A critic agent examines an output against goals, constraints, or quality criteria. It may identify errors, explain what is missing, score the result, and suggest which parts should be revised.
The critic is not automatically correct or independent. Its feedback is only as useful as its evaluation criteria and judgment, so important results still need verification against the actual task requirements.
ELI5
A critic agent is like a second helper whose job is to check the first helper's work. It points out problems and suggests what should be improved.
For example, one agent might build a game scene, while the critic notices that an object is in the wrong place and asks the first agent to move it.
