standardize using `Co-developed-by` to indicate tools used during development

[dims]

context: https://lwn.net/Articles/1032612/

We currently have no visibility of what AI tools were used in a PR by pull request authors.

For authors, who really want to share that information (so reviewers/approvers can scrutinize a PR a bit more), there is no standard way for them to indicate as such.

for example see kubernetes/kubernetes#133659

Just like we use Co-Authored-By to indicate a human participant, the proposal here is to use Co-developed-by for tools.

It is obviously on the PR authors to do the right thing, it's extra metadata that may help us at some point to see adoption of these tools and if some specific tools work better (or not!).

What do folks think?

[dims]

cc @BenTheElder @thockin @aojea @soltysh

[kannon92]

It would be nice to use whatever the bots use when generating commit messages.

I guess we have a AGENTS.md file though so we could suggest to AI to use whatever standard we decide.

In that PR, I did the following:

A. Asked Claude Code to generate a plan of fixing the panic
B. Claude figured out the issue and adding the code
C. Asked Claude to generate the commit messages
D. And I pushed and opened up the PR.

[dims]

@kannon92 i would differ a bit. i'd coax the agents to generate what we wish. If it is not there then it's the responsibility of the PR author to do so.

[BenTheElder]

For authors, who really want to share that information (so reviewers/approvers can scrutinize a PR a bit more), there is no standard way for them to indicate as such.

https://www.kubernetes.dev/docs/guide/pull-requests/#large-or-automatic-edits

To make it easier for reviewers to handle such Pull Requests, please explain how it was generated in the “Special notes for your reviewer” section of the Pull Request description. Reviewers may then be able to reproduce those steps (search/replace, linters) or can start the review with the right expectations (LLMs). Also consider the section about splitting up Pull Requests above.

Ref kubernetes/steering#291

[ffromani]

(I'm talking in general, the change kubernetes/kubernetes#133659 brought as example is straightforwarded and linear and poses no real concern)

My 2c: as approver and reviewer (in this order), I would greatly appreciate a clear declaration of co-development using AI at commit level. My understanding is this complements and enhances our existing guidelines.
I must say I'm still wrestling with the understanding of copyright attribution in the case of large LLM-generated edit.
I would appreciate, for example, clear documentation of how the prompting was done in order to be enabled to assess how much human effort a given contribution had: trivial prompting vs some back and fort vs heavy human intervention.

My very own idealistic scenario is every commit comes with AI-Attribution tag, being the tag Abbreviated Statement in the https://aiattribution.github.io/ format. Should a commit lack AI-Attribution tag, it will be considered 100% human-made.
I fully acknowledge this can be excessive, I'm just reporting as the ideal scenario as I see it now. I'm not advocating for the project to adopt it (yet :) ).

I'm generally in favor of gen-AI use (to be precise I tend to swing between "neutral" to "mildly in favor") but I'm trying to avoid the project to find itself in need to urgently rewrite or revert a piece of code because of copyright issues.

[soltysh]

/committee steering