Agent System Prompt Auditor

It audits 5 dimensions: Role & Persona (is the assistant role defined?), Instructions (are there conflicting rules? is fallback behavior defined?), Safety & Guardrails (are disallowed behaviors specified?), Output Format (is the output format clear?), and Context & Audience (is the audience and context described?).

Without a role definition, the model has no anchor for its persona. "You are a customer support assistant for..." gives the model a clear identity, which makes its responses more consistent and predictable. Missing role definitions are the most common cause of inconsistent agent behavior.

Fallback behavior defines what the agent does when it cannot answer or complete a task. Without it, models may hallucinate answers, give confusing refusals, or behave unpredictably at the edges of their capability. Good fallback: "If you are unsure, say so and suggest where the user can get help."

No. All checks and scores are purely rules-based — keyword detection, pattern matching, and structural analysis. No AI model is involved. The same input always produces the same score.

No. This tool runs entirely in your browser. Your system prompt is never sent to any server.