After months of trial and error with various LLMs, I've finally developed a prompt structure that consistently gives me good results.

I'm sharing it here to see what techniques you all are using.

My current approach:

Context Section

I always start by clearly defining the role and objective:

You are [specific expertise]. Your task is to [clear objective]. Background: [relevant context] Target audience: [who will consume this] 

System Behavior

This part was a game-changer for me:

Reasoning approach: [analytical/creative] Interaction style: [collaborative/directive] Error handling: [how to handle uncertainty] 

Chain-of-Thought

I've found that explicitly requesting step-by-step thinking produces much better results:

- Think through this problem systematically - Consider [specific aspects] before concluding - Evaluate multiple perspectives 

Output Format

Being super specific about what I want:

- Format: [markdown/code blocks/etc] - Required sections: [intro, analysis, conclusion] - Tone: [formal/casual/technical] 

Quality Checks

Adding these has reduced errors dramatically:

- Verify calculations - Check that you've addressed all parts of my question - Confirm your reasoning is consistent 

But I'm curious - what prompt structures work best for you?

Do you use completely different approaches? Any clever tricks for getting more creative responses? Or techniques for specialized domains like coding or creative writing?

Would love to build a collection of community best practices. Thanks in advance!