WHAT PEOPLE WANT ep-07: Teach Your AI Agent Marketing
By Juyel Rana · November 22, 2026 · Updated November 22, 2026

Ask a freelancer or a small business owner what they pay for every month and you'll get the same tired list: a chatbot subscription, an SEO tool, a copywriting app, maybe a scheduler on top. Each one promised to handle the marketing. None of them did the whole job, and the bills kept stacking up.
So the request showing up in public threads this month isn't for another app. It's simpler than that, and harder: teach the AI agent I already have how to do my marketing, and don't charge me again for the lesson.
That is the demand a brand-new GitHub project walked straight into this week.
A free pack that turns a coding agent into a marketing helper
On November 16, a marketer named James Dickerson published Boring Marketing Skills on GitHub: 32 free instruction files, MIT licensed, built for Claude Code, Codex, Cursor and other agents. You install them with one command — npx skills add boringmarketer/boring-marketing-skills — and your agent gets a set of playbooks instead of a blank chat box.
The skills follow a customer, not a channel. There is a start-here router that reads your site and your reviews, writes down your brand files, and points you at the next skill. Then come the working parts: a "get found" group with local SEO planning, service-area pages, keyword research and an ai-search skill that checks whether ChatGPT, Perplexity and Google's AI actually name your business. A "get chosen" group covers reviews, landing pages and lead magnets. "Get the call" handles ads, tracking and the follow-up after a missed call. Email and a monthly performance review close the loop.
Two of the skills are aimed squarely at freelancers and agencies: one finds local businesses losing customers, the other writes the pitch about the one thing that's broken. Every skill can keep a separate folder per client, which is the kind of detail that tells you it was built by someone who does this for a living.
Why it caught on in days, not months
The repo went up on November 16 and was on GitHub's trending list by November 17. By November 22 it had passed 780 stars and 280 forks. Numbers that small don't prove quality — a week-old repo has no track record, and the star count says more about the size of the want than the quality of the answer. But the want is the story here.
Small businesses and freelancers are not short on AI tools. They are short on someone — or something — that does the unglamorous middle of marketing: check the Business Profile against the top three in the map pack, write a month of posts, look at the site on a phone the way a customer would, count whether the ads actually made the phone ring. That work costs either a hire or a stack of subscriptions. A free skill pack that hands the job to an agent you already pay for is exactly the trade people keep asking for.
It also fits a pattern this series has been tracking all month: people want to own the tools they depend on. A public repo you can read, fork and keep is a very different deal from a $49-a-month dashboard that holds your data hostage.
The catches, because there are some
First, the skills are free, but the agent isn't. You still pay your AI provider for the model behind Claude Code or Codex. The pack removes the marketing-tool bill; it doesn't remove the model bill.
Second, two optional extras — Exa for web search and Firecrawl for reading pages — have free tiers and are not required. The README says every skill works with the agent's own web access, and the free path comes first at each step.
Third, one of the suggested tools, boringfunnels, belongs to the author. He says so on the same line every time it appears, and no skill needs it. That kind of disclosure is the bare minimum, and it's worth noting when a project meets it.
Fourth, and this matters most: a skill is a set of instructions, not a guarantee. An agent following a good playbook can still write flat copy, misread your reviews, or state a fact it never checked. The README's own advice is to open an issue when a skill gives bad advice. Mine is simpler — read what it writes before it goes live, the same way you'd read a junior hire's first draft.
How to try it in an afternoon
If you already use Claude Code, Codex or Cursor, the test costs nothing but an hour:
- Install the pack with
npx skills add boringmarketer/boring-marketing-skills, or add it as a Claude Code plugin from the same repo. - Run
/start-here, name your business, and let it read the site and reviews. The first run takes about three minutes. - Pick one skill that matches your weakest spot. If nobody finds you, start with
ai-searchorgoogle-business-profile. If people find you but don't call, start withfirst-screen-check. - Treat the output as a draft. Fix the voice, check the facts, then publish.
If it saves you one subscription or one afternoon a week, it has done its job. If the copy comes out generic, you've learned something cheaper than a monthly plan would have taught you: the agent still needs your judgment, and the skill pack is a starting point, not a replacement for it.
Keep reading on heyjuyel.com
- WHAT PEOPLE WANT ep-02: Own Your AI, Don't Rent It
- WHAT PEOPLE WANT ep-03: Free AI Visibility, No Subscription
- OpenSEO: The Open-Source SEO Tool That Costs $10 a Month
- Crawl4AI: The Free Tool That Shows You Your Website the Way ChatGPT Sees It
- Ranking on Google Doesn't Mean ChatGPT Will Cite You
