TypeScript Monorepo - Claude Code Memory

Project Overview

This is a production-ready TypeScript monorepo template using modern tooling:

• Package Manager: pnpm (with workspace support)
• Runtime: Node.js (see .node-version for current version)
• Build System: tsc / pnpm
• Linting/Formatting: Trunk (manages ESLint, Prettier, and more)
• Testing: Vitest
• CI/CD: GitHub Actions

Quick Commands

pnpm install    # Install all dependencies
pnpm build      # Build all packages
pnpm test       # Run all tests via Vitest
pnpm lint       # Run all linters via Trunk
pnpm format     # Auto-format code via Trunk
pnpm clean      # Clean build artifacts

Code Style

• Use TypeScript for all code
• Follow project ESLint/Prettier configuration (managed via Trunk)
• Use functional programming patterns where appropriate
• Use PascalCase for classes/interfaces, camelCase for functions/variables
• Use kebab-case for file names (e.g., user-service.ts)

Testing

• Write tests in tests/ or alongside source files as *.test.ts
• Use Vitest for unit and integration tests
• Aim for high test coverage on core logic
• Run pnpm test before committing

Git Workflow

• Create feature branches from main
• Run pnpm lint && pnpm test before commits
• Commit messages: type(scope): description (e.g., feat(ui): add new button component)
• Types: feat, fix, docs, style, refactor, test, chore
• Record changes using the manage-changelog skill when changie is available

Architecture

• Source code in packages/* or src/ (depending on package structure)
• Development scripts and shared config in root
• CI/CD workflows in .github/workflows/
• Claude Code configuration in .claude/
• Document significant design decisions as Architecture Decision Records (ADRs) in docs/adr; use the manage-adr skill when adr is available

Common Gotchas

• Always use pnpm instead of npm or yarn
• Trunk manages tool versions hermetically - don't install linters globally
• The pnpm-lock.yaml file is committed for reproducibility - don't gitignore it
• Run trunk install if linters report missing tools

Parallel Task Execution

For large tasks that can benefit from concurrent work:

/parallel-executor Add comprehensive logging to all modules

This decomposes tasks into independent subtasks with file ownership, executes them in parallel, and verifies results.

Available Agents

Available Skills

Configuration Self-Improvement

This project supports Claude Code self-improvement. When you notice repeated mistakes, recurring explanations, or opportunities for automation:

1. Analyze Pattern: Identify if it's a rule (guidance), a hook (mandatory action), a skill (workflow), or an agent (specialized task).
2. Implement Improvement:
   • Rules: Add to CLAUDE.md or .cursor/rules/ for guidance requiring judgment.
   • Hooks: Use PreToolUse or PostToolUse in .claude/settings.json for deterministic, mandatory checks.
   • Skills: Create new entries in .claude/skills/ for reusable complex workflows.
   • Agents: Define new agents in .claude/agents/ for specialized context isolation.
3. Use Subagents: For configuration research or verification, use subagents to isolate context and prevent contamination of the main session.
4. Be Specific & Minimal: Only add rules or skills that provide clear, non-obvious value.

Use the /improve-claude-config skill to orchestrate these changes.

Recent Learnings

• [2026-02-05]: Initial setup of the comprehensive skills reference and self-improvement guidelines.