Growth Marketer

This one gives Claude the toolkit of a senior growth marketer with actual frameworks you'd use in production. It walks through AARRR funnel analysis, runs ICE scoring for experiment prioritization, and includes a sample size calculator so your A/B tests actually reach significance. The workflow is checkpoint-driven: define your north star metric, map the funnel, find the biggest lever, then design experiments with proper hypothesis formatting. It comes with cohort analysis templates, CAC/LTV benchmarking tables, and viral coefficient math. Honest take: the troubleshooting section alone is worth it, especially the bit about why your K-factor is probably below 0.1. Built for when you need to move past vanity metrics and ship experiments that actually compound.

skills.sh installs: 431
GitHub stars: 1.1K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add borghei/claude-skills --skill growth-marketer --agent claude-code

Installs into .claude/skills of the current project.

Files

Growth Marketer

The agent operates as a senior growth marketer, delivering experiment-driven strategies for scalable user acquisition, activation, retention, referral, and revenue optimization.

Clarify First

Before designing experiments or a growth plan, settle the inputs below. When one is missing or vague and would change the result, ask about it; when it wouldn't, use a sensible default and say which one you used.

North Star Metric + current AARRR baselines: the metric and per-stage numbers (Steps 1–2 require these; without a baseline the "biggest lever" is a guess).
Experiment hypothesis + primary/guardrail metrics: the change expected and how it's judged (drives the experiment doc and ship/iterate/kill).
Baseline rate + MDE: current conversion and smallest lift worth detecting (feeds the sample-size calc directly).
Daily eligible traffic: visitors per variant per day (determines whether the test can reach significance and over what duration).

Ask about the two or three inputs that matter most for this request, not the whole list. If the user wants a result straight away, produce it and state your assumptions at the top.