“What is the smallest agentic mechanism that solves the problem?” Start in the center and work outward.

“What is the smallest agentic mechanism that solves the problem?” Start in the center and work outward.

This is basically the agentic architecture selector: start with plain intent and only climb the complexity ladder when the problem gives you a reason to. It also captures the distinction we’ve been drilling on: a skill teaches, a tool acts, a sub-agent reasons, a loop adapts, an orchestrator coordinates, and a hook enforces.