BR

browser-daemon

Maintains a persistent Playwright browser instance for interactive testing and debugging.

Install

mkdir -p .claude/skills/browser-daemon && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/254" && unzip -o skill.zip -d .claude/skills/browser-daemon && rm skill.zip

Installs to .claude/skills/browser-daemon

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Persistent browser automation via Playwright daemon. Keep a browser window open and send it commands (navigate, execute JS, inspect console). Perfect for interactive debugging, development, and testing web applications. Use when you need to interact with a browser repeatedly without opening/closing it.
303 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Intermediate

Key capabilities

  • Polls file-based IPC to accept navigation commands
  • Executes arbitrary JS in a persistent browser instance
  • Captures real-time console logs from the active tab
  • Resizes the viewport dynamically via command-line
  • Maintains browser state across multiple distinct tasks

How it works

Operates as a persistent background daemon that communicates with a headless Chrome browser instance using file-based inter-process communication.

Inputs & outputs

You give it
Command string (navigate/exec/status) directed to client.
You get back
JSON result reflecting the state of the browser daemon.

When to use browser-daemon

  • Debug web frontend interactively
  • Execute browser automation scripts
  • Inspect console logs in real-time

About this skill

Browser Daemon - Persistent Browser Automation

Persistent browser daemon that keeps Chrome open and accepts commands via file-based IPC.

Quick Start

Start the daemon (once per session)

cd ~/.claude/skills/playwright-skill
node browser-daemon.js

Or ask: "Start the browser daemon in the background"

The browser window opens and stays open.

Send commands

Use absolute paths (cleaner, no directory changes):

# Navigate
~/.claude/skills/playwright-skill/browser-client.js navigate "http://localhost:8080/..."

# Execute JavaScript
~/.claude/skills/playwright-skill/browser-client.js exec "document.title"
~/.claude/skills/playwright-skill/browser-client.js exec "document.querySelectorAll('div').length"

# Console logs
~/.claude/skills/playwright-skill/browser-client.js console
~/.claude/skills/playwright-skill/browser-client.js console-clear

# Status
~/.claude/skills/playwright-skill/browser-client.js status

# Resize viewport
~/.claude/skills/playwright-skill/browser-client.js resize 1920 1080

Both scripts have shebangs and are executable, so they can be called directly.

Architecture

Claude (Bash tool)
    ↓
browser-client.js (writes .browser-command)
    ↓
browser-daemon.js (polls, executes, writes .browser-result)
    ↓
Chrome Browser (Playwright-controlled)

IPC Files:

  • .browser-command - Command input (JSON)
  • .browser-result - Command output (JSON)
  • .browser-ready - Ready signal (exists when daemon is running)

Files created/deleted automatically during operation.

Commands

navigate

Navigate to URL and wait for page load.

node browser-client.js navigate "http://localhost:8080/page"

Returns: { success: true, url: string, title: string }

exec

Execute JavaScript in browser context.

node browser-client.js exec "document.querySelectorAll('div').length"

Returns: { success: true, result: any }

console

Get all captured console logs.

node browser-client.js console

Returns: { success: true, logs: [{ type, text, location }] }

console-clear

Clear console log buffer.

node browser-client.js console-clear

Returns: { success: true }

status

Check daemon status and current page info.

node browser-client.js status

Returns: { success: true, url, title, consoleLogsCount }

resize

Resize browser viewport.

node browser-client.js resize 1920 1080

Returns: { success: true, width, height }

Daemon Behavior

  • Launches Chrome (not Chromium) for H.264 codec support
  • Viewport set to full available screen size on startup
  • Automatically captures all console output (log, warn, error, debug, pageerror)
  • Polls for commands every 100ms
  • Lazy restart: When browser is closed and command is sent, daemon detects closure, restarts browser, restores last URL, then executes command

Browser Configuration

chromium.launch({
  channel: 'chrome',      // Use Google Chrome, not Chromium
  headless: false,        // Visible window
  args: ['--start-maximized']
})

Important Quirks

Manual window resizing does NOT work:

  • Playwright controls viewport independently from window
  • Manually resizing browser window does NOT change viewport
  • Window resize does NOT trigger JavaScript resize events
  • Use resize command instead: node browser-client.js resize 1024 768

DevTools overlay:

  • Opening DevTools does NOT resize viewport - overlays on top
  • Does NOT trigger resize events
  • Does NOT change window.innerWidth/Height

Viewport vs Window:

  • window.innerWidth/Height - The viewport (controlled by Playwright)
  • window.outerWidth/Height - The window size
  • Only viewport can be changed via resize command

Use Cases

Interactive development: Keep browser open while testing changes, reload and check console without manual interaction.

Debugging console logs: Track down log sources with automatic source location capture.

Inspecting page state: Query DOM, check element counts, inspect computed styles via JavaScript execution.

Testing workflows: Automate multi-step browser interactions (navigate, fill forms, submit, check results).

Troubleshooting

Daemon not responding:

ls ~/.claude/skills/playwright-skill/.browser-ready
cd ~/.claude/skills/playwright-skill && node browser-daemon.js

Commands timing out:

  • Check if browser window is still open
  • Kill daemon and restart
  • Clean up stuck files: rm -f .browser-command .browser-result

Browser window closed:

  • Kill daemon (Ctrl+C)
  • Restart: node browser-daemon.js
  • Or send any command - lazy restart will trigger automatically

Setup (First Time)

cd ~/.claude/skills/playwright-skill
npm install

This installs Playwright and downloads Chrome browser.

Advanced Capabilities

See META_COMMANDS.md for comprehensive reference of Playwright's meta-level capabilities beyond basic testing:

  • Page events (console, network, dialogs, workers, frames, lifecycle)
  • Content capture (screenshots, PDFs, HTML, video)
  • Network control (interception, mocking, HAR replay, WebSockets)
  • Script injection and Node.js function exposure
  • Performance metrics and garbage collection
  • Chrome DevTools Protocol (CDP) access for low-level browser control
  • Browser context events and meta methods

Integration Notes

When user requests browser interaction:

  1. Check if daemon is running (.browser-ready file exists)
  2. If not, offer to start as background task
  3. Use browser-client.js commands to perform actions
  4. Report results back to user

User can request additional features - the codebase is straightforward and well-documented for extensions.

When not to use it

  • Tasks requiring complete browser isolation or cache clearing
  • Testing non-web-based application interfaces

Prerequisites

playwrightnode.js

Limitations

  • File-based IPC is limited to local development environments
  • Requires manual daemon startup once per session
  • Browser crash impacts all subsequent command tasks

How it compares

It keeps the browser state open between operations, avoiding the performance overhead of launching a new browser session for every single command.

Compared to similar skills

browser-daemon side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
browser-daemon (this skill)59moReviewIntermediate
webapp-testing3533moReviewIntermediate
playwright-browser-automation297moReviewIntermediate
bats97moReviewIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

webapp-testing

anthropics

Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

353585

playwright-browser-automation

lackeyjb

Complete browser automation with Playwright. Auto-detects dev servers, writes clean test scripts to /tmp. Test pages, fill forms, take screenshots, check responsive design, validate UX, test login flows, check links, automate any browser task. Use when user wants to test websites, automate browser interactions, validate web functionality, or perform any browser-based testing.

29146

bats

OleksandrKucherenko

Bash Automated Testing System (BATS) for TDD-style testing of shell scripts. Use when: (1) Writing unit or integration tests for Bash scripts, (2) Testing CLI tools or shell functions, (3) Setting up test infrastructure with setup/teardown hooks, (4) Mocking external commands (curl, git, docker), (5) Generating JUnit reports for CI/CD, (6) Debugging test failures or flaky tests, (7) Implementing test-driven development for shell scripts.

991

textual-ui-developer

massgen

Develop and improve the MassGen Textual TUI by running it in a browser via textual-serve and using Claude's browser tool for visual feedback.

24

examples-auto-run

openai

Run python examples in auto mode with logging, rerun helpers, and background control.

23

workflow-test-fix-cycle

catlog22

End-to-end test-fix workflow generate test sessions with progressive layers (L0-L3), then execute iterative fix cycles until pass rate >= 95%. Combines test-fix-gen and test-cycle-execute into a unified pipeline. Triggers on "workflow:test-fix-cycle".

12

Search skills

Search the agent skills registry