[removal]: Remove AUR from reference and Pipeline

[mathys-lopinto]

What would you like

Hello,

A few days ago, a user filed a deletion request for this package with the following reason:

Duplicate of aur/claude-desktop. This package is an AppImage resulting from a repack of the official .deb as per https://github.com/aaddrick/claude-desktop-debian#installation

The request was approved today by an AUR Package Maintainer and the package has been deleted.

In order to avoid failed pipelines, you need to remove all AUR instructions (and AUR references in the README and other Markdown files).

I still want to thank you for the implementation you made when I asked you for the feature in my first issue.

Mathys,
Former owner of the "claude-desktop-appimage" AUR package.

Use case

---

Existing workarounds

No response

[mathys-lopinto]

@aaddrick, Please take note of this issue.

[github-actions]

Automated draft — AI analysis, not maintainer judgment. This bot won't approve enhancements, prioritize roadmap, or commit timelines. The notes below flag existing surfaces and design questions that may be worth considering before implementation.

Looks like the ask is to permanently remove the AUR publish job and all AUR install references from the CI pipeline and documentation now that claude-desktop-appimage has been deleted from the AUR.

Existing surfaces worth knowing about:

• The update-aur-repo job (lines 868–1001) is the full AUR publish pipeline — already gated off via AUR_PUBLISH_ENABLED != 'false' since 2026-08-01, but the full job body and its SSH/Docker/git-push steps remain and would need removal. (.github/workflows/ci.yml:868-1001)
• Two release-notes generation blocks (primary ~line 297, fallback ~line 355) each embed a ### AUR (Arch Linux) section with yay -S claude-desktop-appimage as an install instruction surfaced on every GitHub Release. (.github/workflows/ci.yml:297-301)
• The ### Using AUR (Arch Linux) section in README.md currently carries an 'Unavailable' notice and kill-switch instructions that would need to be removed or replaced with a permanent removal notice. (README.md:92-103)

Full investigation artifacts attached to the triage workflow run.

---

⚠ This issue was edited after triage began. The
draft above reflects the body as of
2026-08-02T06:26:02Z; the current body may differ.
See input_snapshot.json in the workflow run for
what the bot actually read.

[aaddrick]

Hey!

Already blocked the pipeline, updated the README, and sent this out in response yesterday. Just waiting for my response to clear moderation.

---

I missed the discussion window on PRQ#85209. The request sat without replies
and bertptrs actioned it on 1 Aug, which was the right call given nobody spoke
up. That's on me. Bert pointed me at this list as the normal place to respond,
so I'm making the case here that I should have made there.

I'm asking for claude-desktop-appimage to be reinstated. The stated deletion
reason was: "Duplicate of aur/claude-desktop. This package is an appimage

result of a repack of the official .deb as per
https://github.com/aaddrick/claude-desktop-debian#installation". I want to take
both halves of that, starting with the dates.

The wiki rule is about creating a package that duplicates one already there:
"Check the AUR if the package already exists. If it is currently maintained,
changes can be submitted in a comment for the maintainer's attention. ... Do
not create duplicate packages." The chronology runs the other way here.

• claude-desktop-appimage, first commit 2025-09-16 (mathys-lopinto, "Initial
  commit: claude-desktop-appimage")
• aur/claude-desktop, first submitted 2026-07-05

That's a gap of roughly nine and a half months. I'm not suggesting
kevindiaz314 did anything wrong. Anthropic shipped an official Linux .deb at
the end of June and packaging it was a reasonable thing to do. But whatever the
right resolution is, the older package can't be the duplicate.

It was also live, not abandoned. 119 commits, and the last one was 2026-07-25,
Update to v3.2.1+claude1.24012.9, pushed by CI three days before the request
was filed.

On the suffix itself: the AUR currently carries 399 packages ending in
-appimage, and coexisting with a native or -bin package for the same
upstream is the ordinary case. Three you can check in a minute:

• 86box and 86box-appimage, both maintained by Lili1228
• altus and altus-appimage, different maintainers
• cursor-bin and cursor-appimage

The wiki already has language for the shape of this. It's written about the
official binary repos rather than AUR-vs-AUR, but the principle is the one in
play: "Exception to this strict rule may only be packages having extra features
enabled and/or patches in comparison to the official ones. In such an occasion
the pkgname should be different to express that difference." The suffix is the
differing name. The patches are the difference.

The reason links my own README's Installation section, and it wasn't wrong
about what the page said. That section claimed two patches and didn't mention
virtiofsd at all. Five have been live for a while. It's corrected on main now,
each one listed against the upstream gap it covers.

Here's the one an Arch user runs into directly. Upstream resolves virtiofsd
from two absolute paths, /usr/libexec/virtiofsd and /usr/bin/virtiofsd, and
only falls back to its own bundled copy when /etc/os-release says Ubuntu
22.x. Arch installs virtiofsd at /usr/lib/virtiofsd. So the path resolves
null and Cowork reports "Cowork requires QEMU" on a machine with a complete,
working KVM stack. Our patch drops the os-release gate on the bundled fallback,
and system paths stay preferred. It's filed upstream. Until it lands, that's a
defect an Arch user hits on the native package and doesn't hit on this one.

The other four, briefly:

• org-plugins: upstream's platform switch has cases for darwin and win32, and
  the default returns null. The feature is silently dead on Linux. The patch
  adds a linux case resolving /etc/claude/org-plugins. Filed upstream.
• tray-icon selection: upstream's desktop-environment detector returns kde,
  gnome, or other. Cinnamon lands in "other" and gets a dark glyph on a dark
  panel. The patch threads a launcher-set env var into upstream's own selector
  and doesn't replace its icons. Filed upstream.
• quick-window: adds blur() before hide() on the pop-up window so the main
  window comes back after a quick-entry submit. Works around an Electron focus
  bug on KDE.
• cowork-bwrap: every branch is behind an opt-in env var, so an unflagged
  launch runs upstream's path unchanged. I'd rather say that plainly than count
  it as a real difference.

Worth knowing about the shape of that list: project policy is that a patch has
to justify itself against the official bytes or get deleted, and most of the
old suite did get deleted when Anthropic shipped official Linux builds. These
five are what survived that pass. The launcher around them handles opt-in
native Wayland (which fixes the Quick Entry global shortcut on GNOME),
GPU-fatal crash recovery, stale SingletonLock and Cowork socket cleanup,
rotating user-config backups, and a --doctor mode. Source is all open:
https://github.com/aaddrick/claude-desktop-debian

The fair question, and the one I'd ask from a PM's seat, is what an AppImage
adds on Arch where native packaging is cheap. For someone on plain Arch with a
working PKGBUILD, probably nothing. The difference shows up elsewhere. The file
is relocatable, it runs without root, and it's the practical option on
immutable and atomic setups and inside containers and toolboxes where
installing a package isn't how software gets there. Those are the users who ask
me for it.

One thing I can fix regardless. The PKGBUILD already declares
provides=('claude-desktop') and conflicts=('claude-desktop'), but the
pkgdesc reads "Claude Desktop for Linux - AppImage version from
aaddrick/claude-desktop-debian", which doesn't tell you it's a patched
unofficial build. I'll reword it so that's legible on the package page. If the
-appimage suffix specifically is the objection, I'll take a rename to
whatever you'd prefer.

If reinstatement still isn't the right outcome, I'd like to know which part of
the guidelines it falls on, so I can update the README and stop sending people
to a dead link.

aaddrick

[mathys-lopinto]

@aaddrick I was the one who initially created the AUR package "claude-desktop-appimage", then made the initial commit for adding claude-desktop-appimage to AUR with the pipeline (and was the person who manually updated the AUR package before the commit that added the automation). I'm not the AUR Package Maintainer who deleted it (and I don't understand why it was deleted because it's an improved repack). I'm on holiday and I've also missed the discussion to argue that it's not a duplication, sorry.

[aaddrick]

No worries at all!

Thanks for taking care of that originally. I honestly appreciate it.

I got the original email too, but missed it. Should be fine once my response to the mailing list clears moderation.