codehava-agent-kit 2.0.0 → 3.0.0

package/README.md

@@ -1,49 +1,98 @@
-# Codehava Antigravity Agent Kit 🚀
+# Codehava Agent Kit V3.0 🚀
 
-The ultimate **Vibe Coding** Agent Kit boilerplate designed to turn plain-text ideas into production-ready Web and Mobile applications seamlessly. Codehava Agent Kit is specifically curated to inject superhuman coding capabilities into Cursor, Antigravity, or Claude Code.
+The ultimate **Vibe Coding** Agent Kit — transforms plain-text ideas into production-ready Web and Mobile applications. Injects superhuman coding capabilities into Cursor, Antigravity, or Claude Code via the **SEED + PAUL** execution architecture.
 
-## 🌟 What's Inside?
+---
 
-We've supercharged this kit to be fully autonomous, meaning you **no longer need to memorize skill names or syntax**.
+## What's Inside?
 
-1. **Auto-Discovery Engine (450+ Elite Skills):**
-   This kit ships with ~455 hand-picked, elite-tier skills inside the `.agent/skills/` directory (Frontend, Backend, Security, React, Next.js, and Mobile). Depending on your natural language prompt, the AI will *automatically discover*, load, and route tasks to the best skill seamlessly. 
+### 🌱 SEED-Style Typed Incubation
+Five project types with different rigor levels (Application, API/Backend, Campaign, Utility, Workflow). No more one-size-fits-all PRD — each type gets type-aware discovery questions, constraints, and quality gates before any code is written.
 
-2. **Long-Term Memory Protocol (`MEMORY.md`):**
-   Vibe coding sessions can go on for hours. To prevent your AI from forgetting initial architectural decisions, the kit enforces an autonomous memory state. The AI will quietly maintain a `.agent/MEMORY.md` file to preserve colors, database schemas, and global context permanently.
+### 🔄 PAUL Execution Loop (Token-Efficient)
+**Plan → Apply → Unify** — single-session execution with mandatory reconciliation. No token-heavy parallel subagents (~70% quality). Each session closes with a SUMMARY.md capturing planned vs. actual outcomes. No orphaned work. No state drift.
 
-3. **Vibe Workflow (Skill Chaining):**
-   Say goodbye to rushy, hallucinated code. When you ask the agent to build a new feature, the Agent Kit forces a staged pipeline:
-   `vibe-prd` ➔ `vibe-techdesign` ➔ `coding specialist`.
+### 🤖 Auto-Discovery Engine (600+ Elite Skills)
+~600 hand-picked, elite-tier skills auto-routed based on your natural language request — no memorization required. Keyword-based search with concrete algorithm defined in SKILLS_INDEX.md.
 
-4. **Self-Healing Protocol:**
-   Encountered a syntax error? Let the AI handle it autonomously using `@systematic-debugging` and `@lint-and-validate` skills without asking you "how to fix this stack trace".
+### 🧠 Long-Term Memory Protocol
+`.agent/MEMORY.md` preserves colors, schemas, tech decisions, and constraints across sessions. `.agent/STATE.md` tracks current loop position. `docs/handoffs/` enables zero-context session resumption.
 
-5. **Built-in Vibe MCP Servers:**
-   Pre-configured `mcp_config.json` integrating:
-   - **Brave Search / Tavily:** Real-time documentation web scraping to fight AI hallucinations.
-   - **Sequential Thinking:** An Anthropic-approved system for logical AI braking.
-   - **Figma & Google Stitch:** Direct integration to extract design DNA straight into code.
+### ✅ 4-Level Verification (from GSD)
+Every task verified at 4 levels: **Exists → Substantive → Wired → Data Flows** — before any DONE status is reported. Escalation status (DONE / DONE_WITH_CONCERNS / NEEDS_CONTEXT / BLOCKED) replaces binary pass/fail.
 
-## 📦 Installation & Usage
+### 🔧 Self-Healing Protocol
+Autonomously fixes errors using `@systematic-debugging` and `@lint-and-validate`. Never asks users "how to fix?" — fixes silently, delivers working output.
 
-You can initialize this kit in any empty directory to begin your vibe coding journey:
+### 📊 Model Profiles
+**Quality** (Opus/Opus), **Balanced** (Opus/Sonnet, default), **Budget** (Sonnet/Haiku) — match compute to task complexity.
+
+### 🔌 Built-in Vibe MCP Servers
+Context7 (real-time docs), GitHub, Brave Search, Sequential Thinking, PostgreSQL, Playwright, Sentry, Docker — pre-configured in `mcp_config.json`.
+
+---
+
+## Installation & Usage
 
 ```bash
 npx codehava-agent-kit init
 ```
 
-*This command will install the `.agent/`, `docs/`, `specs/` directories and prompt a successful connection.*
+This creates `.agent/`, `docs/`, and `specs/` directories in your project root.
+
+**Then place:**
+- `.agent/` → your project root
+- `docs/` → your project root
+- `specs/` → your project root
+- `.antigravity/rules.md` → if using Antigravity IDE
+
+---
+
+## The PAUL Loop: How to "Vibe Code"
+
+```
+1. /vibe-plan     Start here for any new project or feature
+                  → Detects project type, asks the right questions
+                  → Outputs PLANNING.md with BDD acceptance criteria
+
+2. /apply         Execute the plan, task by task
+                  → Auto-discovers relevant skills from 600+ library
+                  → Qualify each task before moving on
 
-## ⚡ How to "Vibe Code"
+3. /unify         Close the loop (mandatory after every /apply)
+                  → Documents planned vs. actual outcomes
+                  → Updates STATE.md and MEMORY.md
 
-1. Open the initialized folder in your AI IDE (Antigravity, Cursor, etc.).
-2. Tell it what you want in human language:
-   > *"Tolong buatkan saya aplikasi POS kasir. Gunakan nextjs-app-router dan UI modern dengan dark mode."*
-3. Watch the Magic happen. The AI will Auto-Discover the best matching framework, record the session memory, and generate interactive, animated UI out of the box using `ui-ux-pro-max` guidelines!
+4. /vibe-recap    Context compaction when context gets long
+                  → Creates recap.md → open fresh chat to continue
+```
 
-## 🔧 Excluded Bloat
-We've pruned over 800+ niche/outdated community skills to ensure this NPM module is incredibly light and fast during the initialization process. Your IDE will load context in milliseconds.
+**Example:**
+> "Saya ingin buat aplikasi manajemen inventory untuk toko kecil"
+> → `/vibe-plan` detects Application type → asks 8 targeted questions → produces PLANNING.md
+> → `/apply` executes sprint 1 tasks → qualifies each with 4-level verification
+> → `/unify` documents what was done, deferred issues, tech decisions
+> → repeat for sprint 2
+
+---
+
+## Why SEED + PAUL (not GSD)?
+
+| Approach | Token Cost | Quality | Crash Recovery |
+|----------|-----------|---------|----------------|
+| Parallel Subagents (GSD) | Heavy | ~70% | Lock files needed |
+| Single Session (PAUL) | Efficient | ~95% | HANDOFF files |
+
+PAUL's single-session approach delivers higher quality at lower token cost. SEED's typed incubation prevents hallucinated code from vague specs.
+
+---
+
+## What Was Pruned
+
+Over 800+ niche/outdated community skills were pruned to keep the kit incredibly light and fast during initialization. Only elite, battle-tested skills remain.
+
+---
 
 ## License
-MIT License - Codehava
+
+MIT License — Codehava

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "codehava-agent-kit",
-  "version": "2.0.0",
-  "description": "Codehava Agent Kit - CLI to initialize agent templates, workflows, and skills",
+  "version": "3.0.0",
+  "description": "Codehava Agent Kit - CLI to initialize agent templates, workflows, and skills (SEED+PAUL Architecture)",
   "main": "index.js",
   "type": "module",
   "bin": {

package/templates/.agent/.shared/ACTIVE-template.md

@@ -0,0 +1,44 @@
+# .agent/ACTIVE.md — Project Pipeline Tracker
+
+> Copy file ini ke `.agent/ACTIVE.md` di root project kamu (atau workspace root).
+> AI memperbarui file ini saat /vibe-plan, /graduate, /launch, dan /unify dijalankan.
+> Dibaca oleh /pipeline untuk menampilkan project pipeline view.
+
+---
+
+## Ideation Pipeline
+
+| Project | Type | Status | Created | Notes |
+|---------|------|--------|---------|-------|
+| [nama-project] | [Application/API/Campaign/Utility/Workflow] | [in-progress/ready/graduated] | YYYY-MM-DD | [catatan] |
+
+**Status definitions:**
+- `in-progress` — PLANNING.md ada tapi belum complete
+- `ready` — PLANNING.md pass quality gate, siap di-graduate
+- `graduated` — Sudah dipindah ke apps/, git initialized
+
+---
+
+## Active Builds
+
+| Project | Type | Phase | Loop Status | Last Updated |
+|---------|------|-------|-------------|--------------|
+| [nama-project] | [Type] | Phase [N]/[Total] | [PLAN/APPLY/UNIFY] | YYYY-MM-DD |
+
+**Loop status:**
+- `PLAN` — Plan dibuat, menunggu /apply
+- `APPLY` — Sedang di-apply
+- `UNIFY` — Apply selesai, menunggu /unify
+- `CLOSED` — Loop closed, siap phase berikutnya
+
+---
+
+## Graduated (Completed)
+
+| Project | Type | Graduated | Location |
+|---------|------|-----------|----------|
+| [nama-project] | [Type] | YYYY-MM-DD | apps/[nama-project]/ |
+
+---
+
+> Terakhir diupdate: [YYYY-MM-DD] oleh [workflow yang mengupdate]

package/templates/.agent/.shared/HANDOFF-template.md

@@ -0,0 +1,75 @@
+# HANDOFF — [YYYY-MM-DD] Session [N]
+
+> File ini dibuat oleh AI saat session ditutup di tengah jalan (pause/interrupt).
+> Di session berikutnya, baca file ini PERTAMA sebelum melakukan apapun.
+> Path: `docs/handoffs/HANDOFF-[YYYY-MM-DD].md`
+
+---
+
+## Status Saat Ditutup
+
+- **Closed at:** [HH:MM YYYY-MM-DD]
+- **Reason:** [pause / interrupt / context limit / user request]
+- **Loop status:** [open — belum unify / closed — sudah unify]
+
+---
+
+## Yang Sedang Dikerjakan
+
+- **Active plan:** [path ke PLAN.md]
+- **Current task:** [T00X — nama task]
+- **Task status:** [sedang dikerjakan / baru dimulai / hampir selesai]
+
+**Progress:**
+```
+[T001] ✅ DONE
+[T002] ✅ DONE
+[T003] 🔄 IN PROGRESS — [apa yang sudah selesai, apa yang belum]
+[T004] ⏳ PENDING
+```
+
+---
+
+## Apa yang Sudah Dilakukan di Session Ini
+
+- [Perubahan 1] — [file yang diubah]
+- [Perubahan 2] — [file yang diubah]
+- [Keputusan yang dibuat] — [kenapa]
+
+---
+
+## Masalah yang Diketahui
+
+- [Issue 1] — [deskripsi, severity]
+- [Issue 2] — [deskripsi, severity]
+
+---
+
+## Langkah Selanjutnya (Exact Resume Point)
+
+1. Buka [path ke file yang sedang dikerjakan]
+2. Lanjutkan dari: [titik exact — baris/fungsi/komponen mana]
+3. Setelah itu: [task T00X berikutnya]
+4. Jangan lupa: [hal penting yang mudah terlupakan]
+
+---
+
+## Context Penting untuk Diingat
+
+- [Constraint atau keputusan penting yang relevan]
+- [Pattern atau convention yang sedang diikuti]
+- [Dependency yang belum selesai]
+
+---
+
+## Files yang Dimodifikasi Session Ini
+
+| File | Status | Notes |
+|------|--------|-------|
+| [path/file.ts] | Modified | [apa yang diubah] |
+| [path/file.tsx] | Modified | [apa yang diubah] |
+| [path/file.ts] | Created | [file baru] |
+
+---
+
+**Resume command:** Baca file ini lalu ketik "Resume" atau langsung lanjutkan task.

package/templates/.agent/.shared/MEMORY-schema.md

@@ -0,0 +1,117 @@
+# .agent/MEMORY.md — Architectural Memory
+
+> Copy file ini ke `.agent/MEMORY.md` di root project kamu.
+> AI memperbarui file ini di setiap UNIFY phase jika ada keputusan arsitektur baru.
+> Dibaca di AWAL setiap session sebelum menulis kode apapun.
+
+---
+
+## Tech Stack
+
+| Layer | Pilihan | Versi | Catatan |
+|-------|---------|-------|---------|
+| **Framework** | [Next.js / Vite / Flutter / dll] | [versi] | [catatan import path, dll] |
+| **Language** | [TypeScript / Python / Dart] | [versi] | [strict mode? settings?] |
+| **Database** | [PostgreSQL / MongoDB / Supabase] | [versi] | [via Prisma 7? Drizzle?] |
+| **Auth** | [Better Auth / Clerk / NextAuth] | [versi] | [provider apa?] |
+| **Styling** | [Tailwind + shadcn / dll] | [versi] | [custom config?] |
+| **State** | [Zustand / Jotai / Redux] | [versi] | [global atau local?] |
+| **Queue** | [BullMQ / Inngest / dll] | [versi] | [Redis URL?] |
+| **Deployment** | [Vercel / Railway / Docker] | - | [env staging/prod?] |
+
+---
+
+## Import Paths (Kritis — Jangan Salah)
+
+```typescript
+// Prisma
+import { db } from '@/lib/db'
+import type { User } from '@/generated/prisma'  // BUKAN @prisma/client
+
+// Auth
+import { auth } from '@/lib/auth'
+
+// [Tambahkan custom path lain di sini]
+```
+
+---
+
+## Design Tokens
+
+| Token | Value | Notes |
+|-------|-------|-------|
+| **Primary color** | [#hex] | [nama, konteks penggunaan] |
+| **Secondary color** | [#hex] | [nama, konteks penggunaan] |
+| **Background** | [#hex] | [dark/light default] |
+| **Font heading** | [font-family] | [Google Fonts? CDN?] |
+| **Font body** | [font-family] | [size default?] |
+| **Border radius** | [sm/md/lg / px value] | - |
+| **Dark mode** | [yes/no] | [default dark atau light?] |
+
+**Warna yang DILARANG:** [daftar warna yang tidak boleh dipakai — misal: ungu/violet per Purple Ban]
+
+---
+
+## Database Schema Snapshot
+
+> Update setiap kali ada migrasi significant.
+
+```prisma
+// Tabel utama (snapshot — bukan full schema)
+model User {
+  id        String   @id @default(cuid())
+  email     String   @unique
+  // ...
+}
+
+// [Tambahkan model penting lainnya]
+```
+
+---
+
+## API Conventions
+
+```typescript
+// Response format standar
+{ success: true, data: {...} }
+{ success: false, error: "pesan error" }
+
+// Auth pattern
+// Headers yang dipakai: [Bearer? Cookie?]
+
+// Rate limiting: [ya/tidak, provider apa]
+```
+
+---
+
+## Constraints (WAJIB DIIKUTI)
+
+- **UU PDP:** Jangan log NIK, nomor rekening, password, token
+- **File upload:** Hanya jpg/png/webp, maksimal 5MB
+- **Pagination:** Semua findMany harus ada `take` + `skip`
+- **Soft delete:** Gunakan `deletedAt DateTime?` — jangan hard delete data penting
+- **Secrets:** Selalu di `.env` — tidak pernah di kode
+- [Tambahkan constraint spesifik project di sini]
+
+---
+
+## Anti-patterns (DILARANG)
+
+- ❌ Import dari `@prisma/client` langsung (gunakan `@/generated/prisma`)
+- ❌ String literal untuk queue names (gunakan konstanta QUEUES)
+- ❌ `any` di TypeScript
+- ❌ Query tanpa pagination
+- ❌ Hard delete data user
+- [Tambahkan anti-pattern spesifik project di sini]
+
+---
+
+## Keputusan Arsitektur Terdokumentasi
+
+| Date | Keputusan | Alasan | Berlaku Sampai |
+|------|-----------|--------|----------------|
+| YYYY-MM-DD | [keputusan] | [alasan] | [permanent / sprint N] |
+
+---
+
+> Terakhir diupdate: [YYYY-MM-DD] oleh AI di Phase [N] UNIFY

package/templates/.agent/.shared/STATE-template.md

@@ -0,0 +1,59 @@
+# .agent/STATE.md — Session State
+
+> Copy file ini ke `.agent/STATE.md` di root project kamu.
+> AI akan membaca dan mengupdate file ini di setiap UNIFY phase.
+
+---
+
+## Current Position
+
+- **Phase:** [planning / apply / unify / complete]
+- **Active plan:** [path ke PLAN.md yang sedang dikerjakan]
+- **Last task completed:** [T00X — nama task]
+- **Next task:** [T00X — nama task berikutnya]
+- **Loop status:** [open / closed]
+
+---
+
+## Project Overview
+
+- **Name:** [nama project]
+- **Type:** [Application / API-Backend / Campaign / Utility / Workflow]
+- **Current milestone:** [Milestone N — nama milestone]
+- **Current phase:** [Phase N — nama phase]
+
+---
+
+## Decisions Log
+
+| Date | Decision | Reason | Impact Area |
+|------|----------|--------|-------------|
+| YYYY-MM-DD | [keputusan teknis] | [kenapa] | [komponen/area] |
+| YYYY-MM-DD | [keputusan desain] | [kenapa] | [komponen/area] |
+
+---
+
+## Deferred Issues (Carry Forward)
+
+- [ ] [Issue 1] — Priority: High — Target: [milestone/sprint]
+- [ ] [Issue 2] — Priority: Medium — Target: [milestone/sprint]
+- [ ] [Issue 3] — Priority: Low — Target: [milestone/sprint]
+
+---
+
+## Session History
+
+| Session | Date | Phase | Summary |
+|---------|------|-------|---------|
+| 1 | YYYY-MM-DD | Phase 1 | [1 kalimat ringkasan] |
+| 2 | YYYY-MM-DD | Phase 2 | [1 kalimat ringkasan] |
+
+---
+
+## Context Restoration Checklist
+
+Saat mulai session baru, AI harus:
+- [ ] Baca STATE.md ini
+- [ ] Baca MEMORY.md untuk tech decisions
+- [ ] Baca SUMMARY.md dari phase terakhir di `docs/phases/`
+- [ ] Cek docs/task_on_hand.md jika ada task yang sedang berjalan

package/templates/.agent/.shared/checklists/planning-quality.md

@@ -0,0 +1,69 @@
+# Planning Quality Checklist
+
+Quality gate untuk `/graduate` dan `/launch`.
+Menentukan apakah PLANNING.md cukup kaya untuk graduation yang bersih
+dan inisialisasi PAUL state tanpa tanya ulang.
+
+---
+
+## Universal Checks (Semua Types)
+
+- [ ] **Type metadata** ada di header (Application, API/Backend, Campaign, Utility, Workflow, atau custom)
+- [ ] **Problem statement** spesifik — menyebut masalah, audience, dan kenapa penting
+- [ ] **Design decisions** terdokumentasi — minimal 1 keputusan resolved dengan rationale
+- [ ] **Open questions** section ada (boleh kosong, tapi harus ada section-nya)
+- [ ] **Next actions** terdefinisi — minimal 1 concrete next step
+- [ ] **Skill loadout** dicatat — skill apa dari 600+ library yang relevan
+
+---
+
+## Type-Specific Checks
+
+Referensi `data/{type}/config.md` untuk authority pada required vs optional sections.
+
+- [ ] **Semua required sections** (per config.md) punya konten substantif — bukan hanya header atau placeholder
+- [ ] **Optional sections** either terisi atau explicitly marked N/A
+- [ ] **Depth sesuai rigor** — tight types (Utility) boleh brief; deep types (Application) butuh coverage thorough
+
+### Quick Reference
+
+| Type | Required Sections | Rigor |
+|------|-------------------|-------|
+| Application | 8 dari 10 | deep — coverage thorough |
+| API/Backend | 7 dari 8 | standard — balanced |
+| Campaign | 6 dari 7 | creative — measurable goal wajib |
+| Utility | 6 dari 6 | tight — semua wajib, tapi brief |
+| Workflow | 7 dari 8 | standard — idempotency non-negotiable |
+
+---
+
+## PAUL-Readiness Checks (untuk /launch saja)
+
+Hanya required jika user mau `/launch` (graduate + init sekaligus).
+Skip untuk `/graduate` saja.
+
+- [ ] **Phase breakdown** ada dengan independently shippable milestones
+- [ ] **Setiap phase** punya: apa yang dibangun + cara test + outcome yang diharapkan
+- [ ] **Tech stack decisions** resolved — tidak ada critical "TBD" yang blocking
+- [ ] **Detail cukup** untuk AI derive milestones, phases, dan tasks tanpa tanya ulang
+
+---
+
+## Assessment
+
+| Hasil | Kriteria | Action |
+|-------|----------|--------|
+| **Pass** | Semua universal + semua required type-specific | Proceed dengan graduation |
+| **Warn** | Missing optional sections atau thin content di beberapa area | Note gaps, tanya user: "Mau lengkapi dulu atau lanjut?" |
+| **Fail** | Missing required sections atau tidak ada problem statement | Kembali ke `/vibe-plan` untuk complete ideation |
+
+---
+
+## Cara Pakai
+
+Checklist ini direferensikan oleh:
+- `workflows/graduate.md` — step quality_check sebelum buat app directory
+- `workflows/launch.md` — diwarisi via graduate delegation
+
+Ini bukan scorecard angka — ini qualitative assessment yang AI gunakan untuk
+memutuskan apakah perlu warn user atau proceed langsung.

package/templates/.agent/.shared/types/api-backend/config.md

@@ -0,0 +1,25 @@
+# API/Backend — Configuration
+
+| Setting | Value |
+|---------|-------|
+| Rigor | standard |
+| Demeanor | "Pragmatic dan schema-focused — push clarity pada endpoint contracts dan auth sebelum data model. Resist over-engineering di awal." |
+| Sections | 8 |
+
+## Required Sections (7)
+- Purpose & Consumers
+- Core Endpoints
+- Auth Mechanism
+- Data Schema
+- Error Handling & Validation
+- Testing Strategy
+- Deployment
+
+## Optional Sections (1)
+- Performance & Scale (required jika ada estimasi > 1000 RPM)
+
+## Rigor Notes
+- Auth mechanism harus settled sebelum desain endpoint
+- Error format harus konsisten di seluruh API — decide early
+- Resist adding endpoints yang "mungkin dibutuhkan nanti"
+- Test strategy wajib — API tanpa tests = tech debt langsung

package/templates/.agent/.shared/types/api-backend/guide.md

@@ -0,0 +1,41 @@
+# API/Backend — Conversation Guide
+
+## Section 1: Purpose & Consumers
+**Explore:** API ini untuk apa? Siapa yang akan consume — frontend sendiri, mobile app, third party, atau semuanya? Internal-only atau public? Apa use case utama yang paling sering dipanggil?
+**Suggest:** Kalau hanya untuk frontend sendiri — tRPC atau server actions (Next.js) lebih sederhana dari REST. Kalau public atau third party → REST dengan OpenAPI spec. Kalau butuh real-time → pertimbangkan WebSocket atau SSE dari awal.
+**Depth:** required
+
+## Section 2: Core Endpoints
+**Explore:** Endpoint-endpoint utama yang dibutuhkan MVP? List 5-10 yang paling penting. Mana yang paling kritis — yang kalau tidak ada, API tidak berguna?
+**Suggest:** Group endpoints by resource: /users, /products, /orders. Untuk setiap endpoint: method + path + purpose + response shape. Kalau list terlalu panjang → tanda scope terlalu besar untuk MVP.
+**Depth:** required
+
+## Section 3: Auth Mechanism
+**Explore:** Auth seperti apa yang dibutuhkan? JWT, API Key, OAuth 2.0, atau session-based? Apakah ada rate limiting per user/key? Perlu refresh token? Multi-tenant?
+**Suggest:** API Key: paling simpel untuk internal/B2B. JWT: stateless, baik untuk mobile. OAuth 2.0: hanya kalau memang butuh third-party delegasi. Rate limiting: wajib dari awal, jangan retrofit — pakai Redis atau Upstash.
+**Depth:** required
+
+## Section 4: Data Schema
+**Explore:** Tabel/collection apa yang utama? Relasi antar entitas? Field apa yang paling sering di-query? Ada soft delete requirement?
+**Suggest:** Mulai minimal: tambah field lebih mudah dari hapus field. Selalu ada `createdAt`, `updatedAt`, dan `deletedAt` (soft delete). Index pada field yang sering di-filter. Hindari nested objects di PostgreSQL — normalisasi dulu.
+**Depth:** required
+
+## Section 5: Error Handling & Validation
+**Explore:** Bagaimana error harus di-return? Ada standard error format? Validasi input dilakukan di layer mana — controller atau service? Bagaimana handle external API failure?
+**Suggest:** Standard: `{ success: false, error: "message", code: "ERROR_CODE" }`. Validasi: Zod di entry point (controller/route handler), bukan di service. External API: retry dengan exponential backoff untuk transient errors, circuit breaker untuk persistent failures.
+**Depth:** required
+
+## Section 6: Performance & Scale
+**Explore:** Berapa estimasi request per second? Ada endpoint yang berat (report generation, bulk processing)? Butuh caching? Database query yang perlu dioptimasi?
+**Suggest:** MVP: jangan over-optimize. Tambahkan `take` + `skip` ke semua list endpoints dari awal. Cache dengan Redis hanya kalau ada data yang truly static atau expensive. Background jobs (BullMQ) untuk proses yang > 5 detik.
+**Depth:** optional
+
+## Section 7: Testing Strategy
+**Explore:** Unit test, integration test, atau keduanya? Ada contract testing dengan consumers? Bagaimana test auth-protected endpoints? CI/CD pipeline dengan automated tests?
+**Suggest:** Minimum: integration tests yang hit real database (pakai test DB, bukan mock). Setiap endpoint punya: success case + validation error case + auth error case. Gunakan `supertest` atau built-in framework test utilities.
+**Depth:** required
+
+## Section 8: Deployment
+**Explore:** Deploy ke mana? Containerized dengan Docker? Butuh horizontal scaling? Ada environment variables yang perlu di-manage?
+**Suggest:** Railway atau Fly.io untuk simple deployment. Docker Compose untuk local dev dari awal. Secrets di env vars SELALU — tidak pernah di kode atau git. Health check endpoint `/health` wajib untuk monitoring.
+**Depth:** required

package/templates/.agent/.shared/types/api-backend/skill-loadout.md

@@ -0,0 +1,16 @@
+# API/Backend — Skill Loadout
+
+| Skill | Priority | Why |
+|-------|----------|-----|
+| `api-design-principles` | required | REST conventions + error handling standards |
+| `vibe-techdesign` | required | Tech decisions + schema design |
+| `clean-code` | required | Code quality selalu aktif |
+| `systematic-debugging` | required | Error recovery |
+| `prisma-7-patterns` | required | Jika pakai Prisma 7 |
+| `nodejs-backend-patterns` | recommended | Node.js conventions |
+| `database-design` | recommended | Schema modeling |
+| `auth-implementation-patterns` | recommended | Auth yang benar dari awal |
+| `testing-patterns` | recommended | Integration test patterns |
+| `postgresql-optimization` | optional | Query optimization |
+| `bullmq-specialist` | optional | Background jobs |
+| `api-security-best-practices` | recommended | Security hardening |

package/templates/.agent/.shared/types/application/config.md

@@ -0,0 +1,27 @@
+# Application — Configuration
+
+| Setting | Value |
+|---------|-------|
+| Rigor | deep |
+| Demeanor | "Thorough dan architectural — eksplorasi trade-offs, push clarity pada data model dan deployment sebelum lanjut. Coaching partner, bukan interviewer." |
+| Sections | 10 |
+
+## Required Sections (8)
+- Problem Statement
+- Tech Stack
+- Data Model
+- Auth & Roles
+- UI/UX Utama
+- API Surface
+- Deployment & Infrastructure
+- Phase Breakdown
+
+## Optional Sections (2)
+- Keamanan & Compliance (recommended jika ada PII)
+- Skill Loadout & Tools
+
+## Rigor Notes
+- Jangan rush ke tech stack sebelum problem statement clear
+- Data model adalah foundation — luangkan waktu di sini
+- Phase breakdown harus independently shippable — bukan task list
+- Resistance terhadap scope creep: "Ini untuk MVP atau masa depan?"