What it helps you do
Read-only public-site analysis, starting with one page or a bounded 10-page crawl. No login, credentials, paid provider, source edits or external submissions.
Expected output
- Bounded crawl or page evidence in JSON
- Observed response, metadata and indexability signals
- Three actionable findings plus unknowns
Environment and setup
Upstream documents: Codex, Claude Code, Cursor, Claude Desktop (MCP), Other local-stdio MCP clients.
Upstream documents skill/CLI and MCP integration. No agent install or runtime compatibility test has been performed.
- Node.js >=22.19.0
- npm package seo, reviewed source version 0.2.41
- Local shell and filesystem for CLI mode, or a configured local stdio MCP client
- Browser installation may be needed for rendering/Lighthouse paths; raw HTTP crawl does not establish rendered-page evidence
- Runtime packages include undici, robots-parser, linkedom, libsql, lighthouse and playwright-core; see pinned package.json
Read the pinned original Skill instructions ↗ and setup documentation ↗ before installing. No automatic installation is provided by this site. Review a fixed version and its dependencies before running setup.
Costs and permissions
Apache-2.0 package with no separate subscription for local technical reports. Existing agent/model charges still apply.
Commercial API requirement: No for local technical crawling, sitemap health and audit-page. Google sign-in and paid research providers are optional and excluded from the starter route.
Use a supplied public URL. Do not run seo start, connect Google, add providers, enable enrichment, submit IndexNow or buy anything in this exercise.
- Makes direct HTTP requests to target sites; browser rendering can load third-party page resources. Bound host, pages and depth.
- Writes local caches, settings and report files. Keep them out of the public teaching repository.
- Anonymous telemetry is enabled by default according to PRIVACY.md. Set DO_NOT_TRACK=1 before the first CLI run to opt out.
- The skill tells the agent to install an unpinned global npm package when missing. Do not follow that automatically: review and approve a pinned install first.
- The suite can connect accounts, store tokens, use paid providers, edit source and send IndexNow notifications. These are outside this starter scope.
- Host agents may send tool outputs to their model provider. Local execution does not guarantee offline/private model processing.
A bounded starter prompt
Use the installed local SEO skill to inspect <PUBLIC_URL>. Read-only, no paid APIs, Google sign-in, providers, editing, publishing, IndexNow or automatic installation. Set DO_NOT_TRACK=1. First verify the installed version. Discover and describe the appropriate report before constructing parameters. Bound a crawl to 10 pages, depth 1, the supplied host, no external-link checks, and disable sitemap expansion for that crawl. For single-page review use audit-page. Save complete JSON, then summarize three findings with evidence and limits. Stop on authentication, denial or unclear scope.
This prompt is an original teaching template. It has not been executed with the third-party Skill.
Documented CLI examples · Not run here
Inspect installed CLI only. documented/not-run.
DO_NOT_TRACK=1 seo --version
Discover and describe single-page report. documented/not-run.
DO_NOT_TRACK=1 seo reports describe audit-page --json
Bounded raw crawl after reviewing installed help. upstream-documented options; composed example/not-run.
DO_NOT_TRACK=1 seo crawl https://example.com --max-pages 10 --max-depth 1 --no-sitemap --no-external --json --output crawl.json
Check installed help and respect target access/rate limits. No claim of successful execution.
Limits of the result
- Guaranteed indexation, rankings or traffic
- A complete site audit from a small crawl
- Proving real-user Core Web Vitals from static HTML
- Automatic changes to robots, canonical or noindex without confirming intent
Source reviewed and Not run describe different evidence. No indexing, ranking, traffic, installation safety or blanket runtime compatibility is established here.
One exact CLI path was executed
We ran seo 0.2.41 audit-page on two original synthetic HTML documents over loopback HTTP. Read the real run, complete JSON and limits. This does not establish assistant integration, broad crawling or sitemap-report behavior.
Source and license
Maintainer: Ian Nuttall. Reviewed version: 0.2.41.
Commit: 5b794fe9c8addbc227c49262bb6ac79ad8429452
License: Apache-2.0. Pinned LICENSE and package.json reviewed. Project names and artwork have a separate trademark policy.
License source ↗ · Original repository ↗