Skip to content

LLM Agent Design Architecture

Main Navigation Home Preface Five layers

Skills

MCP

English

English

Appearance

Sidebar Navigation

Preface

Questions and scope

Part I

Constraint summary

Part II

Five layers

Placement

Part III

Skills

Skills procedures and examples

Skill Design Guide

How to Create Skills

Distilling Skills from Conversations

How to Use Skills

Skill Use Cases

Anti-patterns

MCP vs Skills

Showcase

MCP

MCP procedures and catalogue

Catalog

Security

Semantic Layer

Development

Doctrine

Memory

Agent

Agent topics

Agent Taxonomy

Sub-agents

Sub-agent vs Skills

Quality Gate Pattern

Multi-Agent / Agent Teams

What is A2A

Agent Identity

Part IV

Patterns

Limits

Physical world

Prompt decomposition

Strategy and rollout

Composition Patterns

MCP Family

Deterministic Verdicts

Local LLM Workspace Mapping

Harness Engineering Mapping

Hooks

Proposal vs. Binding

Permission vs. Authority

Weight vs. Context Specialization

Routing vs. Cascading

Agent Loop Patterns

Discovery vs. Production

Loop Engineering

Appendix

FAQ (3-line answers)

Scope (see Preface)

MCP vs Skills

Agent/Sub-agent/Skill/MCP comparison

Information

Overview

Architecture Map

Workflows

Development Phases

Patterns Overview

Translation

Spec Verification

Compliance

Dev Support

Doc Generation

Multi-Agent

Issue→Deploy Autonomy (Meta + Sub-agent)

Glossary

Outputs

Old paths (Concepts)

Overview

Vision (→ Constraint Summary)

Reference Sources (→ Placement)

Architecture (→ Five Layers)

AI Design Patterns (→ Patterns)

Solving AI Limitations (→ Limits)

Physical AI (→ Physical world)

Doctrine & Intent (→ Doctrine)

Memory & Knowledge (→ Memory)

Prompt Decomposition (→ Prompt decomposition)

On this page

To the Preface

This page has moved to the Preface. You will be redirected automatically.

Edit this page

Last updated:

Pager

Previous pageOutputs

Next pageVision (→ Constraint Summary)

Released under the MIT License.

To the Preface ​

To the Preface