Skip to content

31 free marketing skills for AI agents land on GitHub

A one-day-old MIT repository gives Claude Code, Codex and Cursor a staged marketing workflow, from brand voice to ad tracking.

Draftship Editorial3 min read

A free marketing skill pack for coding agents

On 2026-11-17 the GitHub trending list carried boringmarketer/boring-marketing-skills, a repository created the day before on 2026-11-16. In that one day it collected 331 stars and 147 forks, according to its own listing. The package is a set of free marketing skills for AI agents, MIT licensed, written mainly in Python, and aimed at Claude Code, Codex, Cursor and comparable tools.

What the pack actually contains

The skills are grouped by the stage at which a customer meets a business, not by channel. The opening group holds a router called start-here plus brand-voice, positioning-angles, audience-research and competitive-intel. A discovery group covers google-business-profile, local-seo-plan, service-area-pages, ai-search, site-audit, keyword-research, seo-content and programmatic-seo. Conversion skills cover reviews, first-screen-check, landing-page, direct-response-copy and lead-magnet. Paid media gets paid-ads, google-ads-tracking, ads-incrementality-check, meta-ads, speed-to-lead and creative. Then come email-sequences, newsletter and performance-review, two agency skills (find-local-clients, pitch-the-fix) and three self-promotion skills (social-content, long-form-article, content-atomizer).

The author is James Dickerson, who says the skills came out of real campaigns. The README names three findings behind them: that many local businesses running ads promising same-day or round-the-clock service do not state that promise on the first screen of the page the ad opens (hundreds of local sites checked on a phone, Nov 2026); that a Performance Max campaign credited with most of an account's conversions showed no movement in the phone log when paused; and that one broad campaign brought cheaper leads than the same budget split across three city campaigns.

How you install and run it

There are two routes. Any agent can run npx skills add boringmarketer/boring-marketing-skills. Claude Code users can add the marketplace and install the plugin instead. After that you type /start-here and name the business. The skill reads the site and the reviews, writes the brand files the other skills depend on, and tells you which skill to run next. The first run takes about three minutes, per the README. Each client can get a separate folder by telling the router "new client".

What it changes for people building apps

Shipping a page is rarely the hard part; deciding what the page should say is. A pack like this moves the research and copy work into the same agent session where the site is being built. Every skill is designed to run on the agent's own web access with no paid tool. Optional services are listed at the step where they help, with the free path named first: Exa for search, Firecrawl for crawling, DataForSEO for keyword data, Google Ads API, Meta's Ads MCP server, ElevenLabs for voiceover. The README also discloses that boringfunnels, one of the suggested tools, belongs to the author, and states that no skill requires it.

Limits and open questions

The headline says 31 skills and the body says 32; the README does not reconcile the two. The star and fork counts cover a single day of a repository created on 2026-11-16, so they say little about whether the output is any good. The three campaign findings are asserted without a published method or sample size. Output quality will depend on the agent's browsing ability and on how much of the business is legible online. Cursor is named as a supported agent but the documented install paths are the npx command and the Claude Code plugin. Nothing in the source benchmarks these skills against a human marketer or tests them outside local service businesses.

What it means for building apps with AI

Skills like these produce briefs, copy and plans, not deployed pages. The handoff is a prompt: brand files and a page draft that a builder turns into something live. Draftship's guide to writing effective prompts covers that step, since the quality of a generated page tracks how specific the input is. The full skill list and install instructions are in the repository's README.

Primary sources

This article was drafted with AI assistance from the primary sources listed below and checked automatically against them. An editor hasn't reviewed it yet.