coreyhaines31 is attracting attention with +165 stars today because it treats marketing knowledge as something AI agents can execute—not just something they can summarize.
The project focuses on reusable skills for Claude Code and other AI agent workflows, covering:
- Conversion rate optimization (CRO)
- Copywriting and landing-page messaging
- SEO research and content planning
- Analytics interpretation
- Growth engineering experiments
The interesting architectural idea is the separation between an agent’s general reasoning ability and domain-specific operating procedures. Instead of repeatedly explaining how to audit a landing page or structure an SEO brief, you can load a focused skill and make the workflow more consistent.
A minimal setup could look like this:
git clone https://github.com/coreyhaines31/marketing-skills.git
cd marketing-skills
# Inspect available skills before connecting them to an agent
find . -maxdepth 2 -type f | sort
For a Claude Code-style workflow, I would start with a narrow instruction rather than enabling every skill at once:
Use the CRO skill to review this page.
Output:
1. The three biggest conversion blockers
2. Evidence from the page content
3. One low-effort experiment per blocker
4. A measurable success metric
This approach reduces context noise and makes the agent’s output easier to evaluate. If a skill averages 1,000 tokens of instructions and replaces five separate prompts of 500 tokens, the workflow can save roughly 1,500 tokens per task while improving repeatability.
Before using these workflows in production, keep two trade-offs in mind:
- Recommendations still need evidence. Marketing skills can produce polished hypotheses, but analytics, user research, and experiment results should remain the source of truth.
- Context quality determines output quality. Provide real page copy, funnel data, audience details, and business constraints instead of relying on generic prompts.
For developers building AI-assisted growth workflows, this repository is a practical example of turning expert playbooks into composable agent capabilities.
Top comments (0)