codex-config

Production-ready configuration for OpenAI Codex CLI — 33 custom skills, security hooks, and opinionated instructions for professional software development.

What's Included

~/.codex/
├── config.toml              # Main configuration (model, hooks, agents, features)
├── instructions.md          # Global coding standards & rules
├── install.sh               # One-command installer
├── scripts/
│   └── command-validator/
│       └── validate.sh      # PreToolUse security hook
└── skills/                  # 33 custom skills
    ├── code-reviewer/       # Expert code review
    ├── security-reviewer/   # OWASP Top 10 detection
    ├── typescript-reviewer/ # TS type safety & patterns
    ├── build-error-resolver/# Fix build errors fast
    ├── database-reviewer/   # PostgreSQL optimization
    ├── apex/                # APEX methodology workflow
    ├── tdd-workflow/        # Test-driven development
    ├── commit/              # Conventional commits
    ├── create-pr/           # Auto-generated PR descriptions
    └── ... (24 more)

Quick Install

git clone https://github.com/h41k0/codex-config.git ~/.codex-config
cd ~/.codex-config
./install.sh

The installer will:

1. Backup your existing config.toml
2. Install the configuration with paths adjusted to your $HOME
3. Copy all skills to ~/.codex/skills/
4. Set up the security hook

Skills Overview

Agents (auto-triggered)

Skill	Trigger
$code-reviewer	After writing/modifying code
$security-reviewer	Auth, user input, API endpoints
$typescript-reviewer	Any TS/JS code changes
$build-error-resolver	Build fails or type errors
$database-reviewer	SQL, migrations, schemas
$explore-codebase	Need to understand code structure
$explore-docs	Need library documentation
$websearch	Quick web search

Workflows (invoked explicitly)

Skill	Purpose
$apex	Analyze-Plan-Execute-Validate methodology
$code	Explore → Plan → Execute → Validate
$oneshot	Ultra-fast focused implementation
$ultrathink	Deep thinking for complex problems
$tdd-workflow	Test-driven development (80%+ coverage)
$ralph-loop	Autonomous coding loop

Git Operations

Skill	Purpose
$commit	Conventional commit messages
$create-pr	Auto-generate PR description
$merge	Context-aware conflict resolution
$fix-pr-comments	Implement review feedback
$review	Comprehensive PR review

Fix & Cleanup

Skill	Purpose
$fix-errors	Batch fix ESLint + TypeScript errors
$fix-grammar	Fix grammar preserving formatting
$simplify	Review code for reuse and efficiency

Design Patterns & Guides

Skill	Purpose
$api-design	REST API patterns
$backend-patterns	Node.js architecture
$frontend-patterns	React/Next.js patterns
$frontend-design	Production-grade UI design
$deployment-patterns	CI/CD and Docker
$prompt-creator	LLM prompt engineering
$openai-api	OpenAI SDK patterns
$security-review	Security checklist

Security Hook

The validate.sh PreToolUse hook provides:

• BLOCK (rm -rf patterns) — completely denied
• WARN (sudo, kill, dd, mkfs, fdisk, chmod, chown, su, killall) — requires approval

Configuration Highlights

• Model: GPT-5.5 with high reasoning effort
• Sandbox: workspace-write (safe default)
• Multi-agent: 8 concurrent threads, 3 depth max
• Theme: Dark
• Sound: Notification on session end

Instructions (Coding Standards)

The instructions.md file enforces:

• Immutability (no mutation)
• KISS / DRY / YAGNI principles
• 80%+ test coverage with TDD
• Security-first (OWASP, no hardcoded secrets)
• Conventional commits
• TypeScript best practices (no any, Zod validation)
• Code review after every change

Customization

Change model

Edit config.toml:

model = "gpt-5.5"
model_reasoning_effort = "high"  # none, low, medium, high, xhigh

Add a new skill

mkdir -p ~/.codex/skills/my-skill/agents
# Create SKILL.md with frontmatter (name + description) and body
# Create agents/openai.yaml with interface metadata

Disable a skill's auto-trigger

In agents/openai.yaml:

policy:
  allow_implicit_invocation: false

Credits

Ported from a battle-tested Claude Code configuration with 9 agents, 23 skills, 15 rules, and security hooks. Adapted for Codex CLI's TOML/SKILL.md format.

License

MIT