dotnet-unit-testing

# .NET Unit Testing Agent generates or updates repository-style MSTest unit tests for .NET code. ## When to Use | User prompt | Use | |---|---| | User asks for new .NET unit tests | Use this skill | | User asks to extend tests after a code change | Use this skill | | User asks to align tests with repository conventions | Use this skill | ## When Not to Use | User prompt | Route | |---|---| | User asks for an audit only | Use `test-anti-patterns`, `grade-tests`, or `assertion-quality` | | User asks to run tests | Use `run-tests` | | User uses xUnit or NUnit intentionally | Use a framework-matched skill or preserve local framework conventions | ## Required Inputs | Input | Required | Description | |---|---|---| | Production class or method | Yes | Agent needs the target behavior. | | Existing tests | No | Agent reads them to match local patterns. | | Dependency graph | No | Agent reads constructor or factory dependencies when isolation is unclear. | ## Repository Conventions | Area | Agent action | Pass | |---|---|---| | Test project | Agent places tests in the mirrored `Tests/**` project structure. | Namespace and path align to the production class. | | Test class | Agent uses `{ProductionClass}Tests`. | File and class naming stay predictable. | | Test method | Agent uses `Method_Scenario_Result`. | Method name communicates behavior. | | Private fields | Agent uses local camelCase patterns and avoids underscore prefixes when neighboring tests do. | Fields match repository style. | | Assertions | Agent prefers specific MSTest assertions. | Failure messages stay readable. | ## Coverage Pattern Table | Category | Agent expectation | Pass | |---|---|---| | Happy path | Agent covers normal success behavior. | Main outcome is asserted. | | Edge case | Agent covers null, empty, boundary, or alternate valid input when applicable. | Boundary behavior is explicit. | | Failure path | Agent covers dependency failure or validation error when applicable. | Failure surface is asserted. | | Data flow | Agent verifies important transformed values or collaborator inputs. | Captured values prove behavior. | | Guard clauses | Agent covers argument validation for public APIs. | Invalid input throws or rejects as designed. | ## Dependency Pattern Table | Dependency shape | Agent default | Agent avoids | |---|---|---| | External boundary | Agent mocks DB, HTTP, queue, or remote service dependencies. | Agent avoids live infrastructure in unit tests. | | Stable framework helper | Agent uses real helpers such as `Options.Create(...)` or `NullLogger<T>.Instance` when sufficient. | Agent avoids unnecessary mocks. | | Pure value object | Agent instantiates the real object. | Agent avoids mocking DTOs or records. | | Sealed helper without interface | Agent constructs the real helper with mocked inner boundaries when practical. | Agent avoids brittle wrappers solely for tests. | ## Workflow | Step | Agent action | Test | Pass | |---|---|---|---| | 1 | Agent reads the target production code and sibling tests. | Review source and nearby tests. | New tests match repository patterns. | | 2 | Agent maps direct dependencies and isolates only the needed boundaries. | Review constructor and collaborators. | Test setup contains only relevant mocks. | | 3 | Agent writes focused tests for happy, edge, and failure behavior. | Review test list. | Public behavior coverage is balanced. | | 4 | Agent uses specific assertions and data capture where behavior needs proof. | Review assertions and verifies. | Assertions prove outcomes, not implementation trivia. | | 5 | Agent references `references/test-patterns.md` or `references/workflow-testing.md` when the target matches those scenarios. | Review supporting guidance. | Pattern choice matches repository examples. | ## Output Requirements | Output | Requirement | |---|---| | Generated tests | Agent returns compilable MSTest code. | | Rationale | Agent names only the repository conventions that affected the result. | | Validation plan | Agent identifies the smallest relevant build or test command when execution is in scope. | ## Verification Checklist - [ ] Agent mirrored repository naming and placement conventions. - [ ] Agent covered happy, edge, and failure behavior where applicable. - [ ] Agent mocked only direct external boundaries. - [ ] Agent used specific MSTest assertions. - [ ] Agent kept the skill under the token budget. ## Common Pitfalls | Pitfall | Agent fix | |---|---| | Agent copies a generic template blindly | Agent reads sibling tests first. | | Agent mocks every collaborator | Agent keeps stable helpers real when they do not hide behavior. | | Agent writes tests against internals only | Agent anchors assertions on observable behavior. | | Agent omits failure-path coverage | Agent adds one meaningful negative path for public APIs. | ## References - [references/test-patterns.md](references/test-patterns.md) - [references/workflow-testing.md](references/workflow-testing.md)