding agents about Lightdash’s YAML configuration, metrics, dimensions, and best practices. They’re powered by agentskills.io and provide agents with the knowledge they need to:

• Generate correct YAML configurations for metrics and dimensions
• Follow Lightdash naming conventions and best practices
• Understand table relationships and joins
• Use the latest Lightdash features and syntax

Supported tools

Lightdash skills work with popular AI coding assistants:

• Cursor - AI-powered code editor
• Claude Code - Anthropic’s coding assistant
• Codex - OpenAI’s code generation model
• Antigravity - AI development platform
• GitHub Copilot - GitHub’s AI pair programmer (supports skills in .github/skills and .claude/skills directories). Learn more about Copilot agent skills.
• Other tools that support the skills format

Installing skills

Use the Lightdash CLI to install skills for your preferred AI coding agent.

Prerequisites

• Lightdash CLI installed

Install globally

Install skills globally to use them across all your projects (recommended, since the version of skills is tied to the version of the Lightdash CLI, which is typically installed globally):

# Install globally for Claude (default)
lightdash install-skills --global

# Install globally for Cursor
lightdash install-skills --agent cursor --global

# Install globally for Codex
lightdash install-skills --agent codex --global

Global skills are installed to your home directory (e.g., ~/.claude/skills/).

Install for your project

Install skills at the project level for team collaboration:

# Install for Claude (default)
lightdash install-skills

# Install for Cursor
lightdash install-skills --agent cursor

# Install for Codex
lightdash install-skills --agent codex

This creates a skills directory at your git root:

• .claude/skills/ for Claude (also works with GitHub Copilot)
• .cursor/skills/ for Cursor
• .codex/skills/ for Codex

Using GitHub Copilot? Run lightdash install-skills (the default command) to install skills to .claude/skills/. GitHub Copilot automatically reads skills from this directory.

Install to a custom path

Specify a custom installation path:

lightdash install-skills --path ./my-project

Using skills with your agent

Once installed, your AI coding agent automatically picks up the Lightdash skills. You can then ask your agent to:

• Create new metrics and dimensions
• Generate YAML configurations for dbt models
• Fix validation errors in your semantic layer
• Refactor existing metric definitions
• Add joins between tables

Example prompts

Create a revenue metric that sums the amount field from the orders table

Add a customer_segment dimension based on the lifetime_value field

Generate the Lightdash YAML configuration for my users dbt model

The agent uses the installed skills to generate correct, up-to-date Lightdash configurations.

Keeping skills updated

Run lightdash --version to check if your skills are up to date. This command returns the relevant skill update command based on where your skills are currently installed and whether they’re outdated.

lightdash --version

This ensures your agent always has the latest skill definitions with the newest Lightdash features and best practices.