Review open PRs for domain identity (CCI + getDomainProof family)

[linear]

What

Hayk's domain-identity track has three PRs open across two repos. Sit with the diffs, validate the security model end-to-end, leave the actionable comments. This is a thread that touches identity binding, RPC surface, and the SDK builder all at once — needs one reviewer who can see all three at the same time so the contract between layers stays consistent.

PRs in scope

• node Verify website ownership for domain identities #897 — getDomainProof RPC + domain verification (CCI side) — Verify website ownership for domain identities #897
• sdks Decoding how crosschain transactions worked in the Legacy Frontend #91 — addDomainIdentity + domain identity types — Add domain identity (website ownership) support sdks#91
• wallet-extension Mapping out how crosschain transactions work post-demoswork #92 — provider method + approval popup + test page — https://github.com/kynesyslabs/wallet-extension/pull/92

What to focus on

Already-known high-severity points from the first review pass (need to be confirmed addressed or pushed back on):

• node Verify website ownership for domain identities #897 — the signed proof message has no hostname / sender / nonce binding. An attacker who fetches the victim's .well-known/demos-cci.txt once can re-host it on their own domain and the same message + type + signature verifies. Either bind hostname + sender + a server-issued nonce/timestamp into the signed payload, or this binding proves nothing about which domain you actually control.
• node Verify website ownership for domain identities #897 — the RPC is authRequired: false with no per-IP rate limit. Even with the SSRF guard, every node becomes an open HTTPS GET proxy. Either gate the RPC or rate-limit it.
• sdks Decoding how crosschain transactions worked in the Legacy Frontend #91 — the domain claim ships its own flat web2 sub-context and re-signs the legacy dw2p message rather than aligning with the new ClaimReference + dacs-*:v1: scheme shipped on feat/sr4-cci-channel-transcript. Aligning keeps us shipping ONE identity format across SR-4 and DACS-3; not aligning means we ship two.

Acceptance

• Each of the three PRs has either an approval or a request-changes comment with the concrete blockers
• Domain proof verifier binds hostname + sender + nonce into the signed payload
• getDomainProof RPC is gated or rate-limited
• sdks Decoding how crosschain transactions worked in the Legacy Frontend #91 aligns with the new ClaimReference / dacs-*:v1: scheme OR has a documented reason it cannot

Source

Earlier review summary on these PRs is in conversation context; concrete file:line citations there too.