<p>This paper investigates a multi-resource job shop scheduling problem with buffer capacity constraints aimed at minimizing the makespan. We first establish a mixed-integer linear programming model to formulate the problem. Afterward, we develop an improved salp swarm algorithm (ISSA) to solve the problem. This algorithm uses a two-layer encoding method and designs a priority weight-based decoding approach to generate high-quality individuals. To obtain excellent solutions, we propose a combinatorial heuristic that includes a first in first out and a high priority weight out rules. For evolutionary operators, we design an oscillatory crossover operator and a spiral fly mutation operator to improve the exploration and exploitation capabilities of the ISSA. We further propose a new cooperation way between leaders and followers and design a nonlinear mathematical formula to generate followers for the purpose of achieving a trade-off between the exploration and exploitation capabilities of the algorithm. Additionally, we create a novel role of explorer to seek new and promising solutions and feed back the solution’s information to leaders, which aims to prevent the algorithm from getting trapped in local optima. The computational results demonstrate that the ISSA is efficient.</p>