Poorly damped natural subsynchronous oscillations in IBR-dominated power systems may emerge or disappear as operating conditions change, requiring stability assessment tools capable of tracking the corresponding system dynamics. This becomes particularly challenging for black-box IBRs, whose dynamic responses vary with the operating point and must be efficiently updated for near-real-time assessment. This paper presents a frequency-domain digital twin (DT) framework to address this challenge. The proposed modular bottom-up approach characterises IBR dynamics offline through dynamic frequency scans across their operating range. Parametric frequency-response models are then evaluated at the prevailing operating point and interconnected with the rest of the system representation to reconstruct the overall system dynamics. The resulting DT enables tracking of critical oscillatory modes and provides early warning of emerging poorly damped operating conditions. The framework is demonstrated on a modified IEEE 39-bus system, showing its ability to capture the evolution of critical modes as the operating point changes. Computational scalability is further assessed using modified IEEE 118- and 300-bus systems, quantifying the reconstruction requirements as system size and number of IBRs increase. Overall, the proposed framework provides a scalable basis for broader stability assessment applications while accounting for operating-point-dependent dynamics.