Most prospects who book a call with us have the same scar tissue. They paid an SEO agency a thousand a month for a year, got two unedited blog posts dropped into a Google Doc, and a monthly ranking report full of keywords nobody searches for. When they asked the agency to explain the methodology, the answer was some version of "trust us, we know SEO." The methodology turned out to be a junior content writer working from a template and a Google Doc that hadn't been updated since 2019.
I bring this up because the wedge we use in that conversation is structurally hard to copy. Ours is on GitHub. The five skills that produce every piece of content and every audit we ship are open source, dated, citable, and readable in an evening. A prospect can read them tonight and sign tomorrow if the substance holds up, which is the order I'd want if I were on the other side of the table.
The black box problem
The default agency posture is opacity, which works until it doesn't. The 30-page strategy deck, the proprietary "framework" with the trademark symbol on it, the methodology that's always described but never shown, the monthly report that tracks vanity metrics because the harder ones would be embarrassing. None of it is malicious. Most agencies were good at 2018 SEO, they hired juniors during the cheap-capital years, and the playbook fossilized while Google's SGE and ChatGPT and Perplexity rewrote what content even needs to do. The retainer kept paying because nobody could verify what was being shipped.
The wedge for me is that the methodology being public is its own answer to the trust problem. A prospect who's been burned doesn't have to take my word for it. They can read the code, watch how it produces output on a sample, run the slop detector on a piece of their existing content, and decide before any money changes hands whether the framework is worth signing for. The contract becomes a formality once the verification has happened upstream.
The five skills, what each one does
Each of these is a working Claude Code skill, published under my GitHub handle, used in production on every Space & Story engagement. The accuracy of the descriptions below is auditable against the repos themselves.
1. mahmouds-seo-writer
An autonomous SEO content writer and auditor. It drafts a post with the target keyword in the title and H1, generates a meta description, drops in two to three internal links to related pages, two to three external links to authoritative sources, Article schema, FAQ schema where it fits, and an auto-generated OG image at 1200 by 630. It also runs as an auditor against existing URLs, which is most of what we do for prospects in week one. The output is publish-ready before any human edits, which sounds like a flex and is closer to a starting line. The human edits are where the taste lives.
2. mahmouds-seo-guide-v3
The knowledge base the writer pulls from. 19 domain reference files distilled from 1,000+ expert sources, covering SEO strategy, AEO and GEO, AI search visibility, content writing and auditing, keyword research, technical SEO, schema, link building, analytics, local SEO, customer research, competitor research, conversion copy, and marketing psychology. The point of the knowledge base is that the writer isn't guessing about what works in 2026; it's pulling from material that's been read, dated, and condensed by a human who's been doing the work for a decade.
3. mahmouds-writing-voice
A voice codex and audit framework. It catches the AI tells that competitors don't catch: the hedging stack ("it may be possible that this could potentially…"), the over-formality, the weak verbs, the "as an AI" cadence that bleeds into ghostwritten content. It also standardizes prose style across projects, which matters when one operator is shipping copy for a builder, a clinic, and a contractor in the same week. The builder's site shouldn't read like the clinic's site shouldn't read like the contractor's site, and the codex is what makes that consistent without me sitting at a keyboard 18 hours a day.
4. ai-slop-detector
A universal prose audit on two axes. The AI-Slop axis answers "does this read like AI wrote it?"; the Comprehension axis answers "can a fresh reader follow this?" It catches 45 AI-slop patterns, ~150 vocabulary tells, ~33 formatting tells, 35 comprehension patterns, and 8 readability metrics, all density-scored so one stray buzzword doesn't sink an otherwise-clean piece. Every blog post we ship runs through it. This one ran through it before you read it, which is probably the most honest disclosure I'll make in a thousand words.
5. mahmouds-reddit-strategist
An autonomous Reddit strategist for AI-citation seeding. Reddit drives roughly 40% of the source data underneath LLM answers, which is why most AI-search wins happen on threads rather than blog posts. The skill drafts comments and posts that survive moderator scrutiny while compounding into AI citations over a 90-day window. We don't post anything automatically; the operator reviews everything before it goes live. The skill is the assistant, not the agent of record.
The three quality gates every piece passes through
The production flow is more boring than the tooling suggests, which is usually true of operating models that work. Every piece of content we ship gets three passes before it goes live:
- mahmouds-seo-writer drafts the piece against the 1,000+ source knowledge base, with schema, internal and external linking, meta description, and the OG image generated.
- mahmouds-writing-voice catches the AI tells the writer didn't catch, because the writer is producing volume and the voice audit is enforcing consistency, and those are different jobs that need different scanners.
- ai-slop-detector does a two-axis scan on the revised piece. Anything that fails either gate gets flagged with quoted instances and severity, and the offending sentences get rewritten before publish.
That's the difference between content that compounds in Google's SGE and ChatGPT search and content that gets downranked because the model recognized it as AI-generated within the first paragraph. The competitor agency next door is shipping unedited ChatGPT output. We're shipping copy that passes a human-grade voice audit and an AI-slop scan on every piece, which is the moat that compounds whether or not the prospect notices on day one.
The 280-source Marketing OS sits above the skills
The skills are the daily-driver tools. The umbrella framework is the Agentic Marketing Operating System, published April 2026 at mahmoudhalat.com/marketing-os, licensed CC-BY 4.0 because I'd rather it be useful than mine. Eight domains, a 32-cell maturity matrix from Crawl to Fly, 42 named case studies (Anthropic plus Clay killing their top legacy data provider, Vercel going under-1% to 10% of new signups from ChatGPT in six months, Refine Labs' $50M HIRO over 24 months, SaaStr's 20-agent stack with 1.2 humans running 70K emails a month, CVS Health and Simile's 100K+ agentic twins on 2.9M consented responses, Klarna's re-balance, and so on).
One of the eight domains is AI Search and Answer Visibility, which is the reason AEO ships on every site we build by default rather than as a $1,000-to-$5,000-a-month add-on the way competitors price it. The Marketing OS is the strategy layer; the five skills are the execution layer. A client who wants the why reads the framework, and a client who wants the what reads the skills. Both are public and dated and live at URLs anyone can cite.
Why we open-sourced the whole thing
I get asked this on every other call, usually with some version of "aren't you just giving competitors a playbook." Four reasons it's the right call:
- Verification before signing. A skeptical prospect can read the exact framework that will touch their content tonight, run the slop detector on a piece of their existing copy in the morning, and sign by lunch. The agency next door can't offer that because there's no framework to show, and the prospect feels the difference even when they can't name it.
- Compounding credibility. The repos are dated, the commits are public, the updates are citable. When the field shifts (which it does, monthly), the methodology moves with it and the movement is visible. Competitors can claim they've updated their playbook; we can show the diff.
- It removes a major objection before it gets raised. "Trust us" is the hardest sell in a market full of burned prospects. "Read it tonight" is one of the easiest. The black box is the friction; transparency is the lubricant.
- Competitors can copy the skills, but they can't copy the operator running them. Taste is non-delegable. The skills produce drafts; a human with a decade of healthcare informatics and a $100K Lovable Shipped grand prize against 5,800 builders decides what ships. The agency that copies the skills tomorrow still has to find someone who knows when to override them.
The way I put it on calls is shorter than this section:
"Most agencies sell you a black box and say 'trust us.' Mine is on GitHub. You can read it tonight before you sign anything tomorrow."
It lands because it's true and because it short-circuits the entire trust conversation. The prospect doesn't have to take a position on whether agencies in general are trustworthy. They just have to spend an evening reading code, and if the substance holds, the rest of the sales process is paperwork.
Where to go from here
If you want to read the skills, they're all at github.com/MahmoudHalat. The Marketing OS is at mahmoudhalat.com/marketing-os, CC-BY 4.0, fork it if it's useful. The case studies on the studio's portfolio show what happens when the methodology is applied to real client work over a 90-day window, which is the timeframe most of the citation and ranking lifts compound on.
If you'd rather skip ahead and have us run the methodology against your existing site, the free audit is at /audit. We pull the Lighthouse scan, the cite-met.com AI-citation report, the isitagentready audit, and a slop scan on a recent piece of your copy. You get the full report whether you end up signing or not, which is the same logic that runs the open-source decision: I'd rather have you read the work than take my word for it.