FrameCraft Operations

Program dashboard

Re-verified 2026-07-24 by a 16-cluster, 91-item code audit against Anthony's FULL architect handoff (JOHN-HANDOFF + 301 map + personalization + size-spec + SBF content + CFS QA), with every DONE/PARTIAL claim adversarially re-checked (14 over-optimistic claims corrected). Headline: 38 DONE / 36 PARTIAL / 8 NOT-STARTED / 6 BLOCKED / 2 DEFERRED. The #1 non-negotiable — server-rendered BODY content for AI crawlers — is genuinely DONE on all three stores. Most 'PARTIAL' items are one-store-only or behind an off flag, not missing work. The launch-critical gaps: (P0-legal) the personalization consent gate exists as a proofread checkbox but is NOT bound to the order record (no timestamp/text captured as a line attribute) and the ToS 'Personalized and Custom-Text Items' clause is absent; (P0-SEO) the 301 priority tier is done but the long tail hard-404s — Rule 6 catch-all, Rule 4 moulding 1:1s, Rule 5 chains, ~11 category .html + ~17 CMS remaps, and .html/case normalization are all missing; (P0-commerce) the Shopify variant ID + high-priced placeholder still pending. Newly-surfaced, not previously tracked: GTM (GTM-WNC3937K) ships on ZERO flagship pages (CPF has no analytics; store-b only fires with an unset env; only CFS is wired); store-b omits Organization+WebSite JSON-LD site-wide; flagship+store-b frame PDPs have no 120-char meta-description floor (store-b ships 18 four-word taglines); store-a/store-b have NO loading.tsx skeletons at all; voice/brand lint fails on the stores (em-dashes, ~30-40 banned words, exclamations, an emoji); /samples OG image is 1.69MB; the welcome.custompictureframes.com preservation rule is documented NOWHERE in the repo. Two-source reviews avoided the conflation trap (seller single-sourced, product scope distinct) but the live Reviews.io widgets aren't wired. Customer accounts are a full real vertical slice, live only on CFS; CPF/SBF need the same-site /bff auth proxy + flag flip for 'one account across all brands'. RB2B / Clarity / Klaviyo / 3D-spinner / loyalty-vendor are all correctly deferred to post-launch. Full report: docs/reference/LAUNCH_READINESS_AUDIT_2026-07-24.md; reusable rules for future storefronts: docs/reference/migration-playbook/. — PROGRESS SINCE THE AUDIT (2026-07-24, PRs open on develop): (P0-SEO) the 301 long tail is CLOSED — PR #449 adds the Rule 4/5 + category/CMS rows (redirects.js +116, all 54 destinations curl-verified 200) plus a Rule 6 catch-all + case-normalization in middleware, so no legacy .html hard-404s (6 new tests, chains terminate at 200). GTM: the flagship injection is merged (8dace7e); PR #450 corrects the container so flagship no longer reports CPF traffic into CFS's container — it now needs its OWN CPF container id (provisioning, not code). Cross-brand accounts: the same-origin /bff proxy + an origin-aware OAuth redirect are built (PR #61 api + PR #451 monorepo, 109 + 1366 tests) — CPF/SBF now need only the Shopify redirect-URI registration + the enableAccounts flip. RB2B: the pixel plumbing landed dark on /business (PR #450), pending the team id + counsel wording. CPF checkout branding verified fully wired (code-identical to CFS). The top remaining launch blockers are unchanged: the store-a/store-b @sentry SSR crash, secret rotation, and the ToS/counsel signoff. — ALSO 2026-07-25 (more PRs open on develop): the flagship designer-API gap is CLOSED — PR #457 wires the 5 missing routes designers call (design-recommendations + upscale, both behind flags that are ON for flagship, plus proxy-image/frame-images/objects-upload-from-url); flagship now has 7 of 9 (business-quote + asset correctly absent). Designer image bugs fixed — PR #453 resolves raw image paths in ChallengeCoin/BulletinBoard/Heart (thumbnails + default previews were 404ing on the CDN); a scan of all ~30 specialty designers found no others. Mobile preview fixed — PR #454 + #455 replace the scroll-coupled shrink that collapsed the phone preview to a ~200px sliver with a fixed, viewport-capped height across all 29 specialty designers. The cpf-assets-not-uploaded blocker is largely cleared: the full out/cdn-assets staging set (flagship 410 + shared 55) was uploaded to R2 (add-only) and verified 200 — residual is a few never-staged assets (e.g. currency lifestyle) + confirming NEXT_PUBLIC_CDN_* on the Vercel deploys. Blockers still needing owner action: the @sentry store-a/b SSR crash, secret rotation, ToS/counsel signoff. — PROGRESS 2026-07-26 (pre-launch USER-EXPERIENCE SIMULATION, 2 rounds): ran two deep simulations against a LOCAL production build of the flagship + the PRODUCTION api (the deployable state — fixes land on develop unmerged, so a prod sim would re-find already-fixed issues), each rebuilt + Playwright/axe re-swept after fixing. ROUND 1 → PR #459 (MERGED to develop, merge-commit 775b187): 301 fixes incl. legacy MIXED-FRACTION size decode (11-3-4x16-1-2 → 11¾×16½, was dumping to /picture-frames), homepage <title> brand token, word-boundary meta truncation, removal of 'designer is being wired up' placeholder copy from 48 indexable pages, dedup of the ARIA-referenced duplicate <h1> id on ~75 pages, 4 hero images repaired from existing assets (7 more need client photography), first a11y batch (custom-mats criticals, specialty contrast ×31, main+canvas designer nested-interactive). ROUND 2 → PR #461 (OPEN on develop, 2 commits/56 files): a deeper sim caught what round 1 missed — (P1) HYDRATION crash on /framed-bulletin-boards (inline <style>{`…`} with double-quoted selectors → `"`→&quot; SSR-escape mismatch → React #425/#422; fixed via dangerouslySetInnerHTML + unquoted selectors); (P1) SOFT-404 (invalid/unbuilt /frames/sizes/* & /resources/* returned HTTP 200 not-found bodies → added dynamicParams=false, now real 404s); nested-interactive REMOVED from ALL remaining ~22 specialty designers + the nameplate header (closes the deferred-12 a11y debt); single <h1> across the 75-page frame-profile-hero (mobile copy → styled <p>) + 13 designer h1→h2 + blog markdown #→h2 + canvas depth Tabs→role=radiogroup (killed aria-valid-attr-value); 10 mat sliders + 4 icon buttons named; carousel tabIndex; aria-hidden decorative edges; star role=img; samples cards→label+checkbox; bulletin img→button; clampMetaDescription() public in @framecraft/seo funneling ~40 over-160 metas; mobile overflow (newspaper/blog/magazine). Verified: @framecraft/seo + packages/ui + flagship tsc clean, 102 flagship + 125 ui tests pass, jsx-a11y build warnings 10→0, hydration pageerrors 0 / mobile overflow 0 / multi-h1 75→0 on re-sweep. DOCUMENTED, deliberately not code-changed (need a design decision or are non-issues): link-in-text-block (~455 inline text-primary links across 68 files — wants a shared underlined prose-link style; a blanket change would wrongly underline nav/card links; low real impact as links are colour-coded); rotating search-placeholder contrast (passes ~6:1 at rest, only axe-flagged mid-fade = animation artifact); /scarf-and-sash-framing ~44px mobile overflow (minor). Adversarial verification also OVERTURNED 2 tempting fixes: /unfinished-picture-frames 404 is CORRECT (the real legacy URL already 301s), and ticket-frames' meta length was HTML-entity inflation, not truly long. Reviews.io PHASE 1 foundation built + gated OFF (env kill-switch + grounded REST client + clampMetaDescription; runbook at docs/reference/flagship-cpf/REVIEWSIO_INTEGRATION_RUNBOOK.md) on feat/reviewsio-integration — HELD pending review import + the public Store ID. Launch blockers still needing OWNER action are UNCHANGED: publish the Shopify store (cpf-headless still password-protected), restore the prod DB, set NEXT_PUBLIC_SITE_ORIGIN + verify canonicals/noindex, the @sentry store-a/b SSR crash, secret rotation, ToS/counsel signoff, 7 client hero photos. Detail: PR #459 (merged) + #461 (open); docs/reference/flagship-cpf/PRELAUNCH_SIMULATION_2026-07-26.md. — PROGRESS 2026-08-04 (client QA batch #195-#210 + an independent PRE-LAUNCH AUDIT, all landed DIRECT TO master): shipped Brian/Anthony's issue run — signature circle mat cut as a square hole so the accent read as a black box behind a round photo (both mat clip paths branched heart-vs-rect only); heart insert photos curated to the 16 whose composition survives the top-centre notch (every rejected one puts a head or a graduation cap exactly there); a $5 MINIMUM on the non-glare upgrade (Anthony's ruling) — the old T7a 'does not reproduce' verdict was a two-point sample at 16x20/32x40, and below ~16x20 the delta is cents that the round-up-to-.99 erases, so a 4x6 sonogram quoted $22.99 either way; needlework/diamond/cross-stitch/paint-by-number insert images stopped stretching (object-fit fill -> cover) with aspect-matched selection cutting worst-case crop 59%->25%; oversize mat filtering (32x40 vs 40x60) — flagship had FABRICATED 40x60 SKUs for the two texture mats by copying the 32x40 SKU, and 6 designers that can reach an oversize sheet never filtered (Currency/Stamp via CUSTOM sizes, TicketStub only on a narrow moulding — all three found by writing the guard, not by reading the code); hyphenated sizes ('16-1/16' parsed as 1, and the quiet one, '16-1/2' as 8 — a plausible wrong size straight to cart); the shadow-box colour cards, whose 9 CTAs all pointed at a bare /shadow-box-frames that mounted NO designer. — Then checked the independent 2026-08-03 pre-launch audit line by line against the code. Six findings CONFIRMED and fixed (commit 99c02e8): the sitemap listed zero product pages (73 now), which exposed that catalogue ids are LEGACY slugs so getFramePdpHref was emitting redirect sources — the audit's separate ~52-internal-301s finding, same root cause — plus six blog slugs already listed while redirecting away; titles losing their brand was ONE clamp cutting mid-brand with '|' missing from the separator strip, not six typos; garbled alt text was NOT the upstream name field the audit blamed but pre-rendered alts in our own frames.json generated against a superseded naming scheme (106 fixed on flagship, 106 on each of store-a/store-b, which the audit never checked); Product schema now inherits brandConfig's 4.8/4,280 (301 of 330 pages) and /jersey-frames gained the offers it never had; icons + apple-touch-icon on 541/541 pages; the auth-gated photo fetch no longer 401s for anonymous visitors. THREE audit claims were WRONG and deliberately not actioned: the mat multi-pack 'blocker' is Ben's own 2026-07-16 cutover decision (its 'truth build' is the abandoned v2 source; acting on it would roughly halve bulk mat revenue), the 36x48-vs-40x60 pricing anomaly does not reproduce across any of 84 SKUs, and the alt-text root cause was misdiagnosed. Net: the audit is reliable on symptoms and wrong on causes 3 for 3 — two of its prescriptions would have been wasted or harmful. REMAINING from it: the jersey/vinyl/graded-card base-price offsets (the one real blocker, needs Anthony's ruling), the /specialty-frames hero (absent from R2 — an upload, not code), alt-string LENGTH (1,642 of 2,694 over the 125-char window, worse than the audit's ~100), and 29 pages still hand-rolling Product schema.