AI Prompts for Prompt Engineering

Here is a prompt I am using: [PASTE PROMPT]. It is producing [DESCRIBE UNSATISFYING OUTPUT]. Analyze what is weak or ambiguous about my prompt. What is missing, what is unclear, and what would a more experienced prompt engineer add or change? Then rewrite the prompt with your suggested improvements.

I want to prompt an AI to [DESCRIBE TASK]. Here is my current prompt: [PASTE PROMPT]. What context about [AUDIENCE, FORMAT, CONSTRAINTS, TONE, PURPOSE] would most improve the output? Rewrite the prompt with the most important context added, and explain why each addition helps.

I want AI output in a specific format for [DESCRIBE TASK]. The format I need is: [DESCRIBE FORMAT, e.g. JSON, bulleted list, table, specific structure]. Write a prompt that gives clear format instructions that are specific enough to produce consistent output across multiple runs.

I am building an application that uses an AI for [DESCRIBE USE CASE]. Write a system prompt that establishes the AI's role, tone, constraints, and response format. The system prompt should produce consistent behavior regardless of what the user asks and should gracefully handle requests that fall outside the intended use case.

I want to use few-shot examples to get more consistent output for [DESCRIBE TASK]. Here are three examples of the input-output pairs I want: [PASTE EXAMPLES]. Write a complete prompt that includes these examples in a format that teaches the AI the pattern I want, and explain the best way to structure few-shot examples for this type of task.

This prompt produces good output sometimes and bad output other times: [PASTE PROMPT]. The bad outputs look like: [DESCRIBE BAD OUTPUTS]. Diagnose why this is inconsistent. Is it an ambiguity in the instructions, a missing constraint, or a context gap? Write a more robust version that produces consistently good output.

My prompt is too vague and produces generic results, but when I add too much detail it becomes rigid. The current prompt: [PASTE PROMPT]. I want output that is [DESCRIBE IDEAL OUTPUT QUALITIES]. Find the right level of specificity: add enough constraints to eliminate generic outputs without over-constraining the response.

This prompt is producing outputs that include fabricated information: [PASTE PROMPT]. The hallucinations look like: [DESCRIBE EXAMPLES]. What prompt techniques reduce hallucination for this type of task? Rewrite the prompt to encourage the AI to be accurate and to explicitly signal when it is uncertain or does not have reliable information.

I need consistently structured long-form output for [DESCRIBE TASK]. My current prompt: [PASTE PROMPT]. The AI loses structure halfway through or the quality degrades in longer outputs. What prompt strategies help maintain consistency and quality across longer outputs? Rewrite the prompt to address this.

I am using [MODEL NAME, e.g. Claude, GPT-4, Gemini] for [DESCRIBE TASK]. Here is my current prompt: [PASTE PROMPT]. What are the specific characteristics of this model that I should be aware of when writing prompts? Are there patterns this model responds to particularly well or poorly? Rewrite the prompt optimized for this model.

I need the AI to reason through a complex problem: [DESCRIBE PROBLEM]. Write a chain-of-thought prompt that instructs the AI to show its reasoning step by step before reaching a conclusion. Explain when chain-of-thought prompting helps and when it is unnecessary, and what the limitations are for this specific problem type.

I need the AI to complete a task with multiple distinct steps: [DESCRIBE TASK AND STEPS]. Write a prompt that breaks this into a clear sequence the AI can follow without losing context between steps. Should this be a single prompt with step-by-step instructions, or multiple separate prompts? Explain the trade-offs.

I want the AI to take on a specific expert role for [DESCRIBE TASK]. Write a role assignment prompt that gives the AI a clear persona, expertise level, communication style, and constraints. Explain what makes a role assignment effective versus what makes it produce unhelpful role-play behavior.

I want the AI to evaluate [DESCRIBE CONTENT] against specific criteria. Write a prompt that gives the AI clear, unambiguous criteria and a structured output format for the evaluation. The prompt should produce consistent, actionable assessments rather than vague feedback.

I want the AI to generate content and then critique its own output for [DESCRIBE TASK]. Write a two-step prompt: first generation, then structured self-critique. The critique should focus on: [CRITERIA, e.g. accuracy, completeness, clarity]. How do I structure this to get honest self-critique rather than defensive justification?

I want to extract structured data from [DESCRIBE UNSTRUCTURED INPUT, e.g. emails, documents, web pages]. The output format I need is: [DESCRIBE FORMAT]. Write a prompt that reliably extracts the right data, handles missing or ambiguous values, and outputs consistent structure. Include instructions for handling cases where the required information is not present.

I need to classify [DESCRIBE CONTENT] into these categories: [LIST CATEGORIES WITH DEFINITIONS]. Write a prompt that produces consistent, accurate classification. Include: how to handle borderline cases, what to do with content that fits multiple categories, and how to output confidence or uncertainty.

I want to use AI to generate code for [DESCRIBE TASK]. Write a prompt that produces code that is: correct, follows best practices for [LANGUAGE], includes error handling, has meaningful variable names, and is ready to use without significant modification. What context does the AI need about my project to produce usable code?

I want to summarize [DESCRIBE DOCUMENT TYPE] consistently. The summaries should be: [DESCRIBE FORMAT, LENGTH, FOCUS]. Write a prompt that produces summaries that capture the most important information, maintain a consistent structure, and are appropriate for [DESCRIBE AUDIENCE]. Include instructions for handling very long documents.

I want to create a reusable prompt template for [DESCRIBE RECURRING TASK]. The task varies in [DESCRIBE VARIABLE PARTS]. Design a template with clear placeholders, instructions for what to fill in each placeholder, and any guardrails that prevent the template from being used incorrectly. The template should be usable by someone who did not write it.