KMP Code One Task

This documentation was generated with the assistance of AI. Please report any inaccuracies.

Carry out a single task’s implementation, tests, and checkbox update in a Kotlin Multiplatform (Android + iOS) app — nothing else. This is the narrowest execution unit for a kmp task in the iru-code pipeline: iru-kmp-code-one-task-group calls it once per task in a bucket (potentially many at once, in parallel agents via iru-isolated-skill-executor).

Purpose

iru-kmp-code-one-task is the Kotlin Multiplatform counterpart of iru-android-code-one-task and iru-swift-code-one-task. It detects the project shape at runtime (the iru-setup-kmp-app layout of shared/
androidApp/ + XcodeGen-generated iosApp/, or the JetBrains wizard’s single composeApp/ module; which Android plugin the shared module uses; shared Compose Multiplatform UI vs native UI per platform; iOS host capability), decides which module and source set(s) the task belongs to (commonMain first), and loads only the matching files from its own reference/ directory — mandatory rules, code style, and project structure always; testing, KDoc, shared logic, Compose Multiplatform or native UI, iOS interop, and security as the task requires.

It then re-checks the current code state (reading every actual of an expect, and the Swift callers of any public declaration), implements exactly what the task specifies, writes or updates tests in the matching test source set, and compile-checks every target the touched source sets feed. Two things set it apart from the Android and Swift equivalents: a commonMain change must compile for both Android and iOS, and Swift is a first-class consumer of the shared module’s public API. If the task did not stop on a blocker it immediately checks off its own checkbox (and its sub-tasks') in implementation_plan.md with a "group validation pending" note. For a task that adds or changes a screen, component or list it applies the catalog’s UI contract mapped onto KMP (see Building user interfaces): it first writes the Screen › Region › Component tree (inputs, events, visual states, reuse) and reports it back, builds leaves first with stateless components over the shared view model’s immutable state, uses Lazy* containers with key/contentType (and SwiftUI List/lazy stacks) for every collection, and keeps navigation in the one graph. It also writes the UI tests: runComposeUiTest tests on the Android host (Robolectric) and the iOS simulator for shared Compose UI, or Robolectric Compose / ImageRenderer tests for native UI, capturing .ui-screenshots/ across the .ui-quality.yml directions × themes with an --android/--ios variant through the module’s screenshot helper, plus the accessibility check. It never adds Robolectric, a test dependency or a test target to make a UI test fit; missing wiring is a blocker to report. It does not capture a quality baseline, add license headers, run tests, coverage or linters, run Dokka, link the iOS framework, or boot a simulator — all of that happens once per bucket in iru-kmp-code-one-task-group.

Purpose group

Kotlin Multiplatform Development Tooling

Plugin

Shipped by the iru-kmp plugin of the ai-catalog marketplace (install with /plugin install iru-kmp@ai-catalog; see Plugins).

Invocation

/iru-kmp-code-one-task <task description>

Inputs

Argument Required Description Default

Task description

Yes

The task’s own text, including its exact implementation_plan.md checkbox line(s) and its sub-tasks' text, plus any relevant "Current code state" context.

None — always invoked with a specific task’s text, normally by iru-kmp-code-one-task-group.

Outputs

  • Production code in the right module and source set (commonMain first, then androidMain, iosMain, androidApp, or Swift in iosApp), with actual`s for every target of any new `expect, KDoc on every public/internal declaration, and updated Swift callers when a public signature they use changed.

  • New or updated tests in the matching source set: commonTest, the Android host-test set, iosTest, androidApp/src/test, or iosApp XCTest — written but not run.

  • For a UI task: UI tests and screenshot-capture calls (component/screen/flow, states, platforms) in commonTest, androidApp/src/test or iosApp/iosAppTests, written but not run, and the Screen › Region › Component decomposition.

  • A regenerated Xcode project (xcodegen generate) when iosApp/project.yml changed; the generated framework, header, and .xcodeproj are never edited by hand.

  • implementation_plan.md with this task’s checkbox (and its sub-tasks') checked and a "group validation pending" note naming the module(s), source sets, and files touched; left untouched if the task stopped on a blocker.

  • A short summary back to the caller: layout, Android plugin and UI mode detected; module(s)/source sets touched and reference files read; files touched and tests added/updated per source set; compile-check result per target (passed, failed and fixed, failed outside scope, or unavailable); public API changes visible to Swift; any deviation from reference/ and pre-existing mandatory-rule violations noticed; for a UI task, the decomposition and the UI tests and captures added; and whether the task stopped on a blocker.

Execution flow

flowchart TD A[Start /iru-kmp-code-one-task] --> B["Step 0: detect layout, Android plugin,\nUI mode, iOS project, tooling, host"] B --> C{"KMP module found?"} C -- no --> C1[Report blocker: task mis-tagged] C -- yes --> D["Step 1: decide module + source sets;\nread reference/README.md and only\nthe files the task needs"] D --> E["Step 2: re-read current code;\nall actuals of an expect;\nSwift callers of public API"] E --> F["Step 3: implement exactly the task\n(commonMain first, then actuals,\nthen app/Swift callers;\nUI tasks: decompose first)"] F --> G["Step 4: write/update tests in the\nmatching source set; UI tasks: UI tests +\nscreenshot captures (not run)"] G --> H["Step 5: one Gradle invocation to\ncompile-check every target fed;\nxcodegen + build-for-testing if Swift changed"] H --> I{"Compile failure?"} I -- in this task's code --> F I -- outside scope --> J["Re-run once; note in report"] I -- no --> K J --> K{"Step 6: blocker or\nneeds user decision?"} K -- yes --> L["Ask user / report blocked;\nleave checkbox untouched"] K -- no --> M["Step 7: check off this task's\ncheckbox with 'group validation\npending' note"] M --> N["Step 8: report outcome"] L --> N

Dependencies

Invokes

None. iru-kmp-code-one-task does not invoke any other catalog skill or spawn any agent — it only reads its reference/ files, edits the code, test, and plan files the task names, and runs the Gradle wrapper (and, on macOS, xcodegen/xcodebuild) directly for compile-checks.

Invoked by

  • KMP Code One Task Group — Step 2, once per task in a bucket (concurrently, in a single batch of Agent calls, when the group is marked parallelizable; one at a time otherwise), via the iru-isolated-skill-executor agent, passing that task’s full text including its exact checkbox line(s). That caller re-reads implementation_plan.md afterward to confirm the checkbox this skill already flipped landed, then replaces the "pending" note with the final validation outcome.

Related agents

None spawned by this skill itself; it is run inside an iru-isolated-skill-executor agent by its caller.

Source

SKILL.md on GitHub — the file this page was generated from.