android-run-device

Audit and manage the in-app localization of a Capacitor / web-shell app, whose UI strings live in JavaScript and HTML rather than in .strings, .xcstrings or strings.xml. Use whenever such an app needs a language checked, added or fixed, when a screen shows the wrong language or an empty label, when asked to find untranslated or hardcoded UI text, or when handling RTL in a bilingual app ("שפה לא נכונה", "תרגום באפליקציה"). It finds the failure this architecture hides: a lookup that falls back instead of failing, so a missing translation ships silently as the wrong language. For an Xcode project whose strings really are in String Catalogs, use localization-i18n instead — that skill has nothing to work on here. It covers the web layer's strings; native Apple localization is localization-i18n's job.

Tighten existing product copy — App Store / Google Play listing text, marketing and product sites, README marketing sections, in-app strings, replies to store reviews — when it reads clumsily: long sentences, stacked clauses, the same point made twice, throat-clearing openers, jargon the buyer does not know, or a translation that reads as a translation rather than as the language. Use when the user says copy is convoluted, wordy, clunky, unprofessional, hard to read, sounds machine-written, "מסורבל", "מנוסח רע", "נשמע כאילו מודל כתב", or asks to rewrite, tighten, sharpen or polish existing text in any language. This skill EDITS text that already exists; it does not decide what to say. Run it before delivering copy to a store or deploying a site.

Build an app's product profile — the one document every piece of copy about it is later derived from: store listings, site pages, campaigns. Use when onboarding a new app, before writing any copy for it, or when the user asks to "analyze my app", "write up what my app does", "build the marketing profile", "document the app's advantages" ("נתח את האפליקציה", "בנה פרופיל"). It reads the app's source like a product analyst, interviews the owner for what code cannot reveal, researches the real competitors, and proposes a feature and advantage list for the owner to edit before anything is written. Writes `profile.md`, an owner-facing translation beside it, and prepares the media folder. It does NOT capture media (appstore-media does), does NOT write store listing text (store-metadata-writer does), and never writes the app's name or subtitle — those live in the app's README, owned by app-identity.

Decide an app's NAME for both stores, and own the project README as the source of truth for the name, subtitle / short description, and detected feature list. Use when the user wants to name or rename an app, asks "what should I call my app", wants to change the name under the icon or in a store listing, or wants the README's feature list written or refreshed ("תן שם לאפליקציה", "עדכן את ה-README"). Run it when the app is feature-complete, and always BEFORE app-store-metadata, play-store-metadata, aso-keywords and appstore-media, which all read the name and feature list from here. It scans the code for features and monetization, proposes ASO-aware candidates, and DISCUSSES them: it never picks a name on its own. Once decided it writes the display name into the Apple build settings and into Android's strings.xml, and owns the README's version label and changelog. It does NOT write the store metadata files or the keyword field — app-store-metadata, play-store-metadata and aso-keywords own those.

Set up the app factory in a repo, or say why its skills are not loading. Use when a store or release skill that should exist is missing, when the user asks "why aren't the skills loading here", "set up the factory", "enable the tracks", "which tracks does this repo need" ("למה אין לי את הסקילים", "תפעיל את המסלולים"), or after installing the marketplace for the first time. It reads the repo to decide what it is, proposes the tracks that follow, and writes the repo's own .claude/settings.json. It does NOT install plugins, does NOT change the machine's global settings, and does NOT commit — enabling a track is a fact about one repo.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling or beautifying any web UI; "תעצב את הדף", "שפר את העיצוב"). Generates creative, polished code and UI design that avoids generic AI aesthetics. It does NOT decide a page's structure — `page-builder` does that first — and does NOT audit the result (`web-design-guidelines`).

Verify an Apple app is submission-ready and ship it — the FINAL App Store / Mac App Store step: confirm the content is already in place (signing, compliance, metadata, screenshots, app record), then archive, upload, complete the App Store Connect website steps, and submit. Use this when the user wants to release, publish, submit or upload an iOS or Mac build ("תעלה לאפסטור", "תשלח לבדיקה"), asks "how do I upload or submit to the App Store", or is at the end of the process and wants the verify → upload → submit checklist. This skill does NOT produce the listing copy, screenshots, compliance fixes, or signing — those are owned by their focused skills (app-store-metadata, appstore-media, app-store-review-compliance, apple-credentials, code-signing-provisioning). It VERIFIES their output is present and valid, then builds, uploads, and submits (including the website-only steps fastlane can't do).

Set up direct (outside the Mac App Store) distribution for a macOS Xcode app: a dedicated "Direct" target — sandbox off, hardened runtime on, a DIRECT flag, direct-only frameworks linked to it alone — wired to Sparkle auto-updates with EdDSA signing keys, a SUFeedURL, a "Check for Updates" menu item and an appcast signing script — and, for selling that build, a licence key from a hosted checkout, kept out of the App Store build. Use whenever the user wants automatic updates for a Mac app sold outside the App Store ("הפצה ישירה", "עדכונים אוטומטיים"), add Sparkle, create a second non-MAS target alongside an App Store build, set up an appcast, sell the direct build with a licence key ("מפתח רישיון", Lemon Squeezy), or ask how to keep Sparkle or licensing out of the App Store build. Pairs with `notarize-and-distribute`, which builds the signed DMG this skill's appcast points at.

Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.

Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".

Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.

This skill should be used when the user wants to "write agent code", "build an agent with ADK", "add a tool", "create a callback", "define an agent", "use state management" — in a project that needs ADK (Agent Development Kit) API patterns and code examples. It provides a quick reference for agent types, tool definitions, orchestration patterns, callbacks, state management, the graph Workflow API, and reference recipes to study. Do NOT use for scaffolding (use google-agents-cli-scaffold) or deployment (use google-agents-cli-deploy).

Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.

Generate AI music on RunComfy via the `runcomfy` CLI — a smart router across the music-model catalog. Routes to ElevenLabs AI Music Generation (premium 44.1 kHz stereo vocal tracks, 5 s–5 min, $0.0083/s) and ACE Step / ACE Step 1.5 (StepFun-AI open-weights, tag-driven composition, multilingual lyrics, $0.0002–0.0003/s, ~27× cheaper), plus ACE Step audio-inpaint (regenerate a time range inside an existing track) and ACE Step audio-outpaint (extend a track before or after). Picks the right model for the user's actual intent — premium vocal hook, cheap background music library, multilingual pop song, repair a bad chorus, lengthen a 30 s draft into a 2 min cut — and ships each model's documented prompting patterns plus the minimal `runcomfy run` invoke. Triggers on "generate music", "make a song", "AI music", "background music", "instrumental track", "soundtrack", "jingle", "theme music", "royalty-free music", "compose", "music with lyrics", "extend music", "fix this song", "inpaint music", or any explicit ask to generate or edit music.

创建 lark-cli 的自定义 Skill。当用户需要把飞书 API 操作封装成可复用的 Skill（包装原子 API 或编排多步流程）时使用。