swarming

Orchestrate parallel worker agents with inbox-driven messaging and file reservation coordination.

Updated Mar 27, 2026
One-click install
npx skills add https://github.com/quanpersie2001/pulse --skill swarming
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarming
Source: https://github.com/quanpersie2001/pulse/tree/main/plugins/pulse/skills/swarming
Command: npx skills add https://github.com/quanpersie2001/pulse --skill swarming

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Swarming ensures multi-agent execution does not stall by continuously tending a coordination runtime, enforcing startup acknowledgments, monitoring worker reports, handling file conflicts, and escalating only when human judgment is required.

Core Features & Use Cases

  • Orchestrator-only role: Launches bounded worker subagents, monitors their inbox-driven lifecycle, and never implements beads directly.
  • Robust monitoring loop: Poll-act-repeat inbox and topic checks, live bead graph verification, silence ladder reminders, and coordinator handoffs when context is high.
  • Coordination primitives: Startup acknowledgments with two-name identity, completion and blocker reports, file reservation coordination, and overseer broadcasts for shared corrections.
  • Use Case: Run a coordinated swarm to implement a phase of feature work where multiple workers must reserve files, report completion, and avoid conflicting edits while the orchestrator ensures continuous progress.

Quick Start

Start a coordinated swarm for the current phase by initializing the coordination runtime, spawning bounded workers with the provided worker prompt template, and keep polling inboxes and the bead graph until execution completes.

Frequently Asked Questions about swarming

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

FAQPage Schema
How do I coordinate parallel worker agents without stalling execution?

Multi-agent execution coordination requires an orchestrator-only role that launches bounded worker subagents, monitors their inbox-driven lifecycle, and handles file conflicts to prevent idle swarms during phase execution.

How does inbox polling work for worker coordination in a multi-agent swarm?

Inbox polling for worker coordination uses a poll-act-repeat loop that continuously checks worker inboxes and topics, applies silence ladder reminders, and verifies the live bead graph until execution completes.

What is file reservation coordination and when do I need it for parallel agents?

File reservation coordination is needed when multiple parallel workers must reserve files, report completion, and avoid conflicting edits while an orchestrator ensures continuous progress throughout the phase execution.

How do I handle file conflicts when multiple worker agents edit the same files?

File conflict resolution during multi-agent execution is handled by the orchestrator through file reservation coordination, ensuring workers avoid conflicting edits and escalating to human judgment only when necessary.

Does this multi-agent orchestrator implement beads directly or just monitor workers?

The orchestrator never implements beads directly; it only launches bounded worker subagents, monitors their startup acknowledgments with two-name identity tracking, and generates coordinator handoffs when context is high.

What's the best way to generate coordinator handoffs when context limits are high?

Coordinator handoffs are generated automatically by the orchestrator during the monitoring loop when context is high, ensuring continuous coordination and preventing idle swarms during phase execution.