agent loops — tool_use and the request/tool/respond cycle4 / 9

stop_reason, tool_use, tool_result — the loop every agent runs

Read the editor. The fake model returns tool_use on calls 1 and 2, then end_turn on call 3. How many times does the loop iterate before exiting?

Your prediction

⌘↵ runs the editor.

Booting Python…

Output

[promptdojo:~]$ _