What checks should be completed before an increment is considered finished?

What checks should be completed before an increment is considered finished?

An increment is finished when its agreed user outcome works, its important failure cases are handled, and the team has evidence to support that conclusion. A coding agent saying “done” or a successful build is useful information, but neither proves the feature satisfies the product requirements.

For wedding-detail editing, completion means more than displaying a form. The saved result, cancellation behavior, access rules, and existing onboarding flow all matter.

Check the acceptance criteria directly

Walk through the increment using realistic test data:

• Open the editor and confirm saved information is displayed.
• Save valid changes and verify them after refreshing.
• Cancel an unsaved change and confirm the original value remains.
• Submit invalid input and inspect the feedback.
• Confirm an unrelated account cannot update the wedding.

Also check relevant loading, empty, and network-failure states. If the increment includes collaboration, exercise each affected role.

Review the implementation and evidence

Inspect the diff for unrelated changes, accidental secrets, missing server checks, and undocumented assumptions. Run the repository checks appropriate to the change, including targeted tests and the required build or lint steps.

Record what ran and what it demonstrated. A test skipped because a service was unavailable should be reported as unverified, not counted as a pass. Reviewers need enough detail to reproduce an actionable finding.

Close the documentation loop

Resolve blocking findings and rerun the checks affected by those fixes. Update project status, document changed API or schema behavior, and list any explicitly deferred work.

Prepare a coherent commit with a description of the completed behavior. Where deployment is part of the increment, verify the deployed version too; a local result does not establish production readiness.

This shared definition of completion helps engineering, review, QA, and future AI sessions agree on what can safely serve as the foundation for the next feature.