Projects are inherently subject to various uncertainties and disruptions, which are among the main causes of project completion delays and reduced reliability of the initial schedule. However, in real-world project environments, activities are not equally affected by disruptions, and ignoring these differences can lead to inefficient resource allocation and unnecessary cost overruns. This study proposes a novel approach to assessing activity vulnerability and vulnerability-based resource acquisition in the multi-mode resource-constrained project scheduling problem with stochastic activity durations. In the first stage, 11 network-based, temporal, and resource-based indicators are extracted and weighted using the Shannon entropy method. The vulnerability level of each activity is then calculated as a weighted combination of these indicators. Based on this, a heuristic algorithm is developed to acquire renewable resources, prioritizing activities with higher vulnerability and determining the purchase quantity as a function of the vulnerability index, resource tightness, and maximum resource requirement. Once the purchase quantities are determined, the proposed mixed-integer linear programming (MILP) model is solved to make scheduling and resource allocation decisions under uncertainty. Computational results show that the vulnerability-based resource acquisition strategy reduces makespan compared to no acquisition, while the two-stage heuristic-exact approach significantly reduces computational time compared to solving the MILP model directly. These findings highlight that identifying vulnerable activities plays a key role in enhancing the robustness of project scheduling.