Found, or found but not chosen?
The skill asks this before it checks anything. A site that search engines cannot crawl or index has a technical problem. A site that appears but loses the click has a content and relevance problem. The answer decides where the audit spends its time.
It still covers all four areas, because one cannot make up for another. Strong articles on a site with crawl errors underperform, and so does a spotless site with thin pages and no links.
Crawling and indexing come first.
If a page cannot be crawled, it cannot be indexed, and it cannot rank. So the audit opens with robots.txt, looking for a stray line that blocks whole sections, often left over from a staging site. It compares the XML sitemap with what Search Console reports as indexed, hunts for crawl traps and orphaned pages, and flags anything more than three clicks from the homepage.
Next come canonical tags, unintended noindex tags, HTTPS and mixed content, and redirect chains longer than two steps. Core Web Vitals are measured against the skill’s targets: Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, and layout shift under 0.1.
Titles, headings, and the words on the page
On each important page, the skill checks that the title is unique and fits about 50 to 60 characters, with the keyword early. It expects exactly one H1, clean URLs with hyphens, and alt text that describes the image rather than stuffing keywords.
The content review looks for two pages chasing the same keyword and for pages thinner than the top five results for their query. It also catches near duplicates, such as a location template repeated with only the city name swapped. It also checks for bylines, author pages, cited sources, and up to date facts.
Links from other sites
The off-page review counts referring domains rather than raw backlinks, reads the anchor text for signs of manipulation, and compares your links with three to five competitors. Domains that link to several competitors but not to you become the outreach list.
The action list at the end
Findings land in four tiers. Priority one is for problems to fix immediately: blocked crawlers, key pages noindexed by mistake, broken critical pages, and redirect loops. Priority two belongs in the next sprint, priority three within 30 to 60 days, and priority four is ongoing work such as content expansion and link building.
The full audit method on GitHub
Every check and tier described here comes from the seo-audit skill file in the Marketing Skills OS repository. It is released under the MIT License, “Copyright (c) 2026 ScayverGraphix.” The skills follow the Agent Skills spec published by vercel-labs/skills.
Run it against real data.
The skill expects exports from Search Console, PageSpeed Insights, a crawler such as Screaming Frog, and Ahrefs or Semrush. It treats an audit without data as an opinion. Mention an SEO audit or a prioritized SEO action plan, and the agent picks the skill up without being named.
The project readme links this skill with schema and ai-seo. Add all three with npx skills add scayver/marketing-skills --skill seo-audit schema ai-seo.
Priorities with a shared direction
An audit can produce a long list with no shared direction. Our Search Engine Optimization service connects technical health, page content, site structure, and measurement in one search plan, starting with the work that deserves attention first.