Marketing

How Kim Barrett Built an Advertising Skills Library for AI Agents

Built by realkimbarrett · · Source: github

Quick answer

Kim Barrett published an open source library of advertising skills for Claude Code, OpenClaw and other AI agents. The skills cover customer research, direct response copy, ad creative, funnels and QA, and they chain together into full campaign flows.

Result: 789 GitHub stars and 72 forks on the repo. The README gives no ad performance numbers.

What they built

Kim Barrett published advertising-skills, an open source (MIT) library of skills for AI agents. The repo describes it as advertising skills for OpenClaw, Claude Code and other agents, covering direct response, paid ads, funnels and copy systems. The README also lists Cursor, Windsurf, Codex and custom agent frameworks as places it works.

According to the README, most AI marketing output is generic, and the library is Barrett's attempt to fix that by writing down the thinking behind ads, offers and funnels as skills an agent can follow. It says the library is for media buyers, founders running paid ads, agencies, and people building AI marketing systems.

How it works

Each skill lives in its own folder with a SKILL.md file. The README says each one defines when to use it, what inputs it needs, how it should reason, what it outputs, and what it should not do. Installation is copying the skills into the .agents/skills/ folder of an agent that follows the Agent Skills pattern.

The skills are grouped into five categories:

  • Foundations: avatar-extraction, voice-of-customer-miner, offer-extraction, proof-harvester, positioning-core
  • Copy Chief: a layer based on Eugene Schwartz, with skills like schwartz-awareness-mapper, mechanism-builder, headline-matrix and objection-crusher
  • Operator OS: execution skills like scroll-stopping-creative, ad-angle-multiplier, landing-page-architect, email-sequence-architect and performance-diagnosis
  • QA: generic-language-killer, claim-checker, consistency-auditor, final-polish
  • Orchestrators: ready made flows for a full funnel campaign, a booked call campaign and a campaign diagnosis

The skills are meant to run in a chain. The README's example for a booked call campaign starts with avatar and offer extraction, moves through awareness mapping, mechanism, ad angles, creative and the conversion path, and finishes with objection handling and a pass that strips out generic language. The repo includes an example output for that flow in its examples folder.

The roadmap mentions a bigger Copy Chief library, more funnel orchestration, creative testing systems, and platform layers for Meta, TikTok and YouTube.

The result

The repo shows 789 stars and 72 forks. The README does not share ad performance numbers, client results or case studies, so there is no evidence in the source of how campaigns built with these skills perform.

How to build something similar

  • Pick a field you know well and write down how you make decisions in it, one decision per skill, instead of one giant prompt.
  • Give every skill the same structure: when to use it, inputs, reasoning steps, outputs, and things it must not do.
  • Add a QA skill or two that check the output for vague language and unsupported claims before anything ships.
  • Write an orchestrator skill that runs the others in order, and include a full worked example so people can see what the chain produces.

The original post

Kim Barrett · View on github

Want to build something like this?

Join the Shipping Skool community free. Paid members get live bootcamp calls every week.

Join Shipping Skool — Free

More use cases

← All AI agent use cases