Technical debt indexes are metrics for assessing the quality of a software system. Both academic and commercial tools have begun to provide computations of such indexes based on design violations and smells (e.g., cycles among system elements). When computing a debt index for a given project, a common use case is that engineers look at the index values for spotting design issues that negatively affect system evolution and quality. In this context, those smells being critical for the system architecture should be promptly identified so as to evaluate proper remediation actions. However, the interpretation of an index value in terms of problematic smells is usually a manual and labor-intensive task for engineers. To help with this task, we propose a tool called Sen4Smells that performs an automated sensitivity analysis for a given debt index based on the evolution of both the index values and the corresponding smells across (past) system versions. The Sen4Smells output is a ranking of smells that, due to their variations or instability, are major contributors to the debt index, and thus, can impact on architecture quality. Sen4Smells is designed as a pipeline that combines information from existing tools for smell detection, predefined debt index formulas, and the Sobol method for sensitivity analysis. As a demonstration of the tool functionality, we briefly present implementations for the Arcan and Sonargraph tools with their respective debt indexes.