Guide
Retesting agents on every change
By Yuntong Zhang, Co-founder & Head of Research, Verigrey · PhD, National University of Singapore1 min readUpdated
New LLM, new prompt, new tool call — any of them can silently break assurance. How to trigger a new pass automatically in CI/CD.
A policy that held on Monday can silently fail on Tuesday after a routine model upgrade or a one-line prompt edit. Point-in-time assurance reports don’t catch that — continuous retesting does.
This guide covers wiring an assurance run into CI/CD so any change to the model, prompt, or tool surface automatically triggers a new pass, with a worked example for a typical agent deployment pipeline.