SEO-Kami 神

The God of SEO skills — earned, not boasted.

SEO-Kami is a current (2026), evidence-tagged SEO/GEO/AEO Agent Skill. The name is earned, not boasted: it blends the working parts of the best open-source SEO skills, re-verifies every volatile fact against primary 2025–2026 sources, and passes the whole through three hardening rounds — so it ships fundamentals and frontier in one audit without confusing the two. Every line it returns carries an evidence tier — PROVEN, CORRELATED, CONSENSUS, or SPECULATIVE — so proven SEO never wears the same robes as AI-search hype. The godlike part isn't a ranking promise; it's the refusal to present a metric it didn't measure, or to sell a tactic the data can't back.

Adopt SEO-Kami (神)

Already have the skill, or can your agent see this repo? Drop it in, then paste any one-liner below. Your agent installs it, reads SKILL.md, routes to the right checks, and answers with every finding tagged by evidence tier — proven SEO separated from AI-search hype.

Install (one line):

cp -R SEO-Kami ~/.claude/skills/seo-kami

Then paste one of these:

• Adopt this. — installs SEO-Kami and runs a baseline audit on our project, no extra setup.
• Analyze our SEO state compared to SEO-Kami. — gap report measuring our site against the skill's evidence-tagged standards.
• Audit https://our-url.com with SEO-Kami and give me a prioritized fix plan. — full SEO/GEO/AEO crawl, scored, with the highest-leverage fixes first.
• Compare our SEO to SEO-Kami and flag anything that's hype, not proven. — same audit, but it calls out every SPECULATIVE tactic instead of selling it.

Every line that comes back is tiered — PROVEN, CORRELATED, CONSENSUS, or SPECULATIVE — so you always know what's earned and what's a bet. No paid API keys required; optional PAGESPEED_API_KEY and GITHUB_TOKEN just unlock field data and higher rate limits. MIT-licensed by MEGZO_tech.

Why it exists / the fusion

SEO-Kami is a synthesis, not a fork. It reuses ideas and structure from prior open-source skills, rewrites all content as original prose, and updates every volatile fact to 2026. Full attribution lives in NOTICE.md.

Source	License	What SEO-Kami took from it
claude-seo	MIT	The router/skill shape and the idea of script-backed checks invoked from a single SKILL.md entry point.
agentic-seo	MIT	The audit-as-orchestrated-pipeline pattern — discrete checks (technical, schema, links) feeding one report.
seo-audit-skill	MIT	Concrete technical-audit checks and report structure for a crawl-and-grade pass.
seo-geo-aeo	no license	Design reference only. Used to shape the GEO/AEO topic coverage. No code or text was copied, because it carries no license; treated as inspiration, not source.

The shared blind spot across all four: they present tactics without distinguishing Google-confirmed facts from practitioner folklore, and their GEO/AEO advice leans on backlinks and structured-data tricks instead of entity and earned-media signals. SEO-Kami's evidence tiers and entity/mention weighting are the fix.

What makes it different

• Evidence tiers on every recommendation. Nothing ships without a tier. CWV thresholds and schema deprecations are PROVEN; citation-correlation numbers are CORRELATED (correlation, not proven causation); llms.txt as a ranking/citation lever is SPECULATIVE and flagged as hype where claimed otherwise.
• Entity and earned-media weighting. GEO scoring follows the 2025 evidence: branded web mentions correlate ~0.664 with AI-Overview brand visibility vs ~0.218 for backlinks (~3:1); Muck Rack found ~82% of AI citations come from earned media. The skill scores brand-entity consistency and off-site presence, not just link counts.
• Zero-click measurement. It frames success as impressions / visibility / AI-citation share / brand-mention growth rather than sessions alone, because ~58-68% of Google searches end without a click and AI-Overview queries run ~83% zero-click.
• Built-in finding verification. finding_verifier.py re-checks claims before they reach the report, so findings are grounded in what was actually observed rather than asserted.
• Optional GitHub-SEO. github_seo_audit.py audits a repo's discoverability (README, topics, metadata) when you point it at a GitHub project.
• Zero mandatory paid APIs. Standard-library-first Python. Everything core runs with no keys. Optional keys (PAGESPEED_API_KEY, GITHUB_TOKEN) unlock field-data and higher-rate-limit paths but are never required.

Install

SEO-Kami is an Agent Skill (agentskills.io spec). Install it by copying the skill folder into your skills directory.

• Claude Code: copy this folder to ~/.claude/skills/seo-kami.
• Claude.ai / Cursor / Codex: copy it into that tool's skills directory.

cp -R SEO-Kami ~/.claude/skills/seo-kami

Then invoke it by asking your agent to audit a URL, for example: "Audit https://example.com with SEO-Kami." The agent reads SKILL.md, routes to the relevant references and scripts, and runs the checks.

Quickstart

One command does the whole audit — fetch, run every check, merge, score, and write the report:

python3 scripts/seo_kami.py https://example.com --report md --out report.md

Without --report it prints the JSON envelope; add --no-cwv/--no-links to skip the slower network checks. Any single check also runs standalone:

python3 scripts/seo_kami.py https://example.com            # JSON findings
python3 scripts/schema_check.py <url>                      # one dimension
python3 scripts/hreflang_check.py <url> --reciprocal       # i18n, with return-tag check

If a URL returns non-200 or non-HTML, the audit stops with one honest "not auditable" finding rather than inventing results from an error page.

Layout

SEO-Kami/
  SKILL.md                  # router / entry point the agent reads first
  README.md
  LICENSE                   # MIT
  NOTICE.md                 # attribution to the source skills
  references/               # 16 evidence-tagged docs (CWV, schema, GEO/AEO, E-E-A-T,
                            #   multilingual, content-strategy, content-brief, ...)
  scripts/
    seo_kami.py             # orchestrator
    technical_audit.py
    schema_check.py
    cwv_check.py
    geo_aeo_scan.py
    entity_check.py
    links_audit.py
    hreflang_check.py       # deterministic hreflang / i18n validator
    finding_verifier.py
    report_build.py
    github_seo_audit.py
    seo_common.py
    lib/
      safe_http.py          # SSRF-guarded fetch + per-redirect re-validation
      sanitize.py           # prompt-injection sanitiser for crawled content
  agents/
    seo-verifier.md         # adversarial subagent: verify findings before reporting
  assets/
    schema-templates.json   # JSON-LD templates for types that still earn rich results
  tests/
    test_smoke.py           # offline smoke tests for the scripts

Requirements

• Python 3.8+ (standard-library-first; no install step for core checks).
• requests is optional — scripts fall back to the standard library when it is absent.
• No API keys required. Optional environment variables:
  • PAGESPEED_API_KEY — enables PageSpeed Insights / CrUX field-data lookups for Core Web Vitals.
  • GITHUB_TOKEN — raises GitHub API rate limits for github_seo_audit.py.

License

MIT. See LICENSE and the attribution in NOTICE.md.