Purpose : The primary objective of this research is to present the Slime Mould Algorithm with Foraging Risk (SMAFR), an enhanced version of the Slime Mould Algorithm (SMA). This development aims to address the inherent limitations of many metaheuristic algorithms, such as slow convergence and susceptibility to local optima. Design/methodology/approach : Drawing inspiration from the risk-related foraging behaviour observed in natural slime moulds, the SMAFR incorporates these behaviours into the algorithm's search mechanics. Through environmental parameters and mechanisms that mimic the slime mould's response to risk when seeking food, the algorithm is designed to balance exploration and exploitation. Findings: The SMAFR demonstrated improved efficiency over the original SMA. When rigorously assessed against three constrained real-world engineering problems, SMAFR exhibited superior performance in comparison to six other well-established metaheuristic algorithms. Statistical analyses of the results validated the effectiveness and robustness of the SMAFR in a variety of optimization scenarios.  Research limitations/implications : While SMAFR exhibited strong performance in the tested scenarios, it's crucial to recognize the No Free Lunch (NFL) theorem's implications, suggesting that no single optimization technique is universally superior. The performance of the SMAFR, like all algorithms, may vary based on the specific optimization problem. Practical implications : The SMAFR offers an advanced solution for a range of optimization problems, especially those that require a balance between exploration and exploitation. Its design, influenced by real-world biological behaviours, lends it an edge in certain optimization scenarios, making it a viable option for professionals and researchers in relevant fields. Originality/value: The incorporation of foraging risk behaviour, inspired by actual slime mould behaviour, into a metaheuristic algorithm is the cornerstone of the SMAFR's originality. This unique approach not only enhances the algorithm's efficiency but also brings it closer to mimicking real-world adaptive behaviours. The presented work thus offers significant value by providing an innovative solution to tackle prevalent challenges in optimization algorithms.