Prompt Template Builder
Fill a prompt template from spreadsheet rows and get one prompt per row.
Build a structured system prompt for a chatbot, assistant or custom GPT.
The System Prompt Generator assembles a production-ready system prompt from the sections that hold up over a long conversation — role, audience, task, rules, prohibitions, tone, answer format and what to do when the answer is not known — in markdown, XML tags or plain labels.
A system prompt is the part of a conversation the user never sees and the model never forgets. It is where the assistant learns who it is, who it is talking to, what it must always do and what it must never do — and a vague one is why a support bot starts inventing refund policies on the tenth message.
This generator lays the prompt out as the sections that actually hold up in production: a role and audience, the job itself, rules and prohibitions written as short imperatives, a tone and answer format chosen from tested wording, and an explicit instruction for what to do when the answer is not known. Add the facts the model must stick to and one example exchange, and the prompt is assembled as you type.
Choose markdown headings, XML tags or plain labels for the structure. Claude tends to follow XML-tagged sections closely, most models read markdown well, and plain labels suit a Modelfile or a field with no formatting. Presets for a support agent, a code reviewer, a research assistant, a copy editor and a JSON extractor give you a working starting point.
Rules and prohibitions are kept as short imperatives, which models keep applying on the tenth turn, not only the first.
Choose a tone and an answer format and get phrasing that is specific enough to enforce.
Say plainly, ask one question, escalate or return null — the choice that stops an assistant inventing answers.
XML tags for Claude, markdown for most models, plain labels for Modelfiles.
Support agent, code reviewer, research assistant, copy editor and JSON extractor.
Pick the closest job, or start from blank. Every field stays editable.
Say who the assistant is, who it is talking to and what it is there to do. The task field matters most.
One per line. Short imperatives such as "Quote the source for every figure" hold up better than paragraphs.
Paste it into the system field of the API, a custom GPT, an assistant or a Modelfile.
The system prompt sets standing instructions for the whole conversation: identity, rules, tone and format. The user prompt is each individual request. Models give system instructions more weight and keep applying them across turns.
As long as it needs to be and no longer. Most production prompts land between 150 and 800 words. Remember it is sent with every message, so every sentence is billed on every turn — cut anything that does not change behaviour.
Both work. XML tags make section boundaries unambiguous, which Anthropic recommends for Claude, and they are useful when the prompt also contains documents. Markdown headings read naturally for GPT and Gemini. Consistency matters more than the choice.
Because models imitate what they are shown more reliably than what they are told. One short example fixes tone, length and format faster than several extra rules.
No. The prompt is assembled in your browser. Nothing is sent to a model, an API or a server by this page.
What people who use the System Prompt Generator think of it. Reviews come from visitors and are posted as written.
No ratings yet
Yours would be the first.
0 reviews
Newest first
No reviews yet
Be the first to rate the System Prompt Generator. It takes half a minute and no account.
Write the first reviewFill a prompt template from spreadsheet rows and get one prompt per row.
Write valid tool-calling JSON schemas without hand-editing braces.
Work out what an AI feature will cost per call, per day and per month.
Estimate how many tokens your text costs before you send it.
Split long documents into overlapping chunks for embeddings and RAG.
Generate robots.txt rules to block GPTBot, ClaudeBot, CCBot and other AI crawlers.
Build a Stable Diffusion negative prompt from the problems you keep getting.
Turn a rough idea into a structured prompt that gets a usable answer.