GitHub Copilot Pro - 30 Days

Copilot generates generic code that doesn't follow your project's conventions, tech stack, or folder structure.

What to do next: Create a .github/copilot-instructions.md file at the root of your repository. This file is read on every Copilot chat or agent request. Include these key sections:

<code class="language-markdown"># MyApp
E-commerce platform for small businesses.

## Tech stack
- Backend: Flask + SQLAlchemy (Postgres)
- Frontend: Astro + Svelte + TypeScript
- Testing: Pytest, Vitest, Playwright

## Coding guidelines
- Always use type hints in Python
- Use ES modules, not CommonJS
- Unit tests required before PR

## Project structure
- server/routes/ : API endpoints
- client/src/components/ : Svelte components
- scripts/ : Dev and deploy scripts</code></pre>

Bonus: You can ask Copilot agent mode to generate this file by prompting it to "onboard this repository" and inspect the codebase first.

You accept the first Copilot inline suggestion out of habit, but it's often not the best option.

Execution steps: Use keyboard shortcuts to browse alternatives:

• Alt+] (Option+]) — see the next suggestion
• Alt+[ (Option+[) — see the previous suggestion
• Ctrl+Enter — open the completions panel showing up to 10 alternatives at once

This is especially useful for algorithm implementations where Copilot might offer a readable but slow approach first, while suggestion #3 uses a more efficient data structure.

If none of the suggestions work, refine your prompt. Change a vague comment like # sort list to # sort the list of names alphabetically, case-insensitive and trigger Copilot again. Small wording changes can dramatically shift the quality of suggestions.

Copilot suggests code that ignores your existing models, utility functions, or type definitions in other files.

What to do next: Copilot uses open editor tabs ("neighboring tabs") as context. Before working on a feature, open all related files as tabs. For example, if you're writing a new API route that uses your User model, open models.py, validators.py, and your route file side by side.

Also, explicitly write your imports before asking Copilot to generate code:

<code class="language-python">import pandas as pd
from myapp.models import User
from myapp.validators import validate_email

# Now Copilot knows to use these specific imports
# and will generate code using pandas, User model, etc.</code></pre>

Close unrelated files when switching tasks. Copilot's context window is limited, and irrelevant open files can "distract" it, leading to confused or off-target suggestions.

Issue:
Copilot Pro access may look unstable when extension auth and browser account auth are mixed.

What to do:
1. Sign out from duplicate GitHub sessions in unused browsers.
2. Reauthorize Copilot from the same account used in VS Code.
3. Verify billing status first, then test with a small code completion task.

Result:
Copilot response becomes consistent and sign in loops are reduced.

Symptom:
Copilot Pro shows repeated auth popups and does not stay signed in.

Fix path:
1. Sign out from duplicate GitHub sessions in other browsers.
2. Reconnect Copilot extension with the billing owner account.
3. Restart editor and validate with one small completion task.

Expected outcome:
Auth state remains stable and extension features recover.