A new linear-time runtime verifier for large language model (LLM) conversations has been developed to enhance accuracy by maintaining an explicit dependency graph, addressing issues with context manipulation. The verifier demonstrates improved performance over existing models in various benchmarks, ensuring more reliable conversational outcomes.