A coding beginner shipped a game to 4 stores by directing AI agents: what worked, what broke

Description and use

A beginner developer used AI agents to build and test a mobile game, but real‑device bugs and a store rejection due to a test ad ID surfaced. After fixes, the game went live on Google Play and Apps in Toss, and is pending review on the App Store.

Problem

A beginner developer used AI coding agents to ship a mobile game to multiple app stores, but faced real-device bugs and store rejection issues that automated tests did not catch.

Outcome

The project completed 528 commits over 17 days, expanded from 66 to 229 TypeScript files, and added ~1,480 automated tests. Four app stores were submitted; Google Play and Apps in Toss went live, ONE store listed, and the App Store remains in review. Real‑device bugs were identified and fixed, and store rejection was resolved by removing a test ad ID string at build time.

Workflow

• 1. Create AGENTS.md guardrail file with hard rules
• 2. Run read‑only audit agents to report problems
• 3. Split work by file area, give each implementation agent a git worktree
• 4. Merge changes after each round
• 5. Run simulated player bots for balance metrics
• 6. Perform persona QA with agents playing in a real browser
• 7. Use AI‑generated assets and CC0 sound packs
• 8. Submit to app stores
• 9. Address real‑device bugs and store rejections
• 10. Add automated checks for rejected strings

Evidence

All 1,450 automated tests were passing. I didn't write the code, and I didn't write the tests. AI coding agents did.

Source

The joystick froze after an ad. iOS doesn't deliver the touch-end event to the web view while a native full-screen ad is up, so the game believed the old finger was still down and ignored every new touch.

Source

On October 7th the Apps in Toss review rejected my bundle with (my translation): "a mini-app bundle to be released can't contain a test ad group ID."

Source