swarm

Orchestrate large tasks by spawning background Workflows across multiple workers.

Updated Feb 22, 2026
One-click install
npx skills add https://github.com/mbriggsy/ai-learning-journey --skill swarm-mbriggsy
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm
Source: https://github.com/mbriggsy/ai-learning-journey/tree/main/projects/skills/swarm
Command: npx skills add https://github.com/mbriggsy/ai-learning-journey --skill swarm-mbriggsy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This skill enables a lightweight coordinator to decompose complex tasks, launch a background Workflow to execute heavy lifting across multiple workers, and collate only the final results back in the main window, keeping the user interface responsive and focused on supervision.

Core Features & Use Cases

  • Distributed coordination: decomposes tasks into parallel work units, coordinating dozens to hundreds of workers when needed.
  • Workflow-based execution: leverages the Workflow tool to run heavy tasks in background contexts with per-worker configuration (model, effort) and ensures final results are returned in a structured form.
  • Guardrails & verification: supports adversarial verification and strict final-result synthesis to prevent unvetted outcomes.
  • Use Case: a product-manager uses swarm to delegate validation and analysis tasks across multiple agents, then reviews a concise final report while the system handles the internal orchestration.

Quick Start

Swarm an analysis task by delegating the work to a Workflow-based swarm and reviewing the final results produced by the workers.

Frequently Asked Questions about swarm

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

FAQPage Schema
How do I delegate parallel tasks to multiple background agents without blocking my main workflow?

Distributed task coordination decomposes complex work into parallel units, coordinating dozens to hundreds of workers when needed. It leverages background Workflows to execute heavy tasks and collates only structured final results back to the main window.

What is adversarial verification in distributed agent workflows?

Adversarial verification in distributed workflows is a guardrail mechanism that strictly synthesizes and validates final results to prevent unvetted outcomes. It ensures that outputs produced by parallel background workers are checked before being returned to the coordinator.

How do I orchestrate a large analysis task across hundreds of workers?

You orchestrate large tasks by explicitly decomposing them into parallel work units and launching a background Workflow with per-worker configuration like model and effort. The system handles internal orchestration and returns a concise final report.

Can I configure individual models and effort levels for each worker in a parallel workflow?

Yes, you can apply per-worker configuration specifying the model and effort level for each background worker in the swarm. This allows you to allocate resources appropriately based on the complexity of each decomposed task unit.

When should I not use a fire-and-forget swarm for task delegation?

Avoid fire-and-forget swarm delegation when tasks require live steering or continuous interactive intervention. This approach is designed for scale and parallelism where heavy lifting runs autonomously in background contexts without real-time user input.

How does a coordinator track background workflow execution for auditability?

The coordinator tracks execution for auditability by maintaining a durable war diary that records internal orchestration steps, explicit task decomposition, and adversarial verification checks taken by the background workers.