"What is CrewAI's defining pattern?"	Role-based sequential crews: define agents with role+goal+backstory; assign tasks; CrewAI orchestrates (sequential output→input or parallel). The crew structure EMERGES from roles, not explicit edges.	harness-engineering::dd12::recall
"CrewAI vs LangGraph — what's the key tradeoff?"	CrewAI: emergent (define roles → crew emerges). Simpler, faster setup, less explicit. LangGraph: explicit (draw nodes+edges → graph IS code). More control, fully auditable. Simplicity vs control.	harness-engineering::dd12::analysis
"CrewAI scores 33/55. Where does it win and lose?"	Wins Module 1.3 (Subagents): 4/5 — sequential crews are a clean multi-agent pattern. Loses Module 5 (Sandbox: 1/5 — none) and Module 8 (State: 2/5 — limited). Framework for orchestration, not production harness.	harness-engineering::dd12::analysis
"What's the multi-agent security risk in CrewAI?"	Role-based crews inherit all agents' COMBINED permissions. A compromised agent influences downstream agents via task outputs (injection propagates through the crew). Per-agent capability scoping (Module 2.4) is essential but NOT built-in.	harness-engineering::dd12::analysis
"When is CrewAI the right choice?"	Prototyping multi-agent patterns. Learning orchestration. Fast crew setup for non-production tasks. Move to LangGraph or Agents SDK for production.	harness-engineering::dd12::application