How to teach GitHub Copilot about your codebase | Tutorial for Beginners

by GitHub

GitHub explains how to make GitHub Copilot work better in a real project by giving it durable context about your repo. It covers adding custom instructions under .github, defining repeatable “skills” for tasks like testing, using custom agents for bigger workflows, and connecting MCP servers such as Playwright.

Overview

Why context matters for AI coding

The video focuses on improving Copilot’s output by providing project-specific context (conventions, scripts, and code style) so Copilot can follow the same standards as the rest of the codebase.

Custom instructions and .github folder setup

GitHub shows how to set up custom instructions in the repository (under the .github folder) so Copilot can consistently apply your project’s standards when generating or modifying code.

Repeatable procedures with agent skills

The tutorial introduces creating repeatable “skills” that Copilot can reuse for common tasks (for example, running tests in a consistent way).

Using custom agents for complex tasks

The video covers assigning custom agents to handle more complex or multi-step tasks, so work can be delegated to an agent configured for that kind of workflow.

Connecting MCP servers for external capabilities

GitHub explains how to connect MCP servers (Model Context Protocol servers) to extend Copilot with external tools and capabilities. One example mentioned is connecting Playwright to enable browser testing workflows.

The four levels of Copilot extensibility

The episode closes by summarizing Copilot extensibility as a set of levels, tying together: