Reviewer Agent

Critics focus on correctness; reviewers focus on deliverable quality. Separating the roles prevents one agent from trading accuracy for style.

The reviewer is the copy editor at a newspaper — facts are already verified; they ensure the headline sings and the layout is clean.

Scenario

Critic-approved support reply → Reviewer shortens to 3 sentences, adds greeting, removes jargon, ensures CSAT-friendly tone.

Solution

In Multi-Agent Systems, apply Reviewer Agent to this scenario: Critic-approved support reply → Reviewer shortens to 3 sentences, adds greeting, removes jargon, ensures CSAT-friendly tone. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Reviewer Agent (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Reviewer Agent happens in the code.

1reviewer = Agent(  # key line for Reviewer Agent2    instructions="Polish for customer-facing tone. Keep facts unchanged. Max 150 words.",3)4final = reviewer.run_sync(f"Draft reply: {approved_draft}").output  # key line for Reviewer Agent

• Treating Reviewer Agent as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for reviewer agent

Quick recap — the most important points from this module.