AI Bots & Prompting Prompts

Improve this prompt: [paste your current prompt]. Diagnose: what is vague or ambiguous, what context is missing, what format instructions are needed, and where the model is likely to 'hallucinate' or go off-track. Rewrite it using best practices: role + context + task + constraints + output format. Explain each improvement and why it will produce better results.

I want to build a few-shot prompt that teaches Claude/GPT to [describe the task — e.g., 'classify customer emails by intent']. Create 5 high-quality input-output examples that demonstrate the pattern. The examples should: cover edge cases, be diverse, demonstrate the exact output format I want, and collectively teach the model the decision logic. Then assemble the full few-shot prompt.

Convert this straightforward prompt into a chain-of-thought (CoT) prompt that forces the AI to reason step by step before answering: [paste original prompt]. The CoT version should: instruct the model to think aloud, break the problem into explicit reasoning steps, reduce the chance of wrong shortcuts, and produce a more reliable answer. Show the before/after and explain the reasoning behind the CoT structure.

I keep getting bad outputs when I ask for [describe task]. Specifically, the model keeps: [describe the unwanted behaviour — too long, adds disclaimers, goes off-topic, uses wrong format, etc.]. Write a revised prompt that uses explicit negative constraints to prevent these issues, while preserving the quality of the desired output. Include do/don't lists and format examples.

I need Claude/GPT to perform as [a specific expert: e.g., 'a senior data analyst', 'a startup lawyer', 'an experienced copywriter']. Write a role prompt that: defines the persona's expertise level, their communication style, what they prioritise, how they handle uncertainty, and what they should explicitly NOT do (avoid overstepping into areas like giving financial advice). Test the prompt with this question: [test question].