BACKGROUND
This project was an operational reference document for a content team using AI tools across ideation, structuring, drafting, and editing tasks. The goal was to give every team member a tested, documented set of prompts with clear guidance on when to use each one, when not to, and what human review was required after.
I organized the library across six categories covering thirteen prompts, each presented as a card with the prompt text, a list of variables to replace before sending, a use-when and not-for field, and a human review requirement. The final entry was a meta-prompt for diagnosing why a prompt was underperforming, which embedded the principle that if a prompt was producing poor output the issue might be that AI was not the right tool for the task at all.
The result was a library where the most important instruction appeared on the opening page: these prompts produce outputs, not finished content, and the human review field on each card is the minimum standard, not the ceiling.
