Add copilot instructions for issues and reviews

[datejada]

@abelsiqueira @gnawin I have created this file to start assigning the good-first-issues to Copilot. We now have Copilot licenses at TNO, so we can start using them more to help with minor issues and possibly initial reviews.

I followed the instructions here: https://docs.github.com/en/copilot/how-tos/configure-custom-instructions/add-repository-instructions

I also took inspiration from here: https://github.com/NREL-Sienna/InfrastructureOptimizationModels.jl/blob/main/.claude/Sienna.md

Please feel free to add/modify/delete the instructions as you consider best.

Thanks!

Related issues

There is no related issue.

Checklist

• I am following the contributing guidelines
• [N.A.] Tests are passing
• Lint workflow is passing
• [N.A.] Docs were updated and workflow is passing

[abelsiqueira]

I like this, great idea. I have some general comments before making changes to the content itself.

Maybe we should change it to AGENTS.md. The downside is that it is not automatically read by all agents yet - the website says it works for "Coding Agent from GitHub Copilot" - so would need to either:

1. Add only AGENTS.md and hope that Claude improves their agent to read AGENTS.md automatically, or that people using Claude add "ALWAYS READ AGENTS.md" to their global config.
2. Add CLAUDE.md saying "READ AGENTS.md" as well as AGENTS.md. Possibly one for every new "rogue" agent.

I prefer option 1.

The other thing on my mind is whether we should split the instructions into multiple files with task-specific info, such as writing-code.md, commit-and-pr.md, etc. - we can ask the agent to come up with the subdivisions 😉.

Claude also has some additional info that we can merge here as well. I can ask it to make a pr with them, after we discuss the above.

[datejada]

@abelsiqueira thanks, I prefer option 1 as well. It should be fine for people with other agents to point to these instructions in their prompts or configurations.

Regarding the different files, I am not sure whether it is beneficial for the Agent to split the file, or if it is more for the developers' organization. Do you have any intuition about it? I am okay with both options 😉

[abelsiqueira]

Some people split it to avoid adding unnecessary context to the agent, but I don't have any real intuition on how much it matters at the moment. We can start with a single file and split later, if we think it's useful.

[datejada]

@abelsiqueira updated to AGENTS.md 😉 Please check the content now and don't hesitate to add/modify/delete the instructions as you consider best.

[abelsiqueira]

Thanks @datejada. I checked this and used claude to compare with CLAUDE.md from BestieTemplate.jl and with the developer documentation. I selected what I think were the most relevant modifications and made the suggestions below.

I don't know how strict you want to be here. Should we focus on getting something good to start and update it later, or do you want to refine as much as possible for a few days? I focused on the first.

[datejada]

@abelsiqueira thanks for the exhaustive review 🤓 it has improved a lot. I will ask the Sienna people to get more insights from their side. I liked all your suggestions and implemented them in the latest commit. Please check if it okay now and don't hesitate to comment if you come up with something new 😃

EDIT: I think it is better focus on getting something good to start and update it later

I will also wait for @gnawin comments before merging 😉