// the find
thatrebeccarae/claude-marketing
A full marketing department for Claude Code. Skill packs for Klaviyo, Shopify, GA4, Looker Studio, paid media, and more. Audit, optimize, and report using natural language.
A library of 56 markdown-based 'skills' (plus a handful of scripts) that inject marketing-domain knowledge — ad platform checklists, email/CRM playbooks, SEO frameworks, benchmark tables — into Claude Code so it can answer marketing questions with more specific context than the base model has. It's aimed at marketing consultants and in-house teams already using Claude Code who want canned frameworks for paid media, Klaviyo, Shopify, GA4, and reporting work, not at engineers looking for a code library.
The GA4 pipeline (agents/ga4-monitor, ga4-gap-analyzer) is the one place with real logic — compare-events.js actually diffs live GA4 data against a JSON Schema-validated spec instead of just prompting an LLM to pretend it did. The multi-tool conversion story is legitimately useful: one canonical skill source gets mechanically converted to Cursor .mdc, Windsurf rules, Copilot instructions, and Gemini @import format via scripts/convert.sh, so you're not maintaining five copies by hand. Shipping SKILL.md/REFERENCE.md/EXAMPLES.md as three separate files per skill is a sane split — decision logic stays separate from the benchmark tables and API schemas that actually go stale.
The 'scored audits' (74-check A-F grades for Google Ads, 46-check grades for Meta) are just instructions telling Claude to produce a score — there's no verification harness checking the LLM actually ran 74 checks or that the grade means anything consistent between runs. Almost the entire repo is markdown; despite being tagged Python, there's no real test suite visible for the 56 skills themselves, only a couple of validator scripts for the GA4 agent. The README is as much a funnel for the author's paid Notion guide, Substack, and personal brand as it is documentation, which makes it hard to tell where the free tool ends and the upsell begins. Freshness is a real risk — ad platform rate limits, API schemas, and 'industry benchmarks' baked into REFERENCE.md files have no visible update mechanism beyond a one-time 'tested March 2026' date in frontmatter, so these will quietly rot.