Instrumentations and GenAI: Add copilot review instructions, update agents.md

[lmolkova]

Add review instructions: according to claude it ignores .github instructions when authoring, and github review bot ignores agents.md 🤷‍♀️ maybe it will change some day, but for now:

• agents.md focus on authoring
• copilot instructions on review
• there is some duplication

would appreciate any feedback on how to do it better

[MikeGoldsmith]

Looks good. I wish there was a better way to define these once and link to them from the relevant places but doesn't look like it's possible right now.

[lmolkova]

@Kludex I think you know how to do it right, would you be interested in reviewing this? thanks a lot!

[Kludex]

According to https://docs.github.com/en/copilot/how-tos/configure-custom-instructions/add-repository-instructions#creating-custom-instructions it doesn't ignore AGENTS.md. Why do you say it's ignoring it? 🤔

Is this toggle enabled? https://docs.github.com/en/copilot/how-tos/configure-custom-instructions/add-repository-instructions#enabling-or-disabling-custom-instructions-for-copilot-code-review

[lmolkova]

According to https://docs.github.com/en/copilot/how-tos/configure-custom-instructions/add-repository-instructions#creating-custom-instructions it doesn't ignore AGENTS.md. Why do you say it's ignoring it? 🤔

Is this toggle enabled? https://docs.github.com/en/copilot/how-tos/configure-custom-instructions/add-repository-instructions#enabling-or-disabling-custom-instructions-for-copilot-code-review

@Kludex thank for the feedback!

this is according to copilot docs - https://docs.github.com/en/copilot/reference/custom-instructions-support

The review agent does not seem to be using agents.md, but copilot coding agent does.

Some quotes from me chatting with copilot

Copilot code review reads repository‑wide and path‑specific custom instruction files but does not use AGENTS.md (agent instructions are for cloud agents). See support matrix in Custom instructions support.

Add AGENTS.md when you expect cloud agents to act (detailed build/test/run steps, error workarounds, and explicit “trust this” instructions). This improves cloud agent reliability but is not required for Copilot code review.