Agentic SDLC

Overview

Every team is adding coding agents to how they build software. Few have thought about where in the lifecycle they help most, and what has to change around them so the speed actually reaches the team. This module looks at both. In the hands-on part you build the same feature twice with a coding agent, once straight from a prompt and once with a spec and tests written up front, then compare the results and see how to test what the agent built.

What you'll cover

• ▸The inner and outer loop of development, and how agents speed up the inner loop so the slow part moves elsewhere
• ▸Why fast but shallow agent work turns into technical debt: no shared standard, too much to review, code nobody fully understands, and unclear requirements
• ▸Writing specs and tests before any code exists, so the agent has no gaps to guess at
• ▸Giving agents what they cannot work out from the code alone: a context file in the repo, and skills for tasks you repeat
• ▸Reviewing agent code: an automated first pass, checks that must pass before anything merges, and human review kept for the judgement calls

What you'll be able to do

• ✓Find where agents speed up your team's lifecycle and where the bottleneck moves to
• ✓Write specs and tests up front so an agent has no ambiguity to fill in
• ✓Build the same feature with an agent from a prompt alone and from a spec and tests, and compare the results
• ✓Set up a repo for agents, with a context layer, skills for recurring tasks and gates before anything merges