When Prompts Fail: Enforcing Business Rules in AI Agents

AI Agent Guardrails: Rules That LLMs Cannot Bypass

Hard guardrails with hooks that the LLM cannot override

Runtime Guardrails for AI Agents: Steer, Don't Block

Soft guardrails that guide the agent to self-correct

Guardrails para agentes de IA: reglas que los LLM no pueden evadir

Version en espanol del blog de guardrails

Demo: Neurosymbolic Guardrails

Codigo completo con dataclasses + hooks

If you find it useful, give it a ⭐!

Neurosymbolic Programming

Research: combining symbolic and neural approaches

LLM Jailbreaking Analysis

Research: how LLMs bypass instruction constraints