Check it

Judge the agent's work by evidence, not by its own report.

  1. Stop 1, next

    What "done" means, up front acceptance criteria

    Think of it when you are about to hand an agent a task and "done" is still a feeling.

    StartAbout 3 minutes. The word for it comes at the end.

  2. Stop 2, next

    An agent that did not write it reviewer agent

    Think of it when an agent says its own change is correct and nobody else has looked.

    StartAbout 3 minutes. The word for it comes at the end.

  3. Stop 3, next

    A test for the agent itself eval

    Think of it when you change a skill, a rules file or the model, and need to know whether the agent got better or worse.

    StartAbout 3 minutes. The word for it comes at the end.