Spring Boot Web Application Specification Generator

This skill generates a comprehensive specification document (Markdown) that serves as a blueprint for building a monolith Spring Boot 3 web application with server-rendered views. The spec is intended to be followed by a developer or a coding agent to produce a fully functional project scaffold.

The specification does NOT generate code. It produces a detailed, opinionated technical document describing every layer of the application — from Maven configuration to JTE layouts to security filter chains — so that implementation becomes a mechanical exercise.

Technology Stack

Core Stack (Always Included)

These are the fixed versions the spec targets. Do not deviate unless the user explicitly requests different versions.

Optional Integration Versions

Include in the version table only when the corresponding integration is selected.

Core Dependencies

The spec must include these in the Maven configuration section (always):

• Spring Web (starter-web)
• Spring Modulith (core, events-api)
• JTE Spring Boot Starter (gg.jte:jte-spring-boot-starter-3)
• JTE precompiler (gg.jte:jte-maven-plugin with precompile goal at process-classes phase, output to target/classes so precompiled templates are included in the Spring Boot fat JAR)
• Lombok
• Spring Boot DevTools
• MapStruct (with annotation processor)
• frontend-maven-plugin (for Node/Vite build)

Conditional Dependencies

If Database = MongoDB:

• Spring Data MongoDB (spring-boot-starter-data-mongodb)
• MongoDB Driver
• Spring Modulith starter-mongodb (spring-modulith-starter-mongodb)

If Database = PostgreSQL or MySQL:

• Spring Data JPA (spring-boot-starter-data-jpa)
• PostgreSQL driver (org.postgresql:postgresql) or MySQL driver (com.mysql:mysql-connector-j)
• Flyway (org.flywaydb:flyway-core) for schema migration
• Spring Modulith starter-jpa (spring-modulith-starter-jpa)

If Auth = Keycloak:

• Spring Security (spring-boot-starter-security)
• Spring Security OAuth2 Client (spring-boot-starter-oauth2-client)

If Auth = Spring Security (form login):

• Spring Security (spring-boot-starter-security)

If Scheduling = yes:

• Spring Quartz (spring-boot-starter-quartz)
• If Database = MongoDB: io.fluidsonic.mirror:fluidsonic-mirror-quartz (MongoDB job store for Quartz)
• If Database = PostgreSQL/MySQL: Quartz JDBC job store (built-in)

If Scheduling = yes AND Spring Batch = yes:

• Spring Batch (spring-boot-starter-batch)

If Scheduling = yes AND Spring Batch = yes AND Remote Partitioning = yes:

• Spring Batch Integration (spring-batch-integration)
• Spring Integration AMQP (spring-integration-amqp)
• Spring Boot AMQP Starter (spring-boot-starter-amqp)

If Messaging = yes:

• Spring Boot AMQP Starter (spring-boot-starter-amqp)

(shared with Remote Partitioning — if both are selected, include the dependency once)

If i18n = yes:

• No extra dependency — uses Spring's built-in `MessageSource` (part of spring-context,

already on the classpath via starter-web). Resource bundles under src/main/resources/i18n/, a CookieLocaleResolver, and a LocaleChangeInterceptor provide locale resolution and switching; a small Localizer helper exposes messages to JTE templates.

If Reporting = yes:

• JasperReports (net.sf.jasperreports:jasperreports:7.0.3) — report engine with JRDesign API for programmatic layout
• JasperReports Fonts (net.sf.jasperreports:jasperreports-fonts:7.0.3)
• OpenPDF (com.github.librepdf:openpdf:2.0.4) — PDF export engine for JasperReports 7.x
• Apache POI OOXML (org.apache.poi:poi-ooxml:5.4.1) — XLSX export support

When the Skill Triggers

Generate the spec when the user provides an application name and version that corresponds to one of the custom applications defined in CLAUDE.md. The skill reads all required inputs from the project's context files — no interactive Q&A is needed for the core inputs.

The user invokes this skill by specifying the target application and version, for example:

• /specgen-spring-jpa-jtehtmx hub_middleware v1.0.3
• /specgen-spring-jpa-jtehtmx hub_middleware v1.0.3 module:Location Information
• /specgen-spring-jpa-jtehtmx "Hub Middleware" v1.0.3

The skill then locates the matching context folder and reads all input files automatically.

Version Gate

Before starting any work, resolve the application folder first (see Input Resolution below), then check CHANGELOG.md in the application folder (<app_folder>/CHANGELOG.md):

1. If <app_folder>/CHANGELOG.md does not exist, skip this check (first-ever execution for this application).
2. If <app_folder>/CHANGELOG.md exists, scan all ## vX.Y.Z headings and determine the highest version using semantic versioning comparison.
3. Compare the requested version against the highest version:

• If requested version >= highest version: proceed normally.
• If requested version < highest version: STOP immediately. Print: "Version {requested} is lower than the current application version {highest} recorded in <app_folder>/CHANGELOG.md. Execution rejected." Do NOT proceed with any work.

Input Resolution

This skill uses standardized input resolution. Provide:

Application Folder Resolution

The application name is matched against root-level application folders:

1. Strip any leading <number>_ prefix from folder names (e.g., 1_hub_middleware → hub_middleware)
2. Match case-insensitively against the provided application name
3. Accept snake_case, kebab-case, or title-case input (all match the same folder)
4. If no match found, list available applications and stop

Auto-Resolved Paths

Example Invocations

• /specgen-spring-jpa-jtehtmx hub_middleware v1.0.3 (all modules)
• /specgen-spring-jpa-jtehtmx hub_middleware v1.0.3 module:Location Information (one module)
• /specgen-spring-jpa-jtehtmx "Hub Middleware" v1.0.3

Version Filtering

When a version is provided, only include user stories, NFRs, and constraints from versions <= the provided version. For example, if v1.0.3 is specified:

• Include items tagged [v1.0.0], [v1.0.1], [v1.0.2], [v1.0.3]
• Exclude items tagged [v1.0.4] or later
• Version comparison uses semantic versioning order

Module Filtering

When module:<name> is provided:

• Only generate the SPEC.md for that specific module
• Other existing module spec files remain untouched
• SPECIFICATION.md (root) gets a partial update — only that module's entry in the TOC

is added or updated; all other TOC entries are preserved as-is

Gathering Input

The specification is driven by six input sources that are read from the project's context files. The skill does NOT ask the user for database, authentication, scheduling, or messaging choices — it determines these automatically from the context.

Input 1: Application Name (from CLAUDE.md)

From CLAUDE.md (already loaded in context), locate the target application under the Custom Applications section. Extract:

• Application name: The section heading (e.g., "Hub Middleware", "HC Adapter")
• Application description: The description paragraph below the heading
• Dependencies: The "Depends on" list — this is the primary source for determining

optional components (see Determining Optional Components)

The application name is used to derive:

• Artifact ID: Kebab-case of the application name (e.g., hub-middleware)
• Group ID: com.bestinet.urp (project-level constant)
• Base package: com.bestinet.urp.<artifactid_no_hyphens> (e.g., com.bestinet.urp.hubmiddleware)

Input 2: User Stories (from PRD.md)

Read <app_folder>/context/PRD.md. This file contains all user stories organized by module. Extract:

• System modules: Modules under the # System Module heading (e.g., User, Notification,

Activities, Audit Trail). These become system-level modules in the spec.

• Business modules: Modules under the # Business Module heading (e.g., Location

Information, Corridor, Employer). These become business-level modules.

• User stories per module: Each ### User Story section contains tagged items like

[USHM00108] As a user, I want to.... These define the functional requirements for each module's service interface and page controllers.

The user stories directly inform:

• Which CRUD operations each module's service must expose
• Which page controllers and JTE views are needed
• Which form fields and validation rules apply

Important: Items with strikethrough (~~text~~) are deprecated — do NOT include them as active requirements. Instead, list them in the "Removed / Replaced" subsection of the traceability table (see spec-template.md) so that developers can see what was removed and which version removed it. If a deprecated item has a replacement (e.g., USHM00015 replaced by USHM00222), note the replacement ID.

• Version tags per item: Each ### User Story, ### Non Functional Requirement,

and ### Constraint section contains one or more version blocks formatted as [v1.0.x]. Items listed under each version tag belong to that version. The skill must track the version tag for each item (user story, NFR, constraint) and carry it through to the generated specification's traceability section. When a version block explicitly lists "Removed ... from previous version", record those removals in a "Removed / Replaced" subsection with the version that removed them and the replacement ID (if any).

Input 3: Non-Functional Requirements (from PRD.md)

Within the same PRD.md, each module has a ### Non Functional Requirement section with tagged items like [NFRHM0120]. These inform:

• Data storage strategies (e.g., "stored in JWT token", "stored in the database")
• Integration patterns (e.g., "API call to SSO service", "sent asynchronously")
• Performance constraints (e.g., "automatically deleted after 30 days")
• Security requirements (e.g., "require user to relogin")

NFRs should be mapped to specific technical decisions in the spec — for example, an NFR stating "stored in JWT token" confirms stateless auth, while "sent asynchronously" confirms event-driven processing.

Input 4: Constraints (from PRD.md)

Within the same PRD.md, each module has a ### Constraint section with tagged items like [CONSHM042]. These define hard boundaries that the spec must enforce:

• Field-level restrictions (e.g., "can only change name and phone number")
• Feature boundaries (e.g., "only 2 delivery channels supported")
• Scope limitations (e.g., "does not manage any user, permissions and roles")

Constraints are embedded directly into the relevant module blueprint — they inform service interface contracts, validation rules, and UI form configurations.

Input 5: Module Model (from model/ folder)

Read <app_folder>/context/model/MODEL.md first as the index, then read the individual module model files in each module subfolder.

MODEL.md provides:

• Summary table of all modules with collection counts and design decisions
• Links to each module's detailed model files

Per-module files (e.g., model/location-information/model.md):

• Complete document/entity field definitions with types and constraints
• Embedded vs. referenced relationships
• Collection/table names
• Index specifications
• Audit field patterns

Per-module schema (e.g., model/location-information/schemas.json):

• JSON schemas defining exact field types, required fields, and validation rules

Per-module diagram (e.g., model/location-information/document-model.mermaid):

• Visual representation of document structure and relationships

The module model directly maps to:

• MongoDB documents or JPA entities (field-for-field, not placeholder)
• Repository methods and query patterns
• MapStruct mapper definitions
• DTO structures matching the actual module fields
• Service interface method signatures

• Version tracking: The MODEL.md summary table includes a "Versions" column listing

which versions each module participates in (e.g., "1.0.0, 1.0.1, 1.0.3"). Per-module model.md files may also include version annotations on fields and indexes. The skill must carry these version tags into the generated specification.

Input 6: HTML Mockup Screens (from mockup/ folder)

Read <app_folder>/context/mockup/MOCKUP.html first as the index page, then read the HTML files organized by role in subfolders.

MOCKUP.html provides:

• Application identity (name, version, short name)
• Design tokens: fonts, colors, spacing from Tailwind config
• List of user roles and their screen sets
• Setup instructions for the mockup server

Role-specific subfolders (e.g., mockup/hub_administrator/content/):

• Individual HTML screens for each page in the application
• Screen layout: which components are used (tables, forms, cards, tabs, modals)
• Navigation structure from sidebar HTML files
• Data display patterns (list pages, detail pages, create/edit forms)

IMPORTANT — Role folders inform access control, NOT URL paths. The role-specific folder structure (e.g., mockup/hub_administrator/content/corridor.html) determines:

1. Which role can access the page → @PreAuthorize("hasRole('HUB_ADMINISTRATOR')")
2. Which sidebar navigation items appear for each role

It does NOT determine the URL path. The URL path is always module-based:

• @RequestMapping("/corridor") — NOT @RequestMapping("/hub_administrator/corridor")
• Fragment URL: @RequestMapping("/corridor/fragments") — NOT @RequestMapping("/api/content/hub_administrator/corridor")

Shared partials (e.g., mockup/partials/):

• header.html — Header bar layout and elements
• footer.html — Footer layout
• sidebar-<role>.html — Per-role navigation menus
• shell.html — Page shell/wrapper structure

The mockup screens directly map to:

• JTE page templates (one per HTML screen)
• JTE fragment templates (for htmx partial updates)
• Page controller endpoints (one per screen)
• View model classes (with fields matching the data shown in each screen)
• Sidebar navigation items per role
• Form field layouts and validation display
• Design tokens for Tailwind configuration (colors, fonts from MOCKUP.html)

• Version tracking: The MOCKUP.html index page shows a version tag on each screen

card (e.g., v1.0.0, v1.0.3). Individual mockup screens may include version annotations. The skill must associate each mockup screen with its version and carry this through to the generated specification.

PRD.md Extended Sections

Before determining optional components, check PRD.md for the following extended sections and extract their content for use throughout specification generation:

Architecture Principle Extraction

If PRD.md contains an # Architecture Principle section, read it and extract architectural patterns as a structured context object. These patterns serve as primary signals for optional component determination and specification content:

If the section is absent, proceed with existing CLAUDE.md-only detection.

Design System Extraction

If PRD.md contains a # Design System section with a file reference (e.g., [DESIGN_SYSTEM.md](reference/DESIGN_SYSTEM.md)):

1. Resolve the path relative to PRD.md and read the referenced file
2. Extract design tokens: color palettes, typography, component patterns
3. Include a "Design System Integration" subsection in SPECIFICATION.md under "Application Configuration" specifying:

• Tailwind CSS custom theme configuration (colors, fonts from design system)
• JTE layout template design token variables
• CSS variable definitions for dynamic theming

1. In per-module SPEC.md, reference which design tokens apply to specific components (e.g., status badges use accent-success color for Active)

If the section is absent, use design tokens from MOCKUP.html Tailwind config (existing behavior).

High Level Process Flow Extraction

If PRD.md contains a # High Level Process Flow section:

1. Parse all named process flows and their ordered steps
2. Each process flow becomes an authoritative source for the messaging pipeline sections in per-module SPEC.md:

• Each step maps to a specific service method, event listener, or message handler
• Error paths generate corresponding exception handlers and dead-letter queue configurations
• ACK/NACK patterns become outbound message publisher specifications with defined message schemas

1. Include a "Process Flow Implementation" subsection in each affected module's SPEC.md that maps flow steps to service methods
2. Generate an event catalog in SPECIFICATION.md listing all domain events derived from process flows

If the section is absent, derive messaging patterns from NFRs only (existing behavior).

Determining Optional Components

Instead of asking the user, the skill determines optional components by analyzing the dependencies listed in CLAUDE.md, the # Architecture Principle section in PRD.md (if present), and cross-referencing with PRD.md NFRs and constraints.

Database Detection

First check PRD.md `# Architecture Principle`: If it explicitly mentions a database type (e.g., "document based database", "MongoDB", "relational database", "MySQL"), use that as the primary signal.

Fallback to CLAUDE.md: Examine the "Depends on" list in CLAUDE.md for the target application:

Also check CLAUDE.md's database section for the exact database name, and read ENVIRONMENT.md (in the project root) for host, port, and credentials to use in the spec's application configuration.

Authentication Detection

If Auth = Keycloak, also extract from CLAUDE.md:

• Keycloak version from "Hub Single Sign On" section
• Keycloak realm: Default derived from project name
• Keycloak client ID: Default <artifact-id>-web
• Keycloak issuer URI: Default http://localhost:8180/realms/<realm>
• Keycloak roles: Infer from mockup sidebar roles (e.g., hub_administrator → HUB_ADMINISTRATOR)

Messaging Detection

If Messaging = yes, also extract the RabbitMQ version from the corresponding message queue section in CLAUDE.md.

Scheduling Detection

Scheduling is determined from PRD.md content:

If Scheduling = yes, further determine:

• Spring Batch = yes if NFRs mention "batch processing", "ETL", "bulk import/export",

or processing large volumes of data with reader/processor/writer patterns

• Remote Partitioning = yes if NFRs mention "distributed processing", "horizontal

scaling", or "partitioned batch jobs"

Reporting Detection

Reporting is determined from PRD.md content:

If Reporting = yes, the spec includes:

• JasperReports with JRDesign API for fully programmatic report layout (no .jrxml templates)
• ReportDefinition interface with buildDesign() method for modules to implement
• ReportDesignHelper utility class with static builders for common layout patterns
• ReportService orchestrating compile → fill → export via JRBeanCollectionDataSource
• Report registry persisted in the database
• Report page controller with parameter form and download endpoint
• JTE templates for report list and parameter form pages
• Multi-format export: PDF (OpenPDF), XLSX (Apache POI), CSV

Internationalisation Detection

Internationalisation is determined from PRD.md content:

If i18n = yes, the spec includes:

• An I18nConfig providing a MessageSource, a CookieLocaleResolver, and a LocaleChangeInterceptor
• Resource bundles under src/main/resources/i18n/ (messages.properties + messages_<locale>.properties)
• An app.i18n config block (default-locale, supported-locales) under the app: namespace
• A Localizer helper exposed as a global @ModelAttribute so JTE templates resolve message keys
• A language switcher fragment (htmx-driven, hitting ?lang=<locale>)
• Bean-validation message localization wired to the same MessageSource

If i18n = no, do NOT generate a locale/language switcher or any message-bundle scaffolding — JTE templates use literal copy. This is the existing "do NOT generate UI without backend support" constraint (see Constraints below).

Summary of Determination

After analyzing all inputs, produce a determination summary before generating the spec. Present it to the user for confirmation:

Optional Component Determination:
- Database:      MongoDB (from CLAUDE.md → depends on Hub Database)
- Authentication: Keycloak (from CLAUDE.md → depends on Hub Single Sign On)
- Scheduling:    yes (from PRD.md → NFR mentions automatic deletion)
  - Spring Batch: no
  - Remote Partitioning: no
- Messaging:     yes (from CLAUDE.md → depends on Hub to HC/SC Adapter Message Queue)
- Reporting:     yes (from PRD.md → Report module with Report interface NFR)
- Internationalisation: yes (en, ms) (from PRD.md → English + Bahasa Malaysia required)

If the user disagrees with any determination, allow them to override before proceeding.

Required Inputs

After determination, these values are needed. Most are derived automatically:

Auto-derived from context files:

• Application name: From CLAUDE.md section heading
• Artifact ID: Kebab-case of application name
• Group ID: com.bestinet.urp
• Base package: com.bestinet.urp.<artifactid>
• Application description: From CLAUDE.md description
• Modules: From PRD.md module headings + model/MODEL.md
• Database: Auto-determined (see above)
• Authentication: Auto-determined (see above)
• Scheduling: Auto-determined (see above)
• Messaging: Auto-determined (see above)
• Database name/credentials: From ENVIRONMENT.md (root-level file with local environment credentials)
• User roles: From mockup sidebar files
• Design tokens: From MOCKUP.html Tailwind config

Auto-derived from CLAUDE.md (Port Allocation table):

• Server port: Look up the application's port from the Port Allocation table in the Custom Applications section of CLAUDE.md. Do NOT hardcode a default — the port MUST match the allocated port for this application.

Optional (use sensible defaults if not found in context):

• Default theme: Default light (supports light/dark)
• Log level: Default INFO for application, WARN for frameworks
• Default locale: Default en (only relevant if i18n = yes)
• Supported locales: Derived from PRD.md language requirements (e.g., en, ms); default [en]

Generating the Specification

Once inputs are gathered from context files and optional components are determined, generate the specification as a multi-file output split by module. Read the spec template at references/spec-template.md for the exact structure and content of each section. The template is the authoritative guide — follow it closely.

The specification is split into two categories:

1. Root `SPECIFICATION.md` — Contains the Table of Contents, shared infrastructure,

and application-level sections that apply across all modules.

1. Per-module `<module-name>/SPEC.md` — Each module gets its own folder with

a self-contained specification file covering that module's complete blueprint.

This split enables a coding agent to:

• First, scaffold the shared infrastructure from SPECIFICATION.md
• Then, implement each module independently by picking up its <module>/SPEC.md

Important: The generated spec must use real module data from the context files, not generic placeholders. Specifically:

• Modules must use the actual module names from PRD.md and MODEL.md

(e.g., locationInformation, corridor, employer — not module1, module2)

• Document/Entity fields must match the actual fields defined in the module model

files (e.g., model/location-information/model.md), not placeholder fieldOne/fieldTwo

• Service interfaces must expose methods matching the actual user stories (e.g., if

a story says "view provinces by country", the service needs listProvincesByCountry())

• Page controllers must map to the actual mockup screens (e.g., if

hub_administrator/content/location_information.html exists, there must be a matching controller endpoint and JTE page). The controller URL is module-based (e.g., /location-information), NOT role-prefixed (e.g., NOT /hub_administrator/location-information). The mockup's role folder determines @PreAuthorize annotations, not URL structure.

• Sidebar navigation must match the mockup sidebar files per role. Sidebar hrefs use

module-based paths (e.g., /corridor, /quota) — the role determines which items appear in the sidebar, not the URL prefix

• Form fields must match what the mockup screens display
• Design tokens (colors, fonts) must match the MOCKUP.html Tailwind configuration
• Version tags: Every user story ID, NFR ID, constraint ID, and mockup screen in

the traceability section must include its version tag (e.g., USHM00228 [v1.0.3]). This enables incremental implementation by version. ALL traceability sub-tables (User Stories, NFRs, AND Constraints) MUST include the `| Version |` column. Do not omit the Version column from any table — even if a module only has v1.0.0 items.

• Removed / Replaced items: The traceability section must include a "Removed / Replaced"

subsection listing any deprecated items from previous versions — showing the removed ID, the version that removed it, the replacement ID (if any), and a brief reason. This ensures developers can see what was removed and understand the evolution of requirements. If a module has no removed items, include the subsection with _None._ to make it explicit that nothing was removed.

Output Structure

<app_folder>/context/specification/
├── SPECIFICATION.md                    ← TOC + shared/application-level specs
├── location-information/
│   └── SPEC.md                         ← Module blueprint for Location Information
├── corridor/
│   └── SPEC.md                         ← Module blueprint for Corridor
├── employer/
│   └── SPEC.md                         ← Module blueprint for Employer
├── ...                                 ← One folder per module from PRD.md

What Goes in SPECIFICATION.md (Root)

The root file contains the TOC and all shared/application-level sections. These are the sections that a coding agent implements first before any module work:

#### 1. Project Overview Project metadata, application description, technology stack summary, the complete list of user roles extracted from mockup sidebar files, and the Module Index — a table listing every module with a link to its <module>/SPEC.md file.

#### 2. Maven Configuration Complete pom.xml structure with all dependencies (core + selected conditional), plugin configurations (MapStruct annotation processor, Spring Boot Maven plugin, frontend-maven-plugin for Vite build, maven-clean-plugin to delete on-demand jte-classes/ folder, jte-maven-plugin with precompile goal at process-classes phase outputting to ${project.build.directory}/classes), and property management.

#### 3. Application Configuration (conditional content varies) A single application.yml (no profile-specific files like application-dev.yml or application-prod.yml) covering database connection (MongoDB URI or JDBC datasource depending on selection), auth settings (Keycloak/OAuth2 if selected, or form login if selected), JTE configuration (with JTE_DEV_MODE and JTE_PRECOMPILED env vars), scheduling config (if selected), theme defaults, and logging configuration. All environment-sensitive values (ports, hostnames, credentials, URIs) MUST use Spring's `${ENV_VAR:default}` syntax to allow externalization via environment variables while keeping sensible defaults for local development. Do NOT use Spring profiles or profile-specific YAML files — environment differences are handled entirely through environment variables (e.g., via .env file locally or system environment variables in deployment).

#### 3a. Application-Specific Configuration (app: namespace)

All application-owned configuration — any config specific to THIS application, as opposed to the Spring/Java ecosystem — MUST live under the top-level app: key in application.yml. NEVER place application-specific keys under Spring framework namespaces (spring.*, server.*, management.*, logging.*) — those are reserved for framework configuration.

Grouping rule. Organise the app: tree by module:

• Cross-cutting values (version, shared feature flags, shared timeouts, CORS,

shared security settings, shared messaging infrastructure, shared object-storage settings) sit directly under app.* with no module prefix.

• Per-module values are grouped under app.<module-kebab-case>.*, one block

per module that needs runtime config. A module with a single config value still gets its own block — do not flatten.

• NEVER place module-specific keys at the YAML root (e.g., top-level

notification:, batch-job:, audit-trail:). They MUST be nested under app:.

Naming rule. YAML keys use kebab-case (from-address, retry-limit, max-retry). Spring Boot's relaxed binding maps them to camelCase Java fields automatically. Do NOT use camelCase or snake_case in YAML.

Binding rule. For every app.* subtree (cross-cutting OR per-module), bind once via @ConfigurationProperties on a record in the corresponding module's config subpackage — or, for cross-cutting values, in the application-level config subpackage. Do NOT inject individual values via @Value("${app....}") scattered across beans or templates. Bind once at the module boundary and inject the typed record.

Environment-override rule. Every app.* leaf value MUST use Spring's ${ENV_VAR:default} syntax so it can be overridden per environment without editing YAML. Every referenced ${ENV_VAR} MUST appear in the .env file generated in section 3b.

Example structure:

app:
  # Cross-cutting (no module prefix)
  version: ${APP_VERSION:1.0.0}
  cors:
    allowed-origins: ${APP_CORS_ALLOWED_ORIGINS:http://localhost:3000}
  security:
    public-paths:
      - /actuator/health
      - /login
      - /error

  # Per-module blocks — one per module with runtime config
  notification:
    email:
      from-address: ${NOTIFICATION_FROM_ADDRESS:[email protected]}
      retry-limit: ${NOTIFICATION_RETRY_LIMIT:3}
    cleanup:
      retention-days: ${NOTIFICATION_RETENTION_DAYS:30}
  batch-job:
    cleanup:
      retention-days: ${BATCH_JOB_RETENTION_DAYS:90}
  audit-trail:
    cleanup:
      retention-days: ${AUDIT_TRAIL_RETENTION_DAYS:365}

*Example cross-cutting record (`app.`):**

@ConfigurationProperties(prefix = "app")
public record AppProperties(String version, Cors cors, Security security) {
    public record Cors(List<String> allowedOrigins) {}
    public record Security(List<String> publicPaths) {}
}

*Example per-module record (`app.<module>.`):**

@ConfigurationProperties(prefix = "app.notification")
public record NotificationProperties(
    Email email,
    Cleanup cleanup
) {
    public record Email(String fromAddress, int retryLimit) {}
    public record Cleanup(int retentionDays) {}
}

Enable binding at the application root with @ConfigurationPropertiesScan on the main application class, or register each record individually via @EnableConfigurationProperties on a @Configuration class in the owning module.

Mandatory cross-cutting values:

Additional cross-cutting keys may be added as the application grows, but app.version is always present.

`app.version` specifics. The .env file must include APP_VERSION={version} with the actual version value. The pom.xml <version> element MUST also be set to the same version value (e.g., 1.0.3). When multiple versions were provided during skill invocation, use the highest one. The application MUST expose this version in the footer of every page. Expose it by injecting AppProperties into a @ControllerAdvice-annotated class that registers appVersion as a global @ModelAttribute, then read it from the shared footer layout (footer.jte or footer.html) as ${appVersion} and render it as v{version} (e.g., v1.0.3). Do NOT use @Value("${app.version}") in controllers or templates.

#### 3b. .env File Generation from ENVIRONMENT.md Generate a .env file at the project root by reading ENVIRONMENT.md from the project root. The .env file maps ENVIRONMENT.md credential and platform values to the environment variable names referenced in application.yml. The spec must define the complete .env content with actual values from ENVIRONMENT.md.

Process:

1. Read ENVIRONMENT.md from the project root
2. Extract credential values from ENVIRONMENT.md (# Supporting 3rd Party Applications

for database hosts, ports, usernames, passwords, plus Keycloak/RabbitMQ); read toolchain paths (JDK, Maven, Node.js) from DEVTOOL.md

1. Map each value to the corresponding ${ENV_VAR} name used in application.yml
2. Generate the .env file with KEY=value pairs

Example `.env` output (derived from ENVIRONMENT.md):

# Database
DB_HOST=localhost
DB_PORT=3306
DB_NAME=hub_supp
DB_USERNAME=root
DB_PASSWORD=B3st1n3t@2025

# Authentication (Keycloak)
KEYCLOAK_HOST=http://localhost:8180
KEYCLOAK_REALM=urp
KEYCLOAK_CLIENT_ID=hub-middleware-web

# Messaging (RabbitMQ)
RABBITMQ_HOST=localhost
RABBITMQ_PORT=5672
RABBITMQ_USERNAME=guest
RABBITMQ_PASSWORD=guest

# Server
SERVER_PORT=<port from CLAUDE.md Port Allocation table>

# Platform
JAVA_HOME=C:\Users\rashidee.rashid.BESTINET\.jdks\azul-21.0.9
MAVEN_HOME=C:\Users\rashidee.rashid.BESTINET\apache-maven-3.9.12

Rules:

• Only include variables that are actually referenced in application.yml
• Use actual values from ENVIRONMENT.md — never use placeholders or TODO
• If ENVIRONMENT.md does not exist or a value is not found, use sensible defaults for local

development (e.g., localhost, default ports)

• The .env file is gitignored (already covered in .gitignore)

#### 4. Build & Tooling Vite configuration, Tailwind CSS setup (using design tokens from MOCKUP.html), PostCSS config, frontend directory structure, and Maven integration for building JS/CSS assets.

#### 4b. .gitignore Generate a .gitignore file at the project root that excludes all generated, downloaded, and environment-specific files from version control. The spec must include the complete .gitignore content. At minimum, include:

# Build output
target/

# IDE
.idea/
*.iml
.vscode/
.settings/
.project
.classpath
*.swp
*~

# OS
.DS_Store
Thumbs.db

# Frontend
frontend/node_modules/
frontend/.vite/
src/main/resources/static/dist/

# JTE generated
jte-classes/

# E2E
e2e/node_modules/
e2e/test-results/
e2e/playwright-report/
e2e/visual-baselines/

# Logs
*.log
logs/

# Environment
.env
.env.*

Conditional additions:

• If Database = PostgreSQL/MySQL: No additional entries needed (Flyway migrations are committed)
• If Scheduling = yes AND Spring Batch = yes: Add *.tmp for batch temp files
• If Reporting = yes: No additional .gitignore entries needed (report layouts are programmatic Java code, not external template files)

#### 5. Package Structure The complete directory tree rooted at the base package. The structure follows Domain-Driven Design with Spring Modulith conventions, adapted for server-rendered web with page and fragment subpackages. Use actual module names from the context. Read references/modulith-patterns.md for the detailed module layout and inter-module communication rules.

#### 6. Security Configuration (conditional — include only if Auth != none) If Auth = Keycloak: OAuth2 Client setup with Keycloak OIDC, SecurityFilterChain with oauth2Login(), KeycloakGrantedAuthoritiesMapper to extract roles from ID token claims, CSP nonce filter, automatic redirect to Keycloak login, and public endpoint configuration. Session-based (Spring default). Read references/security-patterns.md for the full security architecture.

If Auth = form: Spring Security form login configuration, SecurityFilterChain with formLogin(), UserDetailsService implementation backed by the selected database, password encoding with BCrypt, role-based access control, CSP nonce filter, and remember-me support.

#### 7. Layouts (JTE) Three layouts: MainLayout (header/sidebar/footer for authenticated pages), AuthLayout (split-screen for login), ErrorLayout (centered error display). Each with Java view model class and JTE template. Layout structure should match the mockup shell/partials. Note: If Auth = none, AuthLayout may be omitted or simplified.

#### 8. Fragments Reusable JTE fragments: HeaderFragment, SidebarFragment, FooterFragment, ThemeSwitcherFragment. Each with view model and template. Sidebar must include the actual navigation items from the mockup sidebar files, organized per role.

#### 9. UI Components (Tailwind-based) Pure Tailwind CSS components (no DaisyUI): Alert, Avatar, Badge, Breadcrumb, Button, Card, Drawer, Dropdown, FormControl/Input/Select/Textarea/Checkbox/Radio/Toggle, Loading, Menu, Modal, Navbar, Pagination, Progress, Stats, Steps, Table, Tabs, Toast, Tooltip. Each with Java view model and JTE template using Tailwind utility classes.

#### 10. Frontend JS Structure Alpine.js stores (toast, modal, drawer, theme), htmx extensions (loading states, toast errors), ES module organization, and CSP nonce integration.

#### 11. Styles Tailwind CSS organization: base (reset, typography), components (buttons, forms, cards, tables, modals, navigation), layouts (sidebar, navbar, footer responsive), themes (light/dark CSS custom properties). Use the actual design tokens (colors, fonts) from MOCKUP.html.

#### 12. Authentication Pages (conditional — include only if Auth != none) If Auth = Keycloak: No custom login page needed. Spring Security automatically redirects unauthenticated users to Keycloak login. Keycloak handles the login form, social login providers, and credential management. After successful authentication, Keycloak redirects back with an authorization code which Spring exchanges for tokens.

If Auth = form: Login page with email/password form, CSRF token, remember-me checkbox. Logout endpoint. Optionally a registration page if user self-registration is desired.

#### 13. Data Access (conditional content varies) If Database = MongoDB: MongoDB repositories per module with Spring Data MongoRepository. DTO mapping via MapStruct. Pagination helpers with PaginationAware mixin. All list endpoints paginated. Use actual collection names from the module model.

If Database = PostgreSQL/MySQL: JPA repositories per module with Spring Data JpaRepository. JPA entities with @Entity/@Table annotations extending BaseEntity. BaseEntity uses @UuidGenerator(style = UuidGenerator.Style.TIME) with @Column(columnDefinition = "CHAR(36)") and private String id. Flyway migration scripts. DTO mapping via MapStruct. Pagination helpers. All list endpoints paginated. Use actual table names from the module model. MySQL UUID rule: All UUID primary keys and foreign keys use CHAR(36) in DDL. In JPA entities, map with columnDefinition = "CHAR(36)" (not length = 36) to pass Hibernate ddl-auto: validate.

If Database = none: In-memory data structures or stubs. The spec should note that a database integration can be added later.

#### 14. Error Handling & Exceptions Base WebApplicationException with status/code/userMessage. GlobalExceptionHandler returning error pages for full requests and toast fragments for htmx requests.

#### 15. Theming Light/dark theme switching via cookie. Server reads cookie and passes theme to layout view model. Alpine store syncs toggle with cookie and data-theme attribute.

#### 16. Pagination Support PaginationRequest, PaginatedResult, and the shared pagination JTE component.

#### 17. Logging Strategy SLF4J with Logback configuration, correlation ID (from JWT if auth is enabled, or request-scoped UUID otherwise), MDC context filter, per-module log levels.

#### 18. Scheduling and Batch Processing (conditional — include only if Scheduling = yes) Quartz scheduler configuration with job store matching the selected database (MongoDB job store via fluidsonic-mirror if MongoDB, JDBC job store if relational, in-memory if no database). If Spring Batch = no, Quartz runs standalone with direct service-call jobs. If Spring Batch = yes, Quartz triggers Spring Batch jobs with chunk-oriented reader/processor/writer processing. If Remote Partitioning = yes (requires Spring Batch), also includes Spring Batch remote partitioning with RabbitMQ for horizontal scaling of batch jobs across multiple worker nodes. Read references/batch-patterns.md for the full scaffolding spec (Quartz-only, Quartz+Batch, and remote partitioning patterns).

#### 19. Event-Driven Architecture Spring Modulith application events for inter-module communication. Read references/modulith-patterns.md for details.

#### 20. Messaging (RabbitMQ Pub/Sub) (conditional — include only if Messaging = yes) Standalone RabbitMQ publisher/consumer services for inter-system communication. Topic exchange for event broadcasting, direct exchange for point-to-point commands. Read references/messaging-patterns.md for the full messaging architecture.

#### 21. MapStruct Usage Per-module MapStruct mapper conventions and mapping flow patterns.

#### 22. Testing Strategy Overview of testing approach — unit tests, module tests, security test utilities (JWT mocking if Keycloak, or @WithMockUser if form login), view model tests.

#### 23. Reporting (JasperReports JRDesign API) (conditional — include only if Reporting = yes) JasperReports infrastructure with fully programmatic report layout via JRDesign API — no .jrxml XML templates. Report layouts are built entirely in Java code using JasperDesign, JRDesignBand, JRDesignStaticText, JRDesignTextField, and JRDesignField classes. Includes ReportDefinition interface with buildDesign() method for modules to implement, ReportDesignHelper utility with static builders for common patterns (A4 portrait/landscape, column headers, detail bands, page footers), ReportService orchestrating compile → fill → export via JRBeanCollectionDataSource, ReportRegistry for auto-discovering and persisting report definitions at startup, report page controller with parameter form and download endpoint, JTE templates for report list and parameter form, multi-format export (PDF via OpenPDF, XLSX via Apache POI, CSV). Modules register reports by creating @Component classes implementing ReportDefinition — each report builds its layout programmatically and calls module services to produce DTOs (never repositories directly), preserving Spring Modulith module boundaries. Read references/jasper-patterns.md for the full reporting architecture.

#### 24. Internationalisation (i18n) (conditional — include only if i18n = yes) Spring's built-in MessageSource localization. Covers: an I18nConfig declaring a ResourceBundleMessageSource (basename i18n/messages, UTF-8, fallback disabled), a CookieLocaleResolver (cookie LOCALE, default from app.i18n.default-locale), and a LocaleChangeInterceptor (param lang) registered through a WebMvcConfigurer; resource bundles under src/main/resources/i18n/ (messages.properties as default/en plus messages_<locale>.properties per supported locale, with module keys namespaced by module slug, e.g. location-information.title); an app.i18n block under the app: namespace (default-locale, supported-locales) bound to an I18nProperties record; a Localizer helper registered as a global @ModelAttribute("t") via @ControllerAdvice (mirroring how appVersion is exposed) so JTE templates resolve keys with ${t.msg("location-information.title")} against LocaleContextHolder.getLocale(); a language switcher JTE fragment (htmx hx-get to ?lang=<locale> with hx-refresh); and bean-validation localization by pointing a LocalValidatorFactoryBean at the same MessageSource. The language switcher is the ONLY locale UI element and MUST NOT be rendered when i18n = no.

What Goes in Each <module>/SPEC.md (Per-Module)

For EACH module from PRD.md and MODEL.md, create a folder named after the module (kebab-case, e.g., location-information/) and generate a SPEC.md inside it.

Each module SPEC.md is a self-contained blueprint that a coding agent can pick up and implement independently (after the shared infrastructure is in place). It must include:

• Header with module name and a back-reference to the root SPECIFICATION.md
• Traceability: User story IDs, NFR IDs, constraint IDs, collection/table names,

mockup screen filenames

• Service interface with methods derived from user stories
• DTOs with fields matching the module model's document/entity fields
• Exception class for module-specific errors
• Document/Entity class with fields from model/<module>/model.md and

model/<module>/schemas.json

• Repository with query methods matching user stories' data access patterns
• MapStruct mapper with mappings matching actual field names
• Service implementation with full CRUD logic
• Page controllers with endpoints matching mockup screens for that module
• Fragment controllers for htmx partial updates
• View models with fields matching what mockup screens display
• JTE templates (list, detail, create, edit, row fragments)
• Message keys (only if i18n = yes) — the module's display strings added to each

i18n/messages_<locale>.properties bundle under a module-slug key namespace (e.g., location-information.title, location-information.create), referenced from this module's JTE templates via ${t.msg("location-information.title")} instead of literal copy

• Complete code samples for every component — continuous and copy-pasteable

Separating UI Layer from Messaging/Async Pipeline:

When a module has BOTH user-facing screens (user stories) AND async processing NFRs (e.g., RabbitMQ message consumption, message validation, ACK publishing, forwarding), the SPEC.md MUST clearly separate these into distinct sections:

1. UI Layer sections — service methods for read-only queries (search, getById,

getHistory), page controllers, fragment controllers, JTE templates. These are driven by user stories (USHMxxxxx).

1. Messaging Pipeline sections — message consumer (@RabbitListener), message

validator, ACK publisher, forward publisher, queue configuration, domain events, processIncomingMessage() service method. These are driven by NFRs (NFRHMxxxx).

MQ listener placement (mandatory). Every @RabbitListener class generated for a module MUST be placed inside that module's internal/ package ({{base}}.{{module}}.internal), package-private, alongside the page/fragment controllers. Shared messaging infrastructure (RabbitMQMessagingConfig, RabbitMQPublisher, MessageConverterConfig, cross-module event/command DTOs) lives in shared/messaging/, but that package must contain no @RabbitListener classes. See references/messaging-patterns.md § "Consumer Service (module-internal)" for the canonical pattern.

This separation enables the implementation orchestrator to track and implement each concern independently. A module may have its UI layer fully implemented while its messaging pipeline remains pending (e.g., waiting for an upstream adapter). The IMPLEMENTATION_MODULE.md checklist tracks each user story and NFR individually, so the status accurately reflects what is done vs. what remains.

See references/spec-template.md Section "Module SPEC.md Template" for the exact template structure.

Changelog Append

After all specification files are successfully generated, append an entry to CHANGELOG.md in the application folder (<app_folder>/CHANGELOG.md):

1. Read <app_folder>/CHANGELOG.md. If it does not exist, create it with:

   # Changelog

   - This file tracks all skill executions by version for this application.
   - The highest version recorded here is the current application version.
   - Skills MUST NOT execute for a version lower than the highest version in this file.

   ---

1. Search for a ## {version} heading matching the current version.
2. If the section exists: append a new row to its table.
3. If the section does not exist: insert a new section after the --- below the context header and before any existing ## vX.Y.Z section (newest-first ordering), with a new table header and the first row.
4. Row format: | {YYYY-MM-DD} | {application_name} | specgen-spring-jpa-jtehtmx | {module or "All"} | Generated Spring Boot web application technical specification |
5. Never modify or delete existing rows.

Output Format

The generated specification is a folder of files, not a single document:

<app_folder>/context/specification/
├── SPECIFICATION.md                    ← Root: TOC + shared/application-level specs
├── <module-1>/
│   └── SPEC.md                         ← Module blueprint (self-contained)
├── <module-2>/
│   └── SPEC.md
├── <module-N>/
│   └── SPEC.md

• Root folder: <app_folder>/context/specification/
• Root file: SPECIFICATION.md — contains the Table of Contents (with links to

each module's SPEC.md), all shared infrastructure sections (1–9), and all application-level cross-cutting sections (10–24, excluding module blueprints)

• Module folders: One folder per module from PRD.md, named in kebab-case

(e.g., location-information/, corridor/, employer/)

• Module files: Each SPEC.md is self-contained with full code samples for that

module's module — service, DTOs, document/entity, repository, mapper, controllers, views, and JTE templates

Design goal: A coding agent can implement the application module-by-module:

1. First pass: implement shared infrastructure from SPECIFICATION.md
2. Subsequent passes: pick up each <module>/SPEC.md independently and implement it

Sample code is mandatory. Every component described in any spec file must include a complete, self-explanatory code sample inside a fenced code block (Java, JTE, JavaScript, YAML, or HTML as appropriate). The sample code must be continuous (not fragments stitched with "// ..." gaps) and usable as a direct reference by a coding agent or developer. A section that only has bullet-point descriptions without accompanying code is incomplete. The code samples in references/spec-template.md and the other reference files show the expected level of detail — reproduce that level in the generated spec, substituting the actual project values from the context files.

Constraints

These constraints are non-negotiable. Every code sample in the generated spec must follow them. If any constraint is violated, the spec is incorrect.

Universal Constraints (Always Apply)

Do NOT use Lombok's `@Data` annotation — anywhere. Use @Getter and @Setter explicitly on every class that needs accessors. @Data generates equals(), hashCode(), and toString() which cause problems with persistent documents/entities, lazy-loaded collections, and circular references. The spec must use @Getter @Setter consistently.

Do NOT use field injection. Every Spring-managed bean must use constructor injection. Use Lombok's @RequiredArgsConstructor with private final fields. Never use @Autowired on fields. This applies to services, controllers, configurations, filters, and any other Spring component.

Do NOT use DaisyUI. All UI components must be built with pure Tailwind CSS utility classes. No DaisyUI plugin, no DaisyUI class names (no btn, card, navbar, alert, badge, modal, drawer, menu, dropdown, tabs, toast, tooltip, table, stats, steps, progress, avatar, breadcrumbs, swap, join, indicator, input, select, textarea, checkbox, radio, toggle, range, file-input, form-control, etc.). Build equivalent components using Tailwind utility classes only.

Use MapStruct for all object mapping. Define MapStruct mappers per module with @Mapper(componentModel = "spring"). Avoid handwritten mapping beyond simple passthroughs.

URL routing must be module-based, NOT role-prefixed. Controller @RequestMapping paths use only the module name (e.g., /corridor, /quota). Never embed role names in URL paths (e.g., NOT /hub_administrator/corridor). Role-based access control is enforced via @PreAuthorize annotations on controllers, derived from the mockup's role folder structure. Fragment controllers use /{module}/fragments (e.g., /corridor/fragments), NOT /api/content/{role}/{module}.

All data listings must be paginated. Every page and fragment that renders a list must use the shared pagination component. Never render unbounded lists.

Use real module data from context files. The spec must use actual field names, collection names, module names, and user stories from the context files. Generic placeholder names (fieldOne, module1, etc.) are only acceptable in the shared infrastructure sections (layouts, components, error handling) — never in modules.

Alpine.js v3 requires `'unsafe-eval'` in CSP `script-src`. Alpine.js v3 uses new Function() for expression evaluation. The CspNonceFilter MUST include 'unsafe-eval' in the script-src directive alongside the nonce. Without this, Alpine.js expressions will be silently blocked and all x-data, x-show, @click etc. will fail. The CSP font-src must also include data: for bundled font data URIs.

Tailwind CSS v4 requires an explicit `@source` directive to scan JTE templates. In app.css, add @source "../../../jte"; after the @import "tailwindcss"; line. Without this, Tailwind v4's automatic content detection only scans the frontend src/ directory and misses all utility classes used in JTE template files, resulting in a minimal CSS output missing most styling.

Do NOT generate UI elements without backend support. Every interactive UI element in the spec (buttons, dropdowns, toggles) must be backed by a requirement in PRD.md or a functional backend endpoint. Specifically: do NOT include a locale/language switcher unless i18n = yes (i.e., PRD.md has internationalisation requirements). When i18n = yes, the language switcher and its backing MessageSource/LocaleResolver infrastructure are specified in Section 24 (Internationalisation); when i18n = no, omit all of it. Do NOT include features "for future use" — they will be non-functional and confuse users.

HTMX navigation in sidebar and header fragments MUST use direct module URLs, NOT /api/content/... prefixed paths. The pattern is: hx-get="${item.getHref()}" with hx-select="#content-area" and hx-target="#content-area" to extract the content area from the full page response. Never use /api/content/ as a URL prefix — this pattern creates a dependency on a non-existent content API and causes navigation failures.

Sidebar navigation must be filtered by the current user's roles. The LayoutService (or equivalent) that builds NavItem lists MUST check the authenticated user's roles and only include menu items that the user is authorized to access. The role-to-menu mapping is derived from the mockup sidebar files (each role folder defines which menu items that role sees). Never show all menu items to all users — this exposes unauthorized functionality in the UI.

Conditional Constraints

If Auth = Keycloak:

• Use OAuth2 Client (not Resource Server). The application uses oauth2Login() with

Authorization Code flow. After successful login, the OidcUser principal is stored in the HTTP session (Spring Security default session management).

• Do NOT manage users and roles in the application. The application does not create,

update, or delete users or roles. Keycloak (the external identity provider) owns all user and role management. The application only reads roles from the OIDC ID token to enforce access control. There are no user registration endpoints, no role assignment endpoints, and no user entity/document in the database.

If Auth = form:

• The application manages its own user authentication. A UserDetailsService backed by

the selected database provides user lookup. Passwords must be encoded with BCrypt. Session management uses Spring Security defaults (not stateless).

If Scheduling = yes and Database = MongoDB:

• Use `io.fluidsonic.mirror` for Quartz MongoDB job store. Instead of the default

in-memory or JDBC job store, Quartz is configured to persist job metadata in MongoDB using the fluidsonic-mirror-quartz library.

If Scheduling = yes and Database = PostgreSQL/MySQL:

• Use the built-in Quartz JDBC job store with the selected relational database.

If Messaging = yes:

• Queue and exchange names for pub/sub messaging (app.*) must not collide with batch

partition queues (batch.partition.*). Use distinct prefixes for each concern.

If Reporting = yes:

• Report definitions must call module services to produce DTOs, never repositories

directly. This preserves Spring Modulith module boundaries.

• Report layouts are built programmatically using JRDesign API — no .jrxml XML templates.

Each ReportDefinition implements buildDesign() to construct the JasperDesign in Java.

• Report DTO field names must match JRDesignField names declared in buildDesign()

exactly (JavaBean getter convention).

• Use ReportDesignHelper utility methods for common layout patterns (title bands, column

headers, detail bands, page footers) to keep report implementations concise.

• Report implementations must be @Component beans implementing ReportDefinition so

that ReportRegistry auto-discovers them at startup.

If Scheduling = yes and Remote Partitioning = yes: -