Back to directory

Compose Agent Skill

Compose Agent Skill is an AI coding agent skill that grounds Jetpack Compose guidance in real source from the androidx and Compose Multiplatform repositories, covering Android, desktop, iOS and web targets.

Compose Agent Skill

This skill addresses a specific failure mode of AI coding tools writing Jetpack Compose code: code that compiles but gets details wrong, such as incorrect remember usage, unstable recompositions, wrong modifier ordering, deprecated navigation patterns or invented APIs that do not exist, because the model is guessing at behavior instead of checking it.

It gives the AI assistant 24 reference guides covering major Compose topics, including Compose Multiplatform, Android TV, Material 3 motion, design-to-code workflows, animation, Paging 3 and navigation migration, plus six source files pulled directly from the androidx/androidx and compose-multiplatform-core repositories so the agent can check how an API actually behaves before suggesting it, including picking the correct state primitive, applying stability annotations, using type-safe navigation routes, and using expect and actual declarations correctly in shared multiplatform code.

It installs as a plugin for Claude Code, Copilot CLI or Codex CLI, with new versions published as GitHub releases carrying migration notes. It is written in Kotlin.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.