mcpField
F
flutter-performance

flutter-performance

github.com/thienqoDevelopment946 installsUpdated 1d ago
Repository

About flutter-performance

flutter-performance is a skill for Flutter developers investigating performance issues. It guides reproducible measurement, symptom-based diagnosis, and before-and-after verification.

Key features

  1. 01Reproducible profile-mode measurements
  2. 02Diagnosis by performance symptom
  3. 03Frame, CPU, and memory analysis
  4. 04Startup and app-size investigation
  5. 05Before-and-after verification

Use cases

  1. 01Investigating frame jank or excess rebuilds
  2. 02Tracing slow startup or memory growth
  3. 03Comparing release app sizes
  4. 04Verifying a performance change

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/thienqo/flutter-skills --skill flutter-performance

After installing, type/flutter-performancein Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

evals/
cases.json
references/
memory-startup-and-size.md
profile-and-frame-analysis.md

Details

From the project README

Flutter Performance

Measure a reproducible flow before changing code. Debug-mode timing is not release-performance evidence.

Establish evidence

Record the target device, Flutter version, build mode, flow, dataset, and metric. Reproduce in profile mode on representative hardware when the issue is user-facing.

Diagnose by symptom

  • Frame jank: inspect UI/raster timing, expensive build/layout/paint, shader or image work.
  • Excess rebuilds: inspect state subscription scope and widget identity before adding const mechanically.
  • Slow startup: trace initialization and defer work not required for first useful frame.
  • Memory growth: compare repeated-flow snapshots and find retained ownership, listeners, controllers, or caches.
  • Large binaries: use Flutter's app-size analysis and compare like-for-like release artifacts.
  • CPU-heavy Dart: use timeline/CPU profiling and consider isolates only after locating the work.

Prefer removing or moving work over caching everything. Bound caches and preserve correctness.

Load profile and frame analysis for frame, CPU, image, rebuild, or web-rendering investigations. Load memory, startup, and size for retention, first-useful-frame, or artifact-size work.

Verification

Repeat the same scenario after the change and report before/after evidence with variance. Run correctness tests because performance changes often alter lifecycle or caching behavior.

Sources

Comments

Markdown supported