fan-out

Coordinate parallel OST solution exploration across multiple agents with fan-out and fan-in orchestration.

Updated Apr 7, 2026
One-click install
npx skills add https://github.com/haabe/tic-tac-toe --skill fan-out-haabe
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: fan-out
Source: https://github.com/haabe/tic-tac-toe/tree/main/.claude/skills/fan-out
Command: npx skills add https://github.com/haabe/tic-tac-toe --skill fan-out-haabe

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Enable parallel exploration of multiple OST solutions by fan-out and fan-in orchestration to efficiently identify the best options and document decisions.

Core Features & Use Cases

  • Lead agent identifies parallelizable opportunities and plans work items
  • Fan-Out (spawn workers) distributes tasks with read-only context, isolation, and time bounds
  • Fan-In (collect results) aggregates findings, updates scores, runs bias checks, and logs decisions
  • Enforces rules: workers do not modify the canvas or progress diamonds; lead agent performs the bias check
  • Suitable for 2+ competing solutions and multiple research streams needing simultaneous evaluation

Quick Start

Load the canvas opportunities, identify parallelizable solutions, and spawn workers for each solution.

Frequently Asked Questions about fan-out

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

FAQPage Schema
How do I orchestrate parallel exploration across multiple agents?

Parallel exploration is orchestrated by a lead agent that identifies parallelizable solutions, distributes tasks via fan-out, and aggregates findings via fan-in to select winners and log decisions.

When do I need fan-out and fan-in workflow for agent tasks?

You need fan-out and fan-in workflow when evaluating 2 or more competing OST solutions simultaneously, requiring isolated agents to explore each option in parallel before aggregating results.

What is the best way to prevent parallel agents from modifying shared canvas state?

The best way to prevent modification is applying guardrails that restrict worker agents to read-only context, ensuring they cannot modify the canvas or progress diamonds during parallel exploration.

How do I run a bias check on aggregated agent results?

To run a bias check, the lead agent performs the evaluation during the fan-in phase, aggregating collected findings, updating scores, and validating results before logging the final decision.

Does parallel exploration require isolated context for each worker agent?

Yes, parallel exploration requires distributing tasks with read-only context and isolation, establishing time bounds to ensure workers operate independently without interfering with each other.

Can I use this orchestration pattern for multiple research streams needing simultaneous evaluation?

Yes, this orchestration pattern is suitable for multiple research streams needing simultaneous evaluation, coordinating parallel tasks to efficiently identify the best options and document decisions.