typescript-backend-project-setup

name	typescript-backend-project-setup
description	Sets up NX monorepo for TypeScript backend projects optimized for AI-assisted development. Delegates to NX commands where possible, patches configs as last resort. Triggers on: 'set up typescript backend project', 'create backend project', 'initialize typescript backend', 'create monorepo', or when working in an empty project folder.
version	3.0.9

NX Monorepo TypeScript Backend Project Setup

🚨 DO NOT USE PLAN MODE. This skill IS the plan. Follow the steps exactly as written.

⚠️ Check NX docs for latest conventions: https://nx.dev/docs/getting-started/start-new-project NX evolves quickly. Verify these instructions against current NX best practices before use.

Set up NX monorepo for TypeScript backend projects with maximum type safety, strict linting, 100% test coverage, and AI-optimized project structure.

Contents

1. Phase 1: Define Project Context - Gather requirements
2. Phase 2: Create NX Workspace - Run NX generator
3. Phase 3: Install Dependencies - Add plugins and tools
4. Phase 4: Create Initial Projects - Generate packages and apps
5. Phase 5: Add Claude Code Integration - Copy AI guardrails and docs
6. Phase 6: Enforce Strict Standards - Patch configs
7. Phase 7: Establish Coding Conventions - Add skill content
8. Phase 8: Activate Git Hooks - Enable pre-commit checks
9. Phase 9: Verify Setup - Confirm everything works
10. Phase 10: Document Architecture - Optional interview

When This Activates

• User requests: "set up typescript backend project", "create backend project", "initialize typescript backend", "create monorepo"
• Working in an empty or near-empty project folder
• User asks for backend project scaffolding or boilerplate

Template Location

This skill uses a template located at: typescript-backend-project-setup/template/

The template contains only files NX cannot create: Claude Code integration, documentation structure, and git hooks.

Before starting, ask the user for the full path to the claude-skillz repository so you can locate the template.

Setup Procedure

Phase 1: Define Project Context

Ask the user:

1. Workspace name - What should this monorepo be called? (lowercase, hyphens ok)
2. Domain description - Brief description of what this project does
3. Claude-skillz path - What is the full path to the claude-skillz repository on your system?
4. Target directory - Where should the project be created? (defaults to current directory)
5. Initial packages - List any publishable packages to create (e.g., "query, builder, cli")
6. Initial apps - List any applications to create (e.g., "api, docs")

Phase 2: Create NX Workspace

Priority: Commands > Installs > Patch files (last resort)

Run the NX workspace generator:

npx create-nx-workspace@latest [workspace-name] --preset=ts --pm=pnpm --nxCloud=skip --interactive=false

This creates:

• nx.json - NX configuration
• tsconfig.base.json - Base TypeScript config
• package.json - Root package with NX scripts
• pnpm-workspace.yaml - Workspace definition
• .gitignore - Standard ignores

Patch .gitignore - Add test-output:

Add test-output to .gitignore (vitest coverage output):

test-output

Checkpoint: Verify nx report shows NX version.

Phase 3: Install Dependencies

Add testing and code quality tools:

# Add NX plugins
nx add @nx/vitest
nx add @nx/eslint
nx add @nx/node  # Required for creating applications

# Install testing dependencies
pnpm add -D vitest @vitest/coverage-v8

# Install ESLint dependencies (required for strict config)
pnpm add -D typescript-eslint @nx/eslint-plugin eslint-plugin-functional

# Install git hooks
pnpm add -D husky lint-staged

Adding @nx/vitest. Provides integrated test runner with coverage reporting.

Adding @nx/eslint. Provides consistent linting across all projects.

Adding @nx/node. Required for creating Node.js applications.

Adding husky and lint-staged. Provides pre-commit verification gate.

Phase 4: Create Initial Projects

If user specified packages in Phase 1, create them:

# For each package (publishable library with vitest)
nx g @nx/js:library packages/[pkg-name] --publishable --importPath=@[workspace-name]/[pkg-name] --bundler=tsc --unitTestRunner=vitest

If user specified apps in Phase 1, create them:

# For each app (node application - vitest NOT supported, use none)
nx g @nx/node:application apps/[app-name] --unitTestRunner=none

🚨 IMPORTANT:

• @nx/js:library supports --unitTestRunner=vitest
• @nx/node:application only supports --unitTestRunner=jest|none (NOT vitest)

After creating projects, run nx sync to update TypeScript project references.

Phase 5: Add Claude Code Integration

Copy template files (only what NX can't create):

Claude Code Integration:

cp -r [claude-skillz-path]/typescript-backend-project-setup/template/CLAUDE.md [target-directory]/
cp -r [claude-skillz-path]/typescript-backend-project-setup/template/AGENTS.md [target-directory]/
cp -r [claude-skillz-path]/typescript-backend-project-setup/template/.claude [target-directory]/

Adding CLAUDE.md. Provides AI context, commands, and project conventions.

Adding .claude/settings.json. Provides permission guardrails and hook configuration.

Adding .claude/hooks/block-dangerous-commands.sh. Prevents destructive git operations (--force, --hard, --no-verify).

Documentation Structure:

cp -r [claude-skillz-path]/typescript-backend-project-setup/template/docs [target-directory]/
cp [claude-skillz-path]/typescript-backend-project-setup/template/repository-setup-checklist.md [target-directory]/

Adding docs/conventions/. Provides coding standards and workflow documentation.

Adding docs/architecture/. Provides system design and domain terminology templates.

Adding docs/project/. Provides project vision and planning templates.

Git Hooks:

cp -r [claude-skillz-path]/typescript-backend-project-setup/template/.husky [target-directory]/

Adding .husky/pre-commit. Provides pre-commit verification (lint, typecheck, test).

Custom ESLint Rules:

cp -r [claude-skillz-path]/typescript-backend-project-setup/template/.eslint-rules [target-directory]/

Adding .eslint-rules/no-generic-names.js. Custom rule that bans generic names (utils, helpers, service, manager) in filenames and class names.

Make scripts executable:

chmod +x [target-directory]/.claude/hooks/block-dangerous-commands.sh

Replace placeholders in copied files:

Placeholder	Replace With
{{WORKSPACE_NAME}}	User's workspace name
{{WORKSPACE_DESCRIPTION}}	User's domain description
{{DOMAIN_NAME}}	User's workspace name (used as context name in glossary)
{{DOMAIN_DESCRIPTION}}	User's domain description

Files with placeholders:

• CLAUDE.md
• docs/conventions/codebase-structure.md
• docs/architecture/domain-terminology/contextive/definitions.glossary.yml
• docs/project/project-overview.md

Phase 6: Enforce Strict Standards

These patches add our strict standards to NX-generated configs.

Patch nx.json - Add lint dependency to build/test:

Add to targetDefaults.build.dependsOn:

"dependsOn": ["lint", "^build"]

Add to targetDefaults.test:

"dependsOn": ["lint"]

This ensures AI gets immediate lint feedback on any change.

Patch tsconfig.base.json - Add strict TypeScript flags:

Add these to compilerOptions:

{
  "noUncheckedIndexedAccess": true,
  "noImplicitOverride": true,
  "noFallthroughCasesInSwitch": true,
  "noPropertyAccessFromIndexSignature": true,
  "noUnusedLocals": true,
  "noUnusedParameters": true,
  "noImplicitReturns": true,
  "exactOptionalPropertyTypes": true,
  "verbatimModuleSyntax": true
}

Patch eslint.config.mjs - Add strict rules:

IMPORTANT: Completely overwrite eslint.config.mjs with this exact content (do not merge, do not patch - replace the entire file):

import nx from '@nx/eslint-plugin';
import tseslint from 'typescript-eslint';
import noGenericNames from './.eslint-rules/no-generic-names.js';

const customRules = {
  plugins: {
    custom: {
      rules: {
        'no-generic-names': noGenericNames,
      },
    },
  },
};

export default tseslint.config(
  ...nx.configs['flat/base'],
  ...nx.configs['flat/typescript'],
  ...nx.configs['flat/javascript'],
  {
    ignores: ['**/dist', '**/out-tsc', '**/node_modules', '**/.nx', '*.config.ts', '*.config.mjs', '*.config.js', 'vitest.workspace.ts'],
  },
  customRules,
  {
    files: ['**/*.ts', '**/*.tsx'],
    rules: {
      // Custom rule: no generic names
      'custom/no-generic-names': 'error',

      // No comments - forces self-documenting code
      'no-warning-comments': 'off',
      'multiline-comment-style': 'off',
      'capitalized-comments': 'off',
      'no-inline-comments': 'error',
      : ,

      
      : [
        ,
        {
          : ,
          : ,
        },
      ],
      : ,
      : ,

      
      : ,
      : ,
      : ,
      : ,
      : ,

      
      : [, { :  }],

      
      : [
        ,
        {
          : [
            { : [, ], :  },
            { : [, ], :  },
            { : [, ], :  },
            { : [, ], :  },
            { : [, ], :  },
          ],
        },
      ],

      
      : [, { : , : , :  }],
      : [, ],
      : [, ],

      
      : [
        ,
        {
          : ,
          : [],
        },
        {
          : ,
          : [],
          : [, ],
        },
        {
          : ,
          : [],
        },
        {
          : ,
          : [],
          : ,
        },
        {
          : ,
          : [],
        },
        {
          : ,
          : [],
        },
        {
          : ,
          : ,
        },
      ],
    },
  },
  {
    : [, ],
    : {
      : {
        : ,
        : ..,
      },
    },
  }
);

This enforces:

• No generic names - Custom rule bans utils, helpers, service, manager, etc. in filenames and class names
• No let - Only const allowed via no-restricted-syntax
• No type assertions - Fix the types, don't cast
• No generic folders - Bans imports from utils/, helpers/, common/, shared/, core/
• Complexity limits - Max 400 lines, max depth 3, cyclomatic complexity 12
• No inline comments - Forces self-documenting code
• No any types - Anywhere
• Naming conventions - camelCase for variables/functions, PascalCase for types

Patch package.json - Add scripts and lint-staged:

Add these to the root package.json:

{
  "scripts": {
    "build": "nx run-many -t build",
    "test": "nx run-many -t test",
    "lint": "nx run-many -t lint",
    "typecheck": "nx run-many -t typecheck",
    "verify": "nx run-many -t lint,typecheck && nx run-many -t test --coverage",
    "prepare": "husky"
  },
  "lint-staged": {
    "*.ts": ["eslint --fix"]
  }
}

Patch vitest.config.mts files - Add 100% coverage thresholds:

For EACH project created in Phase 4 that has a vitest.config.mts, add thresholds to the coverage block:

coverage: {
  reportsDirectory: './test-output/vitest/coverage',
  provider: 'v8' as const,
  thresholds: {
    lines: 100,
    statements: 100,
    functions: 100,
    branches: 100,
  },
},

This enforces 100% test coverage - tests will FAIL if coverage drops below 100%.

Phase 7: Establish Coding Conventions

Copy content from claude-skillz skills to the docs:

Testing conventions:

• Read: [claude-skillz-path]/writing-tests/SKILL.md
• Write to: docs/conventions/testing.md

Adding docs/conventions/testing.md. Provides test naming, assertion patterns, and edge case checklists.

Software design conventions:

• Read: [claude-skillz-path]/software-design-principles/SKILL.md
• Write to: docs/conventions/software-design.md

Adding docs/conventions/software-design.md. Provides object calisthenics, fail-fast, and dependency inversion patterns.

Phase 8: Activate Git Hooks

Initialize husky, then overwrite the default pre-commit with our version:

cd [target-directory]
npx husky init
# husky init creates a default pre-commit - overwrite it with ours:
cp [claude-skillz-path]/typescript-backend-project-setup/template/.husky/pre-commit .husky/pre-commit

Phase 9: Verify Setup

# Check NX is working
nx report

# View empty workspace
nx graph

Review the repository-setup-checklist.md and ensure all items are checked.

Checkpoint: All verification commands pass. Ready for first commit.

Phase 10: Document Architecture (Optional)

Offer to interview the user to fill in placeholder content:

Architecture Overview (docs/architecture/overview.md):

• "What systems does this interact with?"
• "Who are the primary users?"
• "What are the main technical components?"

Domain Terminology (docs/architecture/domain-terminology/contextive/definitions.glossary.yml):

• "What are the key terms in this domain?"
• "How would you define [term] to a new team member?"

Project Overview (docs/project/project-overview.md):

• "What problem are you solving?"
• "Who are the users and what do they need?"
• "What are your non-negotiable principles?"
• "What are the project phases?"

---

Summary

This skill creates an NX monorepo using a command-first approach:

1. create-nx-workspace for foundation
2. nx add for plugins
3. pnpm add for dependencies
4. Copy template files (only what NX can't create)
5. Patch configs (last resort)
6. Verify setup

The result provides:

• Maximum type safety - strict tsconfig, TypeScript project references
• Strict linting - no comments, naming conventions, no mutation
• 100% test coverage - with sensible excludes
• NX orchestration - caching, affected commands, dependency graph
• AI guardrails - protected configs, blocked dangerous commands
• Scalable structure - apps and packages pattern with workspace:* dependencies

For adding projects after setup, see docs/conventions/codebase-structure.md.

---

Verification Mode

Trigger: "verify typescript setup", "check project setup", "audit monorepo config"

Use this to verify an existing repo has all required configurations.

Step 1: Discover all projects

find . -name "package.json" -not -path "*/node_modules/*" -not -path "*/.nx/*"

Step 2: ESLint Config Verification

Read eslint.config.mjs and verify it contains:

• Import of .eslint-rules/no-generic-names.js
• Rule custom/no-generic-names: 'error'
• Rule no-restricted-syntax with VariableDeclaration[kind="let"] selector
• Rule @typescript-eslint/consistent-type-assertions with assertionStyle: 'never'
• Rule no-restricted-imports with patterns for utils, helpers, common, shared, core
• Rules max-lines: 400, max-depth: 3, complexity: 12

If any missing: List what's missing and offer to fix.

Step 3: Vitest Config Verification

For EACH project discovered in Step 1:

1. Check if vitest.config.mts exists in that project directory
2. If exists, read it and verify coverage.thresholds contains:
   • lines: 100
   • statements: 100
   • functions: 100
   • branches: 100

If any project missing thresholds: List which projects are non-compliant and offer to fix.

Step 4: Git Hooks Verification

• .husky/pre-commit contains lint-staged and verify
• package.json has lint-staged config

Step 5: Gitignore Verification

• .gitignore contains test-output on its own line

Step 6: Report

Output a summary:

✓ ESLint: All rules configured
✗ Vitest: 2/6 projects missing coverage thresholds
  - apps/eclair
  - apps/docs
✓ Git Hooks: Configured
✓ Gitignore: test-output ignored

If failures: Offer to fix all issues automatically.