TrapVLA: Trapping Vision-Language-Action Models in Configured Failure Modes | Startup Networx | Startup Networx