A structured configuration for Claude Code that provides a complete development workflow with specialized agents, commands, and quality controls.

~/.claude/
├── agents/           # Specialized review and research agents
├── commands/         # /dev.* workflow commands
├── skills/           # Reusable skill definitions
├── templates/        # Spec and plan templates
├── hooks/            # Auto-formatting hooks
├── constitution.md   # Guiding principles
├── statusline.sh     # Custom status line
└── settings.json     # Claude Code settings

This config implements a structured development workflow:

/dev.spec  →  /dev.plan  →  /dev.tasks  →  /dev.work
   ↓             ↓              ↓             ↓
 WHAT/WHY       HOW          TODOs        Execute

Specialized agents for different review and research tasks:

Agents are automatically invoked by Claude Code based on their descriptions, or you can request them explicitly (e.g., "use the python-reviewer agent to review this code").

The constitution.md defines advisory principles that guide all /dev.* commands:

1. Test-First Development - Write tests before implementation
2. Measurable Outcomes - Define quantifiable success criteria
3. Simplicity - Prefer the simplest solution
4. Clarification Over Assumption - Ask when unclear
5. Interview-Driven Discovery - Research before specifying
6. Security by Design - Address security during design

These are advisory - commands warn about violations but don't block progress.

Auto-formatting hooks that run after file edits:

• format-markdown.sh - Formats .md files with Prettier
• format-python.sh - Formats .py files with Ruff

Both hooks check for tool availability and skip silently if not installed.

The templates/ directory contains starter templates used by /dev.* commands:

• spec-template.md - Feature specification template (user stories, requirements, security considerations)
• plan-template.md - Technical plan template (architecture, implementation phases, test strategy)

A file-based TODO tracking system organized by git branch:

todos/
├── feature-auth/
│   ├── 001-pending-p1-fix-login.md
│   └── 002-ready-p2-add-tests.md
└── bugfix-query/
    └── 001-ready-p1-optimize-sql.md

Each TODO is a markdown file with YAML frontmatter tracking status, priority, dependencies, and work logs.

# Clone to your preferred location
git clone https://github.com/BonJarber/claude-config.git ~/dotfiles/claude-config

# Symlink individual directories/files
ln -s ~/dotfiles/claude-config/agents ~/.claude/agents
ln -s ~/dotfiles/claude-config/commands ~/.claude/commands
ln -s ~/dotfiles/claude-config/skills ~/.claude/skills
ln -s ~/dotfiles/claude-config/templates ~/.claude/templates
ln -s ~/dotfiles/claude-config/hooks ~/.claude/hooks
ln -s ~/dotfiles/claude-config/constitution.md ~/.claude/constitution.md
ln -s ~/dotfiles/claude-config/statusline.sh ~/.claude/statusline.sh

# Backup existing config
mv ~/.claude ~/.claude.backup

# Clone directly
git clone https://github.com/BonJarber/claude-config.git ~/.claude

Browse the repo and copy individual files/directories that interest you.

The included settings.json enables:

• Extended thinking mode
• Custom status line showing model + git branch
• Post-edit formatting hooks
• Frontend design plugin

Merge with your existing settings or use as-is:

# View current settings
cat ~/.claude/settings.json

# Or merge specific keys into your existing settings

The formatting hooks are optional and require:

• Markdown formatting: npm install -g prettier
• Python formatting: uv tool install ruff (or pip install ruff)

Hooks skip silently if tools aren't available.

Create a new .md file in agents/ with YAML frontmatter:

---
name: my-agent
description: When to use this agent
model: inherit # or: sonnet, opus, haiku
color: blue # terminal color for agent output
tools: Read, Glob, Grep, Edit # optional: restrict available tools
permissionMode: default # or: acceptEdits (auto-accept file changes)
---

You are an expert in...

Create a new .md file in commands/:

---
description: What this command does
argument-hint: "[optional argument description]"
skills: [skill-name] # optional: skills this command uses
allowed-tools: Bash(git:*) # optional: tool restrictions
---

# Command Name

Instructions for Claude...

Use $ARGUMENTS to reference user-provided arguments.

Commands are invoked with /command-name (filename without .md).

This configuration emphasizes:

• Structured thinking - Spec before plan, plan before code
• Quality gates - Specialized reviewers catch issues early
• Simplicity - Solve current problems, not hypothetical ones
• Transparency - TODO tracking with full work logs
• Flexibility - Advisory principles, not rigid rules

MIT - Use however you like.