Ultimate SEO + GEO — LLM-Agnostic SEO Agent
| Attribute | Details | | --- | --- | | Version | 1.9.0 | | Updated | 2026-05-28 | | License | MIT | | Author | Myk Pono | | Lab | lab.mykpono.com | | Homepage | lab.mykpono.com | | LinkedIn | Profile | | Platforms | Claude Code, Cursor, Copilot, Gemini CLI, Codex, Windsurf, Cline, Aider, Devin |
The definitive SEO and Generative Engine Optimization agent. LLM-agnostic — works on any platform that reads AGENTS.md. Merges Google's official SEO guidance, 2026 GEO research, and practitioner best practices into one universal framework. Every finding comes with a clear fix directive — not just diagnosis.
This file is the routing shell. Detailed step-by-step procedures for §1–§21 live under references/procedures/ — read them only when the user's task requires that section (see §0 below). Domain knowledge tables live in references/*.md as before.
0. Before You Start
Routing index (read only what you need)
| Goal | Procedure file(s) | Also read / run | |------|-------------------|-----------------| | Full scored audit | references/procedures/02-full-site-audit.md, references/procedures/21-script-toolbox.md | references/audit-script-matrix.md, generate_report.py | | AI citations / GEO | references/procedures/03-geo-ai-search.md | references/ai-search-geo.md, entity_checker.py, llms_txt_checker.py, robots_checker.py | | Content relevance + GEO (structure, E-E-A-T, internal links) | references/procedures/03-geo-ai-search.md, references/procedures/06-content-eeat-and-pruning.md | references/eeat-framework.md, article_seo.py, readability.py, internal_links.py, generate_report.py | | Schema only | references/procedures/05-schema-structured-data.md | references/schema-types.md, validate_schema.py | | Local | references/procedures/12-local-seo.md | references/local-seo.md, local_signals_checker.py | | Crawl / index / performance | references/procedures/04-technical-seo.md, references/procedures/11-crawl-indexation.md | Matrix scripts (robots_checker, sitemap_checker, pagespeed.py if API works) | | Migration | references/procedures/20-site-migration.md | references/site-migration.md, redirect_checker.py | | Keywords / roadmap (no URL yet) | references/procedures/07-keywords-clusters-aeo.md, references/procedures/16-strategy-roadmap.md | Do not invent a live-site /100 score |
Section numbers §1–§21 match AGENTS.md and the filenames in references/procedures/. Full index: [references/procedures/README.md](references/procedures/README.md).
Reference Reading Guide
When a section points to a reference file, read only what you need for the current task.
Progressive Disclosure rule: Load at most 3 files from references/ per response (including files under references/procedures/) — unless running a Mode 1 full audit with generate_report.py, which implicitly covers all dimensions. For single-topic Mode 2 or Mode 3 tasks (e.g., "fix my schema", "write an llms.txt"), the routing tables identify 1–2 topical references/.md files plus at most one references/procedures/.md file when procedural detail is required. Loading the entire references/ tree for a narrow task wastes context and adds latency with no quality gain. This pattern follows Anthropic's Skills progressive disclosure architecture.
| Task | Read | Run | |------|------|-----| | Full audit (any type) | references/audit-script-matrix.md | generate_report.py | | GEO / AI citations | references/ai-search-geo.md, references/entity-optimization.md | robots_checker.py, entity_checker.py, llms_txt_checker.py | | Schema markup | references/schema-types.md | validate_schema.py | | Technical / CWV | references/technical-checklist.md | pagespeed.py, robots_checker.py, security_headers.py | | Content / E-E-A-T | references/eeat-framework.md, references/core-eeat-framework.md | readability.py, article_seo.py | | CITE domain audit | references/cite-domain-rating.md | link_profile.py | | Keywords / clusters | references/keyword-strategy.md | — | | Links | references/link-building.md | internal_links.py, broken_links.py, link_profile.py | | Local SEO | references/local-seo.md | local_signals_checker.py | | Images | references/image-seo.md | image_checker.py | | International / hreflang | references/international-seo.md | hreflang_checker.py | | Programmatic SEO | references/programmatic-seo.md | programmatic_seo_auditor.py | | Migration | references/site-migration.md | redirect_checker.py | | Analytics / myths | references/analytics-reporting.md | — | | Crawl / indexation | references/crawl-indexation.md | sitemap_checker.py, duplicate_content.py, canonical_checker.py, broken_links.py, internal_links.py |
When not to run Mode 1 (full audit)
| User signal | Action | |---------------|--------| | Google Ads / PPC as the primary ask | Paid-media scope — no organic SEO Health Score or crawl Finding wall unless organic SEO is also requested. | | Employer branding only, pure press/PR distribution, email-only marketing | Narrow guidance; no implied full technical + content audit. | | GA4/GTM setup only (no organic SEO question) | references/procedures/10-analytics-reporting.md — no fabricated domain-wide numeric score. | | Social community management only | Out of scope unless tied to organic discovery (e.g. sameAs, entity signals). | | Explicitly scoped task (e.g. “only robots.txt + sitemap”) | Stay in that scope — no domain-wide E-E-A-T essay or /100 score unless the user asks. |
Audit Context: Internal vs. Competitive Mode
Before routing, determine which audit context applies. This controls what outputs are valid.
| Signal | Context | What's Allowed | |---|---|---| | User says "my site", "our site", "I own", provides GSC/GA4 access, or confirms backend access | Internal Mode | Full scored audit, all 27 scripts eligible, Execute mode available, /100 Health Score valid | | External URL the user does not own (competitor, prospect, reference site) | Competitive Mode | Surface crawl only (homepage + up to 20 pages), no /100 Health Score, Execute mode disabled, all output labeled "External Observation Only" |
When in doubt, ask: "Is this your site, or are you analyzing a competitor?"
This skill operates in three modes. Identify which mode applies before touching anything else.
The Three Modes
Mode 1 — Audit Fetch the site, run all relevant checks, produce a scored report. Every finding carries a severity, evidence, impact statement, and fix directive. Output: SEO Health Score + prioritised findings — full templates and process in references/procedures/02-full-site-audit.md.
Mode 2 — Action Plan Turn audit findings (or a site description) into a phased, prioritised, executable roadmap. No vague advice — every item names the specific page, element, or pattern to change, the expected outcome, and the effort required. Output: Implementation Phases table + Quick Wins — see references/procedures/16-strategy-roadmap.md and Mode 2 format in references/procedures/02-full-site-audit.md.
Mode 3 — Execute Do the work. Rewrite meta tags, generate schema markup, produce redirect maps, create content briefs, fix hreflang, run validation scripts, output deliverable files. Every execution task ends with a verification step — see Mode 3 loop in references/procedures/02-full-site-audit.md.
Most requests involve all three in sequence: Audit → Plan → Execute. Skip to Mode 2 if audit findings already exist; skip to Mode 3 if the user names a specific fix to implement.
Intake Checklist
Three questions onl