App Store Screenshots Skill: Five Agent Skills and What Each One Builds

The App Store screenshot skill with the most GitHub stars is ParthJadhav's app-store-screenshots, which builds a screenshot editor inside your project and exports store-sized PNGs without any image model. The others split into skills that hand your captures to Google's Nano Banana Pro image model, a skill that only writes rules for compositing real captures, and a skill that writes a screenshot brief and stops there.

When we checked Google on October 21, 2026, the first page for this search mixed GitHub repos, skill directory listings, a Reddit thread, a hosted MCP server, and a Medium review of web tools. Most of those results describe a skill in a sentence or two, which isn't enough to tell you what the thing will do to your repo. A skill is a folder with a SKILL.md file of instructions, sometimes with scripts beside it, that your agent reads when a request matches its description. So two skills with nearly the same name can do completely different work: one writes a Next.js app, another calls an image model three times per slide, and a third never touches an image. We read the README and SKILL.md of each skill below on October 21, 2026, and the star counts are from GitHub that day.

Skill	What it adds	Where images are made	Needs	You end up with
app-store-screenshots (ParthJadhav, 7,484 stars)	A Next.js screenshot editor	In your browser, from the local editor	Node and a package manager	A ZIP of PNGs per device, size, and locale
aso-appstore-screenshots (Adam Lyttle, 1,792 stars)	Python scripts and a frame template	Pillow locally, then Nano Banana Pro	Pillow, SF Pro Display Black, a Gemini MCP server	1290 x 2796 iPhone PNGs
app-store-screenshot-studio (gitethanwoo, 22 stars)	Workflow docs and a validation script	Nano Banana Pro	A Nano Banana Pro skill with a Gemini key, XcodeBuildMCP	Validated composites, packaged for upload
store-screenshot-mockups (iamngoni, 6 stars)	Markdown rules only	Whatever your agent has	Nothing specific	Raw captures and framed mockups in a store/ folder
screenshot-optimization (aso-skills, 2,193 stars)	Markdown guidance	Nowhere	Nothing	A written plan and design brief

ParthJadhav's app-store-screenshots: an editor in your repo

This is the one to look at first if you want control over every slide. You install it with npx skills add ParthJadhav/app-store-screenshots, which works for Claude Code, Codex, Cursor, and the other agents the skills CLI supports. When you ask for screenshots, the agent asks about your app, source captures, platforms, locales, and style, then scaffolds a Next.js project with a canvas, a sidebar of screens, and an inspector for copy and layout. You run the dev server, edit in the browser, and press Export bundle to download a ZIP organized by platform, device, resolution, and locale. Its device list is the widest here: iPhone, iPhone Duo, iPad, Apple TV, Apple Watch, CarPlay, Mac, Android phone and tablet, and the Google Play feature graphic. The deck lives in app-store-screenshots.json, so you can commit it and rebuild the same set after a fresh clone. The tradeoff is that you now maintain a small web app, and the design work is yours to do in its editor (it ships 18 named styles to start from).

Adam Lyttle's skill sends a Pillow scaffold to Nano Banana Pro

Adam Lyttle's ASO skill starts somewhere the others don't: it reads your codebase to find three to five benefits worth selling, then rates your simulator captures and pairs each one with a benefit. A Python script builds a plain scaffold with the headline, a device frame, and your capture, and Nano Banana Pro (through a Gemini MCP server) redraws it with a photorealistic frame and elements that break out of the phone, three versions per slide. The skill then crops the pick to 1290 x 2796. It's written for Claude Code and iPhone, it expects the SF Pro Display Black font at a fixed path on your Mac, and the repo hasn't been updated since March 30, 2026. Because the image model repaints the whole slide, compare each final against the capture it came from before you upload anything, since App Review checks that screenshots show the app as it really is. The guide to rejected App Store screenshots covers what Apple flags.

App Store Screenshot Studio adds capture and compliance checks

gitethanwoo's skill covers more of the pipeline around the image model. It writes a storyboard, captures raw screens by driving the simulator through XcodeBuildMCP (its repo now lives at getsentry/MobileBuildMCP), sends them to a separate Nano Banana Pro skill (which needs a GEMINI_API_KEY), and runs a Pillow script that checks every file against Apple's sizes. Its rules forbid prices in overlays, mentions of other platforms, and anything but demo data. You install it by cloning into ~/.claude/skills, and its last commit was January 7, 2026. If you'd rather capture by hand, the guide to App Store screenshots from the Simulator has a script that does that part without any MCP server.

store-screenshot-mockups keeps the real UI on top

iamngoni's skill is the smallest: a SKILL.md, two reference files, and a Codex agent config, with no scripts. Its core rule is that the app screen in an upload candidate must be the exact capture, placed as a single layer under a headline and frame. Anything an image model drew goes in a separate image-gen-direction folder as reference only. Since it brings no renderer, the agent composites with whatever tools it already has, and results depend on that. The repo has no license file, so ask the author before reusing its text.

screenshot-optimization writes the brief

The aso-skills collection has around 40 App Store Optimization skills, and screenshot-optimization is the one for screenshots. Install only that one with npx skills add appeeky/aso-skills --skill screenshot-optimization. It plans a ten-slot set, writes overlay copy and a design brief, and audits competitors' screenshots, but it doesn't produce an image. Some of its numbers have no source, such as a line saying the first three screenshots decide 80% of the conversion decision, so treat those as the author's estimate. Pair it with one of the skills above or a designer, and compare its copy against the headline formulas drawn from top-chart apps.

Before you install one

Read the SKILL.md before you run a skill, because your agent will follow it and run its scripts with the permissions your agent already has. Check which API keys it expects and whether your captures leave your machine: the two Nano Banana skills send them to Google, while ParthJadhav's renders locally. Then check the output. The free size checker reads exported files in your browser and names the display slot each one fits, and the screenshot size tables list what App Store Connect accepts per device.

Where Screenies fits

Screenies is a remote MCP server rather than a skill, and your agent connects to it with one command (claude mcp add --transport http screenies https://screenies.app/mcp in Claude Code), so nothing is added to your repo and generation runs on Screenies from the captures and app description your agent uploads. A set your agent starts through the server spends the same credits as one started in the dashboard, bought as one-time packs on the pricing page, and Screenies has no Apple Watch, Mac, or Apple TV output. The Claude Code screenshot workflow walks through the setup and the prompts.