This document is a complete technical guide for building a reusable, consultancy‑level AI‑assisted development capability using Augment Code, with a concrete case study of migrating the NVT Long‑Term Yield (LTY) web application from a Vue 2 stack to a React + Azure Functions architecture. It explains the hierarchy of Prompts, Rules, Skills, and Agents, the agentskills.io SKILL.md format, directory conventions, discovery precedence, and version‑control strategies for reusable skill libraries. It details how to configure the Model Context Protocol (MCP) for Atlassian tools (Confluence, Jira) so that Auggie can read/write live project data, and outlines Agent modes (interactive vs. auto), checkpoint safety, and token‑cost considerations. A side‑by‑side comparison with Google Antigravity highlights differing execution models while confirming skill‑format compatibility. The guide then enumerates a three‑layer skill taxonomy (framework, workflow, project‑specific) and provides concrete SKILL.md examples for React‑Azure patterns, data‑table heat‑maps, cascading filters, view‑mode toggles, agricultural charting, and NVT‑specific API contracts. It maps the reverse‑engineered NVT backend endpoints to Azure Functions, recommends caching, authentication migration, and MeiliSearch replacement, and proposes additional skills (URL‑state, accessibility, export, auth). Finally, it offers best‑practice recommendations for building, maintaining, and evolving a consultancy‑grade skill repository, including feedback loops, token budgeting, and distribution via Git and install scripts.
Topics
Hierarchy of Prompts, Rules, Skills, and Agents in Augment Code
Defines the four abstraction levels, their scopes, persistence, and how they compose to enable reusable AI‑assisted development.