Vertical Agent Design

Design principles for building accurate, domain-specific (vertical) AI agents that customers pick because they are more accurate. Use when designing or reviewing an agent's tools, system prompt, context layout, or tool architecture — deciding how many tools to expose, what belongs in the system prompt vs. loaded on demand, how to shape tool inputs/outputs, or why an agent is inaccurate or context-bloated. Core idea: structure the agent's context like a CPU memory hierarchy (L1/L2/L3). Adapted from Peter Wang's essay "Building a Good Vertical Agent."

jordangaston Updated 0 repo stars

File contents

jordangaston/claude-plugins commit 6ae6e5932f

Frequently asked questions

npx get-skill add jordangaston/vertical-agent-design