/unity-test

name: unity-test
description: "Writes missing tests and runs them via MCP. Identifies untested code, creates EditMode/PlayMode tests, executes via run_tests, reports results."
user-invocable: true
args: scope

/unity-test — Write and Run Tests

Write tests for the project and execute them via MCP.

Scope

If the user specified a scope: test **$ARGUMENTS** If no scope: identify the most critical untested code paths.

Workflow

Use the `unity-test-runner` agent to:

Step 1: Assess Test Coverage

1. Find existing test assemblies (`.asmdef` files with test references) 2. If no test assemblies exist, create them:

• `ProjectName.Tests.Editor` — EditMode tests (fast, no scene)
• `ProjectName.Tests.Runtime` — PlayMode tests (full lifecycle)

3. Identify scripts with public APIs that lack tests 4. Prioritize: gameplay logic > systems > utilities

Step 2: Write Tests

For each untested class/method:

• **EditMode test** if it's pure logic (no MonoBehaviour lifecycle needed)
• **PlayMode test** if it involves MonoBehaviour, physics, or scene state
• Naming: `MethodName_Condition_ExpectedResult`
• Arrange-Act-Assert pattern
• Clean up GameObjects in TearDown

Step 3: Run Tests

run_tests → execute all tests (or specific fixture if scoped)
read_console → get test results

Step 4: Report

Present results:

• Total: X passed, Y failed, Z skipped
• For failures: test name, expected vs actual, stack trace, suggested fix
• New tests created: list with file paths
• Coverage gaps: what still needs testing

Test Priority

1. **Game state logic** — health, damage, scoring, inventory 2. **Input processing** — movement calculation, ability activation 3. **Data systems** — save/load, serialization, configuration 4. **Edge cases** — zero health, empty inventory, null references