Expo Performance Audit — Security skill for Claude Code

Expo Performance Audit

An installable agent skill for reviewing an Expo app before release. It inspects project code and configuration for performance risks, then guides an evidence-based report with actions the developer can review. The `SKILL.md` format can be used by compatible coding agents, including Codex, Claude Code, and Cursor.

What it covers

• Lists: virtualization, row rendering, keys, and places where FlashList may be worth measuring.
• Images: caching, remote thumbnail sizing, and repeated large image use.
• Re-renders: context updates, unstable props, and work that needs profiler evidence.
• Animations: JS-driven paths and UI-thread alternatives.
• Startup: Hermes configuration, heavy imports, bundle analysis, and work before first paint.
• Device validation: a release-build checklist for a physical lower-end Android phone.

The audit labels static findings as candidates until the relevant code or runtime behavior confirms them. It cannot establish server image dimensions, frame drops, or real-device performance from source code alone.

Install with `npx skills`

From this project directory, install the local [`expo-performance-audit`](expo-performance-audit/) skill using the [Skills CLI](https://github.com/vercel-labs/skills):

npx skills add ./expo-performance-audit

The CLI lets you choose a supported agent. By default it installs for the current project; add `-g` to install for your user account across projects. If you publish this repository, the CLI can also install from its Git URL or a direct link to the skill directory.

Ask your agent, for example: `Use the expo-performance-audit skill to review /path/to/my-expo-app before release.` The skill performs a read-only review unless you also ask for fixes.

Run the scanner directly

The bundled scanner uses Python 3 and the standard library. It produces a JSON inventory of review candidates:

python3 expo-performance-audit/scripts/scan.py /path/to/my-expo-app

Its