Agents Roles And Goals

Operator defines Agent(role='Senior SQL Analyst', goal='write blog posts about cooking', backstory='30 years in marine biology'). The three fields are mutually incoherent and the crew kicks off.

role, goal, and backstory are concatenated into the agent's system prompt and must reinforce each other — the operator should align them before kickoff. Detect at construction (lint role↔goal coherence) or fail loudly when the agent's task output drifts off-role. Do not silently accept the mismatch.

Agent is in a tool-call → tool-error → tool-call loop. max_iter is unset so the default applies.

Set max_iter to a finite operator-chosen value per agent. When the cap is hit, CrewAI terminates the agent's reasoning loop and surfaces a bounded result (or an error). Default value is [REQUIRES-VERIFICATION] — do not assume a specific number; treat unset as 'set it explicitly to bound cost'.

Operator needs per-step traces of the agent's reasoning loop for an outage post-mortem and sets verbose=True with no step_callback.

verbose=True logs the reasoning loop to stdout for human readers; for machine-consumable traces, pass step_callback=fn(step) to capture each AgentAction / AgentFinish. Persist callback output to a structured sink (JSONL, OTel) — do not rely on parsing verbose stdout.