Probabilistic risk assessment (PRA) tools have been in use for over six decades, providing essential data to support risk-informed decision making. However, like all tools, PRA tools must keep pace with advances in computing technology. Here, we propose a systematic methodology to diagnose and enhance PRA tools. The diagnostics phase of this methodology consists of model generation, benchmarking, standard profiling, and deeper profiling. This phase results in representative PRA models, tool performance assessments, identification of code hot spots needing improvement, and a verification platform for comparing PRA tools. The diagnostics findings guide an improvement strategy that may involve optimization, parallel computing, or a combination of both. Demonstration results show speedups of up to five times for a single model, underscoring the significant impact of utilizing available resources for large PRA models. Although the demonstration focuses on the open-source quantification engine SCRAM-CPP, the methodology can be adapted to other PRA tools with minimal effort.