Did you know the Databricks Assistant now supports Agent Skills? If your team has common, repeatable workflows, this feature is one you need to explore.
Skills provide the Databricks Assistant with a specific set of instructions to handle tasks customized for your organization. Imagine automating tasks like:
- Creating a "genie room" with your team's exact standards.
- Setting up a data pipeline that adheres to your internal best practices.
- Generating metadata files in your organization's required format.
- Styling and formatting a dashboard with a consistent look.
How to Add a Skill
It’s easy to get started. Navigate to the agent skills folder within the Databricks Assistant settings. From there, you can either build a new skill from scratch or import an existing one directly from a GitHub repository.
Check out these examples for inspiration:
How to Create a Skill
A skill is defined in a mandatory markdown file named skill.md. In this file, you'll outline the specific steps and resources the assistant agent must use to achieve the desired outcome. You can also include supporting resources in the skill's folder, such as utility scripts and templates. For a detailed guide, please review the repositories linked above.
How to Use the Skill
Once your skill is created and added, using it is simple. Just type your request into the Assistant while it is in agent mode, and it will automatically identify and use the appropriate skill.
Tip: If the agent isn't finding the correct skill, you can provide hints and context within your workspace instructions file.-----For complete documentation on skills, visit the official Databricks page: Databricks Assistant Skills Documentation.
Have you developed any innovative skills for the Databricks Agent, let us others know in the comments below.
