Skip to main content

Advanced GitHub Copilot Techniques

This document collects concise, practical techniques for getting better results from GitHub Copilot when writing, refactoring, and reviewing code.

1) Crafting effective prompts​

  • Start with a short goal: "Implement X that does Y for Z".
  • Specify language, framework, and constraints: "TypeScript, Node 18, no external deps."
  • Add desired API surface: function names, signatures, and comments.

Example:

// Implement: validateUser(input)
// - Node 18
// - Accepts {email, password}
// - Returns {valid: boolean, errors?: string[]}

2) Use stepwise and chain-of-thought style prompts​

  • Ask for a plan first, then request implementation.

Example:

// Step 1: outline algorithm in 3 steps
// Step 2: implement the function from the outline

3) Prompt templates for common tasks​

  • Implement feature: provide signature, examples, edge cases.
  • Refactor: provide original code and ask for clearer naming and tests.
  • Tests: provide function contract and ask for unit tests.

Template (implementation):

"Implement <function> with signature <sig>. Requirements: <reqs>. Examples: <cases>."

4) Provide context files and surrounding code​

  • Copy-paste small related files (or key snippets) so Copilot preserves style and types.

5) Use inline comments and TODOs​

  • Write TODOs describing intent and expected behavior; Copilot fills the body.

Example:

// TODO: return items sorted by priority, stable sort
function sortItems(items) { }

6) Verify and iterate​

  • Ask Copilot to explain its changes or produce tests.
  • Run generated code, fix failing tests, then prompt for fixes.

7) Refactoring safely​

  • Request small, focused refactors.
  • Ask for a diff only and unit tests to validate behavior.

Example prompt:

Refactor `processOrder(order)` to extract validation and calculation into helpers. Return a git-style diff and new unit tests.

8) Working with multi-file projects​

  • Provide key files and describe where new code should live.
  • Ask for file paths in outputs so you can paste them directly.

9) Security and dependency guidance​

  • Ask Copilot to list security considerations and unsafe patterns for produced code.

10) Performance and complexity​

  • Specify performance targets (O(n), memory limits) in prompts.

11) Using Copilot for docs and examples​

  • Generate README sections, usage examples, and API docs from code comments or signatures.

12) Prompt examples (quick)​

  • Generate unit tests for <function> using Jest.
  • Add TypeScript types for this JavaScript module.
  • Convert this callback API to Promise-based and update callers.

13) Pro tips​

  • Keep prompts minimal but explicit.
  • Use small reproducible examples when possible.
  • Iterate: ask for tests, then fixes, then improvements.

This file is a living reference—add templates and snippets that match your team's stack and style.