Applies the *nix Agent design philosophy to agent tool interfaces. Use when asked to design, review, or refactor how an AI agent exposes tools — especially when considering function-calling vs CLI approaches, tool interface reduction, output truncation/overflow, binary guards, stderr handling, or two-layer execution/presentation architecture. Trigger phrases: "design agent tools", "single run tool", "CLI for agents", "tool interface design", "agent tool architecture", "function calling vs CLI",