Offscript: Agentic Auditing of Instruction Adherence in LLMs

Large Language Models (LLMs) and generative search systems are increasingly used for information seeking by diverse populations with varying preferences for knowledge sourcing and presentation. While users can customize LLM behavior through custom instructions and behavioral prompts, no mechanism exists to evaluate whether these instructions are being followed effectively. We present Offscript, an agent-based automated auditing tool that efficiently identifies potential instruction-following failures in LLMs. In a pilot study analyzing custom instructions sourced from Reddit, Offscript detected potential deviations from instructed behavior in \(84.6\%\) of conversations, \(22.2\%\) of which were confirmed as material violations through human review. Our findings suggest that agentic auditing serves as a viable approach for evaluating compliance to behavioral instructions related to information seeking.