> **Up front:** I am a hobbyist, not a professional software developer. This
> report was researched and written with the help of an AI assistant while I
> was building a third-party shell plugin. I have tried to make it accurate
> and to include the evidence behind every claim, but please read the
> technical details with that in mind. If anything here is wrong or unhelpful,
> tell me and I will correct it or close it.

## What happens

The Omarchy shell sets no accessibility roles or names on any of its surfaces.
With an assistive technology running and Qt's AT-SPI bridge active, the bar,
application menu, clipboard picker, emoji picker, notifications, OSD, lock
screen and every panel publish an item tree with no roles and no names, so a
screen-reader user cannot identify or operate any part of the shell.

## What I expected

Interactive surfaces carry `Accessible.role` and `Accessible.name`, which Qt
Quick supports out of the box through the `Accessible` attached property, so
assistive technology can name and operate them.

## How to reproduce

```
grep -rn 'Accessible\.' /usr/share/omarchy/shell
```

No output, exit 1. That is across 102 `.qml` files and 34,914 lines.

## Evidence

The UI kit is hand-built, which is a perfectly reasonable choice, but it means
no component inherits a default role from Qt Quick Controls. Of the 33
components in `/usr/share/omarchy/shell/Ui/`, only two have a Controls type as
their root:

| Component | Root type |
|---|---|
| `TextField.qml` | `TextField` (Controls) |
| `PanelToolTip.qml` | `ToolTip` (Controls) |
| `Button.qml` | `BorderSurface` |
| `Dropdown.qml`, `MultiSelect.qml`, `SearchableDropdown.qml` | `Item` |
| `NumberField.qml` | `Column` |

Five of these import `QtQuick.Controls` for internal use but do not derive from
it, so they get no default role. In particular `Ui/Button.qml` is a
`BorderSurface` plus a MouseArea, which means clickable things are not buttons
to assistive technology.

`Ui/ConfirmDialog.qml` builds its Cancel and confirm buttons as a bare
`BorderSurface` containing a `Text`, with no accessibility properties. That one
matters beyond the shell: a third-party plugin that uses `ConfirmDialog` cannot
annotate those buttons from outside, so the plugin cannot fix it either.

## Why I am suggesting it be fixed at the source

A plugin can label its own surfaces and nothing else. Roles and names on the
shell's own components, and especially on the shared `Ui/` kit where one change
covers every consumer, would make every plugin's surfaces reachable at once.

## One important caveat, so nobody chases the wrong thing first

While investigating this I found that **nothing a Quickshell window declares
reaches AT-SPI at all on my machine**, so adding roles and names here would not
be observable until that is resolved. Measured with the bridge active, checking
window mapping from the compositor, and selecting the bus connection by
`ToolkitName` rather than by process name:

| What was run | Window it creates | Children published |
|---|---|---|
| `qml probe.qml` (plain Qt) | xdg-toplevel | 1 |
| `quickshell` with a `PanelWindow` | layer surface | 0 |
| `quickshell` with a `FloatingWindow` | xdg-toplevel | 0 |

So Qt's bridge works fine on this system; no Quickshell window publishes a
tree, and it is not specific to layer-shell surfaces. I am reporting that
separately. I mention it here only so this issue is not judged by testing it
before that is fixed.

## A related gotcha for anyone debugging accessibility on Omarchy

`envs.lua` sets `QT_QPA_PLATFORMTHEME=gtk3` for the session. With the bridge
active, the shell process then has **two** connections on the accessibility
bus, both reporting the process name `quickshell`:

| Connection | ToolkitName | Version | ChildCount |
|---|---|---|---|
| first listed | `gtk` | 3.24.52 | 0 |
| second | `Qt` | 6.11.2 | 0 |

The GTK one is empty and is listed first, so anyone diagnosing this who selects
by name rather than by `ToolkitName` will read the wrong connection and draw
the wrong conclusion. It cost me a day. Not necessarily a bug, but worth
knowing about.

## System details

Arch Linux, Hyprland session. quickshell 0.3.1-1, qt6-base 6.11.2-3,
at-spi2-core 2.60.6-1. Shell launched by
`/usr/share/omarchy/bin/omarchy-launch-shell` as
`quickshell -n -p /usr/share/omarchy/shell`. No screen reader installed, so
everything above was observed through the AT-SPI bus directly rather than
through a reader.