Skip to content

AGENTS.md

Latest commit

 

History

History
39 lines (29 loc) · 2.4 KB

AGENTS.md

File metadata and controls

39 lines (29 loc) · 2.4 KB
Raw

AGENTS.md

Commands

  • ./gradlew build - Build all modules
  • ./gradlew test - Run all unit tests
  • ./gradlew testDebugUnitTest - Debug unit tests only
  • ./gradlew :MODULE:testDebugUnitTest - Single module (e.g. :payments-core, :paymentsheet)
  • ./gradlew connectedAndroidTest - Instrumentation tests (requires device)
  • ./gradlew detekt - Static analysis
  • ./gradlew :dokkaGenerate - API docs (outputs to docs/)

GitHub Issues — use gh CLI with GH_HOST=github.com prefix (export for reads, inline for writes). Always use --state all when searching. Check existing issues before investigating user reports.

Internal Tools — Jira: MOBILESDK, RUN_MOBILESDK | Trailhead space: mobile-sdk

Architecture

Multi-module Android library for payment processing and financial services.

Core: payments-core (API models, Stripe client) → payments (high-level APIs) → paymentsheet (pre-built UI) Shared: stripe-core (utilities), payments-model (data models), payments-ui-core (shared UI) Specialized: financial-connections, identity, connect, 3ds2sdk Infra: example, paymentsheet-example, payments-core-testing, lint, screenshot-testing

Key Patterns

  • Kotlin coroutines for async; Jetpack Compose + traditional Views
  • Dagger/Hilt DI in some modules; binary-compatibility-validator for API compat
  • Gradle with shared deps (dependencies.gradle), AGP 8.13.x, Kotlin 2.3.x
  • Detekt for static analysis, Paparazzi for screenshot testing
  • No defaults for internal code: public APIs give parameters defaults (= null, = false) for ergonomic construction; non-public code (internal or @RestrictTo) omits defaults on both model fields and function parameters to force explicit decisions at each call site

Testing — MUST invoke the relevant skill (in .agents/skills/) before writing any test code:

  • write-unit-tests — unit test structure, runScenario pattern, Turbine Flow testing, Truth assertions
  • create-fake — fake implementations with Turbine call tracking
  • compose-tests — Compose UI tests with composeRule, Robolectric, node assertions
  • network-tests — NetworkRule integration tests with testBodyFromFile and fixture patterns

Build validation — invoke delegate-low-reasoning-work before running routine Gradle checks, formatting, static analysis, or generated-output validation. Delegate when a low-cost subagent is available.