design: dev-dashboard revamp umbrella (analysis + Claude-Design prompts)

.gitignore

@@ -15,3 +15,12 @@ dist/
 # CI output land here and embed machine-local absolute paths. Never commit it.
 # If a specific artifact ever needs to be durable, re-add it with a `!` negation.
 .llm/tmp/
+
+# design-sync scratch output (synthetic React package + canvas bundle);
+# rebuild with `deno task design:sync`, never commit.
+.ds-sync/
+# local-preview copies of the generated bundle so `screens/*.html` resolve
+# `../_ns_*` when served from resources/design/dashboard; copy from
+# .ds-sync/bundle after each build, never commit.
+resources/design/dashboard/_ns_runtime.js
+resources/design/dashboard/_ns_styles.css

.llm/runs/dashboard-design--orchestrator/briefs/codex-thread-ids.md

@@ -0,0 +1,16 @@
+# ddr-codex-ux — Codex implementation thread
+- **Thread / session id:** `019f5355-4cc0-7213-ac0f-fa5b9767c940`
+- **Rollout:** `/home/codex/.codex/sessions/2026/07/12/rollout-2026-07-12T00-38-47-019f5355-4cc0-7213-ac0f-fa5b9767c940.jsonl`
+- **Worktree:** `/home/codex/repos/ns-ddr-codex`
+- **Branch:** `design/ddr-s3-codex-ux` (NO upstream by design).
+- **Push rule:** explicit refspec only — `git push origin HEAD:refs/heads/design/ddr-s3-codex-ux`.
+- **Requested route:** provider=openai · model=gpt-5.6-sol · effort=max
+- **Observed route:** provider=openai · model=gpt-5.6-sol · effort=low
+- **Route verdict:** mismatch (effort)
+- **Runtime:** approval=never · sandbox=dangerFullAccess
+- **Brief (staged):** `/home/codex/ddr-codex-ux-brief.md`
+## Steering (same thread — never a second send-message-v2 at this worktree)
+```bash
+codex exec resume 019f5355-4cc0-7213-ac0f-fa5b9767c940 -- "<follow-up>"
+```
+_Written by `.llm/tools/agentic/codex/launch-codex-slice.ts`._

.llm/runs/dashboard-design--orchestrator/briefs/codex-ux-brief.md

@@ -0,0 +1,70 @@
+use harness
+
+## SKILL
+
+netscript-harness · netscript-doctrine (read-only grounding). This is an ANALYSIS-ONLY slice:
+you change NO product code. All output goes under
+`/home/codex/repos/ns-ddr-codex/.llm/runs/dashboard-design--orchestrator/analysis/`.
+
+# Adversarial UX/DX pass — NetScript Dev Dashboard prototype (design revamp run)
+
+You are the adversarial UX/DX evaluator for the dev-dashboard design revamp
+(umbrella PR #685, branch `design/dev-dashboard-revamp`). Your SOLE role is user/developer
+experience. You must answer two questions, hard and honestly:
+
+1. **HOW FAR AHEAD OF THE COMPETITION IS THIS DASHBOARD?** Rate the current prototype
+   against the best-in-class dev consoles (Encore dev dashboard + Flow, Temporal Web UI,
+   Inngest, Trigger.dev, Appwrite console, Directus, Supabase Studio, Convex dashboard,
+   the new Aspire "Deck" React dashboard — see
+   `.llm/runs/dashboard-design--orchestrator/reference/aspire-deck-research.md`).
+   Per screen and overall: where does it lead, where does it merely match, where does it
+   trail? Be adversarial: a screen that "holds gates" can still be mediocre UX.
+2. **HOW MUCH OF NETSCRIPT'S HIGHLIGHT FEATURES DOES IT COVER?** Inventory NetScript's
+   differentiating capabilities (from the repo: packages/, plugins/, docs/architecture/
+   doctrine, and `.llm/runs/dashboard-design--orchestrator/design-project/feedback/POC-ground-truth.md`)
+   and score prototype coverage. Name every highlight feature that is invisible or
+   under-sold in the prototype (e.g. polyglot task runtimes, contract duality, typed route
+   contracts, runtime-config versioning, durable streams, auth projections, AI tool
+   registry, plugin contribution axes).
+
+## Your inputs (all committed on your branch `design/ddr-s3-codex-ux`)
+
+- `.llm/runs/dashboard-design--orchestrator/screen-catalog.md` — ground-truth catalog.
+- `.llm/runs/dashboard-design--orchestrator/screenshots/*.png` — 17 full-page captures.
+- `.llm/runs/dashboard-design--orchestrator/improvement-brief.md` — the six owner axes;
+  your findings must be organized against them.
+- `.llm/runs/dashboard-design--orchestrator/design-project/` — prior adversarial review
+  (13 screens), POC ground truth, v2 design prompts, rescope research.
+- `.llm/runs/dashboard-design--orchestrator/prototype/` — the prototype source; you can
+  re-render it: serve that dir over HTTP, open `NetScript Dev Dashboard.dc.html#/<route>`
+  (route list in the catalog), wait ~2.5 s (loads React from unpkg).
+- Reference apps (read-only, ALREADY CLONED):
+  - `/home/codex/repos/refs/netscript-start` — playground dashboard: study
+    `apps/playground/routes/(dashboard)/dashboard/{plugin,framework}` routing + UX.
+  - `/home/codex/repos/refs/eis-chat` — production-grade chat app: frontend routing and
+    AI-surface patterns.
+
+## Seed findings from the owner (treat as confirmed, expand on them)
+
+(a) The playground's jobs / polyglot tasks / sagas / streams routing experience is MUCH
+better than the prototype's flat 15-route hash router — identify exactly which routing/
+navigation/list-detail patterns should be stolen and adapted.
+(b) The prototype's AI surface (`ai` route) is underwhelming vs the reference chat app.
+The owner does NOT want a generic chat: they want AI capability distributed as actions,
+dynamic triggers, context augmentation, and embedded assists across panels. Assess the gap
+concretely against what the chat app ships.
+
+## Deliverables (commit to your branch, run dir only)
+
+1. `analysis/codex-ux-dx-verdict.md` — the two headline answers with a scored
+   competitive matrix (screen × competitor), the highlight-feature coverage table
+   (feature → covered/undersold/absent → evidence), and a ranked list of the 15 highest-
+   leverage UX/DX changes for the revamp (each: what, why, which axis, which screen(s)).
+2. `analysis/codex-routing-steal-list.md` — concrete routing/UX patterns from the two
+   reference apps worth stealing (cite file paths), adapted to the dashboard's screens.
+
+Rules: analysis only — no changes outside `.llm/runs/dashboard-design--orchestrator/`;
+never mention the reference apps' names as required public naming (internal grounding is
+fine in these analysis docs); do not push to main; commit on `design/ddr-s3-codex-ux` and
+push that branch when done. Honest artifacts; never fabricate evidence; cite file paths
+for every claim about the repos.

.llm/runs/dashboard-design--orchestrator/briefs/glm-design-brief.md

@@ -0,0 +1,51 @@
+use harness
+
+## SKILL
+
+Analysis-only design slice: you change NO product code. All output goes under
+`.llm/runs/dashboard-design--orchestrator/analysis/` in this worktree
+(`/home/codex/repos/ns-ddr-glm`, branch `design/ddr-s4-glm-design`).
+
+# Design/UX critique pass — NetScript Dev Dashboard prototype
+
+You are a principal product designer and DX specialist reviewing a developer-dashboard
+prototype for the NetScript framework (a Deno/Fresh full-stack framework whose dev
+dashboard is a DX satellite orbiting the .NET Aspire dashboard and Scalar API docs).
+
+## Inputs (read these files first)
+
+- `.llm/runs/dashboard-design--orchestrator/screen-catalog.md` — ground-truth catalog of
+  all 15 screens (routes, purpose, states).
+- `.llm/runs/dashboard-design--orchestrator/improvement-brief.md` — the owner's six
+  improvement axes (binding).
+- `.llm/runs/dashboard-design--orchestrator/design-project/feedback/README.md` and
+  `POC-ground-truth.md` — prior adversarial review verdict + the real data model.
+- `.llm/runs/dashboard-design--orchestrator/prototype/assets/proto.css` and `ns-ext.css`
+  — the actual component CSS (tokens, density, motion) if you want visual-language
+  evidence.
+- Screenshots exist as PNGs under `.llm/runs/dashboard-design--orchestrator/screenshots/`
+  (you may not be able to view images; the catalog describes each precisely).
+
+## Deliverable
+
+Write `.llm/runs/dashboard-design--orchestrator/analysis/glm-design-pass.md` in markdown:
+
+1. **Overall design verdict** — visual language, density, hierarchy, typography,
+   color/tone usage, dark mode; what works, what is weak; scored /10 with justification.
+2. **Per-screen critique** for every screen in the catalog (layout, information
+   hierarchy, affordance clarity, state design), each with 3–5 concrete, actionable
+   redesign proposals.
+3. **Axis-by-axis proposals** for the six owner axes (especially routing hierarchy /
+   information architecture, write-action flows, distributed AI surface,
+   plugin/extension visibility) — be specific: name components, layouts, interaction
+   patterns.
+4. **Ten "wow" ideas** that would put this dashboard visibly ahead of best-in-class dev
+   consoles (Temporal, Inngest, Encore, Appwrite, Supabase Studio), each with a
+   one-paragraph design spec.
+
+Ground every point in the catalog and CSS evidence; do not invent screens that are not
+documented.
+
+When the document is written: `git add` it, commit on branch `design/ddr-s4-glm-design`
+with message "docs(run): GLM 5.2 design/UX critique pass", and push the branch to origin.
+Touch nothing else. Your final chat message: 5-line summary of your strongest proposals.