Draft New Prompt

Create high-quality prompts from scratch using proven engineering patterns.

Workflow

1. Analyze Task Requirements

Identify the core elements of your prompt need:

• Purpose: What should the prompt accomplish?
• Context: What background information is needed?
• Constraints: Any specific rules or limitations?
• Output format: How should results be structured?
• Complexity level: Simple task or multi-step workflow?

2. Select Appropriate Pattern

Based on the prompt-engineering skill, choose the right approach:

• Chain-of-Thought (CoT): For reasoning-heavy tasks
• Few-shot learning: When examples help clarify intent
• Role-based: For specialized expertise (e.g., "You are a security auditor...")
• Template-based: Following standardized structure from prompt-library

3. Apply Complexity Decision Framework

Use the Signal-to-Noise Rule from prompt-engineering skill:

• Markdown-first: Default for most prompts (fewer tokens, Claude-native)
• Hybrid XML/Markdown: Only when complexity triggers met:
  • Data Isolation: >50 lines of raw data
  • Constraint Weight: Rules that MUST NEVER be broken
  • Internal Monologue: Complex reasoning requiring step-by-step

4. Create High-Quality Prompt

Structure the prompt following best practices:

• Use # Context section for background
• Use # Assignment section for specific instructions
• Include concrete examples when helpful
• Keep instructions clear and unambiguous
• Apply appropriate pattern from prompt-library templates

5. Output to File

Write the completed prompt to:

• PROMPT.md in the current directory
• Include YAML frontmatter if needed
• Follow template structure from prompt-library skill

Key Resources

• prompt-engineering skill: Theory, patterns, and optimization techniques
• prompt-library skill: Templates and taxonomy
• Template: Use single-prompt.md from prompt-library skill

Output

Complete prompt saved to PROMPT.md with:

• Clear instructions
• Appropriate structure (Markdown/XML)
• Examples if needed
• Ready for immediate use