fan-out

Orchestrate parallel agent tasks with worktree isolation and preflight validation.

489|254|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/me2resh/apexyard --skill fan-out-me2resh
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: fan-out
Source: https://github.com/me2resh/apexyard/tree/main/.claude/skills/fan-out
Command: npx skills add https://github.com/me2resh/apexyard --skill fan-out-me2resh

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents slow, serial execution by enabling you to spawn multiple specialized agent calls concurrently while avoiding common concurrency pitfalls like file collisions and missing active tickets.

Core Features & Use Cases

  • Parallel agent execution: Decomposes a request into independent tasks and runs each with the most suitable agent type.
  • Worktree isolation: Uses per-agent git worktrees so code-writing agents don’t race in the same workspace.
  • Safety guardrails: Refuses fan-out when tasks need an active ticket or when tasks share write targets or have sequential dependencies.
  • Result collection + merge-back: Summarizes outcomes and guides ordered merge-back for worktree branches.

Quick Start

Run /fan-out with a comma-separated list of 2-5 independent tasks so the system spawns specialized agents in parallel and reports results.

Frequently Asked Questions about fan-out

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

FAQPage Schema
How do I run multiple agent tasks in parallel without causing file collisions?

You can run parallel agent tasks safely by using worktree isolation, which assigns each code-writing agent a separate git worktree to prevent write-target collisions and workspace race conditions.

What is the best way to execute independent codebase refactors concurrently?

The best way to execute independent codebase refactors concurrently is to decompose the request into file-independent and context-independent units, allowing specialized agents to execute in parallel with deterministic concurrency constraints.

When should I not use parallel agent execution for software development tasks?

You should not use parallel agent execution when tasks require an active ticket, share write targets, or have sequential dependencies, as the system applies safety guardrails and refuses fan-out under these conditions.

How do I merge back branches created by isolated parallel agents?

To merge back branches created by isolated parallel agents, the system summarizes task outcomes and guides ordered merge-back procedures for the worktree branches to ensure safe integration of parallel work.

Can I spawn multiple specialized agents from a single user request?

Yes, you can spawn multiple specialized agents from a single user request by issuing a comma-separated list of two to five independent tasks, which the system executes concurrently using a single-message multi-Tool-call constraint.