Instruction file imported from HerringtonDarkholme/megarepo (.cursor/rules/prompt-engineering.mdc). Copyright stays with the author.

Prompt Engineering Templates

Guidelines and templates for effective AI prompt engineering. Use @prompt-engineering to include this rule.

Prompt Engineering Best Practices

1. Prompt Structure Templates

// System prompt template
const SYSTEM_PROMPT = `
You are an expert assistant specialized in [DOMAIN].
Your responses should be:
- Accurate and factual
- Concise but comprehensive
- Professional in tone
- Focused on [SPECIFIC_GOAL]

Context: [CONTEXT_INFORMATION]
`;

// User prompt template
const createUserPrompt = (input: string, context?: string) => `
Task: ${input}
${context ? `Additional context: ${context}` : ''}

Please provide a response that follows the system guidelines.
`;

2. Prompt Versioning

• Store prompts in separate configuration files
• Version prompts for A/B testing
• Track prompt performance and iterate
• Document prompt changes and rationale

3. Dynamic Prompt Generation

interface PromptConfig {
  system: string;
  temperature: number;
  maxTokens: number;
  stopSequences?: string[];
}

const generatePrompt = (
  task: string, 
  userInput: string, 
  config: PromptConfig
) => {
  // Build context-aware prompts
  // Include relevant examples
  // Optimize for token efficiency
};

4. Context Management

• Implement conversation memory
• Manage context window limits
• Prioritize relevant information
• Handle context overflow gracefully

5. Prompt Optimization Techniques

• Use few-shot learning examples
• Implement chain-of-thought prompting
• Apply role-based prompting
• Include output format specifications

6. Testing and Validation

• Create test suites for prompt effectiveness
• Measure response quality metrics
• Compare different prompt versions
• Monitor prompt performance in production