QuestionQ28

VerifiedEvaluating and deploying agentic workflows

You are designing an automated code-refactoring assistant with the open-source Agent Development Kit (ADK) library. The solution requires a multistep process in which a generator agent drafts modernized code, a critic agent evaluates the draft against strict security rules, and a refiner agent modifies the code until the critic agent approves the output or an exit condition is met. You need to select an agentic workflow architecture that natively supports this continuous generation, evaluation, and revision cycle.

What should you do?

Explanation

An ADK LoopAgent repeatedly runs its subagents in order, enabling the critic to evaluate the code and the refiner to revise it. The max_iterations setting bounds execution, while a tool that sets tool_context.actions.escalate can stop the loop when the critic approves. Reaching the iteration limit does not itself mean the code has been approved. (adk.dev)

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!