OptionalallowExplicit escape hatch to accept an RPC URL whose origin differs from the location the card was fetched from.
This is intentionally permissive: it is what lets a compromised or
misconfigured card-hosting endpoint redirect all subsequent A2A
traffic for this agent, including whatever credential material the
request-forwarding path carries, to a different origin than the one
that was actually configured and fetched. The A2A spec allows a card
to point RPC at a different origin than where the card itself is
served from (e.g. static card hosting separate from the API
backend), which is the legitimate case this exists for -- only set
this when that's a shape you actually expect. When set to true, a
loud warning is logged for each RPC URL this affects.
Defaults to false: every RPC URL must share the origin the card
was fetched from.
OptionalallowExplicit escape hatch to accept a plaintext http:// RPC URL on a non-loopback host.
This is intentionally insecure: an RPC URL that isn't https and isn't
on this machine can be intercepted or altered by anything on the
network path, even when its origin matches the card's own. Only set
this for a card source you know serves plain http on a host you
trust (e.g. a private, internal service reached over a channel
that's secured some other way). When set to true, a loud warning is
logged for each RPC URL this affects.
Defaults to false: RPC URLs must be https, or http on a loopback
host (localhost, 127.0.0.0/8, ::1, 0.0.0.0, or ::).
Options controlling how a fetched agent card's RPC URL(s) are validated against the location the card was fetched from. See
validateCardRpcTargetsfor what each check defends against and why both default to failing closed.