arXiv:2610.42047v1 Announce Type: new Abstract: Automatic generation control (AGC) stabilizes frequency and scheduled power exchanges among interconnected areas, yet its reliance on communication networks introduces cyber-vulnerabilities. This work examines coordinated AGC cyber-attacks in the presence of remedial action schemes (RAS) and identifies operating conditions under which an attacker can drive grid frequency beyond safe limits while evading RAS. A feedback-based attack policy is used to manipulate AGC signals and steer system frequency toward unsafe trajectories. Then, a large nested-loop parametric sweep---spanning attacked-area inertia relative to its interconnected neighboring area, AGC capacity versus maximum tie-line transfer, AGC update rate, load damping, and bi-directional power transfers---generates over 1.45 million scenarios. These are classified using a decision tree to establish guidelines to distinguish successful from unsuccessful attacks based on frequency deviations. The resulting vulnerability guidelines are validated on a three-area system using phasor-domain simulations with IEEE 9-bus area models, detailed turbine--governor dynamics, and under/over-frequency load and generation shedding. Results show that specific combinations of AGC capacity, inertia distribution, and pre-attack tie-line loading increase the likelihood of successful attacks, offering operators actionable criteria for identifying vulnerable operating conditions.