This work presents a novel method for translating natural language mission requests into structured robotic skill specifications, within the SAIRS framework (Symbiotic AI for Robotics in Industrial Settings). Grounded in a modular ontology of robotic capabilities, the system uses Large Language Models (LLMs) with prompt-based guidance to perform intent recognition and slot-filling. The architecture supports natural language input and generates executable, context-aware mission plans by referencing domain-specific knowledge. The paper describes the design of the interpretation module, skill ontology, and prompt strategies used to ensure robustness. A two-stage evaluation assessed both natural language synthesis and mission plan generation, with Meta’s LLaMA-3.3-70B-Instruct achieving the best performance. Results demonstrate the feasibility of using LLMs for adaptive, human-aligned robotic planning and lay the groundwork for future extensions in multi-agent collaboration and user adaptation.