/ ディレクトリ / プレイグラウンド / marketingskills
● コミュニティ coreyhaines31 ⚡ 即起動

marketingskills

作者 coreyhaines31 · coreyhaines31/marketingskills

Growth, CRO, copywriting, SEO, analytics skills bundle for Claude — an in-house marketing team as a plugin.

marketingskills is a bundle of Claude Code skills covering the full marketing surface: conversion rate optimization, growth experiments, copywriting voice, SEO analysis, analytics event taxonomy. Instead of hiring a fractional CMO, you get opinionated skills that operate on your site/app.

なぜ使うのか

主な機能

ライブデモ

実際の動作

準備完了

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": {
    "marketingskills-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/coreyhaines31/marketingskills",
        "~/.claude/skills/marketingskills"
      ],
      "_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": {
    "marketingskills-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/coreyhaines31/marketingskills",
        "~/.claude/skills/marketingskills"
      ],
      "_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": {
    "marketingskills-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/coreyhaines31/marketingskills",
        "~/.claude/skills/marketingskills"
      ],
      "_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": {
    "marketingskills-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/coreyhaines31/marketingskills",
        "~/.claude/skills/marketingskills"
      ],
      "_inferred": true
    }
  }
}

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

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "marketingskills-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/coreyhaines31/marketingskills",
        "~/.claude/skills/marketingskills"
      ]
    }
  ]
}

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": {
    "marketingskills-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/coreyhaines31/marketingskills",
          "~/.claude/skills/marketingskills"
        ]
      }
    }
  }
}

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

claude mcp add marketingskills-skill -- git clone https://github.com/coreyhaines31/marketingskills ~/.claude/skills/marketingskills

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: marketingskills

Run a proper CRO experiment cycle on a landing page

👤 Growth teams, founders optimizing funnels ⏱ ~60 min intermediate

使うタイミング: Your landing page converts at 2%, and you want to A/B test your way to 4%.

前提条件
  • Skills installed — git clone https://github.com/coreyhaines31/marketingskills ~/.claude/skills/marketingskills
  • Current page source accessible — Either in repo or via URL
フロー
  1. Hypothesis
    Use marketingskills/cro. Analyze /landing/index.html. Propose 3 high-impact experiment hypotheses with expected lift.✓ コピーしました
    → Hypotheses reference specific page elements and behavioral data
  2. Design A/B
    For hypothesis 1, design the variant. Sample size? Runtime? Kill criteria?✓ コピーしました
    → Complete experiment spec with stats
  3. Readout
    Experiment ran 14 days, B beat A by 18% on signup. Interpret — is this real? What's next?✓ コピーしました
    → Sober readout with statistical and business reasoning

結果: An experiment shipped with discipline, not vibes.

注意点
  • Peeking at data mid-experiment and calling early — Pre-commit to sample size; skill enforces this in prompts
組み合わせ: filesystem

Refresh landing-page copy in your brand voice

👤 Founders between copywriter hires ⏱ ~30 min beginner

使うタイミング: Copy feels generic, you know your voice, you don't have time to write it all.

前提条件
  • Brand voice samples — 2–3 paragraphs you know sound right
フロー
  1. Pin voice
    Use marketingskills/copy. Here are 3 voice samples: <samples>. Profile my voice.✓ コピーしました
    → Voice profile: vocabulary, rhythm, attitude
  2. Rewrite
    Rewrite /landing/index.html headline, subheadline, and CTAs in my voice. Three variations each.✓ コピーしました
    → Variants that sound like the samples
  3. Pressure test
    Which variant survives the "would I actually say this out loud" test? Defend.✓ コピーしました
    → One winner picked with reasoning

結果: Copy that sounds like you wrote it on a good day.

注意点
  • Voice drift when asking for length changes — Re-pin the voice samples before every round
組み合わせ: humanizer-skill

組み合わせ

他のMCPと組み合わせて10倍の力を

marketingskills-skill + humanizer-skill

Rewrite then humanize to strip AI tells

Copy-rewrite, then run humanizer.✓ コピーしました
marketingskills-skill + geo-seo-claude-skill

SEO and GEO together

SEO gaps from marketingskills, GEO fixes from geo-seo-claude.✓ 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
cro_hypotheses page, data? Early-stage CRO 0
copy_rewrite text, voice_samples Copy work 0
seo_gaps site, competitors Content strategy 0
event_taxonomy app_description Analytics setup 0

コストと制限

運用コスト

APIクォータ
N/A
呼び出しあたりのトークン
Varies — copy is cheap, full SEO audits expensive
金額
Free (MIT)
ヒント
Use copy_rewrite in small batches — headline-at-a-time

セキュリティ

権限、シークレット、影響範囲

最小スコープ: filesystem-read
認証情報の保管: None
データ送信先: None

トラブルシューティング

よくあるエラーと対処法

Generic hypothesis output

Provide behavioral data — heatmaps, session recordings, funnel numbers — or acknowledge you're hypothesizing blind

Copy still sounds AI

Run humanizer-skill as a second pass

Event taxonomy conflicts with existing events

Provide current schema; skill will propose additions, not a full rewrite

代替案

marketingskills 他との比較

代替案代わりに使う場面トレードオフ
Hire a fractional CMOStakes are high and you want human judgmentExpensive; slower
Point-tool SaaS (Optimizely, VWO)You want full experimentation platformsPaid; less integrated with your code

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