Migrate

stardust:migrate

Apply the target spec authored by direct, the visual canon written by prototype --prep, and the brand-module catalog extracted during prepare-migration to every page in the inventory. Produces a self-contained, deployable static HTML site under stardust/migrated/. Per-page, incremental, idempotent.

migrate is the final stardust phase. Output is platform- agnostic HTML — downstream conversion (AEM EDS, a CMS, a framework) is the job of a separate plugin that consumes migrated/ plus DESIGN.json plus the per-page _meta.json sidecars.

Inputs

• <slug> — optional positional. Migrate just this page. Without it, migrate every page whose status is directed, prototyped, or approved (and not stale).
• --all — migrate every page including stale ones.
• --force — re-migrate every page even when the idempotent skip would skip them.
• --require-approved — refuse to migrate any non-approved page. Default behaviour migrates directed pages too (using Path A′ or Path B per reference/template-and-module-rendering.md); this flag flips approval-gating on.
• --strict-canon — refuse approvals that conflict with canon. Default logs the deviation and continues. Useful for projects where canon discipline matters more than per-template flexibility.
• --clean — delete assets previously bundled but no longer referenced from stardust/migrated/assets/. Off by default (migrate is additive). Implies --force: every page is re-rendered so the run's bundledAssets Set is the complete union of currently-referenced assets — otherwise --clean would risk deleting assets still referenced by idempotent-skipped pages. See reference/asset-bundling.md § Stale asset cleanup.
• --pin-timestamp <ISO8601> — pin the migrate-provenance timestamp so re-runs without source changes produce byte- identical HTML. Default re-uses the current wall clock, which is fine for normal use; CI deployment fingerprinting may want the pin.

The mobile-adapt audit, content-sourcing scan, and placeholder refusal are all mandatory gates — there is no --skip-* or --allow-* flag to bypass them. If a gate refuses a page, the remediation is to fix the proposed file (re-prototype, edit inline, or run an impeccable command) and re-invoke migrate.

Setup

0. Playwright re-probe (mandatory first step). --no-save playwright installs from earlier phases are pruned by any later real npm i (extract SKILL.md § Setup → --no-save installs are ephemeral). Before any rendering step, probe node -e "import('playwright').then(()=>process.exit(0))" from the project root and re-install (npm i -D playwright --no-save --legacy-peer-deps) on failure.

1. Run the master skill's setup (skills/stardust/SKILL.md § Setup). Flow guard. If stardust/state.json exists without flow and the ask is a migration (a URL plus "migrate" / "to EDS" / "re-platform"), do not run: print the two-flow table from the master skill § Two migration flows and hand back to its routing — the flow is chosen and stamped there before any sub-skill runs (skills/stardust/reference/state-machine.md § Flow keys): replica unless the ask names a redesign. (Recorded: migrate <url> as the first command of two same-design migrations led to a hand-built compiler tuned by eye instead of the replica gate.)

2. Verify stardust/state.json exists with at least one directed page.

3. Verify project-root DESIGN.md and DESIGN.json exist with DESIGN.json.extensions.canon populated.

4. Verify stardust/canon/ exists with at least header.html, footer.html, canon.css.

   Canon auto-bootstrap (when steps 3–4 find no canon). The documented prototype → migrate → deploy happy path does not run prepare-migration, so a first migrate legitimately arrives with no canon (observed on 4 of 6 e2e sites, where every run had to derive canon by hand to proceed — this is the fix). When canon is absent and at least one approved prototype exists, do not stop: run the canon write-back inline from the first approved prototype (the canon-author, default home) per ../prototype/reference/canon-extraction.md § Five-step procedure — extract header.html / footer.html / canon.css to stardust/canon/, pin tokens + compositional moves to DESIGN.json.extensions.canon, and record canon.source: "auto-bootstrap: <slug>". This is exactly what prototype --prep does on first approval; migrate performs it on demand so the core pipeline never dead-ends. Only stop and recommend $stardust prepare-migration when canon is absent and no approved prototype exists (there is nothing to derive canon from). Under state.json.handsOff the bootstrap is automatic and logged; interactively, surface it as a one-line notice before proceeding.

5. Verify stardust/direction.md has an active (not pending) direction.

6. Read state.json.pages[] and partition into:

   • inScope: status directed, prototyped, or approved, stale: false (or --all / explicit <slug>).
   • skipped: everything else, with reason captured.

7. Validate provenance on every in-scope page. Call validateProvenance(page) per skills/stardust/reference/state-machine.md § Provenance validation for every page in inScope. Abort with the helper's error when any page lacks live-render evidence — migrating a synthesized page record produces deployable HTML that misrepresents the source site, the exact failure mode that motivated the validator. Surface Provenance OK on N pages in the migrate-plan output before Phase 1.

8. Mobile-adapt audit on every Path A / Path A′ source. For every page whose render branch consumes a proposed or archetype HTML file (Path A, Path A′ per reference/template-and-module-rendering.md § Render path selection), run the audit per skills/prototype/SKILL.md § Mobile-adapt audit:

   • <meta name="viewport" content="width=device-width, ..."> present, width not pinned to a fixed pixel value.
   • At least one @media (max-width: ...) rule.
   • At least one mobile-targeted breakpoint at ≤ 640px.

   Refuse pages that fail — the audit is mandatory; there is no skip flag. The user fixes the proposed file (re-prototype or chat-driven impeccable command) and re-invokes migrate. Record the audit result per page in the migrate report and in the post-render _meta.json#audit.adapt sidecar. Path B (unique-renders) skips the audit because adapt hasn't run on those pages — a Path B page that needs mobile coverage gets it via $impeccable adapt invoked separately by the user. Surface this distinction in the report so it's not read as a silent skip.

Procedure

Phase 1 — Plan

Dynamic-surface precondition (safety net). If stardust/dynamic-features.md is missing, the hand-run flow (extract → direct → prototype → migrate) never passed a pre-import gate: run the stardust dynamics skill Phases 1–3 now (extract --dynamics for reach if needed, detector on the archetypes, triage draft, curate) before rendering any page. Never import a site as static without a decision per dynamic row. Per page, rows of the inventory that touch it become contentDeviations[] kind: "dynamic-dependency" entries (reference/content-preservation.md § Dynamic dependencies).

Gated-archetype precondition (flow: replica). Before rendering any sibling-tier page, read stardust/replica/progress.json: the page type's archetype must have a gate result at every configured breakpoint that is pass: true, or over the bar with every residual carrying a cause (../replica/reference/source-fidelity-gate.md § Residual logging format — a documented residual is a pass with an asterisk). An archetype never gated, or over the bar with no residual entries, blocks its page type: report the archetype slug and $stardust replica <archetype>, and render nothing for that type. The same rule guards rollout Setup; the published-origin re-gate is unchanged. Thresholds are the gate's.

Print the plan and wait for confirmation when the scope is large:

migrate plan
============

In scope: 127 pages
  Path A  (approved)                6 pages: home, news/post-housing-summit, news, ...
  Path A' (template-applied)      118 pages: 84 article, 5 listing, 11 program, 2 form, 16 static
  Path B  (unique)                  3 pages: 404, search, faq

Skipped:  0 stale, 0 unscoped

DESIGN.md sha:    1a2b3c4
DESIGN.json sha:  5d6e7f8
Canon shas:       header:7g8h9i  footer:9i0j1k  css:1k2l3m

Output:           stardust/migrated/  +  per-page _meta.json sidecars
Idempotent skip:  enabled (run with --force to override)

Reply "go" to proceed.

For 1-3 pages or <slug> invocation, skip the confirmation.

Phase 2 — Per-page render

In the replica flow a sibling renders only after its archetype is deployed and gated at the published origin (../replica/reference/handoff-contract.md § 3 row C, unit C-archetype; #126); the render is the encoder's input, its gate is the row after PUT. For each page in scope, follow reference/migration-procedure.md and reference/template-and-module-rendering.md:

• Idempotent skip check first (sha-compare across designMd, designJson, sourceCurrent, sourceProposed, canonShas, archetypeSource).
• Placeholder gate (Path A and Path A′ only — pages with a proposed file or archetype). Refuse when [data-placeholder] elements or non-empty _provenance.unsourcedContent[] are present — the user fills the missing content in the proposed file before re-invoking migrate. No bypass flag.
• Render branch selection (LLM judgment per T&M § Render path selection): A / A′ / B. Declare the page's fidelityTier from the branch — A → archetype (craft-gated), A′ → sibling (canon-fork, the cheap default for breadth — variance-probed once per template before cloning, reference/fidelity-tiers.md § Sibling variance probe; deltas become variant classes, never per-page forks), B/bodyless → thin — per reference/fidelity-tiers.md. Record fidelityTier, archetypeSource, template (the archetype's slug; the archetype's own sidecar leaves it null), modules[] (the page's block ids — composite sections only) and gatesPassed[] in _meta.json so coverage shows what was craft-gated vs cloned and rollout groups and dedups from the sidecars, without a census.
• Render per the chosen branch's procedure in T&M.
• Canon application — chrome injection, canon.css injection, deviation logging.
• Module rendering — render module instances via stardust/canon/modules/<id>.html; bespoke slots logged with data-bespoke.
• Apply content-preservation rules per reference/content-preservation.md. Internal-link rewriting always emits migrated-tree paths; missing slugs flagged broken.
• Content-count acceptance per reference/fidelity-tiers.md § Content-count acceptance: compare role-classified node counts (headings, body/list nodes, CTAs, images) between the captured source page JSON and the rendered result. A count drop in any class not covered by a logged contentDeviations[] entry fails the page — dropped-content importer bugs must surface here, while the importer is still cheap to fix, not at a downstream fidelity gate. Record the pass in _meta.json#gatesPassed[] as "content-count".
• Compose <head> metadata per reference/metadata-and-jsonld.md (five categories; page-type-driven JSON-LD).
• Validate per T&M § Validation contracts. Strict contracts refuse the page; soft contracts log and continue.
• Compute output path per migration-procedure.md § Output path mapping.
• Asset bundling. Scan the final HTML for asset references (six detection shapes per reference/asset-bundling.md § Detection), copy each unique referenced subpath from stardust/current/assets/<subpath> to stardust/migrated/assets/<subpath> (preserving subdir structure), then rewrite every reference to the root-relative form /assets/<subpath>. Cross-page dedup uses a module-level Set seeded from state.json.migrate.bundledAssets[]. Missing source assets warn-and-skip per § Edge cases; the bundle stays internally consistent.
• Media reconciliation. For every image not bundled to same-origin (reused source-CDN URLs under Mode A image-reuse), decide optimize/keep/rewrite/omit per reference/media-reconciliation.md. Cross-origin <img> kept as source URLs must skip createOptimizedPicture (it drops the ?v= key and corrupts the rendition); broken URLs are repaired (missing ?-delimiter, wrong host) or omitted, never shipped as about:error. rollout re-runs the authoritative network resolve at delivery (media-reconcile.mjs).
• Cinematic sibling (when <slug>-cinematic.html exists). Migrate consumes the STATIC prototype only — the cinematic layer is never merged. Copy the motion assets (lenis.min.js, lenis.min.css) from stardust/prototypes/ to stardust/migrated/assets/motion/ (idempotent) for downstream consumers (deploy/rollout decide whether to wire them), and record cinematic-variant-not-consumed in the page's _meta.json#migrationDecisions[].
• Write the migrated index.html and the _meta.json sidecar in the same directory. Provenance block as first child of <head>. Record assetsBundled (count of unique asset refs on this page) in _meta.json.

The mechanics of this phase are scripted: node skills/migrate/scripts/migrate.mjs render <slug…|--all> (the project copy runs from stardust/scripts/migrate/) builds the page map, places each page at its URL-literal path, rewrites internal links, bundles assets, composes <head> (provenance, :root, canonical and JSON-LD defaults), validates strictly and writes the sidecar, skipping pages whose input shas are unchanged. The judgments it cannot make are recorded on the sidecar with gate, deviation, decision, variant and modules on the same script; it never advances pages[].status.

Recorded units. A whole-site render is bookkept as units in stardust/migrate/progress.json, the shape of rollout's C-deliver unit ledger (../replica/reference/handoff-contract.md § 3 row C): units.<name>: {status: pending|running|done|failed, kind: plan|render|assets|report, templates[], pages: n, startedAt, endedAt, verdict}. One plan unit (Phase 1), one render unit per template cluster of at most ~8 siblings (migrate.mjs render <slug…> takes the cluster's slug list), one assets unit when Phase 3 rehosts media, one report unit (Phase 4). Per unit, in this order: the progress.json write → the checkpoint commit → the next unit; a unit end is a safe resume point, not a stop. A session resumed inside migrate reads the file and runs only the units not done; a unit left running re-renders its slugs — the driver is idempotent, unchanged pages are skipped. A recorded run rendered 31 siblings through five parallel builders as one unit in one session — its most expensive — with no resume point between the plan and the final report.