🛫 App Store Preflight
Catch App Store rejections before Apple does.
A Claude Code skill + multi-agent workflow that reviews your app against every App Store Review Guideline and every App Store Connect submission requirement, marks each one ✅ or ❌ with evidence from your code, checks your App Store Connect listing in the browser, and then fixes what fails.
Why
A rejection costs days: wait in the queue, get a one-line guideline number, guess what it means, fix, resubmit, wait again. Most rejections come from things you could have checked: a missing purpose string, no in-app account deletion, a paywall without Restore Purchases, privacy labels that don't match the SDKs you ship, no demo account for the reviewer, placeholder text, a required-reason API missing from the privacy manifest.
App Store Preflight checks all of it in one command.
What makes it different
- Nothing is invented. Every check quotes Apple word for word: the App Review Guidelines, App Review Help, App Store Connect Help and Apple's developer documentation. The scraped source texts ship in
source/, and CI fails if any of the 10,916 quoted passages doesn't appear verbatim in them.
- Every guideline, not a top 10. All 132 numbered guideline sections (1 to 5.6.4), plus privacy manifests and required-reason APIs, App Privacy labels and ATT, metadata limits and screenshot specs, age rating, export compliance, EU DSA trader status, in-app purchase and subscription setup, account deletion, Sign in with Apple, purpose strings, TestFlight and build statuses, agreements, tax and banking, pricing and availability, App Clips, Game Center, in-app events and custom product pages: 1,774 checks in total.
- It reads your app first. A profiler works out what your app actually does (login? subscriptions? UGC? kids? health? location? AI?). Guidelines that don't apply are marked N/A with a reason, not padded into the report.
- Failures are cross-examined. For every ❌, a separate skeptic agent tries to prove it wrong (the deletion flow is on another screen; the SDK ships its own manifest; the purpose string is set in build settings). Only failures that survive are reported.
- It checks App Store Connect too. With Claude in Chrome or Playwright MCP, a read-only browser agent goes through your listing and compares it with the code: privacy labels vs the data your code collects, IAP product IDs in code vs products in App Store Connect, demo account present when login is required, age rating vs content, screenshots for every required size.
- Backtested on real rejections. Run against an app build that App Review rejected under guideline 5.6 (a paywall "Close" button that opened a discount offer instead of closing), preflight fails that build with
file:line evidence and passes the fixed resubmission.
- Then it fixes. Confirmed code failures are fixed on a new git branch, one fixer at a time, followed by a rebuild and independent re-verification of every fix. App Store Connect changes are proposed and applied only with your approval. It never submits for you.
Install
As a Claude Code plugin (recommended):
/plugin marketplace add ilan-uzan/app-store-preflight
/plugin install app-store-preflight@app-store-preflight
As a standalone skill:
git clone https://github.com/ilan-uzan/app-store-preflight.git
ln -s "$PWD/app-store-preflight/skills/app-store-preflight" ~/.claude/skills/app-store-preflight
Requirements: Claude Code, Python 3.9+ (standard library only), macOS with Xcode for the build/Simulator steps. Optional: Claude in Chrome or the Playwright MCP for the App Store Connect pass, and AXe (brew install cameroncooke/axe/axe) so the Simulator pass can tap through your app.
Use
From your app's folder:
/app-store-preflight
or just ask: "Will my app get rejected?", "Run an App Store preflight", "Check my app against the App Store guidelines before I submit".
It asks four questions (audit or audit + fix, App Store Connect via browser or not, Simulator run or not, standard or thorough), then runs in the background. Watch it with /workflows.
以上为 README 的开头部分。其余内容请阅读原文。 查看完整 README ↗
