Prompt Engineering Patterns

Master advanced prompt engineering techniques to maximize LLM performance, reliability, and controllability in production.

When to Use This Skill

• Designing complex prompts for production LLM applications
• Optimizing prompt performance and consistency
• Implementing structured reasoning patterns (chain-of-thought, tree-of-thought)
• Building few-shot learning systems with dynamic example selection
• Creating reusable prompt templates with variable interpolation
• Debugging and refining prompts that produce inconsistent outputs
• Implementing system prompts for specialized AI assistants
• Creating new skills using comprehensive development frameworks
• Updating existing skills with improved structure and content
• Generating high-quality LLM commands using structured guidelines
• Applying universal patterns for command consistency and reliability
• Creating new agents using comprehensive guidelines and pattern frameworks
• Updating existing agents with quality standards and validation checklists
• Applying agent pattern selection for consistent agent architecture

Core Capabilities

• Few-Shot Learning: Dynamic example selection and construction
• Chain-of-Thought Prompting: Structured reasoning elicitation
• Prompt Optimization: Iterative refinement and performance measurement
• Template Systems: Reusable prompt components with variable interpolation
• System Prompt Design: Model behavior and constraint specification
• Skill Creation Frameworks: Comprehensive guidelines for producing high-quality skills based on repository analysis
• Command Generation: Structured frameworks for creating production-ready LLM commands with proper layout patterns and writing conventions
• Universal Pattern Integration: Applying structural guidelines (progressive disclosure, quality assurance, cross-referencing) to prompt design for better reliability and maintainability
• Agent Creation Frameworks: Comprehensive guidelines for producing high-quality agents based on repository analysis with pattern selection, quality assurance, and validation standards

Quick Start

# Basic prompt engineering pattern
prompt = "Convert this natural language query to SQL: Find all users who registered in the last 30 days"

For advanced implementations with templates and examples, see the reference guides below.

Available Resources

Use these guides based on your specific needs:

• "I need to implement few-shot learning" → Load references/few-shot-learning.md for example selection strategies and dynamic retrieval
• "How do I add step-by-step reasoning to prompts?" → Load references/chain-of-thought.md for CoT prompting techniques
• "My prompts produce inconsistent results" → Load references/prompt-optimization.md for systematic refinement workflows
• "I want to create reusable prompt templates" → Load references/prompt-templates.md for variable interpolation and modular components
• "How do I design effective system prompts?" → Load references/system-prompts.md for behavior specification and constraints
• "How do I ensure prompt safety and security?" → Load references/prompt-safety.md for input validation and jailbreak prevention
• "I need to work with images or audio in prompts" → Load references/multimodal-prompting.md for cross-modal reasoning techniques
• "How do I create multi-step workflows?" → Load references/prompt-chaining.md for agent orchestration and complex chains
• "I want to evaluate and compare prompt performance" → Load references/evaluation-frameworks.md for automated testing and A/B experimentation
• "How do I integrate universal patterns into my prompts?" → Load references/prompt-templates.md for structural guidelines and integration techniques
• "I need to create new skills" → Load references/skills/skill-creation-guidelines.md for comprehensive skill development frameworks
• "How do I create new commands" → Load references/commands/command-creation-guidelines.md for structured command generation with universal patterns
• "How do I select agent patterns" → Load references/agents/pattern-selection.md for decision framework choosing agent structures
• "How do I create comprehensive agents" → Load references/agents/comprehensive-agent-guidelines.md for complex multi-domain agent structure
• "How do I create simplified agents" → Load references/agents/simplified-agent-guidelines.md for focused narrow-domain agent structure
• "How do I create specialized agents" → Load references/agents/specialized-agent-guidelines.md for domain-specific workflow agent structure
• "How do I ensure agent quality" → Load references/agents/quality-assurance-checklist.md for validation standards and checklists
• "How do I style agent specifications" → Load references/agents/writing-style-guidelines.md for tone, voice, and formatting standards
• "How do I choose agent models" → Load references/agents/model-selection-guidelines.md for complexity-based model selection
• "How do I write example interactions" → Load references/agents/example-interactions-guidelines.md for representative user prompt guidelines

Best Practices

1. Start simple, add complexity only when needed
2. Test extensively on diverse, representative inputs
3. Treat prompts as code with proper versioning
4. Monitor performance metrics in production
5. Document why prompts are structured as they are

Common Pitfalls

• Starting with complex prompts before trying simple ones
• Using examples that don't match the target task
• Exceeding token limits with excessive examples
• Leaving room for multiple interpretations

Assets and Tools

• references/few-shot-learning.md: Deep dive on example selection and construction
• references/chain-of-thought.md: Advanced reasoning elicitation techniques
• references/prompt-optimization.md: Systematic refinement workflows
• references/prompt-templates.md: Reusable template patterns
• references/system-prompts.md: System-level prompt design
• references/prompt-safety.md: Input validation, jailbreak prevention, and content policies
• references/multimodal-prompting.md: Image/text combinations and cross-modal reasoning
• references/prompt-chaining.md: Multi-step workflows and agent orchestration
• references/evaluation-frameworks.md: Automated testing, benchmarking, and A/B testing
• references/skills/skill-creation-guidelines.md: Comprehensive frameworks for creating high-quality skills
• references/commands/command-creation-guidelines.md: Structured guidelines for generating production-ready LLM commands
• references/agents/pattern-selection.md: Decision framework for selecting appropriate agent structural patterns
• references/agents/comprehensive-agent-guidelines.md: Complete structure guidelines for complex multi-domain agents
• references/agents/simplified-agent-guidelines.md: Structure guidelines for focused narrow-domain agents
• references/agents/specialized-agent-guidelines.md: Structure guidelines for domain-specific workflow agents
• references/agents/quality-assurance-checklist.md: Comprehensive validation standards for agent specifications
• references/agents/writing-style-guidelines.md: Tone, voice, tense, and formatting standards for agents
• references/agents/model-selection-guidelines.md: Complexity-based criteria for Claude model selection
• references/agents/example-interactions-guidelines.md: Guidelines for creating representative user interaction examples
• assets/prompt-template-library.md: Ready-to-use prompt templates for common tasks, including universal pattern integration
• assets/few-shot-examples.json: Curated example datasets for various domains

Success Metrics

Track accuracy, consistency, latency, token usage, success rate, and user satisfaction.

Implementation Guide

1. Start with simple prompts, test, then add complexity
2. Use the reference guides above for specific techniques
3. Experiment with the provided templates and examples
4. Implement automated evaluation and optimization

Quality Assurance Checklist

Before deploying prompts in production:

• Test on diverse, representative inputs (not just edge cases)
• Validate outputs meet quality standards and consistency requirements
• Measure performance metrics (accuracy, latency, token usage)
• Document prompt rationale and expected behavior
• Include fallback mechanisms for unreliable prompts
• Test for adversarial inputs and jailbreak attempts
• Monitor performance in production with automated alerts