Home Marketplace Claude Code Skills for Marketing Ops
Claude Code Skills
for Marketing Ops
Generic AI drafts are the easy part. The hard 80% of marketing is operator discipline — platform character limits, UTM naming conventions, honest claims, one consistent voice on every draft. These 8 agent skills build that discipline into every task: repurposing that writes each platform variant from the source asset, ad test matrices with RSA limits enforced at write time, intent-first SEO briefs, CSV-to-scorecard reporting, scored landing-page teardowns, competitor analysis, and a brand-voice linter that gates every draft. Plain Markdown — drop into ~/.claude/skills/ and it's installed.
Instant download · Commercial licence, unlimited client work · Free updates for life · 7-day refund
The problem
The drafts were never
the hard part.
Anyone can get an AI to write a LinkedIn post. What generic prompting skips is everything that makes marketing output usable: the X variant that blows past 280 characters, the Google ad headline that's 34 characters when the limit is 30, three campaigns tagged utm_source=LinkedIn, linkedin and li so the analytics never reconcile, and five drafts in five different voices because nobody wrote the voice down.
That's the unglamorous 80% — and it's exactly what a skill can encode. Each of these eight is a written procedure the agent follows every time: the limits, the naming conventions, the honesty rules, the voice. The discipline stops living in someone's head.
Fixes — the boring 80%
What's inside
Eight skills,
operator discipline.
content-repurposer — every variant from the source
Turns one source asset — a case study, a blog post, a launch note — into platform-native variants for LinkedIn, X, newsletter and more. Its cardinal rule: every variant is written from the source asset, never from another variant. Copies of copies drift; this forbids them. Platform character limits are checked at write time, not after the reject.
Limits — enforced at write
ad-variant-writer — test matrices, not ad soup
Builds structured ad test matrices — angles × hooks × proofs — instead of "give me 10 ads". RSA character limits (30-char headlines, 90-char descriptions) are enforced as the copy is written, and every claim must trace to something true about the product. No invented statistics, no "award-winning" unless there's an award.
Claims — traceable only
seo-brief-builder — intent before keywords
Writes content briefs that start from search intent — what the searcher is actually trying to do — before touching keywords, headings or word counts. The brief specifies the questions the piece must answer and the proof it must contain, so the writer (human or agent) can't produce a page that ranks for nothing.
Then — structure & proof
analytics-reporter — CSV in, scorecard out
Hand it a raw CSV export from your ad platform or analytics tool and it produces a scorecard: what moved, what didn't, and what to do next — with the arithmetic shown. It reports numbers that are in the data and refuses to editorialise beyond them.
Out — decisions + math
utm-governor — one naming convention, forever
Defines and enforces a single UTM naming convention — lowercase, fixed vocabularies for source and medium, campaign name patterns — and tags every link accordingly. The skill that ends LinkedIn vs linkedin vs li and makes attribution reports reconcile.
Makes — reports reconcile
landing-page-critic — a scored teardown
Reviews a landing page against a fixed rubric — message match, headline clarity, proof, friction, call to action — and returns a scored teardown with specific rewrites, not "consider improving your copy". Run it before traffic is bought, not after it bounces.
Timing — before spend
competitor-teardown — structured, not vibes
A repeatable procedure for analysing a competitor's positioning, offer, pricing signals and messaging — organised into a comparison you can act on. Same structure every time, so teardowns from different weeks are actually comparable.
So — comparable over time
brand-voice-linter — the gate every draft passes
Turns your brand voice into a written, checkable standard — words you use, words you never use, sentence rhythm, claims policy — and lints every draft against it. This is the skill the other seven call before anything ships, so one voice survives a hundred drafts.
Keeps — one voice
README, licence and changelog
Install instructions for Claude Code (~/.claude/skills/) and a note that matters: each skill is plain Markdown and reads as a standalone playbook in any Claude interface — claude.ai, the API, or as team documentation. No code to run, nothing to install beyond copying files.
Works in — any Claude
Who it's for
People who ship marketing
without a marketing team.
Solo founders doing their own marketing
You're the CEO and the marketing department. These skills are the process discipline of a senior marketer, applied to every draft, without the hire.
Freelance marketers & consultants
Commercial licence covers unlimited client work. The voice linter and UTM governor make every client account run to a written standard from day one.
Small teams drowning in channels
One case study becomes every platform's version, correctly, in one pass — and every draft leaves in the same voice regardless of who prompted it.
Not a fit if you want a magic growth button. These skills encode process — they make AI marketing output disciplined and usable. They don't invent your strategy, your product's proof points, or your audience.
Honestly
What this can't do.
Skills are process, not strategy. They make execution disciplined — limits respected, voice consistent, links tagged. Deciding what to market, to whom, and with what offer is still your job.
You need a Claude to run them in. Built for Claude Code, where they install as agent skills; readable as playbooks in any Claude interface. Your Anthropic subscription or API usage is billed by Anthropic — there are no fees to me.
The voice linter needs your voice. It ships with the framework and worked examples; the first run is filling in your own words-we-use and words-we-never-use lists. That's an hour of thinking no tool can do for you.
Can't — choose your strategy
Questions
Before you buy.
What is a Claude Code skill, exactly?
A skill is a Markdown file of instructions that Claude Code loads automatically when a task matches it — a written procedure the agent follows, rather than a prompt you paste each time. Install is copying a folder into ~/.claude/skills/. Because skills are plain Markdown, each one also reads as a standalone playbook in claude.ai or any other Claude interface — you can use these without Claude Code at all, just less automatically.
Why does AI marketing content sound generic, and how do these help?
Because generic prompting produces the average of everyone's marketing. The fixes are process: write every platform variant from the source asset instead of from another variant (copies of copies drift), lint every draft against a written brand voice, and only make claims that trace to something true. Those three rules are the content-repurposer and brand-voice-linter skills — applied automatically instead of remembered occasionally.
Do these work with ChatGPT or other AI tools?
They're written for Claude and install as agent skills in Claude Code. The process content — the repurposing rules, RSA limits, UTM conventions, the voice-linting rubric — is plain-English procedure, so you could adapt it to another tool by hand, but the automatic loading and agent behaviour is Claude Code-specific.
How do the skills keep ad copy inside platform character limits?
The limits are written into the skills as hard rules — Google RSA headlines at 30 characters and descriptions at 90, platform post limits for repurposing — and the agent counts as it writes rather than after. You stop discovering length violations at upload time, which is where most AI-written ad copy dies.
Is this a course or a tool?
Neither — it's working process documentation that an agent executes. There are no videos and no software to run: 11 files, 8 of them skills totalling 2,479 words of procedure, plus README, licence and changelog. If you read them and never install Claude Code, you've bought a concise marketing-ops playbook; if you do install them, the playbook runs itself.
Can I use these for client work?
Yes. The licence is single-operator commercial and covers unlimited client projects — run every client's content, ads and reporting through them. You may not resell or redistribute the files themselves. If it isn't useful, reply to your receipt within seven days for a full refund — and you can keep the files.
Install once. Every draft disciplined.
Instant download. Commercial licence, unlimited client work. Free updates for life. Seven-day refund, no justification needed.
Written from marketing ops run in production, not theory
Where this came from
We market our own products across LinkedIn, X, Instagram and YouTube — the repurposing rules, the UTM convention and the voice linter are the ones running our own pipeline. Read why AI marketing content sounds generic — and the process that fixes it, or if you'd rather have marketing automation built for you, that's the day job.