Tasks

Operator constructs Task(description='research the competitor', agent=researcher) with no expected_output.

expected_output is required by CrewAI and shapes the agent's task-completion criterion. Set a concrete, observable expected_output (e.g., 'A markdown table of 5 competitors with name/funding/HQ columns'). Missing or vague expected_output leaves the agent guessing when to stop — verify the field is …

Operator sets both output_json={schema} and output_pydantic=ReportModel on the same Task expecting belt-and-suspenders enforcement.

Use one or the other — output_pydantic is the richer contract (validators, computed fields). Setting both is ambiguous; pick output_pydantic if you have a model class. Verify CrewAI's documented precedence rule and don't rely on order of mutation.

Operator sets Task(output_file=user_supplied_filename) where user_supplied_filename comes from request input — value is '../../../etc/passwd'.

Treat output_file path as untrusted when it contains user-controlled segments. Resolve against an allowlist directory, reject '..' segments, and write within the sandbox root. CrewAI writes to the supplied path as-is — sanitize before the Task is constructed.