<< All versions

Skill v1.0.0

currentLLM-judged scan95/100
yizong2-cloud/jiongyuan-app/easybangumi-ui-test
──Details
PublishedOctober 23, 2026 at 03:33 PM
Content Hashsha256:3da1eb8825be9296...
Git SHA5d583e78e95d
──Files
Files (1 file, 3.9 KB)
SKILL.md3.9 KBactive
SKILL.md · 90 lines · 3.9 KB

version: "1.0.0" name: easybangumi-ui-test description: Use when testing Pure Pure Bangumi (EasyBangumi) Android features through real adb-driven UI flows, including app launch, app data reset, local mock HTTP servers on Mac, device-to-host port mapping, UI hierarchy dumps, reusable page navigation, and evidence collection for manual or scripted feature verification.


Easybangumi Ui Test

Overview

This skill standardizes real-device testing for Pure Pure Bangumi with adb, local mock services, and repeatable UI evidence collection. Use it when the task is broader than one page or when you need reusable setup, navigation, and validation steps.

Quick Start

  1. Run the common helper script from the repo root:

./skills/easybangumi-ui-test/scripts/easybangumi_test.sh

  1. If you need a local mock server, start it with:

./skills/easybangumi-ui-test/scripts/easybangumi_test.sh serve /path/to/mock-dir 18080

  1. If the Android device must access a Mac-local service through 127.0.0.1, use:

./skills/easybangumi-ui-test/scripts/easybangumi_test.sh reverse 18080

  1. Launch and inspect the app:

./skills/easybangumi-ui-test/scripts/easybangumi_test.sh clear com.heyanle.easybangumi4.debug ./skills/easybangumi-ui-test/scripts/easybangumi_test.sh launch com.heyanle.easybangumi4.debug ./skills/easybangumi-ui-test/scripts/easybangumi_test.sh dump-ui

Workflow

1. Preflight

  • Confirm exactly which package to test: usually com.heyanle.easybangumi4.debug.
  • Prefer the current workspace build when validating current code.
  • Decide whether the test needs clean state:

clear if onboarding, config, or repository state matters.

  • Decide whether the test needs a mock service:

use reverse, not forward, when the app on the device should reach a Mac-local service via 127.0.0.1.

2. Execute the Real Flow

  • Prefer explicit am/monkey launch over tapping a launcher icon when multiple app variants exist.
  • Use dump-ui before important actions and after each state change that matters.
  • Drive the UI with tap, text, and swipe.
  • If onboarding or permissions interrupt the target flow, either complete them manually or document the exact blocking screen before continuing.

3. Validate Outcomes

  • Validate both device-side behavior and host-side evidence.
  • For network-backed flows, check the local mock server logs for requested paths.
  • For install/import flows, verify both the final dialog/toast state and the downstream page state after dismissing dialogs.
  • Record unexpected behavior as:

trigger, observed result, expected result, likely code area.

4. Clean Up

  • Stop any local mock server started for the test.
  • Remove temporary adb reverse mappings with unreverse.
  • Keep test artifacts only if they are useful for follow-up debugging.

Navigation

Read references/navigation.md when the task depends on page entry paths, common bottom-tab routes, or source-management navigation.

Test Strategy

Read references/test-strategy.md when you need a reusable checklist for scenario design, boundary analysis, or result reporting.

Script

Use scripts/easybangumi_test.sh for common setup and adb actions. Prefer the script over retyping raw commands when the action is generic and repeatable.

Supported subcommands:

  • serve <dir> [port]
  • stop-serve [port]
  • reverse <port>
  • unreverse <port>
  • clear <package>
  • launch <package>
  • dump-ui [output-file]
  • tap <x> <y>
  • text <value>
  • swipe <x1> <y1> <x2> <y2> [duration-ms]
  • screenshot [output-file]
  • current-focus

When To Escalate

  • If multiple installed packages share the same app name and launcher icon.
  • If the flow requires manual login, device-level permissions, or document picker interaction that cannot be inferred safely.
  • If observed UI text conflicts with expected code behavior and you need to decide whether to trust code or runtime.
All versions