seo-technical-audit（SEO technical 审核）

Technical SEO Audit

Load client context first. See seo-keywords rules/client-setup.md.

File Organization

technical/
├── audit-{date}.json              # Full audit results
├── sitemap-urls.txt               # All discovered URLs
├── homepage.md                    # Homepage content
├── crawl-data/                    # Raw crawl data
├── schema-markup/                 # Generated JSON-LD files
│   ├── homepage.jsonld
│   └── {page-slug}.jsonld
├── core-web-vitals-{date}.json    # CWV results
├── internal-links.json            # Link graph
├── cannibalization.json           # Cannibalization report
└── speed-audit-{date}.json        # Speed optimization report

Full Audit Workflow

Run these sub-audits in parallel where possible, then compile into a single scored report.

1. Crawlability & URL Discovery

# Discover all URLs
firecrawl map {domain} -o .seo/clients/{slug}/technical/sitemap-urls.txt

# Check robots.txt
curl -sI "{domain}/robots.txt" -o .seo/clients/{slug}/technical/crawl-data/robots-headers.txt
firecrawl scrape "{domain}/robots.txt" -o .seo/clients/{slug}/technical/crawl-data/robots.txt

# Check XML sitemap
firecrawl scrape "{domain}/sitemap.xml" -o .seo/clients/{slug}/technical/crawl-data/sitemap.xml

Check for:

• Robots.txt blocks important pages
• Sitemap exists and is referenced in robots.txt
• Sitemap URLs match crawled URLs (missing pages = orphans or blocked)
• Status codes: sample 20-30 URLs with curl -sI -o /dev/null -w "%{http_code}" {url}
• Redirect chains: follow redirects with curl -sIL {url} — flag chains > 2 hops

2. HTTP Headers Audit

For homepage + 5-10 key pages:

curl -sI "{url}" -o .seo/clients/{slug}/technical/crawl-data/headers-{slug}.txt

Check for:

• X-Robots-Tag — any noindex/nofollow directives in headers
• Link: <url>; rel="canonical" — HTTP canonical vs HTML canonical conflicts
• Content-Type — proper charset (UTF-8)
• Cache-Control / Expires — caching configured
• Strict-Transport-Security — HTTPS enforcement
• Content-Security-Policy — security header present
• Redirect type: 301 (permanent) vs 302 (temporary) — 302s waste link equity

3. Meta Tags Audit

Scrape 10-20 key pages:

firecrawl scrape {url} --format markdown,links -o .seo/clients/{slug}/technical/crawl-data/page-{slug}.json

For each page check:

4. Schema Markup Validation

Scrape page HTML to find existing structured data:

firecrawl scrape {url} --html -o .seo/clients/{slug}/technical/crawl-data/html-{slug}.html

Check for:

• JSON-LD blocks in <script type="application/ld+json">
• Microdata attributes (itemscope, itemtype, itemprop)
• Valid schema.org types for page type
• Required properties present for each schema type
• No errors that would prevent rich results

5. JSON-LD Generation

Generate schema markup based on page type. Save to schema-markup/ directory.

LocalBusiness:

{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "{business_name}",
  "image": "{logo_url}",
  "address": { "@type": "PostalAddress", "streetAddress": "", "addressLocality": "", "addressRegion": "", "postalCode": "" },
  "telephone": "{phone}",
  "url": "{domain}",
  "openingHours": "Mo-Fr 08:00-17:00",
  "priceRange": "$$",
  "sameAs": ["{social_urls}"]
}

Article:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "{title}",
  "author": { "@type": "Person", "name": "{author}" },
  "datePublished": "{date}",
  "dateModified": "{modified_date}",
  "publisher": { "@type": "Organization", "name": "{brand}", "logo": { "@type": "ImageObject", "url": "{logo}" } },
  "image": "{featured_image}",
  "description": "{meta_description}"
}

FAQPage:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    { "@type": "Question", "name": "{question}", "acceptedAnswer": { "@type": "Answer", "text": "{answer}" } }
  ]
}

Also generate when appropriate: Product, HowTo, BreadcrumbList, WebSite (with SearchAction), Organization, Review/AggregateRating.

6. Internal Linking Analysis

# Get links from key pages
firecrawl scrape {url} --format links -o .seo/clients/{slug}/technical/crawl-data/links-{slug}.json

Repeat for homepage + top 20 pages. Build a link graph and check:

• Orphan pages: pages with 0 internal links pointing to them
• Link depth: pages should be reachable within 3 clicks from homepage
• Link distribution: flag pages with < 3 internal links or > 100
• Anchor text: over-optimized (same exact-match anchor) vs descriptive
• Broken internal links: 404s found in link extraction
• Nofollow on internal links: shouldn't nofollow your own pages

7. Core Web Vitals

If PageSpeed Insights API key is available:

curl -s "https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url={url}&strategy=mobile&key={api_key}" -o .seo/clients/{slug}/technical/crawl-data/psi-{slug}.json

If no API key, instruct user to get one (free) at https://developers.google.com/speed/docs/insights/v5/get-started.

Metrics to extract:

• LCP (Largest Contentful Paint): Good < 2.5s, Poor > 4.0s
• CLS (Cumulative Layout Shift): Good < 0.1, Poor > 0.25
• INP (Interaction to Next Paint): Good < 200ms, Poor > 500ms
• FCP (First Contentful Paint): Good < 1.8s
• TTFB (Time to First Byte): Good < 800ms

8. Page Speed Optimization Playbook

Based on CWV results, provide specific fixes:

9. Image Optimization Audit

From scraped pages, check all <img> tags:

• Alt text: present on all images, descriptive (not "image1.jpg")
• File format: using modern formats (WebP, AVIF) vs legacy (PNG, JPG)
• Dimensions: width and height attributes set (prevents CLS)
• Lazy loading: loading="lazy" on below-fold images
• Responsive: srcset attribute for different screen sizes
• File size: flag images likely > 200KB without compression

10. Mobile Friendliness

Check from scraped HTML:

• <meta name="viewport" content="width=device-width, initial-scale=1"> present
• No fixed-width elements (tables, iframes without responsive wrapper)
• Font sizes readable (base >= 16px)
• Tap targets adequate (buttons/links >= 48px)
• No horizontal scroll indicators

11. SERP Feature Eligibility

For each key page, assess which rich results it could qualify for:

12. Crawl Budget Analysis

From the full URL list, identify pages wasting crawl budget:

• Thin content: pages with < 300 words (scrape sample to estimate)
• Duplicate content: URLs that are parameter variants of the same page
• Faceted navigation: filter/sort URLs creating combinatorial explosion
• Tag/category pages: low-value taxonomy pages
• Pagination: proper rel=next/prev or load-more implementation

Recommend: noindex tag pages, consolidate thin content, use canonical tags for parameter URLs, implement robots.txt blocks for faceted navigation.

13. International SEO / Hreflang

If client has languages or regions in config:

• Check all pages for <link rel="alternate" hreflang="x-default"> tag
• Verify hreflang values match ISO 639-1 (language) and ISO 3166-1 (region) codes
• Confirm reciprocal tags (if page A references page B, page B must reference page A)
• Check URL structure: recommend subdirectory (/en/, /fr/) over subdomain or ccTLD for most cases
• Verify <html lang=""> attribute matches page language
• Check for missing return tags (most common hreflang error)

14. Cannibalization Detection

Compare all indexed pages to identify keyword overlap:

# Get all page titles from sitemap
firecrawl scrape {url} --format markdown -o .seo/clients/{slug}/technical/crawl-data/page-{n}.md

Detection method:

1. Extract title tag and H1 from each page
2. Identify pages with similar/identical title keywords
3. Search Google for those keywords via firecrawl and check if multiple client pages appear
4. If 2+ pages from the same domain rank for the same query = cannibalization

Recommendations per case:

• Merge: if both pages are thin, combine into one comprehensive page with 301 redirect
• Differentiate: if pages serve different intents, adjust titles/content to target distinct keywords
• Canonical: if one is clearly stronger, canonical the weaker to the stronger

Scoring

Calculate overall technical SEO health score (0-100):

Output: overall score, category scores, prioritized fix list with effort (low/medium/high) and impact (low/medium/high) for each issue.