Unlike standard agents that rely on the model's native tool_calling capabilities, this agent uses a Structured JSON Prompting architecture. This allows even "weaker" or smaller models to effectively drive external tools by generating machine-readable instructions within their standard text output.