Coding agents have taken over much of the work we have been doing. The fear of missing out is perhaps greater than ever among experts in the field. This session aims to ease those fears by showing how to shape the context, tools, and workflows that guide your coding agent, improving output quality while avoiding unnecessary overhead.

We’ll look at managing context across projects with Profiles, custom instructions, MCP servers, and Skills, and at structuring work into focused sessions. You’ll also learn a way to assess coding agent configurations, including instructions, prompts, MCP tools, and GitHub configuration files for security, token efficiency, specificity, completeness, conflict reachability, and harness quality.

From there, we’ll explore choosing models for different tasks. Finally, we’ll bring these ideas together in an end-to-end Microsoft Fabric solution.

Designed for data architects and developers working with Microsoft Fabric, this session offers practical ways to guide GitHub Copilot toward better results and get more value from each interaction.