Skip to main content
Version: Next

CaseWriter Best Practices

Start with trustworthy evidence​

  • Select the exact source version used by the delivery team.
  • Keep stable revisions as document versions instead of unrelated duplicate uploads.
  • Remove navigation, footer, and unrelated content from URL previews.
  • Combine sources only when they describe the same scope without contradiction.
  • Never use production secrets, credentials, or personal data as generation context.

Keep generation focused​

  • Choose one to three relevant platforms.
  • Select a small set of test types that matches the run's purpose.
  • Use the minimum target count needed to cover the scope.
  • Run AI-Analyze and inspect the test-design mind map for important gaps.
  • Save presets only for configurations that are genuinely reusable.

Write executable cases​

  • Give each case one clear behavioral objective.
  • State preconditions as required starting state.
  • Use one observable action per step where practical.
  • Pair every action with a concrete expected result.
  • Avoid unsupported values, routes, roles, or business rules.
  • Keep priority and complexity aligned with risk and effort.

Treat AI output as a draft​

Generated cases begin in Pending Review for a reason. Read the complete case, compare it with the evidence, remove hallucinated behavior, and resolve duplicates. Generation quality and repository approval are separate concerns.

Regenerate with specific instructions​

Describe the exact deficiency, such as a missing boundary, wrong actor, unsupported result, or required error path. Then compare Original and AI Regenerated versions before choosing Keep Original, Duplicate, or Use Regenerated.

Preserve the review gate​

  • Approve only after reading the entire case.
  • Add actionable comments when rejecting.
  • Reopen approved cases before editing them for changed behavior.
  • Do not bulk-approve merely because cases came from the same run.
  • Save only approved cases to the repository.

Maintain a predictable repository​

  • Organize by stable product or feature boundaries.
  • Search before creating a new test set.
  • Avoid using deep hierarchy for temporary statuses or releases.
  • Use consistent labels for cross-cutting dimensions.
  • Revalidate cases when source requirements change.
  • Remove duplicates carefully and only after checking downstream use.

Import and export safely​

  • Use the current Excel template.
  • Test an import with a small sample first.
  • Treat CSV import as TestRail CSV unless the UI explicitly adds another format.
  • Verify field mappings, step order, and destination sets after import.
  • Remember that exporting generated cases does not approve or publish them.

Operational checklist​

Before generation:

  • Correct project and source version
  • Valid model and integrations
  • Focused platforms and test types
  • Valid 100% distribution totals
  • No sensitive input data

Before approval:

  • Evidence-aligned content
  • Executable steps and observable results
  • Correct priority and complexity
  • No duplicates or sensitive output
  • Useful reviewer history

Before repository save:

  • Every selected case is Approved
  • Correct destination test set
  • Repository organization remains understandable
  • Saved cases verified in the repository