seo-audit-skill MCP — ユースケース, インストール & ライブデモ

なぜ使うのか

主な機能

108 audit rules across 12 categories
Core Web Vitals via Lighthouse integration
Structured data (JSON-LD) validation
Security header check (CSP, HSTS, X-Frame-Options)
Accessibility basics (alt text, ARIA, heading order)
Output as markdown or JSON for downstream tooling

ライブデモ

実際の動作

seo-audit-skill.replay ▶ 準備完了

0/0

インストール

クライアントを選択

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "seo-audit-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/seo-skills/seo-audit-skill",
        "~/.claude/skills/seo-audit-skill"
      ],
      "_inferred": true
    }
  }
}

Claude Desktop → Settings → Developer → Edit Config を開く。保存後、アプリを再起動。

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "seo-audit-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/seo-skills/seo-audit-skill",
        "~/.claude/skills/seo-audit-skill"
      ],
      "_inferred": true
    }
  }
}

Cursor は Claude Desktop と同じ mcpServers スキーマを使用。プロジェクト設定はグローバルより優先。

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "seo-audit-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/seo-skills/seo-audit-skill",
        "~/.claude/skills/seo-audit-skill"
      ],
      "_inferred": true
    }
  }
}

Cline サイドバーの MCP Servers アイコンをクリックし、"Edit Configuration" を選択。

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "seo-audit-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/seo-skills/seo-audit-skill",
        "~/.claude/skills/seo-audit-skill"
      ],
      "_inferred": true
    }
  }
}

Claude Desktop と同じ形式。Windsurf を再起動して反映。

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "seo-audit-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/seo-skills/seo-audit-skill",
        "~/.claude/skills/seo-audit-skill"
      ]
    }
  ]
}

Continue はマップではなくサーバーオブジェクトの配列を使用。

~/.config/zed/settings.json

{
  "context_servers": {
    "seo-audit-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/seo-skills/seo-audit-skill",
          "~/.claude/skills/seo-audit-skill"
        ]
      }
    }
  }
}

context_servers に追加。保存時に Zed がホットリロード。

claude mcp add seo-audit-skill -- git clone https://github.com/seo-skills/seo-audit-skill ~/.claude/skills/seo-audit-skill

ワンライナー。claude mcp list で確認、claude mcp remove で削除。

ユースケース

実用的な使い方： seo-audit-skill

How to run your first full SEO audit and prioritize fixes

👤 Site owners and SEO practitioners who inherit a site with unknown SEO health ⏱ ~30 min beginner

使うタイミング： New project, new site, or pre-launch checklist.

前提条件

Skill installed — git clone https://github.com/seo-skills/seo-audit-skill ~/.claude/skills/seo-audit-skill
Node 18+ — nvm install 18

フロー

Run full audit

Run seo-audit-skill on https://example.com — all 108 rules, output markdown.✓ コピーしました

→ Report with per-rule pass/fail and a category summary
Prioritize by impact

Group the failures into 'ship this week' (high impact, low effort) vs 'backlog'. Explain the cutoff.✓ コピーしました

→ Ranked fix list with rationale
Generate fix PRs

For the top 5 'ship this week' items, draft the code changes in a PR-ready diff.✓ コピーしました

→ Diffs for meta tags, headers, and structured data

結果： A sprint-sized SEO fix list with drafted diffs.

注意点

Lighthouse scores vary run-to-run — Run 3 times, take median; skill supports --runs=3
Rules flag false positives on JS-heavy SPAs — Use --render=js to enable JS rendering; costs more time

組み合わせ： trae-skill

Run just the Core Web Vitals checks in CI

👤 Performance engineers wiring SEO to CI ⏱ ~20 min intermediate

使うタイミング： You want a CI gate that blocks regressions in LCP/CLS/INP.

フロー

Scope to CWV category

Run seo-audit-skill --category=cwv on our staging URL.✓ コピーしました

→ Just the CWV rules with values
Wrap in CI

Generate a GitHub Actions step that runs this and fails if LCP > 2.5s.✓ コピーしました

→ YAML snippet with threshold logic

結果： A CI gate that catches perf regressions before prod.

注意点

CI runner is slower than users — Use field data via CrUX where possible; treat lab data as a regression detector, not absolute target

Validate and fix structured data (JSON-LD)

👤 Developers wanting rich results in SERPs ⏱ ~25 min intermediate

使うタイミング： Your pages aren't getting rich snippets despite having JSON-LD.

フロー

Scope to structured-data rules

Run seo-audit-skill --category=structured-data on these 5 URLs.✓ コピーしました

→ Per-URL list of schema errors and missing required fields
Generate corrected JSON-LD

For each error, produce the corrected JSON-LD snippet keyed to the page type.✓ コピーしました

→ Per-page snippets ready to paste

結果： Schema.org-valid JSON-LD deployed; ready to re-check via Google Rich Results Test.

注意点

Schema valid but Google still doesn't show rich results — Validity is necessary, not sufficient. Check eligibility guidelines for the specific rich result type.

Audit and set security headers

👤 Developers wanting baseline security + SEO signal ⏱ ~45 min intermediate

使うタイミング： securityheaders.com gave you an F and you want to understand why.

フロー

Run security category

Run seo-audit-skill --category=security on https://example.com.✓ コピーしました

→ CSP, HSTS, X-Frame-Options, Referrer-Policy status
Generate headers config

Produce the nginx and Cloudflare Worker configs to set the missing headers safely (report-only CSP first).✓ コピーしました

→ Copy-paste configs with report-only CSP

結果： A secure header config rolled out behind report-only mode.

注意点

Enforcing CSP breaks the site — Always start with Content-Security-Policy-Report-Only; watch reports for 1-2 weeks before enforcing

組み合わせ： claude-code-owasp-skill

組み合わせ

他のMCPと組み合わせて10倍の力を

seo-audit-skill + trae-skill

TRAE scaffolds SEO basics, seo-audit verifies them

Scaffold SEO with TRAE, run seo-audit to verify, fix gaps.✓ コピーしました

seo-audit-skill + claude-code-owasp-skill

Security-header subset of SEO overlaps with OWASP checks

Run seo-audit security category, then OWASP skill for deeper security review.✓ コピーしました

ツール

このMCPが提供する機能

ツール	入力	呼び出すタイミング	コスト
full_audit	url: str, runs?: int, render?: 'js'\|'static'	First audit on a site	0
category_audit	url, category: 'cwv'\|'security'\|'structured-data'\|...	Focused checks or CI	0
diff_fix_generator	report, top_n?	After audit, to turn findings into PRs	Claude tokens

コストと制限

運用コスト

APIクォータ: None. Runs locally via headless Chromium.
呼び出しあたりのトークン: Report interpretation uses 5-15k Claude tokens depending on site size
金額: Free
ヒント: Use --category= to avoid running all 108 rules when you only need a subset.

セキュリティ

権限、シークレット、影響範囲

認証情報の保管： No credentials

データ送信先： Fetches the target URL and its assets. Does not send report to any third party.

Running against authenticated/intranet URLs requires you to pass cookies — don't hardcode session tokens.

トラブルシューティング

よくあるエラーと対処法

Puppeteer fails to launch

Install Chrome dependencies: apt-get install libgbm1 libnss3 (Linux) or run npx puppeteer browsers install chrome

確認： node -e 'require("puppeteer").launch().then(b=>b.close())'

Rules flag modern JS framework features as problems

Use --render=js so the audit runs post-hydration

All CWV rules report N/A

Lighthouse lab mode; on slow connections it times out. Increase --timeout.

代替案

seo-audit-skill 他との比較

代替案	代わりに使う場面	トレードオフ
Google Lighthouse CLI	You only need the canonical CWV/perf subset	No structured-data or SEO-specific rules
Screaming Frog	You need enterprise-grade crawling of large sites	Paid, desktop app, not LLM-integrated

その他

リソース

📖 GitHub の公式 README を読む

🐙 オープンな issue を見る

🔍 400以上のMCPサーバーとSkillsを見る