bld-optimize-applisted

# bld-optimize-app — measure the shipped app, then fix what the numbers point at `/bld-optimize-react` reads your **source** and tells you what looks wrong. This skill loads your **shipped app in a real browser** and tells you what *is* wrong for someone on a mid-range phone. Different instrument, different failures, neither replaces the other. | | Reads | Catches | Misses | |---|---|---|---| | `/bld-optimize-react` | source files | hooks bugs, dead code, unsafe patterns | anything that only appears once bundled and served | | **`/bld-optimize-app`** | the running page | LCP, layout shift, blocking JS, contrast, crawlability | logic errors, anything below the fold, anything behind a login | Run this one **after** the app is deployed or at least built for production, and after `/bld-optimize-react` has cleared the obvious source-level defects. ## The rule that makes every number meaningful or worthless > **Never audit a dev server.** `npm run dev` serves unminified bundles, no tree-shaking, a HMR client, and source maps. It scores 30 points below the same code built for production. A Lighthouse run against `next dev` is not a pessimistic measurement, it is a measurement of something you are never going to ship. Three valid targets, best first: | Target | Command | What the number means | |---|---|---| | **Live production URL** | already up | The real thing, including your CDN and edge cache. This is the number that counts. | | **Local production build** | `npm run buil