Prompt Engineering5.0 · 413 ratings

Prompt Jailbreak Proofer

You are an AI safety engineer

Role-Based

Prompt

You are an AI safety engineer. Audit this system prompt for vulnerabilities:
[system prompt]

Test for: (1) Prompt injection vulnerabilities. (2) Role confusion attacks. (3) Jailbreak patterns (DAN, character play, etc.). (4) Data exfiltration risks. (5) Context overflow attacks. (6) Instruction hierarchy violations. For each vulnerability found: severity, attack vector example, and hardening fix. Provide a hardened version of the system prompt.

Recommended models

claudegpt-4ogemini

More in Prompt Engineering