revenuecat-testing — скилл для Claude Code
Скилл revenuecat-testing помогает составить план тестирования встроенных покупок и подписок в RevenueCat, проверить события в песочнице и разобрать сбои.
MIT ★ 674 обновлён 2026-10-29
/plugin marketplace add evanca/flutter-ai-rules
/plugin install flutter-ai-skills@flutter-ai-rulesЧто делает revenuecat-testing
Скилл выстраивает сквозное тестирование RevenueCat вокруг проверяемых сигналов: событий панели управления (INITIAL_PURCHASE, TRANSFER, PRODUCT_CHANGE), значений period_type и логов отладки. Вместо проверки «приложение не падает» агент проверяет конкретные состояния в истории пользователя.
Инструкция покрывает шесть направлений: жизненный цикл подписок (триалы, апгрейды, отмены), идентификацию аккаунтов, матрицу восстановления и переноса покупок, пейволлы, веб-биллинг и проверку динамических предложений перед запуском экспериментов.
В iOS песочнице оформить возврат нельзя из-за ограничений App Store, поэтому событие отмены там приходится ждать до суток или проверять сценарии через Google Play.
Какие задачи решает
- составить пошаговый план QA для встроенных покупок и подписок
- проверить корректность переноса покупок между учётными записями
- найти причину ошибки Invalid Product Identifiers в логах
- подготовить пейволлы к проведению экспериментов
Примеры запросов агенту
- Составь план тестирования подписок и триалов в RevenueCat для Flutter-приложения
- Пользователь восстановил покупку под новым ID, но подписка не перенеслась: разбери логи
- Проверь, готова ли конфигурация пейволла к запуску эксперимента в RevenueCat
Как установить revenuecat-testing
1Через маркетплейс плагинов
/plugin marketplace add evanca/flutter-ai-rules/plugin install flutter-ai-skills@flutter-ai-rulesКоманда ставит плагин flutter-ai-skills целиком, со всеми его скиллами
2Через Skills CLI (Vercel)
npx skills add evanca/flutter-ai-rules --skill revenuecat-testing -a claude-code3Вручную: положить файлы
~/.claude/skills/revenuecat-testing/SKILL.mdПапка скилла целиком; для одного проекта — .claude/skills/revenuecat-testing/
1Через Skills CLI (Vercel)
npx skills add evanca/flutter-ai-rules --skill revenuecat-testing -a codex2Вручную: положить файлы
.agents/skills/revenuecat-testing/SKILL.mdДля всех проектов — ~/.agents/skills/
1Через Skills CLI (Vercel)
npx skills add evanca/flutter-ai-rules --skill revenuecat-testing -a cursor2Вручную: положить файлы
.cursor/skills/revenuecat-testing/SKILL.mdCursor читает и .agents/skills/, и .claude/skills/
1GitHub CLI
gh skill install evanca/flutter-ai-rules skills/revenuecat-testing2Вручную: положить файлы
.github/skills/revenuecat-testing/SKILL.md1Через Skills CLI (Vercel)
npx skills add evanca/flutter-ai-rules --skill revenuecat-testing -a gemini-cli2Командой платформы
gemini skills install https://github.com/evanca/flutter-ai-rules --path skills/revenuecat-testingВручную: положить файлы
.devin/skills/revenuecat-testing/SKILL.mdСтарый путь .windsurf/skills/ тоже читается
Claude.ai
Упакуйте папку в zip и загрузите в Settings → Capabilities → Skills
По шагам, с командами для каждой платформы: Как установить скилл в Claude Code: пять способов
Вопросы о revenuecat-testing
Как установить revenuecat-testing в Claude Code?
Для установки через маркетплейс выполните /plugin marketplace add evanca/flutter-ai-rules, а затем /plugin install flutter-ai-skills@flutter-ai-rules.
Что делает скилл revenuecat-testing?
Он формирует матрицу сценариев для проверки подписок, помогает настроить песочницу и разобрать ошибки при восстановлении покупок по событиям RevenueCat.
Можно ли протестировать возвраты подписок в песочнице iOS?
В sandbox iOS нельзя оформить возврат средств напрямую: App Store отправляет запросы через поддержку Apple, а событие отмены может приходить с задержкой до 24 часов.
Исходный текст
SKILL.md · MIT
--- name: revenuecat-testing description: "Use when testing RevenueCat purchases/subscriptions, setting up sandbox testing, debugging a purchase/restore/trial, verifying entitlements or events, checking an app is ready for a RevenueCat experiment, or writing an IAP QA plan." license: MIT --- # RevenueCat Testing Help a developer systematically verify their RevenueCat integration works before shipping — turning "I added RevenueCat, does it actually work?" into a concrete, checkable QA plan. The full use-case matrix, methods, platform differences, and the verifiable signal for each case live in `references/use_cases.md`. Read it for exact event names and preconditions; this file is the operating guide. **The organizing principle:** every RevenueCat behavior has a **verifiable signal** — a specific dashboard event (`INITIAL_PURCHASE`, `TRANSFER`, `EXPIRATION`, `PRODUCT_CHANGE`, `CANCELLATION`), a `period_type` value, or a CustomerInfo/debug-log state. A test isn't "did the app not crash"; it's "did the expected event appear in the customer's history with the right fields." Always tie a test to its signal, or it isn't really testing anything. ## How to use this skill 1. **Get the setup facts**: platform(s) (iOS / Android / cross-platform / web billing), SDK (native, RN, Flutter, etc.), whether they've enabled debug logs and sandbox accounts yet, and what specifically they're trying to verify or why they think something is broken. 2. **Establish prerequisites first** — most "purchases don't work" reports are config problems visible before any purchase: - Debug logs on *before* `configure()` (iOS `Purchases.logLevel = .debug`; Android `setLogLevel(LogLevel.DEBUG)`), and check for **"Invalid Product Identifiers"** and error-level logs. Fix those before touching purchase flows. - Sandbox environment set up; multiple sandbox test IDs ready if any account-switching is in scope. - Point them at the debug overlay (`debugRevenueCatOverlay()` / `DebugRevenueCatBottomSheet`) to preview offerings and run test purchases quickly. 3. **Map their goal to use cases** from the reference and produce the plan (below). **Reference implementation (especially for Flutter):** RevenueCat ships an official **Purchase Tester** sample app at `github.com/RevenueCat/purchases-flutter/tree/main/revenuecat_examples/purchase_tester`. It's a runnable app exercising the flows this skill tests — dedicated screens for product change (`product_change_testing_screen.dart`), paywalls and paywall-footer, customer center, virtual currency, winback offers, and custom paywall-impression testing — plus an end-to-end integration test at `integration_test/app_test.dart`. Point Flutter users there to (a) run a known-good build to isolate whether a bug is in their code vs. their store/RevenueCat config, and (b) model their own `integration_test` widget tests on `app_test.dart`. For non-Flutter SDKs, RevenueCat has equivalent Purchase Tester apps in each SDK repo. ## The six testing domains Pull the specific rows from `references/use_cases.md`; here's the shape so you know what to cover: 1. **Subscription lifecycle** — purchase, free trial/intro offer (`period_type` = `TRIAL`/`INTRO`), renewal, upgrade/downgrade (`PRODUCT_CHANGE`), cancellation, expiration, refund. **Platform gotcha to always flag:** iOS sandbox refunds are not possible — the App Store routes users to Apple support and the `CANCELLATION`/`CUSTOMER_SUPPORT` event can take ~24h; Google Play refunds are dashboard-driven. 2. **Configuration** — anonymous vs identified App User IDs (verify via debug log + CustomerInfo), offering display, and login/logout — where the key risk is **unintended data merging** between accounts. 3. **Restoring flow** — the account-switching/transfer matrix. This is the highest-value, most-bug-prone area: `syncPurchases()`, restore-to-new-ID, the three-ID conflict case (transfer succeeds for the empty ID, errors for the one that already owns the sub), and transfer-disabled behavior. The signal throughout is whether a `TRANSFER` event fires (or correctly does *not*). 4. **Paywalls** — display & purchase, localization (device/store-account country → language + localized price), and intro-offer eligibility display for fresh users. 5. **RevenueCat Billing (web)** — purchase, trial auto-conversion, expiration. 6. **Before an experiment** — the app must render the SDK's *current* offering (or the placement's current offering), never a hard-coded one, or customers don't split evenly between variants; and the installed SDK must meet the feature-gate minimum for anything the experiment relies on. Setting up and running the experiment itself is the RevenueCat AI Toolkit's `revenuecat-experiments` skill. ## Deliverables - **QA test plan** — a checklist of the use cases relevant to their app (skip web billing if they're mobile-only, skip trials if they have none, add the pre-experiment checks when an experiment is planned), each row as: *action → expected event/signal → where to see it*. This is the primary output; make it copy-pasteable into a QA doc or issue tracker. - **Debugging a specific failure** — diagnose against the reference: is it a prereq problem (invalid product IDs, logs off, wrong sandbox account)? a wrong-signal expectation (e.g. expecting `TRANSFER` when transfer is disabled)? a platform difference (iOS refund can't be sandbox-tested)? Name the expected event and where it should appear, then work backward from what they actually see. - **Restore/transfer walkthrough** — when the user is confused about account switching, walk the specific ID configuration and the expected `TRANSFER`-or-error outcome for each ID, since this is the area the docs spend the most care on. - **Sandbox setup guidance** — the prerequisites section, adapted to their platform and SDK. Tie every recommendation to the verifiable signal in `references/use_cases.md` — the event cheat-sheet at the bottom of that file maps each event to what it proves.
© Ivanna Kaceviča · MIT · evanca/flutter-ai-rules