Computational thinking (CT) is important for students because it is one of the 21st century’s skills. Reverse engineering pedagogy (REP) can improve students’ CT due to its ability to develop students’ cooperativity, algorithmic thinking, creativity, and problem-solving in discipline education. Thus, this study aimed to explore the effect of REP on primary school students’ CT skills in STEM learning activities. A total of 101 fifth graders in a primary school participated in the study for one semester (16 weeks), including 51 students in the experimental group (EG) with REP, and 50 students in the control group (CG) with the demonstration method (DM). The computational thinking scale (CTS) was used to measure the CT skills of students in the pretest and posttest. The Wilcoxon signed-rank test and the Mann-Whitney U test were used to analyze the data. The results verified that REP had a fine effect on the improvement of students’ CT skills compared to the DM. The findings can provide some ideas for researchers to develop students’ CT skills in STEM learning activities. Teachers can use different teaching methods to reasonably arrange teaching activities to develop primary school students’ CT skills.