Week 1 Workbook

Lecture Notes Template

Write your own definitions:

  • Agentic loop:
  • stop_reason:
  • tool_use:
  • end_turn:
  • Coordinator agent:
  • Subagent isolation:

Reflection Prompts

  1. Why is parsing assistant text for completion a reliability problem?
  2. What information must be returned to the model after a tool call?
  3. When should a workflow use deterministic enforcement instead of prompt guidance?

Lab Worksheet

Scenario: Customer support resolution agent

Architecture sketch:

  • User input:
  • Required tools:
  • Preconditions:
  • Escalation triggers:
  • Handoff fields:

Loop pseudocode:

Subagent strategy:

Quiz Corrections

For any missed question:

  • What I chose:
  • Correct answer:
  • Why my answer was wrong:
  • What distinction I missed:

Week 1 Test Review

  • Strongest answer I wrote:
  • Weakest answer I wrote:
  • One concept to revisit this weekend: