Skip to content

exercise-09-custom-instructions.md

Latest commit

 

History

History
92 lines (61 loc) · 3.72 KB

exercise-09-custom-instructions.md

File metadata and controls

92 lines (61 loc) · 3.72 KB

Exercise 09 — Create Custom Instructions for Your Language

Duration: 3 minutes
Copilot Feature: Workspace Instructions (copilot-instructions.md)
Goal: Set up always-on coding standards that Copilot follows in every conversation.

Background

A copilot-instructions.md file in .github/ is loaded automatically into every Copilot Chat session. Unlike custom agents (which you manually select), instructions are always active — they set the baseline for how Copilot writes code for your project.

Think of it as your team's coding standards, permanently injected into Copilot's context.

Step 1 — Decide Your Language & Framework

Choose the tech stack you want to build the ITMS application with:

Option Language Framework Database
A TypeScript / Node.js Express or Fastify PostgreSQL or SQL of your choice
B Python FastAPI or Django REST PostgreSQL or SQL of your choice
C Java Spring Boot PostgreSQL or Oracle or SQL of your choice
D C# / .NET ASP.NET Core PostgreSQL or SQL Server or SQL of your choice

The rest of the workshop will use whichever you pick. Option A (TypeScript/Node.js) is the default reference if you're unsure.

Step 2 — Generate Your Instructions File

Switch back to the default Copilot agent (not a custom agent).

Copy and paste this prompt, replacing [YOUR CHOICE] with your selection:

I am building a REST API application using [YOUR CHOICE — e.g., TypeScript with Express and YOUR CHOICE DATABASE].

Create a .github/copilot-instructions.md file with workspace-wide coding standards for this project.

Include instructions covering:
1. Language & framework conventions (naming, file structure, module organization)
2. API design rules: versioning (/api/v1/), consistent response envelope { success, data, error, meta }
3. Error handling: custom error classes, centralized error middleware, proper HTTP status codes
4. Security: always validate input with a schema library, never trust user input, use parameterized queries only
5. Database: use migrations for schema changes, use a query builder or ORM — no raw string SQL queries
6. Testing: every new function must have a unit test, integration tests for all API endpoints
7. Logging: structured JSON logs with request ID, user ID, and operation name
8. Code style: no console.log in production code, no TODO comments without ticket numbers, functions under 30 lines
9. Documentation: JSDoc/docstring for all exported functions
10. Git: conventional commits (feat:, fix:, docs:, test:, chore:)

Step 3 — Verify the File

Open .github/copilot-instructions.md and check:

  • File exists at .github/copilot-instructions.md
  • Contains your chosen language/framework
  • Includes rules about input validation and parameterized queries
  • Has error handling conventions
  • Mentions the /api/v1/ URL convention

Step 4 — Test That Instructions Are Active

Close and reopen Copilot Chat. Send this test prompt in the local agent:

Write a simple function that takes a user ID and returns user details from the database.

Observe that Copilot:

  • Uses your chosen language and SQL database
  • Uses parameterized queries (not string concatenation)
  • Includes error handling
  • Follows the response envelope format you defined

Key Takeaway

Instructions set the floor for code quality. Once written, you never have to repeat "use TypeScript", "use parameterized queries", or "follow REST conventions" — Copilot applies these automatically to every piece of code it generates in this workspace.

Next: Exercise 10 — Plan Mode for Implementation