chefmrfrizzle · chefmrfrizzle · Aug 17, 2026 · Aug 17, 2026 · Aug 17, 2026 · Aug 17, 2026
@@ -0,0 +1,7 @@
+node_modules/
+dist/
+.DS_Store
+implementation-desktop-1440x1024.png
+implementation-mobile-390x844.png
+.vercel
+.env*
@@ -0,0 +1,2 @@
+fund=false
+audit=false
@@ -0,0 +1,4 @@
+{
+  "d1": null,
+  "r2": null
+}
@@ -0,0 +1,16 @@
+# Prototype Instructions
+
+Run the local server yourself and open the preview in the browser available to this environment. Do not give the user server-start instructions when you can run it.
+
+Before making substantial visual changes, use the Product Design plugin's `get-context` skill when the visual source is unclear or no longer matches the current goal. When the user gives durable prototype-specific design feedback, preferences, or decisions, record them in `AGENTS.md`.
+
+When implementing from a selected generated mock, treat that image as the source of truth for layout, component anatomy, density, spacing, color, typography, visible content, and hierarchy.
+
+## Selected Valoris direction
+
+- The approved Reviewer Sandbox direction is Option 1: a guided, five-step desktop review workspace with a persistent task sidebar, role switcher, explicit orange DEMO/SYNTHETIC warning, and supporting/challenging evidence presented side by side.
+- Keep the visual language restrained and institutional: white and pale-gray surfaces, dark navy type, vivid blue actions, orange warning states, compact data density, subtle borders, and small radii.
+- Preserve the public-static boundary: no backend, authentication, uploads, private materials, production components, scientific-acceptance claims, or independence claims. Every fixture datum must carry a DEMO or SYNTHETIC label, and cryptography remains BLOCKED_UNVERIFIED.
+- The top Perspective control and the four “I’m reviewing as” buttons are two synchronized inputs for the same active reviewer role. The top control must be a functional selector on mobile and desktop, not a focus-only shortcut.
+
+Build app UI in `src/`. Keep `.openai/hosting.json`, `worker/index.js`, `scripts/prepare-sites-build.mjs`, and `tests/sites-worker.test.mjs` intact so the same local prototype can be handed to Sites. Before a Sites handoff, run `npm run build` and `npm run test:sites`; the build must leave `dist/client/index.html`, `dist/server/index.js`, and `dist/.openai/hosting.json`.
@@ -0,0 +1,74 @@
+# Vercel Deployment Record
+
+- **Stable URL:** https://valoris-reviewer-sandbox.vercel.app
+- **Deployment ID:** `dpl_5vv3pjqr7D8Gr4jcmrdjZ7kWT8fY`
+- **Vercel project:** `halalmfs-projects/valoris-reviewer-sandbox`
+- **Target:** Production — promoted from validated preview `dpl_3YjYLjAWMLs5aEN8DUAjwP46JqYj`.
+- **Source commit:** `2abf130c878875ca7a51cd6d1dd61efded7c15fd`
+- **Status:** `READY`
+- **Artifact:** Prebuilt static Vite output from `dist/client`; no functions, API routes, backend, authentication, or upload endpoints.
+
+## Verification
+
+- Stable alias returned HTTP 200 without an authentication bypass.
+- Deployed HTML title and metadata match the Reviewer Sandbox build.
+- Security headers are active: restrictive Content Security Policy, denied framing, disabled camera/microphone/geolocation, no-referrer policy, MIME sniffing disabled, and Vercel HSTS.
+- Vercel marks the deployment `noindex`, appropriate for a reviewer demonstration.
+- Browser-rendered mobile verification passed at 390 × 844 CSS pixels: document width 390, no horizontal page overflow, single-column evidence layout, static task panel, active step visible, and zero console warnings/errors.
+- Browser evidence: `qa/deployed-mobile-390x844.png`.
+- Mobile interaction checks passed for role switching, the ADR dialog, `BLOCKED_UNVERIFIED` cryptography, and step navigation.
+- Vercel runtime error scan returned no logs; this is expected for a static deployment with no functions.
+
+## Boundaries
+
+The deployment remains DEMO/SYNTHETIC. It makes no scientific, security, cryptographic, production-readiness, or independence claim. Local Vercel link metadata and environment files are ignored and are not committed.
+
+## Reviewer handoff preview — 2026-08-17
+
+- **Preview URL:** https://valoris-reviewer-sandbox-aeybsb0lk-halalmfs-projects.vercel.app
+- **Deployment ID:** `dpl_3YjYLjAWMLs5aEN8DUAjwP46JqYj`
+- **Target:** Preview; the stable production alias was not promoted or changed.
+- **Source commit:** `2abf130c878875ca7a51cd6d1dd61efded7c15fd`
+- **Status:** `READY`
+- **Build artifact:** `index-CAorDIqP.js` and `index-wMgKUQ3e.css` from the static Vite output.
+- **Live verification:** Exact deployed title and synthetic boundaries rendered successfully. At 390 × 844 CSS pixels, the role-aware step-5 handoff generated for Materials Science, all three handoff actions remained reachable, document width stayed 390 with no horizontal overflow, and the console had zero warnings or errors.
+- **Production check after preview:** `https://valoris-reviewer-sandbox.vercel.app` remained deployment `dpl_Ff6fq9pDQx5LyAiPi8QKKxVvkMqi`, target `production`, status `READY`, and returned HTTP 200.
+
+## Production promotion — 2026-08-17
+
+- **Promoted preview:** `dpl_3YjYLjAWMLs5aEN8DUAjwP46JqYj`.
+- **Resulting production deployment:** `dpl_5vv3pjqr7D8Gr4jcmrdjZ7kWT8fY` at `https://valoris-reviewer-sandbox-fv1mxtivr-halalmfs-projects.vercel.app`.
+- **Stable public alias:** https://valoris-reviewer-sandbox.vercel.app returned HTTP 200 and resolved to the resulting production deployment with status `READY`.
+- **Artifact identity:** production served `index-CAorDIqP.js` and `index-wMgKUQ3e.css`, matching the validated preview artifact.
+- **Pre-promotion regression gate:** 7/7 sandbox and review-packet tests passed, 4/4 static-host packaging tests passed, and the Vite static build completed successfully.
+- **Desktop gate:** 1440 × 1024 CSS pixels, no horizontal overflow, exact DEMO/SYNTHETIC and BLOCKED_UNVERIFIED boundaries present, and zero console warnings/errors.
+- **Mobile gate:** 390 × 844 CSS pixels, no horizontal overflow, role-aware Cryptography handoff generated, all three handoff actions reachable, and zero console warnings/errors.
+- **Static boundary:** no backend, authentication, uploads, production components, file inputs, form targets, or Vercel Functions were present. The deployed artifact remained a static Vite bundle.
+- **Security headers:** restrictive Content Security Policy, framing denied, camera/microphone/geolocation disabled, no-referrer, MIME sniffing disabled, and HSTS remained active.
+- **Post-deploy error scan:** Vercel returned no runtime logs, expected for this static deployment.
+- **Promotion boundary:** the deployment remains a DEMO/SYNTHETIC reviewer interface and does not accept scientific, security, cryptographic, production-readiness, or independence policy.
+
+## Mobile overlap correction — 2026-08-17
+
+- **Validated source commit:** `048828823e91d83b882cdf54fe47a6cd5a90e7cd`.
+- **Validated preview:** `dpl_Ao99wJAbqzMKBquVr61X1oCvhKFn` at `https://valoris-reviewer-sandbox-jiuzsae9m-halalmfs-projects.vercel.app`, target `preview`, status `READY`.
+- **Resulting production deployment:** `dpl_D9U1SqbUZN7ZXLyfJjzezcJtANic` at `https://valoris-reviewer-sandbox-eabn3anzq-halalmfs-projects.vercel.app`, target `production`, status `READY`.
+- **Stable public alias:** https://valoris-reviewer-sandbox.vercel.app returned HTTP 200 and resolved to the resulting production deployment.
+- **Artifact identity:** preview and production serve `index-FM-UCdAY.js` and `index-BXtRsd-W.css`. Downloaded production assets exactly matched the promoted prebuilt artifact: JS SHA-256 `4f8b86f503be6fe91370dd4fc93ad9ec5ab99439af054c4768479fad8baf5f69`; CSS SHA-256 `60d5630b783cddb11b1cea98246111dbb3ac4aac7521e42c940121c54333e280`.
+- **Regression gate:** 8/8 sandbox and review-packet tests passed, 4/4 static-host packaging tests passed, the Vite static build passed, and the repository baseline passed.
+- **Responsive gate:** all four roles passed at 390 × 844; Materials Science passed at 320 × 800; desktop passed at 1440 × 1024. The brand, perspective control, global warning, and active walkthrough step remained contained with no document overflow or console warnings/errors.
+- **Reported-versus-fixed evidence:** `qa/mobile-overlap-source-vs-fixed.png`; focused corrected captures are in `qa/mobile-header-fixed-390x844.png`, `qa/mobile-header-fixed-320x800.png`, and `qa/desktop-header-fixed-1440x1024.png`.
+- **Headers and observability:** restrictive CSP, framing denial, disabled camera/microphone/geolocation, no-referrer, MIME-sniffing protection, HSTS, and `noindex` remained active. The post-deploy runtime error scan returned no logs, expected for this static deployment.
+- **Boundary confirmation:** the promoted artifact remains static and synthetic, keeps cryptography `BLOCKED_UNVERIFIED`, and introduces no backend, authentication, uploads, API routes, Vercel Functions, private materials, or scientific/independence claims.
+
+## Functional perspective selector — 2026-08-17
+
+- **Validated source commit:** `32f7e3047c20ba2b0806c1386b0ad7b359d401f2`.
+- **Validated preview:** `dpl_BXDijKNvLS87RM7nzXuuCdUnmbXV` at `https://valoris-reviewer-sandbox-cy1ipox2b-halalmfs-projects.vercel.app`, target `preview`, status `READY`. Browser access remained deployment-protected; Vercel-authenticated retrieval verified the exact artifact before promotion.
+- **Resulting production deployment:** `dpl_HtfKgVR5R9pvhjbxmYm5xkSWrWJK` at `https://valoris-reviewer-sandbox-p9rm6in6v-halalmfs-projects.vercel.app`, target `production`, status `READY`.
+- **Stable public alias:** https://valoris-reviewer-sandbox.vercel.app returned HTTP 200 and resolved to the resulting production deployment.
+- **Artifact identity:** production serves `index-_9axtI6i.js` and `index-D7JI5rEz.css`, exactly matching the promoted prebuilt preview. JS SHA-256: `221cd60bef6f62a405c964cafc260ab2a6c311fd827fcc0396c94613527c6a04`; CSS SHA-256: `3bc359bbf0e6d5f722474672dd4ff332f0439e19c61c3a31fafae7ccb455f816`.
+- **Interaction gate:** on the public alias at 390 × 844, the top selector changed to Cryptography and synchronized the lower active role plus ADR-0012 packet; the lower Materials Science role then synchronized the top selector plus ADR-0010/ADR-0014 packet. `DEMO / SYNTHETIC` and `BLOCKED_UNVERIFIED` remained visible, with no horizontal overflow or runtime errors.
+- **Regression and accessibility gate:** 9/9 sandbox and review-packet tests passed, 4/4 static-host packaging tests passed, the Vite static build and repository baseline passed, and the final automated WCAG A/AA scan reported zero violations.
+- **Headers and observability:** restrictive CSP, framing denial, disabled camera/microphone/geolocation, no-referrer, MIME-sniffing protection, HSTS, and `noindex` remained active. The post-deploy runtime error scan returned no logs, expected for this static deployment.
+- **Boundary confirmation:** this interaction correction adds no backend, authentication, uploads, persistence, API routes, Vercel Functions, private materials, or scientific/independence claims; cryptography remains `BLOCKED_UNVERIFIED`.
@@ -0,0 +1,19 @@
+# LearningEpisode — Public Reviewer Sandbox
+
+- **Trigger:** The founder selected visual direction 1 for a public, static, synthetic Reviewer Sandbox.
+- **Inputs:** The selected 1487 × 1058 design image; the existing Valoris UI system; the onboarding architecture; the current UX audit; the public architecture; ADR-0006, ADR-0007, and proposed ADR-0010.
+- **Invariants:** No backend, authentication, uploads, executable schemas, private materials, production components, scientific-acceptance claim, or independence claim. All fixture data is DEMO/SYNTHETIC. Cryptography stays BLOCKED_UNVERIFIED.
+- **Output:** A five-step, ten-minute React walkthrough with four reviewer perspectives, side-by-side supporting and challenging evidence, bounded ADR packets, limitations, and structured local feedback.
+- **Failure observed:** At 390 CSS pixels the horizontal walkthrough initially opened on steps 1–2 while active step 3 was offscreen.
+- **Correction:** Center the active step on initial responsive render and after navigation; retain the current step in view without changing desktop layout.
+- **Handoff extension:** Added a role-aware, session-only review summary with explicit conflict and blocked-decision fields, copy and JSON download actions, and a clean-session reset. No timestamp, reviewer identity, transport, or storage was introduced.
+- **Failure observed during extension:** The download action revoked its temporary object URL immediately after the synthetic link click, creating a race for slower browser download handlers.
+- **Correction during extension:** Attach the download link to the document, remove it after activation, and defer URL revocation to the next task so the browser can consume the synthetic JSON reliably.
+- **Reported-device failure:** User-supplied iOS captures showed the mobile perspective control obscuring the product name and the horizontally centered walkthrough clipping the previous step label.
+- **Reported-device correction:** At ≤600 CSS pixels, place the brand, perspective control, and warning in explicit grid rows; show one complete active walkthrough step; reserve automatic horizontal centering for the 601–860-pixel tablet layout.
+- **Regression protection:** A source-level test now requires the mobile single-column header, contained perspective control, and active-only phone walkthrough rules.
+- **Interaction failure:** The top Perspective affordance visually implied a dropdown but only moved keyboard focus to the first lower role button, so it could not change reviewer viewpoint.
+- **Interaction correction:** Replace the focus shortcut with a native controlled selector that shares `roleId` with the lower role buttons. Both entry points now update the selected label, active role, task language, and ADR packet.
+- **Accessibility correction:** The interaction audit exposed a serious automated contrast violation in the orange warning and evidence metadata. Darker semantic tokens reduced the follow-up WCAG A/AA result to zero violations; gradient-backed contrast checks remain machine-incomplete and were visually reviewed.
+- **Evidence:** Desktop and mobile browser captures, a combined reported-versus-corrected comparison, synchronized selector/button checks at 320, 390, and 1440 CSS pixels, zero runtime errors, passing static build, 9/9 sandbox and review-packet tests, and 4/4 static-host packaging tests in `design-qa.md`.
+- **Promotion decision:** The static prototype was published to a dedicated Vercel project for founder and reviewer inspection. Vercel assigned the first deployment to its production target automatically; this makes the static interface reachable but does not imply protocol production readiness, policy acceptance, scientific acceptance, or independence assurance.
@@ -0,0 +1,32 @@
+# Valoris Reviewer Sandbox
+
+A public-safe, static, synthetic walkthrough for evaluating the Valoris reviewer experience.
+
+## Boundaries
+
+- Every fixture datum is labeled `DEMO` or `SYNTHETIC`.
+- Cryptographic choices remain `BLOCKED_UNVERIFIED`.
+- The sandbox makes no scientific, security, production, or independence claim.
+- It has no backend, authentication, uploads, executable schemas, private materials, workers, payments, or production infrastructure.
+- Feedback remains in browser memory unless the reviewer explicitly copies or downloads the DEMO handoff.
+- The generated handoff is role-aware, includes conflict and blocked-decision fields, preserves the non-approval boundaries, and can be reset without persistence.
+
+## Local commands
+
+```bash
+npm install
+npm run dev
+npm run build
+npm run test:sandbox
+npm run test:sites
+```
+
+The selected design direction and implementation constraints are recorded in `AGENTS.md`. Visual verification is recorded in `design-qa.md`.
+
+## Vercel preview
+
+`vercel.json` deploys only `dist/client` as a static site and applies restrictive browser security headers. Vercel link metadata and local environment files remain ignored and must never be committed.
+
+Live mobile-ready deployment: https://valoris-reviewer-sandbox.vercel.app
+
+Deployment evidence is recorded in `DEPLOYMENT.md`.