What the term really means
A primary instruction defining the role, rules, limits and behaviour of an AI model or agent.
How it works in practice
A system prompt sets application role and priorities, such as never inventing order status and always using an approved source.
The decision to make before implementation
It must not contain secrets or act as the sole safeguard because users and external data may try to override it.
How to verify that it works
Version the instruction and test conflicts, refusals, languages and leakage after every model or tool change. Compare results with an agreed baseline and review routine cases, difficult exceptions and human hand-offs separately. A practical System Prompt test should have an owner, a review date and a recorded example of an outcome the team will not accept.
The PAR HOUSE Agency approach
We approach System Prompt from the workflow rather than a tool demonstration. It must not contain secrets or act as the sole safeguard because users and external data may try to override it. We then build a small measurable scope, record assumptions and expand only after quality review.
