You don't need a hundred support prompts. You need five you trust and actually reuse.
Keep a negative list
Ask for the reasoning, then the result. When the model shows its steps you can see exactly where to push back.
- Lower the reading level, then check what broke
- Compare two outputs instead of trusting one
- Ask for three angles, then cut to the one worth shipping
Cut before you polish
Treat the first output as a draft, not a verdict. The prompt does the typing; you keep the judgment about what ships.
- Name the audience and the single goal before you generate
- Save the prompt that worked so the next person reuses it
- Ask for three angles, then cut to the one worth shipping
Ask for the reasoning, then the result. When the model shows its steps you can see exactly where to push back.