AGENTSOURCE

The Shelf / App Store / App Store Screenshot Playbook

App Store

App Store Screenshot Playbook

The conversion half of ASO: 7-slot story arc, competitor atlas, captions that get read.

The job: keywords get you impressions. This gets you installs.

A user lands on your listing and decides in under seven seconds, looking at exactly three screenshots and maybe two seconds of a muted video. Most listings lose that window with a "Welcome to AppName" hero and captions nobody reads. This skill runs the decision like an operator: pull what the winning apps in your category actually ship, find the consensus and the breakout, then build your seven frames against it.

What's on the tag:

  • The competitor screenshot atlas. The iTunes Lookup API exposes every app's screenshot URLs; a documented URL trick upgrades them to full resolution. Your agent downloads the top-ranked sets, views them, and answers seven questions per competitor — hero composition, caption style, framing, palette, magic-moment placement — until the category's visual baseline and its one breakout are on the table.
  • The 7-slot story arc: hero → differentiator → magic moment → two use cases → Pro tease → CTA close. Plus the variations for apps without paywalls and the rule about which slot is the only one allowed to show a price.
  • Caption formulas that survive the scroll. Four to eight words, verb first, outcome over feature — with a master pattern table and the anti-pattern list ("Powerful features" is noise; "AI fills in the details" is a caption).
  • A preview-video playbook with the 4-phase script structure and hook patterns by category, plus a paste-ready script template.
  • Custom Product Pages A/B testing with real discipline: one variable, 14-day windows, enough impressions to mean something. Including the documented indie case where the founder's self-made screenshots beat the designer-polished set — the reason the skill makes you test before you swap.
  • Creative localization for eight markets, with the currency and date-format tables that keep a German or Japanese screenshot from quietly torching your credibility.
  • Simulator capture recipes: true-resolution capture, the 9:41 status-bar override, and how to stage app states a real user would need days to reach.

Who it's issued to: iOS developers whose analytics show people finding the listing and not installing, and anyone building a screenshot set without a designer on payroll.

Why not a free directory download? Free skill files tell you screenshots matter. This one was run against live listings — the thumbnail-contrast lesson, the full-res URL trick, the anti-tracing line that keeps App Review off your back — and it's maintained as Apple moves the limits. Buy once, keep the latest version in your locker.

FIELD REPORT real output, not a promise

Two artifacts from the playbook's own documented material: the consensus/breakout read the atlas step produces for a habit-tracker category, and a per-slot recommendation in the exact handoff format the skill outputs.

Atlas output — habit-tracker category (Step D)

After downloading 7 competitors' screenshot sets via iTunes Lookup and running the 7-question analysis:

Consensus: floating phone with brand-color gradient background, 5-word bold-only caption, slot 1 = hero with the full app shown, slot 3 = streaks/calendar visual as the magic moment. 6/7 use the 9:41 status bar. 0/7 show pricing. 0/7 use human photos.

Breakout (the top-ranked indie): no device frame, full-bleed app screen, caption ABOVE the screenshot in a soft pastel, slot 1 = a near-empty state with a single completed habit rather than a maximalist dashboard. This is what "honest indie" reads as in the category — and it's the top converter.

Per-slot recommendation (Step E format)

Slot 1 (Hero):
  Consensus: floating phone + brand gradient
  Breakout: full-bleed app screen, no frame
  Recommendation: adopt the breakout. Reason: this app's UI is already distinctive,
  so showing it bare IS the differentiator. Caption: "Build a habit that sticks" —
  outcome-led, 5 words.

The same structure repeats for all seven slots, so the plan hands straight to a designer (or a rendering pipeline) without interpretation.

The A/B lesson baked into Step 5

From the documented indie habit-tracker case the skill cites: the founder A/B-tested professionally designed screenshots against his original self-made set via Custom Product Pages — and the self-made set won on conversion. The playbook therefore hard-codes the rule: never swap to polished creatives without running the CPP test first. The polish bias is real, and the data gets the veto.

SERVICE RECORD living gear — updated as the factory learns

v1.0.0 — 2026-07-17

First issue. Ported from the factory's internal skill: sanitized for general use, methodology intact, field report captured from a real run.

Every update ships free to owners — your locker always serves the latest version.

QUESTIONS

Do I need design tools or a designer?›

No. The skill produces the plan — slots, screens, captions, per-slot rationale — plus capture commands for real screens. Render the frames in Figma, fastlane, or any framing tool; the specs and compliance rules are included.

Does it really pull competitors' actual screenshots?›

Yes. The iTunes Lookup API exposes every listed app's screenshot URLs; the skill swaps the CDN dimensions for full resolution, downloads the sets in order, and walks a 7-question analysis across them.

Is the A/B testing part against Apple's rules?›

No — it uses Custom Product Pages, Apple's own supported mechanism, with the setup flow and the statistical discipline to read results honestly.

What agents does it run on?›

Written for Claude Code (which can also view the downloaded PNGs directly), works in claude.ai Projects and Codex/AGENTS.md setups. Plain markdown throughout.