// Use when measuring a Sentry performance span locally with an agent-device replay flow on iOS simulator or Android emulator.
Provides coding standards for React Native — performance patterns, consistency rules, and clean React architecture. Use when writing, modifying, or reviewing code.
Records iOS/Android native MP4 evidence for test/repro flows extracted from an Expensify GitHub PR or issue. Use when the user asks to "record the flow for PR
Drive iOS and Android devices for the Expensify App - testing, debugging, performance profiling, bug reproduction, and feature verification. Use when the developer needs to interact with the mobile app on a device.
Analyze Sentry issues, spans, crashes, and performance metrics. Use when user requests check in Sentry, asks about performance metrics and spans or asks about crash rates.
Test the Expensify App using Playwright browser automation. Use when user requests browser testing, after making frontend changes, or when debugging UI issues
Onyx state management patterns — useOnyx hook, action files, optimistic updates, collections, and offline-first architecture. Use when working with Onyx connections, writing action files, debugging state, or implementing API calls with optimistic data.
| name | measure-telemetry-span |
| description | Use when measuring a Sentry performance span locally with an agent-device replay flow on iOS simulator or Android emulator. |
| argument-hint | <span-name> [runs] [platform] [--boot] |
| allowed-tools | Bash(.claude/skills/measure-telemetry-span/measure.sh) Read Grep Glob |
Pattern: from repo root, run one command with a span name and platform → stdout is a small summary table (avg / min / max + sample ms list). The script measures whatever Git checkout is currently active: it never runs git checkout or otherwise switches branches. To compare this branch with main (or any other revision), check out each commit/branch in turn—or use two worktrees/clones—and run measure.sh separately, then compare the two printed summaries.
.claude/skills/measure-telemetry-span/measure.sh <span-name> [runs] [platform] [--boot]
| Argument | Default | Description |
|---|---|---|
<span-name> | — | Must match # @tag sentry-<span-name> on a flow under .claude/skills/agent-device/flows/tests/ (sentry-tagged QA scenarios live there; the script searches flows/ recursively, so flows/macros/ is also scanned for completeness). |
[runs] | 10 | Measured replays after one warmup inside the script. |
[platform] | ios | ios or android — must match the simulator/emulator you use. |
--boot | off | Before open, runs agent-device boot --platform <platform> so a simulator/emulator is started when nothing was connected (adb devices empty, etc.). |
To pick a specific Android AVD or iOS simulator, use the same global flags agent-device already supports (for example --device "Pixel_7_API_34") on boot and on later commands — either run agent-device boot --platform android --device "…" yourself before measure.sh, or rely on agent-device config (~/.agent-device/config.json). --boot inside this script only passes --platform through to boot.
Environment: APP_ID overrides app bundle (default com.expensify.chat.dev). If the flow declares # @param KEY …, set AD_KEY to pass -e KEY=VALUE to replay.
Output: table + Samples: …ms line; stderr has progress (Using flow:, runs, optional reset).
| Must have | Notes |
|---|---|
agent-device (global install, version per repo agent-device skill) | |
Metro on 8081 (npm run start) | |
| Dev build on device | |
| iOS | agent-device react-devtools attached so Hermes console.debug reaches logs. |
| Android | Span line visible in adb logcat at debug once you verify manually. |
If you see no Android device (adb devices empty): append --boot to the measure command, or run manually first:
agent-device boot --platform android (optional --device "<AVD name>"). For iOS, agent-device boot --platform ios or agent-device ensure-simulator --boot when you need a created simulator instance.
[Sentry][<SpanName>] Ending span (<N>ms) via console.debug.# @tag sentry-<SpanName> (same name, case-sensitive).@reset <path.ad> (run by the script after warmup and each measured replay; if absent, the script relaunches the app instead so each run starts from @pre); @param keys overridable via AD_* (passed as -e KEY=VALUE to replay).RUNS matching log lines from the capture. That matches one sample per measured replay only if each replay emits one such line for this span name. Extra matches (duplicate logs, nested/sub-spans with the same message pattern, noisy startup logging) can shift which samples are included—fix the app logging or tighten the grep if that happens.@reset and loop stabilitymeasure.sh replays the same tagged flow every iteration. Treat @reset as “return to a known anchor,” not a second copy of the whole scenario:
@reset at a macro path so one file stays the source of truth..claude/skills/agent-device/SKILL.md (Metro, dev build, device boot, iOS + DevTools for console.debug) before blaming selectors.Optional: keep a tiny markdown table in your team notes mapping SpanName → one-line intent + @pre anchor; the span name still drives which .ad is picked — no need to repeat long repro prose in every chat.
| Symptom | Action |
|---|---|
No captured runs | iOS: DevTools. Android: log level / package. Retry after clearing log pipeline. |
SESSION_NOT_FOUND / empty adb devices | Use --boot on the measure script, or `agent-device boot --platform android |
Fewer samples than runs | Span not emitted or flow flaky — agent-device replay <flow> --debug; fix selectors (ad-flow-author). |
For another app bundle, export APP_ID before the command.