We study the problem of local causal discovery learning which identifies direct causes and effects of a target variable of interest in a causal network. The existing constraint-based local causal discovery approaches are inefficient, since these approaches do not take a triangular structure formed by a given variable and its child variables into account in learning local causal structure, and hence need to spend much time in distinguishing several direct effects. Additionally, these approaches depend on the standard MB (Markov Blanket) or PC (Parent and Children) discovery algorithms which demand to conduct lots of conditional independence tests to obtain the MB or PC sets. To overcome the above problems, in this paper, we propose a novel Efficient Local Causal Discovery algorithm via MB (ELCD) to identify direct causes and effects of a given variable. More specifically, we design a new algorithm for Efficient Oriented MB discovery, name EOMB. EOMB not only utilizes fewer conditional independence tests to identify MB, but also is able to identify more direct effects of a given variable with the help of triangular causal structures and determine several direct causes as much as possible. In addition, based on the proposed EOMB, ELCD is presented to learn a local causal structure around a target variable. The benefits of ELCD are that it not only can determine the direct causes and effects of a given variable accurately, but also runs faster than other local causal discovery algorithms. Experimental results on eight Bayesian networks (BNs) show that our proposed approach performs better than state-of-the-art baseline methods.