Prompts set the task, the context, and the shape of the expected answer. Structure, examples, and small debug loops keep that answer reliable.
Topic | Lessons | Description | Time |
|---|---|---|---|
5 | Five prompt parts, roles, specificity, weak-strong pairs, and a budget-checked builder. | 40 min | |
4 | Examples and staged reasoning in prompts: benefits, costs, ordering, self-checks, and structured output. | 30 min | |
4 | Debug failing prompts: name the failure class, isolate edits, and hold gains with a small regression set. | 30 min |
An isolated edit maps each change to its effect, so the kept version holds only edits that helped. The Prompt Debugging topic covers the one-change rule and small regression sets.