feat(relay): rendezvous-relay service — 7 packages + plans (contracts/transport/agent/control-plane/e2e/auth/web)

Multi-tenant reverse-tunnel service ("ngrok for Claude Code" with E2E): a
host-agent dials OUT to an operator-run relay; external devices reach the host
THROUGH the relay, routed by per-tenant subdomain, forwarding ciphertext only
(the relay never sees plaintext). Lets a customer reach their own self-hosted
web-terminal from anywhere with zero networking setup.

Packages — all tsc-strict + vitest green (656 tests), cross-package integration verified:
- relay-contracts: frozen shared contracts (mux frame codec, data model,
  capability token, E2E envelope, pairing) — the src/types.ts analog
- term-relay:   native WS mux + stateless data plane (subdomain routing, ciphertext forward)
- agent:        host-agent (pairing, per-host Ed25519 + mTLS dial-out, forwards to 127.0.0.1:3000)
- control-plane: accounts/hosts registry, pairing-code flow, routing table, provisioning
- relay-e2e:    browser<->agent E2E (X25519 ECDH through relay, AEAD, anti-replay, recoverable replay key)
- relay-auth:   Passkey/WebAuthn, capability tokens, per-host certs, deny-by-default tenant isolation
- relay-web:    browser login + Web Crypto E2E + client-side preview rendering

Security invariants INV1-15 enforced; cross-tenant isolation CI tripwire live
(.github/workflows/relay-tripwire.yml). Design + implementation-level plans in
docs/PLAN_RELAY_*.md and docs/EXPLORE_RELAY_SERVICE.md.

NOTE: generated autonomously per the reviewed plans. The security-critical
packages (relay-e2e, relay-auth) REQUIRE expert security audit before any real
deployment — passing tests prove self-consistency, not resistance to attackers.
Base app (src/, public/) unchanged; concurrent desktop work left uncommitted.

control-plane/src/ca/csr.ts

@@ -0,0 +1,53 @@
+/**
+ * CSR proof-of-possession (INV14). Models a PKCS#10 CSR as `embeddedPub(32) || sig(64)` where
+ * `sig = Ed25519(privateKey, CSR_CHALLENGE || embeddedPub)`. Verifying the signature against the
+ * embedded pubkey proves the requester holds the matching private key (the exact property a real
+ * PKCS#10 self-signature provides) — using builtin crypto only.
+ *
+ * INTEGRATION SEAM: production parses a real PKCS#10 DER (`@peculiar/x509`) and checks the
+ * self-signature. Swapping it in does not change `signHostLeaf`'s check ORDER or reject path.
+ */
+import { ed25519Sign, ed25519Verify, type KeyObject } from '../util/crypto.js'
+
+const CSR_CHALLENGE = new TextEncoder().encode('relay-cp/csr-pop/v1|')
+
+/** Test/agent helper: build a CSR proving possession of `privateKey` for `embeddedPub`. */
+export function buildCsr(privateKey: KeyObject, embeddedPub: Uint8Array): Uint8Array {
+  const message = concat(CSR_CHALLENGE, embeddedPub)
+  const sig = ed25519Sign(privateKey, message)
+  return concat(embeddedPub, sig)
+}
+
+export interface CsrParts {
+  readonly embeddedPub: Uint8Array
+  readonly sig: Uint8Array
+}
+
+/** Parse the modelled CSR bytes. Throws on wrong length. */
+export function parseCsr(csr: Uint8Array): CsrParts {
+  if (csr.length !== 96) throw new Error('malformed csr')
+  return { embeddedPub: csr.subarray(0, 32), sig: csr.subarray(32, 96) }
+}
+
+/**
+ * Verify CSR proof-of-possession: the embedded pubkey's private key signed the challenge.
+ * Independent of any registry state (a forged signature is refused even for a registered key).
+ */
+export function verifyCsrPoP(csr: Uint8Array): { ok: boolean; embeddedPub: Uint8Array } {
+  let parts: CsrParts
+  try {
+    parts = parseCsr(csr)
+  } catch {
+    return { ok: false, embeddedPub: new Uint8Array(0) }
+  }
+  const message = concat(CSR_CHALLENGE, parts.embeddedPub)
+  const ok = ed25519Verify(parts.embeddedPub, message, parts.sig)
+  return { ok, embeddedPub: parts.embeddedPub }
+}
+
+function concat(a: Uint8Array, b: Uint8Array): Uint8Array {
+  const out = new Uint8Array(a.length + b.length)
+  out.set(a, 0)
+  out.set(b, a.length)
+  return out
+}