Copilot code review now supports AGENTS.md files. 💡 Here's how to customize for more context-aware reviews. ▶️

@github Works in the AGENTS.md demo repo, but let's see how it handles a legacy 2k-line schema 😬

@github I already have Codex.md and Gemini.md look at Agents.md for the instructions. Haven't played with Copilot, so couldn't test that.

@github 👀 Wait til it reads our 500-line legacy workflow and just quits the PR review.

@github Yes. This is the right direction for reviews. A repo has house rules. If the review agent does not read them, you just get generic feedback faster.

@github having a central file for instructions is way better than trying to prompt every single review individually.

Want to see AI Factories in action? Watch ServeTheHome’s latest video breakdown to see how Supermicro and NVIDIA deliver turnkey AI infrastructure at scale.

@github How do AGENTS.md files compare to skills and copilot-instructions.md in practice, same effectiveness or different use cases?

@github Added an AGENTS.md and Copilot still complained about my variable names. 😬

@github Can't wait for AGENTS.md to argue with my .github/CODEOWNERS in the PR comments. 🤖

@github Works in the greenfield demo repo, but let's see how AGENTS.md handles a 10-year-old legacy monolith. 🤔

@github ok but does it actually enforce style conventions or just check for a file existing

@github Subdirectory too or only root level?

@github Ok but it's the end of the month and nobody here has any credits left

@github It doesn’t feel right to bloat that file with all code review instructions 🤔

Our delivery comes with a genius. Go learns what you like, builds your cart in seconds, and gets it all to your door in as fast as 15 minutes.

@github AGENTS.md is most useful when it turns repo knowledge into review policy, not just extra context: invariants, risky files, non-goals, test gates, migration rules, rollback paths. The hard part is keeping it current; stale instructions create confident but wrong reviews.

@github Only if the agent remembers to update AGENTS.md or CLAUDE.md every time the project changes — or if that behavior is built into the workflow. Either way, in complex projects, I still find misalignment over time and need manual reviews.

@github not sure 'context-aware' holds in practice i used CLAUDE.md on my flutter project and the reviews still missed obvious cross-file state bugs. does AGENTS.md actually change what gets flagged, or just how the feedback is phrased?

@github x.com/m0nle0z/status… yeahs

@github I think you guys should focus on the upcoming models and make this feature more affordable.

AGENTS.md is indeed a great idea - a project guide specifically for AI Agents. By incorporating norms and context, the quality review by Copilot should be able to be further enhanced. Over 60,000 projects have already been using it, and I'm planning to give it a try myself. Has anyone already started working on it? Share your experiences!

@github This is actually useful. Context-aware reviews can save a lot of back and forth.

@github Works in the demo repo, but let's see how many AGENTS.md files it takes to break my CI loop. 👀

@github anything that keeps copilot from suggesting code from 2018 is welcome

@github Who is using copilot?

@github ok AGENTS.md in review is actually useful

@github Yeah. Took a year long for QA

@github only now？