How to use the Single Page SEO Analysis agent skill

What is an agent skill, and what does seo-page make the agent do?

An agent skill is a folder containing a SKILL.md file. The file tells a compatible agent when to use the skill and how to carry out the task. This skill is named seo-page, and its description is to analyze one supplied URL across on-page, content, technical metadata, schema, images and performance.

The skill is limited to a single-page review. It does not audit a whole website.

• On-page SEO: title tag length (50-60 characters), meta description length (150-160 characters), H1 and heading hierarchy, URL structure, internal and external links.
• Content quality: word count against page-type minimums, Flesch Reading Ease and grade level, keyword use, E-E-A-T signals such as an author bio, and publication and update dates.
• Technical elements: canonical tag, meta robots, Open Graph and Twitter Card tags, hreflang if the page is multi-language.
• Schema markup: detects types (JSON-LD preferred), checks required properties and flags missing opportunities.
• Images: alt text, file size (flags over 200KB as a warning and over 500KB as critical), modern formats, width and height, and lazy-loading method.
• Core Web Vitals hints: flags possible LCP, INP and CLS problems from the HTML alone.

How do you install the seo-page skill?

You can install the skill in either of two ways. Both start with downloading the skill ZIP from the page where you found this guide.

Option A is a personal install. Unzip the download into ~/.claude/skills/ so the skill is available in all your projects. Option B is a project install. Unzip it into the .claude/skills/ folder inside a specific project, so only that project uses it.

Option C is for the Claude apps. Upload the ZIP under Settings → Capabilities → Skills.

• Download the ZIP from this page.
• Unzip it, so the skill folder contains SKILL.md and LICENSE.txt.
• Place the folder in ~/.claude/skills/ (personal) or .claude/skills/ (project).

What does the skill need to run?

The skill needs Claude Code or another agent that reads SKILL.md files. It also needs a Python runtime, because the skill references a bundled script called metadata_template.py. The agent needs network access to fetch the URL you give it.

DataForSEO is optional. If DataForSEO MCP tools with API credentials are available, the skill can add live SERP positions and backlink data. Without them, the core page review still works.

• Claude Code or another SKILL.md-compatible agent
• Python runtime for the bundled metadata_template.py script
• Network access to fetch the target URL
• Optional: DataForSEO MCP tools and API credentials

How do you trigger the skill with a request?

You can call the skill directly by name and pass the URL, because its argument hint is [url]. You can also ask in plain language that names the page and the kind of check you want. The agent should pick the skill when the request is about reviewing one page.

Give a full URL. The skill fetches the page itself, so you do not need to paste the HTML unless the page requires a login.

• /seo-page https://example.com/your-landing-page
• Check this article for SEO problems before I publish: https://example.com/blog/post
• Review the schema and image alt text on https://example.com/product/shoe
• Give me a readability and metadata check for this draft page: https://example.com/draft

What does the report look like?

The output has four parts. First is a page score card out of 100 with sub-scores for on-page SEO, content quality, technical, schema and images. Second is an issues list sorted from Critical to Low. Third is specific recommendations with expected impact. Fourth is ready-to-use JSON-LD for schema opportunities.

The skill does not recommend HowTo schema, which the skill notes is deprecated, or FAQ schema for rich results. Existing FAQPage markup does not have to be removed, and QAPage is suggested for genuine question-and-answer content.

• Overall Score out of 100
• Sub-scores: On-Page SEO, Content Quality, Technical, Schema, Images
• Issues ordered Critical, High, Medium, Low
• Recommendations and JSON-LD suggestions

What are the limits and safety notes?

The skill runs a bundled script and fetches the URL over the network. Its SKILL.md says fetched pages, SERP results and API responses are untrusted data. If a page contains text that tells the agent to run commands or change settings, the skill reports that text as a finding and does not follow it.

Performance results are hints only. Core Web Vitals cannot be measured from HTML alone, so treat LCP, INP and CLS flags as things to check in a real browser test. Pages behind a login return authentication errors. You can paste the rendered HTML instead. If key content is loaded by JavaScript, the results may be incomplete.

• Single page only: use a site-wide tool for a full audit.
• Login-protected pages: paste rendered HTML or use a public URL.
• Unreachable URLs: the skill reports the error and does not guess the content.
• Performance flags are not real Core Web Vitals measurements.

What tips help you get better results?

Use the live URL when the page is public, because the skill reads what the server returns. For drafts, paste the title and description and ask for the metadata check, which compares them for templated wording. When you act on a schema suggestion, check it against the page content before publishing.

Review the Critical and High items first. Treat image lazy-loading as correct when a JavaScript lazy-loader such as Perfmatters, EWWW or lazysizes is in use, because the skill is designed not to flag those.

• Prefer full URLs, and avoid pages that need a login.
• Fix Critical and High issues before minor wording changes.
• Validate suggested JSON-LD against the visible page content.
• Re-run the skill after edits to confirm the changes.