Xc Console

name: xc-console description: Automate F5 Distributed Cloud web console operations through browser automation using mcp__chrome-devtools MCP tools. Handles multi-provider authentication (Azure SSO, Google, Okta, SAML, native username/password), detecting session expiry and navigating login flows. Warns when VPN is required. Use when creating HTTP/TCP load balancers, origin pools, WAF policies, deploying cloud sites (AWS/Azure/GCP), managing DNS zones, configuring service policies, or executing any F5 XC GUI-based tasks. Triggers on: F5 XC console, GUI automation, browser automation, login, SSO, authenticate, tenant management, visual configuration, Web App and API Protection, WAAP. allowed_args: true

F5 Distributed Cloud Console Automation Skill

Expert in automating F5 Distributed Cloud web console operations through browser automation using the mcp__chrome-devtools MCP tools.

Overview

This skill uses the mcp__chrome-devtools__* MCP tools which provide Chrome browser automation. These tools provide:

• ✅ Works with your existing browser session (preserves authentication)
• ✅ Provides real-time visual feedback (watch Claude navigate in real time)
• ✅ Uses natural language instructions (no low-level scripting required)
• ✅ Automatically handles existing login state and cookies
• ✅ Multi-provider authentication - detects Azure SSO, Google, Okta, SAML, and native login
• ✅ VPN detection - warns when tenant requires VPN access

MCP Tools Available

Prerequisites

Before using this skill, ensure you have:

1. Chrome DevTools MCP Server

The Chrome DevTools MCP server is configured in .mcp.json and starts automatically:

# Verify MCP server is available:
npx -y @anthropic/mcp-server-chrome-devtools@latest

# The server provides browser automation tools

2. F5 XC API Credentials (for validation)

Set these environment variables for CLI-based verification:

export F5XC_API_URL="https://nferreira.staging.volterra.us"
export F5XC_API_TOKEN='2SiwIzdXcUTV9Kk/wURCJO+NPV8='

3. Authentication

You should already be logged into the F5 XC tenant in your Chrome browser. The skill leverages your existing session and handles authentication automatically if session expires.

Multi-Provider Authentication

This skill automatically detects and handles multiple authentication methods:

Detection Triggers

The skill detects login requirements when:

• URL redirects to login page (login*.volterra.us, login.microsoftonline.com, accounts.google.com, *.okta.com)
• Page contains "Sign in", "Go to login", or "Session expired" messages
• Connection times out (may require VPN)

Auto-Flow Sequence

1. Navigate to F5 XC tenant URL
2. Wait for page load (detect connection failures → warn about VPN)
3. Check URL and page content using mcp__chrome-devtools__take_snapshot
4. Identify authentication type:
   a. Native login → Inform user to enter credentials, wait
   b. SSO redirect → Find SSO button, click, wait for provider
   c. Already logged in → Skip to step 6
5. Handle SSO provider:
   - If cached session → auto-redirect back
   - If credentials needed → inform user, wait for confirmation
6. Verify F5 XC console loaded (look for workspace cards)
7. Continue with original navigation task

Login Example

/xc:console login https://nferreira.staging.volterra.us/ and navigate to WAAP

# Claude will:
# 1. Get browser context with tabs_context_mcp
# 2. Navigate to tenant URL
# 3. Detect auth type (native, SSO, or already logged in)
# 4. Handle accordingly (inform user or auto-complete)
# 5. Navigate to Web App and API Protection workspace
# 6. Take screenshot to confirm

See ./authentication-flows.md for detailed workflow steps.

Quick Start

Basic Navigation

# Launch Claude Code (Chrome DevTools MCP starts automatically)
claude

# Then provide natural language instructions:
"Navigate to https://nferreira.staging.volterra.us and tell me what you see on the home page"

# Claude will:
# 1. Navigate to the URL using mcp__chrome-devtools__navigate_page
# 2. Wait for page to load
# 3. Take a snapshot using mcp__chrome-devtools__take_snapshot
# 4. Describe what it sees

Form Filling

claude

"Navigate to the HTTP Load Balancers page at https://nferreira.staging.volterra.us.
Then click the 'Add HTTP Load Balancer' button.
Fill in the form with:
- Name: my-test-lb
- Namespace: production
- Domains: test.example.com

But stop before submitting - I want to review first."

Data Extraction

claude

"Navigate to the HTTP Load Balancers list page.
Extract all load balancer names, their namespaces, and domains.
Save the results as a JSON array."

Core Capabilities

Navigation

• Navigate to any URL within the F5 XC console
• Click menu items, buttons, and links using text or selectors
• Switch between tabs and manage tab groups
• Handle redirects automatically (including Azure SSO redirect)

Form Interaction

• Fill text inputs, textareas
• Select dropdown options
• Check/uncheck checkboxes
• Add/remove items from lists
• Complete multi-step forms

Content Reading

• Extract text content from pages
• Read DOM structure and elements
• Take screenshots for visual verification
• Read console logs (useful for debugging)

Debugging

• Inspect network requests (see API calls)
• Read console errors and warnings
• Analyze page structure
• Verify element visibility and properties

Session Management

• Automatically uses existing Chrome session
• Preserves authentication state across providers (Azure, Google, Okta, SAML)
• Handles session expiry with automatic auth type detection
• Warns when VPN connection is required
• Provides clear messages when manual authentication needed

Common Workflows

Workflow 1: Create HTTP Load Balancer

claude

"I want to create an HTTP load balancer in F5 XC.

Please:
1. Navigate to https://nferreira.staging.volterra.us
2. Find and click the 'HTTP Load Balancers' page
3. Click 'Add HTTP Load Balancer' button
4. Fill in:
   - Name: demo-lb
   - Namespace: production
   - Domains: demo.example.com
   - Protocol: HTTPS with Automatic Certificate
5. Look for an 'Origin Pool' field and let me know what options are available

Don't submit yet - just show me the form filled in."

Workflow 2: Explore Console Structure

claude

"Help me inventory the F5 Distributed Cloud console.

Navigate to https://nferreira.staging.volterra.us and:
1. Look at the main left sidebar menu
2. For each top-level menu item, tell me:
   - The menu item name
   - Any submenus
   - What page appears when clicked

Take screenshots of key pages so I can see the structure.
Organize the results as a hierarchical list."

Workflow 3: Verify with CLI Integration

# First, use the console to create something
claude

"Navigate to HTTP Load Balancers page and create a new LB named 'cli-test' in 'default' namespace.
Don't submit yet - just tell me the form is ready."

# Then verify with CLI
f5xcctl configuration list http_loadbalancer -n default

# You should see the newly created resource in the list

Advanced Patterns

Taking Screenshots for Reference

claude

"Navigate to the HTTP Load Balancers creation form and take a screenshot.
Save it so I can see the exact form layout and field names."

Handling Authentication Issues

When Claude encounters a login page, CAPTCHA, or other security challenge:

• Claude will pause and describe what it sees
• You manually handle the authentication (log in, solve CAPTCHA)
• Tell Claude to continue with the task

claude

"Try to navigate to https://nferreira.staging.volterra.us"

# If you get: "I see a login page. Azure SSO button is visible."
# You manually click the SSO button or provide credentials in your browser
# Then tell Claude: "I've logged in, continue with the task"

Extracting Structured Data

claude

"Navigate to the HTTP Load Balancers list page.
For each load balancer shown, extract:
- Name
- Namespace
- Status
- Created date (if visible)

Format as a JSON array and save to lb-list.json"

Key Files in This Skill

Documentation Integration

This skill is designed to work alongside official F5 XC documentation:

• See documentation-index.md for links to docs.cloud.f5.com
• Consult console-navigation-metadata.json for detailed form field information
• Review workflows/ directory for step-by-step task guides

Integration with Other F5 XC Skills

This skill works seamlessly with:

f5xc-cli Skill (Query & Verify)

Use f5xcctl to validate console actions:

# After creating something in the console:
f5xcctl configuration get http_loadbalancer demo-lb -n production

# Compare what the console shows vs what the API returns

f5xc-terraform Skill (Infrastructure as Code)

Use Terraform to deploy the same resources as code:

# The console skill helps you:
# 1. Understand the UI workflow
# 2. See all available options
# 3. Learn the resource structure
# Then use f5xc-terraform to automate it

Best Practices

1. Break Large Tasks into Smaller Steps

# Instead of asking Claude to do everything in one go:
# ❌ "Create a complete load balancer with origin pool, health checks, and WAF policy"

# ✅ Do this in phases:
claude
# Phase 1: "Create origin pool named backend-pool"

# Phase 2: "Create HTTP LB named my-app pointing to backend-pool"

# Phase 3: "Add WAF policy to my-app LB"

2. Take Screenshots for Reference

claude

"Take screenshots of these pages and save them:
1. HTTP Load Balancers list page
2. HTTP Load Balancer creation form
3. Origin Pool creation form

I want to see the exact layout and field names."

3. Verify Console State with CLI

# After using console:
f5xcctl configuration list http_loadbalancer --all-namespaces --output-format json

# Compare with what console showed to ensure consistency

4. Be Specific with Instructions

# ❌ Too vague:
# "Create a load balancer"

# ✅ Specific:
# "Create HTTP load balancer named demo-lb in namespace 'production'
#  with domain example.com pointing to origin pool backend-pool.
#  Use HTTPS with automatic certificate."

Troubleshooting

Chrome DevTools MCP Not Available

# 1. Verify MCP server is configured in .mcp.json:
cat .mcp.json

# 2. Test the server manually:
npx -y @anthropic/mcp-server-chrome-devtools@latest

# 3. Check Claude Code MCP status:
# The server should start automatically when needed

Session Expired

When Claude detects session expiry:

# Claude will identify the authentication type:
# - Native login (username/password): You'll be asked to enter credentials
# - SSO (Azure, Google, Okta): Claude attempts auto-login if cached
# - Connection timeout: Claude warns about VPN requirement
# After authenticating, tell Claude to continue

Form Fields Not Found

# Claude will describe what it sees
# Ask Claude to:
# 1. Take a screenshot
# 2. Describe all visible input fields
# 3. Look for the field by label text or placeholder

"Take a screenshot of the form.
Then find and fill the field labeled 'Domain' with 'example.com'."

Navigation Paths Changed

The console UI may change. If Claude can't find expected buttons:

# Claude will explore and find the new location
"The button location seems to have changed.
Explore the page and find the 'Create' or 'Add' button, then click it."

Security & Permissions

This skill operates within your existing Chrome browser session, so:

• ✅ Uses your existing SSO login (Azure, Google, Okta - no re-authentication if cached)
• ✅ Respects your browser's cookie storage and session state
• ✅ Cannot access other browser tabs or extensions
• ✅ Never enters credentials on your behalf (security policy)
• ⚠️ Can only interact with pages you have permission to access
• ⚠️ Should only be used with trusted instructions (avoid pasting untrusted prompts)

For sensitive operations:

• Review Claude's actions before it submits forms
• Use the preview/review step before final submission
• Verify critical operations with f5xcctl CLI afterward

Getting Help

Debugging Claude's Navigation

claude

"I notice you took a wrong turn. Let me help.
Take a screenshot and describe what page you're on.
Then tell me what button or link you see that matches 'HTTP Load Balancers'."

Understanding Form Structure

claude

"Navigate to the form page and analyze its structure.
For each form field, tell me:
- The label text
- The input type (text, select, checkbox, etc.)
- Whether it's required
- Any visible validation hints"

Learning Console Workflows

claude

"Walk me through the steps to create an HTTP load balancer from scratch.
Assume I have:
- A namespace named 'production'
- An origin pool named 'backend-pool'

Show me each page I'd need to visit and what I'd fill in."

Deterministic Navigation (v2.2)

This skill uses pre-crawled metadata for deterministic browser automation. The plugin ships with pre-crawled metadata that works out of the box. Crawling is optional - use it to refresh stale data or update after F5 XC console UI changes.

Selector Priority Chain

The v2.2 metadata includes stable selectors that work across browser sessions, not just session-specific refs:

How It Works

Before v2.2 (Session-Specific Refs):

Claude: Uses ref_27 from metadata
Risk: Refs change between browser sessions
Result: ~70% success rate

After v2.2 (Stable Selectors):

Claude: Uses data_testid > aria_label > text_match fallback
Uses: mcp__chrome-devtools__click with stable selector from snapshot
Result: ~95% success rate across sessions

Metadata Structure (v2.2)

Each element now includes both refs and stable selectors:

{
  "add_button": {
    "ref": "ref_27",
    "text": "Add HTTP Load Balancer",
    "selectors": {
      "data_testid": null,
      "aria_label": "Add HTTP Load Balancer",
      "text_match": "Add HTTP Load Balancer",
      "css": "button:has-text('Add HTTP Load Balancer')"
    }
  }
}

URL Sitemap

The url-sitemap.json file provides complete route mapping:

• Static routes: Fixed paths like /web/home, /web/workspaces/...
• Dynamic routes: Paths with variables like /namespaces/{namespace}/...
• Workspace mapping: Shorthand aliases (waap → /web/workspaces/web-app-and-api-protection)
• Resource shortcuts: Quick navigation (http-lb → full path with namespace variable)

Fallback Strategy

When navigating, Claude uses this priority:

1. Try data_testid selector (most stable)
2. Try aria_label selector
3. Try text_match with find()
4. Try css selector
5. Try session-specific ref (may be stale)
6. Report mismatch for metadata update

Crawl Command

To refresh the metadata (optional):

/xc:console crawl https://nferreira.staging.volterra.us/

See crawl-workflow.md for the detailed crawl process.

Angular CDK/Material Interaction Patterns

CRITICAL: The F5 XC Console uses Angular with Angular CDK/Material components. Some UI elements require specific interaction patterns that differ from standard accessibility-based clicking.

The Overlay Problem

Action menus (⋮ buttons) open Angular CDK overlays whose menu items do NOT appear in accessibility snapshots. This is because:

• Overlays are rendered in a separate container (.cdk-overlay-container)
• The overlay content is outside the main DOM tree that snapshots capture
• Standard click() may not trigger Angular change detection

Solution: JavaScript Event Dispatch

When clicking items in dropdown menus, use evaluate_script with full MouseEvent sequence:

mcp__chrome-devtools__evaluate_script({
  function: `(menuText) => {
    const overlay = document.querySelector('.cdk-overlay-container');
    if (!overlay) return { success: false, error: 'No overlay' };

    const items = overlay.querySelectorAll('[role="menuitem"], button, .mat-menu-item');
    for (const item of items) {
      if (item.textContent.includes(menuText)) {
        // Full event sequence required for Angular Material
        ['mousedown', 'mouseup', 'click'].forEach(type => {
          item.dispatchEvent(new MouseEvent(type, {
            bubbles: true,
            cancelable: true,
            view: window
          }));
        });
        return { success: true, clicked: menuText };
      }
    }
    return { success: false, error: 'Not found: ' + menuText };
  }`,
  args: ["Delete"]
})

When to Use Each Pattern

Full Reference

See browser-interaction-patterns.md for:

• Complete code examples for all patterns
• Troubleshooting common issues
• Test resource naming conventions
• E2E workflow templates

State Detection Capabilities (v2.3)

This skill includes runtime detection scripts for discovering tenant state:

RBAC Permission Detection

Detect read-only vs editable permissions at runtime:

# Claude automatically detects permission state when navigating
/xc:console navigate to HTTP Load Balancers in namespace p-ashworth

# Returns permission state:
# - canEdit: false
# - canDelete: false
# - canCreate: false
# - viewOnly: true
# - lockedActions: ["Add", "Edit Configuration", "Clone Object", "Delete"]

Detection Patterns:

Script: scripts/detect-permissions.js

Subscription Tier Detection

Detect Standard vs Advanced subscription features:

# Claude scans workspace cards for subscription badges
/xc:console check subscription tier

# Returns subscription state:
# - tier: "standard" | "advanced" | "enterprise"
# - badges: ["Limited Availability", "New", "Early Access"]
# - gatedFeatures: ["API Discovery", "Bot Defense Advanced"]

Badge Types:

Script: scripts/detect-subscription.js

Module Initialization Detection

Detect which workspaces/modules need initialization:

# Claude checks workspace About page for service status
/xc:console check module status for web-app-scanning

# Returns module state:
# - initialized: true
# - status: "enabled"
# - action_available: "Explore"

Status Indicators:

Script: scripts/detect-modules.js

Detection Patterns File

All detection patterns are documented in detection-patterns.json:

• No PII or tenant-specific data
• Generalized patterns that work across any F5 XC tenant
• Machine-readable format for automated detection

Usage in Workflows

Claude automatically uses state detection when:

1. Navigating to resource lists (checks RBAC)
2. Opening forms (checks if Add/Edit buttons are available)
3. Entering workspaces (checks initialization state)
4. Scanning home page (checks subscription badges)

This enables conditional workflow execution - Claude adapts its automation based on detected permissions and features.

Current Status

Metadata v2.3.0 (State Detection):

• ✅ Skill directory structure created
• ✅ SKILL.md written with comprehensive instructions
• ✅ Multi-provider authentication (Azure, Google, Okta, SAML, native)
• ✅ VPN detection and warning
• ✅ Console crawl scripts with stable selector extraction
• ✅ Crawl workflow documented (crawl-workflow.md v2.3)
• ✅ URL sitemap with static/dynamic routes (url-sitemap.json)
• ✅ Stable selectors (data-testid, aria-label, text_match, css)
• ✅ Selector priority fallback chain
• ✅ Metadata ships with plugin (crawl is optional)
• ✅ RBAC permission detection (scripts/detect-permissions.js)
• ✅ Subscription tier detection (scripts/detect-subscription.js)
• ✅ Module initialization detection (scripts/detect-modules.js)
• ✅ Detection patterns file (detection-patterns.json)
• ✅ Crawl phases 7-10 for state detection workflow

Next Steps

1. Run Initial Crawl

   /xc:console crawl https://your-tenant.volterra.us/
   

   Populate selectors for all elements across workspaces.

2. Validate Cross-Session Navigation Test deterministic navigation without refs (selectors only).

3. Validate with CLI

   f5xcctl configuration list http_loadbalancer --all-namespaces
   

For detailed API-driven management: See the f5xc-cli and f5xc-terraform skills.

For console documentation mapping: See documentation-index.md (coming soon).

For specific task workflows: See workflows/ directory (coming soon).