Module 6 · Project: an agent task test set
Project: a ten-task test set for your agent
In short
Build ten test tasks (ordinary, messy, traps) with tickable conditions, take a two-run baseline, then rerun and compare task by task before every change.
After this lesson you can
- Build a test set of ten tasks with expected outcomes
- Run the test set before changing the prompt, tools or model
Sign in to read this lesson
The title and summary are open to everyone. The full lesson is for signed-in learners.