Agent-Skills.md

Agent Skills: Tabz MCP - Browser Automation

Controls Chrome browser: takes screenshots, clicks buttons, fills forms, downloads images, inspects pages, captures network requests, checks console errors, debugs API issues. Use when: 'screenshot', 'click', 'fill form', 'download image', 'check browser', 'look at screen', 'capture page', 'check for errors', 'debug network', 'API failing', 'console errors'. Provides MCP tool discovery for 70 tabz_* browser automation tools.

UncategorizedID: GGPrompts/TabzChrome/automating-browser

Repository

GGPrompts/TabzChrome
GGPromptsLicense: MIT
12010

Install this agent skill to your local

pnpm dlx add-skill https://github.com/GGPrompts/TabzChrome/tree/HEAD/plugins/conductor/skills/automating-browser

Skill Files

Browse the full folder contents for automating-browser.

Download Skill

Loading file tree…

plugins/conductor/skills/automating-browser/SKILL.md

Skill Metadata

Name
automating-browser
Description
"Controls Chrome browser: takes screenshots, clicks buttons, fills forms, downloads images, inspects pages, captures network requests, checks console errors, debugs API issues. Use when: 'screenshot', 'click', 'fill form', 'download image', 'check browser', 'look at screen', 'capture page', 'check for errors', 'debug network', 'API failing', 'console errors'. Provides MCP tool discovery for 70 tabz_* browser automation tools."

Tabz MCP - Browser Automation

Overview

Control Chrome browser programmatically via the Tabz MCP server. 70 tools for screenshots, interaction, network debugging, and more.

Tool Discovery (MCP-CLI Mode)

# List all tabz tools
mcp-cli tools tabz

# Search for specific tools
mcp-cli tools tabz | grep screenshot
mcp-cli grep "network"

# Get tool schema before calling (REQUIRED)
mcp-cli info tabz/tabz_screenshot

# Call tool
mcp-cli call tabz/tabz_screenshot '{}'

Browser Debugging (Common Issues)

Check Console Errors

mcp-cli info tabz/tabz_get_console_logs
mcp-cli call tabz/tabz_get_console_logs '{"level": "error"}'

Debug Network/API Issues

# 1. Enable capture BEFORE triggering the action
mcp-cli info tabz/tabz_enable_network_capture
mcp-cli call tabz/tabz_enable_network_capture '{}'

# 2. Trigger the action (click button, navigate, etc.)

# 3. Get failed requests (status >= 400)
mcp-cli info tabz/tabz_get_network_requests
mcp-cli call tabz/tabz_get_network_requests '{"statusMin": 400}'

# Or filter by URL pattern
mcp-cli call tabz/tabz_get_network_requests '{"urlPattern": "api/"}'

Screenshot for Visual QA

mcp-cli info tabz/tabz_screenshot
mcp-cli call tabz/tabz_screenshot '{}'
# Returns file path - use Read tool to view the image

Check Page State

mcp-cli info tabz/tabz_get_page_info
mcp-cli call tabz/tabz_get_page_info '{}'
# Returns URL, title, loading state

mcp-cli info tabz/tabz_get_element
mcp-cli call tabz/tabz_get_element '{"selector": "#error-message", "includeStyles": true}'

Performance Profiling

mcp-cli info tabz/tabz_profile_performance
mcp-cli call tabz/tabz_profile_performance '{}'
# Returns: DOM node count, JS heap size, event listeners, timing metrics

DOM Tree Inspection

mcp-cli info tabz/tabz_get_dom_tree
mcp-cli call tabz/tabz_get_dom_tree '{"maxDepth": 3}'
# Or focus on specific element
mcp-cli call tabz/tabz_get_dom_tree '{"selector": "main", "maxDepth": 5}'

Code Coverage Analysis

mcp-cli info tabz/tabz_get_coverage
mcp-cli call tabz/tabz_get_coverage '{"type": "js"}'
# Shows used vs unused bytes per file

Tool Categories

| Category | Tools | Purpose | |----------|-------|---------| | Tab Management | list_tabs, switch_tab, rename_tab, open_url | Navigate tabs | | Tab Groups | create_group, add_to_group, ungroup_tabs | Organize tabs | | Windows | list_windows, create_window, tile_windows | Window management | | Audio | speak, list_voices, play_audio | TTS notifications | | Page Info | get_page_info, get_element, get_dom_tree | Inspect content | | Interaction | click, fill | Click/type | | Screenshots | screenshot, screenshot_full | Capture visuals | | Downloads | download_image, download_file | Save files | | Network | enable_network_capture, get_network_requests | Debug APIs | | Console | get_console_logs, execute_script | Debug JS | | Performance | profile_performance, get_coverage | Diagnose slowness | | Emulation | emulate_device, emulate_geolocation | Responsive testing |

Tab Groups (Parallel Workers)

When multiple Claude workers run in parallel, each MUST create their own named group:

# Create unique group for this worker
mcp-cli info tabz/tabz_create_group
mcp-cli call tabz/tabz_create_group '{"tabIds": [123, 456], "title": "ISSUE-ID: Research", "color": "blue"}'

# Add more tabs later
mcp-cli call tabz/tabz_add_to_group '{"groupId": 12345, "tabIds": [789]}'

# Cleanup when done
mcp-cli call tabz/tabz_ungroup_tabs '{"tabIds": [123, 456, 789]}'

Group colors: grey, blue, red, yellow, green, pink, purple, cyan

Quick Patterns

Screenshot:

mcp-cli call tabz/tabz_screenshot '{}'

Click:

mcp-cli call tabz/tabz_click '{"selector": "button.submit"}'

Fill form:

mcp-cli call tabz/tabz_fill '{"selector": "#email", "value": "test@example.com"}'

Switch tab:

mcp-cli call tabz/tabz_list_tabs '{}'  # Get tab IDs (large integers like 1762556601)
mcp-cli call tabz/tabz_switch_tab '{"tabId": 1762556601}'

TTS notification:

mcp-cli call tabz/tabz_speak '{"text": "Done!", "priority": "high"}'

Important Notes

  1. Always check schema first: Run mcp-cli info tabz/<tool> before mcp-cli call
  2. Tab IDs: Chrome tab IDs are large integers (e.g., 1762556601), not 1, 2, 3
  3. Network capture: Enable BEFORE the page makes requests
  4. Screenshots: Return file paths - use Read tool to view
  5. Selectors: CSS selectors - #id, .class, button[type="submit"]
  6. JSON quoting: Use single quotes around JSON, double quotes inside

Heredoc for Complex JSON

For complex nested JSON, use heredoc to avoid escaping issues:

mcp-cli call tabz/tabz_execute_script - <<'EOF'
{"code": "document.querySelector('button').click()"}
EOF

Resources

See references/workflows.md for more patterns.