Citizen-Built Agents Degrade Silently: A Continuous-Assurance Framework for Low-Code Agent Sprawl
Natan Levy and Harel Berger identify a reliability gap created as non-engineering staff build agents through low-code, no-code, and conversational environments: what looks to the user like a simple productivity artifact actually depends on changing models, tools, retrieval sources, permissions, prompts, schedules, and external services, any of which can cause silent degradation long after deployment with no user edit involved. They propose a lightweight framework combining dependency mapping, readiness contracts, scheduled checks, diagnostics, and lifecycle governance, and present a prototype auditor with scenario-based assessment. For anyone running scheduled agent pipelines, the readiness-contract framing is directly transplantable.
↳ Follow the thread