Crawl Audit Fixtures

Title and H1 both say 'vs' / 'comparison', 300+ words, and no <table> or role=table anywhere.

Twelve <img> tags and one <video>, none declared in /sitemap.xml; plus a <figure> with no <figcaption>.

og:image declared with no width, height or alt; an article with a published time but no modified time, author or twitter:creator.

WebSite without url, Person without name, SiteLinksSearchBox without potentialAction. Absence is never a finding — each type must be declared and wrong.

One JSON-LD block typed ["WebApplication", "Product"] with an Offer and a brand, and no shippingDetails, hasMerchantReturnPolicy, review or aggregateRating.

The same block typed WebApplication only, with no brand. Nothing about shipping, returns, reviews or ratings may be reported.

Hub for the #154–#205 batch: one raw HTML fixture per issue code, plus the canonical targets and redirect hops they need.

Four outbound destinations on four independent hosts, each verified to ANSWER 404. An unreachable host is recorded as unchecked, not broken.

HTTPS page loading a script, a stylesheet and an iframe over http://. Active subresources only — an http:// image would raise #65 instead.

rel=manifest declared with no href. A page with NO manifest link is silent — the check grades configuration, not adoption.

The first rel=canonical on the page is an absolute http:// URL.

No <h1> anywhere in the document — PageHeader, the only component that renders one, is deliberately not used.

URL path carries uppercase letters and underscores. The URL must be crawled, not merely linked — the audit skips URLs it cannot resolve to a urlId.

Links the same-domain http:// form of /http-scheme-page. pageUrl is request.url, the URL as queued, so the finding survives an edge redirect to https.

A single 303 hop. Valid codes are 301/302/307/308; 303 is deliberately excluded. One hop, so it does not also raise #34.

/chain-hop-1 -> /chain-hop-2 -> /status-200, chainLength 2. Both hops are 301, so it does not also raise #28.

Each level links only the next. Five levels so the chain crosses maxDepth 3 whether the scan is seeded here or at the site root.

#38 is live on this host — /robots.txt declares no Sitemap: line, and /sitemap.xml is still found by the conventional-path probe. #26 and #314 contradict #37/#41 and the robots.txt fixtures respectively, so they are raised by the second deployment of this same source (wrangler deploy --env no-crawl-config), where both files answer 404.

Static file: no <meta charset> and no http-equiv content-type. Next.js injects a charset into every rendered page, so this cannot be a route.

Static file declaring windows-1252. The other branch of the same if/else as #312, so it needs its own document.

Static file with no viewport tag. The guard tests the CONTENT, so an empty content attribute would land here too.

Static file declaring width=1024, initial-scale=0.8 — short both required directives.

Static file with no icon, shortcut icon or apple-touch-icon. The root layout gives every rendered page a favicon link.

Static file whose content sits in a div with role=main. The selector matches tag names, so the role does not count. The root layout wraps every page in <main>.

Opens with a > summary and no # title (malformed), lists two URLs that 404 (broken links), and lists four of ~50 crawled pages (coverage). Deliberately NOT linked as an anchor — the crawler fetches it at the origin root.

Two unusable lines inside the User-agent: * group. parseRobotsTxt skips both without ending the group, so every existing directive still applies.