seo-site-audit

SEO site audit

Start with the main report. It joins the available first-party evidence and returns a small set of follow-up commands.

Through MCP, call seo_run_report with report id workflow-diagnose-property. Call seo_describe_report first when its parameters are not already known. Use the CLI when MCP is not available:

seo report --project <project> --json
seo report --site sc-domain:example.com --json

Read the main report

1. Check dataStatus, source completeness, skipped sections, warnings, and caveats before interpreting findings.
2. Read the narrative and priority queue against their evidence references.
3. Use nextCommands to choose a small number of follow-ups. Do not run every report by default.
4. Keep missing, unavailable, filtered, partial, capped, and complete evidence separate. A partial source cannot support an all-clear.

Add technical crawl evidence

Run a crawl when the user asks for technical coverage or the main report identifies a technical evidence gap:

seo crawl --project <project> --max-pages 500 --save --json
seo crawl https://example.com --max-pages 500 --save --json

Read summary, topFixes, warnings, and caveats first. Check summary.pageLimitReached before making sitewide claims. Reuse the saved report for detail instead of crawling again:

• Report id top-fixes returns a bounded queue.
• Report id affected-urls returns URLs for one rule.
• Report id explain-issue explains an unfamiliar rule.
• Report id get-crawl-report returns requested report detail.

Use report id performance-audit for a selected URL when performance evidence is needed. A crawl does not replace Lighthouse lab data or CrUX field data.

Present the result

Give the user the top three to five evidence-backed actions, affected areas, and a verification command for each action. Name every material gap or cap.

Never invent clicks, sessions, rankings, issue counts, indexing state, or sitewide health. Keep advice structural and diagnostic unless the user asks for copy.