Skills

Seo Audit

Run a comprehensive, evidence-based SEO audit of a website covering technical, content, mobile, performance, and conversion factors. Produces a report readable by executives and engineers.

SKILL.md

SEO Audit

A senior-SEO-consultant audit grounded in fetched, observable evidence — not generic advice. Output is two-track: a plain-English executive summary for non-technical readers, and a detailed technical report with code/config snippets for engineers.

When to use

  • The user provides a URL and asks for SEO analysis, ranking help, or a site audit.
  • The user shares Google Search Console screenshots or asks why traffic dropped.
  • The user wants to compare their site against competitors from an SEO angle.

When NOT to use

  • Pure keyword research (no site to audit) → use a keyword-research skill instead.
  • Backlink/off-page-only questions → this audit infers authority signals but does not crawl link graphs.
  • One-off questions about a single SEO concept ("what is canonical?") → just answer directly.

Step 1 — Ask before auditing (max 4 quick questions)

Phrase questions for a non-technical site owner, not a marketer. Skip any you can already infer from the URL or prior conversation.

  1. What do you want this site to do for you? (pick the closest, or describe it)

    • Get people to know about it / build my brand
    • Get sign-ups, members, or community
    • Get donations / financial support
    • Get inquiries from partners, sponsors, or clients
    • Sell something
    • A mix of the above

  2. Where are your visitors? — a specific city, country, region, or anywhere worldwide? This shapes local-SEO recommendations. (If the site title already names a place, infer it and just confirm: "I'll assume Singapore-focused — say if that's wrong.")

  3. Want me to also check site speed? — Real performance data needs ~1 minute of your time: "Open https://pagespeed.web.dev, paste your URL, share the report link or screenshots. Say no and I'll skip the speed section." Front-load this so the user knows what they're opting in to.

  4. (Optional) Competitors — name 1–3 sites you'd like to outrank, or say "you pick."

If the user wants to skip everything, default to: goal = mixed, geography = inferred from the site (state your inference back so they can correct), speed check = ask once before generating the report, competitors = inferred from positioning.

Step 2 — Gather evidence

See references/evidence-gathering.md for the full evidence-gathering protocol (required fetches, renders, scans, and what to do when a fetch fails).

Step 3 — Distinguish measured vs inferred

Tag every finding in the final report:

  • [Measured] — Directly observed in fetched data (e.g., "title tag is 12 chars" — verifiable in HTML).
  • [Inferred] — Best-guess from indirect signals (e.g., "likely losing rankings to X" — no GSC data).
  • [Needs user data] — Requires GSC / Analytics / rank tracker / log files the user didn't share.

Never present [Inferred] as fact. This is the single biggest credibility risk.

Step 4 — Run the framework

Audit these 16 dimensions. For each, produce findings with evidence + severity + plain-English impact.

  1. Executive SEO Summary — maturity, top 3 strengths, top 3 weaknesses, biggest ranking blockers.
  2. Crawlability & Indexability — robots.txt, sitemap, x-robots-tag, orphan risk, AI crawler policy (GPTBot, ClaudeBot, PerplexityBot).
  3. URL Structure & Canonicalization — readability, parameters, canonical tags, HTTP↔HTTPS, www consistency, trailing slashes.
  4. Structured Data — focus on currently rich-result-eligible types: Organization, Product, Article/NewsArticle, Breadcrumb, LocalBusiness, Recipe, Event, VideoObject, Course. Do not recommend FAQ or HowTo schema as growth opportunities — Google restricted rich-result display for both in Aug 2023; they're now only shown for select authoritative sites. Still valid as semantic context, not as a SERP play.
  5. Mobile SEO — content parity desktop↔mobile, hidden important content, CTA visibility, viewport meta. Tap targets: report against WCAG 2.2 SC 2.5.8 (24×24 CSS px), Apple HIG (44pt), Material (48dp) — flag which thresholds fail.
  6. Performance & Core Web Vitals — LCP, CLS, INP with probable causes (LCP image not preloaded, layout shift from web fonts, INP from heavy JS handler / long tasks / hydration). INP is field-only — synthetic tools give an estimate, not a measurement. Prefer CrUX data when available.
  7. On-Page (Metadata) — title intent-match (Google rewrites ~60% of titles since 2021, so length is directional, not a hard rule — aim ≤60 char to avoid truncation but optimize for query/intent match first), description length (≤155 char), duplicates, missing, CTR appeal.
  8. On-Page (Headings & Content) — single H1, hierarchical H2/H3, thin content (<300 words on commercial pages), keyword/intent alignment.
  9. Content Strategy & E-E-A-T — author bylines, citations, original research, topical clusters, gaps.
  10. Internal Linking & Architecture — depth from homepage (>3 clicks = buried), anchor text variety, hub-and-spoke structure.
  11. Image SEO — alt text presence, descriptive filenames, modern formats (WebP/AVIF), explicit width/height (CLS).
  12. Accessibility ↔ SEO overlap — semantic HTML, contrast, focus states; flag where a11y failures also hurt SEO.
  13. Local SEO (if applicable) — NAP consistency, GBP signals, LocalBusiness schema, location pages.
  14. International SEO (if applicable) — hreflang correctness, language switcher, locale URLs.
  15. AI Search Readiness — llms.txt presence and quality, AI crawler allowlist policy (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot), content structured for LLM citation (clear definitions, lists, named-entity clarity, author bios with credentials, dates on every article), Schema.org Author + sameAs for entity grounding.
  16. Conversion Alignment — CTA placement above fold, trust signals, form friction; would SEO traffic actually convert?

Step 5 — Severity rubric

Use these exact criteria — don't grade on vibes.

LevelDefinitionExample
CriticalBlocks indexing or causes major traffic lossnoindex on homepage; robots.txt disallowing /; site not crawlable
HighMeasurable rankings/traffic/conversion impact within 30 days of fixingMissing title tags; LCP >4s on key pages; broken canonical
MediumBest-practice gap; gradual impactThin content; missing schema; weak internal linking
LowPolish; defensiveMissing OG image; minor alt-text gaps

Step 6 — Produce the report

See references/report-template.md for the report structure and output template (header through "Top 5 opportunities"), and references/report-findings-example.md for the detailed findings format, glossary, roadmap, and auto-fix prompt template.

See references/output-and-rules.md for output delivery instructions and hard rules.

More skills

Explore all skills →

Evaluate Requirement

Evaluate product requirements from a URL (Confluence page, Jira ticket, Google Doc, GitHub issue/PR, Figma design), a local file, or pasted text. Produces a structured report: testability check, clarifying questions, notes, out-of-scope items, risk table, test ideas, requirement bugs, and Mermaid mental model diagrams. URL-first — paste any link and the right connector is chosen automatically

→

Ux Audit

Run an evidence-based UX/UI audit of a website with side-by-side desktop vs mobile comparison, covering visual design, responsiveness, WCAG 2.2 accessibility, and conversion. Produces a report readable by designers, engineers, and executives

→

Web Security Audit

Run a non-invasive, evidence-based security audit of a deployed website covering security headers, TLS, information disclosure, third-party scripts, cookies, and privacy/compliance. Scope is defensive and read-only (no exploitation, no fuzzing, no credential testing). Produces a report readable by site owners, developers, and compliance reviewers

→