The problem
Coding agents write mobile code fast but forget the rules that keep an app fast and safe: tokens end up in AsyncStorage, search fields get concatenated into database filters, sign-out leaves the last user's data behind. Existing skills explain Expo APIs; none tell you whether the app is ready for production or who can read whose data.
Approach
Built one plugin that installs in Claude Code, OpenAI Codex and Cursor, with 10 skills, 12 agents and a real-time guard hook. A zero-dependency scanner collects facts locally; in deep mode specialist auditors run in parallel and an adversarial verifier tries to disprove every serious finding. The security audit traces every input to the query it reaches, builds an access matrix from RLS policies and API routes, scans git history for leaked secrets and fills an OWASP Mobile and API Top 10 matrix. It also covers performance, MMKV and secure storage, dependencies, SDK upgrades, store privacy forms and per-folder agent rules.
Outcome
One command gives a scored report with file:line evidence, a GO / NO-GO verdict and a trend across audits. The fix skill only reports success after gates, an independent re-check of each finding and a re-audit pass, in whichever coding agent the team uses.
Metrics
- 3Coding agents supported
- 140+Detection rules
- 10 · 12Skills · agents
