Re: [PATCH 1/4] agents: add unified agent coding assistant configuration

[Sasha Levin]

On Sun, Jul 27, 2025 at 10:16:32PM -0700, Kees Cook wrote:
> On Mon, Jul 28, 2025 at 12:43:52AM -0400, Sasha Levin wrote:
> > On Sun, Jul 27, 2025 at 07:37:31PM -0700, Kees Cook wrote:
> > > On Sun, Jul 27, 2025 at 03:57:59PM -0400, Sasha Levin wrote:
> > > > Create a single source of truth for agent instructions in
> > > > Documentation/AI/main.md with symlinks for all major coding
> > > > agents:
> > > > - CLAUDE.md (Claude Code)
> > > > - .github/copilot-instructions.md (GitHub Copilot)
> > > > - .cursorrules (Cursor)
> > > > - .codeium/instructions.md (Codeium)
> > > > - .continue/context.md (Continue)
> > > > - .windsurfrules (Windsurf)
> > > > - .aider.conf.yml (Aider)
> > >
> > > I *really* don't like this. I use the CLAUDE.md file as my instructions
> > > for my agent. I think all of these should be .gitignore entries.
> >
> > Sorry, I might have misunderstood you: how does it play out if we add
> > these to .gitignore?
>
> Then what claude learns about my workflows and preference can be
> correctly stored in CLAUDE.me (which is how claude is designed to work).
> I would think of it like why we don't ship a debian/ package build tree:
> it's going to be different for everyone. And if you look in .gitignore
> you can already see that /debian/ is there. :) These agent files are for
> developer-specific use, and adding them to .gitignore is the right
> approach (at least for Claude and Gemini). Which reminds me, please
> also include GEMINI.md in your list. :)
>
> > The tool will just end replacing whatever we put in there with something
> > customized that doesn't necessarily correspond to what the community
> > will consider a "standard" set of rules for agents?
>
> Right, and then it will always be a git diff delta and cause pain. For
> the agents that are designed to _write_ to their files, then it needs
> to be in .gitignore.

Okay, I'll update it for v2.

--
Thanks,
Sasha