Claude Code Setup

A battle-tested Claude Code configuration with custom hooks, 7 agents, 25 commands, 30+ skills — designed to work alongside oh-my-claudecode (OMC) plugin.

Built on top of configurations and ideas from:

• everything-claude-code — Anthropic hackathon winner's complete config collection
• andrej-karpathy-skills — Karpathy's coding guidelines as Claude Code skills
• claude-code-tips — Practical tips and the dx plugin

Architecture

This repo provides custom configurations that complement the OMC plugin:

Layer	Source	What It Provides
OMC Plugin	oh-my-claudecode@omc	18 base agents, 27 skills (team, autopilot, ralph, ultrawork, etc.), 11 hook events, CLAUDE.md orchestration prompt
This Repo	claude-code-setup	7 custom agents, 8 custom hooks, 7 rules, 25 commands, 30+ skills, MCP templates

~/.claude/
├── settings.json              # Permissions, env vars, plugins (incl. OMC)
├── agents/                    # 7 custom + 18 OMC agents (merged)
├── rules/                     # 7 global rules (loaded into system prompt)
├── hooks/
│   └── hooks.json             # 8 custom hooks (2 events) — OMC hooks are separate
├── commands/                  # 25 slash commands
├── scripts/
│   ├── hooks/ensure-coauthor.sh  # Git commit Co-Authored-By enforcement
│   ├── *.sh                      # Utility scripts
│   └── ci/                       # 5 CI validation scripts
└── skills/                    # 30+ skill packages (merged with OMC plugin skills)

Quick Start

# 1. Clone and install custom config
git clone https://github.com/YOUR_USERNAME/claude-code-setup.git
cd claude-code-setup
bash install.sh

# 2. Install OMC plugin (in a Claude Code session)
/plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
/plugin install oh-my-claudecode

# 3. Install OMC CLI
npm install -g oh-my-claude-sisyphus@latest
omc install

Use --dry-run to preview changes without modifying anything:

bash install.sh --dry-run

What's Included

Custom Agents (7)

Language/tool specialists that have no OMC equivalent:

Model	Agent	Purpose
Sonnet	e2e-runner	Playwright E2E test generation and execution
Sonnet	refactor-cleaner	Dead code detection (knip, depcheck, ts-prune)
Sonnet	python-reviewer	PEP 8, type hints, ruff/mypy/bandit checks
Haiku	database-reviewer	PostgreSQL optimization, schema review, RLS
Haiku	doc-updater	Documentation and codemap generation
Haiku	go-reviewer	Idiomatic Go, concurrency, error handling review
Haiku	go-build-resolver	Go build/vet error resolution

OMC provides 18 base agents (architect, planner, executor, code-reviewer, etc.) via the plugin.

Custom Hooks (8 hooks, 2 events)

These are custom workflow hooks that complement OMC's orchestration hooks:

Event	Hook	Function
PreToolUse	dev server block	Block npm run dev etc. outside tmux (exit 1)
PreToolUse	tmux reminder	Warn when long-running commands run outside tmux
PreToolUse	ensure-coauthor	Block git commit without Co-Authored-By trailer (exit 2)
PreToolUse	git push review	Reminder before git push
PreToolUse	md file block	Block creation of random .md/.txt files
PostToolUse	PR URL logger	Log PR URL after gh pr create
PostToolUse	Prettier format	Auto-format JS/TS files after edits (async)
PostToolUse	TypeScript check	Run tsc --noEmit after TS edits (async)

OMC provides orchestration hooks (keyword detection, context safety, session management, etc.) via the plugin — no overlap with these custom hooks.

Rules (7)

Rule	Focus
coding-style.md	Immutability, file organization (<800 lines), error handling
git-workflow.md	Conventional commits, PR workflow, feature implementation
karpathy-guidelines.md	Think before coding, simplicity first, surgical changes
performance.md	Model selection strategy, context window management
security.md	Secret management, input validation, OWASP checklist
testing.md	80% coverage minimum, TDD workflow, test types
mcp-priority.md	SearXNG + Crawl4AI local stack preferred over cloud APIs

Commands (25)

Command	Description
/plan	Requirements analysis and implementation plan
/tdd	Enforce TDD (RED, GREEN, REFACTOR)
/code-review	Code review with agent
/build-fix	Fix build errors
/test-coverage	Coverage analysis
/e2e	E2E test generation (Playwright)
/refactor-clean	Dead code cleanup
/checkpoint	Save checkpoint
/verify	Run verification
/orchestrate	Multi-agent orchestration
/sessions	Session management
/learn	Extract patterns
/skill-create	Create skill from patterns
/evolve	Evolve instincts into skills
/python-review	Python code review
/go-review / /go-build / /go-test	Go workflows
/update-codemaps / /update-docs	Documentation
/eval	Session evaluation
/setup-pm	Package manager detection
/instinct-*	Instinct management (status, export, import)

Skills (30+)

Web: frontend-patterns, backend-patterns, coding-standards

Python: python-patterns, python-testing, django-patterns, django-security, django-tdd, django-verification

Go: golang-patterns, golang-testing

Java/Spring: springboot-patterns, springboot-security, springboot-tdd, jpa-patterns, java-coding-standards

Database: postgres-patterns, clickhouse-io

Security: security-review (includes cloud infrastructure security)

Workflow: tdd-workflow, verification-loop, eval-harness, iterative-retrieval, strategic-compact

Learning: continuous-learning-v2 (instinct-based system with observer agent and CLI)

MCP Servers

Global (16 servers, configured in ~/.claude.json)

Server	Type	Description
github	HTTP	GitHub Copilot MCP (PRs, issues, repos)
firecrawl	stdio	Web scraping — firecrawl_agent only (500 credits limited)
supabase	HTTP	Supabase database, auth, storage
vercel	HTTP	Vercel deployment
railway	stdio	Railway deployment
cloudflare-*	HTTP	Docs, Workers builds, bindings, observability
clickhouse	HTTP	ClickHouse analytics
magic	stdio	Magic UI components
filesystem	stdio	Filesystem operations
sequential-thinking	stdio	Structured reasoning chains
context7	stdio	Live documentation lookup
c4ai-sse	SSE	Crawl4AI via Docker (localhost:11235) — primary scraping tool
playwright	stdio	Playwright browser automation

MCP Priority: Local Stack First

The mcp-priority.md rule enforces a SearXNG + Crawl4AI local workflow over cloud APIs:

Search  ->  SearXNG (curl localhost:8888)     free, unlimited, local
Scrape  ->  Crawl4AI (c4ai-sse MCP)           free, unlimited, local
Fallback -> WebSearch (built-in) / Firecrawl   rate-limited, costs money

Use config/claude.json.template as a starting point for your ~/.claude.json.

Permissions

The settings.json includes a carefully tuned permission model:

• Allow: Read, Edit, Write, Glob, Grep, WebFetch, WebSearch, and common CLI tools (git, npm, python, etc.)
• Deny: rm -rf /, rm -rf ~, sudo, reading .env files (12 variants)
• Ask: git push, docker commands
• Default Mode: acceptEdits

Plugins

Plugin	Source	Purpose
oh-my-claudecode@omc	GitHub	Multi-agent orchestration (team, autopilot, ralph, ultrawork)
dx@ykdojo	GitHub	Developer experience (clone, half-clone, handoff, gha)
clangd-lsp	claude-plugins-official	C/C++ language server

Customization

• Add agents: Create .md files in agents/ with frontmatter (name, description, tools, model)
• Add rules: Create .md files in rules/ — they're loaded into every session's system prompt
• Add commands: Create .md files in commands/ — accessible via /command-name
• Add hooks: Edit hooks/hooks.json — custom hooks run alongside OMC plugin hooks
• Add skills: Create directories in skills/ with a SKILL.md file

License

MIT