learning-business-central — posts/copilot-skills-prompts-instructions-en.html
index.html
copilot-skills-prompts-instructions-en.html
posts / vs-code

How to Use Your GitHub Copilot Skills, Prompts and Instructions Across Every VS Code Project

Roberto Corella September 15, 2026 GitHub Copilot · VS Code · Agentic Development

If you work with GitHub Copilot in agent mode across several Business Central projects, you've probably run into this: every time you open a new project, your skills, your prompts and your instructions simply aren't there. You end up copying the whole .github folder over again, and from that point on you're maintaining several copies that slowly drift apart.

This article covers how to centralize all of that material in a single folder on your machine, and how to tell Visual Studio Code to use it in any project you open, without copying anything.

The problem: your setup lives inside each project

When you put together a Copilot workflow — as we did with BC Quality or with the ALDC framework — the usual approach is to create a .github folder inside the project, with its own subfolders for instructions, prompts, agents and skills. It works well, but it has a clear limit: that configuration only exists in that one project.

Start a new project and there's nothing there. Start work for a different client and there's nothing there either. Improve a prompt in one project and none of the others find out about it.

When you work inside a single client's repository, keeping it at the repository level makes complete sense. But if you're an independent consultant or developer jumping between clients, you don't have one fixed repository to leave your way of working in. What you actually need is for your framework to travel with you, not with the project.

The idea: a central folder in your user profile

The solution comes down to two steps:

  1. Create a folder on your machine with all of your Copilot configuration.
  2. Tell VS Code, at the user settings level, where that folder is.

Because you set this at the user level — not the workspace level — the configuration applies to any folder or project you open in VS Code.

Step 1: create the framework folder

Inside your Windows user folder, create a structure like this one. In my case I called it ALDC Framework, inside a folder I use for AL development:

C:\Users\Roberto\DesarrolloAL\ALDC Framework\
├── instructions\
├── prompts\
└── agent-skills\

Move into this folder whatever you used to copy from project to project.

Step 2: point VS Code to that folder

Open the command palette (Ctrl+Shift+P), search for Preferences: Open User Settings (JSON), and add the location for each type of file:

{
  "chat.instructionsFilesLocations": {
    "C:\\Users\\Roberto\\DesarrolloAL\\ALDC Framework\\instructions": true
  },
  "chat.promptFilesLocations": {
    "C:\\Users\\Roberto\\DesarrolloAL\\ALDC Framework\\prompts": true
  },
  "chat.agentSkillsLocations": {
    "C:\\Users\\Roberto\\DesarrolloAL\\ALDC Framework\\agent-skills": true
  }
}

Keep in mind settings.json is a plain text file: you can edit it, add more paths, or turn off a location by setting its value to false. The exact names of these settings can vary between VS Code versions, so if one of them doesn't show up, search the Settings UI filtering by "locations" under Chat.

Step 3: check that it works

Open a project that has no .github folder at all. In the Copilot chat, type / and you'll see your prompts already available, things like create functional specification or create technical specification.

You can also confirm it from Configure Prompt Files in the chat itself: it lists the prompts from your central folder alongside any that belong to the project. From there, just run the prompt and give it the details of your requirement so it can generate, for example, the functional specification.

Why this makes your framework more powerful

Centralizing isn't just about convenience. It has a direct effect on the quality of what the agent produces, and it connects with one of the core ideas of agentic development:

Don't fix what the agent produced — fix the skill or prompt that produced it.

If the agent gets something wrong and you fix its output by hand, that fix only helps once: next time it will fail the same way again. If instead you adjust the skill, the instruction, or the prompt, the mistake disappears for good.

With a central folder, that improvement propagates automatically to every project you work on. Every correction accumulates and the framework gets more reliable the more you use it, instead of being scattered across separate copies spread over every client.

Practical recommendation

If you found this content useful, you can buy me a coffee — it helps me keep writing and recording. Thank you!
main ← back to index