crewai
Build Multi-Agent AI Systems with CrewAI
Creating collaborative AI agent teams requires understanding CrewAI framework patterns. This skill provides expert guidance on agent design, task definition, and crew orchestration for building production-ready multi-agent systems.
Download the skill ZIP
Upload in Claude
Go to Settings → Capabilities → Skills → Upload skill
Toggle on and start using
Test it
Using "crewai". Create a research crew with two agents
Expected outcome:
A basic crew configuration with researcher and writer agents, including YAML config files and Python decorator-based setup, demonstrating agent roles, goals, backstories, and task definitions.
Using "crewai". Show me how to use hierarchical process
Expected outcome:
Complete example with manager_llm configuration, worker agent definitions, and how the manager delegates tasks to specialized agents.
Security Audit
SafeStatic analysis flagged potential issues (external_commands, weak_crypto) but all are false positives. The file is a markdown documentation skill using markdown code block delimiters (backticks) which were misidentified as shell execution. No executable code, network calls, or security risks present. Safe for publication.
High Risk Issues (1)
Medium Risk Issues (1)
Quality Score
What You Can Build
Research and Content Creation Pipeline
Create a crew with researcher, writer, and editor agents to automate research and content production workflows
Data Analysis Team
Build specialized analyst agents for data collection, processing, and visualization tasks
Code Review and Documentation
Automate code review, documentation generation, and quality assurance with specialized agent crews
Try These Prompts
Create a simple CrewAI crew with a researcher agent and a writer agent. The researcher should find information on {topic} and the writer should create a blog post from the research.Set up a hierarchical CrewAI process with a manager agent that coordinates three worker agents: a data collector, an analyst, and a reporter. Use gpt-4o as the manager LLM.
Create a crew where task B depends on the output of task A. Task A should gather data and task B should analyze that data and produce a summary report.
Configure a CrewAI crew with planning enabled. Create three agents: planner, executor, and validator. Enable planning so the crew generates an execution plan before running tasks.
Best Practices
- Define specific, focused agent roles rather than generic ones - use titles like 'Senior Research Analyst' instead of 'Researcher'
- Always specify expected_output for tasks to ensure consistent, usable results
- Start with 3-5 agents and add more only when coordination overhead is justified
Avoid
- Creating vague agent roles without clear expertise areas leads to poor task delegation
- Skipping expected_output definition causes inconsistent results and hard-to-chain tasks
- Using too many agents creates coordination overhead and slower execution without proportional benefit