swarm-mode

Coordinate parallel subtasks in dependency-ordered waves with per-wave verification.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/benediktms/overmind --skill swarm-mode
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm-mode
Source: https://github.com/benediktms/overmind/tree/main/cli/claudecode-plugin/skills/swarm-mode
Command: npx skills add https://github.com/benediktms/overmind --skill swarm-mode

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coordinates parallel subtasks through dependency-ordered waves and a verify/fix loop to ensure safe integration of outcomes.

Core Features & Use Cases

  • Wave-based parallelism: decomposes objectives into waves and dispatches eligible subtasks in parallel within each wave.
  • Integration verification after each wave: runs verification to ensure that all subtasks converge before proceeding to the next wave.
  • Iterative fix loop: on failed verification, re-dispatches only the failing tasks for remediation up to a configurable max_iterations.

Quick Start

Provide a complex objective to the swarm delegate to initiate wave-based parallel execution with per-wave verification.

Frequently Asked Questions about swarm-mode

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

FAQPage Schema
How do I coordinate parallel tasks to solve a complex objective safely?

You can coordinate parallel tasks using wave-based dispatch to decompose objectives into independent subtasks, executing them in dependency-ordered waves and verifying integration before proceeding to the next wave.

What is wave-based parallelism for agent orchestration?

Wave-based parallelism is an orchestration mechanism that dispatches eligible subtasks in parallel within each wave, ensuring all subtasks converge through post-wave verification before subsequent dependent waves execute.

How do I handle failed verification when running parallel agents?

When verification fails, the system triggers an iterative fix loop that re-dispatches only the failing tasks for remediation, retrying up to a configurable maximum number of iterations to ensure safe integration.

When should I use swarm mode for workflow planning instead of sequential execution?

Use swarm mode when your complex objective can be decomposed into independent subtasks requiring eventual integration, allowing dependency-ordered parallel execution to significantly reduce overall processing time.

Can I configure the maximum retry iterations for the fix loop?

Yes, the iterative fix loop operates with a configurable max_iterations parameter, allowing you to define how many times failing tasks are re-dispatched for remediation before the process stops.

Does parallel task orchestration work without external dependencies?

Yes, the swarm-mode orchestration operates with zero external dependencies, relying entirely on its internal wave-based dispatch and verification logic to coordinate complex objectives.