seo-audit（SEO 审核）

When this skill is activated, always start your first response with the 🧢 emoji.

SEO Audit

The SEO audit skill provides a systematic methodology for evaluating a website's search optimization across all dimensions: technical infrastructure, on-page elements, content quality, off-page signals, and AI search readiness. It produces a structured scorecard with PASS/FAIL/WARN ratings and prioritized recommendations so every finding is actionable, not just observed. This is the orchestration layer that connects all 13 specialized SEO skills - use it to diagnose, then hand off to the right skill for each fix.

When to use this skill

Trigger this skill when the task involves:

• Running a full SEO audit across a website or specific section
• Performing a technical-only audit (crawlability, indexing, performance)
• Running a content audit (thin pages, cannibalization, freshness, intent)
• Pre-launch SEO check before a new site or major redesign goes live
• Periodic SEO health check (monthly, quarterly)
• Comparing your site's SEO posture against a competitor
• Producing a structured SEO report for a client or stakeholder

Do NOT trigger this skill for:

• Implementing specific fixes - once audit findings are known, load the specialized skill (e.g., core-web-vitals, schema-markup, link-building) for the actual fix work
• Keyword research phase before content planning - use keyword-research skill instead

Key principles

1. Audit systematically - never skip a category because you assume it is fine. The most damaging SEO issues are often in the areas least recently checked.
2. Prioritize by impact x effort - not all issues are equal. A missing sitemap outranks a missing alt text. Score severity honestly.
3. PASS/FAIL/WARN scoring removes subjectivity - every check has a defined threshold. WARN means partially compliant or approaching a threshold, not "kind of okay".
4. An audit without actionable recommendations is useless - every FAIL and every WARN must link to a specific next action, tool, or specialized skill.
5. Re-audit quarterly to catch regressions - deployments, CMS updates, and content changes silently break SEO. Set a recurring audit cadence.

Core concepts

The 5 audit categories

Scoring methodology

Priority matrix

The audit-fix-verify cycle

1. Audit - run all 35 checks, record status and evidence
2. Prioritize - assign Critical/High/Medium/Low to each FAIL and WARN
3. Assign - link each finding to a specialized skill or tool for the fix
4. Verify - re-run the specific check after the fix is deployed (not a full re-audit)
5. Re-audit - full audit quarterly to catch new issues

Common tasks

Run a full SEO audit

Present the complete scorecard to the user. Fill in Status and Details as you analyze the site. Every row must have a status - never leave a row blank.

Section 1: Technical SEO

Section 2: On-Page SEO

Section 3: Content SEO

Section 4: Off-Page SEO

Section 5: AEO & GEO Readiness

Scorecard summary:

Score interpretation:

• 32-35 PASS: Excellent - monitor and maintain
• 26-31 PASS: Good - address WARNs before they become FAILs
• 18-25 PASS: Needs work - prioritize Critical and High items
• Under 18 PASS: Significant SEO debt - consider a full remediation project

Prioritize audit findings

After completing the scorecard, categorize every FAIL and WARN into the priority matrix:

Critical - fix immediately:

• Checks that block indexing: robots.txt disallowing key pages, no sitemap, canonical loops
• Checks that cause major visibility loss: HTTPS failures, redirect chains on key pages
• Rendering issues causing entire page sections to be invisible to Googlebot

High - fix this sprint:

• Missing or duplicate title tags across more than 10% of pages
• Core Web Vitals failing on high-traffic templates
• Keyword cannibalization on money pages
• Toxic backlinks that could trigger a manual penalty

Medium - fix this quarter:

• Missing alt text, inconsistent heading hierarchy
• Thin content on secondary pages
• Missing OG tags, incomplete schema markup
• Stale content on secondary pages

Low - backlog:

• LLMs.txt optimization
• Minor anchor text imbalances
• Brand mention acquisition on marginal sources

Present findings as a prioritized table:

Generate audit report

Use the full report template in references/audit-report-template.md. The report has four sections:

1. Executive summary - overall score, top 3 findings, business impact statement
2. Scorecard - the full 35-check table with status and evidence
3. Detailed findings - one entry per FAIL/WARN with: what was found, why it matters, how to fix it, and which specialized skill to use
4. Action plan - week 1, month 1, and quarter 1 timeline with assigned owners

Load references/audit-report-template.md to get the full report template with placeholder text and formatting.

Run a quick SEO health check

For a rapid 10-check assessment (15-20 minutes, not a full audit):

A quick health check surfaces showstopper issues only. Any FAIL here is Critical priority. For a complete picture, run the full 35-check audit.

Gotchas

1. GSC data lags by 2-3 days - Google Search Console performance data has a processing delay. If you run an audit on Monday, the most recent data in GSC is from Thursday or Friday. Do not interpret the most recent 2-3 days as a traffic drop - it is always incomplete.
2. Robots.txt check T1 can return false PASS if crawled from the wrong user-agent - A robots.txt that disallows Googlebot but allows all other agents will pass generic crawl tests. Always test with User-agent: Googlebot specifically, or use the robots.txt tester in Google Search Console directly.
3. Canonical to a redirecting URL creates a canonical-redirect chain - If the canonical tag points to a URL that itself 301s to another URL, Google may choose to follow the redirect rather than the canonical, or may not honor either. Canonicals must point to the final destination URL, not an intermediate redirect.
4. Core Web Vitals field data vs lab data diverge - PageSpeed Insights lab data (Lighthouse) and the CrUX field data (real user measurements) frequently disagree. A page can show LCP < 2.5s in lab but fail in field data due to geography, network conditions, or device mix. Always treat field data as the ground truth for audit scoring.
5. Hreflang errors silently break international rankings - Invalid hreflang attributes (missing return links, wrong locale codes, self-referencing inconsistencies) are not reported as errors in GSC - they are silently ignored. International SEO issues must be audited with a dedicated hreflang validator tool.

Anti-patterns / common mistakes

References

For detailed audit checklists with per-check verification methods and fix guidance, load:

• references/audit-checklist-technical.md - Detailed technical SEO checks with how-to-verify steps, tool recommendations, and PASS/FAIL/WARN criteria
• references/audit-checklist-content.md - On-page and content audit methodology with verification methods for all 15 checks
• references/audit-checklist-offpage.md - Off-page and AEO/GEO audit methodology for all 10 checks
• references/audit-report-template.md - Full audit report template with executive summary, scorecard, findings, and action plan sections

Only load a reference file when actively working on that audit category - they are detailed and will consume context if loaded all at once.

For deep fixes on specific categories, load the specialized skill:

• keyword-research - Search intent analysis and keyword opportunity mapping
• schema-markup - Structured data implementation (JSON-LD, FAQ, HowTo, Product)
• core-web-vitals - LCP, CLS, INP optimization and performance budgets
• technical-seo - Crawlability, indexing, rendering, redirect management
• content-seo - Topic clusters, E-E-A-T, cannibalization resolution
• link-building - Backlink acquisition strategy and toxic link removal
• local-seo - Google Business Profile, NAP consistency, local citations
• international-seo - Hreflang, multi-language/region SEO architecture
• ecommerce-seo - Product pages, faceted navigation, category SEO
• programmatic-seo - Page generation at scale, template optimization
• aeo-optimization - Featured snippets, voice search, AI answer optimization
• geo-optimization - Generative engine optimization for ChatGPT, Perplexity, Gemini
• on-site-seo - Framework-specific on-page fixes (Next.js, Nuxt, WordPress, etc.)

Companion check

On first activation of this skill in a conversation: check which companion skills are installed by running ls ~/.claude/skills/ ~/.agent/skills/ ~/.agents/skills/.claude/skills/.agent/skills/.agents/skills/ 2>/dev/null. Compare the results against the recommended_skills field in this file's frontmatter. For any that are missing, mention them once and offer to install: `` npx skills add AbsolutelySkilled/AbsolutelySkilled --skill <name> ` Skip entirely if recommended_skills` is empty or all companions are already installed.