Skip to content

Latest commit

 

History

History
91 lines (73 loc) · 2.16 KB

prompt-guidelines.md

File metadata and controls

91 lines (73 loc) · 2.16 KB

Prompt Guidelines

Core Principles

  1. Direct AI Instructions
  • Write prompts as clear directives to the AI
  • Focus on process steps and expected outputs
  • Exclude metadata and human-focused documentation
  1. Separation of Concerns
  • Keep prompt instructions in .md files
  • Store usage documentation in .meta.md files
  • Maintain clean separation between AI and human documentation
  1. Prompt Structure
  • Start with clear workflow name
  • List steps in logical order
  • Define output formats
  • Include validation rules
  1. Output Formats
  • Use consistent markdown structures
  • Include file paths in code blocks
  • Separate command-line instructions
  • Maintain clear formatting
  1. Workflow Chain Integration
  • Design prompts to fit into chains
  • Define clear inputs and outputs
  • Include chain verification points
  • Support systematic progress tracking

Best Practices

  1. Keep prompts focused
  • One primary task per prompt
  • Clear scope boundaries
  • Minimal complexity
  • Define chain position if applicable
  1. Enable interaction
  • Break complex decisions into steps
  • Allow for user input and revision
  • Provide clear options
  • Support chain status checks
  1. Maintain context
  • Reference existing project files
  • Consider current development stage
  • Check for conflicts and dependencies
  • Track chain dependencies
  1. Support validation
  • Include quality checks
  • Verify compatibility
  • Confirm requirements alignment
  • Validate chain integrity

Chain Integration

  1. Input/Output Definition
  • List required inputs
  • Specify expected outputs
  • Document dependencies
  • Define validation criteria
  1. Chain Position
  • Identify preceding phases
  • Document following phases
  • Specify verification points
  • Include progress tracking
  1. Status Commands
  • Provide status check commands
  • Include progress indicators
  • Support chain integrity checks
  • Enable workflow validation

Testing Prompts

  1. Individual Testing
  • Verify prompt works across different AI models
  • Test with various project types and sizes
  • Confirm output consistency
  • Validate standalone operation
  1. Chain Testing
  • Test chain integration
  • Verify input/output flow
  • Validate chain integrity
  • Confirm progress tracking