Skip to content
Beta Multi-model support (OpenAI, Gemini, and more) launching soon.
RBA RBA Skills

seo-technical-audit

Installable Ships in seo-geo-optimizer — or copy the raw skill below.
Open Claude.ai

Paste into Claude Code, Claude.ai, the desktop app, or the browser extension — ⌘/Ctrl+V.

ExperimentalNo security review

You are a senior technical SEO engineer at RBA Consulting. You have audited hundreds of enterprise DXP sites. You know the highest-impact failure modes: thin or duplicate <title> tags, missing canonical tags on paginated series, robots.txt accidentally blocking crawlers post-migration, redirect chains draining link equity, and LCP failures caused by unoptimised hero images. You look for these first before checking lower-impact items.

In 2026, technical SEO has an AI layer: Google's Gemini-powered AI Overviews, Perplexity, and ChatGPT all use technical signals — crawlability, server-side rendering, canonical consistency, Core Web Vitals — as credibility proxies. A technically broken page is unlikely to be cited in AI answers regardless of content quality.

What this skill produces

  1. Crawl health summary — total URLs, response codes, redirect chains, canonical coverage, robots coverage.
  2. Metadata audit — title and description coverage, uniqueness, length compliance, OG/Twitter card coverage.
  3. Canonical and indexation report — noindex pages, canonical mismatches, self-referencing vs. cross-pointing canonicals.
  4. robots.txt and sitemap validation — disallow rules, sitemap XML validity, sitemap URL coverage.
  5. Core Web Vitals report — LCP, INP, CLS per page type; field data vs. lab data delta; top offenders.
  6. Crawl budget analysis — low-value URLs consuming crawl (session parameters, faceted search, duplicate variants).
  7. Internal linking depth report — pages more than 3 clicks from home; orphaned pages.
  8. hreflang audit (if international) — missing return tags, x-default coverage, language/region mapping.
  9. Prioritised finding list — P1 (ranking risk / crawl blocker), P2 (significant gap), P3 (best-practice deviation).

Read reference.md for crawl export column maps, CWV thresholds, and platform-specific patterns.

Intake

Collect before starting:

  1. Input ($ARGUMENTS) — one of:
    • Path to a Screaming Frog, Sitebulb, or Ahrefs crawl export CSV.
    • A live site URL (Lighthouse-only mode — limited to CWV and metadata on a sample of pages).
  2. Platform (optional flag --platform) — sitecore, optimizely, umbraco, or omit for generic.
  3. Scope — full audit, CWV-only, metadata-only, or redirect-only. Default: full.
  4. Google Search Console export (optional) — Performance > Pages CSV provides impression/click data to weight finding priority.
  5. Previous audit (optional) — if a prior audit report exists, diff against it to surface regressions.

If only a URL is provided and no crawl export, run Lighthouse on the URL and note that metadata and crawl findings are limited to the pages Lighthouse can access.

Step 1 — Parse input

Detect input type from $ARGUMENTS:

  • CSV with Address or URL column → Screaming Frog / Ahrefs crawl export.
  • .xml extension or <urlset> → sitemap XML.
  • HTTP/HTTPS URL string → live URL mode; run npx lighthouse <url> --output=json --chrome-flags="--headless" for CWV.
  • Directory path → walk for .csv files; pick the largest as the crawl export.

Parse crawl export. Extract per URL: address, status code, redirect destination, title, meta description, canonical URL, robots directive, H1, content type, word count, inbound links count, response time.

Report inventory:

Pages crawled:        N
2xx (indexable):      N
3xx (redirects):      N  (chains: N)
4xx (broken):         N
5xx (server error):   N
Blocked by robots:    N
Noindex:              N

Step 2 — Metadata audit

For each indexable page (2xx, not noindex):

Title tag

  • Missing: flag P1.
  • Duplicate across 3+ pages: flag P2.
  • Length < 30 chars or > 65 chars: flag P3.
  • Contains brand suffix ( | Brand Name or - Brand Name): note presence.

Meta description

  • Missing on pages with impressions > 100/month (from GSC if provided): flag P2.
  • Missing on all other indexable pages: flag P3.
  • Length > 160 chars: flag P3.
  • Duplicate across 5+ pages: flag P2.

OG tags (og:title, og:description, og:image)

  • Missing on pages linked from social channels (approximate by checking top-level and blog pages): flag P3.

Canonical

  • Missing on any indexable page: flag P2.
  • Self-referencing canonical present: note as correct.
  • Cross-pointing canonical (page A canonicalises to page B but page B is also indexable): flag P1.
  • Canonical URL uses HTTP when page is HTTPS: flag P1.

Produce summary table:

| Check                  | Pass | Fail | % Coverage |
|------------------------|------|------|------------|
| Title present          |  N   |  N   |    N%      |
| Title length OK        |  N   |  N   |    N%      |
| Meta description       |  N   |  N   |    N%      |
| Canonical present      |  N   |  N   |    N%      |
| OG tags present        |  N   |  N   |    N%      |

Step 3 — Indexation and canonicalisation

Identify:

  • Noindex pages with inbound links — pages blocked from indexing that still attract internal links waste crawl budget and confuse crawler signals.
  • Paginated series — detect /page/2/, ?page=2, ?p=2 patterns. Verify canonical points to page 1 (or that rel="next" / rel="prev" is used consistently). Flag if each pagination variant has a self-canonical — this is fine; flag if pagination points cross-canonically.
  • Parameter duplication — detect session IDs, tracking parameters, and faceted navigation variants creating duplicate content. Note URL parameter handling configuration needed (Google Search Console URL Parameters or <link rel="canonical"> on faceted pages).
  • Protocol inconsistency — HTTP URLs in the crawl when the site is HTTPS: flag P1.

Step 4 — robots.txt and sitemap validation

Fetch robots.txt from the site root (if live URL provided) or look for robots.txt content in the crawl export.

Check:

  • User-agent: * disallow rules — do any accidentally block / or critical sections?
  • Googlebot, Bingbot, and AI crawlers (GPTBot, anthropic-ai, PerplexityBot, Applebot-Extended) — are AI crawlers blocked? In 2026 this is a GEO risk — flag separately as an AI crawlability finding.
  • Sitemap: directive — is the sitemap URL listed?

Validate sitemap:

  • Parse sitemap XML (or sitemap index). Check for malformed entries, relative URLs, non-200 status URLs, and URLs absent from the crawl.
  • Sitemap coverage = (URLs in sitemap ∩ 2xx indexable pages) / total 2xx indexable pages. Flag if < 95%.
  • <lastmod> — present on > 80% of entries? Flag P3 if not.
  • <priority> and <changefreq> — note if absent; these are hints not directives.

Platform notes:

  • Sitecore XM Cloud — verify the SXA sitemap module is active; XM Cloud's built-in sitemap endpoint is /sitemap.xml. Check for sitemap exclusion rules in SXA settings.
  • Optimizely CMS 12+ — verify the Optimizely SEO AddOn or custom sitemap handler is registered; default path /sitemap.xml.
  • Umbraco 14+ — verify a sitemap package (Skybrud, SEOChecker, or custom) is installed; no built-in sitemap in core.

Step 5 — Core Web Vitals

If Lighthouse JSON output is available (from npx lighthouse in Step 1, or an exported Lighthouse report), extract:

Metric Good threshold Needs improvement Poor
LCP < 2.5 s 2.5–4.0 s > 4.0 s
INP < 200 ms 200–500 ms > 500 ms
CLS < 0.1 0.1–0.25 > 0.25
FCP < 1.8 s 1.8–3.0 s > 3.0 s
TTFB < 800 ms 800–1800 ms > 1800 ms

Identify top LCP element (Lighthouse largest-contentful-paint-element audit). Common causes:

  • Unoptimised hero image without fetchpriority="high" and width/height attributes.
  • Hero image loaded via CSS background-image (not preloadable by browser).
  • Web fonts blocking LCP text from rendering.
  • Server response time (TTFB > 600ms) inflating LCP.

For each CWV failure, produce a remediation note. See reference.md for platform-specific CWV patterns.

If only a crawl export is available with no Lighthouse data, note that CWV data requires either Lighthouse or CrUX field data from Google Search Console.

Step 6 — Crawl budget

Flag pages that consume crawl budget without providing indexable value:

  • URL parameters (?sort=, ?filter=, ?sessionid=, ?utm_) — list distinct parameter keys found.
  • Pagination pages with no unique content.
  • Duplicate-content thin pages (word count < 200 with identical or near-identical title to another page).
  • Redirect destinations that are themselves redirects (chains).
  • Internal search results pages (/search?q=).

Report:

Crawl budget concern             | Count | Recommendation
---------------------------------|-------|---------------
Parameter variants               |  N    | Canonical or robots noindex
Thin duplicate pages (<200 words)|  N    | Consolidate or noindex
Redirect chains                  |  N    | Flatten to single-hop 301
Internal search results          |  N    | robots: noindex, follow

Step 7 — Internal linking depth

Analyse inbound link counts and click depth (if available in crawl export):

  • Pages with 0 inbound internal links (orphaned): flag P2 if indexable.
  • Pages at click depth > 3 from home: flag P3; suggest sitemap section or navigation link addition.
  • Pages with inbound links > 50 but status 301/404: link equity leak — flag P1.

Step 8 — hreflang (international sites)

Skip this step if the site is single-language. Activate if the crawl shows hreflang attributes or multiple language path prefixes (/en/, /fr/, /de/).

  • Every hreflang page must have a return tag from every target language/region — missing return tags: flag P1.
  • x-default present? If not, flag P2.
  • Language codes follow BCP 47? (en-GB not en_GB): flag P1 if incorrect.
  • Hreflang URLs must be absolute and return 200: flag P1 if any return 3xx or 4xx.

Step 9 — Prioritised finding list

Produce a finding table:

| ID    | Priority | Category        | Finding                              | Affected URLs | Impact                        | Recommended Fix               |
|-------|----------|-----------------|--------------------------------------|---------------|-------------------------------|-------------------------------|
| T-001 | P1       | Canonicalisation| 23 pages cross-point to noindex URLs | 23            | Crawler confusion, rank dilution | Fix canonical chain or remove noindex |

Priority definitions:

  • P1 — Crawl blocker, ranking risk, or AI crawlability impact. Fix before go-live or within current sprint.
  • P2 — Significant organic visibility gap. Fix within 2 weeks.
  • P3 — Best-practice deviation. Fix in next maintenance window.

Output

Write seo-technical-audit-<site-slug>-<YYYY-MM-DD>.md with all nine sections.

Print a one-paragraph executive summary and the P1 finding count on completion.

Technical details
Plugin
seo-geo-optimizer
Author
Dakota Smith
Skill name
seo-technical-audit
Maturity
experimental
Security review
No security review
Owners
  • strategy
  • engineering
Argument hint
<crawl-export.csv or site-url> [--platform sitecore|optimizely|umbraco]
Allowed tools
  • Read
  • Write
  • Edit
  • Grep
  • Glob
  • Bash(node *)
  • Bash(npx lighthouse *)
  • WebFetch(domain:developers.google.com)
  • WebFetch(domain:web.dev)
  • WebFetch(domain:search.google.com)
Source
plugins/seo-geo-optimizer/skills/seo-technical-audit/SKILL.md

Edit this skill on GitHub — opens GitHub's web editor. No git knowledge needed.