Skip to content
Skills Library
Programming · Testing

browser

/browser

The entry point for all browser work in this estate. Drive a real Chrome via the browsercontainer sidecar over MCP SSE (`browser-gpu`). Use when a task needs a live browser — navigate, click, fill forms, screenshot, read the accessibility tree, run JavaScript in-page, or validate WebGPU/WebGL rendering. Canonical owner of the sidecar connection details. Not for raw CDP scripting (use chrome-cdp); when unsure which browser tool fits, `/route` reaches the unregistered browser-automation router.

Repository
agentbox ↗
Licence
AGPL-3.0
Last repository change
27.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
0
HTML comments
0
URLs in SKILL.md
5
Deprecation

No signals detected. Repository changed on 27.10.2026, without references to retired models or APIs.

Licence

AGPL-3.0, copyleft: przeróbki udostępniane innym muszą mieć tę samą licencję.

What it does and when to use it

The entry point for all browser work in this estate. Drive a real Chrome via the browsercontainer sidecar over MCP SSE (`browser-gpu`). Use when a task needs a live browser — navigate, click, fill forms, screenshot, read the accessibility tree, run JavaScript in-page, or validate WebGPU/WebGL rendering. Canonical owner of the sidecar connection details. Not for raw CDP scripting (use chrome-cdp); when unsure which browser tool fits, `/route` reaches the unregistered browser-automation router.

Original author description

The entry point for all browser work in this estate. Drive a real Chrome via the browsercontainer sidecar over MCP SSE (`browser-gpu`). Use when a task needs a live browser — navigate, click, fill forms, screenshot, read the accessibility tree, run JavaScript in-page, or validate WebGPU/WebGL rendering. Canonical owner of the sidecar connection details. Not for raw CDP scripting (use chrome-cdp); when unsure which browser tool fits, `/route` reaches the unregistered browser-automation router.

How it works

  1. Connection (canonical)
  2. When To Use
  3. When Not To Use
  4. Key Tools
  5. Quick Start
  6. Sidecar Management
  7. Health Check
  8. LaTeX and Diagram Workflows
Original SKILL.md headings · 97 lines, 394 words
SKILL.md preview
---
name: browser
description: >-
  The entry point for all browser work in this estate. Drive a real Chrome via the
  browsercontainer sidecar over MCP SSE (`browser-gpu`). Use when a task needs a live
  browser — navigate, click, fill forms, screenshot, read the accessibility tree, run
  JavaScript in-page, or validate WebGPU/WebGL rendering. Canonical owner of the
  sidecar connection details. Not for raw CDP scripting (use chrome-cdp); when unsure
  which browser tool fits, `/route` reaches the unregistered browser-automation router.
version: 2.0.0
triggers:
  - /browser
  - browse
  - web automation
  - scrape
  - navigate
  - screenshot
---

# Browser Automation Skill

All browser automation runs on the external `browsercontainer` sidecar via
`chrome-devtools-mcp` (40+ tools) over MCP SSE. No local browser is installed
in the agentbox image.

## Connection (canonical)

This is the canonical sidecar-connection block; the `chrome-cdp`, `playwright`,
and `browser-automation` skills point here rather than restating it.

```
MCP SSE:  http://browsercontainer:8931/sse   (chrome-devtools-mcp)
CDP:      browsercontainer:9223              (raw Chrome DevTools Protocol, socat proxy, in-network)
VNC:      localhost:5903                     (visual debugging, Display :2)
```

9222 is the host-mapped port (socat rebinds Chrome's localhost-only :9222 to :9223
in-network); use it only from the host, never from inside agentbox.

Auto-registered at boot as `browser-gpu` in `.mcp.json`. Manual registration:

```bash
claude mcp add browser-gpu --transport sse http://browsercontainer:8931/sse
```

## When To Use

- Page navigation, clicking, typing, form fills
- Screenshots and accessibility snapshots
- JavaScript evaluation in page context
- WebGPU/WebGL rendering validation
- Console log monitoring
- Multi-tab workflows
- Mermaid diagram rendering (via `/render-mermaid` HTTP endpoint; see `mermaid-diagrams` skill)

## When Not To Use

- Fetching page content without interaction -- use `web-summary` or `gemini-url-context`
- API testing without a browser -- use `curl` or `httpx`
- Raw CDP protocol scripting -- use the **chrome-cdp** skill
- QE-grade typed assertions, visual-diff baselines -- use **qe-browser**

## Key Tools

Tool names below are the `browser-gpu` server's (`mcp__browser-gpu__<name>`).

### `take_snapshot` (preferred for LLM interaction)
Returns an accessibility tree -- structured, deterministic, no vision model needed. Element `uid`s from it feed `click`, `fill`, `hover`.

### `take_screenshot`
Capture viewport, full page (`fullPage: true`) or one element (`uid`); `filePath` saves instead of attaching.

### `navigate_page`
Go to a URL (`type: "url"`), or back/forward/reload. `new_page` / `select_page` / `list_pages` manage tabs.

### `click` / `type_text` / `fill` / `fill_form`
Interact with page elements.

### `evaluate_script`
Execute JavaScript in the page context.

### `list_console_messages`
Read browser console output.

## Quick Start

```javascript
navigate_page({ type: "url", url: "https://example.com" })
take_snapshot()
take_screenshot({ filePath: "page.png", fullPage: true })
```

## Sidecar Management

```bash
agentbox.sh browsercontainer up        # start
agentbox.sh browsercontainer health    # check all 5 services
agentbox.sh browsercontainer cdp       # list CDP tabs
agentbox.sh browsercontainer shell     # shell into container
agentbox.sh browsercontainer rebuild   # full rebuild
```

## Health Check

```bash
curl -s http://browsercontainer:8931/health
curl -s http://browsercontainer:9223/json/list | jq '.[].url'
```

## LaTeX and Diagram Workflows

Rendering Mermaid `.mmd` files through the sidecar-backed `mmdc` client and
spot-checking compiled LaTeX/PDF pages via the sidecar are covered in
[`references/latex-diagram-workflows.md`](references/latex-diagram-workflows.md).
Files in this package (2)
  • SKILL.md 4 KB
  • references/latex-diagram-workflows.md 2 KB

Versions

RepositoryScoreLicence
This page · dreamlab-ai/agentbox100AGPL-3.0
vibiumdev/vibium100Apache-2.0
superset-sh/superset100MIT
phnx-labs/agi-cli100Apache-2.0
phnx-labs/browser-cli100Apache-2.0
Embervy-ai/embervy100MIT
BlackBeltTechnology/pi-agent-dashboard100MIT
shibing624/agentica100Apache-2.0
iammrduncan/fxbot100Apache-2.0

Database snapshot: 13.11.2026 · package version c7ff683f9c82 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100chrome-cdp
    Raw Chrome DevTools Protocol scripting against a live Chrome session — direct WebSocket, no Puppeteer. Use when you need low-level CDP access: WebGPU/WebGL GPU testing, console/JS evaluation, screenshots, or performance traces on the browsercontainer GPU sidecar. For…
  • 100playwright
    Browser automation, web scraping, visual testing, and WebGPU validation via the browsercontainer sidecar (chrome-devtools-mcp, 40+ tools). Use for navigating pages, clicking elements, filling forms, taking screenshots, executing JavaScript, accessibility snapshots, and WebGPU…
  • 100browser-cdp
    Use this skill when you need to control a Chrome browser via CDP (Chrome DevTools Protocol) to reuse existing login sessions. Covers: launching Chrome in debug mode, opening URLs, waiting for page load, evaluating JavaScript, taking snapshots, and extracting auth tokens. Trigger…
  • 100web-browser
    Automate and interact with web pages through Chrome or Chromium using the Chrome DevTools Protocol (CDP): navigate, click, fill forms, inspect content, take screenshots, and debug console or network activity. Use when an agent needs a real browser. Prefer headless Chrome unless…
  • 100operating-cdp-cli
    Drives a live browser with the repo's cdp CLI — attaching to a running Chrome/Brave, navigating, clicking, evaluating JavaScript, extracting page state, and running the screenshot-first observe-act-verify loop over cdp --mcp. Use when the task means operating a browser right now…
  • 100browser-cdp
    Use this skill when you need to control a Chrome browser via CDP (Chrome DevTools Protocol) to reuse existing login sessions. Covers: launching Chrome in debug mode, opening URLs, waiting for page load, evaluating JavaScript, taking snapshots, and extracting auth tokens. Trigger…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.