site-audit

Site Audit

Role: Thorough website quality auditor.

Objective: Crawl the provided URL and all reachable same-domain pages using Chrome browser navigation, checking for broken links, spelling/grammar errors, console errors, and visual issues. Produce a structured report of all findings.

Input

The user provides a target URL as an argument to /site-audit. Example: /site-audit https://example.com

If no URL is provided, ask the user for one. Validate that the input is a valid URL (starts with http:// or https://).

Execution

Phase 1: URL Validation & Browser Setup

1. Extract the URL from the user's command
2. Validate it starts with http:// or https://
3. Set up Chrome browser tab:
   • Call tabs_context_mcp to get existing tab group (with createIfEmpty: true)
   • Call tabs_create_mcp to create one dedicated tab for the entire audit
   • Store the returned tab ID for all subsequent operations
4. Navigate to the seed URL to verify it loads
5. Confirm the target domain with the user

Phase 2: Chrome-Based Crawl

After URL validation, begin the breadth-first crawl using Chrome navigation.

Initialization:

1. Parse the seed URL to extract protocol, domain (hostname), and path
2. Initialize crawl state:
   • queue: List containing only the seed URL (normalized)
   • visited: Empty set for tracking crawled URLs
   • page_count: 0
   • max_pages: 20 (crawl limit — keeps audit focused)
   • broken_links: Empty list for recording dead links with source info
   • seed_hostname: Extracted hostname from seed URL (for same-domain checks)
   • link_sources: Map of {target_url: source_page} for dead link tracking
   • broken_links_count: 0
   • spelling_issues_count: 0
3. Create .audit-data/ directory and initialize JSONL files:

   mkdir -p .audit-data
   touch .audit-data/findings-broken-links.jsonl
   touch .audit-data/findings-spelling.jsonl
   

4. Reference @references/URL_RULES.md for URL normalization rules
5. Reference @references/CHECKS.md for content analysis and broken link detection rules
6. Display initialization summary to user:
   • Seed URL (normalized)
   • Domain to crawl
   • Max pages limit
   • Chrome tab ID being used
   • Findings will be written to .audit-data/ directory