Windows App Playbooks

Explorer, Notepad, Photos, Edge via computer_use.

Skill metadata

Source	Bundled (installed by default)
Path	skills/desktop/windows-app-playbooks
Version	1.0.0
Author	VibeOS
License	MIT
Platforms	windows
Tags	windows, desktop, computer-use, explorer, notepad, photos, edge
Related skills	windows-desktop, choose-interaction-surface, computer-use, desktop-operate

Reference: full SKILL.md

信息

The following is the complete skill definition that VibeOS loads when this skill is triggered. This is what the agent sees as instructions when the skill is active.

Windows App Playbooks

GUI procedures for common Windows apps via computer_use. Prefer terminal + file tools for known paths; prefer MCP for GitHub/Notion. Long Operate jobs → desktop-operate (≤2 retries + scorecard).

When to Use

• Browse folders in File Explorer when the user wants a visible UI path
• Draft text in Notepad
• Open/export an image in Photos (light Operate)
• Open a URL or fill a simple form in Edge when browser toolset is off

Prerequisites

• computer_use enabled (vibeos computer-use doctor)
• Load windows-desktop for capture / clipboard basics

How to Run

1. Confirm native Windows app → computer_use (not browser tools unless Edge web content).
2. Capture with SoM/UIA; click by element index.
3. Re-capture after navigation.

Quick Reference

App	Prefer instead when possible
Explorer	terminal / file tools for known paths
Notepad	write_file / editor in workspace
Photos	read_file / vision_analyze for inspect-only
Edge web UI	browser_* toolset

Procedure

Explorer — open a folder

1. Capture Explorer (or start it).
2. Use address bar: click it, type the path, press Enter.
3. Confirm the content pane shows expected items.

Notepad — draft text

1. Capture Notepad; focus the edit control.
2. Type or paste content.
3. Save only if the user asked (Save As dialog → capture again).

Photos — open and export

1. Prefer opening a known path via Explorer or terminal, then capture Photos.
2. For Save / Export / Share: re-capture the dialog; set destination under the task workspace when possible.
3. Do not overwrite the original without human approve.
4. Empty UIA tree → one re-capture; after two stuck tries escalate (desktop-operate scorecard).

Edge — open URL (fallback)

1. If browser_* is available, use that instead.
2. Otherwise capture Edge address bar, type URL, Enter.
3. Prefer browser_snapshot when the toolset is on.

Pitfalls

• UAC / SmartScreen: stop and ask the user.
• Elevated Explorer windows may be unreachable from a normal process.
• Do not store secrets in Notepad on shared machines.
• Web CAPTCHA in Edge/Chrome: use browser_* + challenge-gate, not computer_use on the browser window first.
• Blind coordinate spam burns the Operate retry budget — use element indices.

Verification

• Explorer shows the requested folder.
• Notepad contains the drafted text.
• Photos export exists when requested.
• Edge loaded the intended URL (or browser tools were used instead).