Can I customize the agent prompts for different tasks?

Yes, the decomposition phase generates specific agent prompts and task-specific instructions in the prompts/ directory, which can be reviewed and edited before execution.

How does this skill prevent merge conflicts between agents?

The skill uses git worktree isolation and a contract-first approach, ensuring each agent works in a separate environment with clearly defined boundaries and shared interfaces defined upfront.

When should I use the parallel-agents skill instead of standard Claude Code?

Use it when you have a well-defined Tech Spec or PRD that can be broken down into independent modules, allowing multiple agents to work simultaneously rather than sequentially.

What is the Claude Parallel Orchestrator (cpo)?

cpo is a CLI tool included with this skill that handles the initialization, validation, and execution of multiple Claude agents working in parallel within isolated directory structures.

What is the significance of the integration phase?

The integration phase uses the /parallel-integrate command to check contract compliance, run cross-module tests, and generate a report ensuring all parallel workstreams align correctly.

Parallel Agent Orchestrator: Claude Code Skill for Multi-Agent Dev

The Parallel Agent Orchestrator enables developers to scale their output by breaking down complex Product Requirements Documents (PRDs) or Tech Specs into a structured, multi-agent workflow. Utilizing the 'cpo' (Claude Parallel Orchestrator) CLI, this skill automates the creation of isolated git worktrees, enforces contract-first development through shared domain types, and manages execution 'waves' to ensure logical dependency handling. It is designed for high-velocity teams who need to execute large-scale feature development across independent workstreams without the manual overhead of manual task coordination.

Características Principales

01Multi-wave execution sequencing to manage complex task dependency graphs

02Automated integration reporting and post-execution verification workflows

03Contract-first development enforcement with shared API schemas and domain types

04Isolated execution environments using git worktree management to prevent code conflicts

05Automated PRD decomposition into structured, machine-readable task manifests

062 GitHub stars

Casos de Uso

01Scaling a single large feature request across 5+ concurrent Claude Code agents to accelerate delivery

02Managing complex refactoring projects that require simultaneous changes across different layers of the stack

03Decomposing a monolith architecture into isolated, testable micro-services or modules

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jpoutrin/product-forge parallel-agents

For use in Claude.ai and ChatGPT

Download Skill