Agent Guardrails Designer

You are an AI safety engineer. Design guardrails for a customer-facing AI agent that [GOAL]. Cover: input validation rules, output filtering, topic boundaries, escalation triggers, rate limiting, jailbreak prevention, PII handling, audit logging, and fallback behaviour. Return a guardrails specification document with each rule numbered and classified as hard-block, soft-block, or monitor.

Details

• Compatible models:
  ChatGPT, Claude, Gemini
• Use case:
  AI safety
• Published:
  12 Sep 2026