Prompt Design Principles AI Prompt

Apply structured prompt design principles to improve the reliability and quality of LLM outputs for this task.

Task: {{task_description}}
Model: {{model}} (GPT-4, Claude, Llama, Mistral, etc.)
Output format required: {{output_format}}
Current prompt: {{current_prompt}} (if exists)

1. Anatomy of an effective prompt:

   System prompt (instruction context):
   - State the role: 'You are an expert {{domain}} analyst.'
   - State the task clearly: what should the model do?
   - State the constraints: what should the model NOT do?
   - State the output format explicitly: 'Return a JSON object with fields...'
   - Keep the system prompt focused: one role, one task type per system prompt

   User prompt (the input):
   - Provide the specific input to process
   - Separate instructions from data: use XML tags, triple backticks, or markdown headings
   - Be specific: avoid vague instructions like 'summarize well' — say 'summarize in 3 bullet points, each < 20 words'

2. Clarity and specificity:
   - Vague: 'Analyze this text'
   - Better: 'Identify the main argument, list 3 supporting claims, and note any logical fallacies. Return as JSON: {main_argument: str, supporting_claims: [str], fallacies: [str]}'
   - Always specify: length, format, level of detail, target audience, and any constraints

3. Context and role-setting:
   - Assigning a role improves domain-specific outputs: 'You are a board-certified cardiologist...'
   - Providing context reduces hallucination: tell the model what it needs to know
   - Grounding: 'Based only on the following document:' prevents the model from using outside knowledge

4. Output format specification:
   - For structured data: always specify JSON schema with field names, types, and descriptions
   - For text: specify structure (e.g., 'Use H2 headings for each section, bullet points under each')
   - Use few-shot examples for complex or non-standard formats
   - Add: 'Return only the JSON object and nothing else, no preamble or explanation'

5. Negative instructions:
   - 'Do not include any information not present in the source text'
   - 'Do not use the phrase "In conclusion"'
   - 'Do not make assumptions about data not provided'

6. Iterative refinement:
   - Test the prompt on 10-20 diverse examples before finalizing
   - Review failures: which examples fail and why?
   - Add a clarifying sentence to the system prompt for each failure category

Return: revised system prompt, user prompt template, output format specification, and test plan.