site-architecture
GitHub用于规划网站页面层级、导航结构、URL 模式及内部链接策略，确保用户与爬虫路径一致，避免深度过深或 URL 设计错误。
Trigger Scenarios
Install
npx skills add fcakyon/claude-codex-settings --skill site-architecture -g -y
SKILL.md
Frontmatter
{
"name": "site-architecture",
"license": "MIT",
"description": "Use when planning or restructuring what pages a site has and how they connect: hierarchy, navigation, URL patterns, breadcrumbs, and internal linking. Not for XML sitemaps, which are in seo-audit."
}
Site architecture
Architecture is one decision repeated: how does someone get from the homepage to the page that answers their question, and how does a crawler follow the same path. Depth, navigation, URLs, and internal links are four views of that one structure, so change them together or they drift.
Depth
Aim to put any page that matters within three clicks of the homepage. It isn't a law, but a critical page four or more levels down is a symptom worth chasing.
| Shape | Fits | Costs |
|---|---|---|
| Flat, 2 levels | Small sites, portfolios | Stops scaling once a nav item has 20 children |
| Moderate, 3 levels | Most SaaS and content sites | Usually the right answer |
| Deep, 4 or more | Large catalogs, big docs | Scales, but buries things without strong linking |
Go as flat as the navigation tolerates. When a dropdown passes roughly 20 items, that's the signal to add a level rather than keep the list flat.
Levels: L0 is the homepage, L1 is a primary section (/features, /blog), L2 is a page within it (/features/analytics), L3 and beyond are detail pages (/docs/api/authentication).
Site types as starting points
| Type | Depth | Sections | URL shape |
|---|---|---|---|
| SaaS marketing | 2 to 3 | Home, Features, Pricing, Blog, Docs | /features/{name}, /blog/{slug} |
| Content site | 2 to 3 | Home, Blog, Categories, About | /blog/{slug}, /blog/category/{slug} |
| Ecommerce | 3 to 4 | Home, Categories, Products | /{category}/{subcategory}/{product} |
| Documentation | 3 to 4 | Home, Guides, Reference | /docs/{section}/{page} |
| SaaS plus content | 3 to 4 | Home, Product, Blog, Resources, Docs | /product/{feature}, /blog/{slug} |
| Small business | 1 to 2 | Home, Services, About, Contact | /services/{name} |
URLs
Readable, lowercase, hyphenated, mirroring the hierarchy, with one trailing-slash policy enforced everywhere. Short but still descriptive: /blog/landing-page-conversions beats /blog/how-to-improve-your-landing-page-conversion-rates.
The mistakes that cost the most:
- Dates in blog URLs.
/blog/2026/07/25/titleadds nothing and ages the post visibly. - IDs or query strings carrying content.
/product/12345and/blog?id=123should be slugs. - Over-nesting past what the hierarchy needs.
- Mixing parents for the same kind of page, like
/features/analyticsalongside/product/automation. - Changing a URL without a 301. Every old URL needs one, or the links pointing at it stop counting and anyone who bookmarked it gets a 404. This is the single most common cause of traffic loss after a redesign.
references/patterns.md has the URL pattern per page type, navigation layouts, and the diagram formats to hand back.
Navigation
Primary navigation holds 4 to 7 items, ordered by importance, with the logo linking home and the call to action rightmost. Past 7, people stop reading the list and start hunting.
Footers group into columns: product, resources, company, legal. Sidebars carry within-section navigation for docs and long content. Breadcrumbs mirror the URL path exactly, with every segment linked except the current page, and they pair with BreadcrumbList schema.
Breadcrumbs are the cheapest structural win available: they add internal links on every page, they make hierarchy legible to a crawler, and they can earn a richer result.
Internal linking
- No orphans. Every page needs at least one internal link pointing at it, and the sitemap is not a link.
- Anchor text describes the destination. Never "click here" or "read more".
- How many contextual links a single page carries is a decision for whoever owns that page's format, so recommend the connections worth making rather than a density target.
- Link the pages that matter more often. Inbound internal links are how you tell a crawler what's important.
- Hub and spoke for content clusters: one comprehensive hub, spokes covering sub-topics, each spoke linking back to the hub, the hub linking to all spokes, and spokes cross-linking where a reader would actually want it.
Hub and spoke is what makes a set of posts add up to more than its pages, because it concentrates the signal on the hub rather than spreading it across a dozen equal posts competing with each other.
What to hand back
An ASCII tree of the hierarchy with the URL at each node, a URL map table (page, URL, parent, where it appears in navigation, priority), the redirect list when anything moves, and a Mermaid diagram when the structure is worth seeing rather than reading. references/patterns.md has the formats.
Sources
- Google Search Central, URL structure best practices: https://developers.google.com/search/docs/crawling-indexing/url-structure
- Google Search Central, Redirects and Google Search: https://developers.google.com/search/docs/crawling-indexing/301-redirects
- Google Search Central, Breadcrumb structured data: https://developers.google.com/search/docs/appearance/structured-data/breadcrumb
- Nielsen Norman Group, Flat vs deep website hierarchies: https://www.nngroup.com/articles/flat-vs-deep-hierarchy/
Version History
- 57286f4 Current 2026-08-28 15:21
