Hi Grafana community,
GrafanaGhost raised the question our graduation research at Cairo University measures: when can text in logs steer an assistant, and what stops it? We’re building an open benchmark on AIOpsLab for AI on-call agents, comparing input filtering, privilege separation, a policy check before acting, and health checks with rollback.
We use Grafana Cloud’s free tier (with the Assistant) on our testbed. Could a DevRel or Assistant-team member spare 20 minutes? Anything product-specific we’d only test in our own tenant and report through your security channel first.
Thank you,
Omar Betawy
ReliOps (graduation research project)
Faculty of Computing & AI, Cairo University
20230263@stud.fci-cu.edu.eg