Agent skill

browser-testing

Browser automation, testing, and debugging using Chrome DevTools. Use when users want to take screenshots, test websites, debug UI, inspect network requests, check console errors, or automate browser interactions.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/development/browser-testing

SKILL.md

Browser Testing

Browser automation via Chrome DevTools Protocol using chrome-devtools-mcp.

Prerequisites

  • Bun: curl -fsSL https://bun.sh/install | bash
  • uv: Already installed

Important: UID-Based Interactions

This tool uses element UIDs from snapshots for interactions (click, fill, hover), NOT CSS selectors.

Workflow:

  1. Navigate to page
  2. Take snapshot to get element UIDs
  3. Use UIDs for click/fill/hover

Usage

bash
cd /Users/roy/.claude/skills/browser-testing
uv run python scripts/browser.py <command> [args]

Commands

Navigation

bash
# Navigate to URL
uv run python scripts/browser.py navigate --url https://example.com

# Back/forward/reload
uv run python scripts/browser.py navigate --back
uv run python scripts/browser.py navigate --forward
uv run python scripts/browser.py navigate --reload --ignore-cache

# New page
uv run python scripts/browser.py new-page https://google.com

# List/select/close pages
uv run python scripts/browser.py list-pages
uv run python scripts/browser.py select-page 1 --bring-to-front
uv run python scripts/browser.py close-page 0

# Wait for text
uv run python scripts/browser.py wait-for "Success" --timeout 5000

Debug

bash
# Screenshot
uv run python scripts/browser.py screenshot --save-path ./shot.png
uv run python scripts/browser.py screenshot --full-page --save-path ./full.png

# Snapshot (get element UIDs for interactions)
uv run python scripts/browser.py snapshot
uv run python scripts/browser.py snapshot --verbose

# Eval JavaScript (must be function syntax)
uv run python scripts/browser.py eval "() => document.title"
uv run python scripts/browser.py eval "() => document.querySelectorAll('a').length"

# Console messages
uv run python scripts/browser.py list-console --types error,warn
uv run python scripts/browser.py get-console 123

Input (requires UIDs from snapshot)

bash
# Get UIDs first
uv run python scripts/browser.py snapshot
# Output: uid=1_3 link "Learn more" ...

# Then interact using UIDs
uv run python scripts/browser.py click 1_3
uv run python scripts/browser.py click 1_3 --dbl-click
uv run python scripts/browser.py fill 1_5 "test@example.com"
uv run python scripts/browser.py hover 1_2

# Form filling
uv run python scripts/browser.py fill-form --elements '[{"uid":"1_5","value":"test"},{"uid":"1_6","value":"pass"}]'

# Keyboard
uv run python scripts/browser.py press-key Enter
uv run python scripts/browser.py press-key "Control+A"
uv run python scripts/browser.py press-key "Control+Shift+R"

# Drag and drop
uv run python scripts/browser.py drag 1_10 1_20

# File upload
uv run python scripts/browser.py upload-file 1_15 ./document.pdf

# Dialog handling
uv run python scripts/browser.py handle-dialog accept
uv run python scripts/browser.py handle-dialog dismiss --prompt-text "input"

Emulation

bash
# Network throttling
uv run python scripts/browser.py emulate --network "Slow 3G"
uv run python scripts/browser.py emulate --network "Offline"
uv run python scripts/browser.py emulate --network "No emulation"

# CPU throttling (1-20, 1=normal)
uv run python scripts/browser.py emulate --cpu 4

# Geolocation
uv run python scripts/browser.py emulate --geo "37.7749,-122.4194"
uv run python scripts/browser.py emulate --geo clear

# Viewport resize
uv run python scripts/browser.py resize 1920 1080
uv run python scripts/browser.py resize 375 667

Network

bash
uv run python scripts/browser.py list-network
uv run python scripts/browser.py list-network --types xhr,fetch
uv run python scripts/browser.py get-network --reqid 5

Performance

bash
uv run python scripts/browser.py perf-start --reload --auto-stop
uv run python scripts/browser.py perf-stop
uv run python scripts/browser.py perf-analyze <insight-set-id> DocumentLatency

Session Note

Each CLI command spawns a fresh browser. For multi-step workflows, use Python directly:

python
from fastmcp import Client
from fastmcp.client.transports import StdioTransport
import asyncio

async def test_flow():
    transport = StdioTransport(command="bunx", args=["chrome-devtools-mcp@latest"])
    async with Client(transport) as client:
        await client.call_tool("navigate_page", {"type": "url", "url": "https://example.com"})
        r = await client.call_tool("take_snapshot", {})
        print(r.content[0].text)  # Get UIDs
        await client.call_tool("click", {"uid": "1_3"})

asyncio.run(test_flow())

Didn't find tool you were looking for?

Be as detailed as possible for better results