Claude Code Skills for Marketing: How to Build One

How to create Claude Code skills for marketing: the SKILL.md format, example marketing skills, and where skills live on disk.

How is a skill different from a prompt or a slash command: a comparison table from the post
How is a skill different from a prompt or a slash command

The question this page answers: how do I create Claude Code skills for marketing?

Claude Code skills are folders with a SKILL.md file that teach the AI a specific job. A marketing skill can write ad variants, audit a landing page, or draft a campaign brief on command. Skills load only when asked, so you can install dozens without slowing anything down. Building your first one takes minutes and needs no code.

Most guides on Claude Code skills are written by developers for developers. This is the operator version. I run a local SEO agency, and the automation that changed my week was putting follow-up on autopilot. That is what a skill file did first for me.

TLDR

  • Skills are folders with a SKILL.md file that teach Claude Code one job.
  • A description line controls when the skill loads, so dozens of skills cost nothing until called.
  • The first one takes minutes, needs no code, and starts with one repeated workflow.
  • I run my own outreach on skill files. Roughly 200 form submissions a day go through it.

What are Claude Code skills?

A Claude Code skill is a folder with a SKILL.md file that Claude loads on demand. It is a named job, and the folder holds everything the AI needs to run it. Skills stay off until the work matches, then load in full.

A skill is a named job the AI can run, not a feature you turn on.

The definition comes straight from the Claude Code skills documentation. Two behaviors make skills different from ordinary instructions. First, the load is on demand. Claude reads the description and pulls the skill in only when the task matches. Second, the format is standard. One folder, one SKILL.md file, and the same file works anywhere Claude Code runs.

Claude Code ships with several bundled skills, including /code-review, /debug, and /doctor. The bundled skills section says they run the same way any skill runs, by typing / followed by the name. You do not need to build anything to see the pattern in action.

How do I write a Claude Code skill for marketing?

Start with one task you repeat. Create a folder, add a SKILL.md file, and write the job in plain steps. The description line tells Claude when to use it. No code is required for the basic version.

Write the description like a promise: what it does, and when to call it.

The recipe is the same every time. One folder per skill. One SKILL.md inside it. A description that says what the skill does and when it applies. Instructions after that say how. The Claude Code skills documentation describes the description as the field that decides when the skill loads, so that one line does most of the work.

The first skill I automated was follow-up. My own outreach system pulls in roughly 200 form submissions a day, and every reply used to be written by hand. Now the sequence is a skill file with my tone and the steps in it. I build free websites for prospects before they pay, and that workflow is next in line for a file. The mechanics of the full walkthrough live in how to create Claude Code skills.

Three questions tell me if a workflow earns a skill file. Do I do it weekly? Are the steps the same every time? Can the rules be written down? Three yeses means build the skill. Anything else is a prompt.

What is the SKILL.md format for Claude Code skills?

SKILL.md is a markdown file with YAML frontmatter at the top and instructions below. The frontmatter holds name, description, and optional controls like allowed-tools and disable-model-invocation. Claude reads the description to decide when the skill applies. Everything after the frontmatter is the job itself.

The frontmatter is the trigger. The body is the job.

Here is the format in tutorial form. The file starts with --- on the first line, then the fields, then the closing ---, then the instructions. The Claude Code skills documentation is explicit that the frontmatter is read only when the opening --- is the file’s first line. Put anything before it and the whole file becomes content.

Descriptions have a hard limit. The combined description and when_to_use text is truncated at 1,536 characters in the skill listing, so the key use case goes first. The listing itself scales at 1% of the model’s context window. Keep descriptions short and a shelf of skills stays cheap to carry. I write the description first, before the instructions, because the trigger shapes the job.

The format follows the open Agent Skills specification. That spec is what makes a skill folder portable across the tools that support it, and it defines progressive disclosure in three stages: discovery, activation, and execution. Claude Code accepts the spec fields, and skills built that way load in Claude Code without changes. It is Anthropic’s coding tool, and its skills documentation is part of the broader docs site.

How is a skill different from a prompt or a slash command?

A prompt is a one-time instruction. A slash command is a shortcut you run by hand. A skill is a named job that loads on its own when the work matches its description, with its own instructions and tools attached.

A prompt answers a question. A skill holds the whole job.

Type What it is When it runs
Skill A named job in a folder with its own instructions Loads on its own when the work matches its description
Prompt A one-time instruction you type Runs once, then gone
Slash command A shortcut you invoke by hand Runs when you type /name

The naming rule matters. In a personal or project skill, the command comes from the directory name, not the name field. The name field only sets the display label in listings, per the Claude Code skills documentation. So ad-variants/ in the skills folder is invoked as /ad-variants even when the frontmatter calls it something friendlier.

I reach for a prompt when a task is new and a skill once it repeats. The prompt answers today’s question. The skill answers the next hundred.

What do working Claude Code marketing skills look like?

Working marketing skills cover conversion optimization, SEO, copywriting, analytics, and growth engineering. A single skill can audit a landing page, draft ad variants, or score a piece of content. Public collections on GitHub show the pattern, and you can build your own in minutes.

The fastest route is not writing a skill. It is copying one and editing the description.

The marketing skill collections on GitHub ship ready-made skill files for those five areas. Each one is a folder with a SKILL.md file in the same format the docs teach, so you can install them as-is or copy the pattern into your own. They are MIT-licensed and built to run in Claude Code.

Anyone selling a decade of agent experience is selling you something. I have not deployed agents for paying clients yet. I run them on my own business first, and this site is the proof. The AI SEO team writing this post runs as skill files, one per role, from keyword research to publishing. That is the honest answer to whether the format holds up in real work, and it is why I trust it enough to sell it: the skills bundle packs marketing, SEO, and copywriting skill files, all skills for $99. The fuller picture for a whole marketing operation is AI agents for marketing agencies.

The AutomateReal skills page

AutomateReal skills

The workflow test decides what belongs in a skill file. I use it before building anything new:

Workflow Skill-worthy? Why
Ad variants for one product Yes Repeatable, rule-based, carries your tone
Landing page audit Yes Same checklist every time
One-off campaign brainstorm No Each one is different, so a prompt is enough
Follow-up after a form submission Yes Same structure daily, new details each time

How do I install, store, and list skills in Claude Code?

Personal skills live in ~/.claude/skills, project skills in .claude/skills, and synced claude.ai skills in ~/.claude/skills/synced. Claude picks them up at startup. Type /skills to list what is installed.

Skills are folders. Install means putting a folder in the right place.

Storage is two default paths, per the configuring skills section of the documentation. Personal skills go in ~/.claude/skills. Project skills go in .claude/skills, which means they travel with the repo and load for anyone who clones it. Extra directories work through the --add-dir flag.

Location Scope Loads when
~/.claude/skills Personal Every session, for you
.claude/skills Project Every session in that repo
Any directory Custom You, via --add-dir
~/.claude/skills/synced claude.ai sync When you enable the skill on claude.ai

Two behaviors save real time, and both sit in the configuring skills section. Starting Claude in a subdirectory still picks up skills defined at the root, and --add-dir points it at paths outside the defaults. To list what is installed, run /skills. Synced skills show up there under claude.ai sync, and a local skill with the same name overrides the synced copy, so the two never collide silently.

Plugins ship skills too, and they come namespaced. A skill in my-plugin/skills/ runs as /my-plugin:skill-name, so a plugin can never shadow one of yours. If two project skills share a name, Claude Code namespaces the deeper one by path, for example apps/web/.claude/skills/deploy/SKILL.md becomes /apps/web:deploy. Every skill follows the same shape: ~/.claude/skills/<skill-name>/SKILL.md.

FAQ

Do skills work with Claude Code subagents?

Yes. Add context: fork to the frontmatter and the skill runs in its own subagent with a clean context. The skill content becomes the prompt that drives that subagent, and the result comes back when it finishes.

How do I write a skill that calls a tool or a script?

List the tools in allowed-tools in the frontmatter. Claude then calls them without asking permission during the turn that invokes the skill, and the grant covers MCP tools too. Scripts live in the skill folder and run from the skill body. The grant clears when you send your next message.

Where are Claude Code skills stored on disk?

Personal skills go in ~/.claude/skills. Project skills go in .claude/skills. Synced claude.ai skills go in ~/.claude/skills/synced. Each skill is its own folder with a SKILL.md file inside. Deleting the folder removes the skill.

Are there ready-made Claude Code skills for marketing?

Yes. Ready-made marketing skills cover conversion optimization, SEO, copywriting, analytics, and growth engineering, in the same SKILL.md format you would write yourself. You can install them directly or copy the pattern into your own folders.

The skills bundle is the same idea, all skills for $99.

Can I use the same skill in Claude Code and claude.ai?

Skills you enable on claude.ai download into ~/.claude/skills/synced and load in Claude Code automatically. A local skill with the same name overrides the synced copy. The /skills menu groups synced skills under claude.ai sync, so you can tell where each one came from.

Close

The fastest way to learn the format is to build one skill this week. Pick the workflow you repeat most, write the description like a promise, and let the folder do the rest. If you want help finding that first workflow, a discovery call maps it in about thirty minutes.

The AutomateReal services page

AutomateReal services