Gives an AI coding agent one house style for Jetpack Compose and Compose Multiplatform code, covering architecture, data, features, platform splits and project setup. The agent then writes screens and modules that follow the same patterns instead of drifting from file to file.
Claude Code
> /plugin marketplace add Meet-Miyani/compose-skill> /plugin install compose-kit@compose-kit
Codex
$ npx skills add Meet-Miyani/compose-skill -a codex
Cursor
$ npx skills add Meet-Miyani/compose-skill -a cursor
Gemini CLI
$ npx skills add Meet-Miyani/compose-skill -a gemini-cli
GitHub Copilot
$ copilot plugin marketplace add Meet-Miyani/compose-skill$ copilot plugin install compose-kit@compose-kit
OpenCode
$ npx skills add Meet-Miyani/compose-skill -a opencode
Windsurf
$ npx skills add Meet-Miyani/compose-skill -a windsurf
A Compose skill for AI coding agents: seven agent skills that make Claude Code, Codex, Cursor, Copilot and Gemini write better Jetpack Compose and Compose Multiplatform code. One consistent house style, measured on seven models in real agent CLIs.
Works with Claude Code · Codex · Cursor · OpenCode · Copilot · Gemini CLI · Antigravity
Website: compose.avinya.dev · Install guides per agent · Results
AI agents write Compose that compiles and then drifts: state ends up in the wrong place, every screen gets its own pattern, error handling changes from file to file, and reviews flag perfectly fine code as blocking. Compose Kit gives the agent one house style plus the reasoning behind it, and only loads the parts the current task needs.
BaseViewModel contract, Koin annotations, Navigation 3,
feature-owned data/domain/presentation layers, and typed error handling.| Skill | What it owns |
|---|---|
compose | The entry decision tree (~1.3k tokens): task kind → affected area → exact reference files |
compose-architecture | Module graph, MVI/BaseViewModel contract, error tiers, state ownership, naming, Koin DI, Navigation 3, coroutines |
compose-feature | A screen or feature slice end to end: contract, ViewModel, UI, navigation, DI and tests; review mode |
compose-ui | Composables: Route/Screen split, stability, loading/empty/error states, lists, animation, accessibility, resources, adaptive layouts |
compose-data | Repositories, Ktor, Room, DataStore, Paging 3, offline-first, data-layer tests |
compose-project | New projects, adopting the kit, modules, convention plugins, version catalog, CI and agent hooks |
compose-platform | commonMain vs expect/actual, iOS/Swift interop, desktop and web targets, platform lifecycle |
Also included: project and feature templates, and the guard scripts (compose-architecture/scripts/).
flowchart TD
P["Your prompt"] --> E["compose entry tree · ~1.3k tokens"]
E --> K{"What kind of task?<br/>new feature · change · bug fix · review<br/>conform · setup · question"}
K --> A{"Which areas does it touch?<br/>architecture · feature · ui<br/>data · project · platform"}
A --> F["Load one reference file per affected area"]
F --> C["Code in the house style"]
Install all seven skills. compose is the entry point, and it hands each task to the other six. Step-by-step pages
for each agent are on the website: Claude Code,
Codex, Cursor,
GitHub Copilot,
Gemini CLI and Antigravity,
OpenCode.
Every option below can install for one project (lives in the repo, so your team gets it too) or globally (for you, in every project on your machine).
Needs Node 22.20+. Works with Claude Code, Codex, Cursor, OpenCode, Copilot, Gemini CLI, Antigravity and 70+ other agents.
# This project only (run it in the repo root)
npx skills add Meet-Miyani/compose-skill --skill '*' -a claude-code
# Globally, for every project
npx skills add Meet-Miyani/compose-skill --skill '*' -a claude-code -g
Swap claude-code for your agent: codex, cursor, opencode, github-copilot, gemini-cli, antigravity. Repeat
-a to install for several at once. Leave it out and you get an interactive picker.
| Agent | This project | Global |
|---|---|---|
| Claude Code | .claude/skills/ | ~/.claude/skills/ |
| Codex, Cursor, OpenCode, Copilot, Gemini CLI, Antigravity | .agents/skills/ | the agent's own folder, e.g. ~/.codex/skills/ |
Update later with npx skills update.
Needs GitHub CLI 2.90+. It installs the latest stable release. Pass --agent too: without it, a non-interactive run
installs for GitHub Copilot.
for s in compose compose-architecture compose-feature compose-ui compose-data compose-project compose-platform; do
gh skill install Meet-Miyani/compose-skill "$s" --agent claude-code --scope project
done
Update later with gh skill update --all. Installed with --pin before? Run gh skill update --unpin --all once:
gh skill update skips pinned skills.
Use --scope user to install globally. Agent IDs: claude-code, codex, cursor, opencode, gemini-cli,
antigravity, github-copilot.
/plugin marketplace add Meet-Miyani/compose-skill
/plugin install compose-kit@compose-kit
From a terminal, the same with a scope: user (the default, every project), project (shared through the repo) or
local (this project, just you).
claude plugin marketplace add Meet-Miyani/compose-skill
claude plugin install compose-kit@compose-kit --scope project
Update later from a terminal: claude plugin marketplace update compose-kit, then
claude plugin update compose-kit@compose-kit, then restart Claude Code.
codex plugin marketplace add Meet-Miyani/compose-skill
codex plugin add compose-kit@compose-kit
Codex plugins install for your user, so they're available in every project.
Update later with codex plugin marketplace upgrade: it refreshes the marketplace and the installed plugin.
Add this line to the project's AGENTS.md or CLAUDE.md:
Compose/CMP work: load the compose skill first; it picks the path and the files to read.
The skills also trigger on their own; the line just makes sure the entry tree loads first.
Coming from the old single compose skill (v5.x or the composekit CLI)? Delete the old compose folder from
your agent's skills directory, then install with any option above. The old CLI's update can't fetch v6.
Optional: install the guard scripts into a project to catch drift in CI:
bash <skills dir>/compose-architecture/scripts/install-guards.sh <project>
The v6.1 test: 5 models × 6 new tasks × 3 arms (no kit, v6.0 kit, v6.1 kit) = 90 agentic runs in the real CLIs. Each answer was graded blind by the two vendors other than its own, and an item passes only when both graders pass it.
Items passed of 39 (27 [eng] items written by the independent task author, 12 [kit] items written by the
moderator):
| Model | No kit | v6.0 kit | v6.1 kit | v6.1 − no kit |
|---|---|---|---|---|
| Sonnet 5.5 | 31 (79.5%) | 35 (89.7%) | 34 (87.2%) | +3 (+7.7 pts) |
| GPT-6-Luna | 28 (71.8%) | 35 (89.7%) | 37 (94.9%) | +9 (+23.1 pts) |
| GPT-6-Sol | 30 (76.9%) | 34 (87.2%) | 34 (87.2%) | +4 (+10.3 pts) |
| Gemini 3.8 Flash | 22 (56.4%) | 33 (84.6%) | 34 (87.2%) | +12 (+30.8 pts) |
| Muse Spark 1.3 | 27 (69.2%) | 27 (69.2%) | 29 (74.4%) | +2 (+5.1 pts) |
By task type, all five models together (both graders agree):
| Tasks | No kit | v6.0 kit | v6.1 kit |
|---|---|---|---|
| Conform (2 tasks, 60 items) | 73.3% | 75.0% | 80.0% |
| Review (2 tasks, 75 items) | 78.7% | 94.7% | 94.7% |
| New feature (2 tasks, 60 items) | 58.3% | 80.0% | 81.7% |
All pre-registered rules held, and no model scored below its own no-kit result, so v6.1.0 is the stable release.
Stars on GitHub
A development methodology for coding agents that makes the agent brainstorm before building, follow plans, run parallel work and finish branches in a disciplined way. You get fewer skipped steps and more consistent results from your coding assistant.
@obraHelps you take a change from a rough idea or bug report through to a tested, reviewed result, with each step handled by a focused skill. Its skills cover pinning down domain terms, designing module interfaces, diagnosing hard bugs, and reviewing changes against your repo's coding standards.
@mattpocockGives your coding agent a planning, testing and review routine to follow on every task. It plans before building, checks changes with tests, has a fresh reviewer look for regressions, and saves repeated wins as reusable skills.
@affaan-mDeveloper tools for working on the React codebase itself. Provides skills for extracting error messages, managing and debugging feature flags across release channels, running Flow type checks, and fixing lint and formatting issues before committing.
@facebookGives an AI coding agent one house style for Jetpack Compose and Compose Multiplatform code, covering architecture, data, features, platform splits and project setup. The agent then writes screens and modules that follow the same patterns instead of drifting from file to file. Install it with the command on its page. It includes compose-architecture, compose-data, compose-feature. Its skills do not fire on their own yet. Free and open source.
Is this plugin yours?
Claim it with GitHubSubmit a pluginPromote it