Re: How to fix problems with the sashiko review model

> On Jul 22, 2026, at 8:10=E2=80=AFAM, James Bottomley <James.Bottomley@hans=
enpartnership.com> wrote:
>=20
> =EF=BB=BFOn Wed, 2026-07-22 at 06:43 -0700, Guenter Roeck wrote:
> [...]
>> For Sashiko, fork and clone
>> https://github.com/sashiko-dev/sashiko.git,
>> submit your changes into the clone, and send a pull request. Or at
>> least that is how I do it.
>>=20
>> In the Sashiko source, per-subsystem prompts are located in directory
>> third_party/prompts/kernel/subsystem.
>=20
> OK, now I'm confused.  I thought
>=20
> https://github.com/masoncl/review-prompts
>=20
> was the source of truth for this, but that project seems to have been
> copied into sashiko rather than pulled in as a module.  If you want to
> improve the review prompts do you now have to submit to both (which is
> getting to be a bit wasteful)?

It=E2=80=99s a bit messy now, but Chris and me are working on something more=
 sane.
As of now, Sashiko is using per-subsystem prompts from Chris=E2=80=99s repos=
itory
with some changes around more generic prompts. I periodically pull updates
from Chris=E2=80=99s repository. Also several folks sent prompts changes dir=
ectly to
Sashiko repository, Idk if Chris is merging them back, probably not.

I also experiment with auto-generating prompts based on human feedback
to sashiko reviews, initial results are quite promising.

So as of now, I=E2=80=99d say send a pull request to Chris or me directly, w=
e=E2=80=99ll figure out
a better model going forward.

The hard part is testing: it=E2=80=99s quite easy to confuse a LLM with an u=
nlucky prompt change.
We need a stronger guidance and maybe LLM review for prompts too (sashiko fo=
r sashiko?!).
E.g. Sashiko implements a much smaller set of tools (intentionally) than a g=
eneric coding agent,
so if a prompt contains something like =E2=80=9Crun checkpatch.pl=E2=80=9D o=
r =E2=80=9Csearch lore for =E2=80=A6=E2=80=9D it might be
very confusing to the LLM. When I=E2=80=99m pulling significant prompt updat=
es from Chris=E2=80=99s repository,
I run the full benchmark to make sure we don=E2=80=99t get regressions.

Thanks!=