Free open-source community tool

SEO + AI Growth System Skill

A portable community skill for Codex, ChatGPT, and Claude Code—built to collect what can be observed and label everything else honestly.

SEO operating system, not a magic score

Turn a public site into an evidence-led SEO + AI workspace.

The free package combines an Agent Skill with a Node.js CLI. It crawls a bounded public site, imports optional Search Console CSVs, creates technical and content planning files, and gives your AI agent a strict protocol for verified citation research.

Download v1.1.0 ZIP See installation

20generated workspace files

0required API keys

Node 20+only built-in runtime modules

MITopen community license

The download runs locally. Network requests go only to the public site and competitor homepages you explicitly name.

What it creates

One folder for the work SEO teams repeatedly rebuild

The baseline is useful without paid services. External measurements stay visibly incomplete until you import or verify them.

Site + brand baseline

Executive overview, observed page vocabulary, and a clearly labeled heuristic brand-voice snapshot.

GSC evidence slots

Query and page CSVs import when supplied. Without exports, both remain valid CSV templates marked “not collected.”

Technical audit

Static-HTML crawl findings, robots and sitemap evidence, request failures, canonicals, headings, metadata, images, and JSON-LD inventory.

AI visibility research

Buyer-prompt queue, citation evidence table, and perception protocol—without fabricating model answers or citations.

Competitor + content plan

Named homepage observations, honest share-of-voice placeholders, candidate gaps, comparison formats, and a validation-first brief template.

Reporting rhythm

A dashboard, weekly report, and safe Monday/Wednesday/Friday shell reminders that never spend credits silently.

How it works

Baseline first. Verified research second.

The skill keeps observed facts separate from imported performance data and live AI answer research.

Download

Extract one portable root folder. No npm install is needed.

Crawl

Run the bounded same-origin crawler with an explicit page cap and timeout.

Add evidence

Import GSC CSVs and let an authorized agent verify AI answers with source URLs and dates.

Operate

Validate issues, prioritize gaps, prepare briefs, and update the weekly evidence ledger.

Install + run

Portable across skill-enabled agent workflows

Keep the extracted folder intact. Copy it to the skills location used by your agent, then run the CLI from that folder.

Codex

Copy to your configured Codex skills directory, commonly $CODEX_HOME/skills/seo-ai-growth-system/, then reload the task.

Claude Code

Copy to .claude/skills/seo-ai-growth-system/ in your project, or the user-level Claude skills directory.

ChatGPT / Codex app

Upload the ZIP or add the extracted folder through the workspace Skills interface when that feature is available.

node bin/seo-ai-growth-system.mjs \
  --site https://example.com \
  --out ./example-seo-system \
  --gsc-dir ./gsc-export \
  --competitors https://competitor-a.com,https://competitor-b.com \
  --max-pages 50

Auditing a trusted local fixture? Add --allow-private explicitly. This weakens the private-network boundary and must never be used with untrusted URL input. URL credentials and cross-origin redirects remain blocked.

Privacy + evidence limits

Designed to say “not collected” when the proof is missing.

This is a safe baseline, not a replacement for Search Console, a rendered crawler, rank tracking, field performance data, or dated AI-answer testing.

Local-first
No telemetry, account connection, tracking pixel, or required cloud service.

No secrets
The package does not accept or store API keys, cookies, or login credentials.

Bounded crawling
Same-origin HTML only, path-level robots subset, DNS/private-network checks, manual redirects, 2 MiB response limit, timeouts, and page caps.

No fake metrics
Search volume, rankings, GSC performance, AI citations, and share of voice require named external evidence.

Local CLI versus cloud agents: the CLI processes files on your machine and makes bounded requests only to the public targets you name. If you upload the ZIP, generated reports, or GSC exports to a cloud-hosted agent, that provider receives those files under its own privacy and retention terms.

Verify the immutable v1.1.0 release

SHA-256: 7de9737e5ebdf206546bb163f2b4804d21d70cae5bb915e8d292d508e9db9a4a

Ed25519 public-key fingerprint: SHA256:581e567fb7e569b71f83700afda7ba7dec36521ec1957ed3d97cacdb00c5c453

Signature trust starts from the public key and fingerprint published here, not merely from downloading every verification file from the same server.

Immutable ZIP SHA-256 Signature Public key Verifier Manifest

Questions

Before you use the package

Does it automatically check ChatGPT or Google AI Overviews?

No. The baseline creates a research queue and evidence schema. A skill-enabled agent can perform live research with available tools only when authorized, recording exact prompts, platforms, dates, and source URLs.

Does “keywords.csv” contain rankings or search volume?

No. It contains observed on-page term frequency from fetched HTML. The file labels this directly so it cannot be mistaken for demand or rank data.

Can I share or modify it?

Yes. The source is included under the MIT License. Remove private GSC exports and generated client data before sharing your own output folder.

What does the crawler miss?

It does not render JavaScript, inspect logged-in pages, prove Google indexation, measure field Core Web Vitals, or replace a full enterprise crawler. Those limits are repeated inside the generated reports.

Build your evidence pack locally.

Download the complete signed v1.1.0 skill, CLI, tests, templates, and MIT license in one community-ready ZIP.

Download the free ZIP