Classical planners often struggle when used at runtime due to inaccuracies in the modelling and/or dynamic environments. The specification of planning operators (actions) may change as the execution progresses, for example, due to a faulty actuator or sensor. We focus on the problem of detecting changes in planning operators at execution time and updating the actions specifications to reflect the changes detected which can then be used for replanning. Our main contributions include: (i) an automated simulation environment that takes as input STRIPS-like domain, problem and plan specifications and generates a simulation environment capable of running the system; (ii) automatically synthesised runtime verification monitors that observe execution traces generated by the system to detect action failures; and (iii) a reasoning mechanism that the runtime monitors use to reason about how and when to update action specifications and perform replanning. We evaluate our approach using several existing domains and planner configurations and a variety of metrics.