dbt-labs/dbt-agent-skills

What it solves

This project provides a collection of "Agent Skills" designed to help AI agents (like Claude Code, Cursor, or GitHub Copilot) perform dbt (data build tool) workflows more accurately. It bridges the gap between a general-purpose AI and the specific technical requirements of analytics engineering, such as building models, writing tests, and managing semantic layers.

How it works

Instead of using manual commands, these skills are sets of instructions and resources that an AI agent automatically loads when a user's natural language prompt matches a specific use case. They are compatible with the Agent Skills specification and can be installed via tools like the Vercel Skills CLI, Tessl, or directly within Claude Code.

Who it’s for

Analytics engineers and data professionals who use AI agents to assist with dbt project development, maintenance, and migration.

Highlights

• Analytics Engineering: Capabilities for building/modifying models, debugging errors, and writing unit tests.
• Semantic Layer: Tools for creating metrics and dimensions using MetricFlow.
• dbt Mesh: Support for multi-project setups, governance, and cross-project collaboration.
• Migration Tools: Specialized skills for upgrading dbt-core versions or migrating projects to the Fusion engine.
• Platform Operations: Skills for troubleshooting job failures and configuring the dbt MCP server.