AI NewsOpen sourceAnnouncement
fire-your-seo-agency runs five search optimization methods against a site
fire-your-seo-agency is a Claude Code skill that audits a site as crawlers without JavaScript see it, then applies fixes for classical search, answer engines, generative AI, model knowledge, and Naver. It was created on 26 August and had 380 stars and 98 forks six days later, MIT licensed.
Image: GitHub
Why it mattersThe skill treats the five audiences separately, each with its own acceptance test, which is the first thing a working SEO team has to decide and the thing most all-in-one plugins get wrong.
leopard627 published a Claude Code skill called fire-your-seo-agency on 26 August 2026. The repository had 380 stars and 98 forks by 1 September, under an MIT licence, which is real interest for a search-optimization tool aimed at a specialist audience.
The pitch in the readme is that a paid SEO agency can be replaced by a Claude Code agent, and the price it names is the roughly 500,000 to 3,500,000 Korean won a small Korean business would pay a monthly retainer for. That is between about 400 and 2,500 US dollars a month at current rates. Treat the claim with the caution you would give any tool that promises to replace an agency, since no independent measurement of the outcome is offered in the repository.
What it actually does
The skill defines five audiences, and treats each separately, with its own acceptance test. Classical SEO is the Google and Bing crawlers, and it asks whether crawlers can read and index the page at all. Answer Engine Optimization is the AI answer box above the results, and it asks whether that box cites the site. Generative Engine Optimization is ChatGPT, Perplexity and Claude browsing on behalf of a user, and it asks whether the site is the source that gets quoted. LLM Optimization is the model's own training data, and it asks whether the model knows the brand and knows it correctly. NEO is Naver, and it asks whether Naver's own AI Briefing and Search Advisor systems cite the site. The Naver part matters because Naver handles roughly half of Korean search traffic, and the standard English guides do not cover it.
The audit and fix loop
The audit runs against the site as a simple crawler sees it, without JavaScript, and scores each of the five audiences. The fixes are the standard ones, done by the skill rather than by hand: server-side rendering exposure, sitemaps, structured data, intent-matched pages, and machine-readable formats. The last stage is a measurement loop, in the sense that the skill schedules a re-evaluation and asks whether the fix changed the score.
The Naver part is what separates it from every other tool of this kind. If you serve the Korean market and half your users are on Naver, the tool that only covers Google is by definition covering half your search traffic.
Every serious search team ends up with five distinct methods, one per audience, because a fix for Google can hurt ChatGPT and a fix for ChatGPT can be invisible to Naver. The interesting design choice here is that the skill accepts five acceptance tests at once and refuses to score a page against one audience while ignoring the other four. Whether it works well is the question you would answer with your own site rather than by trusting the readme.
Source
- leopard627/fire-your-seo-agency on GitHub, MIT licence. Star and fork figures read from the GitHub API on 1 September 2026.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
