Signed-off-by is already a custom/formality that is surely cargo-culted by many first-time/infrequent contributors. It has an air of "the plans were on display in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying 'Beware of the Leopard.'" There's no way to assert that every contributor has read a random document declaring what that line means in kernel parlance.
I recently made a kernel contribution. Another contributor took issue with my patch and used it as the impetus for a larger refactor. The refactor was primarily done by a third contributor, but the original objector was strangely insistent on getting the "author" credit. They added our names at the bottom in "Co-developed-by" and "Signed-off-by" tags. The final submission included bits I hadn't seen before. I would have polished it more if I had.
I'm not raising a stink about it because I want the feature to land - it's the whole reason I submitted the first patch. And since it's a refactor of a patch I initially submitted (and "Signed-off-by,") you can make the argument that I signed off on the parts of my code that were incorporated.
But so far as I can tell, there's nothing keeping you from adding "Co-developed-by" and "Signed-off-by Jim-Bob Someguy" to the bottom of your submission. Maybe a lawyer would eventually be mad at you if Jim-Bob said he didn't sign off.
There's no magic pixie dust that gives those incantations legal standing, and nothing that keeps LLMs from adding them unless the LLMs internalize the new AI guidance.
The way you describe it, the developers all did the right thing. You contributed something to the patch, and even if it wasn't in your preferred final form (and it's basically never going to be for a kernel contribution of any significance), you were correctly credited.
If you didn't want to be credited you should have said.
Signed-off-by probably has some legal weight. When you add that to code you are making a clear statement about the origins of the code and that you have legal authority to contribute it - for example, that you asked your company for permission if needed. As far as I know none of this has been tested in court, but it seems reasonable to assume it might be one day.
The problem is they've got a doc that declares "when you say balacalaboozy, you're declaring that a specific set of legal conditions is met. You must say balacalaboozy to proceed."
Newcomers see everyone saying balacalaboozy, so they say it to. It doesn't mean that they have read or agree to the doc that declared its meaning.
LLMs are the world's most sophisticated copycats. Surely they too will parrot balacalaboozy, unless their training is updated to include, understand, and consistently follow these new guidelines.
You can write in AGENTS.md to ask the user for explicit sign off and to explain the document to the user.
> You contributed something to the patch, and even if it wasn't in your preferred final form (and it's basically never going to be for a kernel contribution of any significance), you were correctly credited.
I don't see how the "signed-off-by" attestation constitutes correct credit here. It's claiming that GP saw the final result and approved of it, which is apparently false.
Signed-off-by is a chain. The second person asserts that they delegate to the first person for the parts contributed by the first, and signs off on the ones that were contributed personally.
Hypothetically in court you'd go to the last, ask "did you write this" and only if not go up.