Re: [PATCH v2] docs/process: document AI-assisted patch tags

[Jan Beulich <jbeulich@xxxxxxxx>]

On 19.05.2026 16:20, Cody Zuschlag wrote:
> --- a/docs/process/sending-patches.pandoc
> +++ b/docs/process/sending-patches.pandoc
> @@ -195,6 +195,25 @@ E.g.:
>  
>      Requested-by: Jane Doe <jane.doe@xxxxxxxxxxx>
>  
> +### Assisted-by:
> +
> +This optional tag can be used when AI-assisted tooling contributed to the
> +preparation of a patch.

Noticed only while reading your other email, plus checking the Linux
wording: s/can/should/ at the very least? (Personally I'd make it a
requirement.)

> It is in the format:
> +
> +    Assisted-by: <tool-name>[:<model-version>] [tool...]
> +
> +E.g.:
> +
> +    Assisted-by: Claude:claude-3-opus coccinelle sparse
> +    Assisted-by: ChatGPT:gpt-5.5
> +
> +Basic development tools, such as git, compilers, make, and editors, should
> +not be listed.
> +
> +The person submitting the patch remains responsible for reviewing any
> +AI-generated content and ensuring the contribution is correct, appropriate,
> +and compliant with Xen's contribution and licensing requirements.
> +
>  ### Signed-off-by:
>  
>  This mandatory tag specifies the author(s) of a patch (for each author a
> @@ -209,6 +228,9 @@ E.g.:
>  The author must be a natural person (not a team or just a company) and the
>  `Signed-off-by:` tag must include the real name of the author (no pseudonym).
>  
> +AI tools must not add `Signed-off-by:` tags, as only natural
> +persons can certify the `Developer's Certificate of Origin`.

Linux emphasizes "MUST NOT". I think that's advisable for us as well. I
also wonder whether some of their other wording might be worth taking
verbatim, rather than making our own, slightly different variant.

Nit: Wrap differently to make better use of line length?

Jan