Use Case Foundry
Home

Agent test suite

Test expected, difficult, and unsafe cases before trust

Every starter package includes tests, safe offline integration stand-ins, and a runnable test plan. Each case reports pass, fail, or blocked.

What this establishes

The decision in view.

Principle 1

A check that is declared but never exercised by a case shows as untested — a real gap, not just a missing checklist item.

Principle 2

A missing or unreadable minimum threshold conservatively assumes a 100% pass requirement rather than skipping the check.

Principle 3

Offline testing checks that the starter code is connected correctly. It does not prove performance against a live customer system.

What the test suite contains

  • Expected cases — clean examples based on real records where available.
  • Difficult cases — partial data, new items, mismatches, duplicates, and other exceptions.
  • Pass-or-fail rules — how each result is judged.
  • Blocked risky actions — tests for every tool that can change data or take action.
  • Minimum pilot score — the pass rate required before the pilot can begin.

Offline test results

ResultWhat it means
PassedEvery runnable case passed and the minimum score was met
FailedThe tests ran, but the pass rate was too low
BlockedThe tests had no runnable cases or an integration stand-in was not ready
Unsafe to runA risky action has not been covered by a real test case

Next steps

Go deeper with resources

The starter package the tests check.