Run tests
Provides guidelines for running Unity tests using the run_unity_tests tool. Make sure to use this skill whenever running, executing, or re-running tests on the Unity editor. This includes verifying implementations, debugging test failures, running specific test assemblies, or any task that involves the run_unity_tests tool. Also covers running Play Mode tests on the player for verifying player-only behavior (e.g., #if directives and code stripping). Also covers verifying `[Category("VisualVerification")]` tests after a run by analyzing the saved screenshots. Even if the user just says "run the tests" or "check if it passes", use this skill.From its SKILL.md
npx -y skills add nowsprinting/unity-coding-skills --skill run-testsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 20 stars20 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as Unlicense. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
8.5 KB, as published. Nobody here has run it
Gotchas
- Never call two Unity Editor tools in parallel.
unity_play_control,get_unity_compilation_result,run_unity_tests, andrun_method_in_unitymust be called strictly one at a time — always wait for each call to return before making the next one. Calling them concurrently causes domain-reload conflicts that result in "canceled" or "did not connect within 30 seconds" errors. - When a Unity Editor tool returns
errororcanceled, wait 10 seconds before retrying. Domain reload typically takes several seconds; immediate retry hits the same in-flight reload and fails again. Do not switch tools in the meantime (e.g., callingunity_play_controlto verify state) — that just compounds the multiplexed calls. If the same tool returnserrororcanceledon two consecutive attempts (with the 10-second wait between them), stop and consult the user instead of retrying further.
Run Tests
Before running tests, complete the following steps in order:
- If any code was modified, confirm compilation success using the
get_unity_compilation_resulttool before proceeding. - To determine
assemblyNamesandtestModefor a specific test class, run${CLAUDE_SKILL_DIR}/scripts/resolve-test-target.sh <test-class-cs-path>. The script prints<assemblyName>\t<testMode>(e.g.MyGame.Tests\tPlayMode). Skip this step when running an already-known assembly.
Then use the run_unity_tests tool to run the tests on the Unity editor.
Test execution can take several minutes. Do not re-run while a test is in progress — always wait for it to complete or time out. If a timeout occurs, narrow down the tests using filter settings and re-run.
Performance Test Results
When com.unity.test-framework.performance is installed, each run writes its measurements to Application.persistentDataPath/PerformanceTestResults.json (alongside the TestResults.xml they are parsed from). Resolve that directory with ${CLAUDE_SKILL_DIR}/scripts/get-persistent-data-path.sh <unity-project-root>.
For the package's measurement pitfalls, see the test-writing-guide skill's resources/unity-test-framework-performance.md.
Run Tests on the Player
run_unity_tests runs tests only inside the Unity Editor. Run tests on a player instead when the user asks for it (e.g., "run on player", "standalone", "実機で実行"), or when player-only behavior must be verified (code inside #if !UNITY_EDITOR, Resources-based loading, IL2CPP, player-build hooks such as ITestPlayerBuildModifier). Pick the section below by target platform.
Standalone Player
For a standalone player running on the host OS (macOS, Windows, or Linux): Read ${CLAUDE_SKILL_DIR}/resources/run-on-standalone-player.md. It covers the build-and-run procedure and troubleshooting including Player.log locations.
Other Players
TBD
Visual Verification
A test method carrying [Category("VisualVerification")] has no Assert statements (see test-writing-guide skill), so Passed only means the test ran without throwing — it says nothing about whether the screen actually looks right. That judgment has to come from analyzing the screenshot the test captured.
Only analyze when the run was to confirm a change made in this session — production code, test code, a scene, a prefab, or another asset. Skip the analysis when nothing is under verification: a regression run over unchanged code, or when the user only asked whether the tests pass. When in doubt, skip.
This applies to Editor runs (run_unity_tests) only. A standalone-player run writes Temp/PlayerTestResult.txt, not an XML result file, and on macOS the Player's persistentDataPath differs from the Editor's — see resources/run-on-standalone-player.md. [TakeScreenshot] itself is Play Mode only.
Procedure
- Run:
python3 ${CLAUDE_SKILL_DIR}/scripts/extract-visual-verification.py <unity-project-root>This prints, per[Category("VisualVerification")]test found inApplication.persistentDataPath/TestResults.xml, its NUnitresult, itsDescriptionproperty (the verification aspects), and itsScreenshotpropert(y/ies) (absolute path(s)). See the script's own comments for why an XML library is used instead of grep/awk (class-scoped[Category]lands on the ancestor<test-suite>, not the<test-case>; a test can record more than oneScreenshot). - Check freshness before trusting the output: compare the printed list of "All N test(s) in this result file" against the tests you actually just ran. A filtered run's XML contains only the test-cases that were executed, so a mismatch (extra tests you didn't run, or missing ones you did) means this file is from an earlier run —
TestResults.xmlis overwritten on each run, not appended. Treat the printed age (— N seconds ago) as a secondary signal only; it can't discriminate a stale file from a few minutes ago from a run that legitimately took a few minutes. On either signal of staleness, stop and report it instead of analyzing. - For each test whose
resultisPassed, read the image(s) at itsScreenshotpath(s) and judge them against the aspects listed inDescription. Report each aspect's verdict with what you actually saw in the image. - For a test whose
resultis notPassed, do not analyze — report it as a failure instead (the screenshot, if any, may be from a partial run).
If no TestResults.xml is found (exit code 3)
TestResults.xml under persistentDataPath is written by the com.unity.test-framework.performance package's RunFinished callback, not by Unity Test Framework itself — UTF only writes a result XML for CLI/batchmode runs (to <project-root>/TestResults-<ticks>.xml), never for Test Runner window / MCP-driven runs. If that package isn't installed, fall back:
- Screenshot paths:
[TakeScreenshot]/ScreenshotHelperlogSave screenshot to <absolute path>on every capture. Find the current log file (compare mtimes ofEditor.logandEditor-prev.log— seeresources/troubleshooting-run-unity-tests.md→ "Editor.log vs Editor-prev.log"), then take theSave screenshot tolines that appear after the last- Finished resetting the current domainline (the lifecycle marker for the run you just triggered — same resource, "Tests launched but timed out" table). - Verification aspects: read the
[Description(...)]attribute directly from the test's source file.
Helper script
${CLAUDE_SKILL_DIR}/scripts/get-persistent-data-path.sh <unity-project-root> prints the Editor's Application.persistentDataPath for the current OS (from companyName/productName in ProjectSettings/ProjectSettings.asset, per Unity's docs). extract-visual-verification.py calls it automatically; use it directly only when you need the raw directory.
Rules for Test Failures
If the same test(s) fail on two or more consecutive runs, stop and consult the user rather than continuing to fix.
When consulting, clarify:
- Current failure status: what is failing and the likely cause
- Fix history: what was changed, how many times, and the scope of impact
- Planned approach: what options are being considered next
Troubleshooting
Read the appropriate resource file based on the situation:
- Any Unity MCP tool (
run_unity_tests,unity_play_control,get_unity_compilation_result) is not available or fails with a connection error: Read${CLAUDE_SKILL_DIR}/resources/troubleshooting-run-unity-tests.md - A test fails due to an assertion, constraint, or comparer in the
TestHelpernamespace (excludingTestHelper.UI): Read${CLAUDE_SKILL_DIR}/resources/troubleshooting-test-helper.md - A test fails due to an exception thrown from the
TestHelper.UInamespace: Read${CLAUDE_SKILL_DIR}/resources/troubleshooting-test-helper-ui.md
What ships with it: 10 files
65.8 KB alongside SKILL.md, 5 of them executable
assets/
- PlayerTestRunner.cs9.0 KB
resources/
scripts/
- extract-visual-verification.pyruns6.5 KB
- get-persistent-data-path.shruns1.3 KB
- get-player-settings.shruns892 B
- resolve-assembly.shruns1006 B
- resolve-test-target.shruns1.3 KB
Gives 0 of the 12 instructions most test skills give
Counted across 1,201 of the 2,096 authors here whose files we hold, read 2026-09-06
- Write a failing test before writing codein 43 of 1201, across 36 files
- Run the full test suitein 36 of 1201, across 35 files
- Test only one variable per experimentin 34 of 1201, across 17 files
- Read product marketing context before asking questionsin 34 of 1201, across 14 files
- Mock external dependenciesin 34 of 1201, across 30 files
- Define primary, secondary, and guardrail metricsin 33 of 1201, across 16 files
- Pre-determine sample size before startingin 31 of 1201, across 14 files
- Test behavior rather than implementationin 31 of 1201, across 29 files
- Formulate a hypothesis before designing a testin 30 of 1201, across 13 files
- Document every test hypothesis, variant, and resultin 29 of 1201, across 11 files
- Use descriptive test function namesin 25 of 1201, across 21 files
- Commit to the methodology without stopping earlyin 24 of 1201, across 8 files
Said here and by no other author read
- Confirm compilation success before running tests
- Resolve assembly names and test modes before running
- Run tests using the run_unity_tests tool
- Wait for test execution to complete before re-running
- Wait ten seconds before retrying failed tool calls
- Analyze screenshots for tests with VisualVerification category
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.