feat: add Claude Code setup

.claude/agents/stack-maintainer.md

@@ -0,0 +1,50 @@
+# Stack Maintainer Agent
+
+You are the stack maintainer agent. Your role is to protect the mergeability and security of the Devkit Node stack.
+
+## Responsibilities
+
+### 1. Protect mergeability
+
+- **Prevent risky renames**: Core stack files should stay in their original locations
+- **Avoid structure breakage**: Don't move modules, change folder structures, or rename core files under `lib/` or `config/`
+- **Stable paths**: Ensure downstream projects can merge updates cleanly
+- **Flag risky changes**: Warn about changes that might cause merge conflicts in `lib/services/`, `lib/middlewares/`, or `config/defaults/`
+
+### 2. Sanity-check for security
+
+- **Secret leakage**: Check for accidentally committed secrets, tokens, or credentials
+- **Broad permissions**: Review permission changes for security risks
+- **Dependencies**: Flag suspicious or risky dependency additions
+- **Env vars**: Ensure sensitive config uses `DEVKIT_NODE_*` env vars, not hardcoded values
+- **Auth bypass**: Watch for changes that weaken JWT/Passport validation or policy middleware
+
+### 3. Verify modularity
+
+- **Cross-module coupling**: Flag unnecessary imports between modules
+- **Layer violations**: Ensure controllers don't call repositories directly (must go through services)
+- **Module boundaries**: Keep logic isolated within `modules/{name}/` — controllers, services, repositories, models, policies, routes, tests
+
+## When invoked
+
+- Review proposed changes briefly
+- Flag any concerns with severity:
+  - 🔴 **Critical**: Must fix (security, breakage)
+  - 🟡 **Warning**: Should review (coupling, patterns, layer violations)
+  - 🟢 **Info**: Good to know (suggestions)
+- Be concise — this is a quick sanity check, not a full audit
+
+## What NOT to do
+
+- Don't run workflows or execute commands
+- Don't implement features
+- Don't write code
+- Keep reviews short and focused
+
+## Example review
+
+```
+🔴 Critical: `.env` file was modified (should be git-ignored)
+🟡 Warning: Controller imports repository directly — must go through service layer
+🟢 Info: Consider extracting this validation schema to `lib/helpers/joi.js` for reuse
+```

.claude/settings.json

@@ -0,0 +1,137 @@
+{
+  "mcpServers": {},
+  "permissions": {
+    "files": {
+      "deny": [
+        ".env",
+        ".env.*",
+        "secrets/**",
+        "**/*.pem",
+        "**/*.key",
+        "**/*.enc",
+        "**/credentials.json",
+        "**/credentials.yml",
+        "**/credentials.yaml",
+        "**/*-key.json",
+        "**/*_rsa",
+        "**/*_dsa",
+        "**/*_ecdsa",
+        "**/*_ed25519"
+      ]
+    },
+    "bash": {
+      "allow": [
+        {
+          "command": "git status",
+          "prompt": "Check git status"
+        },
+        {
+          "command": "git diff",
+          "prompt": "View git diff"
+        },
+        {
+          "command": "git diff *",
+          "prompt": "View git diff for specific files"
+        },
+        {
+          "command": "git log",
+          "prompt": "View git log"
+        },
+        {
+          "command": "git log *",
+          "prompt": "View git log with options"
+        },
+        {
+          "command": "git show *",
+          "prompt": "Show git commit details"
+        },
+        {
+          "command": "npm start",
+          "prompt": "Start dev server"
+        },
+        {
+          "command": "npm run debug",
+          "prompt": "Start debug server"
+        },
+        {
+          "command": "npm run prod",
+          "prompt": "Start production server"
+        },
+        {
+          "command": "npm test",
+          "prompt": "Run tests"
+        },
+        {
+          "command": "npm run test:watch",
+          "prompt": "Run tests in watch mode"
+        },
+        {
+          "command": "npm run test:coverage",
+          "prompt": "Generate test coverage"
+        },
+        {
+          "command": "npm run lint",
+          "prompt": "Check code quality"
+        },
+        {
+          "command": "npm run lint:fix",
+          "prompt": "Auto-fix code quality issues"
+        },
+        {
+          "command": "npm run seed:dev",
+          "prompt": "Seed development database"
+        },
+        {
+          "command": "npm run commit",
+          "prompt": "Commit with commitizen"
+        },
+        {
+          "command": "docker-compose up",
+          "prompt": "Start docker-compose services"
+        },
+        {
+          "command": "docker-compose up -d",
+          "prompt": "Start docker-compose in background"
+        },
+        {
+          "command": "docker-compose down",
+          "prompt": "Stop docker-compose services"
+        },
+        {
+          "command": "docker-compose ps",
+          "prompt": "Show docker-compose status"
+        }
+      ],
+      "deny": [
+        {
+          "pattern": "rm -rf *",
+          "reason": "Destructive command"
+        },
+        {
+          "pattern": "sudo *",
+          "reason": "Requires elevated privileges"
+        },
+        {
+          "pattern": "curl * | *sh",
+          "reason": "Arbitrary code execution risk"
+        },
+        {
+          "pattern": "wget * | *sh",
+          "reason": "Arbitrary code execution risk"
+        },
+        {
+          "pattern": "git push *",
+          "reason": "Automatic push not allowed"
+        },
+        {
+          "pattern": "npm publish",
+          "reason": "Package publishing requires explicit approval"
+        },
+        {
+          "pattern": "npm run seed:mongodrop",
+          "reason": "Destructive database operation"
+        }
+      ]
+    }
+  }
+}

.claude/skills/create-module/SKILL.md

@@ -0,0 +1,111 @@
+---
+name: create-module
+description: Create a new feature module by duplicating the canonical `tasks` module template. Use when adding a new module to the application, scaffolding a new domain area from scratch, or generating the boilerplate for a new feature.
+---
+
+# Create Module Skill
+
+Create a new module by copying and renaming the `tasks` template module.
+
+## Prerequisites
+
+- The canonical template module `modules/tasks` must exist
+- You need a name for the new module (kebab-case)
+
+## Steps
+
+### 1. Ask for the module name
+
+Prompt user for the new module name in kebab-case (e.g., `my-feature`, `user-settings`)
+
+### 2. Derive naming conventions
+
+Follow `/naming` for the full reference. Quick summary from the module name (e.g., `my-feature`):
+
+- **kebab-case**: `my-feature` (folder names, file prefixes, routes)
+- **PascalCase**: `MyFeature` (Mongoose model names, class names)
+- **lowerCamelCase**: `myFeature` (variable names, function names, JS exports)
+- **UPPER_SNAKE_CASE**: `MY_FEATURE` (constants)
+
+### 3. Duplicate the module
+
+```bash
+cp -r modules/tasks modules/{new-module-name}
+```
+
+### 4. Rename references
+
+Search and replace the following tokens across the new module:
+
+- `tasks` → `{new-module-name}` (kebab-case)
+- `Tasks` → `{NewModuleName}` (PascalCase)
+- `task` → `{new-module}` (singular kebab-case, if applicable)
+- `Task` → `{NewModule}` (singular PascalCase, if applicable)
+
+Files to check in `modules/{new-module-name}/`:
+
+- File names (controllers, services, repositories, models, schemas, policies, routes, tests)
+- Mongoose model name and collection name
+- Route paths and prefixes
+- Joi validation schemas
+- Policy function names
+- Test descriptions and fixture data
+
+### 5. Show rename plan (if broad changes)
+
+If renaming affects many files, show a brief plan before applying changes.
+
+### 6. Apply renames carefully
+
+Use safe search+replace to avoid false positives:
+
+- Be case-sensitive
+- Match whole words where possible
+- Don't rename unrelated code (e.g., "tasks" in comments about other features)
+
+### 7. Run verify
+
+```bash
+npm run lint
+npm test
+```
+
+### 8. Report results
+
+Provide a summary:
+
+- ✅ Module created at: `modules/{new-module-name}`
+- ✅ Renamed tokens: `tasks` → `{new-module-name}`, etc.
+- ✅ Verification: lint passed, tests passed
+- 📝 Next steps: Customize the schema, update business logic in services. Routes are auto-discovered via glob `modules/*/routes/*.js` — no registration needed
+
+## Module Structure
+
+```
+modules/{new-module-name}/
+├── controllers/
+│   └── {module}.controller.js
+├── services/
+│   └── {module}.service.js
+│   └── {module}.data.service.js
+├── repositories/
+│   └── {module}.repository.js
+├── models/
+│   └── {module}.model.mongoose.js
+│   └── {module}.model.sequelize.js  (optional)
+│   └── {module}.schema.js
+├── policies/
+│   └── {module}.policy.js
+├── routes/
+│   └── {module}.routes.js
+└── tests/
+    ├── {module}.integration.tests.js
+    └── {module}.unit.tests.js
+```
+
+## Notes
+
+- Preserves the layered architecture: controllers → services → repositories → models
+- Follows modularity rules: keeps the module self-contained
+- Policy functions control route authorization — update them for your use case
+- Remove `{module}.model.sequelize.js` if not using SQL

.claude/skills/feature/SKILL.md

@@ -0,0 +1,77 @@
+---
+name: feature
+description: Implement a new feature or modify existing functionality following the project's layered architecture and modularity rules. Use when adding features, modifying existing ones, or ensuring correct isolation within module boundaries.
+---
+
+# Feature Skill
+
+Implement features with strict layer ordering and module isolation.
+
+## Steps
+
+### 1. Identify target module(s)
+
+Before coding, determine:
+
+- Which module(s) does this feature belong to?
+- Default to **ONE module** unless strictly necessary
+- If it spans multiple modules, explain why and minimize coupling
+
+### 2. Apply layer rules
+
+Respect the strict layer order — never skip or reverse it:
+
+```
+Routes → Controllers → Services → Repositories → Models
+```
+
+- **Routes** (`routes/`): Define HTTP endpoints, apply policy middleware
+- **Controllers** (`controllers/`): Handle HTTP req/res, call services, format responses via `lib/helpers/responses.js`
+- **Services** (`services/`): Business logic, call repositories, throw `AppError` for domain errors
+- **Repositories** (`repositories/`): Database access only, use Joi schemas for validation
+- **Models** (`models/`): Mongoose schema definition (`*.model.mongoose.js`) and Joi schema (`*.schema.js`)
+
+### 3. Apply modularity rules
+
+- **Isolate feature logic** inside the module boundary
+- **Avoid cross-module imports** unless absolutely required
+- If shared code is needed:
+  - Place it in `lib/helpers/` or `lib/services/`
+  - Provide **explicit justification** for why it must be shared
+- Keep these inside the module (`modules/{module}/**`):
+  - Controllers, services, repositories, models, schemas, policies, routes, tests
+
+### 4. Use existing helpers
+
+Before writing new utilities, check:
+
+- `lib/helpers/responses.js` — JSend response wrapper
+- `lib/helpers/AppError.js` — Custom error class for domain errors
+- `lib/helpers/errors.js` — Error formatting utilities
+- `lib/helpers/joi.js` — Shared Joi validation helpers
+- `lib/middlewares/policy.js` — Authorization middleware
+- `lib/middlewares/model.js` — Model middleware
+
+### 5. Follow API response format
+
+All API responses must use the JSend wrapper from `lib/helpers/responses.js`:
+
+```js
+// Success
+responses.success(res, 'task list')(data);
+// Error
+responses.error(res, 422, 'Unprocessable Entity', errors.getMessage(err))(err);
+```
+
+### 6. Run verify
+
+```bash
+npm run lint
+npm test
+```
+
+## Notes
+
+- Config is loaded from `config/` and overridable via `DEVKIT_NODE_*` env vars
+- Authentication is handled via Passport JWT middleware — don't reimplement
+- All policies must use `lib/middlewares/policy.js` patterns