{"slug": "mr-cmo", "title": "Mr CMO", "summary": "Mrabbi3 released MR-CMO, a Claude Agent Skill that provides a structured chief marketing officer workflow, including product research, foundational marketing documents, and SEO/GEO/content execution, with a claims boundary to prevent unsupported claims. The skill is available as a plugin or manual install for Claude Code and requires no dependencies or API keys.", "body_md": "**A fractional CMO, as a Claude skill.**\n\nResearch the product from primary sources → write five foundational marketing documents → execute SEO, GEO, content, and distribution against them.\n\nA [Claude Agent Skill](https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview)\nthat gives Claude a working method for acting as your chief marketing officer — not a\nprompt saying \"you are a marketing expert,\" but the actual sequence a good CMO follows,\nplus the discipline that keeps the output true.\n\nAsk Claude to be your CMO, or ask for any single piece of marketing work, and it will:\n\n**Ground itself in what your product actually is**— reading your codebase, schemas, and design tokens where available, not just scraping your homepage.** Write five context documents**into`marketing/`\n\n. Everything afterward references them.**Check whether your funnel can even receive traffic** before making any.**Ship discovery infrastructure**—`llms.txt`\n\n, structured data, technical SEO.**Run a content engine** aimed at territory your competitors aren't defending.**Draft distribution**— and tell you plainly what it can't send for you.\n\nAsk a language model for marketing copy and you get something that reads beautifully and contains claims nobody can support — invented customer counts, uptime figures, certifications you don't hold. It looks like great work right up until a real buyer checks one of the numbers.\n\nMR-CMO is built around preventing that. Its central mechanism is a **claims boundary**:\nan explicit written list of what may and may not be said about your product, derived\nfrom your actual product surface and legal posture, that every downstream asset is\nchecked against. If a claim can't be traced to something real, it doesn't ship.\n\nThe second idea is **sequencing**. Most marketing effort is wasted not on bad tactics but\non good tactics in the wrong order — content driven to a page that doesn't explain the\nproduct, positioning invented before anyone looked at the competitors. The phases are\nordered, and each gates the next.\n\n**Install as a plugin** (recommended — updates come to you):\n\n```\n/plugin marketplace add Mrabbi3/MR-CMO\n/plugin install mr-cmo@mr-cmo\n```\n\n**Or copy the skill in manually.** For one project:\n\n```\ngit clone https://github.com/Mrabbi3/MR-CMO.git /tmp/mr-cmo && \\\n  mkdir -p .claude/skills && \\\n  cp -r /tmp/mr-cmo/skills/mr-cmo .claude/skills/mr-cmo\n```\n\nFor every project:\n\n```\ngit clone https://github.com/Mrabbi3/MR-CMO.git /tmp/mr-cmo && \\\n  mkdir -p ~/.claude/skills && \\\n  cp -r /tmp/mr-cmo/skills/mr-cmo ~/.claude/skills/mr-cmo\n```\n\nStart a new session. Claude picks it up automatically when the work calls for it, or you\ncan invoke it directly with `/mr-cmo`\n\n.\n\nRequires[Claude Code]or any client that supports Agent Skills. No dependencies, no API keys, no build step — it's Markdown.\n\nIt triggers on ordinary requests. You don't need to name it:\n\n```\nBe our CMO. Figure out what we do and build us a marketing plan.\n\nWho are our real competitors and how should we position against them?\n\nWrite me an llms.txt so we get cited by AI assistants.\n\nWe have no blog and no traffic. Where do we start?\n\nDraft a launch plan for the feature we're shipping next month.\n```\n\n**Best first move:** point it at your codebase and ask for the context documents.\nEverything else it does is better afterward, and it only has to happen once.\n\nGive it whatever you have — repo access, the live URL, competitor names, existing brand guidelines. It works from a URL alone, but code access is what lets it describe your product from your data models rather than your marketing copy. That's the difference between a generic plan and a specific one.\n\n```\nyour-project/\n└── marketing/\n    ├── README.md                     Index, strategy in one sentence, the blocker\n    ├── 01-product-information.md     What it is; what may and may not be claimed\n    ├── 02-competitor-analysis.md     The real map, and the gap in it\n    ├── 03-marketing-strategy.md      Positioning, audience order, phasing, goals\n    ├── 04-brand-voice-and-design.md  How it sounds; your real design tokens\n    └── 05-content-strategy.md        Pillars, search and GEO targets, the queue\n```\n\nThen, as the work continues: `llms.txt`\n\n, structured data, page copy, articles, social\ndrafts, and technical SEO changes committed to your codebase.\n\nSix phases, ordered because each is worthless without the one before it.\n\n| Phase | Name | Question it answers |\n|---|---|---|\n| 0 | Ground truth | What is this thing, really? |\n| 1 | Context documents | What may we say, and where can we win? |\n| 2 | Legibility gate | Can the funnel receive traffic yet? |\n| 3 | Discovery infrastructure | Can search engines and AI assistants find and quote us? |\n| 4 | Content engine | What do we publish, in what order? |\n| 5 | Distribution | Where do we show up, and who presses send? |\n\nYou rarely do all six in one session. The skill works out where you are, does that phase properly, and tells you what comes next.\n\n```\nskills/mr-cmo/\n├── SKILL.md                        The method: six phases + standing disciplines\n└── references/\n    ├── context-documents.md        Structure and templates for the five documents\n    ├── positioning.md              The empty-column method; audience ordering\n    ├── seo-geo.md                  llms.txt, structured data, AI-citation tactics\n    ├── content-engine.md           Pillars, cadence, editorial standards\n    └── voice.md                    Voice extraction, banned words, register by surface\n```\n\n`SKILL.md`\n\nloads when the skill triggers; reference files load only when the current\nphase needs them. So asking for an `llms.txt`\n\ndoesn't drag positioning theory into\ncontext.\n\nEven if you never install this.\n\nPositioning is a research output, not a brainstorm. Build a matrix with competitors as\nrows and **stages of your customer's actual journey** as columns — not feature\ncategories, which produce a checklist the biggest competitor always wins. Then find the\ncolumn where every competitor has a gap.\n\nIt's usually adjacent to the category rather than inside it, because categories converge on the same product boundaries and nobody revisits them. That adjacency is what makes it defensible: a competitor can't close it with a feature release.\n\n**→ Full worked example** — a fictional product, its\ncapability matrix, and the wedge that falls out of it.\n\nThe skill opens by telling you what it **cannot** do — it can't publish to accounts it\ndoesn't hold, can't run outreach without a contact database, can't buy media. Tools in\nthis space market themselves as autonomous marketing departments. Knowing the boundary\non day one is worth more than discovering it in week three.\n\n**Does this work outside Claude Code?**\nAnywhere Agent Skills are supported. It's plain Markdown with no dependencies, so you\ncan also just read `SKILL.md`\n\nand paste the method into any assistant.\n\n**What's the difference between the plugin and the manual install?**\nSame skill. The plugin install tracks this repo, so you get updates with\n`/plugin marketplace update mr-cmo`\n\n. The manual copy is a snapshot you own.\n\n**Does it need access to my codebase?**\nNo, but it's much better with it. Reading your schemas tells it what your product\nactually does, unfiltered by your own marketing.\n\n**Will it publish to my social accounts?**\nNo. It drafts; you send. That's deliberate — see honest capability accounting above.\n\n**Is this only for B2B software?**\nThe method generalizes, but the current reference material is strongest for organic\nacquisition on software products. Paid, lifecycle, ecommerce, and local marketing are\nopen contributions — see [CONTRIBUTING.md](/Mrabbi3/MR-CMO/blob/main/CONTRIBUTING.md).\n\n**Won't it just make things up like every other AI marketing tool?**\nThat's the specific failure it's built to prevent, via the claims boundary. It isn't a\nguarantee — verify claims before you publish them. But the method makes every claim\ntraceable to a document you can check, which is a great deal more than \"write me a\nlanding page\" gives you.\n\nIssues and pull requests welcome. The most valuable contribution is a case where the\nguidance produced **bad output** — that's how failure modes get found.\n\nAlso wanted: channels not yet covered (paid, lifecycle, video, ecommerce, local), sharper\nreference material, and worked examples. See [CONTRIBUTING.md](/Mrabbi3/MR-CMO/blob/main/CONTRIBUTING.md).\n\n[MIT](/Mrabbi3/MR-CMO/blob/main/LICENSE). Use it commercially, modify it, redistribute it.\n\nIf this saved you a week of marketing work, a ⭐ helps other people find it.", "url": "https://wpnews.pro/news/mr-cmo", "canonical_source": "https://github.com/Mrabbi3/MR-CMO", "published_at": "2026-09-03 00:46:58+00:00", "updated_at": "2026-09-03 01:22:25.949301+00:00", "lang": "en", "topics": ["ai-tools", "ai-agents", "developer-tools"], "entities": ["Mrabbi3", "MR-CMO", "Claude", "Claude Code", "Anthropic"], "alternates": {"html": "https://wpnews.pro/news/mr-cmo", "markdown": "https://wpnews.pro/news/mr-cmo.md", "text": "https://wpnews.pro/news/mr-cmo.txt", "jsonld": "https://wpnews.pro/news/mr-cmo.jsonld"}}