System Prompt
The system prompt is the standing instruction a language model receives before any user message: it defines the model's role, tone, limits, output format and the tools it may use, and it shapes every answer in the conversation.
No jargon, no math
In plain words, what is it?
Every message you send to an AI lands on top of a hidden layer of instructions. The system prompt is that layer: a block of text the application sets once and the user rarely sees. In it you write who the model is (“you are a procurement assistant”), what it must never do (“never invent part numbers”), how it should format answers, and which tools it can call. The model treats it as the highest-authority context, so a well-written system prompt is the cheapest reliability tool you have: most “the AI went off script” incidents trace back to a vague one. Its weakness: it is still just text. A determined user can attack it (prompt injection), and a foggy instruction (“be concise”) loses against a long, urgent user message. Precision beats length: concrete rules (“answer in at most 5 bullet points”) survive pressure that slogans do not.
Terms that live nearby.
Prompt
A prompt is everything you hand to a language model in one call: instructions, context and the question itself. The model answers by continuing exactly that text, and nothing else.
ReadPrompt Engineering
Prompt engineering is the practice of writing prompts so a language model performs a task reliably: clear instructions, the right context, worked examples of the desired output, and tests against real cases before you trust it.
ReadPrompt Injection
Prompt injection is an attack where hidden instructions are planted inside content an AI reads (an email, a web page, a document), so the model executes a stranger's commands as if they were yours.
Read§ 03
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.