What problem does it solve? Web and Electron apps often ship with slow first paint, bloated JavaScript bundles, oversized images, and input jank, and generic performance advice rarely tells you which fix actually moves LCP, TBT, or keystroke latency. This Skill provides a catalog of techniques proven in measured speed labs, each with attributed before/after numbers and explicit anti-patterns. ## Core Features & Use Cases - Measured technique catalog: Covers boot payload embedding, critical JS graph reduction, content-hash caching, image srcset ladders, cache header rules, and O(n) paste parsing, each with real millisecond deltas. - Anti-pattern guardrails: A lab-proven "what not to do" table lists tempting changes (preloading thumbnails, fetchpriority hints, immutable derivative caching) that measured worse or caused regressions. - Keep/ditch discipline: Scores every iteration on speed, complexity, and maintainability before anything is proposed for the main branch. - Use Case: A developer profiling a slow SPA uses the catalog to embed route-scoped boot payloads, split the JS graph to the first view, and cap image srcset rungs to the CSS slot, verifying each change against a measured noise floor. ## Quick Start Ask the agent to speed up your web app's load time using the measured techniques in this skill, starting with profiling a user-felt metric like LCP.