KMPilot
ThisIsSadeghi/KMPilot
Spec-driven development para KMP com skills de IA integradas ao Claude Code
Ver no GitHubConjunto de skills de IA para Claude Code que implementa desenvolvimento orientado por specs para Kotlin Multiplatform. Automatiza design, construção, testes e revisão de features com arquitetura como restrição, cobrindo Android, iOS e web com Compose Multiplatform.
Instalação
npx convez-ai ThisIsSadeghi/KMPilot1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Design de fluxos de UI com blueprint automático
- Scaffolding de features KMP com arquitetura enforçada
- Testes automatizados em todas as camadas (UI, lógica, integração)
- Auditoria arquitetural e review de features
- Sincronização contínua entre spec e código
Veredito

Skill robusto e bem arquitetado que transforma o ciclo de desenvolvimento KMP em um fluxo determinístico com IA. Reduz ciclos de review e garante consistência arquitetural, mas exige disciplina na adoção do padrão spec-driven e conhecimento profundo de KMP.
README
A tradução para o português deste README ainda está na fila.
KMPilot
AI skills for every phase of KMP development.
Design, build, test, and review features — with the architecture enforced.
Built for Claude Code
https://github.com/user-attachments/assets/ca64c2cb-e530-4e88-88e2-755932dc5493
Documentation · Report Bug · Request Feature
Quick Start
Prerequisites: JDK 21+ · Android Studio · Xcode 15+ (iOS) · Claude Code
1. Install
Replace <MyApp> and <com.acme.myapp> with your own values:
curl -fsSL https://raw.githubusercontent.com/ThisIsSadeghi/KMPilot/main/install.sh \
| bash -s <MyApp> <com.acme.myapp>
MyApp— project name (used for the folder, root project, and Android app label)com.acme.myapp— package prefix / application ID (used fornamespace,applicationId, and source-set package roots)
Clones the template, renames packages, initializes fresh git.
Windows: run the same command inside Git Bash (ships with Git for Windows) or WSL — not PowerShell or
cmd. The installer uses bash + GNUsed/find, both of which are present in Git Bash. iOS targets remain macOS-only.
2. Open in Claude Code
cd MyApp && claude
3. Run skills along the lifecycle
> /ui-designer checkout flow with delivery options # design
> /creating-kmp-feature product detail screen with reviews # scaffold
> /verify-ui productdetail # verify UI vs design
> /feature-test productdetail # test all layers
> /feature-review productdetail # audit architecture
See Skills for the full catalog and The Pattern for the flow.
The Pattern
Spec-Driven Development for KMP. Every feature has a living spec at .claude/docs/{name}/spec.md that stays in sync with the code. Skills are phases of one lifecycle:
/ui-designer → /creating-kmp-feature → /verify-ui → /feature-test → /feature-review
design scaffold verify UI test review
Each skill owns one phase and composes without coordinating. The blueprint from /ui-designer is picked up automatically by /creating-kmp-feature; the spec is regenerated as the code evolves; /verify-ui, /feature-test, and /feature-review audit against it. The architecture is the constraint; the LLM is the executor; the spec is the contract.
KMP is the proof point. The pattern generalizes to any opinionated stack.
Skills
Slash-commands ordered by the lifecycle they cover. Run them inside Claude Code in your KMPilot project.
| Phase | Command | Does |
|---|---|---|
| Design | /ui-designer {name} | Design screens in Google Stitch, produce a Compose blueprint |
| Scaffold | /creating-kmp-feature {prompt} | Build a complete feature from a prompt (uses the blueprint if present) |
| Iterate | /modifying-kmp-feature {prompt} | Apply changes to an existing feature |
| Verify UI | /verify-ui {name} | Audit the implementation against the Stitch design |
| Test | /feature-test {name} | Generate fixtures, repository, ViewModel, and UI tests |
| Review | /feature-review {name} | Audit a feature against the 10 architecture rules |
| Spec | /audit-spec {name} | Regenerate or diff a feature's living spec |
| Coverage | /coverage | Test coverage report |
| Health | /features-health | Status report across all feature modules |
Plus two auto-activated skills: using-design-system (enforces X-components on UI work) and bridging-swift-kotlin (guides iOS SDK integration).
What Gets Generated
Every feature module follows the same Clean Architecture shape:
feature/{name}/
├── data/
│ ├── model/ # @Serializable DTOs
│ ├── remote/ # Ktor Resources
│ ├── datasource/
│ │ ├── {Name}RemoteDataSource.kt # interface
│ │ └── {Name}RemoteDataSourceImpl.kt
│ └── repository/
│ ├── {Name}Repository.kt # interface
│ └── {Name}RepositoryImpl.kt
├── presentation/
│ ├── {Name}ViewModel.kt
│ ├── {Name}UiState.kt
│ ├── {Name}UiModel.kt
│ └── ui/
│ ├── {Name}Screen.kt # ViewModel wrapper
│ └── {Name}ScreenRoot.kt # ViewModel-free, testable
└── di/
└── {Name}Modules.kt # Koin bindings
Plus a living spec at .claude/docs/{name}/spec.md that updates when the code changes, and a full test suite (fixtures, repository, ViewModel, UI) when you run /feature-test.
Project Structure
KMPilot/
├── composeApp/ # Shared entry point
│ ├── BaseAppNavHost # Feature routes registered here
│ └── initKoin # Feature modules registered here
│
├── core/
│ ├── common/ # Either, UiState, BaseViewModel
│ ├── data/ # ApiClient, network config
│ └── designsystem/ # X-components (XButton, XTextField, XScaffold...)
│
├── feature/{name}/ # AI-generated feature modules
│ ├── data/ # Models, DataSource, Repository
│ ├── presentation/ # ViewModel, Screens, Navigation
│ └── di/ # Koin module
│
└── .claude/
├── agents/ # Specialized AI agents
├── commands/ # Slash-command definitions
├── skills/ # Skill workflows
└── docs/{feature}/ # Living specifications (spec.md)
Tech Stack
| Category | Technologies |
|---|---|
| Core | Kotlin · Compose Multiplatform · Coroutines & Flow |
| Network | Ktor · Kotlinx Serialization |
| Persistence | Room · DataStore |
| DI | Koin |
| Navigation | Navigation Compose (type-safe) |
| Testing | Turbine · Mokkery · Kover |
Documentation
| Resource | Description |
|---|---|
| Wiki | Complete reference for agents, skills, and architecture patterns |
| CLAUDE.md | Rules and conventions that AI agents follow |
Contributing
Contributions welcome. See CONTRIBUTING.md for guidelines.
MIT License