System Prompt
The standing instructions a harness sends on every request.
A system prompt is the block of instructions a harness prepends to every model provider request: the role you are playing, the rules you must follow, the tools you may call, the output format. It is not a secret and not magic, it is configuration.
Because it is prepended on every turn, it is also the most reliable place to put a rule you always want applied, and the most expensive: it occupies the context window on every single call. Long system prompts are a common hidden cost, and a common cause of "the model ignores the rule I wrote" when the rule sits too far from the current request.
Related terms
Prompt
The input you give a model. The word covers more than the question you type.
Harness
Everything around the model that turns it into an agent: tools, system prompt, permissions, hooks.
Context Engineering
Deciding what goes into the model's context, and what stays out.
Guardrails
Checks that stop or constrain an action before it happens.
Want this working inside your team?
It starts with one conversation about your workflow.