Ce Agent Native Architecture

This is the architectural playbook for building apps where agents aren't helpers but the primary mechanism for getting things done. Instead of writing features as code, you give Claude atomic tools and define outcomes in prompts. The skill walks you through designing these systems from scratch: how to structure tools for maximum composability, when to graduate from primitives to domain tools, how to inject runtime context into agent prompts, and how to ensure parity between what users can do in your UI and what agents can achieve autonomously. The reference material covers everything from execution patterns and completion signals to mobile-specific challenges like checkpoint/resume. If you're building something where the agent loop is the product, not just an enhancement, this is the mental model.