← ClaudeAtlas

appstronautlisted

Ship a mobile app to the iOS App Store and Google Play end to end. One round of intake questions, then it fills App Store Connect and Play Console through their APIs (browser only for the gaps), runs a rejection preflight built from real rejections, captures store screenshots on simulators/emulators, builds, uploads and (when asked) submits for review. Covers developer-account enrollment, app registration, TestFlight / Play testing and production release for Expo/EAS, React Native, Flutter and native iOS/Android. Stops only for passwords, agreements, 2FA and the final Submit. Use when the user wants to publish, prepare store screenshots or listings, gather store requirements, or fix a store rejection.
neetu-ai-dev/appstronaut · ★ 0 · Web & Frontend · score 72
Install: claude install-skill neetu-ai-dev/appstronaut
# appstronaut A staged wizard to take a mobile app from **nothing → live on the stores**. You (the LLM) are the brain; thin scripts are the hands. `SKILL_DIR` = this file's directory. ## Works in any agent This skill is plain markdown plus shell / Node scripts, so any coding agent that can run commands can follow it (Claude Code, Codex CLI, Cursor, Gemini CLI, GitHub Copilot). Where it names a capability, use your agent's equivalent: | The skill says | Claude Code | Other agents | |---|---|---| | **ask the user** (intake, checkpoints) | `AskUserQuestion` | your question tool, else a numbered list in chat with the recommended answer first — then wait | | **fetch a page** (live specs, fees) | `WebFetch` | your web / fetch tool, else `curl -sL <url>` | | **drive the browser** (console-only steps) | Chrome (`claude --chrome`) | a Playwright / browser MCP server or computer-use tool, else hand the human `console-browser.md` as a click list | | **run a script** | Bash | your shell tool | Host: **macOS** (Xcode for iOS simulators and uploads; Android SDK for emulators). Node 18+ for `scripts/store/`. ## Pipeline mode — "fill everything for me" (preferred) When the user wants the stores set up end to end, run this pipeline instead of the staged wizard. It asks questions **once**, then works unattended. Progress lives in `<project>/store-assets/launch-state.json` (`{"stage": "…", "done": [...], "blocked": [...]}`) so a restarted session resumes where it stopped. ``` A Intake ─