Infrastructure requirements
Environment capabilities and how unavailable fixtures are reported.
Some fixtures depend on infrastructure below the application: multiple origins, real TLS termination, broken certificates, HTTP/2 or HTTP/3, compression control, streaming services, or TLS fingerprint capture at the edge.
Each such fixture declares its needs in environmentCapabilities, and every
deployment publishes the capabilities it actually provides in the manifest
envelope. A test runner intersects the two up front: a fixture whose
requirements are not met in an environment is reported as unavailable — it
is never simulated, because a simulated transport failure would test nothing.
Supporting profile origins (robots variants, llms variants, host
canonicalisation, TLS certificates, protocol edge, fingerprinting, streaming)
are subdomains of the primary site. The manifest’s profiles array lists every
known origin with its base URL, risk level, and per-environment availability,
so crawlers discover profile hosts intentionally rather than by accident.
Locally, every production hostname maps to a .localhost equivalent behind a
reverse proxy, so origin-scoped fixtures are testable without production DNS.