AgentSkillsCN

radar-tinder-mobileapp-testing

针对采用 Expo + React Native、Firebase（身份验证、Firestore、实时/云函数）、地图 API 以及设备端 ONNX 推理技术构建的 Radar Tinder 移动应用，开展测试与验证工作流程。本技能适用于回归测试、功能验证、实时问题排查、排行榜数据不一致、导航漏洞修复，以及 AI 诊断结果的验证。

SKILL.md
--- frontmatter
name: radar-tinder-mobileapp-testing
description: >
  Testing and validation workflows for the Radar Tinder mobile application built with Expo + React Native,
  Firebase (Auth, Firestore, Realtime/Cloud Functions), Maps APIs, and on-device ONNX inference.
  This skill should be used for regression testing, feature verification, realtime issues,
  leaderboard inconsistencies, navigation bugs, and AI diagnostics validation.

Radar Tinder – Mobile App Testing Skill

When to use this skill

Use this skill when:

  • •Adding or modifying features in the mobile app
  • •Investigating leaderboard, points, or ranking inconsistencies
  • •Debugging Firebase Auth, Firestore, or realtime update issues
  • •Verifying radar report & confirmation flows
  • •Testing navigation, nearby radar discovery, or directions logic
  • •Validating on-device ONNX AI inference (offline & online)

Quick Smoke Test (5–10 minutes)

These checks must pass before deeper testing:

  1. •App builds and launches (Expo / EAS)
  2. •Firebase Auth initializes without errors
  3. •Login / signup works (email, Google, Apple if enabled)
  4. •Firestore connection established
  5. •Leaderboard screen loads without crash

If any fail → check logs, Firebase config, and environment variables first.

Test Layers

A) Manual Smoke Tests (Happy Path)

1. Authentication & Profile

  • •Sign up with email
  • •Profile document is created in Firestore
  • •Username is unique
  • •Display name defaults correctly
  • •Avatar update reflects in leaderboard

2. Radar Reporting

  • •Create a radar report (location + type)
  • •Firestore document created
  • •Reporter receives points
  • •UI reflects updated points

3. Confirmations

  • •Second user confirms radar
  • •Confirmation stored in Firestore
  • •Points awarded correctly
  • •Duplicate confirmations prevented

4. Leaderboard & Realtime

  • •Leaderboard updates after report/confirmation
  • •Changes propagate without app restart
  • •Ordering is correct (points / rank)

5. Navigation & Maps

  • •Directions API returns valid route
  • •Nearby radars load from Firestore
  • •Radar alerts appear at correct distances
  • •Map markers update correctly

6. AI Diagnostics (ONNX)

  • •Sample image runs inference successfully
  • •Output labels/confidence are readable
  • •Works offline (no network dependency)
  • •No UI freeze during inference

Firebase-Specific Validation

Firestore Expectations

  • •Collections are structured consistently
  • •Writes are idempotent where needed
  • •Security rules allow valid actions only
  • •No client-side rank calculation

Realtime / Updates

  • •Firestore listeners are scoped correctly
  • •No duplicate subscriptions
  • •Updates clean up on unmount

Auth

  • •Firebase Auth state syncs correctly
  • •Token refresh does not break sessions
  • •Sign-out clears local state

Common Debugging Playbooks

“Leaderboard not updating”

  1. •Firestore listener active?
  2. •Query orderBy / limit correct?
  3. •Security rules blocking reads?
  4. •Local state cache stale?

“Points not awarded”

  1. •Report written successfully?
  2. •Cloud Function triggered?
  3. •Transaction failed silently?
  4. •Duplicate prevention logic firing?

“Login works but profile missing”

  1. •Auth success but Firestore write failed?
  2. •Network race condition?
  3. •Security rules deny create?

Test Data Strategy

Use consistent test fixtures:

  • •2 test users (reporter / confirmer)
  • •2–3 radar reports
  • •1 confirmation
  • •1 sample image for AI
  • •1 directions route fixture

How Copilot should respond when using this skill

For each test or bug request:

  1. •Identify affected flow (Auth / Radar / Confirm / Leaderboard / Maps / AI)
  2. •Provide minimal reproduction steps
  3. •Suggest fix hypothesis
  4. •Recommend tests to add/update
  5. •Output verification checklist

Output Format

Copilot responses should follow:

  • •Reproduction Steps
  • •Expected vs Actual
  • •Root Cause Hypothesis
  • •Fix Plan
  • •Tests Added or Updated
  • •How to Verify