arbitrage

Delegate implementation tasks to codex while Claude handles planning and review workflows.

168|2|Updated Jun 12, 2026

Search Tags:#code review #spec-driven #codex #Claude Code #implementation delegation #token arbitrage

Ready to run

npx skills add https://github.com/blader/arbitrage --skill arbitrage-blader

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: arbitrage
Source: https://github.com/blader/arbitrage/tree/main
Command: npx skills add https://github.com/blader/arbitrage --skill arbitrage-blader

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill addresses the inefficiency of using expensive Claude Code tokens for routine coding tasks, instead leveraging free codex quota for implementation.

Core Features & Use Cases

Token Arbitrage: Distributes planning and high-level decision-making to Claude Code, while delegating implementation to codex using free quota.
Spec-Driven Workflow: Requires detailed specifications before implementation to ensure clarity and prevent unnecessary rework.
Visual Validation Loop: Enables visual validation of UI work through codex, with iterative feedback loops for refinement.

Quick Start

To begin, write a specification for your task in 'SPEC.md', then dispatch the implementation to codex using the command: codex exec --cd <worktree> "/goal <one-line objective; details in SPEC.md>"

Frequently Asked Questions about arbitrage

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I reduce Claude Code token usage for routine implementation tasks?▼

Apply token arbitrage by offloading implementation tasks to codex's free quota while reserving Claude Code for planning and judgment work. You reduce expensive token consumption for routine coding by distributing workloads efficiently.

What is token arbitrage in agentic coding workflows?▼

Token arbitrage in agentic coding distributes planning and high-level decisions to Claude Code while delegating implementation to codex. This maximizes efficiency by leveraging free quota for routine coding tasks and iterative feedback loops.

How do I set up a spec-driven workflow to delegate implementation to codex?▼

Set up a spec-driven workflow by creating a SPEC.md file with detailed task requirements, then running `codex exec --cd <worktree> "/goal <one-line objective; details in SPEC.md>"` to dispatch implementation to codex with clear objectives.

Can I use codex for visual validation of UI work during code review?▼

Yes, codex supports visual validation of UI work through iterative feedback loops. Claude Code delegates UI implementation tasks to codex, enabling visual checks and refinement cycles for high-quality code generation during review.

Does Claude Code work with codex for parallel execution in continuous integration?▼

Claude Code integrates with codex for parallel execution in continuous integration by delegating implementation tasks while Claude handles judgment work. This token arbitrage approach ensures high-quality code through iterative feedback and spec-driven workflows.

Why do I need a SPEC.md file before dispatching tasks to codex?▼

A SPEC.md file is required before dispatching tasks to codex because the spec-driven workflow ensures clarity and prevents unnecessary rework. Codex reads the detailed specifications to implement the one-line objective accurately without ambiguity.