databricks/databricks-agent-skills

What it solves

This project provides a collection of specialized "skills" and plugins for AI coding agents (such as Claude Code, Cursor, GitHub Copilot, and Codex) to make them experts in the Databricks ecosystem. It eliminates the need for the agent to guess how to use Databricks APIs, CLI, and services, providing them with specific guidance on building apps, managing data pipelines, and deploying resources.

How it works

The project implements the Agent Skills Specification, where each "skill" consists of a set of instructions (SKILL.md) and supporting references. These are installed into the agent's local directory or cached via plugin marketplaces.

Depending on the installation method, the project provides:

• Skills: Guidance files that steer the agent's behavior for specific Databricks tasks.
• Plugins: Integration hooks and slash commands (like /databricks:setup and /databricks:doctor) that provide session context, handle authentication failures, and route prompts to the correct skill.

Who it’s for

Developers and data engineers using AI coding agents to build, deploy, and manage resources on the Databricks platform.

Highlights

• Broad Ecosystem Support: Compatible with major AI agents including Claude Code, Cursor, GitHub Copilot, and Codex.
• Extensive Skill Library: Includes stable skills for core CLI, data discovery, Lakeflow Jobs, Model Serving, Vector Search, and Unity Catalog.
• Intelligent Routing: Uses prompt routers and rules to automatically steer agents toward the relevant Databricks skill based on the user's prompt.
• Flexible Installation: Supports installation via the Databricks CLI or directly through agent-specific plugin marketplaces.