agentleFS
torisKRtorisKR/toris-agent/skills/expo-android-performance/SKILL.md

expo-android-performance

Skill2 starsChanged 4 months agoApache-2.0View on GitHub

Diagnose, measure, and optimize Android performance in Expo and React Native applications across startup, JavaScript and UI responsiveness, rendering, lists, memory, network, images and fonts, bundle size, native modules, and package artifacts. Use when an Expo Android app feels slow, drops frames, starts slowly, uses excessive memory or network, has an oversized build, regresses, or needs an Expo-aware performance audit and implementation.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Read execution defaults and select execute, review, or plan.

    line 12
  2. Distinctive:

    Follow the shared mode selection: explicit diagnosis-only requests are read-only.

    line 22
  3. Distinctive:

    Do not manually edit generated native files when config, a config plugin, or another source of truth will overwrite them.

    line 24
  4. Distinctive:

    Use current official Expo, React Native, Android, and package documentation for version-dependent behavior.

    line 27
  5. Distinctive:

    Read evidence and verification before reporting gains.

    line 28
  6. Distinctive:

    Use repeated controlled trials and retain raw evidence.

    line 54
  7. Distinctive:

    When authorized, change the durable layer: JavaScript/TypeScript source, app configuration, build profile, or config plugin as appropriate.

    line 68
  8. Distinctive:

    Do not add memoization everywhere, indiscriminately change architecture/engine, eject/prebuild, install a package, or replace …

    line 72
  9. Distinctive:

    If evidence is noisy or negative, report it.

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Expo Android Performance

Measure the real bottleneck across JavaScript, UI, and native boundaries before changing code. Preserve Expo workflow ownership and prove results in a production-like Android build on representative hardware.

Quick start

Read execution defaults and select execute, review, or plan. In the default execute mode:

  1. inspect the Expo project, dirty state, SDK/runtime, workflow, native-directory ownership, architecture/engine, routes, dependencies, and build profiles;
  2. make the Android scenario reproducible and capture a production-like baseline;
  3. attribute the bottleneck across JavaScript, UI/rendering, native/startup, data, or artifact layers;
  4. implement the smallest fix in the durable source of truth, preserving config-plugin/prebuild ownership;
  5. rerun the same scenario and report comparable measurements, correctness, and a repeatable regression path.

Operating rules

  • Follow the shared mode selection: explicit diagnosis-only requests are read-only; create, improve, optimize, or fix requests against a supplied project default to execution.
  • Detect the actual Expo/React Native workflow and architecture; do not assume managed, prebuild, bare, old, or new architecture.
  • Do not manually edit generated native files when config, a config plugin, or another source of truth will overwrite them.
  • Development-server and debug behavior cannot prove production performance.
  • Preserve correctness, accessibility, navigation, lifecycle behavior, over-the-air update policy, analytics, and unrelated changes.
  • Use current official Expo, React Native, Android, and package documentation for version-dependent behavior.
  • Read evidence and verification before reporting gains.

Workflow

1. Define the scenario

Turn the symptom into a numbered sequence with install/update state, launch type, account/data, route, navigation or gesture, network, duration, and expected behavior. Record target device classes and supported Android versions.

2. Detect ownership and runtime

Inspect:

  • Expo SDK, React Native and React versions, package manager/lockfile, app config, and build profiles;
  • managed, continuous-native-generation/prebuild, or bare workflow;
  • whether android/ is generated, ignored, committed, or intentionally maintained;
  • JavaScript engine, React Native architecture, native modules/config plugins, development client, and update/runtime-version policy;
  • router/navigation, state/data stack, list/image/font/animation libraries, and modules active in the scenario;
  • build variant, ABI, minification/shrinking, source maps/symbols, and artifact type;
  • physical device, Android version, memory/chipset, refresh rate, thermal/power state, and connection.

Determine the durable source of truth before proposing an edit.

3. Capture a production-like baseline

Read measurement playbook. Select metrics tied to the symptom: cold/warm startup, time to usable UI, JS and UI thread stalls, frame timing/jank, interaction latency, render/commit behavior, list throughput, CPU, allocations/retained memory, GC, network waterfalls/payload, image decode/cache, font loading, bundle/module weight, native initialization, package/download/install size, energy, or thermal effects.

Use repeated controlled trials and retain raw evidence. Keep development tooling overhead out of conclusions about production builds.

4. Attribute the bottleneck

  • JavaScript/React: synchronous work, render cascades, unstable props/selectors, context fan-out, serialization, timers, effects, large module evaluation, data transforms, and JS-driven animation.
  • UI/rendering: view hierarchy, layout, overdraw, shadows/clipping, images, list virtualization, gesture contention, animation work, and native view cost.
  • Data/network: waterfalls, retries, payloads, caching, persistence, offline hydration, and request lifecycle.
  • Native/startup: module initialization, splash/font/assets, config plugins, services, engine/architecture boundaries, and platform lifecycle.
  • Artifact/update: bundle composition, dependencies, assets, ABIs, native libraries, source maps/symbols, minification/shrinking, and duplicated resources.

For each hypothesis record supporting evidence, a falsifying experiment, likely user impact, confidence, risk, and effort.

5. Implement safely

When authorized, change the durable layer: JavaScript/TypeScript source, app configuration, build profile, or config plugin as appropriate. Work on one causal area at a time where practical.

Possible evidence-backed actions include deferring noncritical startup, narrowing renders/subscriptions, moving animation work off the JS thread using already-supported architecture, correcting list virtualization, sizing/caching images, batching data or native crossings, removing a request waterfall, lazy-loading real boundaries, or correcting artifact configuration.

Do not add memoization everywhere, indiscriminately change architecture/engine, eject/prebuild, install a package, or replace navigation/rendering infrastructure without evidence and explicit scope.

6. Re-measure and validate

Repeat the same scenario with matching device class, build profile, architecture/engine, data, cache/launch state, network, and measurement method. Compare raw samples, median/range, user-visible behavior, JS/UI attribution, crash/error logs, memory peaks, energy/thermal effects, and artifact changes.

Validate deep links, back/navigation, gestures, lists, loading/error/empty/offline states, background/foreground, updates, accessibility, text scaling, reduced motion, and targeted web/iOS surfaces where the project supports them.

If evidence is noisy or negative, report it; do not cherry-pick.

Deliverable

Provide scenario/environment, ownership/source-of-truth map, baseline, ranked hypotheses, changes or recommendations, comparable before/after results, correctness gates, unverified areas, and a repeatable regression protocol. Do not report a performance win without comparable evidence.

Definition of done

  • The scenario records Expo/React Native runtime, workflow ownership, build profile, and representative device conditions.
  • The bottleneck is attributed with evidence across the correct JS, UI, native, data, or artifact layer.
  • The smallest authorized fix is made in a durable file that prebuild or configuration will not overwrite.
  • Production-like before/after measurements and navigation/lifecycle/accessibility regressions are checked.
  • Noisy, unavailable, or cross-platform evidence is labeled instead of overstated.
  • The handoff lists changed files, measurements, remaining device/platform coverage, and any user-only build or release action.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.