What is the web SEO skill?
The web SEO skill is a SKILL.md in web-quality-skills. It guides an AI coding agent through technical SEO such as robots.txt, sitemaps, canonicals and HTTPS, on-page titles, descriptions and headings, JSON-LD, mobile and hreflang. It separates crawl findings from content quality and tells the agent never to invent ranking weights.
This is the seo folder of addyosmani/web-quality-skills. Its workflow runs Lighthouse SEO and Agentic Browsing checks when available, then inspects what Lighthouse cannot establish: response headers, redirects, robots.txt, sitemap coverage, canonicals and structured-data eligibility.
It keeps technical crawl and index findings apart from content quality and authority. It tells the agent not to invent ranking-factor weights or promise ranking changes, and to report that search engine validation is still pending.
What changes in your agent's output
- Sitemaps follow the stated rules: at most 50,000 URLs or 50 MB each, canonical indexable URLs only, a sitemap index for larger sites.
- Titles and descriptions use 50 to 60 and 150 to 160 characters only as a rough linting proxy, since snippets vary by query and device.
- Pages are not failed solely for having more than one h1 when the HTML is valid.
- Structured data describes visible, accurate content and is validated on the rendered output, with no type added only to chase a rich result.
- AI crawler controls are treated as product-specific, and llms.txt is called optional.
What the skill covers
Sections span crawlability, XML sitemaps, URL structure, HTTPS, title tags, meta descriptions, headings, image SEO, internal linking, JSON-LD, mobile SEO and hreflang. A table lists crawler tokens such as OAI-SearchBot, PerplexityBot and Google-Extended with their documented purpose and the effect of blocking.
Who the web SEO skill is for
Best for
- Pre-launch technical SEO checks on pages an agent builds.
- Teams that want honest limits: the skill lists what it can verify and what it cannot.
- Pairing with fixing-metadata for title, canonical and social tag fixes in code.
Not for
- Keyword research or content strategy. It audits technical and on-page signals.
- Guaranteeing rankings or AI citations. It says a valid llms.txt does not prove any AI product will cite a page.
- Performance work. Link to core-web-vitals for measured speed evidence.
Use it with UiChemy
Pages that UiChemy places into WordPress still need titles, canonicals and structured data from your SEO setup. Run this skill on the published page to see what is missing at the rendered level. Click Get Started on uichemy.com.
How to install Web SEO Skill: Technical and On-Page SEO for AI Agents
git clone https://github.com/addyosmani/web-quality-skills.gitClone the repository
Run
git clone https://github.com/addyosmani/web-quality-skills.git. Each skill is a folder inside the skills directory, and seo is skills/seo.Copy the skill folder
Copy skills/seo into the folder where your agent loads skills, for one project or globally. The README's manual route copies the whole skills directory, which keeps the cross-links between the sibling skills working.
Add live browser tools if you have them
The README says the skills prefer Chrome DevTools MCP for traces and Lighthouse runs when your agent exposes it. Without it they fall back to the Lighthouse CLI, PageSpeed Insights and manual checks.
Ask in plain words
Skills activate when your request matches their description, for example by asking for a quality review of a page.
Check that it loaded
Ask for an SEO review of one page. The agent should run Lighthouse SEO checks or inspect the rendered page, check robots.txt, sitemap and canonicals, and say that search engine validation remains pending.
Key features
- Evidence-led workflowLighthouse SEO and Agentic Browsing results first, then headers, redirects, robots.txt, sitemap and canonical checks.
- Verifiable scope tableLists what the skill can verify, such as crawl controls and rendered metadata, and what it cannot, such as ranking.
- Sitemap rulesMaximum 50,000 URLs or 50 MB, sitemap index, canonical URLs only, accurate lastmod and Search Console submission.
- URL structureHyphens, lowercase, short, HTTPS, and few parameters.
- On-page guidanceTitles, meta descriptions, heading hierarchy, image SEO and internal linking with descriptive anchors and breadcrumbs.
- Structured data disciplineJSON-LD for visible content only, the most specific type, stable identifiers and validation of rendered output.
- Agentic browsing signalsCovers the agent-facing accessibility tree, optional llms.txt and WebMCP registrations, kept separate from ranking claims.
- Crawler control tableDocuments OAI-SearchBot, PerplexityBot, Google-Extended and other crawler tokens with the effect of blocking each.
- Mobile and internationalResponsive design, tap targets, font sizes, hreflang and language declaration.
- Prioritized checklistCritical items such as HTTPS, robots.txt, noindex and unique titles come first.
Frequently asked questions
What is the web SEO skill?
It is the seo folder of addyosmani/web-quality-skills. It guides an AI coding agent through technical and on-page SEO checks using Lighthouse evidence where possible, and it separates crawl findings from content quality and authority.
How do I install it without npx?
Clone the repository with git and copy skills/seo into your agent's skills folder, or copy the whole skills directory. The skill activates when a request matches its description, such as fix meta tags or add structured data.
Will it improve my rankings?
It does not promise that. The skill tells the agent not to invent ranking-factor weights or promise ranking changes, and to say search engine validation is pending after fixes. It improves technical correctness and consistency.
Does it recommend adding llms.txt?
It calls llms.txt optional. A valid file may help compatible tools find curated content, but the skill says a pass does not prove any search or AI product will ingest, rank or cite the page.
Can it block AI crawlers for me?
It documents crawler controls in a table, such as OAI-SearchBot, PerplexityBot and Google-Extended, with the effect of blocking each. It treats the controls as product-specific, so review each one before changing robots.txt.
How is it different from fixing-metadata?
This skill audits the full technical and on-page picture. fixing-metadata is a narrower rules skill for metadata fixes in code. The web quality audit skill runs SEO alongside performance and accessibility.
What license does it use?
The repository is MIT licensed and the SKILL.md frontmatter says MIT. The 48.3K install count on this page comes from a third-party skills listing, not from the repository. The repository README and SKILL.md linked in the sources section are the authority for details.