mjmirza/app-store-compliance
Developer ToolsA pre-submission compliance playbook and scanner that checks iOS and Android apps against App Store and Google Play rejection patterns.
What it is
App Store Compliance Playbook is a reference plus an automated pre-submission guard for iOS and Android apps. The core is a machine-readable taxonomy of rejection patterns in data/rejection-patterns.json, each mapped to a guideline and a concrete fix, and a shell guard that scans a project for signals tied to those patterns. Around it sit platform-by-platform rejection maps for Apple and Google Play, a regulatory deadline database covering the EU AI Act, DSA, COPPA and age laws, and agent skills that plug the check into AI coding tools.
It suits developers about to submit or resubmit an app, and non-developers who need a plain checklist. The website version is at the playbook site.
Quick start
A standalone scan needs only git and bash. Clone the repository, then point the guard at the folder containing your Xcode project or Android app:
git clone https://github.com/mjmirza/app-store-compliance ~/repositories/app-store-compliance
bash ~/repositories/app-store-compliance/agent-os/hooks/app-store-compliance-guard.sh "/path/to/your/app"
The guard changes nothing in your project. Findings are printed with the file and line that triggered them, and the final line tells you where you stand: BLOCKED. means a critical risk was found and each [CRITICAL] entry carries a fix. line; CLEAR. means no critical risk was detected in code; NOT CHECKED. means no project was found at that path or the scan files could not be written. After a clear scan, continue with the account and listing checks in the pre-submission checklist.
For an automatic gate on every submit, the installer wires the guard into a supported AI coding tool so that commands like fastlane, eas submit, gradle bundleRelease and xcrun altool are blocked when a critical risk exists. See the install instructions for the per-tool commands and post-install steps (restart, or approving the hook in Codex).
Other monitors
The repository ships additional Python scripts run from the checkout: monitor.py tracks Apple developer requirement areas against the live RSS feed; monitor-regulatory.py, monitor-android.py, monitor-ai-policy.py, monitor-privacy.py and monitor-security.py watch policy tracks; release-audit.py writes a release-readiness report; accessibility-audit.py performs a static accessibility scan; and deadline-checker.py prints regulatory deadlines in a rolling 90-day window. Several of these use built-in sample announcements unless --live is passed, so their output should be read as worked examples of migration tasks rather than current news.
Constraints
The guard works on macOS, Linux and Windows via Git Bash; tools other than Claude Code also need python3. Static scans cannot see developer account status, accepted agreements, registered package names, review notes, or declared data usage inside the store consoles, which is why the checklist remains part of the flow. The accessibility audit is static and does not compute icon contrast or verify focus order, so VoiceOver, TalkBack and Accessibility Inspector testing is still required for those.
The code and documentation are under the OpenRoots Agent License 2.3: free for individuals, nonprofits, schools, government bodies and companies at or below USD 20 million annual gross revenue, with a 0.5 percent royalty on the using product's revenue above that threshold, capped annually. Reselling the playbook as itself, or training models on it, needs separate agreements. Releases before 24 August 2026 remain under MIT for code and CC BY 4.0 for content.
HotRepo's take
The project reached 11th on GitHub's all-language daily ranking on 19 November 2026, gaining 542 stars that day with 1,473 stars and 110 forks in total. That is a single recorded daily snapshot, so it supports the fact of a sharp one-day spike but not a trend. What stands out technically is the scope of the deliverable: 117 documented rejection patterns, an install harness for a dozen coding agents, and a citation verifier that flags soft-404 pages instead of trusting HTTP 200 responses. The README's own caveat that unverifiable account-level items are surfaced rather than assumed clean is the right default for a check that only runs once at submission time.
Written by AI from the project's README, updated . The project's own docs are the source of truth.