Teaching dbt Wizard to write code like your team

Everyone's answer to inconsistent AI output is a better prompt. Prompts are the wrong layer — they're per-person, per-session, and they disappear the moment the conversation ends. Every team already has conventions: naming standards, testing and logic rules, and style guides. The fix for output that doesn't look like yours was never a better AI model, it was writing your standards down somewhere they get used, which is exactly what Agent Skills are, and what dbt Wizard reads before it writes a line of your code.