Email & lifecycle skills for AI agents

No. They plan and write the emails and check deliverability; you send them from your own email tool, such as Mailchimp, Klaviyo or Resend.

That keeps your lists, consent records and unsubscribes in one place, where your email tool already handles them.

Yes. SMS & WhatsApp marketing builds compliant welcome, abandoned-cart, post-purchase and win-back flows, with the opt-in and timing rules each channel needs.

Retention & churn prevention covers the rest of the customer lifecycle: health signals, cancel flows, save offers and failed-payment recovery.

Yes. All 6 are open source under the MIT licence, so you can use, change and share them, including for client work, without paying or asking permission.

Where a skill adapts someone else's open work, its page credits the source and the licence. Repos with a copyleft licence are only used for ideas, and repos with no licence aren't used at all.

Claude Code installs them from the cmo.so marketplace. Any agent that reads SKILL.md folders, such as Codex or Cursor, can use them by copying the folder into its skills directory.

The core skills also work in Claude, ChatGPT and other MCP clients through the cmo.so MCP server, so you can use whichever assistant you already have.

In Claude Code, add the marketplace with /plugin marketplace add AI-Marketing-OS-LTD/cmo-so-skills, then install the group with /plugin install cmo-so-email@cmo-so. That adds all 6 skills in one go.

Ask in plain words afterwards and the agent picks the skill that fits. For other agents, copy the folders of the skills you want from GitHub into the agent's skills directory.

No. The Email & lifecycle skills work on their own with whatever your agent can already see and do: your files, the pages you share and the data you paste in.

A free cmo.so account only matters if you want Maggie to plan, write and publish for your website too. The core skills connect your agent to it, and every change is shown to you before it happens.

No. Ask for the outcome you want in plain words and the agent picks the right skill, asks for anything it's missing and follows the skill's steps.

If you'd like a starting point, the section above lists an everyday job for every skill on this page with a prompt you can copy. Words in a highlight are the ones to change for your business.

A prompt is one request. A skill is the method behind the job: the questions to ask first, the order of the steps, the quality checks and the format of the result.

That's what makes the answer reliable. The skill stops the agent skipping research, inventing numbers or handing back something you can't use, and you can read every step it will take.

Yes. A skill is a plain Markdown file, so you can edit its steps, add your own checks and examples, and keep your version in your own repo. The MIT licence allows it.

If your change would help others too, suggest it on the skills page or open a pull request on GitHub, and it may be folded into the library for everyone.

No. A skill only guides your own agent; your files, documents and data stay wherever your agent and its tools keep them.

cmo.so only sees what you connect to cmo.so, for example when you use the core skills through the MCP server with your account.

They're maintained in the public cmo-so-skills repo and synced to this site, so new and improved skills show up here as soon as they ship. In Claude Code, update the marketplace to pick them up.

Upvotes and downvotes show us which skills people rely on, and those are the first we improve.

Suggest an open-source repo on the skills page: Maggie reads it, checks the licence and decides whether to add it as a new skill, use it to improve one, or put it to a vote.

You can also browse the other categories below; many jobs span more than one, and the agent can combine skills in a single request.