AI & LLMs
System Prompt
The standing instructions that set an LLM's persona, rules, and constraints for an entire conversation.
Reviewed by the RadarTrek editorial team · June 2026
Every LLM API call is a structured conversation with roles — system, user, and assistant. The system prompt is seen before the user's first message and persists throughout: it defines the model's persona, output format, scope, and any hard constraints. It's the closest thing to "configuring" the model without retraining it.
Why it matters
- —A well-structured system prompt is usually faster and cheaper than fine-tuning for behaviour changes.
- —System prompts can be exploited by prompt injection — they are not a security boundary on their own.
- —Dynamic values (user plan, retrieved context) are typically injected into the system prompt at request time.
Where to learn this
🎓
System Prompts and Roles
How LLMs Actually Work course
This is the exact lesson that covers this term in depth — with examples, diagrams, and a hands-on exercise.