← ClaudeAtlas

ref-sp-web-seolisted

Audit a website's SEO from what is actually observable — served and rendered HTML, robots.txt, sitemaps, canonicals, titles, structured data, Core Web Vitals — separate that from the performance data only Search Console can answer, and design before/after tests that mean something. Use when: asked to check, audit, or improve a site's SEO; diagnosing why pages are not indexed, not ranking, or why organic traffic dropped; reviewing robots.txt, sitemaps, canonicals, titles, or structured data; debugging why a crawler cannot see JavaScript-rendered content; SEO on Next.js, WordPress, or headless WordPress + Next.js (metadata, sitemaps, canonicals, static export); interpreting a Lighthouse or Core Web Vitals report; judging whether an SEO change actually worked; weighing an SEO claim from a leak, a vendor, or a blog; or assessing content quality against Google's rater framework.
swiftpostlabs/agentic-tools · ★ 0 · Web & Frontend · score 73
Install: claude install-skill swiftpostlabs/agentic-tools
# Web SEO ## Purpose Audit a site's SEO using evidence you can actually obtain, state plainly which questions cannot be answered without the owner's Search Console data, and design SEO tests whose results mean something. ## This file goes stale in months, not years Search changes faster than almost anything else this repo documents, and the AI surfaces are changing faster still. **Treat every vendor-specific claim below as dated, not permanent.** Content here was verified **2026-07-12**; a specific that is more than a few months old has a real chance of being wrong, and stating it confidently is worse than looking it up. What is durable: the crawl → render → index → rank → measure pipeline, the observable/not-observable split, the claim-tiering rule, and the test-design discipline. Those are method, and method holds. What rots: thresholds, report names, which directives Google honours, what Search Console exposes, and anything about AI answers. Before asserting one of those, check `./references/sources.md`. This is not boilerplate caution. A freshness pass on this skill's own first draft found two claims that had already been superseded within weeks of being written. **Every dated claim here has a source, and re-verifying is one command.** Read `./references/sources.md` before asserting a specific, and whenever the skill may be stale. It maps each claim to the document it came from, the date it was checked, and what would invalidate it — and it opens with the cheap pa