Skip to content

Improve copilot instructions#7762

Open
achamayou wants to merge 9 commits intomainfrom
improve_copilot_instructions
Open

Improve copilot instructions#7762
achamayou wants to merge 9 commits intomainfrom
improve_copilot_instructions

Conversation

@achamayou
Copy link
Member

@achamayou achamayou commented Mar 24, 2026

Goals:

  1. Expand instruction coverage of the repo, but:
  2. Split more into skills
  3. Condense top-level instructions

- This document provides guidance for AI coding and review agents working in the CCF (Confidential Consortium Framework) repository
- **CCF** is an open-source framework for building secure, highly available, and performant applications focused on multi-party compute and data. It's designed for confidential, distributed systems running on secure hardware.

## Architecture
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think pointing the robot at the docs here is likely more effective, from a 'go read these whenever you need to learn about a thing'.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It could be, but this GitHub blog suggests it's worthwhile: https://github.blog/ai-and-ml/github-copilot/5-tips-for-writing-better-custom-instructions-for-copilot/

Perhaps it is good to have a relatively context-efficient dispatch early on? I don't know for sure.

@cjen1-msft
Copy link
Contributor

As a generic review. I think we should restrict these kind of files to things that the robot often gets wrong (see norms and standards for how to build and run tests).
It seems to be pretty good at getting context for most cases, so unless its often entirely misunderstanding our architecture I'm not sure what the value is.

@achamayou
Copy link
Member Author

As a generic review. I think we should restrict these kind of files to things that the robot often gets wrong (see norms and standards for how to build and run tests). It seems to be pretty good at getting context for most cases, so unless its often entirely misunderstanding our architecture I'm not sure what the value is.

That was also my hunch, but then I read https://github.blog/ai-and-ml/github-copilot/5-tips-for-writing-better-custom-instructions-for-copilot/. I think it's an early dispatch optimisation that leaves more context for the task rather than a "gets things wrong measurably" situation. Same for pushing stuff to skills.

I would be better if there was a way to measure this properly, all suggestions welcome.

@achamayou achamayou marked this pull request as ready for review March 25, 2026 18:31
@achamayou achamayou requested a review from a team as a code owner March 25, 2026 18:31
Copilot AI review requested due to automatic review settings March 25, 2026 18:31
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR expands and restructures Copilot guidance by adding a dedicated Testing skill doc and enriching the main repository instructions with architectural and contributor workflow details.

Changes:

  • Added a new “Testing” skill document describing how to run unit/e2e/SDK tests, labels, coverage, and e2e test patterns.
  • Expanded .github/copilot-instructions.md with architecture details, key directories, build/test/lint pointers, and C++ conventions (naming, JSON, endpoints, logging, KV).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File Description
.github/skills/testing.md New testing skill guide (how to run tests, coverage, and e2e infra/patterns).
.github/copilot-instructions.md Adds architecture overview and links to skill docs; consolidates build/test/lint sections and C++ conventions.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants