agent-native-architecture

Agent-native architecture is a design paradigm where autonomous agents are central to the system, operating in loops to achieve feature outcomes. You need it when building self-modifying systems or applications driven by agentic loops rather than pre-coded functions.

To design tools for an LLM agent, you should build primitive tools that adhere to the principles of parity, granularity, and composability. This allows the agent to dynamically discover capabilities and execute actions within its operating loop.

The best way to architect an AI application for agent outcomes is to use an agent-native design pattern. This involves creating systems where features like organizing files are achieved by agents operating in loops, rather than relying on pre-coded deterministic functions.

System prompt engineering for autonomous agents involves crafting instructions that guide the agent's behavior within its execution loop. This Skill provides practical advice on designing these prompts to ensure agents operate effectively and achieve the desired outcomes.

Yes, you can use agent-native architecture to build a self-modifying system. This approach provides the necessary patterns and execution frameworks where autonomous agents can dynamically adjust system capabilities and behaviors through continuous agentic loops.