@genui/a3-create 0.1.36

package/template/docs/designs/a3-gtm-strategy.md

@@ -0,0 +1,280 @@
+# Design: A3 GTM Strategy — From Framework to Repeatable Motion
+
+Generated by /office-hours on 2026-03-19
+Branch: main
+Repo: generalui/a3
+Status: APPROVED
+Mode: Intrapreneurship
+Promoted: 2026-03-19
+See also: docs/designs/a3-platform-vision.md (CEO scope expansion review)
+
+---
+
+## Product Vision
+
+A3's core thesis: **the best agentic applications are specifically defined, trustworthy, governable,
+and reliable** — not open-ended do-everything assistants that may or may not do the right thing.
+
+The platonic ideal is the optimal combination of:
+
+- **Deterministic code** for structured, critical, compliance-sensitive actions (forms, payments,
+  routing decisions, data writes) — places where LLM unpredictability is unacceptable
+- **Agentic flexibility** for natural language understanding, intent classification, fluid conversation,
+  and handling unpredictable situations within a bounded scope
+
+Most agentic frameworks default to "let the LLM decide everything." A3's thesis is that this is
+wrong for production systems. In healthcare, commerce, finance, and any regulated or high-stakes
+context: **the LLM should handle what humans are naturally good at (conversation, nuance, ambiguity)
+and structured code should handle everything else.**
+
+This is not a limitation of A3 — it IS A3. The Care Agent hybrid (deterministic UI components for
+intake/payments/forms, conversational agents for triage and routing) is the pattern, not an edge case.
+
+---
+
+## Problem Statement
+
+A3 is a production-grade, open-source TypeScript framework for multi-agent AI orchestration.
+The framework is built and proven in production (Experity: 2,000+ clinics, 3.5M sessions, 68% discharge automation).
+The challenge is NOT product/market fit — it's turning one enterprise win into a repeatable GTM motion,
+and anchoring that motion to the core product thesis (governed, trustworthy, deterministic+agentic hybrid)
+rather than generic "agentic AI" positioning.
+
+A3 serves two audiences simultaneously:
+
+1. **Developers** — as an open-source package to adopt and build with
+2. **Enterprise buyers** — as the engine GenUI uses to automate their manual workflows
+
+The current positioning (landing copy, comparison tables, framework-vs-framework framing) is optimized
+for audience #1. But the warm pipeline and the most compelling demand signal are coming from audience #2 —
+and in both cases, the core differentiator is the **governed hybrid architecture**, not the feature list.
+
+---
+
+## Demand Evidence
+
+- **Experity/Care Agent**: 2,000+ urgent care clinics, ~1M patients, 3.5M+ sessions, 68% of patient
+  discharge interactions automated. Jonathan Moss (EVP, Experity): "Staff is doing the work that
+  improves experience and throughput rather than repeating administrative tasks."
+- **Warm pipeline**: One specific prospect — real person, identified role, real professional stakes
+  tied to solving this problem. Has not yet signed, but the conversation is active.
+- **OSS signals**: Community / developer signals exist but are not the primary growth lever right now.
+
+The most important demand signal in this session: **the competition is human labor, not LangChain.**
+These prospects don't have an agentic workflow at all — they have staff doing repetitive,
+multi-step tasks. A3 replaces the task, not the framework.
+
+---
+
+## Status Quo
+
+The warm pipeline prospect's current workflow is staffed — humans handling a repetitive, high-volume,
+multi-step process with no agentic version in existence. "Inefficient and painful" was the description.
+
+This is the pattern: the ROI story for A3 in enterprise is not "switch frameworks" —
+it's "automate what your staff is doing manually right now."
+
+Real cost of the status quo:
+
+- Staff time spent on workflows that don't require human judgment
+- Error rates from manual, repetitive processes
+- Inability to scale without proportionally growing headcount
+
+---
+
+## Target User & Narrowest Wedge
+
+**The developer audience:**
+
+- TypeScript engineers building multi-agent AI products
+- Attracted via OSS: GitHub, docs, DX quality, community
+- Acquisition channel: organic / developer word-of-mouth
+
+**The enterprise buyer (primary GTM focus right now):**
+
+- Ops executives, VPs of Growth/Operations, CTOs at companies with high-volume, multi-step
+  manual workflows in regulated or high-stakes environments
+- Attracted via: Experity case study, direct outreach, partner referrals
+- What gets them to sign: a bounded pilot that automates one specific workflow end-to-end
+  with measurable before/after (time saved, FTE cost, throughput)
+
+**Narrowest wedge:** One workflow, fully automated, measurable ROI.
+Not the platform. Not the full skills architecture.
+The single workflow that hurts the most and maps cleanest to what A3 can do.
+
+---
+
+## Constraints
+
+- GenUI is a development consultancy + framework publisher — not a pure-play sales org
+- A3 must serve both developers (OSS) and enterprises (GenUI clients) — dual-track messaging required
+- No dedicated sales team; GTM runs through the same people who build
+- The CoBuilder Partner Program exists but needs a proven sales motion before it can create leverage
+- The observation gap: GenUI has not yet watched the warm pipeline prospect's manual workflow in detail
+
+---
+
+## Premises
+
+1. **The real competition is human labor, not other frameworks** — for enterprise buyers,
+   A3 is competing against staffed workflows, not LangChain. The comparison table on the
+   landing page is aimed at the wrong buyer. *(agreed)*
+
+2. **The economic buyer is not the developer** — developers are influencers and implementers,
+   but the person who signs is an ops exec measuring FTE cost and automation rate.
+   Two different conversations, two different messages. *(agreed)*
+
+3. **The CoBuilder Partner Program is premature (for now)** — partners need a repeatable sales
+   motion to execute. Until GenUI has proven the second enterprise win, the partner program
+   creates overhead without leverage. *(agreed)*
+
+4. **Healthcare is a wedge, not a destination** — Experity is in healthcare, but "regulated
+   workflows" as the primary positioning limits the TAM unnecessarily. The pattern
+   (manual, high-volume, multi-step workflows) exists in commerce, ops, HR tech,
+   logistics, insurance. *(agreed)*
+
+---
+
+## Approaches Considered
+
+### Approach A: Land the Second Client First
+
+Focus 100% on converting the warm pipeline to a paid pilot.
+Nothing else until there is a second reference customer.
+
+- **Effort:** S
+- **Risk:** Low
+- **Pros:** Real person, real stakes, real problem. Forces the "watch the workflow" step.
+  Second client makes the partner program real.
+- **Cons:** One-at-a-time; doesn't scale. OSS adoption continues to stagnate in parallel.
+
+### Approach B: Two Tracks in Parallel
+
+Run OSS developer community growth (docs, DX, GitHub) alongside enterprise outreach
+(vertical playbooks, warm pipeline). Separate messaging for each audience.
+
+- **Effort:** L
+- **Risk:** Med (split focus on a lean team)
+- **Pros:** Covers both audiences. Community builds long-term moat; enterprise funds now.
+- **Cons:** Two buyers, two messages, one team. Easy to be mediocre at both.
+  OSS is a 12-18 month compounding game.
+
+### Approach C: Vertical Playbook — One More, Done Right
+
+Pick one adjacent vertical. Build a specific playbook: 5 manual workflows that map to A3,
+ROI calculator, reference architecture, pitch narrative. Use it to close the warm pipeline
+AND recruit one anchor partner.
+
+- **Effort:** M
+- **Risk:** Med
+- **Pros:** Builds repeatable sales infrastructure. Gives the CoBuilder Program something
+  concrete to sell.
+- **Cons:** Requires real vertical expertise. Wrong vertical choice = wasted effort.
+
+---
+
+## Recommended Approach: A then C
+
+**Sequence matters.** Close the warm pipeline first (Approach A), then build the vertical
+playbook (Approach C). The second enterprise client teaches you what the playbook should say.
+Building the playbook before you've lived through a second deal is writing fiction.
+
+**How the sequence plays out:**
+
+1. **Now:** Watch the workflow. Sit with the warm pipeline prospect's team and observe the
+   manual process they want to automate — without helping, without guiding. What happens?
+   What's surprising? That observation shapes the pilot scope.
+
+2. **Next 30 days:** Scope and propose a bounded pilot. One workflow, fully automated,
+   measurable before/after. Get it signed.
+
+3. **During the pilot:** Document everything — the workflow map, the handoff points,
+   the schema design, the session state choices. This becomes Playbook Template v1.
+
+4. **Post-pilot:** Turn the second case study + the documentation into the vertical playbook.
+   NOW activate the CoBuilder Partner Program with that playbook as the sales tool.
+
+**For the developer/OSS track (parallel, but lower priority):**
+
+- Improve docs and quickstart experience (already in progress per git log)
+- The Experity case study is the most powerful dev community signal — make it findable
+- Don't split messaging: "A3 is what GenUI uses to build Care Agent-scale systems" is a
+  better dev pitch than "A3 vs LangGraph feature comparison"
+
+---
+
+## Open Questions
+
+1. **What vertical is the warm pipeline in?** If it's healthcare again, the second client
+   confirms a vertical thesis. If it's different, it expands the positioning story.
+
+2. **What does the pilot pricing look like?** Is this a consulting engagement where GenUI
+   builds the automation, or a framework licensing play where the client's team builds with
+   A3 + GenUI support? The answer changes the GTM motion significantly.
+
+3. **Who inside GenUI owns the "watch the workflow" step?** This is not a sales task —
+   it's a product/design research task. Someone needs to go observe before the pilot scope
+   is written.
+
+4. **Is the developer OSS track meant to generate enterprise leads, or is it a separate
+   community-building goal?** If devs are meant to become enterprise champions (bottom-up),
+   the DX and docs investment is justified. If the two tracks are genuinely separate, they
+   need separate owners.
+
+---
+
+## Success Criteria
+
+- **30 days:** Warm pipeline pilot scoped, proposal sent, response received
+- **60 days:** Second enterprise client signed (or clear reason it didn't close)
+- **90 days:** Second case study drafted + one vertical playbook document complete
+- **6 months:** CoBuilder Partner Program has one active partner using the playbook
+- **OSS:** A3 GitHub repo has a clear "here's what this looks like in production" section
+  linking to Experity + second case study
+
+---
+
+## Dependencies
+
+- Warm pipeline: the specific person needs to agree to a workflow observation session
+- GenUI leadership: alignment that "land the second client" is the priority before partner
+  program scaling
+- Experity: permission to use the case study in new vertical outreach (already in A3_overview.md)
+
+---
+
+## The Assignment
+
+**This week:** Ask the warm pipeline contact for a 90-minute session where you can watch
+their team do the workflow you're proposing to automate. Don't demo A3. Don't pitch.
+Just watch. Take notes. What are the steps? Where does information move between people?
+Where does something get looked up? Where does judgment get applied vs. where is it rote?
+
+That observation session is the foundation of a pilot scope that actually fits the problem.
+Without it, you're scoping from assumptions.
+
+---
+
+## What I noticed about how you think
+
+- You said the competition is "humans — the agentic version of the workflow doesn't exist."
+  That's the clearest GTM insight in the whole session, and you said it like it was obvious.
+  It's not obvious. Most framework builders never get there because they're talking to
+  developers, not operators.
+
+- When I challenged the premises, you agreed with all four AND corrected me with new
+  information ("dual-purpose OSS + enterprise"). That's not compliance — that's a
+  well-formed mental model of a hard trade-off. You know what you're holding.
+
+- You chose "A then C" immediately. No hedging, no "it depends." That's the right call
+  and it was fast. People who've thought about sequencing are faster at this question
+  than people who haven't.
+
+- "One specific workflow automated end-to-end" — you didn't say "a demo" or "a proof of
+  concept." You said the thing you'd actually ship. That matters.
+
+- When you pushed back on the design doc, you didn't say "I want to change the strategy."
+  You said "we're missing the core thesis." That's product thinking, not marketing thinking.
+  The governed hybrid architecture — deterministic where it matters, agentic where it helps —
+  is a genuine product insight that most AI framework builders don't have because they're
+  optimizing for demos, not production systems. You named it precisely.

package/template/docs/designs/a3-platform-vision.md

@@ -0,0 +1,276 @@
+# A3 Platform Vision — From Framework to Governed Agentic Standard
+
+Generated by /plan-ceo-review on 2026-03-19
+Branch: main | Mode: Scope Expansion
+Status: ACTIVE
+
+---
+
+## Vision
+
+### The 10x Version
+
+A3 becomes what WordPress became for websites — the default stack for deploying production
+agentic workflows.
+Every serious enterprise implementing agentic AI in healthcare, regulated industries, or
+high-volume commerce uses A3.
+When a developer at a 500-person company needs to automate a patient discharge workflow,
+they find A3, run `npx create-a3-app --template=patient-discharge`, and have a running
+prototype in 45 minutes.
+
+GenUI operates at three tiers:
+
+1. **Framework publisher** — open source, community, standards
+2. **Premium agent publisher** — `@genui/agent-patient-intake`, `@genui/agent-retail-returns`, etc.
+3. **Enterprise integrator** — for organizations that want help through complex implementations
+4. **Certification authority** — "A3 Certified" developer credential for agencies and partners
+
+Revenue compounds: self-serve (product) → partner-built (CoBuilder) → GenUI-built (enterprise).
+
+### Platonic Ideal
+
+**Sarah, VP of Patient Experience at a 200-clinic urgent care chain:**
+Her CTO Googles "A3 agentic framework," finds a site that says "Build what Experity built.
+Here's how."
+He runs `npx create-a3-app --template=patient-discharge`.
+Working prototype in one hour, deployed to their AWS account, demoing to front-desk staff
+two weeks later.
+The feeling: "I didn't need to wait 18 months. I built it in a weekend."
+
+**Marcus, developer at a digital agency with healthcare clients:**
+He takes the A3 certification course (4 hours), becomes "A3 Certified."
+He finds two pre-built `@genui/agent-*` packages that cover 80% of what his clients need.
+GenUI is on the other end of a Slack channel for the remaining 20%.
+The feeling: "I can offer this to my clients, and I'm not alone when it gets hard."
+
+---
+
+## Core Product Thesis
+
+> A3's thesis: the best agentic applications are specifically defined, trustworthy,
+> governable, and reliable — not open-ended do-everything assistants.
+
+The platonic ideal combines:
+
+- **Deterministic code** for structured, critical, compliance-sensitive actions (forms, payments,
+  routing decisions, data writes) — where LLM unpredictability is unacceptable
+- **Agentic flexibility** for natural language understanding, intent classification, fluid
+  conversation, and handling unpredictable situations within a bounded scope
+
+This is A3's permanent differentiator vs. LangChain, LangGraph, and general-purpose AI frameworks.
+They optimize for capability.
+A3 optimizes for governability.
+
+---
+
+## Accepted Scope — This Initiative
+
+### 1. Use-Case Templates
+
+Extend the `create` CLI to support `--template` flags:
+
+```bash
+npx create-a3-app --template=patient-discharge
+npx create-a3-app --template=retail-returns
+npx create-a3-app --template=staff-onboarding
+```
+
+Each template ships pre-wired agent configurations, Zod schemas matching that domain,
+API integration stubs, and `complianceSurface` defaults where applicable.
+Framing: **workflow-first, not industry-first** (like WordPress themes, not industry verticals).
+
+Effort: S (CC: ~4h)
+
+### 2. Analytics / Observability Layer
+
+A3 sessions already have state, schema-validated outputs, agent handoffs, and timestamps.
+Ship a structured analytics events interface with first-class adapters for OpenTelemetry,
+Datadog, and Grafana.
+For deployments without existing observability infra: include a reference dashboard in the
+example app.
+An ops exec should be able to see their own "68% automated" number from their own deployment.
+
+Effort: S–M (CC: ~3h)
+
+### 3. `complianceSurface` Configuration
+
+A configuration flag that activates a bundle of production defaults appropriate for a
+given compliance surface:
+
+```typescript
+createChatSession({
+  complianceSurface: 'hipaa', // or 'soc2', 'gdpr'
+  provider,
+  store,
+})
+```
+
+Activates: PII detection and redaction, audit-trail logging with tamper-evident timestamps,
+configurable data retention TTLs, encrypted session state.
+
+**Critical framing:** "This configures A3's defaults for the HIPAA compliance surface.
+It does not certify your application as HIPAA compliant."
+Requires legal review of the API and documentation before shipping.
+
+Effort: M (CC: ~6h)
+
+### 4. A3 Advisor Agents (First `@genui/agent-*` Packages)
+
+A3 ships with a suite of purpose-built advisor agents — published as installable npm packages.
+These become the first `@genui/agent-*` reference implementations.
+A3 uses A3 to teach A3.
+
+- **`@genui/agent-a3-guide`** — conversational A3 tutorial with codebase context. Ask it
+  "how do I set up routing?" and it walks you through it with examples from your project.
+- **`@genui/agent-a3-deployment`** — helps you choose provider, infrastructure config, and
+  compliance surface for your use case.
+- **`@genui/agent-a3-compliance`** — explains what each `complianceSurface` setting does,
+  what it doesn't cover, and what additional organizational steps are required.
+- **`@genui/agent-a3-audit`** — reads your `a3 audit export` output and explains it in plain
+  language for a non-technical compliance officer or auditor.
+
+Each use-case template (above) ships with `@genui/agent-a3-guide` pre-wired.
+Each agent demonstrates the governed hybrid thesis: specific job, bounded scope, validated output.
+
+Effort: S (CC: ~4h)
+
+### 5. `a3 doctor` — Configuration Linter
+
+A CLI command that analyzes your agent configuration before it hits production:
+
+```bash
+a3 doctor
+# ✓ No circular routing detected
+# ✓ All agents have terminal paths
+# ⚠ agent 'intake': outputSchema has 12 required fields — consider reducing
+# ✗ complianceSurface: 'hipaa' set but phone number field has no PII redaction
+```
+
+Flags: circular routing risks, missing fallback/terminal agents, schema fields likely to
+confuse the LLM, and `complianceSurface`-specific warnings.
+Runs in CI; catches architecture problems at build time.
+
+Effort: S (CC: ~2h)
+
+### 6. `a3 audit export` — Compliance Artifact Generator
+
+A CLI command that generates a structured, timestamped audit report for a session:
+
+```bash
+a3 audit export --session=SESSION_ID --format=hipaa
+# → ./audit-2026-03-19-SESSION_ID.pdf
+# → ./audit-2026-03-19-SESSION_ID.json
+```
+
+The PDF is **human-readable and user-friendly** — designed to be handed to an auditor or
+compliance officer without embarrassment.
+Layout: summary page first (session overview, outcome, violation count), then per-turn
+detail (agent name, action, data extracted with PII masked, routing decision).
+Signed with a tamper-evident hash.
+
+Effort: S (CC: ~2h)
+
+### 7. `a3` CLI Consolidation
+
+The `create` package evolves into the unified `a3` CLI:
+
+```text
+@genui/a3-cli (was: @genui/a3-create)
+├── a3 create --template=<use-case>    # scaffold a new project
+├── a3 doctor                           # analyze agent config, flag issues
+└── a3 audit export --session=<id>     # generate compliance artifact
+```
+
+**One-way door warning:** once the CLI API is published, command signatures create
+user expectations.
+Design the command surface carefully before 1.0.
+Decide now: does `a3` CLI version independently of `@genui/a3` core?
+
+---
+
+## Architecture Diagram
+
+```text
+@genui/a3-cli
+├── a3 create --template=<name>
+├── a3 doctor
+└── a3 audit export
+
+@genui/a3 (core, extended)
+├── complianceSurface config
+├── Analytics events interface
+│   ├── OTel adapter
+│   ├── Datadog adapter
+│   └── Grafana adapter
+└── Audit trail emission
+
+@genui/agent-* (Advisor Agents + future domain agents)
+├── @genui/agent-a3-guide
+├── @genui/agent-a3-deployment
+├── @genui/agent-a3-compliance
+└── @genui/agent-a3-audit
+
+create-a3-app templates/
+├── templates/default/        (existing)
+├── templates/patient-discharge/
+├── templates/retail-returns/
+└── templates/staff-onboarding/
+
+example/ (Next.js app — ships reference dashboard)
+```
+
+---
+
+## Platform Flywheel (18-Month Vision)
+
+```text
+Templates → Developers adopt A3 → Need more agents → @genui/agent-* fills gaps
+     ↑                                                          |
+     └──── Community publishes their own agents ←──────────────┘
+                    (agent package registry / marketplace)
+```
+
+This initiative establishes the foundation.
+The marketplace comes when the community exists.
+
+---
+
+## NOT in Scope (Deferred)
+
+- Full agent package registry / marketplace — 18-month horizon; needs community first
+- GenUI Managed A3 / hosted SaaS offering — requires significant infrastructure
+- A3 certification program — needs a developer community to certify
+- Agent routing live visualizer — Phase 2 delight item
+- `A3 Compatibility Badge` npm interface standard — needs ecosystem participants
+
+---
+
+## Strategic Premises
+
+From the office-hours GTM session (see `docs/designs/a3-gtm-strategy.md`):
+
+- **Competition is human labor, not other frameworks** — enterprise buyers aren't choosing
+  between A3 and LangChain; they're choosing between automated workflows and staffed workflows
+- **Economic buyer is an ops exec, not a developer** — developers influence; ops execs sign
+- **`complianceSurface` (not "compliance mode")** — signals tuning, not certification; legally safer
+- **A3 is taking over the GenUI identity** — the GenUI site is being rebuilt around A3
+  as a first-class citizen
+- **Use-case templates (not verticals)** — workflow-first framing matches the WordPress mental model
+
+---
+
+## Open Questions (Resolve Before Implementation)
+
+1. **`a3 doctor` analysis mode:** Static analysis (parse TypeScript source files) vs. runtime
+   introspection (spin up a test A3 instance)? Static works in CI without side effects;
+   runtime is more accurate.
+2. **`@genui/agent-*` pricing:** Free OSS, freemium, or premium for domain-specific agents?
+   Affects positioning and the CoBuilder partner value prop.
+3. **`complianceSurface` legal review:** Get legal/compliance sign-off on the API design and
+   documentation framing before shipping.
+4. **PDF audit export persona:** Primary reader is compliance officer or auditor — not a
+   developer. Design review required for the PDF layout before implementation.
+5. **Template selection:** Which 2-3 templates ship first? Should match the warm pipeline
+   vertical — don't build `patient-discharge` if the second client is in commerce.
+6. **`a3` CLI versioning:** Does the CLI version independently of `@genui/a3` core?
+   What's the compatibility matrix?

package/template/next-env.d.ts

@@ -0,0 +1,6 @@
+/// <reference types="next" />
+/// <reference types="next/image-types/global" />
+import "./.next/types/routes.d.ts";
+
+// NOTE: This file should not be edited
+// see https://nextjs.org/docs/app/api-reference/config/typescript for more information.

package/template/next.config.mjs

@@ -0,0 +1,15 @@
+import fs from 'node:fs'
+import path from 'node:path'
+
+// Check if we are running in the GenUI A3 monorepo
+const isMonorepo = fs.existsSync(path.join(import.meta.dirname, '..', 'a3.code-workspace'))
+
+/** @type {import('next').NextConfig} */
+const nextConfig = {
+  reactStrictMode: true,
+  turbopack: {
+    root: isMonorepo ? path.join(import.meta.dirname, '..') : import.meta.dirname,
+  },
+}
+
+export default nextConfig

package/template/package.json

@@ -0,0 +1,41 @@
+{
+  "name": "@genui/example",
+  "version": "0.1.0",
+  "scripts": {
+    "build": "next build",
+    "dev": "node scripts/dev.mjs",
+    "start": "next start"
+  },
+  "dependencies": {
+    "@ag-ui/client": "0.0.45",
+    "@ag-ui/encoder": "0.0.45",
+    "@emotion/react": "11.14.0",
+    "@emotion/styled": "11.14.1",
+    "@genui/a3": "latest",
+    "@genui/a3-anthropic": "latest",
+    "@genui/a3-bedrock": "latest",
+    "@genui/a3-openai": "latest",
+    "@mui/icons-material": "7.3.7",
+    "@mui/material": "7.3.7",
+    "next": "16.1.7",
+    "react": "19.2.4",
+    "react-dom": "19.2.4",
+    "react-markdown": "10.1.0",
+    "react-syntax-highlighter": "15.6.1",
+    "remark-gfm": "4.0.1",
+    "@dagrejs/dagre": "2.0.4",
+    "@xyflow/react": "12.10.1",
+    "styled-components": "6.3.8",
+    "zod": "4.3.6"
+  },
+  "devDependencies": {
+    "@types/node": "20.19.0",
+    "@types/react": "19.2.10",
+    "@types/react-dom": "19.2.3",
+    "@types/react-syntax-highlighter": "15.5.13",
+    "typescript": "5.9.3"
+  },
+  "overrides": {
+    "fast-xml-parser": "5.5.6"
+  }
+}

package/template/public/site.webmanifest

@@ -0,0 +1,11 @@
+{
+  "name": "A3 Core Example",
+  "short_name": "A3 Example",
+  "icons": [
+    { "src": "/android-chrome-192x192.png", "sizes": "192x192", "type": "image/png" },
+    { "src": "/android-chrome-512x512.png", "sizes": "512x512", "type": "image/png" }
+  ],
+  "theme_color": "#ffffff",
+  "background_color": "#ffffff",
+  "display": "standalone"
+}

package/template/scripts/dev.mjs

@@ -0,0 +1,29 @@
+import { exec, spawn } from 'node:child_process'
+
+function openBrowser(url) {
+  const quoted = encodeURI(url)
+  const cmd =
+    process.platform === 'win32'
+      ? `cmd.exe /c start ${quoted}`
+      : process.platform === 'darwin'
+        ? `open '${quoted}'`
+        : `xdg-open '${quoted}'`
+  exec(cmd)
+}
+
+const child = spawn('next', ['dev'], {
+  stdio: ['inherit', 'pipe', 'inherit'],
+  shell: process.platform === 'win32', // required for Windows PATH resolution
+})
+
+function onData(data) {
+  process.stdout.write(data)
+  const match = data.toString().match(/Local:\s+(https?:\/\/localhost:\d+)/)
+  if (match) {
+    child.stdout.off('data', onData)
+    openBrowser(match[1])
+  }
+}
+
+child.stdout.on('data', onData)
+child.on('exit', (code) => process.exit(code ?? 0))