AI-DLC Hands-On Workshop

Part 3 - Frameworks

HVE-CORE

AI-DLC Workflows

Specs.md

BMAD

https://github.com/microsoft/hve-core

https://github.com/awslabs/aidlc-workflows

https://github.com/fabriqaai/specs.md

https://github.com/bmad-code-org/BMAD-METHOD

• Research

• Plan

• Implement

• Inception

• Construction

• Operations

• Simple

• FIRE (Fast Intent Rapid Execution)

• Full AI-DLC (Inception, Construction, Operations)

• Multiple modules

• Persona based agents

• Modular and easy to extend

• Party mode 🥳

HVE-CORE

Group 1

Research: Ask rpi-agent to research the design space: "Research what we need to build a company directory API with Copilot. Identify requirements and decision points."

Plan: Generate a plan artifact: ADR for database choice + API structure spec

Implement: Execute the plan: migrations, Express + TypeScript scaffold, CRUD endpoints, JWT auth

AI-DLC Workflows

Group 2

You trigger the flow by saying: "Using AI-DLC ... "

Prompt: Using AI-DLC build a company directory. It needs: a PostgreSQL table (name, industry, website, notes), a REST API in Node.js/Express with TypeScript with full CRUD endpoints, JWT-based authentication (register + login), and a health check endpoint. The API goes in api/. Local PostgreSQL runs in Docker.

Review and approve the execution plan

Let it run Construction; review each artifact in aidlc-docs/ before approving the next stage

Specs.md

Group 3

Init a FIRE workflow for this workshop. You will have access to 3 agents. You invoke each agent depending on the stage that you are in

Prompt: We're building a company directory. It needs: a PostgreSQL table (name, industry, website, notes), a REST API in Node.js/Express with TypeScript with full CRUD endpoints, JWT-based authentication (register + login), and a health check endpoint. The API goes in api/. Local PostgreSQL runs in Docker.

Invoke Fire-Orchestrator with the above prompt to capture the intent

Invoke Fire-Planner to decompose into work items. The framework auto-sizes checkpoints (Autopilot / Confirm / Validate) based on complexity

Invoke Fire-Builder to execute the plan

BMAD

Group 4

When setting up, select BMAD Core Module and BMad Method Agile-AI Driven-Development. You will have a lot of different agents through SKILLS and scripts that BMAD will invoke when triggered

Prompt: Using AI-DLC build a company directory. It needs: a PostgreSQL table (name, industry, website, notes), a REST API in Node.js/Express with TypeScript with full CRUD endpoints, JWT-based authentication (register + login), and a health check endpoint. The API goes in api/. Local PostgreSQL runs in Docker.

Start with PM agent: capture requirements, acceptance criteria, user stories

Move to Architect agent: tech decisions, project structure, API design

Use Party Mode at the Architect stage: bring in the Developer and ask them to challenge the design

Switch to Developer agent: implement

AI-DLC Frameworks.loop

Let's note down our experience