SierraBased on real interviews
Agent loop bug hunt
Review a short tool-calling loop that should return the model's final answer after using its tools.
Unlock with Full AccessThe challenge
The implementation uses Chat Completions to run a tool-calling exchange. Follow how model responses, tool calls and their results move through the loop. Identify the bugs that prevent the exchange from behaving as intended, and explain targeted fixes without redesigning the agent.
What you'll practise
Conversation state
Trace the information passed between the model and the application across turns.
Tool handling
Inspect how requested tool calls are processed and their results returned.
Loop completion
Check how execution moves from intermediate tool work to the final answer.
Agent loop implementationCode included in the full challenge
Your code review
Explain each bug, its effect on the exchange and the proposed fix.
The code, instructions and feedback are included with Full Access.