Raw view: Write the Failing Test First

No invisible, zero-width, or bidi-control characters detected.

Write one failing test for the behavior below. Write the test only. Do not write or sketch the implementation; the whole point is that the test fails for the right reason before any code exists.

<behavior>
{{BEHAVIOR}}
</behavior>

<unit_under_test>
{{UNIT_UNDER_TEST}}
</unit_under_test>

<existing_test>
{{EXISTING_TEST}}
</existing_test>

Language: {{LANGUAGE}}. Framework: {{TEST_FRAMEWORK}}. Tests run with `{{TEST_COMMAND}}`.

Match <existing_test> exactly in file layout, naming convention, import style, fixture/setup pattern, and assertion library. A test that passes but looks foreign to the codebase is a worse outcome than one that needs a tweak, because reviewers will ask for a rewrite.

Rules for the test itself:

1. Test exactly one observable behavior. If <behavior> describes several, pick the smallest one that must be true first and list the rest under "Next tests".
2. Name the test as a sentence about behavior, in the project's naming style: `returns_empty_list_when_no_matches`, `it('rejects expired tokens')`, not `test1` or `testFunction`.
3. Arrange, act, assert, in that order, visually separated. One act. Assertions on the result, not on internals or mocks, unless the behavior is "calls X".
4. Use literal, boring inputs. `"alice@example.com"` beats `generateEmail()`. Hardcode the expected value; do not compute it with the same logic the implementation would use.
5. Mock only what crosses a process boundary (network, clock, filesystem, randomness). If <unit_under_test> does not exist yet, import it as if it did, with the simplest signature that satisfies the test.
6. The test must fail today. Before finishing, trace through it against the current state of <unit_under_test> and name the exact failure you expect (import error, assertion mismatch, exception). If you cannot name a failure, the test is not testing new behavior; rewrite it.

## Output format

**Test file** - one fenced code block containing the complete, runnable test file (or the new test function if it belongs in an existing file, with a comment stating the target file path on line 1).

**Expected failure** - one or two sentences: the command to run, and the exact error or assertion message you expect right now, before implementation.

**Why this test and not another** - 2-3 sentences on the behavior boundary you chose and any ambiguity in <behavior> you resolved (state the assumption you made).

**Next tests** - bullet list of further behaviors from <behavior> not covered here, each as a one-line test name. Write "None" if the behavior was fully covered.