parlor → website screen/route audit (D9 reconciliation)

parlor → website screen/route audit (D9 reconciliation)

Written 2026-10-09 on branch agents-first/recon. This is the audit half of D9: website is the one customer-facing frontend (D1), and anything in parlor still worth having must be moved to website deliberately, before new screens land in the wrong tree.

The surface is screens and routes, and nothing else. Not the fleet, not the libs, not the deploy substrate, not a modernization sweep. The verdict column below is the work; the table is its shape.

Method

Both trees were compared as route families and screens, not file counts:

• src/routes/ — every route file in both trees, paired by path.
• src/app/ — every screen and form module, paired by path, then diffed. Byte-identical files are stated as such; every differing file was read to determine the direction of the difference (which tree is ahead).
• src/lib/, src/components/, src/styles/ — compared for parity only, to make the reasons accurate. They are not the surface and nothing below moves them.

The audit

#	parlor family	website counterpart	Verdict	Reason (one line)
1	/ landing (src/app/page.tsx)	exists	already-in-website	website’s is the marketing home and is ahead: Cafaye brand, W0-P1 pricing, meta description, brand-titles test.
2	/register	exists	already-in-website	register-form.tsx is byte-identical; page wrapper identical.
3	/login	exists	already-in-website	login-form.tsx is byte-identical; page wrapper identical.
4	/forgot-password	exists	already-in-website	forgot-password-form.tsx is byte-identical.
5	/reset-password	exists	already-in-website	reset-password-form.tsx is byte-identical.
6	/verify-email, /verify-email/confirm	exist	already-in-website	both forms byte-identical; page wrappers identical.
7	/accounts	exists	already-in-website	accounts-screen.tsx is byte-identical.
8	/accounts/[accountId]	exists	already-in-website	website is ahead: adds AccountEntitlementPanel (conductor-backed) and a token-gated sign-in check; parlor has nothing website lacks.
9	/invitations/[token]	exists	already-in-website	the only diff is the brand word (parlor → Cafaye).
10	/admin	exists	already-in-website	website is ahead: EntitlementOps (list/grant/revoke Kit entitlements via conductor) is 300+ lines parlor’s console does not have; parlor’s has nothing website lacks.
11	/settings (+ mfa-section.tsx)	exists	already-in-website	both files byte-identical.
12	/terms, /privacy	exist	already-in-website	brand-only diffs (title + placeholder name).
13	/healthz, /readyz	exist	already-in-website	handlers identical.
14	/v1/* BFF + /v1/auth/oauth/$provider/callback	exist	already-in-website	website is ahead: the forwarder also targets conductor; the callback route differs only by brand/test ids.
15	Session shell (src/components/shell/header.tsx)	exists	already-in-website	the only diff is the brand word.
16	Design system (src/styles/tokens.css, src/components/ui/)	exists	already-in-website	tokens.css identical; the ui barrel differs only in export order. See observation (b) on button.tsx.

No family is marked move. Every screen and route family parlor has, website already has — and in every case where the two differ, website is ahead (brand rename, conductor entitlement surfaces, landing pricing, footer, meta). There is nothing left in parlor’s screens or routes that website lacks.

Already-planned, not missing

The future-screen promises in AGENTS-FIRST W2–W5 all land in website by construction, so their absence from parlor is correct, not a gap:

Promise	Packet	parlor has it?
/agents/claim/[token] approval screen	W2-P1	no — correct
/email-providers (+ /cloudflare form)	W3-P3	no — correct
/audit + release-acceptance screens	W3-P3, W5-P2	no — correct
/boxes	W3-P3	no — correct

Secondary observations (outside the screen/route surface, recorded for honesty)

These were found while making the reasons accurate. None is a move; the brief bounds this packet to screens/routes, and each is a component or a doc row, not a screen.

• (a) shadcn primitives. parlor’s badge/separator/tabs are the house-built versions (no radix/cva, SSR-safe) and carry tests; website wears stock shadcn/ui instead, without tests. A component choice website already made — not a screen, and not ours to reverse here.
• (b) ui/button.tsx ink variant + press scale. parlor’s button has them; website’s does not. But no parlor screen uses ink — only parlor’s own test does — and website’s landing sources ink from shadcn/button.tsx, which has it. Moving it would be shipping an unneeded move.
• (c) A stale README row in both trees. Both READMEs list /billing/plans under “What is here today”; neither tree has that route (the billing screens were removed from the product; the READMEs’ own billing sections say so). Shared, not a divergence — but the sunset note below should not be read as the only stale copy.

Conclusion

The audit justifies no moves. Every screen/route family parlor has, website already has, and website is ahead in every divergent case. The reconciliation D9 calls for is therefore complete on the evidence: nothing is worth moving, and the sunset note in the README is the remaining deliverable.