A GitHub account with Copilot access. Copilot Free works for IDE practice; organisation policies and content exclusions need a Business or Enterprise trial
VS Code (or another supported IDE) and working knowledge of at least one programming language and basic Git and pull requests
Week 1 · 3 h
Responsible use, data flow and architecture
Name the risks of generative AI coding tools (insecure code, bias, licensing, outdated APIs) and why every suggestion must be reviewed
Trace how Copilot builds a prompt from open files and context, sends it through the proxy, filters it and post-processes the response
Explain the code suggestion lifecycle, LLM limits such as context window size and stale training data, and who owns the output
Copilot features in the IDE, CLI and github.com (biggest domain)
Use inline suggestions, Copilot Chat with slash commands and participants, Copilot Edits and agent mode, and know the limits of each
Connect an MCP server to agent mode, manage agent sessions and delegate work to sub-agents
Install GitHub Copilot CLI and use it interactively to generate scripts and manage files
Use Copilot code review, pull request summaries, Spaces and Spark, and set review standards with repository custom instructions and reusable prompt files
Configure organisation policies for features and code review, read Copilot audit log events, and manage seats with the REST API
Hands-on: In a sample repo, add a .github/copilot-instructions.md that sets coding standards, use agent mode to add a feature across two files, open a pull request, then request a Copilot code review and generate the PR summary.
Write zero-shot and few-shot prompts, control context with open files and chat variables, and explain how chat history shapes later answers
Use Copilot to refactor legacy code, write documentation, generate sample data and suggest performance and security fixes
Generate unit and integration tests, including edge cases and assertions
Configure content exclusions at repo and organisation level, turn on the public code matching filter, and troubleshoot missing suggestions
Hands-on: Take an untested function, ask Copilot Chat for /tests, then re-prompt with two example tests (few-shot) and compare coverage. Finally add a content exclusion for a secrets folder and confirm Copilot stops using that file as context.