> ## Documentation Index
> Fetch the complete documentation index at: https://lightdash-refactor-full-audit-reorg.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Install Lightdash Skills

> Enable AI-powered development with Claude Code or Cursor by installing Lightdash skills

Installing the Lightdash skills lets you build this track's semantic layer conversationally with an AI coding agent instead of hand-writing YAML. The skills teach your agent everything it needs to know about building in Lightdash — from writing metrics to creating dashboards — so you can describe what you want and have it generate correct, validated configuration.

<Frame>
  <iframe width="100%" height="420" src="https://www.youtube.com/embed/0uUmTLdmtig" frameborder="0" webkitallowfullscreen mozallowfullscreen allowfullscreen />
</Frame>

## Quick install

```bash theme={null}
lightdash install-skills
```

This installs skills for Claude by default. For global installs, custom paths, other AI agents (Cursor, Codex, and more), and keeping skills up to date, see the [Agent skills guide](/workflow/install-agent-skills).

## What it unlocks

With the skills installed, your agent can:

* **Build from scratch.** Describe the report you want and have your agent profile your warehouse, suggest metrics, and build dashboards — no existing dbt files required.
* **Run bulk operations.** Rename tables or metrics and update every dependent chart and dashboard in one pass.
* **Migrate from other tools.** Translate existing metric logic into Lightdash definitions.

<Note>
  When using Claude, prefix your prompts with `/developing-in-lightdash` so it applies the Lightdash skills effectively.
</Note>

<Tip>
  **Working with verified content?** The Lightdash skills are aware of verified charts and dashboards (exposed via the [Lightdash MCP](/agents/lightdash-mcp#data-exploration)), but they won't automatically prefer verified content unless you ask. To treat it as the canonical pattern when building new content, say so in your prompt — for example: *"Reference our verified charts and dashboards as the source of truth when building new content."*
</Tip>