Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

System prompt

A system prompt is the block of instructions an application sends to a language model before the user's messages. It sets the model's role, rules, tone and output format for the whole conversation.

Also called: system message, instructions

By Divya Sharma · Checked Nov 15, 2026 · How we review

Most chat APIs split a request into two parts: a system prompt written by the developer and a list of user and assistant turns. The system prompt is where an application says who the model is acting as, what it may and may not do, and how answers should look. Some APIs call it the system message or simply instructions.

Why it matters

The system prompt is the main lever a builder has over a model's behaviour without fine-tuning. Anthropic's documentation notes that setting a role there changes the model's behaviour and tone, and that even one sentence can make a difference. A support bot, a coding helper and a translator can run on the same model and differ only in their system prompts.

A concrete example

A system prompt for an internal help desk might say: you answer questions about our VPN and laptops, keep replies short, and if you are unsure, tell the user to open a ticket. Every user message is then read in that frame.

What it is not

A system prompt is guidance, not a security boundary. Text that reaches the model through user input, documents or tool results can try to override it, which is the core of prompt injection. Put real limits, such as permissions and data access, in code and guardrails, not only in instructions.

It is also only one part of what the model sees. Context engineering treats the system prompt as one input among retrieved documents, tool definitions and history, all competing for space in the context window.

Related terms

Prompt engineering
Prompt engineering is the practice of writing, structuring and testing the instructions and examples given to a language model so that it produces the output an application needs, without changing the model's weights.
Prompt injection
Prompt injection is an attack where text fed to a language model, typed by a user or hidden in a web page, file or tool result, changes the model's behaviour in ways its developer did not intend.
Context engineering
Context engineering is the practice of choosing and maintaining everything that goes into a language model's context window during a task: instructions, retrieved documents, tool definitions, tool results and conversation history.
Guardrails (AI)
Guardrails are runtime checks placed around a language model that inspect inputs and outputs and block, rewrite or redirect them, so an application stays on topic, avoids harmful content and follows rules the model alone would not reliably keep.
Context window
A context window is the maximum amount of text, counted in tokens, that a language model can consider at once, covering the system prompt, conversation, documents, tool results and the response it writes.

Sources