According to the basic definition of diagnosis, all existing local diagnosis models can be modeled as directed graphs. Consequently, within the structure of local diagnosis based on comparison models, bidirectional edges may exist between some adjacent processors, which results in the absence of the current local diagnostic structure in some directed graphs. Therefore, this article proposes a novel directed tree structure \(\overrightarrow{T}(u,m)\) based on a comparison model. Using this directed tree structure, we have achieved local diagnosis in the unidirectional graphs. Additionally, by applying this structure to unidirectional star graphs, we have discovered that any vertex \(u\) with a directed tree structure in the unidirectional star graph \({\text{US}}_{n}\) (n ≥ 7) is m-locally diagnosable, where \(m = \left\lfloor {\frac{n - 1}{2}} \right\rfloor\) , if \(u\) is an even permutation; \(m = \left\lceil {\frac{n - 1}{2}} \right\rceil\) , if \(u\) is an odd permutation.