FastPlayWright Skill

High-performance browser automation with Fast Playwright MCP. Features token-optimized batch execution, intelligent element selection with fallback, diff det...

Security Scan

The instructions coherently describe a Playwright MCP workflow, but they rely on dynamically running an unvetted npm package via npx while providing no provenance or install spec—this mismatch is potentially risky.

Purpose & Capability

The SKILL.md describes a Playwright MCP-based browser automation capability that matches the skill name and description. However, the skill declares no dependencies or install spec while the instructions explicitly reference running @tontoko/fast-playwright-mcp via npx. That omission (no declared install or provenance for a third-party package) is unexpected for a skill whose functionality depends on that package.

Instruction Scope

The runtime instructions stay within the stated domain (browser automation): they describe using browser_navigate, browser_batch_execute, selectors, snapshot/diff options, and token-optimization parameters. They do not instruct reading system files, extra environment variables, or sending data to unexpected external endpoints. The only external action described is launching the MCP server via npx, which is relevant to the skill's purpose but extends execution to remote package installation.

Install Mechanism

There is no formal install spec, yet the SKILL.md shows an MCP config that runs `npx @tontoko/fast-playwright-mcp@latest`. Invoking npx will fetch and execute code from the npm registry at runtime. The package is a third-party scoped package with no homepage/repo listed in the skill metadata. Running npx @...@latest is high-risk: it dynamically downloads and executes code with no pinned version or provenance, and the skill gives no guidance for vetting or pinning the package.

Credentials

The skill declares no required environment variables, credentials, or config paths. The instructions likewise do not request secrets or unrelated credentials. This is proportionate to its stated purpose.

Persistence & Privilege

The skill does not request always:true and does not claim persistent system configuration changes. It is user-invocable and allows autonomous invocation (platform default), which is normal. There is no evidence it attempts to modify other skills or system-wide settings.

What to consider before installing

This skill purports to use a third-party npm package (@tontoko/fast-playwright-mcp) invoked via `npx ...@latest`, but the registry metadata provides no homepage or source repo. Before installing or using the skill: (1) ask the publisher for the package repository and inspect its code/release history; (2) prefer a pinned version instead of @latest; (3) avoid running in production or with real credentials—test in an isolated environment/VM or container; (4) consider using the official Playwright MCP or a well-known fork with clear provenance; (5) if you must run it, review the package contents for any network calls, credential-stealing code, or arbitrary shell execution. The core functional instructions are coherent, but the lack of provenance and the reliance on dynamic npx execution are the main risks.

Fast Playwright MCP

High-performance browser automation skill using the Fast Playwright MCP server (@tontoko/fast-playwright-mcp). This fork of the Microsoft Playwright MCP provides significant token optimization, batch execution, and enhanced element discovery.

Key Advantages Over Standard Playwright MCP

Feature	Benefit
Token Optimization	70-80% reduction via expectation parameter
Batch Execution	90% token savings for multi-step workflows
Diff Detection	Track only changes, not full snapshots
Enhanced Selectors	Multiple selector types with automatic fallback
Diagnostic Tools	Advanced debugging and element discovery
Image Compression	JPEG format, quality control, resizing

MCP Configuration

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@tontoko/fast-playwright-mcp@latest"]
    }
  }
}

Critical Workflow

Follow these steps in order for optimal results:

1. Navigate First

Use browser_navigate to load the target page before any other operations.

2. Use Batch Execution for Multi-Step Tasks

For 2+ operations, ALWAYS use browser_batch_execute instead of individual tool calls.

3. Optimize Token Usage

Apply expectation parameters to reduce response size.

4. Use Diff Detection for State Changes

Enable diffOptions when tracking changes without navigation.

Token Optimization

The Expectation Parameter

All browser tools support an expectation parameter to control response content:

{
  "includeSnapshot": false,    // 70-80% token reduction
  "includeConsole": false,      // Exclude console messages
  "includeTabs": false,         // Hide tab information
  "includeCode": false,         // Suppress code generation
  "includeDownloads": false     // Exclude download info
}

Snapshot Options

Limit snapshot content for focused analysis:

{
  "snapshotOptions": {
    "selector": ".main-content",  // Capture specific section only
    "maxLength": 2000,            // Limit character count
    "format": "aria"              // Use accessibility tree format
  }
}

Diff Detection

Track only changes between operations:

{
  "diffOptions": {
    "enabled": true,
    "format": "minimal",      // Options: unified, split, minimal
    "threshold": 0.1,
    "maxDiffLines": 50,
    "context": 3
  }
}

Enhanced Selector System

Selector Types (Priority Order)

1. ref - System-generated element ID from previous results (highest priority)
2. css - Standard CSS selectors (#id, .class, tag)
3. role - ARIA roles with optional text (button, textbox, etc.)
4. text - Text content search with optional tag filter

Selector Arrays with Fallback

All element-based tools accept multiple selectors with automatic fallback:

{
  "selectors": [
    { "css": "#submit-btn" },
    { "role": "button", "text": "Submit" },
    { "text": "Submit", "tag": "button" }
  ]
}

Selectors are tried in order until one succeeds. If multiple elements match, returns a candidate list for selection.

Batch Execution

Basic Pattern

Execute multiple operations in a single request:

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      { "tool": "browser_navigate", "arguments": { "url": "https://example.com/login" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#username" }], "text": "[e-mail]" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#password" }], "text": "password123" } },
      { "tool": "browser_click", "arguments": { "selectors": [{ "role": "button", "text": "Login" }] } }
    ]
  }
}

Advanced Configuration

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      {
        "tool": "browser_navigate",
        "arguments": { "url": "https://example.com" },
        "expectation": { "includeSnapshot": false },
        "continueOnError": true
      },
      {
        "tool": "browser_click",
        "arguments": { "selectors": [{ "css": "#submit" }] },
        "expectation": {
          "includeSnapshot": true,
          "snapshotOptions": { "selector": ".result-area" },
          "diffOptions": { "enabled": true, "format": "minimal" }
        }
      }
    ],
    "stopOnFirstError": false,
    "globalExpectation": {
      "includeConsole": false,
      "includeTabs": false
    }
  }
}

Error Handling Options

• continueOnError (per-step): Continue even if this step fails
• stopOnFirstError (global): Stop entire batch on first error

Common Patterns

Navigate and Screenshot

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      { "tool": "browser_navigate", "arguments": { "url": "https://example.com" } },
      {
        "tool": "browser_take_screenshot",
        "arguments": {
          "filename": "homepage.png",
          "fullPage": true,
          "expectation": { "includeSnapshot": false }
        }
      }
    ]
  }
}

Form Filling with Submission

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      { "tool": "browser_navigate", "arguments": { "url": "https://example.com/contact" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#name" }], "text": "John Doe" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#email" }], "text": "[e-mail]" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#message" }], "text": "Hello World" } },
      { "tool": "browser_click", "arguments": { "selectors": [{ "role": "button", "text": "Send" }] } }
    ],
    "globalExpectation": { "includeSnapshot": false }
  }
}

Responsive Design Testing

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      { "tool": "browser_navigate", "arguments": { "url": "https://example.com" } },
      { "tool": "browser_resize", "arguments": { "width": 1920, "height": 1080 } },
      { "tool": "browser_take_screenshot", "arguments": { "filename": "desktop.png" } },
      { "tool": "browser_resize", "arguments": { "width": 768, "height": 1024 } },
      { "tool": "browser_take_screenshot", "arguments": { "filename": "tablet.png" } },
      { "tool": "browser_resize", "arguments": { "width": 375, "height": 667 } },
      { "tool": "browser_take_screenshot", "arguments": { "filename": "mobile.png" } }
    ]
  }
}

Login Flow with Verification

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      { "tool": "browser_navigate", "arguments": { "url": "https://example.com/login" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#email" }], "text": "[e-mail]" } },
      { "tool": "browser_type", "arguments": { "selectors": [{ "css": "#password" }], "text": "secret", "submit": true } },
      { "tool": "browser_wait_for", "arguments": { "text": "Dashboard" } }
    ],
    "globalExpectation": { "diffOptions": { "enabled": true } }
  }
}

Diagnostic Tools

Find Elements

Search for elements using multiple criteria:

{
  "tool": "browser_find_elements",
  "arguments": {
    "searchCriteria": {
      "text": "Submit",
      "role": "button"
    },
    "maxResults": 5,
    "enableEnhancedDiscovery": true
  }
}

Page Diagnostics

Comprehensive page analysis:

{
  "tool": "browser_diagnose",
  "arguments": {
    "diagnosticLevel": "detailed",
    "includePerformanceMetrics": true,
    "includeAccessibilityInfo": true,
    "includeTroubleshootingSuggestions": true
  }
}

Diagnostic levels: none, basic, standard, detailed, full

HTML Inspection

Extract and analyze HTML content:

{
  "tool": "browser_inspect_html",
  "arguments": {
    "selectors": [{ "css": ".content" }],
    "depth": 3,
    "maxSize": 50000,
    "format": "html",
    "optimizeForLLM": true
  }
}

Image Optimization

Compressed Screenshots

{
  "tool": "browser_take_screenshot",
  "arguments": {
    "filename": "screenshot.jpg",
    "type": "jpeg",
    "expectation": {
      "imageOptions": {
        "format": "jpeg",
        "quality": 50,
        "maxWidth": 1280
      }
    }
  }
}

Network Request Filtering

Monitor specific network activity:

{
  "tool": "browser_network_requests",
  "arguments": {
    "urlPatterns": ["/api/"],
    "excludeUrlPatterns": ["analytics", "tracking"],
    "methods": ["GET", "POST"],
    "statusRanges": [{ "min": 200, "max": 299 }],
    "maxRequests": 10,
    "newestFirst": true
  }
}

Tab Management

Multi-Tab Workflow

{
  "tool": "browser_batch_execute",
  "arguments": {
    "steps": [
      { "tool": "browser_tab_new", "arguments": { "url": "https://example.com" } },
      { "tool": "browser_tab_list", "arguments": {} },
      { "tool": "browser_tab_select", "arguments": { "index": 0 } },
      { "tool": "browser_tab_close", "arguments": { "index": 1 } }
    ]
  }
}

Wait Strategies

Wait for Text

{ "tool": "browser_wait_for", "arguments": { "text": "Loading complete" } }

Wait for Text to Disappear

{ "tool": "browser_wait_for", "arguments": { "textGone": "Loading..." } }

Wait for Time

{ "tool": "browser_wait_for", "arguments": { "time": 2 } }

Console Monitoring

Filter Console Messages

{
  "tool": "browser_console_messages",
  "arguments": {
    "consoleOptions": {
      "levels": ["error", "warn"],
      "maxMessages": 10,
      "patterns": ["^Error:"],
      "removeDuplicates": true
    }
  }
}

JavaScript Evaluation

Page-Level Evaluation

{
  "tool": "browser_evaluate",
  "arguments": {
    "function": "() => document.title"
  }
}

Element-Level Evaluation

{
  "tool": "browser_evaluate",
  "arguments": {
    "selectors": [{ "css": "#counter" }],
    "function": "(element) => element.textContent"
  }
}

Dialog Handling

{
  "tool": "browser_handle_dialog",
  "arguments": {
    "accept": true,
    "promptText": "Optional prompt response"
  }
}

File Upload

{
  "tool": "browser_file_upload",
  "arguments": {
    "paths": ["/absolute/path/to/file.pdf"]
  }
}

Drag and Drop

{
  "tool": "browser_drag",
  "arguments": {
    "startSelectors": [{ "css": "#draggable" }],
    "endSelectors": [{ "css": "#dropzone" }]
  }
}

Select Options

{
  "tool": "browser_select_option",
  "arguments": {
    "selectors": [{ "css": "#country" }],
    "values": ["us", "de"]
  }
}

Keyboard Input

{
  "tool": "browser_press_key",
  "arguments": {
    "key": "Enter"
  }
}

Special keys: Enter, Tab, Escape, ArrowUp, ArrowDown, ArrowLeft, ArrowRight, Backspace, Delete, Home, End

Best Practices

Token Efficiency

1. Use batch execution for 2+ operations
2. Disable snapshots for intermediate steps
3. Enable diff detection for state tracking
4. Filter console messages to relevant levels
5. Use selective snapshots with CSS selectors
6. Compress images when quality is not critical

Selector Strategy

1. Use ref from previous results when available
2. Provide fallback selectors for robustness
3. Prefer role-based selectors for accessibility
4. Use CSS selectors for specific targeting
5. Text selectors as last resort

Error Handling

1. Use continueOnError for non-critical steps
2. Set stopOnFirstError: false for best-effort execution
3. Use diagnostic tools when automation fails
4. Check console messages for JavaScript errors

Performance

1. Minimize snapshot size with selectors
2. Use minimal diff format for change tracking
3. Filter network requests to relevant patterns
4. Batch related operations together

Troubleshooting

Element Not Found

1. Use browser_find_elements to discover alternatives
2. Run browser_diagnose for page analysis
3. Check for iframes or shadow DOM
4. Verify page has finished loading

Timeout Issues

1. Use browser_wait_for with specific conditions
2. Increase wait time for slow pages
3. Check network requests for blocked resources

Token Overflow

1. Enable includeSnapshot: false
2. Use snapshotOptions.selector to limit scope
3. Enable diffOptions.enabled: true
4. Reduce maxRequests in network filtering

Tool Reference

Core Automation

• browser_navigate - Navigate to URL
• browser_click - Click element
• browser_type - Type text
• browser_hover - Hover over element
• browser_drag - Drag and drop
• browser_select_option - Select dropdown option
• browser_press_key - Press keyboard key
• browser_file_upload - Upload files
• browser_evaluate - Execute JavaScript
• browser_wait_for - Wait for condition

Batch Operations

• browser_batch_execute - Execute multiple actions

Information Gathering

• browser_snapshot - Capture accessibility snapshot
• browser_take_screenshot - Take screenshot
• browser_console_messages - Get console output
• browser_network_requests - List network requests
• browser_find_elements - Find elements by criteria
• browser_diagnose - Page diagnostics
• browser_inspect_html - HTML content extraction

Tab Management

• browser_tab_list - List all tabs
• browser_tab_new - Open new tab
• browser_tab_select - Switch to tab
• browser_tab_close - Close tab

Navigation

• browser_navigate_back - Go back
• browser_navigate_forward - Go forward

Browser Control

• browser_resize - Resize window
• browser_close - Close browser
• browser_install - Install browser
• browser_handle_dialog - Handle alerts/confirms

Vision (requires --caps=vision)

• browser_mouse_click_xy - Click at coordinates
• browser_mouse_move_xy - Move mouse to coordinates
• browser_mouse_drag_xy - Drag between coordinates

PDF (requires --caps=pdf)

• browser_pdf_save - Save page as PDF