The enormous potential of the agent-based nature-inspired algorithm named the gravitational search algorithm (GSA) remains untapped, so it is the subject of this work. Existing GSA has provided ways to select the super-agents or prune the poor-performing agents. Here, a holistic framework is proposed by introducing a survival boost which re-establishes self-significance of a poor-performing agent and liberates all agents from an infinity trap for acceleration. It restricts the agents within the search space whose violation produces infeasible solutions. The analysis is extended to solve a number of benchmark test functions, an inverse kinematic problem of a 7 DOF robot manipulator, and a voltage regulation problem of an automatic voltage regulator system. The results and comparative studies show better performances in than some of the popular methods from existing literature (e.g., existing GSA, chaotic GSA, different variants of particle swarm optimisation, artificial bee colony, differential evolution, Jaya algorithm).