Skip to content

jevierra/prompt-architecture-patterns

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 

Repository files navigation

Prompt Architecture Patterns

Design patterns for production prompt engineering


These patterns emerged from building and operating a 9-employee AI workforce in daily production since January 2026. They're modeled after Gang of Four design patterns -- problem/solution/example format -- because prompt engineering has matured past the "just write a good prompt" stage. Production systems need architecture, not tricks.

These aren't theoretical. They're extracted from systems running in production daily.

Pattern Catalog

# Pattern Description
01 Identity Separation Decompose AI identity into independent, maintainable components
02 Session Loading Reliable context bootstrapping for stateless AI sessions
03 Rolling Task Protocol Prevent task staleness with time-based detection and forced decisions
04 Attention Routing Surface critical items before routine work
05 Multi-Model Deployment Match model capabilities and cost to task requirements
06 Employee vs Agent Choose the right human-AI collaboration pattern for your use case
07 Persistent Memory Structured file-based memory that survives session boundaries
08 Governance Rules Naming conventions and structural discipline for AI-navigable file systems
09 Cross-Employee Coordination Information sharing across multiple AI employees without confusion
10 Cultural Layer Personality and engagement systems that improve output quality

How to Read These Patterns

Each pattern follows a consistent structure:

  • Problem -- The challenge you'll recognize from your own work
  • Solution -- The architectural approach
  • Implementation -- How to build it
  • Example -- Concrete, production-tested example
  • Anti-Patterns -- Mistakes we made so you don't have to
  • Related Patterns -- How patterns connect to each other

Who This Is For

  • Engineers building AI-powered workflows and tools
  • Teams deploying LLMs in production beyond simple chat interfaces
  • Anyone who has felt the pain of prompts that work in testing but fail in production

Philosophy

Prompt engineering is software engineering. The same principles that make code maintainable -- separation of concerns, clear interfaces, explicit state management -- make prompts maintainable. These patterns are the proof.


prompt-engineering design-patterns ai llm best-practices

About

Design patterns for production prompt engineering, modeled after Gang of Four patterns

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors