Module 08: SDLC

Run a software change through the full lifecycle: problem, design, build, test, review, release, observe, and learn.

SDLC is the professional control system for change.

problem -> requirements -> design -> implementation -> verification
        -> release -> monitoring -> retrospective

Without SDLC, teams confuse activity with progress.

SDLC is like constructing a bridge. You do not start pouring concrete because someone said “make crossing faster.” You survey, design, calculate load, build, inspect, open traffic, monitor cracks, and maintain it.

“Agents can summarize ticket history and draft a response, but a human must approve before anything is sent.”

This is not just an AI feature. It touches:

• Data access.
• Prompt/context construction.
• Model failure.
• Human approval.
• Audit logs.
• Privacy.
• UI state.
• Cost.
• Deployment.

Use an agent for design critique:

Review this SDLC plan for an AI-assisted support response feature. Find missing
requirements, privacy risks, human approval gaps, audit gaps, tests, release
risks, and monitoring requirements.

• What user pain does the feature solve?
• What data should the model see?
• What should never be automated?
• What must be audited?
• What tests prove approval is required?
• What monitoring catches bad model behavior?
• What rollback disables the feature safely?

1. Create portfolio/08-sdlc/design-doc.md.
2. Write a problem statement.
3. Write acceptance criteria.
4. Draw the approval flow.
5. Write a data access policy.
6. Write a test plan.
7. Write a release plan.
8. Write a rollback plan.
9. Ask an agent to critique the plan.
10. Revise it.

• The feature is scoped.
• Human approval is enforceable.
• Privacy and audit are addressed.
• Tests and release gates are defined.
• Rollback is clear.