# AI use cases for DSH Computer Use Win (Yu-tao-Li/dsh-computer-use-win)

Repo: [Yu-tao-Li/dsh-computer-use-win](https://github.com/Yu-tao-Li/dsh-computer-use-win) | Use cases: 4 | Page: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/

A Windows plugin for DeepSeek Harness (DSH) that lets an agent see, click, type, and manage real desktop apps through an MCP server and a PowerShell UI Automation backend.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Type text into a Windows app and confirm it was saved: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/fill-and-verify-desktop-text-entry/
- Search a desktop app and confirm a result with OCR: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/search-app-without-accessibility-tree-via-ocr/
- Run a browser search and have an agent summarize results: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/browser-search-and-summary-with-confirmation/
- Watch an agent's desktop run and stop it with Esc: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/supervise-agent-desktop-run-with-esc-stop/

## How to type text into a Windows app and confirm it was saved with DSH Computer Use Win

Repo: [Yu-tao-Li/dsh-computer-use-win](https://github.com/Yu-tao-Li/dsh-computer-use-win) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Have a DSH agent type into an open Windows app, then read the text back to check the entry landed.

**Who it is for:** An office worker or developer who repeats the same text entry in a desktop app and wants an agent to do it and prove it worked.

**The problem:** Typing into a desktop app by hand is repetitive, and an agent that only sends keystrokes cannot tell whether the text was actually written.

**The outcome:** The agent types the requested text into the target app and reads the content back, so you get a confirmed result instead of an assumed one.

### Steps

1. Right-click the DSH tray icon, choose Open DSH Terminal, and run `dsh plugin --profile desktop add dsh-computer-use-win`.
2. Right-click the tray icon, choose Quit, then reopen DSH Desktop so the plugin loads.
3. Open the target app, such as Notepad, and leave it visible on screen.
4. Give the agent the task and state what must be confirmed, for example that the text appears in the document.
5. Check the agent's final summary and the step records in the conversation to confirm the read-back result.

### Prompt to paste into your agent

```text
Using the dsh-computer-use-win tools, type <text to enter> into the open <app name> window, then read the text back with the snapshot or ocr tools. Report whether the text was found in the window. Verify the read-back text matches the entered text before reporting success.
```

### Caveat

Keystrokes and coordinate clicks use your real mouse and keyboard, so avoid touching them while the agent is working, and input is only reported as sent until the agent confirms the change.

Risk, credentials: Screenshots and text read from the screen can include account names, passwords, or private content, so mask them before sharing logs or filing an issue.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/fill-and-verify-desktop-text-entry/

## How to search a desktop app and confirm a result with OCR with DSH Computer Use Win

Repo: [Yu-tao-Li/dsh-computer-use-win](https://github.com/Yu-tao-Li/dsh-computer-use-win) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Run a search in a Windows app whose controls are hidden, and use on-screen text recognition to confirm the result.

**Who it is for:** A researcher or support worker who needs to run a search in a desktop app that exposes few UI controls, so normal control lookup fails.

**The problem:** Some apps, including custom-drawn interfaces, do not expose their search box or results to UI Automation, so an agent cannot find them by control name alone.

**The outcome:** The agent locates the search field from the visible text, runs the search, and confirms the expected result appears on screen.

### Steps

1. Install the plugin with `dsh plugin --profile desktop add dsh-computer-use-win` from the DSH terminal.
2. Quit DSH Desktop from the tray icon and reopen it.
3. Open Settings → Computer Use and keep the default tool loading, which includes all tools.
4. Open the target app and make sure the search results page is visible.
5. Give the agent the search term and the text that must appear, then review the confirmation in its summary.

### Prompt to paste into your agent

```text
With the dsh-computer-use-win tools, open <app name>, search for <search term>, and confirm that <expected text> appears. If the accessibility tree does not show the search box, use the ocr tool to locate visible text. Report the confirmed text and its location, and state clearly if it was not found.
```

### Caveat

OCR only returns text and screen coordinates and does not invent control IDs, so results depend on what is visible and can shift when the window moves.

Risk, credentials: OCR reads everything visible on screen, so keep account details and private messages out of view while the agent runs.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/search-app-without-accessibility-tree-via-ocr/

## How to run a browser search and have an agent summarize results with DSH Computer Use Win

Repo: [Yu-tao-Li/dsh-computer-use-win](https://github.com/Yu-tao-Li/dsh-computer-use-win) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Open a browser window, run a search, check the results page, and get a short summary of what was done.

**Who it is for:** A knowledge worker who wants an agent to gather search results in a browser and report back, with a check that the results actually loaded.

**The problem:** Browsing and copying results by hand is slow, and an agent that does not check the results page may report a search that never happened.

**The outcome:** The agent opens the browser, enters the query, confirms the result page, and returns a summary of its actions and what it verified.

### Steps

1. Run `dsh plugin --profile desktop add dsh-computer-use-win` in the DSH terminal opened from the tray icon.
2. Quit DSH Desktop from the tray and reopen it so the new plugin loads.
3. Open Settings → Computer Use to check the tool loading option, then leave it on the default.
4. Give the agent the search query and the text that must appear on the results page.
5. Read the final summary and confirm the agent named what it saw on the results page.

### Prompt to paste into your agent

```text
Using the dsh-computer-use-win tools, open a browser, search for <query>, and confirm the results page shows <expected text>. Summarize what you did and what you verified in no more than five sentences. Do not report success unless the expected text was observed.
```

### Caveat

Chromium and Electron apps can ignore background input, so bring the browser window to the front before the agent types, and results depend on the page and the model used.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/browser-search-and-summary-with-confirmation/

## How to watch an agent's desktop run and stop it with Esc with DSH Computer Use Win

Repo: [Yu-tao-Li/dsh-computer-use-win](https://github.com/Yu-tao-Li/dsh-computer-use-win) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README

Keep an on-screen view of what an agent is doing, and halt the run at once with Esc or your own shortcut.

**Who it is for:** A user who is new to letting an agent control their PC and wants clear visual signals plus a reliable way to stop it.

**The problem:** Letting an agent move the mouse and type on your real desktop feels risky without a visible signal and a dependable stop.

**The outcome:** You see the agent's activity through screen indicators, step records, and the final summary, and can stop a run at any moment.

### Steps

1. Install with `dsh plugin --profile desktop add dsh-computer-use-win` from the DSH terminal, then quit and reopen DSH Desktop.
2. Open Settings → Computer Use and check the screen indicator and mouse indicator options.
3. Use the preview option in settings to see how the indicators look on your screen.
4. Start a small, low-risk task and watch the blue edge glow, the notice bar, and the step records.
5. Press Esc to test the stop, then send a new task or say continue when you want the agent to resume.

### Prompt to paste into your agent

```text
Using the dsh-computer-use-win tools, perform <low-risk desktop task> in <app name> and describe each step as you take it. Stop immediately if the user presses Esc, and report the final state as completed, sent but not confirmed, not completed, or stopped.
```

### Caveat

Esc stops queued and in-progress actions, but input already sent to a program cannot be taken back, and a frozen program must respond before it can stop.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/yu-tao-li-dsh-computer-use-win/supervise-agent-desktop-run-with-esc-stop/
