Definition
A system prompt is the standing instruction an application gives a language model before the user ever types anything. It defines the model's persona, its rules, the tools it may use, and the boundaries it must respect. Where each user message is treated as a request, the system prompt is treated as policy, sitting at a higher priority in the conversation so the model follows it across every turn. It is the primary lever developers have to shape behavior without retraining the model.
What goes in a system prompt
Typical contents include the assistant's role ("you are a Bitcoin mining support agent"), formatting requirements, refusal rules, available tool descriptions, and contextual facts the model should treat as authoritative. For an AI agent, the system prompt also enumerates the tools and the loop the model should follow. Because it is prepended to every request, a long system prompt consumes part of the context window and is billed on every call.
The security boundary
The system prompt is meant to be a trust boundary, but it is a soft one. Prompt injection attacks specifically try to make the model disregard its system prompt or reveal it. Treating the system prompt as a secret is therefore weak security; assume a determined user can extract or override it, and put real controls (scoped permissions, human approval) at the tool layer instead.
For sovereign operators running self-hosted inference, the system prompt is fully under your control, no vendor can silently change it, which is a meaningful advantage when you need predictable, auditable model behavior.
In Simple Terms
A system prompt is the standing instruction an application gives a language model before the user ever types anything. It defines the model’s persona, its…