Build to a verifiable check
Build against a concrete, runnable check.
Official basis Give Claude a way to verify its work — show evidence, don't assert success. docs ↗
The prompt
Implement <function>. Example cases: <input → output>, <input → output>. After implementing, run the tests and show me they pass — show the output, don't just assert that it works.Why it works
A check Claude can run turns a session you have to babysit into one you can walk away from. Requiring the actual test output (“show evidence”) closes the trust-then-verify gap where the model claims success it didn’t earn.
Fill in
<function>— what to build.<input → output>— 2–3 concrete example cases.
Follow-ups
- “Add an example for the empty / error case and make it pass too.”