A Survey Maps LLM Agents Onto the Procedural Loop Security Analysts Already Run

arXiv 2608.28490·low signalXBlueskyLinkedInCopy link

arXiv 2608.28490 (2026-08-28, cs.CR/cs.AI) surveys LLM-based agents for software and systems security, framing the fit around the fact that security workflows are already procedural: analysts inspect heterogeneous artifacts, form hypotheses, invoke tools, interpret outputs and revise plans, which maps onto agents that plan, use tools, retain state and revise actions across multi-step work. It covers approaches, applications and assessment across that space. Value here is the assessment dimension, since it sits alongside this same listing day's empirical security work rather than duplicating it.