Crawls a site from its sitemap and audits every page for crawl, index, on-page, content, schema and mobile issues, scored out of 100 with ranked fixes.
SEO skills for Claude Code
Anthropic's terminal coding agent. Install any SEO skill into Claude Code with one command.
Install into Claude Code
$ npx skills add https://seoskills.sh -a claude-codeClaude Code runs in your terminal, which changes what an SEO skill can actually do. It is not limited to describing a fix: it reads the files in front of it, runs the bundled script, calls the API, and writes the change back. Point it at a repo and a technical audit works on your real templates, not a generic checklist. That tool access is why the script-heavy skills in this catalog, the ones that crawl, parse, and hit an API, belong here first.
Most skills install with the single command above. A few ship as a plugin instead, because they share helper scripts across a folder that a one-skill copy would miss, and each skill page flags when to take the author's plugin route.
Start with SEO Audit, the most-installed skill in the catalog, which diagnoses a site end to end and hands you a ranked list of what to fix. From there the work splits by symptom. For page experience, Core Web Vitals reads both field and lab data for LCP, INP, and CLS. For structured data, Schema adds and validates the markup that earns rich results. For answer engines, AI SEO works on getting your pages cited by ChatGPT, Perplexity, and AI Overviews. The first-party skills lean hardest on Claude Code's file access: the Programmatic Template Auditor samples template pages and clusters the near-duplicates before they index, the Striking Distance Opportunity Finder reads your Search Console export for keywords stuck in positions 11 to 20, and the Search Console Cannibalization Analyzer finds the queries where two of your own pages compete.
Run the broad audit first, let it tell you which layer is weakest, then open the focused skill for that layer. Because Claude Code can act on the result, you fix in the same session you diagnosed in.
Install in 4 steps
Pick skills from the catalog
Run this in your project. The CLI lists the skills our index serves, nearly the whole catalog, and asks which ones to install into Claude Code. Add --skill '*' to take all of them.
$ npx skills add https://seoskills.sh -a claude-codeOr install a single skill
Each skill page has its own command. Our skills install by slug from https://seoskills.sh, and skills by other authors install from their GitHub repository so the install counts toward them. Add -a claude-code to either.
$ npx skills add https://seoskills.sh --skill content-decay-predictor -a claude-codeDecide where it lives
By default a skill lands in .claude/skills/ inside the project, so it is committed with the repo and everyone on the team gets it. Add -g to put it in ~/.claude/skills/ instead, where every project can use it.
Ask for the job
Claude Code reads each skill's description and loads the skill when a request matches, so you can just say "audit this site's technical SEO". Claude Code notices new skills in a running session; if the skills folder did not exist when the session started, run /reload-skills or start a new session.
Start with these
Twelve to start with, all built by seoskills.sh. The first five cover the jobs most sites need, from a full-site audit to AI search; the rest run scripts against your Search Console data, crawls and templates, which is where a terminal agent earns its keep.
Diagnoses failing LCP, INP and CLS from field data and Lighthouse, names the elements, files and scripts at fault, and ranks the fixes by savings.
Generates complete JSON-LD for a page from what it shows, checks existing markup against Google's required properties, and flags what it could not find.
Scores a page on how easily ChatGPT, Perplexity, Claude and Google's AI Overviews can quote it, checks AI crawler access, and rewrites the weak passages.
Builds pages at scale from a spreadsheet and one template, with hubs, related links and a sitemap, and holds back thin, duplicate or boilerplate pages.
Samples pages from a programmatic template and uses SimHash to find near-duplicates, thin pages and boilerplate-heavy doorway pages before you scale.
Finds keywords sitting in Search Console positions 11 to 20 with real demand, ranks them by projected click gain, and maps each to its page.
Finds keyword cannibalization in Search Console data: queries where several of your URLs compete and keep swapping positions.
Fits a trend to each page's Search Console clicks and impressions to predict which pages will keep losing traffic, ranked by projected click loss.
Checks that every old URL 301-redirects once to a live equivalent after a migration, and that titles, canonicals, schema and indexability carried over.
Crawls a site for canonical, hreflang, redirect and robots signals that contradict each other and make Google ignore your canonical tags.
Builds your internal link graph, finds closely related pages that aren't linked yet, and suggests contextual links with anchor text.
Questions
- Do I need the skills CLI to use these in Claude Code?
- No. A skill is a folder with a SKILL.md file, so you can copy one into .claude/skills/<name>/ by hand. The CLI saves the copying, brings any scripts and reference files the skill ships with, and keeps the skill updatable.
- How do I update or remove a skill?
- npx skills update refreshes everything you installed, npx skills list shows what is there, and npx skills remove <name> takes a skill out. Add -a claude-code to remove it from Claude Code only.
- Can a skill run commands without asking me?
- Only if its SKILL.md pre-approves them. A few skills grant themselves shell access in their frontmatter, and their pages say "Runs shell commands" next to the security scan result, so read those before you install.
- Do the skills cost anything?
- No, every skill here is free to install. Some call paid APIs such as Ahrefs, Semrush or DataForSEO and need your own key; the skill's instructions say which.