Assessment Bank

Use these questions to test whether the learner can think like an engineer. Answers should include examples, diagrams, or evidence where possible.

• Explain what happens when a user clicks “Submit” on a web form.
• What parts of a system exist outside source code?
• What is the difference between a product, app, service, and system?
• Why is a database not the same thing as a backend?
• What is a boundary? Give three examples.
• What is a failure mode? Give five for a login flow.

• What is the difference between pure logic and a side effect?
• Why should parsing, validation, business rules, and storage be separated?
• What makes a function hard to test?
• What data shape would represent a support ticket?
• What should happen when a function receives invalid input?
• Why do names matter in code?

• What is durable state?
• What is temporary state?
• What should be append-only in an approval workflow?
• What is a migration?
• Why can migrations break production?
• What is an index and what tradeoff does it create?
• What is one database constraint you would add to prevent bad data?

• Explain 401 versus 403.
• Why should a backend never trust the browser?
• What makes an API contract complete?
• What does idempotency mean?
• When should a create endpoint use an idempotency key?
• What error should an API return for invalid status transition?

• What is the difference between a screen and a feature?
• List all states a ticket creation form should handle.
• What is an acceptance criterion?
• How would you prevent duplicate submit?
• How would you explain a feature from UI to database?

• What is the difference between a symptom and root cause?
• Why is reproduction important?
• What should a useful log line include?
• Write a regression test idea for duplicate ticket creation.
• What evidence proves a bug fix?
• What belongs in a postmortem?

• What is a deployable artifact?
• What is environment configuration?
• Why should secrets not be in Git?
• What should a health endpoint prove?
• What is a rollback?
• Why can deployment succeed while user behavior is broken?

• What comes before implementation?
• What should a design doc include?
• What should be reviewed before release?
• What makes a requirement testable?
• How do you know a release worked?
• What should be captured in a retrospective?

• What context should you give an agent before asking for code?
• What makes an agent task too broad?
• What agent output should be rejected?
• How do you verify generated code?
• Why is “the agent said so” not evidence?
• What is one good use of an explorer agent?

• How do you turn vague user pain into requirements?
• Why is a demo not the same as production?
• What makes a workflow map useful?
• What should stay human-controlled in an AI workflow?
• How do you build trust with users after deployment?
• What should a handoff note include?

A user logs in and sees an empty dashboard, but they expect data.

Ask:

• Is this a frontend, API, auth, data, or permissions issue?
• What request would you inspect first?
• What database query would you check?
• What log line would help?
• What would you tell the user while investigating?

Document upload takes 45 seconds.

Ask:

• Should upload and processing be synchronous?
• Where would you add a queue?
• What status should the UI show?
• What metrics would you collect?
• What failure state should exist?

An AI-generated customer response was incorrect.

Ask:

• Was human approval required?
• What data did the model see?
• Is there an audit log?
• Can the feature be disabled?
• What tests or evaluations were missing?
• What should the postmortem include?

After deployment, the API returns 500 on ticket creation.

Ask:

• Did code and schema deploy in the right order?
• What error appears in logs?
• Can rollback work with the new schema?
• Is data corrupted?
• What migration safety practice was missing?

A strong answer names:

• Component.
• Boundary.
• Data involved.
• Failure mode.
• Evidence.
• Fix or next step.