What is leopard627/fire-your-seo-agency?

The repository packages five search-optimization checklists into a Claude Code skill, with Naver coverage that global answer-engine guides skip. It is a checklist runner, not a ranking guarantee, and the README is explicit about that.

Can I use leopard627/fire-your-seo-agency commercially?

Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.

Is leopard627/fire-your-seo-agency still maintained?

Yes. The repository last received commits 22 days ago.

What programming language is leopard627/fire-your-seo-agency written in?

GitHub does not report a main language for this repository.

fire-your-seo-agency: Claude Code SEO and GEO skill

At a glance

What is it?: The repository packages five search-optimization checklists into a Claude Code skill, with Naver coverage that global answer-engine guides skip. It is a checklist runner, not a ranking guarantee, and the README is explicit about that.
Who is it for?: Adopt it if you run Claude Code, own a site whose technical base you can actually change, and want a repeatable audit with a scheduled re-measurement instead of a monthly retainer. Skip it if you need someone to execute content strategy or link acquisition, since the skill explicitly refuses those, and if you cannot act on the findings yourself.
Can I use it commercially?: Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?: Yes. The repository last received commits 22 days ago.
What is it written in?: GitHub does not report a main language for this repository.

Answers come from the project's GitHub data, last synced on September 16, 2026, and from our analysis. They are not legal advice.

DEEP OPEN-SOURCE ANALYSIS

Hysen Labs Editorial · September 17, 2026 · leopard627/fire-your-seo-agency

The retainer it is trying to replace

The pitch is blunt: agencies charge monthly for "AI-era search optimization" and "guaranteed ChatGPT citations", and most of that work is public standards plus a repeatable checklist. The README's framing is that the same checklist executed by an agent is a skill rather than a retainer. That is a positioning claim, not a measured one, and it is worth separating from the artifact itself.

The artifact is a Claude Code skill. It is aimed at people who already run Claude Code, own or control a site, and can change the technical layer: server rendering, sitemaps, meta tags, structured data, robots.txt, llms.txt. If you cannot touch those files, the audit output is a list of things you will hand to someone else. The skill also assumes you can read a scorecard and decide priorities, because the README describes it as proposing priorities rather than executing a fixed plan.

The evidence offered is a single site: Chickenstock, a solo-built Korean stock research service, with 1.54M search impressions in 30 days and 7.4K clicks, plus pages cited by Naver's AI Briefing. That is one operator's result on one domain. It is not a benchmark, and the README does not present it as one.

Five lanes, and why NEO is the differentiator

The skill splits search optimization into five lanes, each mapped to a different engine and a different question. SEO covers Google and Bing crawlers and asks whether they can read and index the content at all. AEO covers Google AI Overviews and Bing Copilot and asks whether the answer box cites you. GEO covers ChatGPT, Perplexity and Claude and asks whether generative systems treat you as the primary source when they browse. LLMO covers the model's own knowledge and asks whether it knows your brand, and knows it correctly. NEO covers Naver search and AI Briefing.

The NEO lane is the honest differentiator. Naver holds a large share of the Korean market, and the README states that global AEO guides ignore it entirely. That gap is real for anyone serving Korean users, and it is the part of this skill that has no obvious equivalent in English-language tooling. For a site with no Korean audience, NEO is dead weight you will read past.

The lanes also overlap in ways the table does not resolve. AEO and GEO both depend on machine-readable structure, and LLMO depends on consistent entity naming that also affects structured data. The README does not describe how conflicts between lanes are resolved when, for example, a brand-name normalization for LLMO changes the schema markup that AEO relies on. The agent is expected to sequence that itself.

How the skill is structured on disk

The repository layout is the mechanism. SKILL.md holds the agent operating procedure, described as audit, implement, measure. The references directory holds one document per lane: seo.md for the technical checklist and real-world traps, aeo.md for answer engine optimization including Bing Webmaster Tools, AI Overviews, Copilot and E-E-A-T, geo.md for AI crawler policy, llms.txt and primary-source positioning, llmo.md for brand entity work, neo-naver.md for Search Advisor, AI Briefing and the blog two-track, and measure.md for the measurement loop.

One detail matters more than it looks. The Korean documents under references are canonical, and the agent reads those. The references/en directory mirrors them in English for human readers. So an English-speaking operator reading the repo is reading a translation that the agent does not use. If the two drift, the agent follows the Korean source. The README does not state whether the mirrors are generated or maintained by hand, which is the thing to check before trusting an English reference doc as the authoritative checklist.

The .claude-plugin directory holds the plugin and marketplace manifests that make the one-command install work. The audit itself is described as reading the site the way a crawler does, without JavaScript, which is the constraint that makes the SEO lane meaningful: if your content only appears after client-side rendering, the audit will score it as absent.

Installing it and running a first audit

The README gives two install paths. The recommended one is the plugin route, which the v1.1.0 release notes list as /plugin install support. Two commands register the marketplace and install the skill:

bash

/plugin marketplace add leopard627/fire-your-seo-agency
/plugin install fire-your-seo-agency@fire-your-seo-agency

The alternative is a git clone, either into a single project or into your personal skills directory so every project can use it:

bash

# As a project skill (this project only)
git clone https://github.com/leopard627/fire-your-seo-agency.git .claude/skills/fire-your-seo-agency

# Or as a personal skill (every project)
git clone https://github.com/leopard627/fire-your-seo-agency.git ~/.claude/skills/fire-your-seo-agency

With the skill present, the README shows the invocation as a slash command with a natural-language target:

text

/fire-your-seo-agency audit my site

What you should expect back is a scorecard before any changes are made. The README's example shows five rows, one per lane, each with a status marker and a piece of evidence. The SEO row in that example reads that the body is server-rendered but 214 detail pages are missing from the sitemap. The AEO row reports no direct-answer first paragraphs and zero FAQ structured data. The GEO row reports no llms.txt and an undecided GPTBot and PerplexityBot policy in robots.txt. The LLMO row reports the brand name spelled three different ways across surfaces. The NEO row reports the site is not registered in Naver Search Advisor.

That example is illustrative, not a template for your output. The point is the shape: evidence per lane, then proposed priorities, then implementation, then a scheduled re-measurement. If you get a scorecard with no evidence attached to a failing lane, the audit did not do what the README describes.

What it will not do, and where that bites

The refusal list is the most concrete part of the README. No buying backlinks, engagement pods or content spam. No guaranteed-ranking promises, framed as no claims without measurement. No keyword stuffing, hidden text or cloaking. The stated philosophy is that AI cites accurate data rather than good writing, so the goal is to become the primary source for a number and let citations follow.

That refusal is a real limitation, not a marketing line. Link acquisition and digital PR are where a large share of competitive SEO effort actually goes, and this skill will not touch them. If your site is technically clean and your competitors outrank you on authority, the audit will come back mostly green and you will have learned little. The skill is diagnostic and structural. It is not a growth channel.

The no-JavaScript crawl is the second failure mode. Sites built as client-rendered single-page applications will score poorly on the SEO lane, and correctly so, but the fix is a rendering change that may be outside what you can do quickly. The skill will tell you the content is invisible to crawlers. It will not migrate your frontend.

There is also a measurement honesty problem the README acknowledges indirectly. The measure.md reference exists because, in the README's phrasing, fixing it is not the finish line. That is a warning that an audit can look successful while nothing changes in search results. The skill schedules a re-measurement, but the README does not document rollback for changes the agent makes to your files. Version control is on you.

How it differs from an SEO audit tool

The closest comparison is a conventional crawling and audit tool such as Screaming Frog or Lighthouse-style checks. The difference is not the checks themselves, many of which overlap with the seo.md checklist. The difference is that a crawler returns a report and stops, while this skill is written as an operating procedure that continues into implementation and a scheduled re-measurement.

That is also where the risk sits. A crawler cannot edit your robots.txt. This skill can, and the README's own example shows an undecided GPTBot and PerplexityBot policy as a finding, which means the agent may end up writing crawler directives. The trade-off is clear: you get execution instead of a PDF, and you accept an agent editing production files with whatever review process you have around Claude Code.

Against a human agency, the comparison is cost and consistency rather than capability. An agency brings judgment about positioning, content and outreach that a checklist does not contain. The README's argument is that a large fraction of what agencies bill for is the checklist part. Whether that fraction is large for your specific site is the question to answer before deciding.

Licence, maintenance and upgrade cost

The licence is MIT, so you can use, modify and redistribute the skill, including commercially, provided the licence and copyright notice are preserved. That is a permissive licence and it is compatible with the README's invitation to keep the retainer money. It says nothing about the content the agent generates for your site, and nothing about any third-party service the audit touches, such as Bing Webmaster Tools or Naver Search Advisor. Those are governed by their own terms, which the repository does not restate.

On maintenance: the repository is not archived, and the last push was on 2026-08-27, which is recent enough that the project is being changed. The v1.1.0 release on the same date added Bing support, AI crawler policy, E-E-A-T, English documentation and the /plugin install path. That release note is the best evidence of what the maintainer considers in scope, and it suggests the reference documents are the part most likely to change as crawler policies and answer-engine behaviour shift.

Upgrade cost is low if you installed as a plugin, since the README presents that route as easy updates. If you cloned into .claude/skills, you are pulling yourself. The subtler cost is that the canonical references are Korean and the English mirrors may lag, so an upgrade can change agent behaviour in ways an English reader will not see in the mirror.

Editorial conclusion

Adopt it if you run Claude Code, own a site whose technical base you can actually change, and want a repeatable audit with a scheduled re-measurement instead of a monthly retainer. Skip it if you need someone to execute content strategy or link acquisition, since the skill explicitly refuses those, and if you cannot act on the findings yourself. Before installing, verify the two things the README leaves open: that your Claude Code build supports the /plugin marketplace commands as written, and that your site's rendering lets a no-JavaScript crawler see the content the audit will score.

Frequently asked questions

Is SEO dead now with AI?

The README treats SEO as one of five lanes rather than something AI replaced. Its SEO lane still asks whether Google and Bing crawlers can read and index the content at all, and the audit reads the site without JavaScript for that purpose. The newer lanes cover answer boxes, generative engines, model knowledge and Naver.

Is SEO still worth it in 2026?

The repository's argument is that the work is worth doing but not worth a monthly retainer, since most of it is public standards and repeatable checklists. It reports that the same playbook produced 1.54M search impressions in 30 days for one Korean site, which is a single operator's result rather than a general benchmark.

What does "SEO agency" mean?

The README frames an SEO or AI-visibility agency as a service charging monthly retainers for AI-era search optimization, guaranteed ChatGPT citations and LLM visibility. It argues that what such services actually do is apply public standards and repeatable checklists, which is the work this skill moves into an agent.

Official sources

Community notes