# Host Computer Use Macos

Backend-specific macOS guidance for `computer_use_remote`.

- Repository: https://github.com/agent0ai/agent-zero (19630 stars)
- Category: Design
- License: see repository
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add agent0ai/agent-zero --skill host-computer-use-macos -g -a claude-code -y`
Codex: `npx skills add agent0ai/agent-zero --skill host-computer-use-macos -g -a codex -y`
Cursor: `npx skills add agent0ai/agent-zero --skill host-computer-use-macos -g -a cursor -y`
Gemini CLI: `npx skills add agent0ai/agent-zero --skill host-computer-use-macos -g -a gemini-cli -y`

## Description

Backend-specific macOS guidance for `computer_use_remote`. Load when a tool result reports backend_family/backend_id `macos` or Accessibility-tree features. Covers AX structural targeting, macOS window actions, and screenshot verification.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Host Computer Use - macOS

Use this after `host-computer-use` when the connected A0 CLI reports the macOS computer-use backend.

Do not use this skill for Linux, Windows, Xpra, Docker, or browser-only tasks. If the backend is not macOS or does not advertise Accessibility-tree support, skip AX actions and follow the generic host computer-use rules.

## macOS AX Targeting

macOS backends can advertise structural Accessibility features:

- `accessibility-tree-snapshot`
- `accessibility-structural-targeting`
- `accessibility-element-click`

When these features are present, prefer structural targeting over pixel clicks for named controls such as buttons, menu items, text fields, sheets, alerts, toolbar items, and sidebar rows.

If the backend also advertises `native-window-list`, `window-state`, `element-index-targeting`, or `background-dispatch`, prefer the generic background loop from `host-computer-use`: `list_windows` -> `get_window_state` -> `element_action`. If those features are absent, use the AX snapshot/action flow below.

With `app-scoped-semantic-targeting`, `element_action` also accepts top-level `pid` or `window_id` to scope semantic matching, or `target.app_name` / `target.bundle_id` to select one app. Use identifiers from `list_windows`; app names and bundle IDs match exactly, ignoring case. Without this feature, use `get_window_state` and its top-level `element_index` for another app/window.

For example, replace the app identifier and control title with observed values:

```json
{
  "tool_name": "computer_use_remote",
  "tool_args": {
    "action": "element_action",
    "target": {
      "bundle_id": "com.example.editor",
      "role": "AXButton",
      "title": "Save"
    },
    "operation": "press",
    "dispatch": "background"
  }
}
```

Use `ax_snapshot` to inspect the frontmost app's bounded Accessibility tree:

The rest: https://github.com/agent0ai/agent-zero/blob/HEAD/plugins/_a0_connector/skills/host-computer-use-macos/SKILL.md