aso-appstore-screenshots-skill: Generate High-Converting App Store Screenshots

App Store screenshots are marketing assets, yet most indie developers ship whatever simulator captures were lying around. aso-appstore-screenshots is a Claude Code skill that treats the screenshot set like a product surface: it reads your codebase to find what actually sells the app, then builds polished, benefit-led screenshots around that message.

Why This Skill Matters

The skill runs a four-phase pipeline: Benefit Discovery analyzes your app's codebase to identify the 3-5 core benefits that drive downloads; Screenshot Pairing reviews your simulator screenshots, rates them, and pairs each with the best benefit; Generation produces the images; Showcase renders a preview of the full set side by side. Instead of asking an image model to invent everything at once, it splits the work between deterministic scaffolding and AI enhancement — which is why a set of screenshots ends up looking like a set.

Installation

Four setup steps from the README. First, add the skill to Claude Code:

claude install-skill github.com/adamlyttleapps/claude-skill-aso-appstore-screenshots

Second, install the Python dependency used by the scaffold generator:

pip install Pillow

Third, the headline font: the skill uses SF Pro Display Black, expected at /Library/Fonts/SF-Pro-Display-Black.otf, installable from Apple's developer fonts page.

Fourth, the AI enhancement stage depends on the Gemini MCP server:

npm install -g @houtini/gemini-mcp

Then register @houtini/gemini-mcp as an MCP server in your Claude Code config (~/.claude/settings.json or the project .mcp.json).

Real Workflow: Ship Screenshots for an App Update

1. From inside your iOS app's project directory, run:

/aso-appstore-screenshots

2. The skill walks you through each phase interactively. In the discovery phase it reads your codebase and proposes the 3-5 core benefits that drive downloads.

3. It reviews the simulator screenshots you provide, rates them, and pairs each capture with the benefit it best demonstrates.

4. Generation is a two-stage pipeline. compose.py builds a deterministic scaffold — exact text positioning, device frame, your simulator capture composited inside. Nano Banana Pro (via Gemini MCP) then enhances the scaffold with a photorealistic device frame, breakout elements, and visual polish.

5. Each benefit folder receives working versions — scaffold.png plus AI variants v1.png, v2.png, v3.png — and approved picks land in final/ cropped to exact App Store dimensions, defaulting to 1290×2796px for iPhone 6.7". A showcase.png lays every screenshot out side by side for a last review before upload.

Because progress is saved to Claude Code's memory system, this does not have to happen in one sitting — you can resume the run in a later conversation.

Tips

• Keep clean simulator captures ready before starting. Pairing works best when it has real screens to rate rather than placeholders.
• Trust the scaffold. The deterministic first stage is what keeps text positioning and layout consistent across a whole set; let the AI stage handle only the polish.
• The final/ folder is upload-ready. Resizing to Apple's dimensions happens automatically, so do not re-export manually.
• If the proposed benefits miss your positioning, correct them at the discovery phase — everything downstream inherits that message.

When Not to Use This

This skill targets the iOS App Store — the README documents iPhone dimensions and an App Store workflow, with nothing for Google Play asset specs. It also carries real dependencies: Pillow, a specific font file, and a Gemini MCP server. If you would rather not wire those up, a screenshot template tool remains the simpler path for a small update.

See the leaderboard for more skills.