Add AGENTS.md - #484
Conversation
a29b986 to
52dd046
Compare
52dd046 to
3b1bdce
Compare
3b1bdce to
1a01ef7
Compare
rgildein
left a comment
rgildein
left a comment
There was a problem hiding this comment.
Overall LGTM, only one small suggestion.
| Always run against a dedicated [Kind](https://kind.sigs.k8s.io/) cluster, not a real dev/prod cluster. | ||
| The Makefile passes `--context kind-$(KIND_CLUSTER_NAME)` to all kubectl calls automatically — no need to switch context globally. |
There was a problem hiding this comment.
User can use any other k8s, e.g. I can spin VM with other k8s and ask Claude to run test on it. I would suggest to rephrase this to "Run agains Kind cluster only or any other user provided".
This is exactly one of reason why I'm using VM in #491, so Claude run commands in VM and do not touch view any of my context.
There was a problem hiding this comment.
I agree that any k8s suffices (also e.g. minikube). However for the same reason I wouldn't make a suggestion to use a dedicated VM. While I do see a benefit for sandboxing coding agents in a VM, I'm not sure if we should make that recommendation within in this project. I'd rather then completely leave it up to a user/contributor on how to run k8s and how to sandbox their coding agent.
That being said, the default development environment uses Tilt, where we explicitly set some "allowed contexts" in https://github.com/ironcore-dev/network-operator/blob/main/Tiltfile.
1a01ef7 to
02c80be
Compare
02c80be to
1323228
Compare
Add AGENTS.md with coding guidelines for AI agents covering project structure, constraints, workflows, API/controller/provider design, logging conventions, testing style, and references. Symlink CLAUDE.md to AGENTS.md as single source of truth. Signed-off-by: Felix Kästner <felix.kaestner@sap.com>
1323228 to
e54db69
Compare
2 participants
No description provided.