claude-marketing

claude-marketing contains 56 open-source Claude Code skills, specialized agents, and autonomous workflows for marketing teams. Paid media, e-commerce, content, strategy, creative, reporting, and development — specialist depth without specialist headcount.

git clone https://github.com/thatrebeccarae/claude-marketing.git

Works on Mac, Windows, and Linux.

Quick Start · Skills · Skill Packs · Agents · Workflows · Multi-Tool Support · How Skills Work · Composing Skills · Example Prompts · Configuration · Contributing · License

56 open-source skills that give Claude Code real marketing expertise — diagnostic frameworks, industry benchmarks, audit checklists, and platform-specific reference data across paid media, SEO, e-commerce, content, CRO, analytics, and strategy. Also works with Cursor, Aider, Windsurf, GitHub Copilot, and Gemini CLI.

Each skill ships three layers: SKILL.md (frameworks and decision trees), REFERENCE.md (benchmarks, API schemas, rate limits), and EXAMPLES.md (worked prompts with expected output). Not prompt templates — implementation-grade tools tested in real engagements.

The catalog came out of a client engagement where the team was drowning in meetings and grew from there. The companion essay on dgtl dept* walks through three of the skills with the exact prompts I run on live engagements, the output they return, and what they replace.

For installing and using them properly, the Claude Marketing — The Complete Guide is the deeper reference — free, in Notion, six pages on install, credentials, the six skills to start with, how the scored audits work, five workflow chains I run on engagements, and the n8n pipeline build. Read it before you start, or keep it open on a second monitor while you work.

Quick Start

# Clone the repo
git clone https://github.com/thatrebeccarae/claude-marketing.git

# Copy any skill
cp -r claude-marketing/skills/google-ads ~/.claude/skills/

# Or install a full pack
python claude-marketing/skill-packs/scripts/setup-paid-media.py

# Then ask Claude Code anything
# "Run a scored audit of my Google Ads account"
# "Audit my Klaviyo flows and tell me what's missing"
# "Optimize this page for AI search citations"

Each skill works independently — install only what you need.

What’s Inside

Skills

All 56 skills live in skills/ — install individually or use a skill pack to set up a related group. The interactive demo site has search, skill details, and install instructions for each.

All skills tested March 2026 with Claude Code v2.1. Each skill’s SKILL.md includes a tested date and tested_with version in its frontmatter metadata.

Paid Media

SEO & AI Search

Content

Growth & Conversion

Strategy & Research

Email & Lifecycle

Analytics

Creative & Design

Reporting & Deliverables

DevOps

Skill Packs

Grouped collections with setup wizards for installing related skills together:

• Content Pack — 3 skills for SEO content, editorial pipelines, and client communication. (seo-content-writer, content-workflow, email-composer)
• Creative & Design Pack — 3 skills for frontend interfaces, technical diagrams, and programmatic video. (frontend-design, tech-diagram, remotion-video)
• Developer Tools Pack — 2 skills for pre-push security scanning and README generation. (safe-push, github-readme)
• DTC Skill Pack — 6 skills that give Claude Code deep expertise in DTC e-commerce marketing. (klaviyo-analyst, klaviyo-developer, shopify, google-analytics, looker-studio, pro-deck-builder)
• Paid Media Pack — 6 skills for cross-platform paid media — Google Ads, Meta Ads, Microsoft Ads, competitive analysis, waste detection, and account structure review. (google-ads, facebook-ads, microsoft-ads, competitor-ads-analyst, wasted-spend-finder, account-structure-review)
• Strategy & Research Pack — 3 skills for market research, ICP development, and research synthesis. (market-research, icp-research, research-digest)

Multi-Tool Support

Skills work with 5 AI coding tools beyond Claude Code — Cursor, Aider, Windsurf, GitHub Copilot, and Gemini CLI. Run install.sh for interactive setup, or follow the per-tool instructions below.

Quick Install

bash scripts/install.sh

Auto-detects your installed tools, lets you choose which to install for.

Per-Tool Quick-Start

Cursor — One rule file per skill, auto-suggested by Cursor agent.

cp integrations/cursor/*.mdc your-project/.cursor/rules/

Aider — Skills load every session via config.

cp integrations/aider/CONVENTIONS.md your-project/
echo "read: CONVENTIONS.md" >> your-project/.aider.conf.yml

Windsurf — All skills in one file, always active.

cp integrations/windsurf/.windsurfrules your-project/

GitHub Copilot — Works in VS Code and on GitHub.com.

cp integrations/copilot/*.instructions.md your-project/.github/instructions/

Gemini CLI — Modular loading via @import.

cp -r integrations/gemini/* your-project/

Compatibility Matrix

Not all skill content translates across formats — some tools have size limits or don’t support file references.

Advanced Usage

To regenerate integrations/ after editing skills:

bash scripts/convert.sh

See CONTRIBUTING.md for adding new skills and keeping converted output in sync.

How Skills Work

A Claude Code skill is a structured domain expertise file that loads into Claude automatically when invoked. Each skill contains diagnostic frameworks, industry benchmarks, and platform-specific reference data that Claude doesn’t have out of the box — the kind of context that turns a general-purpose AI into a specialist.

What a Skill Contains

How Claude Uses Them

1. You ask a question in natural language (e.g., “Audit my Klaviyo flows”)
2. Claude loads the relevant skill’s SKILL.md, which contains the diagnostic framework
3. Claude references REFERENCE.md for benchmarks, thresholds, and platform-specific data
4. If an MCP server is configured (e.g., Klaviyo), Claude pulls live data from your account
5. Claude delivers structured analysis with specific recommendations — not generic advice

[!NOTE] Skills work without MCP servers too. You can paste data, share screenshots, or use the included scripts to export data manually. The MCP server just makes it seamless.

Agents and Workflows

Agents are standalone analysis modules with portable logic — comparison engines, prompt builders, response parsers. Each agent works independently and can be used without n8n.

Workflows are autonomous n8n pipelines that wire agents together on a schedule. n8n runs locally here to save on costs, but cloud hosting works too. Unlike skills (which respond when you ask), workflows run on a schedule and act independently.

Workflows use Claude via the Anthropic API (not Claude Code) — Claude Opus 4.6 with adaptive thinking for deep analysis, Claude Haiku 4.5 for fast classification.

Agents

Standalone agents with portable analysis logic — usable independently or wired together by a workflow.

Each agent works independently. Use GA4 Monitor for one-time audits without Claude or GTM. Use GA4 Gap Analyzer to diagnose issues from any comparison data. Use GTM Implementer to provision resources from any spec.

Workflows

Autonomous n8n pipelines that wire agents together on a schedule. They monitor, analyze, and act — then notify you with results.

GA4-GTM Pipeline — Daily GA4 monitoring + GTM implementation

GA4 tracking gaps cost you conversion data every day. This pipeline detects them automatically and patches them in GTM.

Handles multiple GA4 properties — configure once per property, runs hands-off from there.

See the GA4-GTM Pipeline README for the architecture overview, or the Complete Guide for the full n8n build walkthrough.

Composing Skills

Skills are most powerful when chained together — each skill’s output becomes context for the next. See the full composing guide for how context flows between skills.

Example Prompts

Email & Lifecycle

"Audit my Klaviyo flows and identify which essential flows I'm missing"
"My checkout completion rate is 31% — what's causing the drop-off?"
"Which traffic sources are driving the most conversions this month?"

Paid Media

"Analyze my Google Ads search terms and find wasted spend"
"Pull competitor ads from Meta Ad Library for these 3 brands and find messaging gaps"
"Run a full cross-platform audit across Google, Meta, and Microsoft — give me a unified health score and deck"

Content & Strategy

"Build an ICP for our SaaS product targeting VP of Marketing at mid-market companies"
"Write a 1,500-word SEO article targeting 'email marketing benchmarks for DTC brands'"
"Generate a research brief on AI in email marketing from the last 30 days"

Creative & Reporting

"Create a dark-mode deck summarizing this month's email performance"
"Build an HTML audit report from these Klaviyo metrics"
"Design a landing page for our analytics product — editorial aesthetic, not typical SaaS"

Developer Tools

"Audit this repo's README and score it"
"Run a health check across all my public repos"
"Scaffold a new repo with standard files and CI config"
"Generate release notes from the last 20 commits"
"Check for outdated dependencies across all my Node projects"

Configuration

Setup Wizard Options

API Keys by Platform

[!IMPORTANT] API keys are stored in .env files which are gitignored by default. Never hardcode keys in skill files or commit them to version control. See SECURITY.md for full credential handling guidance.

claude mcp add klaviyo --transport http https://mcp.klaviyo.com/mcp

claude mcp add klaviyo --transport http "https://mcp.klaviyo.com/mcp?read-only=true"

{
  "mcpServers": {
    "klaviyo": {
      "command": "uvx",
      "args": ["klaviyo-mcp-server@latest"],
      "env": {
        "PRIVATE_API_KEY": "${KLAVIYO_API_KEY}",
        "READ_ONLY": "true",
        "ALLOW_USER_GENERATED_CONTENT": "false"
      }
    }
  }
}

Security

All scripts include input validation, path sanitization, SSRF protection, and secure credential handling. Key principles:

• API keys stored in .env files (gitignored), never hardcoded
• Read-only API access by default — write scopes are opt-in
• No PII extraction — skills work with aggregates, not individual customer data
• No persistent storage — analysis runs in-memory, output goes to local files you control

For vulnerability reporting and full security design details, see SECURITY.md.

Troubleshooting

python --version
# or
python3 --version

Documentation

Skill Packs

Other Resources

Contributing

Contributions are welcome — bug reports, documentation fixes, skill suggestions, and new skills. See CONTRIBUTING.md for guidelines on submitting skills, the required file structure, and the PR process.

License

MIT — see LICENSE for details.