Core features

Website health

The development build now observes several things that can make a website unhealthy even while its server still returns HTTP 200.

TLS

Web Fleet records certificate validity, hostname matching, issuer/subject data and expiry. TLS observations use the same guarded network path as monitoring and are refreshed separately from minute-level checks.

DNS

Web Fleet records A, AAAA and CNAME results, normalizes them, rejects private/reserved answers and distinguishes resolution errors from genuine configuration changes.

Pages and links

The crawler follows same-origin pages with hard page/depth/link caps, respects robots.txt rules, discovers sitemaps, persists the page/link graph and compares broken targets with the previous completed crawl. External verification is capped and deliberately more conservative than internal crawling.

site crawl
├── /                  200
├── /docs/             200
│   ├── /docs/install  200
│   └── /docs/api      404  ← new regression
└── /download          301 → /releases

Headers and redirects

Web Fleet records bounded redirect chains and selected HTTP/security headers independently from availability. Required-header expectations are configurable per site so an optional header is not treated as universally mandatory.

Performance

Server-side timings and transfer behaviour come first because they can be measured cheaply and consistently. Web Fleet should never label synthetic server timing as Core Web Vitals.

Audit

CP12 will add optional browser-rendered audits for representative pages. The site-detail UI will expose these under an Audit tab with the latest category scores and findings. History is opt-in; when enabled it also explains regressions across runs.

• Performance signals measured in a real browser context.
• Accessibility checks with concrete failing elements where practical.
• Best-practice and security-hygiene checks.
• Technical discoverability and SEO-style checks without pretending Web Fleet is a search-ranking oracle.
• Fleet-wide surfacing when an audit meaningfully regresses.

Audit is manually triggered rather than scheduled. Operators can run one site or deliberately batch a filtered set; Chromium remains optional for normal monitoring. Audit history is opt-in rather than accumulated by default.