# gemini-computer-use

Runs Gemini 2.5 Computer Use agents with Playwright to drive browser workflows. It captures screenshots, calls UI actions, and prompts for confirmation on risky steps.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode · 1200 installs · ★ 1039
- Author: am-will
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/am-will/codex-skills/tree/HEAD/skills/gemini-computer-use/SKILL.md
- Updated: 2026-07-14
- HTML: https://affpapa.org/en/nevibecode/skills/gemini-computer-use

## What it does

The skill sets up an agent loop using the Gemini 2.5 Computer Use model. The script `scripts/computer_use_agent.py` takes a page screenshot, inspects model actions from `function_call`, and executes them via Playwright. Each step returns a new screenshot and location until the task finishes or hits `--turn-limit`.

You need Python, `google-genai`, and `playwright` with Chromium installed. Playwright's bundled Chromium runs by default. Setting `COMPUTER_USE_BROWSER_CHANNEL` switches to Chrome or Edge, while `COMPUTER_USE_BROWSER_EXECUTABLE` targets a custom binary and takes precedence.

Running these workflows in a sandbox profile or container prevents unintended account modifications. The `--exclude` flag stops forbidden UI operations, and the agent pauses whenever `safety_decision` flags an action as `require_confirmation`.

## Requirements
- Works with: Playwright

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add am-will/codex-skills --skill gemini-computer-use -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/gemini-computer-use/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/gemini-computer-use/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add am-will/codex-skills --skill gemini-computer-use -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/gemini-computer-use/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add am-will/codex-skills --skill gemini-computer-use -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/gemini-computer-use/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install am-will/codex-skills gemini-computer-use
```

2. **Manually: put the files in place**
- File: `.github/skills/gemini-computer-use/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add am-will/codex-skills --skill gemini-computer-use -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/am-will/codex-skills --path skills/gemini-computer-use
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/gemini-computer-use/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add am-will/codex-skills --skill gemini-computer-use -a opencode
```

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/am-will/codex-skills/tree/HEAD/skills/gemini-computer-use/SKILL.md

## When to use it
- automate browser workflows using visual actions from Gemini 2.5
- execute an inspect-action loop with screenshots and Playwright
- prevent risky browser commands using the `--exclude` parameter
- pause for manual user approval on sensitive UI interactions

## Example prompts
- Set up a Gemini 2.5 browser agent using Playwright to extract test data.
- Run `scripts/computer_use_agent.py` with `--turn-limit 6` and human approval checks.
- Configure `COMPUTER_USE_BROWSER_CHANNEL` to launch Chrome instead of bundled Chromium.

## Similar
- [kernel-python-sdk](https://affpapa.org/en/nevibecode/skills/kernel-python-sdk.md) — skill · 353 installs
- [browser-use](https://affpapa.org/en/nevibecode/skills/browser-use.md) — skill · 98200 installs
- [playwright-cli](https://affpapa.org/en/nevibecode/skills/playwright-cli.md) — skill · 195700 installs
- [browser-trace](https://affpapa.org/en/nevibecode/skills/browser-trace.md) — skill · 3000 installs
- [autobrowse](https://affpapa.org/en/nevibecode/skills/autobrowse.md) — skill · 2800 installs
- [agentic-browser-testing](https://affpapa.org/en/nevibecode/skills/agentic-browser-testing.md) — skill · 1400 installs

From the same set: [am-will/codex-skills](https://affpapa.org/en/nevibecode/by/am-will/codex-skills.md) — 13 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md