Agent Swarm Workflow

dicklesworthstone/agent_flywheel_clawdbot_skills_and_integrations

This is Jeffrey Emanuel's actual production workflow for running parallel agent swarms on real projects. You get the exact prompts he uses to coordinate multiple Claude instances through NTM (tmux session manager), Agent Mail (message passing), Beads (task graph), and BV (prioritization). The core insight is that agents are fungible generalists, so coordination matters more than specialization. Use this when you're past planning and ready to execute with multiple agents working simultaneously. The prompts cover everything from initial setup to self-review loops to handling context compaction. It's opinionated and specific, which makes it actually useful if you're building something similar.

skills.sh installs: 169
GitHub stars: 73
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add dicklesworthstone/agent_flywheel_clawdbot_skills_and_integrations --skill agent-swarm-workflow --agent claude-code

Installs into .claude/skills of the current project.

Files

Agent Swarm Workflow — Parallel Implementation

Core Insight: Every agent is fungible and a generalist. They all use the same base model and read the same AGENTS.md. Simply telling one it's a "frontend agent" doesn't make it better at frontend.

The swarm is distributed, robust, and self-organizing through Agent Mail and Beads.

Prerequisites

Before starting a swarm:

Comprehensive plan created (see planning-workflow skill)
Polished beads ready (see beads-workflow skill)
AGENTS.md configured with all tool blurbs
Agent Mail server running (am or ~/projects/mcp_agent_mail/scripts/run_server_with_token.sh)
NTM available for session management