Government Dev Track
Development for a government agency with AI coding agents. Deploy a Next.js reference app to Azure Government, AWS GovCloud or Google Cloud, then produce SBOMs, write control statements, modernize legacy code, and add AI features a person can overrule
The full track: the 13 foundation modules below, some marked Quick read, Optional or Reference, then AI-Accelerated Government Development (37 lessons).
The shared modules cover CLI fundamentals. Skip what you already know, focus on tool setup and orchestration, then dive into the applied module: setup plus 36 lessons on a reference app any agency can land on its own cloud and code host, ending in the modernization and AI work agencies are doing now.
First session
The 3 lessons, about 85 minutes, to take first; each is also part of the full track below.
- 1. The Dev Workbench 15 min
Set up a bash or zsh terminal (WSL2 on Windows) and an AI CLI tool for this track.
- 2.
Confirm which AI tools and model endpoints your agency approved, before using one.
- 3.
The reference app, your platform, and the project memory file the agent reads.
First session ยท 3 lessons ยท about 85 min in all
Foundation Modules
Choosing Your AI Surface & Workflow
Command Line Basics
Your Agent In Depth: Desktop App & CLI
Planning & Orchestration
Ship It
The Prompting Playbook
Agentic Engineering
Cheat Sheets (Reference)
Applied: AI-Accelerated Government Development
Setup plus 36 lessons across 8 workflow lanesPlatform Foundation
Approved tools and endpoints, the reference app and its platform seam, provider-neutral sign-in, the app shell, and role enforcement: the base every feature runs on
Feature Development
Define and integrate a feature, errors as a contract, database access with no typed password, mutations with visible failure, provenance, and audience-tiered views
Quality & Conformance
Vitest, Playwright, accessibility to Section 508 and Title II, and bringing an off-pattern module up to standard
Infrastructure & Delivery
One OpenTelemetry path, a code-host workflow with no long-lived tokens, one image deployed anywhere, and the pipeline stages any CI can run
The Work Agencies Do Now
SBOMs and build evidence, control statements a human signs, reading and strangling a legacy system, keeping its database while the app is replaced, agent-driven parity testing, and AI features a caseworker can overrule
More Legacy Stacks and App Types
The same read, record and prove method for Oracle and Db2, Java EE, .NET Framework and Access, then the systems around the intake: the nightly batch export, case workflows, a public status portal and the daily report
Land It on Your Platform
Read the cloud lesson your agency uses, then the code-host lesson: the same seam landed on Azure Government, AWS GovCloud or Google Cloud Assured Workloads, and on GitHub, GitLab, Azure DevOps or Bitbucket
Capstone
The intake rebuild end to end, checked by someone who did not build it