dbt has released a set of agent skills aimed at enhancing AI capabilities in data work. These skills allow AI agents, like Claude Code and OpenAI's Codex, to adopt best practices from the dbt community. By integrating these skills, coding agents shift from generalists to specialists in analytics engineering, improving their ability to handle tasks like building dbt models, troubleshooting job failures, and migrating projects. The development of these skills draws from extensive real-world experience and feedback from dbt experts. The skills repository includes tools for analytics engineering, creating metrics with MetricFlow, managing platform operations, and migrating projects to the dbt Fusion engine. The skills vary in complexity, with some covering entire workflows while others focus on specific tasks. Users can add these skills to their agents and start leveraging them immediately. The article emphasizes the importance of context for AI agents, which skills provide by embedding procedural knowledge and company-specific insights. There’s an ongoing evolution in how these skills integrate into agent workflows. As community feedback expands, the skills will become more refined and effective. The article contrasts skills with the dbt MCP server, clarifying that while both enhance AI functionality, they serve different purposes: MCP focuses on tool access, while skills supply contextual knowledge. This distinction highlights the complementary nature of the two, underscoring the need for both in maximizing AI performance in data projects.