Evidence-first authority page
What Is a System Prompt?
The distinction between a system prompt and a user prompt, and why system prompt design matters more for agents than for simple chat.
Direct answer
A system prompt is an instruction set provided to a model before any user input, establishing its role, behavioral constraints, and available tools — distinct from the user prompt, which is the specific request or message the model responds to; for agents, the system prompt typically also defines tool-use conventions and stopping behavior, making its design a first-order factor in reliability.
Last reviewed 2026-08-29
This page remains noindex while its page-level source receipts are completed and verified. It is available for review, not presented as publication-ready authority.
System prompt vs. user prompt
The system prompt sets the stage before any conversation begins — who the model is acting as, what it should and shouldn't do, what tools are available and how to use them. The user prompt is the actual question or request within that established context. The same user prompt can produce very different behavior depending on the system prompt it's paired with.
Why this matters more for agents than for simple chat
An agent's system prompt typically also has to specify tool-use conventions (when to call which tool), stopping conditions (when the task is complete), and error-handling behavior (what to do if a tool call fails) — a poorly specified system prompt is a common root cause of an agent that behaves inconsistently, distinct from any issue with the underlying model's capability.
Practical design guidance
Keep the system prompt focused on genuinely stable, session-wide instructions, and put task-specific or frequently-changing detail in the user prompt or retrieved context instead — a system prompt that tries to cover every possible edge case becomes hard to maintain and can dilute the instructions that actually matter most.