Harald Kirschner from Microsoft demonstrates advanced context engineering techniques in VS Code, showing developers how to unlock the full capabilities of GitHub Copilot for deeply context-aware AI code generation.

Your codebase, your rules: Customizing Copilot with context engineering

Speaker: Harald Kirschner, Principal Product Manager, Microsoft

Overview

This video delves into optimizing GitHub Copilot within Visual Studio Code by applying context engineering principles. Developers will gain insights into making Copilot more context-aware—tailoring it to not only their codebase, but team workflows and architectural patterns.

Key Topics

Default Copilot Context in VS Code:
- Understand what Copilot “sees” out of the box and the scope of its context window.
Core Techniques
- Using custom instructions and prompt engineering to refine Copilot’s output for specific projects.
- Applying Plan Mode and context engineering to adapt Copilot’s responses to your workflow.
Example Workflow
- How to implement a custom test-driven development (TDD) flow using Copilot and context customization.
Scaling Context Engineering Enterprise-Wide
- Frameworks for implementing context-aware practices across large codebases and teams with Copilot Enterprise.
- Ready-to-use templates for sharing project-wide context and standardized instruction sets.
Getting Started
- Practical recommendations for enabling advanced Copilot features in enterprise environments and setting up your context framework.

Chapters

00:00 Welcome to context engineering
01:02 What Copilot sees: The default context in VS Code
04:59 Core techniques: Using instructions, prompts, and Plan Mode
12:18 Example workflow: Custom test-driven development (TDD) flow
15:35 Conclusion and getting started

Additional Resources

Takeaways

Practical strategies for customizing Copilot’s context-awareness
Framework for scaling Copilot’s capabilities in enterprise settings
Templates and recommendations for immediate adoption in professional projects