launch-frame
A Claude skill for designing beautiful, production-ready app store screenshots and preparing release assets for Google Play, Amazon Appstore, and other platforms.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/rifaque--launch-frame)HTML
<a href="/listings/rifaque--launch-frame"><img src="/api/listings/rifaque--launch-frame/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Requirements: Python 3.9+ with Pillow (`pip install Pillow`). For rendering, either `pip install playwright` or an installed Chrome, Edge or Chromium. For automatic capture, whichever platform tooling you already have: Android SDK platform-tools (adb) plus emulator/AVDs (the Android CLI is used when present), Xcode simulators on macOS, and Playwright/Chrome for web. Optional: Claude in Chrome (or another browser tool) for upload assistance.
**As a plugin (Claude Code)**
/plugin marketplace add Rifaque/launch-frame /plugin install launch-frame@launch-frame
**As a personal skill** (clone, then link the skill folder):
git clone https://github.com/Rifaque/launch-frame.git # macOS / Linux ln -s "$PWD/launch-frame/skills/launch-frame" ~/.claude/skills/launch-frame # Windows (PowerShell) New-Item -ItemType Junction -Path "$HOME\.claude\skills\launch-frame" -Target "$PWD\launch-frame\skills\launch-frame"
Copying the folder instead of linking also works. Run `/reload-skills` (or restart) if it doesn't appear.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 16 lines
- {
- "name": "launch-frame",
- "owner": {
- "name": "RSX Develops"
- },
- "metadata": {
- "description": "LaunchFrame: app store screenshot design, validation and release handoff skill."
- },
- "plugins": [
- {
- "name": "launch-frame",
- "source": "./",
- "description": "App store screenshot and release asset production from an app's real UI, with store-spec validation and upload handoff."
- }
- ]
- }
README.md · README · 146 lines
- # LaunchFrame
- An Agent Skill for Claude Code that turns an app's real screenshots into store-ready listing assets:
- designed to match the app, sized and validated for each store, and handed off in a release folder
- that someone else (or another agent) can upload from.
- How a release works:
- 1. It inspects the repository to work out which targets the app really supports. Each target is graded
- Confirmed, Probable, Unknown or Unsupported, based on manifests, build files and Xcode settings, not
- keywords.
- 2. It checks what this machine can capture with: SDKs, emulators, simulators and connected devices.
- 3. It captures authentic screenshots automatically, preferring emulators and simulators. It navigates
- via deep links or taps grounded in the live UI tree, and records where every source came from.
- 4. It reads the app's design evidence: theme code, tokens, fonts, icons, brand docs and the captures.
- 5. It plans a story from features it can prove and composes each image individually around the real UI.
- 6. It renders exact-size files, reviews them visually, fixes problems, and validates them against a dated
- profile of official store requirements.
- 7. It writes upload instructions per store. When you ask, it can also do the console upload through
- browser tools.
- ## Supported stores
- | Store | Device sets |
- |---|---|
- | Google Play | Phone, 7" and 10" tablet, Chromebook, foldable, Android TV / Google TV, Wear OS, Automotive OS, Android XR, feature graphic, TV banner, hi-res icon |
- | Apple App Store | iPhone (large and 6.1" classes, Duo), iPad 13", Mac, Apple TV, Vision Pro, Apple Watch |
- | Amazon Appstore | Fire Tablet (screenshots, icons, promo image), Fire TV (screenshots, icon, background) |
- | Others | Samsung Galaxy Store, Huawei AppGallery, Microsoft Store, etc. via a profile override written from their official docs at release time (none bundled) |
- The store rules live in `skills/launch-frame/assets/platform-profiles.json`, checked against official
- documentation on the date in that file. Store rules change, so the skill re-checks the official pages at
- the start of each release when it has web access.
- ## Install
- Requirements: Python 3.9+ with Pillow (`pip install Pillow`). For rendering, either
- `pip install playwright` or an installed Chrome, Edge or Chromium. For automatic capture, whichever
- platform tooling you already have: Android SDK platform-tools (adb) plus emulator/AVDs (the Android CLI
- is used when present), Xcode simulators on macOS, and Playwright/Chrome for web. Optional: Claude in Chrome (or
- another browser tool) for upload assistance.
- **As a plugin (Claude Code)**
- ```
- /plugin marketplace add Rifaque/launch-frame
- /plugin install launch-frame@launch-frame
- ```
- **As a personal skill** (clone, then link the skill folder):
- ```
- git clone https://github.com/Rifaque/launch-frame.git
- # macOS / Linux
- ln -s "$PWD/launch-frame/skills/launch-frame" ~/.claude/skills/launch-frame
- # Windows (PowerShell)
- New-Item -ItemType Junction -Path "$HOME\.claude\skills\launch-frame" -Target "$PWD\launch-frame\skills\launch-frame"
- ```
- Copying the folder instead of linking also works. Run `/reload-skills` (or restart) if it doesn't appear.
- ## Use
- From the app's repository, ask in plain language, or invoke it directly (`/launch-frame`, or
- `/launch-frame:launch-frame` when installed as a plugin):
- - "Prepare Google Play phone and TV screenshots for 1.2.0."
- - "Make App Store screenshots for iPhone and iPad, go straight to production."
- - "Validate the assets in store-assets/release-2.0.0."
- - "Upload the Fire TV screenshots from the 1.0.0 release to the Amazon console and save."
- The skill infers the target stores from the project when it can, and asks once when it can't. It
- offers design directions for the first three shots unless you ask it to go straight to production.
- ## Output
- ```
- store-assets/release-<version>/
- ├── README.md status, what's where, re-render commands
- ├── release.json selected targets
- ├── project-profile.md app purpose, audience, evidenced features
- ├── capability-report.md supported targets with evidence and selection decisions
- ├── capture-plan.md every source screenshot: screen, device, method, status
- ├── design-system.md visual DNA evidence, campaign system, storyboard
- ├── validation-report.md
- ├── source-screenshots/ untouched captures + captures.json (provenance)
- ├── working/ editable HTML/CSS compositions, fonts
- ├── previews/ contact-sheet.png, store-scale.png, thumbnails/
- └── google-play/ | apple-app-store/ | amazon-appstore/
- ├── README.md upload location, order, alt text, status
- └── <device>/01.png …
- ```
- Only selected stores and devices get folders.
- ## Scripts
- All in `skills/launch-frame/scripts/`, Python, cross-platform, `--help` on each.
- | Script | Does |
- |---|---|
- | `discover_targets.py` | Grades every device target from repo evidence. Also finds modules, package ids, variants, deep links, existing screenshot/UI-test tooling and prior releases |
- | `init_release.py` | The one way to create a release folder: validates target names against the store profiles, refuses unsupported targets, adds required companions, never touches an existing release; `--out` writes outside a read-only repo |
- | `check_env.py` | Reports SDKs, AVDs, simulators and devices, plus a per-target capture matrix: method, automated or not, setup needed |
- | `capture_android.py` | Boots emulators, cleans the status bar, launches the app, reads the UI tree, does grounded taps and D-pad navigation, and captures at native resolution. Refuses physical devices unless allowed |
- | `capture_apple.py` | Captures from Apple simulators via simctl (macOS only; exits 3 elsewhere) |
- | `capture_web.py` | Captures a running web UI at an exact viewport |
- | `captures.py` | Keeps the provenance manifest (auto-captured, existing or user-provided) and verifies sources stay unedited |
- | `scan_brand.py` | Finds fonts, color tokens, icons and logos in a repo |
- | `extract_palette.py` | Measures colors and light/dark tone from screenshots; WCAG contrast check |
- | `render_html.py` | Renders HTML/CSS compositions to exact-size opaque PNG/JPEG (Playwright or headless Chrome/Edge) |
- | `validate_assets.py` | Checks exports against store profiles: type, size, aspect, file size, alpha, color mode, count, order, duplicates, missing sets. Writes a report that separates errors from warnings and lists what it did not check |
- | `contact_sheet.py` | Builds review sheets of every set, a store-scale sheet and per-asset thumbnails |
- ## Store uploads
- The skill uploads only when you explicitly ask. It works through browser tools in a session where
- you're already signed in, never handles passwords, and asks before replacing existing assets. It
- never submits for review or publishes. It updates the platform README only with what it actually
- confirmed. Without browser control, it gives you the steps instead.
- ## Limitations
- - It needs real captures of the app and won't invent UI. If it can't run the app (no emulator, broken
- build, login or demo data needed), it says exactly what's blocking and asks for the minimum help.
- - Apple capture needs macOS with Xcode. LaunchFrame has no Fire OS emulator path, so Fire TV/Fire tablet
- captures come from a real device (with your approval) or are labelled as Android captures.
- - Android demo mode doesn't clean every status-bar icon on every system image; the skill checks, and
- crops the status bar when needed.
- - The bundled store rules are a dated snapshot. Google doesn't publish TV screenshot dimensions, so
- 1920×1080 is a default there.
- - No device frame artwork is bundled (licensing); frames, if used, are simple CSS.
- - No app preview video production.
- - Design quality depends on the model's judgment plus its review of the rendered images. The
- validator checks the technical rules, not taste.
- ## Development
- ```
- python -m unittest discover -s tests -v
- LAUNCHFRAME_LIVE_SERIAL=emulator-5554 python -m unittest discover -s tests # + live emulator test
- python evals/run_evals.py --list # 20 behavior scenarios
- python evals/run_evals.py --scenario 06-misleading-keywords --model sonnet # strict grading: evals/RUBRIC.md
- ```
- Tests and evals use synthetic fixture repositories and images labelled TEST DATA. Renderer tests are
- skipped when no browser is available. Physical devices are blocked during evals.
- ---
- <sub>Maintained by RSX Develops.</sub>
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.