How to use the Technical SEO Audit agent skill (seo-technical)

What an agent skill is and what this one does

An agent skill is a folder with a SKILL.md file that tells an AI agent how to handle a specific kind of task. Once installed, the agent reads the skill when a request matches its description and follows its checklist.

This skill covers technical SEO: whether search engines can crawl and index pages, how duplicate and thin content is handled, HTTPS and security headers, mobile use, Core Web Vitals, rendering and structured data. It can run a bundled sitemap discovery script against a URL and reports findings by category, including notes on how AI crawlers are managed in robots.txt. It does not cover content strategy, keyword research or backlinks.

• Crawlability: robots.txt, XML sitemap, noindex tags, crawl depth, JavaScript rendering, Googlebot fetch limits
• Indexability: canonical tags, duplicate content, thin content, pagination, hreflang, index bloat
• Security: HTTPS, mixed content, CSP, HSTS, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, back-button hijacking, server stack exposure
• Also covered: mobile, Core Web Vitals, rendering, structured data and IndexNow

Install the seo-technical skill

Download the ZIP for this skill from this page and unzip it. For a personal install, place the skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

You can also upload the skill in the Claude apps under Settings, then Capabilities, then Skills.

The skill relies on the claude-seo plugin root (CLAUDE_PLUGIN_ROOT), so install that plugin alongside this skill. Without it the sitemap script cannot be found.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside the project folder
• Apps: Settings, then Capabilities, then Skills, then upload

Check the requirements before you run an audit

The skill needs Python to run the bundled sitemap_discovery.py script through the claude-seo launcher. It also needs network access to the target website, because the audit fetches and inspects live URLs.

The skill is licensed under MIT, as stated in its LICENSE.txt file. The source does not list any pricing, so check your own costs for the agent and tools you use.

• Python installed
• claude-seo plugin root (CLAUDE_PLUGIN_ROOT) installed alongside this skill
• Network access to the target site
• Permission to audit the site: only audit sites you own or have permission to test

Trigger the skill with natural requests

You do not need a special command. Ask in plain language about a technical SEO audit, crawling, indexing, rendering or site speed. The skill also accepts an optional URL argument, shown as [url] in its settings.

Name the site and the problem you care about. A specific question gets a more focused report than a general request.

• Run a technical SEO audit on https://example.com and group the findings by category.
• Why are my pages not being indexed? Check crawlability and noindex tags on example.com.
• Is my site loading slowly? Look at Core Web Vitals and rendering for example.com.
• Should my robots.txt block GPTBot and Google-Extended? Explain the difference for example.com.

Run the sitemap check

For the XML sitemap check, the skill runs its bundled script through the claude-seo launcher. The command below is the one given in the skill source, with the target URL in place of the placeholder. The skill requires a valid entry in the found results, and reports stale or unsafe robots.txt sitemap declarations separately from working fallback locations.

Understand the AI crawler notes in robots.txt

The skill treats AI crawler management as a technical SEO issue. It lists known crawlers and what each token is used for, and it notes that blocking one crawler often does not affect another. For example, blocking GPTBot prevents OpenAI training but does not govern ChatGPT Search citability, which is controlled by OAI-SearchBot. Blocking Google-Extended does not affect Google Search indexing or ranking.

The skill also warns that you should weigh your AI visibility strategy before blocking any AI search crawler, because blocking it can remove your site from that engine's answers. Treat these notes as guidance, not legal advice, since vendor behaviour can change.

• Google-Extended: Gemini training and grounding, not Search ranking
• GPTBot: OpenAI training only
• OAI-SearchBot: ChatGPT Search citability
• ClaudeBot: Anthropic training only
• Claude-SearchBot: Claude search citability

Safety notes: scripts, network access and server checks

This skill runs scripts and uses the network. It fetches the pages you name, and it reads the normal HTTP response for the audited URL to spot server stack signals such as X-Powered-By headers or WordPress markers. The skill is designed to be read-only: it does not fuzz, brute-force or send exploit payloads.

If a site appears to run a stack with a known critical vulnerability, the skill asks whether you own the site before it requests any other path. It reports the exposure and the patch path, but fixing is the owner's or host's job.

• Only audit sites you own or are authorised to test
• Expect outbound requests from your machine to the target site
• Do not request extra probe paths on sites you do not own

Tips for a useful audit

Give the agent one site and one goal at a time. Canonicalization fixes can take a while to show in search, so the skill advises against judging a fix from an unchanged canonical right after the change. Use the findings as a checklist to prioritise, and verify important changes yourself before you deploy them.

• Ask for findings grouped by category to make triage easier
• Keep key content and structured data within the first 2MB of HTML, since Googlebot fetches only that much
• Remember that Google does not offer manual crawl-rate control; influence crawling through sitemaps, server responsiveness and robots controls
• Ask follow-up questions on any single category, such as Security or Indexability, for more detail