@atlashub/smartstack-cli 3.39.0 → 3.40.0

package/templates/skills/business-analyse/steps/step-02-decomposition.md

@@ -1,387 +1,387 @@
----
-name: step-02-decomposition
-description: Module decomposition - identify modules, dependency graph, topological sort, client checkpoint
-model: opus
-next_step: steps/step-03a1-setup.md
----
-
-> **Context files:** `_shared.md`
-
-# Step 2: Decomposition
-
-## MANDATORY EXECUTION RULES
-
-- ALWAYS present the module structure to the client for validation (BLOCKING checkpoint)
-- NEVER proceed to specification without client approval of module structure
-- ALWAYS sort modules by dependency (foundations first)
-- ALWAYS detect dependency cycles (ERROR if found)
-- This step runs for BOTH application mode and single-module mode (trivial for single)
-
-## YOUR TASK
-
-Decompose the application scope into ordered modules. Identify dependencies between modules and determine the processing order for iterative specification.
-
----
-
-## EXECUTION SEQUENCE
-
-### 1. Read Cadrage Results
-
-```
-// Determine context source based on workflow mode
-IF workflow.mode === "project":
-  projectFeature = ba-reader.findProjectFeature()
-  currentAppCode = projectFeature.metadata.workflow.currentApplication
-  currentApp = projectFeature.applications.find(a => a.code === currentAppCode)
-  appFeature = ba-reader.findFeature(currentApp.featureJsonPath)
-
-  // Scope decomposition to current application only
-  → Read appFeature.cadrage.globalScope (application-specific scope from step-01b)
-  → Read appFeature.cadrage.stakeholders (filtered to this application)
-  → Read appFeature.cadrage.applicationRoles
-  → Read appFeature.metadata.workflow.mode
-
-  Display: "═══ Module decomposition for application: {currentApp.name} ({currentAppIndex + 1}/{total}) ═══"
-ELSE:
-  ba-reader.findFeature({feature_id})
-  → Read cadrage.globalScope (mustHave, shouldHave, couldHave)
-  → Read cadrage.stakeholders (roles, tasks, pain points)
-  → Read cadrage.applicationRoles
-  → Read cadrage.codebaseContext (existing modules found)
-  → Read metadata.workflow.mode
-```
-
-IF status already "decomposed":
-  Display: "Decomposition already completed. Loading next step."
-  Load: steps/step-03a1-setup.md
-  STOP
-
-### 2. Module Identification
-
-**From the scope items and stakeholder needs, identify distinct modules:**
-
-For each mustHave/shouldHave scope item:
-- Group related items into cohesive functional units
-- Each module should be independently deployable
-- Each module maps to a navigation entry at Level 3 (Module)
-
-**ULTRATHINK Challenge:**
-- "Is this one module or two? Could a user need only half of it?"
-- "Are there hidden modules not mentioned in scope? (e.g., Notifications, Dashboard)"
-- "Does the codebase already have modules that should be extended vs. created new?"
-
-**For each module, determine:**
-
-> **STRUCTURE CARD: modules[]** — Field names are EXACT. Include ALL fields below.
-> ```json
-> {
->   "code": "PascalCase",
->   "name": "Human-readable display name (e.g., 'Gestion du temps', 'Ressources Humaines')",
->   "description": "1-2 sentence description",
->   "featureType": "data-centric|integration|ui-centric|workflow|reporting|full-module",
->   "dependencies": ["OtherModule"],
->   "dependents": ["DependentModule"],
->   "status": "pending",
->   "featureJsonPath": null,
->   "priority": "must|should|could",
->   "sortOrder": 0,
->   "entities": ["Anticipated entity names"],
->   "estimatedComplexity": "simple|medium|complex",
->   "anticipatedSections": [
->     { "code": "list", "description": "Main page: grid, create button (navigates to /create page), click to detail", "resources": ["entity-grid", "entity-filters", "entity-create-page", "entity-edit-page"] },
->     { "code": "dashboard", "description": "Module KPIs and overview (if needed)", "resources": ["entity-kpi-cards", "entity-charts"] }
->   ],
->   "detailTabs": ["Informations", "RelatedEntity1", "Historique"]
-> }
-> ```
-> **MANDATORY fields:** `code`, `name`, `description`, `featureType`, `dependencies`, `dependents`, `status`, `priority`, `sortOrder`, `entities`, `estimatedComplexity`
-> **`name` vs `code`:** `code` is PascalCase for file paths/keys (e.g., "GestionTemps"). `name` is the human-readable label shown in navigation and HTML (e.g., "Gestion du temps"). They are ALWAYS both required.
-> **OPTIONAL fields:** `anticipatedSections` (from cadrage coverageMatrix, enriched here with resources)
-> **status:** Always `"pending"` at creation (updated to `"in-progress"` → `"specified"` → `"validated"` by step-03/04)
-> **sortOrder:** Integer from topological sort (0 = foundation layer)
-> **featureJsonPath:** Set to null at creation; updated to actual path when module feature.json is created in step-03
-> **anticipatedSections:** Array of `{ code, description, resources[] }` objects. ONLY functional zones: `list` (always), `dashboard`, `approve`, `import`, `rapport`, `planning`. NEVER `create`, `edit`, or `detail` (those are actions/pages within `list`). Derived from `cadrage.coverageMatrix[].anticipatedSections` aggregated per module, then refined during decomposition.
-> **detailTabs:** Array of tab names for the detail page (reached by clicking a row in `list`). Always includes "Informations". Add relationship tabs based on entity relations.
-
-**Complexity estimation rules:**
-| Criteria | Simple | Medium | Complex |
-|----------|--------|--------|---------|
-| Entities | ≤2 | 3-5 | >5 |
-| Expected UCs | ≤4 | 5-10 | >10 |
-| Integrations | 0 | 1-2 | >2 |
-| Workflows | None | Simple | Multi-step with approvals |
-
-### 3. Present Module List to Client
-
-Display the identified modules as markdown:
-
-```
-## Modules identifiés pour {application_name}
-
-{language == "fr"
-  ? "**Préfixe de table :** `{metadata.tablePrefix}` — toutes les tables de cette application seront préfixées (ex: `{metadata.tablePrefix}Customers`, `{metadata.tablePrefix}Orders`)"
-  : "**Table prefix:** `{metadata.tablePrefix}` — all tables in this application will be prefixed (e.g., `{metadata.tablePrefix}Customers`, `{metadata.tablePrefix}Orders`)"}
-
-| # | Code | Nom | Type | Priorité | Complexité | Sections | Onglets détail | Entités |
-|---|------|-----|------|----------|------------|----------|----------------|---------|
-| 1 | Customers | Clients | data-centric | must | medium | list | Infos, Adresses, Contacts | Customer, Address, Contact |
-| 2 | Products | Produits | data-centric | must | simple | list | Infos, Catégories | Product, Category |
-| 3 | Orders | Commandes | workflow | must | complex | list, approve | Infos, Lignes, Historique | Order, OrderLine |
-| 4 | Invoices | Facturation | workflow | should | medium | list, dashboard | Infos, Lignes, Paiements | Invoice, InvoiceLine, Payment |
-```
-
-> **Sections = functional zones only:** `list` (always), `dashboard`, `approve`, `import`, `rapport`, `planning`.
-> **Onglets detail** = tabs in the detail page (reached by clicking a row in list).
-> NEVER list `create`, `edit`, or `detail` as sections — they are actions/pages within `list`.
-
-Ask via AskUserQuestion:
-```
-question: "Ces modules correspondent-ils à votre vision de l'application ?"
-header: "Modules"
-options:
-  - label: "Validé"
-    description: "La liste des modules est correcte, on peut continuer"
-  - label: "Ajouter un module"
-    description: "Il manque un module dans la liste"
-  - label: "Retirer un module"
-    description: "Un module est en trop ou hors périmètre"
-  - label: "Modifier un module"
-    description: "Renommer, fusionner ou découper un module"
-```
-
-IF modifications requested:
-  Apply changes and re-present the list
-  Loop until "Validé"
-
-### 4. Dependency Analysis
-
-For each pair of modules, determine dependencies:
-
-**Dependency types:**
-| Type | Signal | Example |
-|------|--------|---------|
-| FK | Module A references entity from module B | Order.CustomerId → Customer |
-| Event | Module A produces event consumed by module B | Order.Created → Invoice.AutoGenerate |
-| Shared-entity | Both modules operate on same entity | Both Orders and Invoices reference Customer |
-| Lookup | Module A uses reference data from module B | Products.CategoryId → Categories (if separate module) |
-
-**For each dependency found:**
-```json
-{
-  "from": "Orders",
-  "to": "Customers",
-  "type": "FK",
-  "description": "Order references Customer via CustomerId"
-}
-```
-
-### 5. Topological Sort
-
-Sort modules so dependencies are processed first:
-
-**Algorithm:**
-1. Build adjacency list from dependency edges
-2. Compute in-degree for each module (number of dependencies)
-3. Start with modules having in-degree 0 (no dependencies)
-4. Process in BFS order: remove processed modules, reduce in-degrees
-5. Group into layers (modules with same depth can be processed in parallel)
-
-**Cycle detection:**
-IF a cycle is detected:
-  Display ERROR:
-  ```
-  ⚠ Dépendance circulaire détectée : {Module A} → {Module B} → {Module A}
-  Les modules ne peuvent pas dépendre mutuellement.
-  ```
-  Ask user to resolve: remove one direction of the dependency
-
-**Result:**
-```json
-{
-  "topologicalOrder": ["Customers", "Products", "Orders", "Invoices"],
-  "layers": [
-    { "layer": 0, "modules": ["Customers", "Products"] },
-    { "layer": 1, "modules": ["Orders"] },
-    { "layer": 2, "modules": ["Invoices"] }
-  ]
-}
-```
-
-### 6. Client Checkpoint (BLOCKING)
-
-Display the complete decomposition as markdown:
-
-```
-## Structure de l'application {application_name}
-
-### Graphe de dépendances
-
-```
-Customers ──┐
-            ├──→ Orders ──→ Invoices
-Products ───┘
-```
-
-### Ordre de traitement
-
-| Ordre | Module | Dépend de | Complexité | Sections | Onglets détail |
-|-------|--------|-----------|------------|----------|----------------|
-| 1 | Customers | (aucune) | medium | list | Infos, Adresses, Contacts |
-| 2 | Products | (aucune) | simple | list | Infos, Catégories |
-| 3 | Orders | Customers, Products | complex | list, approve | Infos, Lignes, Historique |
-| 4 | Invoices | Orders, Customers | medium | list, dashboard | Infos, Lignes, Paiements |
-
-### Détail des sections et resources par module
-
-**Module : {module_name}**
-
-| Section | Description | Resources |
-|---------|-------------|-----------|
-| list | Page principale : grille + création + navigation détail | {entity}-grid, {entity}-filters, {entity}-form |
-| {section_2} | {functional zone description} | {resources} |
-
-| Onglet détail | Contenu |
-|---------------|---------|
-| Informations | Fiche principale de l'entité |
-| {relation_tab} | {sous-entités ou entités liées} |
-| Historique | Journal des modifications |
-
-(Repeat for each module)
-
-> Les modules seront spécifiés dans cet ordre. Chaque module sera
-> détaillé avec ses sections, entités, permissions et maquettes,
-> puis validé avant de passer au suivant.
-```
-
-Ask via AskUserQuestion:
-```
-question: "Validez-vous cette structure et cet ordre de traitement ?"
-header: "Validation"
-options:
-  - label: "Validé, on commence"
-    description: "Commencer la spécification par {first_module}"
-  - label: "Modifier l'ordre"
-    description: "Changer la priorité de traitement des modules"
-  - label: "Revenir aux modules"
-    description: "Modifier la liste des modules"
-```
-
-IF "Revenir aux modules": Go back to step 3
-IF "Modifier l'ordre": Re-sort and re-present
-
-### 7. Write Module Registry
-
-Use ba-writer to update master feature.json:
-
-```
-ba-writer.enrichModuleRegistry({
-  featureId: {feature_id},
-  modules: [{module objects with all fields}],
-  dependencyGraph: {
-    edges: [{dependency edges}],
-    topologicalOrder: [{sorted module codes}],
-    layers: [{layer groups}]
-  }
-})
-
-// Initialize workflow loop state
-ba-writer.enrichSection({
-  featureId: {feature_id},
-  section: "metadata.workflow",
-  data: {
-    moduleOrder: [{topological order}],
-    currentModuleIndex: 0,
-    completedModules: [],
-    currentModule: {first module code}
-  }
-})
-
-ba-writer.updateStatus({feature_id}, "decomposed")
-```
-
-### 8. Create Module Directories
-
-For each module in moduleOrder:
-```
-Create: docs/{app}/{module_code}/business-analyse/v1.0/
-```
-
-> Module feature.json files will be created by step-03 when each module is processed.
-
-### 9. Display Summary & AUTOCHAIN
-
-```
-═══════════════════════════════════════════════════════════
-  DECOMPOSITION COMPLETE - {feature_id}
-═══════════════════════════════════════════════════════════
-
-| Aspect | Value |
-|--------|-------|
-| Modules identified | {count} |
-| Dependencies | {edge count} |
-| Dependency layers | {layer count} |
-| First module | {first module name} |
-
-### Processing Order
-{numbered list of modules}
-
-→ Starting iterative specification with {first_module}...
-═══════════════════════════════════════════════════════════
-```
-
-### 9-bis. Team Mode Decision
-
-After module registry is written and client has approved decomposition:
-
-**IF moduleOrder.length >= 2 (TEAM MODE):**
-
-```
-Display:
-═══════════════════════════════════════════════════════════
-  Multi-module détecté ({N} modules) — Activation du Mode Team
-═══════════════════════════════════════════════════════════
-
-  Chaque module sera spécifié par un agent dédié avec un contexte frais.
-  Vous validerez chaque proposition avant sa finalisation.
-
-  Modules : {moduleOrder.join(" → ")}
-═══════════════════════════════════════════════════════════
-```
-
-**Load: `references/team-orchestration.md`**
-→ Follow the team orchestration protocol starting at §1 (Team Setup)
-→ DO NOT load step-03a1-setup.md — agents will load it internally
-→ **STOP HERE** — the team-orchestration.md takes over the workflow
-
-**ELSE (single module — INLINE MODE):**
-
-AUTOCHAIN — DO NOT STOP HERE:
-- Do NOT ask "shall I continue?" or "voulez-vous que je continue?"
-- Do NOT summarize what's coming next
-- Do NOT wait for user input
-- The user ALREADY validated at step 6 — that was the checkpoint
-- IMMEDIATELY load step-03a1-setup.md
-
----
-
-## SINGLE-MODULE MODE
-
-When `workflow.mode = "module"` or only 1 scope item identified:
-
-1. Create a single module entry with code = `{module_name}` from init
-2. No dependency analysis needed
-3. **Simplified checkpoint:** Quick confirmation instead of full validation
-4. Set moduleOrder = [{single module code}]
-5. Proceed to step-03 (inline, no team)
-
----
-
-## NEXT STEP (conditional)
-
-**If 2+ modules:** DO NOT load step-03a1. Load `references/team-orchestration.md` instead.
-**If 1 module:** Load `./step-03a1-setup.md`
-
-> **FORBIDDEN:** Asking any question between decomposition summary and step-03 load.
-> The client checkpoint was at step 6. Do NOT add another one.
-
-> **PROJECT MODE NOTE:** In project mode, this step runs once per application. After all modules for the current application are specified (via the module loop in step-03d), the application loop in `_module-loop.md` will advance to the next application and re-load this step for module decomposition of the next application. Module registries and dependency graphs are written to the APPLICATION-level feature.json (not the project level). The project-level feature.json tracks application status via `ba-writer.updateApplicationStatus()`.
+---
+name: step-02-decomposition
+description: Module decomposition - identify modules, dependency graph, topological sort, client checkpoint
+model: opus
+next_step: steps/step-03a1-setup.md
+---
+
+> **Context files:** `_shared.md`
+
+# Step 2: Decomposition
+
+## MANDATORY EXECUTION RULES
+
+- ALWAYS present the module structure to the client for validation (BLOCKING checkpoint)
+- NEVER proceed to specification without client approval of module structure
+- ALWAYS sort modules by dependency (foundations first)
+- ALWAYS detect dependency cycles (ERROR if found)
+- This step runs for BOTH application mode and single-module mode (trivial for single)
+
+## YOUR TASK
+
+Decompose the application scope into ordered modules. Identify dependencies between modules and determine the processing order for iterative specification.
+
+---
+
+## EXECUTION SEQUENCE
+
+### 1. Read Cadrage Results
+
+```
+// Determine context source based on workflow mode
+IF workflow.mode === "project":
+  projectFeature = ba-reader.findProjectFeature()
+  currentAppCode = projectFeature.metadata.workflow.currentApplication
+  currentApp = projectFeature.applications.find(a => a.code === currentAppCode)
+  appFeature = ba-reader.findFeature(currentApp.featureJsonPath)
+
+  // Scope decomposition to current application only
+  → Read appFeature.cadrage.globalScope (application-specific scope from step-01b)
+  → Read appFeature.cadrage.stakeholders (filtered to this application)
+  → Read appFeature.cadrage.applicationRoles
+  → Read appFeature.metadata.workflow.mode
+
+  Display: "═══ Module decomposition for application: {currentApp.name} ({currentAppIndex + 1}/{total}) ═══"
+ELSE:
+  ba-reader.findFeature({feature_id})
+  → Read cadrage.globalScope (mustHave, shouldHave, couldHave)
+  → Read cadrage.stakeholders (roles, tasks, pain points)
+  → Read cadrage.applicationRoles
+  → Read cadrage.codebaseContext (existing modules found)
+  → Read metadata.workflow.mode
+```
+
+IF status already "decomposed":
+  Display: "Decomposition already completed. Loading next step."
+  Load: steps/step-03a1-setup.md
+  STOP
+
+### 2. Module Identification
+
+**From the scope items and stakeholder needs, identify distinct modules:**
+
+For each mustHave/shouldHave scope item:
+- Group related items into cohesive functional units
+- Each module should be independently deployable
+- Each module maps to a navigation entry at Level 3 (Module)
+
+**ULTRATHINK Challenge:**
+- "Is this one module or two? Could a user need only half of it?"
+- "Are there hidden modules not mentioned in scope? (e.g., Notifications, Dashboard)"
+- "Does the codebase already have modules that should be extended vs. created new?"
+
+**For each module, determine:**
+
+> **STRUCTURE CARD: modules[]** — Field names are EXACT. Include ALL fields below.
+> ```json
+> {
+>   "code": "PascalCase",
+>   "name": "Human-readable display name (e.g., 'Gestion du temps', 'Ressources Humaines')",
+>   "description": "1-2 sentence description",
+>   "featureType": "data-centric|integration|ui-centric|workflow|reporting|full-module",
+>   "dependencies": ["OtherModule"],
+>   "dependents": ["DependentModule"],
+>   "status": "pending",
+>   "featureJsonPath": null,
+>   "priority": "must|should|could",
+>   "sortOrder": 0,
+>   "entities": ["Anticipated entity names"],
+>   "estimatedComplexity": "simple|medium|complex",
+>   "anticipatedSections": [
+>     { "code": "list", "description": "Main page: grid, create button (navigates to /create page), click to detail", "resources": ["entity-grid", "entity-filters", "entity-create-page", "entity-edit-page"] },
+>     { "code": "dashboard", "description": "Module KPIs and overview (if needed)", "resources": ["entity-kpi-cards", "entity-charts"] }
+>   ],
+>   "detailTabs": ["Informations", "RelatedEntity1", "Historique"]
+> }
+> ```
+> **MANDATORY fields:** `code`, `name`, `description`, `featureType`, `dependencies`, `dependents`, `status`, `priority`, `sortOrder`, `entities`, `estimatedComplexity`
+> **`name` vs `code`:** `code` is PascalCase for file paths/keys (e.g., "GestionTemps"). `name` is the human-readable label shown in navigation and HTML (e.g., "Gestion du temps"). They are ALWAYS both required.
+> **OPTIONAL fields:** `anticipatedSections` (from cadrage coverageMatrix, enriched here with resources)
+> **status:** Always `"pending"` at creation (updated to `"in-progress"` → `"specified"` → `"validated"` by step-03/04)
+> **sortOrder:** Integer from topological sort (0 = foundation layer)
+> **featureJsonPath:** Set to null at creation; updated to actual path when module feature.json is created in step-03
+> **anticipatedSections:** Array of `{ code, description, resources[] }` objects. ONLY functional zones: `list` (always), `dashboard`, `approve`, `import`, `rapport`, `planning`. NEVER `create`, `edit`, or `detail` (those are actions/pages within `list`). Derived from `cadrage.coverageMatrix[].anticipatedSections` aggregated per module, then refined during decomposition.
+> **detailTabs:** Array of tab names for the detail page (reached by clicking a row in `list`). Always includes "Informations". Add relationship tabs based on entity relations.
+
+**Complexity estimation rules:**
+| Criteria | Simple | Medium | Complex |
+|----------|--------|--------|---------|
+| Entities | ≤2 | 3-5 | >5 |
+| Expected UCs | ≤4 | 5-10 | >10 |
+| Integrations | 0 | 1-2 | >2 |
+| Workflows | None | Simple | Multi-step with approvals |
+
+### 3. Present Module List to Client
+
+Display the identified modules as markdown:
+
+```
+## Modules identifiés pour {application_name}
+
+{language == "fr"
+  ? "**Préfixe de table :** `{metadata.tablePrefix}` — toutes les tables de cette application seront préfixées (ex: `{metadata.tablePrefix}Customers`, `{metadata.tablePrefix}Orders`)"
+  : "**Table prefix:** `{metadata.tablePrefix}` — all tables in this application will be prefixed (e.g., `{metadata.tablePrefix}Customers`, `{metadata.tablePrefix}Orders`)"}
+
+| # | Code | Nom | Type | Priorité | Complexité | Sections | Onglets détail | Entités |
+|---|------|-----|------|----------|------------|----------|----------------|---------|
+| 1 | Customers | Clients | data-centric | must | medium | list | Infos, Adresses, Contacts | Customer, Address, Contact |
+| 2 | Products | Produits | data-centric | must | simple | list | Infos, Catégories | Product, Category |
+| 3 | Orders | Commandes | workflow | must | complex | list, approve | Infos, Lignes, Historique | Order, OrderLine |
+| 4 | Invoices | Facturation | workflow | should | medium | list, dashboard | Infos, Lignes, Paiements | Invoice, InvoiceLine, Payment |
+```
+
+> **Sections = functional zones only:** `list` (always), `dashboard`, `approve`, `import`, `rapport`, `planning`.
+> **Onglets detail** = tabs in the detail page (reached by clicking a row in list).
+> NEVER list `create`, `edit`, or `detail` as sections — they are actions/pages within `list`.
+
+Ask via AskUserQuestion:
+```
+question: "Ces modules correspondent-ils à votre vision de l'application ?"
+header: "Modules"
+options:
+  - label: "Validé"
+    description: "La liste des modules est correcte, on peut continuer"
+  - label: "Ajouter un module"
+    description: "Il manque un module dans la liste"
+  - label: "Retirer un module"
+    description: "Un module est en trop ou hors périmètre"
+  - label: "Modifier un module"
+    description: "Renommer, fusionner ou découper un module"
+```
+
+IF modifications requested:
+  Apply changes and re-present the list
+  Loop until "Validé"
+
+### 4. Dependency Analysis
+
+For each pair of modules, determine dependencies:
+
+**Dependency types:**
+| Type | Signal | Example |
+|------|--------|---------|
+| FK | Module A references entity from module B | Order.CustomerId → Customer |
+| Event | Module A produces event consumed by module B | Order.Created → Invoice.AutoGenerate |
+| Shared-entity | Both modules operate on same entity | Both Orders and Invoices reference Customer |
+| Lookup | Module A uses reference data from module B | Products.CategoryId → Categories (if separate module) |
+
+**For each dependency found:**
+```json
+{
+  "from": "Orders",
+  "to": "Customers",
+  "type": "FK",
+  "description": "Order references Customer via CustomerId"
+}
+```
+
+### 5. Topological Sort
+
+Sort modules so dependencies are processed first:
+
+**Algorithm:**
+1. Build adjacency list from dependency edges
+2. Compute in-degree for each module (number of dependencies)
+3. Start with modules having in-degree 0 (no dependencies)
+4. Process in BFS order: remove processed modules, reduce in-degrees
+5. Group into layers (modules with same depth can be processed in parallel)
+
+**Cycle detection:**
+IF a cycle is detected:
+  Display ERROR:
+  ```
+  ⚠ Dépendance circulaire détectée : {Module A} → {Module B} → {Module A}
+  Les modules ne peuvent pas dépendre mutuellement.
+  ```
+  Ask user to resolve: remove one direction of the dependency
+
+**Result:**
+```json
+{
+  "topologicalOrder": ["Customers", "Products", "Orders", "Invoices"],
+  "layers": [
+    { "layer": 0, "modules": ["Customers", "Products"] },
+    { "layer": 1, "modules": ["Orders"] },
+    { "layer": 2, "modules": ["Invoices"] }
+  ]
+}
+```
+
+### 6. Client Checkpoint (BLOCKING)
+
+Display the complete decomposition as markdown:
+
+```
+## Structure de l'application {application_name}
+
+### Graphe de dépendances
+
+```
+Customers ──┐
+            ├──→ Orders ──→ Invoices
+Products ───┘
+```
+
+### Ordre de traitement
+
+| Ordre | Module | Dépend de | Complexité | Sections | Onglets détail |
+|-------|--------|-----------|------------|----------|----------------|
+| 1 | Customers | (aucune) | medium | list | Infos, Adresses, Contacts |
+| 2 | Products | (aucune) | simple | list | Infos, Catégories |
+| 3 | Orders | Customers, Products | complex | list, approve | Infos, Lignes, Historique |
+| 4 | Invoices | Orders, Customers | medium | list, dashboard | Infos, Lignes, Paiements |
+
+### Détail des sections et resources par module
+
+**Module : {module_name}**
+
+| Section | Description | Resources |
+|---------|-------------|-----------|
+| list | Page principale : grille + création + navigation détail | {entity}-grid, {entity}-filters, {entity}-form |
+| {section_2} | {functional zone description} | {resources} |
+
+| Onglet détail | Contenu |
+|---------------|---------|
+| Informations | Fiche principale de l'entité |
+| {relation_tab} | {sous-entités ou entités liées} |
+| Historique | Journal des modifications |
+
+(Repeat for each module)
+
+> Les modules seront spécifiés dans cet ordre. Chaque module sera
+> détaillé avec ses sections, entités, permissions et maquettes,
+> puis validé avant de passer au suivant.
+```
+
+Ask via AskUserQuestion:
+```
+question: "Validez-vous cette structure et cet ordre de traitement ?"
+header: "Validation"
+options:
+  - label: "Validé, on commence"
+    description: "Commencer la spécification par {first_module}"
+  - label: "Modifier l'ordre"
+    description: "Changer la priorité de traitement des modules"
+  - label: "Revenir aux modules"
+    description: "Modifier la liste des modules"
+```
+
+IF "Revenir aux modules": Go back to step 3
+IF "Modifier l'ordre": Re-sort and re-present
+
+### 7. Write Module Registry
+
+Use ba-writer to update master feature.json:
+
+```
+ba-writer.enrichModuleRegistry({
+  featureId: {feature_id},
+  modules: [{module objects with all fields}],
+  dependencyGraph: {
+    edges: [{dependency edges}],
+    topologicalOrder: [{sorted module codes}],
+    layers: [{layer groups}]
+  }
+})
+
+// Initialize workflow loop state
+ba-writer.enrichSection({
+  featureId: {feature_id},
+  section: "metadata.workflow",
+  data: {
+    moduleOrder: [{topological order}],
+    currentModuleIndex: 0,
+    completedModules: [],
+    currentModule: {first module code}
+  }
+})
+
+ba-writer.updateStatus({feature_id}, "decomposed")
+```
+
+### 8. Create Module Directories
+
+For each module in moduleOrder:
+```
+Create: docs/{app}/{module_code}/business-analyse/v1.0/
+```
+
+> Module feature.json files will be created by step-03 when each module is processed.
+
+### 9. Display Summary & AUTOCHAIN
+
+```
+═══════════════════════════════════════════════════════════
+  DECOMPOSITION COMPLETE - {feature_id}
+═══════════════════════════════════════════════════════════
+
+| Aspect | Value |
+|--------|-------|
+| Modules identified | {count} |
+| Dependencies | {edge count} |
+| Dependency layers | {layer count} |
+| First module | {first module name} |
+
+### Processing Order
+{numbered list of modules}
+
+→ Starting iterative specification with {first_module}...
+═══════════════════════════════════════════════════════════
+```
+
+### 9-bis. Team Mode Decision
+
+After module registry is written and client has approved decomposition:
+
+**IF moduleOrder.length >= 2 (TEAM MODE):**
+
+```
+Display:
+═══════════════════════════════════════════════════════════
+  Multi-module détecté ({N} modules) — Activation du Mode Team
+═══════════════════════════════════════════════════════════
+
+  Chaque module sera spécifié par un agent dédié avec un contexte frais.
+  Vous validerez chaque proposition avant sa finalisation.
+
+  Modules : {moduleOrder.join(" → ")}
+═══════════════════════════════════════════════════════════
+```
+
+**Load: `references/team-orchestration.md`**
+→ Follow the team orchestration protocol starting at §1 (Team Setup)
+→ DO NOT load step-03a1-setup.md — agents will load it internally
+→ **STOP HERE** — the team-orchestration.md takes over the workflow
+
+**ELSE (single module — INLINE MODE):**
+
+AUTOCHAIN — DO NOT STOP HERE:
+- Do NOT ask "shall I continue?" or "voulez-vous que je continue?"
+- Do NOT summarize what's coming next
+- Do NOT wait for user input
+- The user ALREADY validated at step 6 — that was the checkpoint
+- IMMEDIATELY load step-03a1-setup.md
+
+---
+
+## SINGLE-MODULE MODE
+
+When `workflow.mode = "module"` or only 1 scope item identified:
+
+1. Create a single module entry with code = `{module_name}` from init
+2. No dependency analysis needed
+3. **Simplified checkpoint:** Quick confirmation instead of full validation
+4. Set moduleOrder = [{single module code}]
+5. Proceed to step-03 (inline, no team)
+
+---
+
+## NEXT STEP (conditional)
+
+**If 2+ modules:** DO NOT load step-03a1. Load `references/team-orchestration.md` instead.
+**If 1 module:** Load `./step-03a1-setup.md`
+
+> **FORBIDDEN:** Asking any question between decomposition summary and step-03 load.
+> The client checkpoint was at step 6. Do NOT add another one.
+
+> **PROJECT MODE NOTE:** In project mode, this step runs once per application. After all modules for the current application are specified (via the module loop in step-03d), the application loop in `_module-loop.md` will advance to the next application and re-load this step for module decomposition of the next application. Module registries and dependency graphs are written to the APPLICATION-level feature.json (not the project level). The project-level feature.json tracks application status via `ba-writer.updateApplicationStatus()`.