Web metadata

Skill Neznakometz/StackForge/ui/skills/web-metadata

Page metadata for web — title/description, canonical, Open Graph, Twitter cards, JSON-LD structured data, robots/sitemap. Apply when building or auditing pages, landing pages, and shareable links (SEO and social previews).From its SKILL.md

Install

npx -y skills add Neznakometz/StackForge --skill web-metadata

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.2 KB, 503 tokens by cl100k_base, as published. Nobody here has run it

Web metadata (SEO & social)

Sources: HTML living standard, Open Graph protocol (ogp.me), schema.org, Google Search Central.

Core (every page)

Unique <title> (~50–60 chars) and <meta name="description"> (~150–160 chars) per page — not a single global default.
<link rel="canonical" href="…"> — absolute URL; prevents duplicate-content splitting. One canonical per page.
<meta name="viewport" content="width=device-width, initial-scale=1">, correct <html lang>, charset utf-8.
<meta name="robots"> — index,follow for public pages; noindex for private/staging/duplicate. Ship a robots.txt + sitemap.xml.

Social previews

Open Graph (Facebook, LinkedIn, Slack, etc.): og:title, og:description, og:type, og:url (canonical), og:image (absolute URL, ~1200×630, <5MB), og:site_name, og:locale.
Twitter/X cards: twitter:card (summary_large_image), twitter:title, twitter:description, twitter:image. Falls back to OG if omitted, but set twitter:card explicitly.
Test previews before shipping (e.g. social debuggers); a broken/missing og:image makes shared links look unfinished.

Structured data

JSON-LD (<script type="application/ld+json">) with schema.org types relevant to the page: Organization/WebSite (home), Article/BlogPosting, Product, BreadcrumbList, FAQPage. Match the visible content (no markup for things not on the page).
Validate against schema.org / rich-results expectations.

Framework notes

Next.js: the Metadata API (export const metadata / generateMetadata) — per-route, not hardcoded in one layout.
Nuxt: useHead / useSeoMeta per page.
Don't hardcode strings — use i18n + data; dynamic pages generate metadata from their data.

Anti-patterns

Same title/description on every page; missing canonical; relative og:image; JSON-LD that doesn't match the page; noindex accidentally left on production.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.