SEOmator SEO Audit Skill

What it is

SEOmator SEO Audit Skill is a command-line SEO auditor plus the skill that lets a coding agent run it. The tool, seomator, fetches a page or a crawl of your site and checks it against 373 rules in 20 categories. They span technical SEO, Core Web Vitals, links, structured data, security headers, accessibility and AI search readiness. You get a score out of 100, a grade and a fix line for each problem.

1. You install the seomator tool and the skill
2. You ask for an audit of a full web address
3. Your agent runs seomator audit in the llm format
4. The tool fetches the pages and checks 373 rules
5. Your agent lists the failures first, each with a fix

Who it suits

Site owners, developers and marketers who want a technical SEO check without leaving their coding agent. It also suits teams that want the same checks in CI, and agencies that audit a staging site before launch. It audits one page unless you ask for a crawl. It needs an address it can fetch, and the skill prefers a live site to a local dev server.

Where to use it

Any coding agent that can run a command and read its output. The agent guide names Claude Code and Cursor, and the README’s skill section is written for Claude Code.

• Claude Code: the skill goes into .claude/skills/ in the project, or into ~/.claude/skills/ with -g
• Cursor, Codex and other agents: the same skills command, with the agent’s name after -a
• Any terminal: the tool alone, with no agent
• Your CI: the README’s GitHub Actions and GitLab CI examples
• A desktop app: run from source, from the same repository

How to set it up

Get it saves a zip of Mindket’s setup notes in your downloads folder, with every route the README and its agent guide give.

1. Install the seomator tool. You run the npm line in a terminal. It adds the seomator command, which does the audits.
2. Add the skill to your agent. You run the npx line. The skills tool asks which agents get it. The skill is a text file that tells your agent when to run seomator.
3. Check the setup, then ask. Run seomator self doctor, start a new agent session and type the prompt from the Example.

npm install -g @seomator/seo-audit

npx skills add seo-skills/seo-audit-skill

seomator self doctor

To skip the questions and install the skill for Claude Code in every project, the agent guide gives this form.

npx skills add seo-skills/seo-audit-skill --skill seo-audit -g -a claude-code -y

How to use it

Ask in plain words, or run the tool yourself. Your agent picks the flags from your question, and the skill gives it this guide.

Example

Your new site is nearly ready, and you want to know what to fix first.

Audit https://example.com and tell me what to fix first

Your agent runs the tool, reads its XML report and lists the failed checks first, each with the tool’s fix line.

If the page is set to noindex, a score cap keeps the score at 20 or below, and the skill tells your agent to fix that before anything else.

For newcomers and experts

If you are new. A skill is a text file that tells your agent when to use a tool. Here the tool is seomator, which you install once. Start with one page, then add --crawl when you want the whole site. A check marked not measured did not run in that audit, and it does not change the score.

For experts. Exit codes let a script tell a failed run from a low score. 0 is success. 1 means the run worked and scored under 70. A 1 with empty stdout means the command line was rejected. 2 means the command failed. With --format json or llm, a failure still writes an error document with a code such as dns, tls or timeout.

seomator compare with --fail-on-regression exits 1 when the score dropped or a rule got worse, and it never fails for two audits measured differently. --staging waives the noindex and robots.txt caps for a site that is blocked on purpose, and lists what it waived.

Good to know

The tool audits and reports. The skill also lets your agent go on to fix your site, and it tells the agent to give you a summary of every change.

• Files: if the folder has no seomator.toml, the skill has your agent create one with seomator init -y first
• Pre-approval: in Claude Code, the skill’s allowed-tools line lets Claude run seomator commands without asking during that turn, as Claude Code’s skills page describes
• Stored audits: each audit is saved by default in ~/.seomator/audits.db on your computer, and --no-save skips it
• Requests: by default every request identifies itself as SEOmatorBot/3.0 with a link to the repository, unless crawler.user_agent in seomator.toml changes it
• Page text: the llm report wraps text quoted from the audited site as untrusted data, and the report’s own security notice tells your agent never to follow it as instructions
• INP: it is never scored, since it needs a real person’s interaction. The README’s --simulate-interaction flag gives only a synthetic reading that stays unscored
• Maker: SEOmator, MIT licensed. Mindket lists it and is not SEOmator