# GEO Skill Pack Product page: https://www.winstondigitalmarketing.com/products/geo-skill-pack/ Price: $497 USD // product / geo skill pack # The Skills we run on client GEO work, packaged for your agency. Three Claude Skills that produce the deliverables of a generative-search engagement: a 22-point AI visibility audit that builds its own Excel workbook, a content rewrite that comes out as a branded Word document, and the client strategy deck. Your colors, your agency name, unlimited clients. $497, one payment [ Buy the Skill Pack ](https://winstondigital.lemonsqueezy.com/checkout/buy/a4b02724-eb04-4035-b782-dafb7b898e1f) [Or have us build yours ](/services/ai-marketing/custom-gpts-and-skills/) Checkout runs on Lemon Squeezy. They handle the payment and the tax, and the files download straight after you pay. Refunds within 14 days, see the terms. ## Every SEO agency is getting the same question. What are you doing about AI search? Most of the answers are a blog post and a slide. The ones that hold up are a deliverable somebody paid for. This pack is three Claude Skills that produce those deliverables. It is the working version of the process we run on client engagements, packaged so you can run it on yours. ## What is in the box. // geo-tech-audit ### The audit You give it a domain. It works through 22 checks across 8 sections: whether AI crawlers can reach the site, whether they can read it once they do, whether the structured data says what the business is, whether the brand exists anywhere a model can verify it. Then it builds a two-tab Excel workbook. Tab one is the full audit. Tab two is the client's to-do list, grouped by priority, with an owner on every row. // geo-content-rewrite ### The content rewrite Point it at a live URL. It applies a 13-point chunking framework: key takeaways at the top, headings rewritten as the questions people type, an answer capsule under each one, comparison tables, an FAQ, schema recommendations. It comes out as a Word document with your client's brand on it, because a .docx survives a client review process and a chat transcript does not. // geo-strategy-deck ### The strategy deck Fifteen to twenty slides: site-wide recommendations plus a before-and-after for the client's five best blog posts, with the implementation checklist, schema reference, and content priority tracker as spreadsheets. The deck gets approved. The checklist gets executed. One more file carries across all three. brand-config.json holds your agency name, your fonts, and nine hex colors. Set it once and every workbook, document, and slide comes out in your palette with your name on it. ## What a prompt does not give you. A prompt gets you a first draft. These are procedures. Each skill carries its own reference material, the code patterns that produce clean output, and the specific failure modes already written down: the setting that turns every table cell black, the one that breaks tables in Google Docs, the duplicate-run bug that prints every paragraph twice. Those took real hours to find. They are in the files, so they cost you none. ## What it does not do. It does not promise citations in an AI answer. Nobody can promise that. The mechanisms are not published, they change, and the same prompt returns different sources on different days. What it does is remove the reasons a system would skip a site: content that only exists after JavaScript runs, no structured data, no entity anywhere a model can verify the brand against, paragraphs that cover four ideas so they answer none. Every deliverable in the pack is framed as "here is what is broken and here is the fix" rather than "do this and you will be cited." If you want a tool that guarantees a spot in ChatGPT's answer, buy that from someone willing to say it. ## Who it is for, and who it is not for. Agencies and in-house teams who already sell SEO and need a scope of work instead of an opinion. It assumes you know SEO. It does not assume you know how to get a formatted .docx out of an LLM without the tables turning black. It is not for you if you want a tool that scores a site automatically and emails you a PDF. These skills run inside Claude and expect a person with judgment driving them. You write the audit findings. The skill runs the checks and formats the result. The plain-language finding is the part clients pay for, and it is the part that needs your name on it. ## What is included. - Three skills, installable in Claude Code, Claude Desktop, or claude.ai - The workbook build script and a working example findings file to copy - brand-config.json : your colors, fonts, and agency name, set once - Install guide and a quickstart for the first run - One agency license, unlimited clients, unlimited use ### Requirements - Claude Code, Claude Desktop, or claude.ai with skills enabled - Python 3 with openpyxl for the audit workbook - Node with pptxgenjs and docx for the deck and the documents ### License One agency, unlimited clients. Install it on every machine your team uses, run it for as many engagements as you want, sell the deliverables at whatever price you set, and modify anything in it. What you cannot do is resell or republish the pack itself. The work you produce with it is yours. The pack is not. The frameworks reflect how generative search behaves as of the version date, that behavior changes, and the pack is sold without warranty. It is a process for removing the technical and structural reasons a system would skip a site, not a guarantee of a result. ## Related reading. - [Claude Skills for SEO teams: the in-house build guide](/playbooks/claude-skills-for-seo-teams/). What a skill is and which workflows to package first, if you would rather build your own. - [The complete GEO audit methodology](/playbooks/the-complete-geo-audit-methodology/). The framework the audit skill automates. - [The internal audit skill this pack came from](/our-work/winston-geo-tech-audit-skill/). ## Connecting it to real data. GEO work is measurement-poor. There is no Search Console for ChatGPT, you cannot see the queries, and the click that used to prove the work happened is the exact thing an AI answer is built to prevent. So the job is assembling a picture from sources that were never designed for it. The pack ships a reference on doing that with connectors instead of browser tabs: the two free ones that change what Claude can do, Google's official read-only Analytics server, and the fact that no official Search Console server exists whatever the published advice implies. Then honest pricing on the paid tools, including which ones are a rounding error at audit volume and which you should not buy at all. Plus the client-data rules that apply when the account is not yours. It also states the finding to expect before you run a single audit. For most categories the cited sources are large publishers, Reddit, and established directories. Not your client. That is not a failed audit, it is the result, and it means the lever is off-site authority rather than another page on their site. Saying that plainly is worth more than another list of on-page fixes, and it is what most competitors will not tell them. ## Want the Skills we run on client work? $497, one payment. One agency license, unlimited clients. If you would rather we build a skill around your own process instead, that is the other door. [ Buy the Skill Pack ](https://winstondigital.lemonsqueezy.com/checkout/buy/a4b02724-eb04-4035-b782-dafb7b898e1f) [Or have us build yours ](/services/ai-marketing/custom-gpts-and-skills/) ## Frequently asked questions ### What is in the GEO Skill Pack? Three Claude Skills that run the work rather than describe it: a technical GEO audit, a content rewrite, and a strategy deck. Each carries its own reference frameworks, the code patterns that produce clean output, and the specific failure modes already written down, including the setting that turns every table cell black and the duplicate-run bug that prints every paragraph twice. It also ships a brand-config file so every deliverable comes out in your agency's name, an install guide, a quickstart, and a walkthrough. It is a folder of files you install, not software you log into. ### Will this get my client cited in ChatGPT? No, and anyone promising that is selling you something they cannot deliver. The retrieval mechanisms are not published, they change, and the same prompt returns different sources on different days. What the pack does is remove the reasons a system would skip a site: content that only exists after JavaScript runs, no structured data, no entity anywhere a model can verify the brand against, paragraphs covering four ideas so they answer none. Every deliverable is framed as what is broken and how to fix it, rather than as a promise about an outcome nobody controls. ### Can I sell the deliverables to my clients? Yes. The licence is one agency, unlimited clients. Install the skills on every machine your team uses, run them for as many engagements as you want, sell what they produce at whatever price you set, and modify anything: edit the frameworks, change the slide order, rewrite the prompts, add your own checks. What you cannot do is resell or redistribute the pack itself, publish the skill files or reference frameworks publicly, or repackage the contents into another product for sale. The work you produce with it is yours. The pack is not. ### Can I put my own branding on the output? That is what brand-config.json is for. Set your agency name, URL, the prepared-by line, your logo path, fonts and colours once, and every deck and document the skills produce comes out in your branding rather than ours. Nothing in a client-facing deliverable says Winston unless you want it to. ### How is this different from a prompt library? A prompt gets you a first draft. These are procedures. Each skill carries its reference material, the formatting code that produces a clean .docx or deck, and the failure modes that took real hours to find written down so they cost you none. The difference shows up on the tenth run rather than the first: a prompt library gives you ten different-looking outputs, and a skill gives you the same deliverable every time, which is what you can actually put a scope of work and a price against. ### What do I need to run it? Claude, and a person with SEO judgment driving it. The skills run inside Claude and expect someone who knows the work; they do not score a site automatically and email you a PDF. You write the audit findings, the skill runs the checks and formats the result. The plain-language finding is the part clients pay for and the part that needs your name on it. ### Who is it not for? Anyone wanting a tool that audits a site unattended and produces a report without a practitioner involved. It assumes you already sell SEO and know how to interpret what the checks return. What it does not assume is that you know how to get a properly formatted document out of an LLM without the tables turning black, which is the part it solves.