Github: github.com/mykpono/ultimate-seo-geo
https://github.com/mykpono/ultimate-seo-geo
Is This the Best SEO + GEO Agent Skill for LLMs?
Ultimate SEO + GEO is an agent skill you can run, not a PDF checklist. It bundles procedures, 31 Python scripts, and regression tests so outputs stay structured when models drift. It is MIT licensed: github.com/mykpono/ultimate-seo-geo.
Search advice is cheap to generate. Search work that holds up in production is not. You need evidence, prioritization, and deliverables your team can ship. This skill is how I make that normal inside the AI tools developers already use.
Before I shipped it, I reviewed the strongest open agent SEO plugins and skills I could find, traced what each one did well, and noted where real audits still failed in practice: vague outputs, weak verification, GEO treated as an afterthought, scripts that did not line up with procedures.
Then I added my own SEO and GEO practice on top: how to score findings, when to narrow scope, what “done” looks like for schema, crawl, and AI citation work.
The result is deliberate synthesis, not a greenfield guess. Proven patterns from the ecosystem, plus tighter routing, broader module coverage, bundled tooling, HTML and Excel reporting, eval fixtures for regression, and GEO alongside classic SEO as a first class concern. The repo Credits section names the projects that informed architecture and research. Everything above that layer is what I integrated so the agent behaves like a practitioner, not a blog post.
Ultimate SEO + GEO follows the AGENTS.md pattern: one portable instruction set serious coding agents can load without locking you to a single vendor. It works with Claude Code (including the marketplace plugin), Cursor, GitHub Copilot, OpenAI Codex, Gemini CLI, Windsurf, Cline, Aider, and other AGENTS.md compatible tools. You are not buying a chat gimmick. You are installing a portable SEO and GEO operator next to your repo.
Everything routes through three modes. Most sessions use all three in order.
Audit. Fetch the site, run the checks that matter, return prioritized findings with severity, evidence, and a clear path to fix. No wall of generic tips.
Plan. Turn findings into a phased roadmap: what to do first, what it costs in effort, what it buys in impact, and who should own it.
Execute. Output real artifacts: JSON-LD, meta rewrites, redirect maps, robots guidance, validation runs. Then verify so you are not guessing.
If you already have an audit, start at planning. If you know the exact fix, jump to execution. The routing stays explicit so the agent does not expand scope when you only asked for schema or llms.txt.
SEO coverage spans 21 modules: technical health (including Core Web Vitals and crawl or index signals), on-page structure, E-E-A-T and content quality, Schema.org with deprecation awareness, JSON-LD generation and validation, keywords and topic clusters, internal and external links, local and international SEO, programmatic safeguards at scale, site migrations, and analytics oriented diagnostics when organic visibility is on the line.
GEO treats AI surfaces as first class: Google AI Overviews and AI Mode, ChatGPT style search, Perplexity, Bing Copilot. You get citability oriented structure, entity and brand signals, AI crawler rules in robots.txt, llms.txt, and current licensing patterns such as RSL 1.0. This is built for the world where ranking is not only ten blue links.