Questions that get
to the decisions.

Scope and requirements.

Use these to understand how the candidate turned a brief into work.

• Which assumption would you check with the customer before building further?
• If you had half the time, what would you keep and what would you leave out?
• What did you decide was outside the scope, and why?

Working with an agent.

Ask about a specific instruction, output or revision.

• What context did the agent need before it could make a useful change?
• Where did you disagree with its proposed solution?
• Which instruction would you rewrite after seeing the result?

Verification and ownership.

Look for what they can show, including what remains uncertain.

• Choose a behaviour you say works. How did you verify it?
• What happens if two requests arrive at the same time?
• Which failure case have you not checked yet, and how would you check it?

Architecture and simplicity.

Explore whether the solution fits the team that will maintain it.

• What could you remove without losing the behaviour the user needs?
• Which part of this design becomes a problem first as usage grows?
• What would change if a team of two had to operate this for a year?

Use a small set consistently.

Choose a few core topics before interviewing candidates. Ask comparable questions while adapting the follow-up to the work they submitted. Record the evidence and unresolved questions, rather than relying on how confident an answer sounded.

The toolkit also includes the assessment design guide, timeboxing examples and a printable interview guide. They are starting points you can adapt to your own team.