Execute the SvelteKit app migration (DEC-0043)

Migrate chrome-light app surfaces to SvelteKit; docs stay on Astro/Starlight (two apps, one origin) decided the topology — chrome-light surfaces move to a new SvelteKit app (recommended @dossier/app); docs stay on Astro + Starlight (recommended rename @dossier/site → @dossier/docs, docs-only); both unified on one Vercel origin via a /knowledge/:path* rewrite; a shared @dossier/okf-view leaf extracted so both apps read the KB from one source of truth. This atom tracks the execution.

Why a task, not a fix-in-place

This is multi-day, multi-phase build work (DEC-0043 estimates ~4–6 focused days, dominated by the 488-line landing port) requiring the Principal Forward Deployed Engineer's judgment — the phased plan, the preview-URL proving step, and the single apex-domain cutover are the FDE's to drive. The log-auditor surfaced it from the decision and files it so the follow-up is durable, owned, and trackable rather than a line in a reply. confidence: inferred (agent-derived from the decision's named follow-ups, not human-curated).

Status (2026-06-17 — done): LIVE on production, every surface unified on one origin

The Principal Forward Deployed Engineer (orchestrating the Principal SvelteKit Engineer + Astro Starlight Engineer) executed the full migration of plans/sveltekit-migration.md, all gate-verified, and both Vercel projects are now live on production — confirmed by the user this turn ("everything looks good on prod, sites are loading"). This atom moves review → done and DEC-0043 is promoted asserted → verified. See Dossier — Decision & Audit Log (2026-06-17, "DEC-0043 VERIFIED").

• MET (built + verified): @dossier/okf-view extracted and consumed by both apps (Phase 0); @dossier/app renders every chrome-light surface (Phases 1–4); docs stripped to /knowledge/* (90 pages) with Starlight retained and @dossier/site → @dossier/docs applied — package name only, dir stays packages/site (Phase 5); zero nav flash — startViewTransition 0× on client nav, proven in-browser (the DEC-0043 trigger, dead); theme continuity via the kept starlight-theme key + app.html no-FOUC script; pixel-faithful landing port; full suite 385 passed / 1 skipped and all gates green.
• MET (the cutover — DEC-0043 promotion gate, LIVE): both projects deployed — app on dak-dossier.vercel.app, docs on dak-dossier-docs.vercel.app. Live curl confirms /, /board, /graph, /api/subscribe served by the app and /knowledge/, /decisions/*, /tasks/* proxied + styled from the docs on one origin. The unifier is the guarded catch-all in packages/app/vercel.json — not the originally-planned /knowledge/:path* rewrite, which 404'd the top-level atom routes (/decisions/* etc., not /knowledge/*) and left Starlight's root-level hashed assets unstyled; corrected in commit 61b74d0. No custom apex domain exists — production is the Vercel project domain, which satisfies the gate; a later custom-domain attach is trivial.
• DEFERRED fast-follow (now its own task, still open): Svelte-ify the board's ~270 lines of vanilla JS carried verbatim in onMount (Phase 7) — split out to Svelte-ify the board's ~270 lines of vanilla-JS interactivity (the DEC-0043 / Phase-7 fast-follow), NOT dropped.
• Surfaced during execution (filed separately): re-home the Buttondown email templates now /api/subscribe moved → Re-home the Buttondown email templates from packages/site to @dossier/app (the subscribe endpoint moved); decide whether @dossier/docs keeps the now-unused @vercel/analytics → Decide whether @dossier/docs keeps @vercel/analytics now that the landing (its only consumer) moved to @dossier/app; the generate-landing skill + DEC-0037 prose point at the old content/landing path → folded into Fix the stale round-trip-test path in the generate-landing SKILL.md (points at a file that no longer exists).
• Decisions taken mid-build (recorded same-turn): Lenient KB-atom reader in @dossier/okf-view (readKbAtoms) — faithful getCollection reproduction for the SvelteKit app (the lenient readKbAtoms reproducing Astro's getCollection profile) and Persistent read-only chrome via an (app) route-group layout — stop the header re-animating on intra-app nav (the (app) route-group layout so the header doesn't re-animate on intra-app nav).

Guardrails (from DEC-0043)

• No big-bang. Move the apex domain exactly once, at the final cutover, after the full SvelteKit app is proven on a preview URL.
• Two-way door. The rewrite + domain assignment are trivially reversible (subdomain split kept as a fallback).
• Pixel-faithful landing. Port the 488-line landing per the Recalibrate the Dossier brand identity — demote color, promote type + restraint + craft — then build the showcase landing content-model invariants.
• Stay on the stack. SvelteKit is Vite-based — the VoidZero/Vite TypeScript lean (CLAUDE.md) holds; no new build paradigm.
• Sovereignty preserved. Both apps stay derived, replaceable surfaces over the OKF git repo (Adopt OKF as Dossier's canonical knowledge format); design tokens (Establish the design system and the UX-engineering function) consumed unchanged.

Board globbed before filing — no open task covered the SvelteKit migration.