You have 2 options for this solution, both relying to GitHub Copilot advanced features instead of GitHub Copilot built-in function

GitHub Models or Azure OpenAI API call

The GitHub Models will allow you to use Azure OpenAI via Restful, so consider if your task solving up to 50 files only or intended for more than hundreds to thousands, you probably have another choice except using LLM to automate your task. I'm not quite sure if the feature is GA yet but you may need to join the waitlist. Once you joined you can benefit all requests with no extra charge as usual when you making request to either Azure OpenAI or OpenAI, absolutely it would have some limitation but still sufficient for your task.

GitHub Copilot Workspace (Preview)

The same with GitHub Models, you may need to join waitlist. This coolest feature will help you reading GitHub repository, brainstorming your problem and generating the plan to the code. For example:

I'll pick this public repository, including folder of XML Samples https://github.com/zynksoftware/samples/tree/master/XML%20Samples

Access https://copilot-workspace.githubnext.com/ or same name Extension in VSCode, pick the repo

I will start brainstorming firstly

It will prompt to current behavior and proposed behavior by Copilot

After that, you can generate the plan

Once you feel it's ok to go, click Implementation and wait for it

Finally, create PR for all 36 files are changed, you can check more here https://github.com/zhenyuan0502/samples/pull/1/files

If you can access Copilot Workspace, check more snapshot here https://copilot-workspace.githubnext.com/zynksoftware/samples?s=fc07, from my end the website keeps loading till memory out, so maybe it would be fixed in the future for preview these files online. Whereas the VS Code can download to local with no problem occured