swarm

Bootstrap CAST agent teams from swarm config files with isolated git worktrees.

8|1|Updated Mar 20, 2026
One-click install
npx skills add https://github.com/ek33450505/claude-agent-team --skill swarm-ek33450505
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm
Source: https://github.com/ek33450505/claude-agent-team/tree/main/skills/swarm
Command: npx skills add https://github.com/ek33450505/claude-agent-team --skill swarm-ek33450505

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill automates the tedious, error-prone process of creating and managing multi-agent CAST swarms by provisioning isolated git worktrees, seeding each teammate with role identity, applying CAST quality gates, and recording the swarm to cast.db for observability.

Core Features & Use Cases

  • Parallel teammate orchestration: Spawn multiple agents in parallel with isolated worktrees and role-specific prompts to accelerate code, review, or research tasks.
  • Operational safety & lifecycle: Seed teammates with spawn preambles, summarize swarm manifests, provide merge and teardown commands, and surface when all teammates reach Status: DONE.
  • Observability & audit: Emit a JSON manifest and log swarm events to cast.db so teams can trace roles, models, merge strategies, and worktree locations.
  • Use Cases: Rapidly prototype a fullstack feature with parallel code authors, run focused multi-reviewer security reviews, or spin up research teams for comparative investigations.

Quick Start

Invoke the swarm skill with a named team config and task, for example: /swarm fullstack-team "implement dark mode toggle".

Frequently Asked Questions about swarm

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

FAQPage Schema
How do I orchestrate multi-agent CAST teams for parallel code generation?

Multi-agent CAST orchestration is handled by bootstrapping agent teams from a named swarm configuration file. This spawns parallel teammates within isolated git worktrees, seeding each with role-specific prompts to accelerate code generation tasks.

What is a git worktree and why use it for multi-agent development?

A git worktree provides an isolated branch checkout, used here to isolate parallel agent teammates. This isolation prevents file conflicts during multi-agent code generation, allowing simultaneous work on different aspects of a feature.

Do I need YAML parsing support to bootstrap a swarm?

Yes, YAML parsing support is required to read the named swarm configuration file. The bootstrap process parses this configuration to provision teammates, define merge strategies, and establish worktree paths.

How are multi-agent swarm events logged for observability?

Multi-agent swarm events are logged to cast.db for full observability and audit. The system also emits a JSON manifest detailing swarm members, worktree paths, roles, and merge strategies to trace activity.

Can I run multi-reviewer security pipelines using isolated agent worktrees?

Yes, you can run focused multi-reviewer security pipelines by spawning parallel agents. Each agent operates in an isolated git worktree with a specific role to review code securely before applying merge strategies.

What is the best way to tear down parallel agent worktrees after completion?

The best way to tear down worktrees is using the provided merge and teardown commands after all teammates reach Status: DONE. The system summarizes swarm manifests and manages the lifecycle safely.