Defining the System Message The [System Prompt] acts as the set of rules for the AI. While the user prompt changes with every message, the system prompt remains static, anchoring the model's persona. For example, setting your system to 'You are a senior JavaScript developer providing clean, annotated code' ensures that every subsequent response adheres to these standards without needing constant reminders. Persona Constraints You can assign roles to improve the quality of technical output. By defining the AI as a 'critical technical reviewer,' you encourage the model to look for potential edge cases or bugs rather than just providing a 'happy path' response. This is a core concept in [Role Prompting]. Best Practices Always keep system instructions clear and prioritized. Use clear headings such as '# Instructions' and '# Persona'. Review official documentation from Microsoft's Azure AI Guidelines for handling system instructions at scale. FAQs Does the system prompt count against token limits? Yes, it is part of the context window. Can I change the system prompt halfway through? Most APIs allow this, but it may cause the model to 'reset' its behavior. How specific should a persona be? Provide just enough detail to guide the tone without confusing the core task.
Lesson 3: System Prompts and Role-Based Engineering
“Setting the persona and behavioral boundaries for your AI.”
