Two teams at a bank each put a prompt for the same fee-waiver decision into the prompt library, and the two prompts contradict. What should the library do?
Select an answer to reveal the explanation.
Short Explanation
Two clocks in one tower will not tell you the time if they disagree. You keep one approved fee-waiver prompt and retire the other. Leaving both active means the library itself is the conflict.
Full Explanation
A prompt library earns trust by holding one approved wording for a decision. These two fee-waiver prompts contradict, so the library should keep the approved prompt and retire the other, with the retirement recorded so nobody pastes the old wording back into an agent. Leaving both active and letting the last paste win makes runtime order the policy, and two teams can still ship opposite waivers. Deleting the library and moving copies into email removes the owner, the version, and the approval in one step, which is how the conflict started. Training a custom model on both prompts treats the contradiction as data to imitate, so the model can waive and refuse on the same facts. One owner should sign the surviving prompt so the two teams cannot each claim their wording is still current. Exam caveat: retirement is part of the library, not a silent delete. The retired wording should be identifiable so an audit can see what an older agent used. Operational check: run the same fee-waiver facts through the agent twice and confirm both runs follow the single approved prompt.