playbooks

home / skills / a5c-ai / babysitter / crewai-setup

crewai-setup skill

/plugins/babysitter/skills/babysit/process/specializations/ai-agents-conversational/skills/crewai-setup

This skill configures CrewAI agents with roles, goals, tasks, and collaboration patterns to orchestrate reliable multi-agent workflows.

npx playbooks add skill a5c-ai/babysitter --skill crewai-setup

Review the files below or copy the command above to add this skill to your agents.

Files (2)
SKILL.md
1.3 KB
---
name: crewai-setup
description: CrewAI multi-agent orchestration setup for collaborative AI systems
allowed-tools:
  - Read
  - Write
  - Edit
  - Bash
  - Glob
  - Grep
---

# CrewAI Setup Skill

## Capabilities

- Configure CrewAI agents with roles and goals
- Define tasks and expected outputs
- Set up crew orchestration patterns
- Implement agent collaboration workflows
- Configure memory and knowledge sharing
- Design hierarchical agent structures

## Target Processes

- multi-agent-system
- plan-and-execute-agent

## Implementation Details

### Core Components

1. **Agents**: Define roles, goals, backstories, and tools
2. **Tasks**: Specify descriptions, expected outputs, and assigned agents
3. **Crews**: Orchestrate agents with process types
4. **Tools**: Custom tool integration for agents

### Process Types

- Sequential: Linear task execution
- Hierarchical: Manager-led coordination
- Consensus: Agent voting and agreement

### Configuration Options

- LLM selection per agent
- Tool assignment
- Memory configuration
- Delegation settings
- Verbose/debug modes

### Best Practices

- Clear role definitions
- Appropriate task granularity
- Proper tool assignment
- Monitor agent interactions
- Handle failures gracefully

### Dependencies

- crewai
- crewai-tools

Overview

This skill configures CrewAI multi-agent orchestration for collaborative AI systems, letting you define agents, roles, goals, tasks, and crew orchestration patterns. It focuses on deterministic, resumable workflows with options for memory, tools, and delegation. Use it to build predictable multi-agent plans that can be monitored and resumed after interruptions.

How this skill works

The skill inspects and generates structured configurations for agents, tasks, crews, and tools. It sets role definitions, expected outputs, LLM selections, memory settings, and delegation parameters. It also produces orchestration patterns (sequential, hierarchical, consensus) and debugging options to run, monitor, and resume multi-agent workflows. Configurations integrate with crewai and crewai-tools to instantiate agents and manage runtime interactions.

When to use it

  • Designing a coordinated multi-agent workflow with clear roles and goals
  • Setting up hierarchical or manager-led orchestration for complex plans
  • Implementing consensus-based decisions or voting among agents
  • Attaching custom tools and memory to specialized agents
  • Creating resumable, deterministic automation pipelines for development tasks

Best practices

  • Define concise, actionable agent roles and backstories to guide behavior
  • Break work into appropriately granular tasks with clear expected outputs
  • Assign the right tools and LLM per agent based on capability and scope
  • Enable memory and knowledge sharing sparingly to avoid state bloat
  • Monitor interactions and add failure-handling strategies and retries

Example use cases

  • Orchestrate a code review crew where reviewers, a synthesizer, and a QA agent collaborate sequentially
  • Build a hierarchical product planning crew with manager agents assigning and validating subtasks
  • Run a consensus-based architecture decision process where agents propose and vote on options
  • Automate an end-to-end feature delivery pipeline with tool-backed agents for testing and deployment
  • Create resumable research workflows that preserve agent memory between sessions

FAQ

Can I assign different LLMs to different agents?

Yes. Configure LLM selection per agent to match model strengths to tasks.

How does resumability work?

Workflows persist crew state and agent memory so orchestration can resume deterministically after interruptions.

When should I use consensus vs hierarchical patterns?

Use consensus for democratic decisions requiring agreement; use hierarchical when a manager needs control and coordination.