We wrote comprehensive GPT prompts that guide the tool in each analysis. These prompts are founded on our extensive experience in network analysis. By integrating them into our PacketSafari Analyzer platform, we can decode packets using API calls. This way, AI Shark can gather more information for the parts of the trace that are interesting. Generally, the context windows of modern LLMs are still too small to analyze very large files. We implemented several tricks to prevent supplying too much irrelevant data to AI Shark, ensuring that we do not overload the context window with unnecessary information.