System prompts are instructions, not vaults
Explain why system prompts should not store secrets or enforce authorization by themselves.
The move: keep prompts behavioral, not secret-bearing. System prompts are good at setting role, tone, scope, and output format. They are poor at keeping secrets. A determined user may ask for the prompt directly, infer it indirectly, or cause the model to reveal pieces through a strange task. Treat the exact wording as potentially observable. What belongs in a system prompt A role, task boundary, formatting rule, refusal style, source policy, and reminder that untrusted content is evidence rather than instruction. What belongs outside the prompt Credentials, tokens, database names, private pricing logic, user role mappings, security exceptions, and anything…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in