fork

Run the same implementation task in parallel with Claude Code and OpenAI Codex across git worktrees.

2|Updated Mar 18, 2026
One-click install
npx skills add https://github.com/pmatos/skills --skill fork-pmatos
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: fork
Source: https://github.com/pmatos/skills/tree/main/fork
Command: npx skills add https://github.com/pmatos/skills --skill fork-pmatos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrates the same implementation task across two AI models (Claude Code and OpenAI Codex) in parallel, enabling objective comparison and selection of the superior result while keeping work isolated.

Core Features & Use Cases

  • Spawns two isolated git worktrees branching from the same HEAD to run Claude Code and Codex in parallel.
  • Ensures identical prompts and environment for fair comparison and reproducibility.
  • Provides a structured best-of evaluation and optional merge workflow to adopt the winning approach.

Quick Start

Provide the task prompt and I will initialize parallel worktrees, run Claude Code and Codex side-by-side, and return the best-of result.

Frequently Asked Questions about fork

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

FAQPage Schema
How do I run Claude Code and Codex CLI in parallel on the same task?

To run Claude Code and Codex CLI in parallel, you spawn isolated git worktrees from a common HEAD and execute the shared prompt in each. This produces two branches with commits and a diff summary for objective model comparison.

What is the best way to compare AI code generation models safely?

The best way to compare AI code generation models safely is using parallel git worktrees. This isolates Claude Code and OpenAI Codex outputs in separate branches, enabling structured best-of evaluation without risking your main codebase.

Do I need both Claude Code and OpenAI Codex installed to compare implementations?

Yes, you need both Claude Code and OpenAI Codex CLI installed. The parallel dual-model implementation requires both CLIs to execute the shared prompt simultaneously across dedicated git worktrees for a fair comparison.

Can I merge the winning branch after a parallel model comparison?

Yes, you can merge the winning branch after a parallel model comparison. The workflow provides a structured best-of evaluation and an optional merge process to adopt the superior result from the selected git worktree into your main branch.

When should I use parallel worktrees for architectural exploration?

Use parallel worktrees for architectural exploration when you need safe isolation and objective comparison. Running complex prompts through Claude Code and Codex simultaneously validates different solutions before committing to a specific architectural approach.

Does running dual-model code generation require a shared prompt?

Yes, dual-model code generation requires a shared prompt. Providing identical prompts and environments to both Claude Code and Codex CLI ensures fair comparison and reproducible results across the two generated branches.