System Prompt Generator
Pick a role, audience, tone, output format, and constraints, and get a clean, well-structured system prompt you can paste into Claude, ChatGPT, or any LLM API. Updates live as you type.
Few-shot example 1 (optional - shows the model exactly what good output looks like)
Few-shot example 2 (optional)
Why each section matters:
- Role
- Anchors the model's identity and expertise. A specific role ("senior code reviewer") produces sharper, more consistent behavior than a vague one ("helpful assistant").
- Audience
- Calibrates vocabulary and depth. The same answer should look very different for an executive than for a beginner, and the model can only adjust if you tell it who is reading.
- Tone and style
- Keeps voice consistent across every response, which matters for anything user-facing or on-brand.
- Output format
- Makes responses predictable and parseable. Explicit format rules are far more reliable than hoping the model picks the right structure.
- Constraints
- Guardrails like "no speculation" and "cite sources" measurably reduce hallucinations, and "refuse out-of-scope" keeps a production assistant on task.
- Examples
- Few-shot examples are the strongest steering tool available. One good input/output pair often does more than a paragraph of instructions.