2. Prompt and Context Engineering#
The chapter 1 agent writes a good memo for the question you scripted. This chapter makes it work for real clients, in two parts taught in the same session.
2.1 Prompt Engineering. Write the standing instructions the model reads first: role, rules, format and examples. Then find where a prompt stops working.
2.2 Context Engineering. Decide what the model reads on each call, including earlier turns, have code put it there, and hand independent jobs to sub-agents with a clean desk.
Each part has its own Colab notebook.
Watch the lecture#
The session 2 recording, covering both parts. Watch it before or alongside 2.1 and 2.2; the slides are linked below the video.