Testing Coroutines With Runtest logo

Testing Coroutines With Runtest

Community
skydoves
testing-coroutines-with-runtest

Use this skill to test suspend functions and coroutine-using classes on the JVM with kotlinx-coroutines-test. Covers runTest, TestScope, StandardTestDispatcher vs UnconfinedTestDispatcher, virtual time via TestCoroutineScheduler (advanceTimeBy, advanceUntilIdle, runCurrent), the canonical MainDispatcherRule wrapper for Dispatchers.setMain/resetMain, the TestResult Promise contract on KMP, and the runBlockingTest -> runTest migration. If the user mentions runTest, runBlockingTest deprecated, advanceUntilIdle, advanceTimeBy, TestScope, StandardTestDispatcher, UnconfinedTestDispatcher, MainDispatcherRule, "Module with the Main dispatcher is missing", "test hangs forever", dispatchTimeoutMs, viewModelScope test, or a ViewModel that posts state from a coroutine, use this skill.

Overview

Publisherskydoves
Repositoryandroid-testing-skills
Skill nametesting-coroutines-with-runtest
Stars
338
Forks
14
Bundled files
Instructions only
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by skydoves on GitHub. Read the source before you install it.

Installation

Install the Testing Coroutines With Runtest AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/skydoves/android-testing-skills/tree/main/jvm-tests/coroutines/testing-coroutines-with-runtest
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/skydoves/android-testing-skills.git /tmp/android-testing-skills
mkdir -p .claude/skills
cp -r /tmp/android-testing-skills/jvm-tests/coroutines/testing-coroutines-with-runtest .claude/skills/testing-coroutines-with-runtest
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Testing Coroutines With Runtest in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Testing Coroutines With Runtest on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Testing Coroutines With Runtest is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Testing Coroutines With runTest — Virtual Time Without Wall Clock

runTest is the only correct entry point for testing suspend functions and viewModelScope-backed code on the JVM. This skill nails down the TestScope contract, the two TestDispatcher flavors, the shared TestCoroutineScheduler virtual clock, and the MainDispatcherRule plumbing that keeps Dispatchers.setMain from leaking between tests. Flow-specific assertions live in ../testing-flows-with-turbine/SKILL.md.

When to use this skill

  • The class under test exposes a suspend fun or launches into viewModelScope / lifecycleScope.
  • The developer reaches for runBlocking { … } and the test hangs, or a 10-minute delay makes the test slow.
  • A ViewModel emits Loading -> Success from a coroutine and the test needs to assert each intermediate state.
  • The build emits the deprecation runBlockingTest is deprecated. Use runTest or runTest(... dispatchTimeoutMs = ...) errors.
  • The test fails with IllegalStateException: Module with the Main dispatcher is missing because viewModelScope routed work through Dispatchers.Main.
  • The developer needs to advance virtual time (advanceTimeBy(5.seconds), advanceUntilIdle()) to trigger a timeout / retry / debounce.

When NOT to use this skill

  • The test asserts Flow emissions across time. Use ../testing-flows-with-turbine/SKILL.md (Turbine handles cancellation, hot vs cold, awaitItem/awaitComplete).
  • The test exercises Compose's MainTestClock rather than kotlinx.coroutines.test. Use ../../../compose/synchronization/controlling-the-test-clock/SKILL.md.
  • The test runs on an emulator/device through AndroidJUnit4. JUnit4 setup itself is ../../runner/configuring-junit4-on-android/SKILL.md.

Prerequisites

  • Gradle: testImplementation("org.jetbrains.kotlinx:kotlinx-coroutines-test:1.10.x") (docs/CORPUS.md §G.1). MUST be testImplementation only — never implementation. The library installs Dispatchers.setMain plumbing that is unsafe to ship to production.
  • @file:OptIn(ExperimentalCoroutinesApi::class) (or per-call) on test files. currentTime, advanceTimeBy, advanceUntilIdle, runCurrent, and the dispatcher constructors are still @ExperimentalCoroutinesApi on 1.10.x (R5).
  • For ViewModel tests, an injected CoroutineDispatcher (a "DispatcherProvider") so the SUT can route work onto the same TestCoroutineScheduler the test body advances. Hard-coded Dispatchers.IO / Dispatchers.Default will NOT see virtual time.
  • For KMP libraries that compile to JS / wasm, use the single-expression form fun foo() = runTest { … } so the platform TestResult is returned (R5; see "TestResult contract" below).

runTest signature

kotlin
public fun runTest(
    context: CoroutineContext = EmptyCoroutineContext,
    timeout: Duration = 60.seconds,
    testBody: suspend TestScope.() -> Unit
): TestResult

(docs/CORPUS.md §G.4.) Three things to internalize:

  1. timeout is the whole-test deadline, NOT a per-dispatch quiescence timeout. The deprecated dispatchTimeoutMs overload had different semantics — see migration below.
  2. After testBody returns, runTest waits for every child coroutine launched on TestScope to complete. Children launched on TestScope.backgroundScope are auto-cancelled at end of test instead. Use backgroundScope for hot-flow collectors.
  3. Uncaught exceptions in children are aggregated and rethrown as a single failure at the end of the test.

TestScope members

kotlin
sealed interface TestScope : CoroutineScope {
    val testScheduler: TestCoroutineScheduler   // shared virtual clock
    val backgroundScope: CoroutineScope         // children auto-cancelled at test end
    val currentTime: Long                       // virtual ms elapsed
    val testTimeSource: TimeSource.WithComparableMarks
}

Plus library extensions on TestScope: advanceTimeBy(delay), advanceTimeBy(durationMillis), advanceUntilIdle(), runCurrent(). (R5.)

TestDispatcher truth table

DispatcherBehaviorPick when
StandardTestDispatcherQueues continuations on the scheduler. Nothing runs until runCurrent()/advanceTimeBy(...)/advanceUntilIdle(). Default for runTest.Asserting intermediate states (Loading -> Success); precise virtual-time control; race-free repro.
UnconfinedTestDispatcherEager: a launch { } runs synchronously up to its first real suspension.Hot StateFlow/SharedFlow collector setup so the first emission is observed without explicit runCurrent().

(docs/CORPUS.md §G.5.)

Default to StandardTestDispatcher. Reach for UnconfinedTestDispatcher only when collector eagerness genuinely matters and you can defend why; it hides ordering bugs that production code under Dispatchers.Default would lose.

TestCoroutineScheduler — the virtual clock

TestCoroutineScheduler is the single source of virtual time shared by every TestDispatcher participating in a test.

APIEffect
runCurrent()Drains tasks already due at currentTime. Does NOT advance the clock.
advanceTimeBy(delta)Advances currentTime by delta, runs every task whose deadline is reached.
advanceUntilIdle()Loops runCurrent + advance until the queue is empty. "Run everything to completion."
currentTimeRead-only virtual milliseconds elapsed.

(R5.)

CRITICAL: Dispatchers.IO, Dispatchers.Default, and any newSingleThreadContext(...) are real thread pools and do NOT participate in the virtual clock. delay() inside withContext(Dispatchers.IO) uses real wall-clock time. The fix is dependency injection — pass a CoroutineDispatcher and substitute mainRule.dispatcher (or StandardTestDispatcher(testScheduler)) in tests.

Dispatchers.setMain / resetMain — required for ViewModels

Android's Dispatchers.Main is a HandlerContext over the main Looper that does not exist on a JVM unit test. Anything launched via viewModelScope, lifecycleScope, or flowOn(Dispatchers.Main) will throw IllegalStateException: Module with the Main dispatcher is missing without Dispatchers.setMain.

Canonical MainDispatcherRule (androidx)

Verbatim from androidx/testutils/testutils-ktx/src/jvmMain/kotlin/androidx/testutils/MainDispatcherRule.jvm.kt (docs/CORPUS.md §G.3):

kotlin
class MainDispatcherRule(
    private val dispatcher: CoroutineDispatcher,
) : TestRule {
    @OptIn(ExperimentalCoroutinesApi::class)
    override fun apply(base: Statement?, description: Description?) =
        object : Statement() {
            override fun evaluate() {
                Dispatchers.setMain(dispatcher)
                try { base!!.evaluate() } finally { Dispatchers.resetMain() }
            }
        }
}

TestWatcher flavor — when you need to expose the dispatcher

kotlin
@ExperimentalCoroutinesApi
class MainCoroutineRule(
    val dispatcher: TestDispatcher = StandardTestDispatcher(),
) : TestWatcher() {
    override fun starting(description: Description?) {
        super.starting(description); Dispatchers.setMain(dispatcher)
    }
    override fun finished(description: Description?) {
        super.finished(description); Dispatchers.resetMain()
    }
}

Usage

kotlin
@OptIn(ExperimentalCoroutinesApi::class)
class MyVmTest {
    @get:Rule val mainRule = MainCoroutineRule()

    @Test fun loadsItems() = runTest(mainRule.dispatcher) {
        val vm = MyViewModel(repo, mainRule.dispatcher)
        vm.load()
        runCurrent()                              // see Loading state
        assertEquals(Loading, vm.state.value)
        advanceUntilIdle()                        // run network/db
        assertEquals(Success(items), vm.state.value)
    }
}

TestResult contract — JVM vs JS

kotlin
public expect class TestResult
PlatformActualCaller obligation
JVMUnitAnything works.
NativeUnitSame.
Kotlin/JS, wasmPromise<Unit>Test function MUST return runTest { … }.

(R5.)

For Android-only projects, you can write @Test fun foo() { runTest { … } } because TestResult collapses to Unit. For a KMP module, that swallows the Promise<Unit> and the JS framework cannot await the test — write @Test fun foo() = runTest { … } (single-expression form) every time. PREFERRED: always use the single-expression form to avoid surprises if the module ever ships KMP targets.

Migration from runBlockingTest

1.5.x and older1.7+
runBlockingTest { … }runTest { … }
TestCoroutineScopeTestScope
TestCoroutineDispatcherStandardTestDispatcher / UnconfinedTestDispatcher
pauseDispatcher { } / resumeDispatchergone — explicit runCurrent / advanceTimeBy
cleanupTestCoroutines()unnecessary — runTest handles cleanup
runTest(... dispatchTimeoutMs = X) { … }runTest(... timeout = X.milliseconds) { … } (semantics changed)

runBlockingTest is deprecated with WARNING; the dispatchTimeoutMs overload is deprecated with ERROR (R5; docs/CORPUS.md §G.2). The migration replacement is mechanical, but the timeout semantics are not identical: old dispatchTimeoutMs was per-dispatch quiescence, new timeout is the whole-test deadline.

Patterns

Pattern: WRONG — runBlocking to test a suspend function

kotlin
// WRONG
@Test fun loadsItems() = runBlocking {
    val vm = MyViewModel(repo)
    vm.load()
    Thread.sleep(2_000)                          // wait for the launch
    assertEquals(Success(items), vm.state.value)
}
// WRONG because: runBlocking has no virtual time and no exception aggregation. The
// delay in load() runs in real time; the launch may not have completed; uncaught
// exceptions in children are silently swallowed.
kotlin
// RIGHT
@Test fun loadsItems() = runTest(mainRule.dispatcher) {
    val vm = MyViewModel(repo, mainRule.dispatcher)
    vm.load()
    advanceUntilIdle()                           // virtual time flushes everything
    assertEquals(Success(items), vm.state.value)
}

Pattern: WRONG — assert before advancing

kotlin
// WRONG
@Test fun loadsItems() = runTest(mainRule.dispatcher) {
    val vm = MyViewModel(repo, mainRule.dispatcher)
    vm.load()
    assertEquals(emptyList(), vm.state.value.items)   // "passes" trivially
}
// WRONG because: with StandardTestDispatcher (the default), nothing runs until you
// advance. The launch inside vm.load() is queued but has not executed. The test
// passes for the wrong reason — the items list is empty because load() never ran.
kotlin
// RIGHT
@Test fun loadsItems() = runTest(mainRule.dispatcher) {
    val vm = MyViewModel(repo, mainRule.dispatcher)
    vm.load()
    advanceUntilIdle()
    assertEquals(items, vm.state.value.items)
}

Pattern: WRONG — withContext(Dispatchers.IO) under runTest

kotlin
// WRONG: production code
class Repo {
    suspend fun fetch() = withContext(Dispatchers.IO) {
        delay(2_000)                              // real wall-clock time
        api.fetch()
    }
}
// WRONG because: Dispatchers.IO is a real thread pool. delay(2_000) takes 2 seconds
// of real time. advanceTimeBy(2_000) does NOT skip it. The whole test runs in real
// time and may exceed the 60s runTest timeout.
kotlin
// RIGHT: production code
class Repo(private val ioDispatcher: CoroutineDispatcher = Dispatchers.IO) {
    suspend fun fetch() = withContext(ioDispatcher) {
        delay(2_000)
        api.fetch()
    }
}
// RIGHT: test
@Test fun fetchesAfterTwoSeconds() = runTest(mainRule.dispatcher) {
    val repo = Repo(ioDispatcher = mainRule.dispatcher)   // shares scheduler
    val deferred = async { repo.fetch() }
    advanceTimeBy(2_001)                                  // virtual; instant in real time
    assertEquals(expected, deferred.await())
}

Pattern: WRONG — collect on TestScope (test hangs)

kotlin
// WRONG
@Test fun observes() = runTest {
    val seen = mutableListOf<Int>()
    vm.uiState.collect { seen += it }            // never returns
    assertEquals(listOf(0, 1), seen)             // unreachable
}
// WRONG because: vm.uiState is a hot StateFlow. collect on the test's TestScope
// never completes; runTest waits for child completion and hangs until the 60s
// timeout fires. Use backgroundScope (auto-cancelled at end of test) — or Turbine.
kotlin
// RIGHT — backgroundScope
@Test fun observes() = runTest(mainRule.dispatcher) {
    val seen = mutableListOf<Int>()
    vm.uiState
        .onEach { seen += it }
        .launchIn(backgroundScope)
    advanceUntilIdle()
    vm.refresh()
    advanceUntilIdle()
    assertEquals(listOf(0, 1), seen)
}

For Flow assertions specifically, prefer Turbine — see ../testing-flows-with-turbine/SKILL.md.

Pattern: the "two schedulers" trap

kotlin
// WRONG
@get:Rule val mainRule = MainCoroutineRule()       // scheduler A

@Test fun broken() = runTest {                      // scheduler B (auto-created)
    val vm = MyViewModel(repo, mainRule.dispatcher) // posts to A
    vm.load()
    advanceUntilIdle()                              // flushes B only — A's work pending
    assertEquals(Success(items), vm.state.value)    // FAIL: still Loading
}
// WRONG because: by default runTest creates its own StandardTestDispatcher with a
// fresh TestCoroutineScheduler. mainRule's dispatcher has a different scheduler;
// advanceUntilIdle() only flushes the test body's scheduler. (Recent kotlinx-
// coroutines-test versions do try to reuse Main's scheduler when Main was replaced
// with a TestDispatcher BEFORE runTest starts, but the defensive contract — pass
// the dispatcher explicitly — never breaks across versions.)
kotlin
// RIGHT — share the scheduler explicitly
@Test fun ok() = runTest(mainRule.dispatcher) {
    val vm = MyViewModel(repo, mainRule.dispatcher)
    vm.load()
    advanceUntilIdle()
    assertEquals(Success(items), vm.state.value)
}

(R5 gotcha #4.)

Mandatory rules

  • MUST use runTest { … }, not runBlocking { … } or the deprecated runBlockingTest { … }.
  • MUST depend on kotlinx-coroutines-test from testImplementation only. MUST NOT put it on implementation.
  • MUST apply @file:OptIn(ExperimentalCoroutinesApi::class) (or per-call) before using currentTime, advanceTimeBy, advanceUntilIdle, runCurrent, StandardTestDispatcher(...), or UnconfinedTestDispatcher(...).
  • MUST install Dispatchers.setMain(testDispatcher) (via MainDispatcherRule/MainCoroutineRule) for any ViewModel/viewModelScope test. Reset in the rule's finally / @After.
  • MUST pass mainRule.dispatcher into runTest(...) so the TestScope and Main share a single scheduler. Otherwise advanceUntilIdle() only flushes one of them.
  • MUST inject any CoroutineDispatcher your production code uses for IO/Default work. Hard-coded Dispatchers.IO will run in real time even under runTest.
  • MUST launch hot-flow collectors on TestScope.backgroundScope (or use Turbine). Never collect directly on the TestScope — the test hangs to the 60 s timeout.
  • MUST NOT assert state without first calling advanceUntilIdle() (or runCurrent() for the "right now" microtask queue). With StandardTestDispatcher, queued work has not run yet.
  • MUST NOT rely on runTest(... dispatchTimeoutMs = X) — deprecated with error. Use timeout = X.milliseconds; remember the semantics changed (whole-test deadline, not per-dispatch quiescence).
  • PREFERRED: the single-expression form @Test fun foo() = runTest { … } for KMP correctness on JS/wasm where TestResult is Promise<Unit>.
  • PREFERRED: default to StandardTestDispatcher. Use UnconfinedTestDispatcher only for hot-flow collector setup where eagerness is the point.
  • PREFERRED: kotlinx-coroutines-debug on the test classpath so runTest timeouts include a coroutine stack dump.

Verification

  • ./gradlew :module:test passes with no Module with the Main dispatcher is missing exception.
  • No runBlocking { … } or runBlockingTest { … } calls remain in test sources (grep -rn 'runBlocking[^T]' src/test; grep -rn 'runBlockingTest' src/test).
  • No dispatchTimeoutMs parameter remains (grep -rn 'dispatchTimeoutMs' src/test).
  • Every ViewModel/viewModelScope test class declares a @get:Rule MainCoroutineRule (or equivalent) and passes its dispatcher into runTest(...).
  • Every hot-flow collector in tests uses backgroundScope or flow.test { … } (Turbine), never bare collect on TestScope.
  • Production classes that use Dispatchers.IO / Dispatchers.Default accept an injectable CoroutineDispatcher and the test substitutes mainRule.dispatcher.
  • @file:OptIn(ExperimentalCoroutinesApi::class) is on every test file using advanceTimeBy / advanceUntilIdle / runCurrent / currentTime.
  • KMP modules (commonTest) use fun foo() = runTest { … } single-expression form.

References

  • JetBrains — kotlinx.coroutines.test: https://kotlinlang.org/api/kotlinx.coroutines/kotlinx-coroutines-test/
  • JetBrains — runTest: https://kotlinlang.org/api/kotlinx.coroutines/kotlinx-coroutines-test/kotlinx.coroutines.test/run-test.html
  • JetBrains — TestDispatcher: https://kotlinlang.org/api/kotlinx.coroutines/kotlinx-coroutines-test/kotlinx.coroutines.test/-test-dispatcher/
  • Android Developers — Test Kotlin coroutines on Android: https://developer.android.com/kotlin/coroutines/test
  • Android Developers blog — Coroutines on Android, Part III (testing): https://medium.com/androiddevelopers/coroutines-on-android-part-iii-real-work-2ba8a2ec2f45
  • androidx/testutils/testutils-ktx/src/jvmMain/kotlin/androidx/testutils/MainDispatcherRule.jvm.kt — canonical MainDispatcherRule (docs/CORPUS.md §G.3).
  • androidx/pdf/.../MainCoroutineRule.kt — TestWatcher flavor that exposes the dispatcher.
  • androidx/navigationevent/...NavigationEventDispatcherTest.kt — runTest(UnconfinedTestDispatcher()) + launchIn(backgroundScope) pattern.
  • androidx/room3/.../InvalidationTrackerTest.kt — injecting TestScope().coroutineContext into the SUT.
  • Sibling: ../testing-flows-with-turbine/SKILL.md — Flow assertions, awaitItem/awaitComplete, hot vs cold cancellation.
  • Sibling: ../../mocking/mocking-with-mockito/SKILL.md — mocking suspend functions with mockito-kotlin.
  • Sibling: ../../mocking/mocking-with-mockk/SKILL.md — coEvery/coVerify for suspend.
  • Sibling: ../../runner/configuring-junit4-on-android/SKILL.md — JUnit4 plumbing surrounding runTest.
  • Sibling: ../../robolectric/using-robolectric-correctly/SKILL.md — when the SUT needs Android framework on the JVM.
  • Cross-set: ../../../fundamentals/strategies/applying-testing-strategies/SKILL.md — small/medium/big test sizing.
  • Cross-set: ../../../fundamentals/strategies/organizing-test-source-sets/SKILL.md — where coroutine tests belong.
  • Cross-set: ../../../compose/synchronization/controlling-the-test-clock/SKILL.md — Compose's MainTestClock (separate from TestCoroutineScheduler).
  • Cross-set: ../../../instrumentation/runner/running-instrumented-tests-with-androidjunit4/SKILL.md — when the same SUT runs on a device.
  • Research: tasks/research/R5-coroutines-test-turbine.md.

Frequently asked questions

What does the Testing Coroutines With Runtest AI skill do?

Use this skill to test suspend functions and coroutine-using classes on the JVM with kotlinx-coroutines-test. Covers runTest, TestScope, StandardTestDispatcher vs UnconfinedTestDispatcher, virtual time via TestCoroutineScheduler (advanceTimeBy, advanceUntilIdle, runCurrent), the canonical MainDispatcherRule wrapper for Dispatchers.setMain/resetMain, the TestResult Promise contract on KMP, and the runBlockingTest -> runTest migration. If the user mentions runTest, runBlockingTest deprecated, advanceUntilIdle, advanceTimeBy, TestScope, StandardTestDispatcher, UnconfinedTestDispatcher, MainDis...

Why use Testing Coroutines With Runtest on TypingMind?

Because you install it once and use it with any model. Testing Coroutines With Runtest is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Testing Coroutines With Runtest in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/skydoves/android-testing-skills/tree/main/jvm-tests/coroutines/testing-coroutines-with-runtest. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Testing Coroutines With Runtest?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Testing Coroutines With Runtest?

As many as you like. As long as a model supports skills, you can use Testing Coroutines With Runtest with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Testing Coroutines With Runtest AI skill free?

Yes. It is published on GitHub by skydoves under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇