audit-site（审核现场）

audit-site

Perform a comprehensive site audit: crawl, capture enrichment, analyze observations, and produce a findings report.

When to Use

Use this skill when the user wants a thorough analysis of a website — not just downloading it, but understanding its technology stack, identifying issues, and producing actionable findings.

Full Audit Workflow

Phase 1: Configure for the Target

First, understand the site and configure appropriately:

update_settings({
  settings: {
    maxConcurrent: 4,
    crawlDelay: 0.5,
    stripTrackingParams: true
  },
  policy: {
    scopeMode: "sameDomain",
    maxDepth: 5,
    respectRobotsTxt: true,
    includeSupportingFiles: true,
    downloadCrossDomainAssets: true,
    autoUpgradeHTTP: true
  }
})

Adjust based on site size:

• Small site (<100 pages): maxDepth: 10, maxConcurrent: 8
• Medium site (100-1000): maxDepth: 5, maxConcurrent: 4
• Large site (1000+): maxDepth: 3, maxPagesPerCrawl: 500

Phase 2: Crawl

start_crawl({ url: "https://example.com" })

Monitor until complete:

get_crawl_status()
// Repeat with `since` parameter for efficient polling

Phase 3: Capture Enrichment (if crawlio-agent available)

If the Chrome extension is running, enrichment data (framework detection, network requests, console logs) is captured automatically during the crawl and appended to the observation log.

Phase 4: Analyze Observations

Query the observation timeline for patterns:

// Get all observations for the target
get_observations({ host: "example.com", limit: 200 })

// Check framework detection
get_observations({ host: "example.com", source: "extension", limit: 50 })

// Review crawl results
get_observations({ op: "crawl_done" })

// Check for errors
get_errors()
get_failed_urls()

Phase 5: Create Findings

For each issue or insight identified, create a finding with evidence:

create_finding({
  title: "...",
  url: "https://example.com",
  evidence: ["obs_xxx", "obs_yyy"],
  synthesis: "..."
})

Phase 6: Generate Report

Compile findings into a structured report:

get_findings({ host: "example.com" })

Present as a summary with:

• Technology Stack: Framework, rendering mode, CDN
• Issues Found: Grouped by category (security, performance, SEO, errors)
• Site Structure: Tree overview, orphaned pages, broken links
• Recommendations: Prioritized action items

Audit Checklist

Use this checklist to ensure thorough coverage:

Technology

• Framework detected (React, Vue, Next.js, etc.)
• Rendering mode identified (SSR, SSG, CSR, ISR)
• CDN/hosting identified
• Third-party services cataloged

Performance

• Page count and crawl duration noted
• Large files identified (images, videos, scripts)
• External resource dependencies mapped

Security

• HTTPS enforcement checked
• Mixed content identified
• Security headers reviewed (if captured in network data)

Content

• Failed URLs reviewed and categorized
• Redirect chains identified
• Missing resources noted

Structure

• Site tree reviewed for organization
• Depth distribution analyzed
• Cross-domain assets cataloged

Example Report Structure

# Site Audit: example.com

## Summary
- **Pages crawled**: 142
- **Failed**: 4 (3 auth-required API routes, 1 timeout)
- **Framework**: Next.js (App Router, ISR)
- **Total size**: 23.4 MB

## Findings (5)

### Security
1. **Mixed content on 3 pages** — HTTP images loaded on HTTPS pages
   Evidence: obs_a3f7b2c1, obs_b4e8c3d2

### Performance
2. **Unoptimized images** — 8 images over 500KB without modern formats
   Evidence: obs_c5f9d4e3

### SEO
3. **Missing meta descriptions** — 12 pages have no meta description
4. **Duplicate titles** — 3 pairs of pages share identical titles

### Errors
5. **Console errors on product pages** — TypeError in cart widget
   Evidence: obs_d6g0e5f4

## Recommendations
1. Convert images to WebP/AVIF format
2. Fix mixed content by updating CDN URLs to HTTPS
3. Add unique meta descriptions to all pages
4. Fix TypeError in cart widget JavaScript

Tips

• Save the project after crawling so you can revisit: save_project({name: "example.com audit"})
• For recurring audits, load the project and recrawl to compare: load_project({id: "..."})
• Use export_site({format: "folder"}) to archive the downloaded content
• Combine with crawlio-agent browser interaction tools for authenticated areas