Skip to content

Module 2: Copilot Best Practices

Duration: ~25 min total · 3 lessons

You identified your failures. Now you fix them — by building the structure that prevents them.

Module 2 covers the two practices that shape every Copilot interaction: Instructions and Agents. You'll also experience the difference between Plan Mode and Agent Mode directly, so you know when to reach for each one.

Lessons

#TitleFocusTime
L2.1Instructions That Actually WorkWorkspace + file-specific instructions~7 min
L2.2Agents and Decision PatternsRole-specific agents, when to use them~8 min
L2.3Skills and Prompt FilesReusable prompt workflows (deferred)
L2.4Plan Mode vs. Agent ModeTwo-step feature development~10 min

Note: L2.3 (Skills) is a deferred session — it is not part of the workshop today. A brief reference page is provided for context.

What You'll Build

By the end of Module 2, you will have:

  • A workspace instructions file (.github/copilot-instructions.md) that sets the global context for all Copilot interactions.
  • A file-specific instructions file (.github/instructions/config.instructions.md) that activates only for config files.
  • A Code Reviewer agent (.github/agents/code-reviewer.agent.md) that checks output against your project standards.
  • A Testing Advocate agent (.github/agents/testing-advocate.agent.md) that generates tests with defined coverage goals.
  • One feature built via Plan Mode → Agent Mode — so you've experienced the divide firsthand.

Module Outcome

When the module is done, the failures you mapped in L1.2 under "Requirements gap" and "Technical mismatch" will have a structural fix in place — not a better prompt, but enduring context files that persist across sessions.


← Module 1 Overview · Start with L2.1 →

Spec-Driven Development Workshop — structured AI development with GitHub Copilot