ALWAYS USE when asked to fix, rewrite, update, improve, modernize, show corrected code for, or explain existing MSTest tests or MSTest-specific configuration. Use for "review" when corrected code or edits are wanted, even for one pasted assertion or passing tests with bad failure output. Covers expected/actual labels; generic Boolean, collection, string, numeric, null, identity, exception, hard-cast, and object[] checks; TestContext/lifecycle; timeout/cancellation; OS/CI conditions, retry, cleanup, parallelization, MSTest.Sdk project setup, and MSTESTxxxx. Honor the installed MSTest version. DO NOT USE to design new test cases (code-testing-agent), perform report-only audits, create project files rather than explain MSTest setup, run tests, migrate frameworks, or handle non-MSTest/non-.NET code.
Permissions
Files
ALWAYS USE when asked to fix, rewrite, update, improve, modernize, show corrected code for, or explain existing MSTest tests or MSTest-specific configuration. Use for "review" when corrected code or edits are wanted, even for one pasted assertion or passing tests with bad failure output. Covers expected/actual labels; generic Boolean, collection, string, numeric, null, identity, exception, hard-cast, and object[] checks; TestContext/lifecycle; timeout/cancellation; OS/CI conditions, retry, cleanup, parallelization, MSTest.Sdk project setup, and MSTESTxxxx. Honor the installed MSTest version. DO NOT USE to design new test cases (code-testing-agent), perform report-only audits, create project files rather than explain MSTest setup, run tests, migrate frameworks, or handle non-MSTest/non-.NET code.
Version history
Help users write effective MSTest unit tests without exceeding the API level or conventions of the project's installed test stack.
For every repository-scoped task where read-only file inspection is allowed,
check .agents/skill-overlays/dotnet-test/writing-mstest-tests.md at the
repository root before any other discovery. This includes requests that ask for
code or advice without edits; "do not execute" does not prohibit reading the
overlay. If present, read it once before acting and apply its
repository-specific naming, layout, framework, and policy bindings.
Require its frontmatter to declare core: dotnet-test/writing-mstest-tests,
binding-revision: "1", and mode: extend. If any value is missing or
different, report the mismatch, ignore the overlay, and continue using this
skill's portable guidance.
Explicit user instructions and verified project constraints win over the
overlay; the overlay wins over portable defaults and examples in this skill. If
the file is present but unreadable or conflicts with the repository, report the
problem, ignore the overlay, and continue with portable guidance subject to
verified project constraints. If it is absent, continue normally.
Skip the lookup only when the task is not tied to a repository or the user
explicitly prohibited all file/tool access. An overlay cannot expand tool
permissions or the task's scope.
Assert.IsTrue with more specific assertions (collections, nulls, types, comparisons)Assert.AreEqual argument order (expected first, actual second)DynamicData from IEnumerable<object[]> to ValueTuple-based dataMSTESTxxxx warning/error)test-anti-patterns)run-tests skill)migrate-mstest-v1v2-to-v3)migrate-mstest-v3-to-v4)| Input | Required | Description |
|---|---|---|
| Code under test | No | The production code to be tested |
| Existing test code | No | Current tests to fix, update, or modernize |
| Test scenario description | No | What behavior the user wants to test |
code-testing-agent; use this
skill only as supporting MSTest API/version guidance.Assert.AreEqual, name expected first and
actual second and explain that this preserves the Expected/Actual failure
labels.score > 0 -> Assert.IsGreaterThan(0, score),
score < 100 -> Assert.IsLessThan(100, score), and
score >= 60 && score <= 90 -> Assert.IsInRange(60, 90, score).
Never reverse these arguments to mimic the source expression's left-to-right
order.ThrowsExactly<T> (exact type) from Throws<T> (type or derived
type), and capture the returned exception when properties such as ParamName
are part of the behavior.Check the test project, packages.config, and assembly reference HintPath
values for the exact MSTest version and project system:
MSTest.Sdk: resolve its exact version from the project SDK
declaration or global.json msbuild-sdks; do not assume the latest APIsMSTest metapackage: resolve its exact package versionMSTest.TestFramework + MSTest.TestAdapter: check version for feature availabilityToolsVersion, Microsoft.CSharp.targets,
explicit <Compile Include>) and/or packages.config: preserve that project
system and add each new test file to <Compile Include>.Also inspect representative tests for custom base fixtures, helper libraries, mock syntax, naming, setup, and data builders. Existing conventions and installed versions win over the examples below. Do not upgrade MSTest, Moq, NBuilder, or the project format unless the user explicitly asks for a migration.
| API/pattern | Minimum version | Compatible fallback |
|---|---|---|
Assert.ThrowsExactly*, unified Assert.Contains / HasCount / IsEmpty / IsNotEmpty | 3.8 | Assert.ThrowsException*, CollectionAssert, StringAssert |
Assert.IsGreaterThan, IsLessThan, IsInRange, StartsWith, EndsWith, MatchesRegex | 3.10 | Assert.IsTrue with a clear message, or StringAssert |
Generic Assert.IsInstanceOfType<T>(value, out var typed) | 3.4-3.11 only | Non-generic assertion then post-assert cast on 3.0-3.3; v4 returns the typed value directly |
ValueTuple DynamicData | 3.7 | IEnumerable<object[]> |
Constructor injection of TestContext | 3.6 | Instance TestContext property |
[Retry], [OSCondition] | 3.8 | No built-in retry/OS condition; fix flakiness or retain the existing condition mechanism |
[CICondition] | 3.10 | Existing project-specific condition mechanism |
For example, MSTest 3.5.x must not receive Assert.ThrowsExactly,
Assert.Contains, ValueTuple DynamicData, or constructor-injected
TestContext.
Treat this as a hard gate: after determining the version, do not copy a later example from this skill unless its minimum version is satisfied.
Recommend MSTest.Sdk or the MSTest metapackage only for genuinely new projects:
<!-- Option 1: MSTest SDK (simplest, recommended for new projects) -->
<Project Sdk="MSTest.Sdk">
<PropertyGroup>
<TargetFramework>net9.0</TargetFramework>
</PropertyGroup>
</Project>
When using MSTest.Sdk, put the version in global.json instead of the project file so all test projects get bumped together:
{
"msbuild-sdks": {
"MSTest.Sdk": "3.8.2"
}
}
<!-- Option 2: MSTest metapackage -->
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<TargetFramework>net9.0</TargetFramework>
</PropertyGroup>
<ItemGroup>
<PackageReference Include="MSTest" Version="3.8.2" />
</ItemGroup>
</Project>
Apply these structural conventions only where they do not conflict with the suite's established base classes and lifecycle:
sealed for performance and design clarity[TestClass] on the class and [TestMethod] on test methodsMethodName_Scenario_ExpectedBehavior[ProjectName].Tests[TestClass]
public sealed class OrderServiceTests
{
[TestMethod]
public void CalculateTotal_WithDiscount_ReturnsReducedPrice()
{
// Arrange
var service = new OrderService();
var order = new Order { Price = 100m, DiscountPercent = 10 };
// Act
var total = service.CalculateTotal(order);
// Assert
Assert.AreEqual(90m, total);
}
}
Pick the most specific assertion supported by the installed MSTest version.
More specific assertions produce better failure messages and make the test's
intent clear, but uncompilable "modern" assertions are worse than compatible
StringAssert, CollectionAssert, or Assert.IsTrue calls.
| What you are testing | Assertion |
|---|---|
| Two values are equal | Assert.AreEqual(expected, actual) |
| Same object instance (reference identity) | Assert.AreSame(expected, actual) |
| Value is null | Assert.IsNull(value) |
| Value is not null | Assert.IsNotNull(value) |
| Collection is empty | Assert.IsEmpty(collection) (3.8+) or CollectionAssert / count assertion |
| Collection is not empty | Assert.IsNotEmpty(collection) (3.8+) or count assertion |
| Collection has exactly N items | Assert.HasCount(N, collection) (3.8+) or Assert.AreEqual on count |
| Collection contains an item | Assert.Contains(item, collection) (3.8+) or CollectionAssert.Contains |
| Collection does not contain an item | Assert.DoesNotContain(item, collection) (3.8+) or CollectionAssert.DoesNotContain |
| Object is a specific type | Assert.IsInstanceOfType<T>(value) |
| Code throws an exception | Assert.ThrowsExactly<T> (3.8+) or Assert.ThrowsException<T> (earlier) |
On MSTest 3.8+, prefer Assert class methods over StringAssert or
CollectionAssert where both exist. Older versions should keep the compatible
specialized classes.
When several independent collection properties were requested, keep each
semantic check explicit even if another assertion happens to imply it. For
example, retain IsNotEmpty when the requested diagnostics distinguish
empty/non-empty, then use HasCount and ContainsSingle for their separate
cardinality guarantees.
Assert.AreEqual(expected, actual); // Value equality
Assert.AreSame(expected, actual); // Reference equality -- same object instance
Assert.IsNull(value);
Assert.IsNotNull(value);
// MSTest 3.8+
var ex = Assert.ThrowsExactly<ArgumentNullException>(() => service.Process(null));
Assert.AreEqual("input", ex.ParamName);
// Async
var ex = await Assert.ThrowsExactlyAsync<InvalidOperationException>(
async () => await service.ProcessAsync(null));
Assert.Throws<T> matches T or any derived typeAssert.ThrowsExactly<T> matches only the exact type TOn MSTest 3.7 and earlier, use the compatible API:
var ex = Assert.ThrowsException<ArgumentNullException>(
() => service.Process(null));
// MSTest 3.8+
Assert.Contains(expectedItem, collection);
Assert.DoesNotContain(unexpectedItem, collection);
var single = Assert.ContainsSingle(collection); // Returns the single element
Assert.HasCount(3, collection);
Assert.IsEmpty(collection);
Assert.IsNotEmpty(collection);
On earlier versions use CollectionAssert.Contains,
CollectionAssert.DoesNotContain, and Assert.AreEqual(expectedCount, collection.Count).
Replace generic Assert.IsTrue with specialized assertions -- they give better failure messages:
| Instead of | Use |
|---|---|
Assert.IsTrue(list.Count > 0) | Assert.IsNotEmpty(list) |
Assert.IsTrue(list.Count == 0) | Assert.IsEmpty(list) |
Assert.IsTrue(list.Count() == 3) | Assert.HasCount(3, list) |
Assert.IsTrue(x != null) | Assert.IsNotNull(x) |
Assert.IsTrue(x == null) | Assert.IsNull(x) |
Assert.AreEqual(a, b) for same instance | Assert.AreSame(a, b) -- reference identity |
Assert.IsTrue(!list.Contains(item)) | Assert.DoesNotContain(item, list) |
list.Single(predicate) + Assert.IsNotNull | Assert.ContainsSingle(list) |
Assert.IsTrue(list.Contains(item)) | Assert.Contains(item, list) |
// MSTest 3.10+
Assert.Contains("expected", actualString);
Assert.StartsWith("prefix", actualString);
Assert.EndsWith("suffix", actualString);
Assert.MatchesRegex(@"\d{3}-\d{4}", phoneNumber);
On earlier versions use StringAssert.Contains, StringAssert.StartsWith,
StringAssert.EndsWith, and StringAssert.Matches.
MSTest 3.x is not one API level. Pick the form supported by the installed minor version:
// MSTest 3.0-3.3
Assert.IsInstanceOfType(result, typeof(MyHandler));
var typed = (MyHandler)result; // Safe because the assertion stops a mismatch.
// MSTest 3.4-3.11 -- out parameter
Assert.IsInstanceOfType<MyHandler>(result, out var typed);
typed.Handle();
// MSTest 4.x -- returns the proven value directly
var typed = Assert.IsInstanceOfType<MyHandler>(result);
Assert.IsGreaterThan(lowerBound, actual);
Assert.IsLessThan(upperBound, actual);
Assert.IsInRange(low, high, actual);
[TestMethod]
[DataRow(1, 2, 3)]
[DataRow(0, 0, 0, DisplayName = "Zeros")]
[DataRow(-1, 1, 0)]
public void Add_ReturnsExpectedSum(int a, int b, int expected)
{
Assert.AreEqual(expected, Calculator.Add(a, b));
}
On MSTest 3.7+, prefer ValueTuple return types over
IEnumerable<object[]> for type safety. Keep IEnumerable<object[]> on older
versions.
Tuple element names document which position maps to which test parameter, and
tuple element types catch incompatible values at compile time. They do not
make DynamicData position-independent, and swapping two same-typed elements
can still compile. Do not claim otherwise. When rows need custom display names
or metadata rather than only typed positional data, use TestDataRow<T> on
MSTest 3.8+.
[TestMethod]
[DynamicData(nameof(DiscountTestData))]
public void ApplyDiscount_ReturnsExpectedPrice(decimal price, int percent, decimal expected)
{
var result = PriceCalculator.ApplyDiscount(price, percent);
Assert.AreEqual(expected, result);
}
// ValueTuple -- preferred (MSTest 3.7+)
public static IEnumerable<(decimal price, int percent, decimal expected)> DiscountTestData =>
[
(100m, 10, 90m),
(200m, 25, 150m),
(50m, 0, 50m),
];
When you need metadata per test case on MSTest 3.8+, use TestDataRow<T>:
public static IEnumerable<TestDataRow<(decimal price, int percent, decimal expected)>> DiscountTestDataWithMetadata =>
[
new((100m, 10, 90m)) { DisplayName = "10% discount" },
new((200m, 25, 150m)) { DisplayName = "25% discount" },
new((50m, 0, 50m)) { DisplayName = "No discount" },
];
FixtureBase<TSut> or established [TestInitialize] lifecycle rather than rewriting the fixture architecture incidentally.[TestInitialize] only for async initialization, combined with the constructor for sync parts[TestCleanup] for cleanup that must run even on failureTestContext via constructor only on MSTest 3.6+; otherwise use the instance property.[TestClass]
public sealed class RepositoryTests
{
private readonly TestContext _testContext;
private readonly FakeDatabase _db; // readonly -- guaranteed by constructor
public RepositoryTests(TestContext testContext)
{
_testContext = testContext;
_db = new FakeDatabase(); // sync init in ctor
}
[TestInitialize]
public async Task InitAsync()
{
// Use TestInitialize ONLY for async setup
await _db.SeedAsync();
}
[TestCleanup]
public void Cleanup() => _db.Reset();
}
[AssemblyInitialize] -- once per assembly[ClassInitialize] -- once per classTestContext property injection: Constructor -> set TestContext property -> [TestInitialize]TestContext: Constructor (receives TestContext) -> [TestInitialize][TestCleanup] -> DisposeAsync -> Dispose -- per test[ClassCleanup] -- once per class[AssemblyCleanup] -- once per assemblyUse TestContext.CancellationToken with
[Timeout(milliseconds, CooperativeCancellation = true)] when the installed
MSTest version exposes the token directly (3.11+). On MSTest 3.6.4-3.10, use
TestContext.CancellationTokenSource.Token with cooperative cancellation
instead. A plain [Timeout] does not establish that the framework token will
stop in-flight work. On older versions, use a test-owned
CancellationTokenSource where cancellation itself is under test.
// MSTest 3.11+
[TestMethod]
[Timeout(5000, CooperativeCancellation = true)]
public async Task FetchData_ReturnsWithinTimeout()
{
var result = await _client.GetDataAsync(_testContext.CancellationToken);
Assert.IsNotNull(result);
}
For retry, conditional execution, or parallelization requests, read Advanced MSTest patterns. Apply its version gates and keep the real tested operation, assertion, and cleanup.
For an MSTESTxxxx request, read
Advanced MSTest patterns, apply the
version-compatible fix for that rule, and rebuild to confirm the diagnostic is
gone. Prefer fixing the code over suppressing the rule.
When the user asked for repository edits and did not prohibit execution, run the
narrowest affected dotnet test command after editing. A successful process with
no discovered-test count is not verification. Require the intended test cases to
be discovered and pass.
If compilation exposes a directly coupled source issue that prevents the corrected existing suite from running (for example, a missing namespace import in the supplied production file), make only that minimum fix and rerun. Do not upgrade packages or broaden the modernization. Report the actual test count and the fixes made; never present unrun or output-free tests as passing. In the final handoff, map every requested modernization to the exact corrected construct and cite the passing test command. Do not rely on a generic "modernized" summary when expected/actual order, exact type checks, data discovery, or class shape were explicit requirements.
In these kits
More from @dotnet
Works with
Claude, Codex, Cursor & more