Skip to main content

swarm-coordination

Orchestrate multi-agent swarms for complex parallel task execution

Informações da origem

Repositório
ruvnet/agentic-flow
Última atividade na origem
4 de janeiro de 2026 às 20:11
Idioma detectado do SKILL.md
inglês
Estrelas
816
Forks
178

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

Exibindo SKILL.md

SKILL.md
Instruções da origem · Visualização somente leitura
name
swarm-coordination
description
Orchestrate multi-agent swarms for complex parallel task execution
version
1.0.0
invocable
true
author
agentic-flow
capabilities
["swarm_init","agent_spawn","task_orchestration","topology_optimization"]
# Swarm Coordination Skill Coordinate multi-agent swarms for parallel task execution with intelligent topology selection. ## Quick Start ```bash # Initialize a mesh swarm with 5 agents npx agentic-flow@alpha swarm init --topology mesh --agents 5 # Spawn specialized agents npx agentic-flow@alpha swarm spawn --type researcher --name "Research Agent" npx agentic-flow@alpha swarm spawn --type coder --name "Code Agent" # Orchestrate a task npx agentic-flow@alpha swarm orchestrate "Implement feature X with tests" # Check swarm status npx agentic-flow@alpha swarm status ``` ## Topologies ### Mesh (Default) - All agents communicate directly - Best for: Collaborative tasks, code review - Latency: Low - Scalability: Medium ### Hierarchical - Tree structure with coordinator - Best for: Large projects, delegation - Latency: Medium - Scalability: High ### Ring - Sequential communication - Best for: Pipeline processing, CI/CD - Latency: Higher - Scalability: Medium ### Star - Central hub coordinates all - Best for: Simple coordination - Latency: Low - Scalability: Low ## Agent Types | Type | Description | Use Case | |------|-------------|----------| | researcher | Deep analysis | Requirements gathering | | coder | Implementation | Feature development | | tester | Quality assurance | Test creation | | reviewer | Code quality | PR review | | architect | System design | Architecture decisions | | coordinator | Task routing | Complex workflows | ## MCP Tools ```javascript // Initialize swarm mcp__claude-flow__swarm_init({ topology: "mesh", maxAgents: 8 }) // Spawn agent mcp__claude-flow__agent_spawn({ type: "coder", name: "Feature Dev" }) // Orchestrate task mcp__claude-flow__task_orchestrate({ task: "Implement OAuth", strategy: "parallel" }) ``` ## Best Practices 1. **Right-size your swarm**: Start with 3-5 agents 2. **Choose topology wisely**: Match to task structure 3. **Use Claude Code Task tool**: For actual agent spawning 4. **Monitor status**: Check for bottlenecks 5. **Clean up**: Destroy swarm when done
Ver no GitHub