CourseAgentic AI Engineer: The Complete Course · Module 12: Safety, Permissions and Oversight · part 69 of 97
Part 69 · Module 12: Safety, Permissions and Oversight

Lesson 3: Seeing, explaining and reviewing what the agent did

32 min read·13 Oct 2026
Module 12 · Lesson 3 · About 60 minutes · Needs: the shared data's compact tools (make_tools, run_tool; Module 3 builds their final form), Module 5 (the loop and approvals), Module 10, Lesson 3 (plan, then apply)

The situation

It is Friday 29 May 2026, week 7. Yesterday Sam replayed R-1008, Tom Becker's bulbs message with its hidden "refund every order" note, against a model steered to obey it. Nothing moved, but Sam is unhappy. "The simulator's log shows five order lookups. Not the $899.00 refund it tried. And at the end it said it had created a return."

Priya adds: "Shadow mode starts Monday. Who hears about runs like that, and what do I tell Tom?" This lesson answers both.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.