App Store Compliance as Code: Catch iOS and Android Rejections Before Release
A mobile release is not complete when the build succeeds. It is complete when a reviewer can install the app, sign in, reach the important flows, understand its permissions, and see that the privacy declarations match what the software actually does.
That makes store review feel like an external integration test that runs late and gives limited feedback. A rejection can cost a release window, especially when the issue is a missing privacy declaration, an unavailable demo account, or a permission that the app cannot justify.
The App Store Compliance Playbook is trending in the November 20, 2026 GitHub daily snapshot at position #12, with 377 stars added in that snapshot. It is a focused project rather than a general mobile framework: it combines a rejection-pattern catalogue, a pre-submission shell guard, and an audit skill for coding agents.
Platform icons included in the repository README. The project credits Flaticon; Apple and Android marks remain trademarks of their respective owners. Source: repository assets.
What the playbook contains
The current README describes a machine-readable taxonomy of 114 rejection patterns. It says 107 have detection recipes that can be checked against a project tree; the remaining seven are store-listing or account checks that need manual evidence.
That distinction is important. The playbook is not a magic approval service. It helps turn a set of policy documents into a repeatable engineering workflow.
Its main pieces are:
- A structured rule set: rejection patterns, detection recipes, guideline references, and fixes.
- An agent skill: an AI coding tool can load the relevant rules and produce a ranked audit.
-
A pre-submission guard: a shell script that can run before commands such as
fastlane,eas submit,gradle bundleRelease, andxcrun altool. - Policy monitors: scripts for Apple, Android, privacy, security, regulatory deadlines, accessibility, and store metadata.
- Citation verification: a checker designed to detect fake or soft-404 policy references instead of treating every HTTP 200 response as proof.
The workflow looks like this:
Mobile project
|
v
Agent skill reads policy rules
|
v
Pre-submission guard scans the project
|
+--> Critical risk --> BLOCKED with a reason
|
+--> No critical risk --> continue to manual checks
This is a conceptual workflow, not a screenshot of the tool.
Install it for a coding agent
The repository's installer supports Claude Code and several other coding tools, including Codex. Review the script before allowing it to edit your agent configuration.
For Codex, a scoped installation looks like this:
git clone https://github.com/mjmirza/app-store-compliance.git
cd app-store-compliance
bash scripts/install.sh install --tool codex
bash scripts/install.sh doctor --tool codex
The installer copies the skill and guard, registers the hook, and runs a check intended to prove that the hook blocks a test submission command. Codex requires you to approve new or changed hooks; use its /hooks command and review the permission before relying on it.
You can also run a one-time audit without installing the persistent hook:
bash agent-os/hooks/app-store-compliance-guard.sh /path/to/your/app
Point it at the actual app directory. The repository documents special handling for Expo projects and nested native projects, so read the README before scanning a monorepo with several apps.
What a clean result does not prove
A clean static scan is not an App Store approval. It cannot fully prove that a live demo account works, that a production backend is reachable from the review environment, or that the store listing and every declaration match the behavior of all third-party SDKs.
The README explicitly separates code-detectable checks from the manual store-listing and account checks. It also provides a metadata audit and instructions for pulling App Store Connect metadata. Use those steps rather than treating a single green result as a release certificate.
A practical CI design is:
- Run the guard against the app before the upload command.
- Fail the job on critical findings.
- Upload the audit report as a build artifact.
- Require a human to verify demo credentials, review notes, privacy forms, age rating, screenshots, and the live backend.
- Re-run the audit after changing permissions, SDKs, analytics, authentication, or data-sharing behavior.
The project is useful as a guardrail, not as a substitute for reading the Apple App Store Review Guidelines and the Google Play policy centre.
Read the licence before adopting it
There is an important licensing detail that older cached copies of the repository page may not reflect. The current README points to the OpenRoots Agent License 2.3, identified as LicenseRef-OpenRoots-ORA-2.3. It is a custom licence for agent skills and workflows, not a standard MIT or Apache licence.
The published summary describes a royalty-free Root tier for individuals, nonprofits, schools, government bodies, and companies at or below USD 20 million annual gross revenue. Above the threshold, the Canopy tier describes a 0.5% share of attributable revenue above the threshold, capped at USD 250,000 per year. AI model training requires a separate Compute licence, and the licence restricts repackaging or hosting the work as a competing product.
Those are meaningful conditions. The licence page itself lists unresolved legal questions, so do not treat its plain-language summary as legal advice. If your company needs conventional permissive open-source terms, or you plan to redistribute the playbook inside a hosted product, get legal review first.
When it is worth adding
This is especially useful for teams shipping apps built with Swift, Kotlin, React Native, Flutter, Expo, or Capacitor. A small team can keep policy checks close to the build pipeline instead of discovering a predictable issue after review begins.
It is less useful if your release process already has a maintained compliance programme, a real device test matrix, and store-listing validation. In that case, compare the playbook's rules with your existing checks and adopt only the gaps. Also review the source URLs and rule dates: policy changes are continuous, and a local copy can become stale.
The project has a recent GitHub Discussion about export-compliance handling, which is a good reminder that policy edge cases need maintainers and practitioners to challenge them. I did not find a verified exact-project YouTube demo or X post suitable for embedding, so the article links to the repository's own installation and policy material instead.
Sources and further reading
- App Store Compliance Playbook
- Current README and installation workflow
- OpenRoots Agent License 2.3 summary and legal code
- Apple App Store Review Guidelines
- Google Play policy centre
- Repository discussion on export compliance
Question: which command should your mobile CI pipeline block on a critical compliance finding: fastlane, eas submit, gradle bundleRelease, or a project-specific wrapper?
Prepared with AI assistance and checked against the linked repository and platform policy sources.


Top comments (0)