Token导航 LogoToken导航

Agent Skill · 研究检索

finalrun-test-runner(最终运行测试运行程序)

用于辅助测试设计、自动化测试、用例整理和回归验证。它适合让 Agent 编写单元测试、端到端测试、测试计划或根据失败日志定位问题。使用时需要确认项目测试框架、运行命令和夹具数据,避免为了通过测试而改坏真实逻辑;涉及浏览器或外部服务时,应区分本地模拟、测试环境和生产环境。

finalrun-test-runner 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为调用外部服务;许可证信息还不明确,第三方扫描有未通过项,按安装量在研究检索类的 73,675 个 Skill 里排第 64,068。

总安装 306周安装 13GitHub Stars 公开资料未说明下载 107核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:finalrun-test-runner(最终运行测试运行程序)
来源仓库:https://github.com/final-run/agent-skills
仓库路径:skills/finalrun-test-runner
安装命令:
npx skills add https://github.com/final-run/agent-skills --skill finalrun-test-runner
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/final-run/agent-skills --skill finalrun-test-runner

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于辅助测试设计、自动化测试、用例整理和回归验证。finalrun-test-runner 属于研究检索类 Skill,可作为该场景下的辅助能力补充。适合编写单元测试、端到端测试或根据失败日志定位问题。使用时需确认项目测试框架、运行命令和夹具数据,避免误改逻辑。涉及浏览器或外部服务时,应区分本地模拟与生产环境。

安装前建议核对来源仓库和权限范围,确保安全合规使用。

来源仓库内路径:skills/finalrun-test-runner

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/final-run/agent-skills --skill finalrun-test-r…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要调用外部服务,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

finalrun-test-runner 怎么安装?

在终端执行 npx skills add https://github.com/final-run/agent-skills --skill finalrun-test-runner。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

finalrun-test-runner 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

finalrun-test-runner 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

finalrun-test-runner 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

finalrun-test-runner 用的人多吗?

公开数据是累计安装 306 次,近一周安装 13 次,下载 107 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

finalrun-test-runner 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
finalrun-test-runner

最终运行测试运行程序 · 本页

306—调用外部服务未明确未通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

FinalRun Test Runner

Orchestrate FinalRun MCP tools to run mobile app tests on cloud or local devices.

MCP Preflight

Before any test operation:

  1. Run ping.
  2. If ping fails, the FinalRun MCP server may not be installed or configured. Ask the user to verify their MCP setup.
  3. Resume this workflow only after MCP is healthy.

Core Concepts

  • Test: A single automated flow (one scenario).
  • Test Suite: An ordered group of tests executed together on a device.
  • Test Runs: Historical execution records of tests or suites that already ran.

Your Only Job

  1. Find the test or suite to run
  2. Select the target device(s) — cloud or local
  3. Map the app binary — use an existing app, or upload a new one if not available
  4. Confirm the run plan with the user
  5. Execute the test run
For creating tests, see finalrun-generate-test. For updating or deleting tests, see finalrun-update-test.

Required Inputs

  • Run Test: test name + appMapping + one compatible device target (cloud or local)
  • Run Test Suite: suite name + appMapping + one compatible device target (cloud or local)

User Input & Credentials

Tests often require user-specific data such as login credentials, form values, environment URLs, or account details. Follow these rules:

  1. Never guess or fabricate credentials, emails, passwords, API keys, or environment-specific values.
  2. Ask the user before proceeding if any required test input is unknown — this includes but is not limited to:

- Login credentials (username, email, password) - Form field values (addresses, phone numbers, payment details) - Environment-specific URLs or endpoints - Account-specific data (user IDs, org names, project names)

  1. Ask early — identify required inputs during Step 1 (finding the test) and resolve them before Step 4 (confirming the run plan).
  2. Include provided values in the run plan confirmation so the user can verify them.
Not asking the user for unknown inputs is a blocker — do not skip this step.

Platform Compatibility Rule (Strict)

  • Android app upload -> Android devices only
  • iOS app upload -> iOS devices only
  • Never map Android app to iOS device, or iOS app to Android device
  • For platform and autoSelectPlatform parameters, use these exact values: Android or IOS
  • In prose, "iOS" and IOS refer to the same platform; use IOS only when setting MCP arguments.

Workflow Steps

Step 1 — Find the Test or Suite

Search for the test or suite to run:

Use MCP tool: list_tests
Arguments: { "search": "<test name keyword>" }

Or for suites:

Use MCP tool: list_test_suites
Arguments: { "search": "<suite name keyword>" }

If the test or suite doesn't exist, follow the finalrun-generate-test workflow to create it first.

Step 2 — Select Target Devices

Cloud devices:

Use MCP tool: list_supported_devices
Arguments: { "platform": "Android" }  # or "IOS"

Note the requirementId for the target device(s).

Local devices:

Use MCP tool: list_local_devices
Arguments: {}

Note the uuid for the target device. Ensure local prerequisites are met (see Local Run Prerequisites section).

Step 3 — Resolve App Mapping

Search for the app binary to use:

Use MCP tool: available_apps
Arguments: { "search": "<app name>", "platform": "Android" }  # or "IOS"
  • If app exists — if the user specified an upload, use that exact appId + appUploadId; otherwise, ask the user to choose the upload from the returned platform-compatible list
  • If no app is available — upload a new binary following the Upload Routing Rule below

Platform inference:

  • If app has uploads for only one platform — platform is already determined, select matching devices automatically without asking the user
  • If app has uploads for both platforms — ask the user which platform to run on

Enforce platform compatibility: Android app → Android device, iOS app → iOS device.

Step 4 — Confirm the Run Plan

Present the run plan to the user for confirmation:

  • Test or suite name to run
  • Target device(s) and platform
  • App binary version (appId + appUploadId)
  • Cloud or local execution

Do not start the run until the user confirms the plan.

Step 5 — Execute the Run

Cloud — Single Test:

Use MCP tool: run_test_by_name_on_devices
Arguments:
  testName: "<test name>"
  devices: [{ "cloudRequirementId": "<requirement-id>" }]
  appMapping: { "<appId>": "<appUploadId>" }

Cloud — Test Suite:

Use MCP tool: run_test_suite_by_name_on_devices
Arguments:
  testSuiteName: "<suite name>"
  devices: [{ "cloudRequirementId": "<requirement-id>" }]
  appMapping: { "<appId>": "<appUploadId>" }

Local — Single Test:

Use MCP tool: run_test_locally
Arguments:
  testName: "<test name>"
  deviceUUID: "<device-uuid>"
  appMapping: { "<appId>": "<appUploadId>" }

Local — Test Suite:

Use MCP tool: run_test_suite_locally
Arguments:
  testSuiteName: "<suite name>"
  deviceUUID: "<device-uuid>"
  appMapping: { "<appId>": "<appUploadId>" }

To stop a running local test:

Use MCP tool: stop_local_test_run
Arguments: { "testRunId": "<test-run-id>" }

Local Run Prerequisites

Before executing a Local Test Run or Local Test Suite Run:

  • Android: Ensure adb is installed. Verify by running which adb. If not present, install via Homebrew (brew install android-platform-tools) or Android Studio.
  • iOS: Ensure xcrun is available (mac only). Verify by running which xcrun. If not present, run xcode-select --install.

Upload Routing Rule

When user asks to upload a new build, always ask:

  • "Should this upload go to an existing app or a new app container?"
  • "Confirm platform: Android or iOS?"

Then follow:

  1. Existing app path

- Call available_apps - Filter/select by platform (Android or IOS) - Pick target app node by appId, then read its appName - Call create_app_version(appName, filePath) using that exact appName

  1. New app path

- Confirm platform (Android or IOS) - Call create_app(name, appKnowledge?) -> get appId - Call create_app_version(appName, filePath) to upload first version

App Mapping

When running a test or suite, appMapping is required.

Identity rules:

  • appId identifies the app container (the same app record that includes appName).
  • appUploadId identifies the uploaded version/build under that app.

Canonical format:

{ "appId": "appUploadId" }

Use available_apps response to get valid pairs.

Device Targeting

For cloud runs, each device target must include exactly one:

  • cloudRequirementId — from list_supported_devices, targets a specific cloud device
  • autoSelectPlatform (Android or IOS) — auto-assigns any available device of that platform

For local runs:

  • Use list_local_devices to discover connected devices
  • Provide deviceUUID from the discovered device

Decision flow:

  1. Ask: Local or Cloud run?
  2. Cloud → Ask: auto-select or specific device?

- Auto-select → use autoSelectPlatform (derived from app platform) - Specific → use list_supported_devices → pick cloudRequirementId

  1. Local → use list_local_devices → pick deviceUUID

Two-Phase Confirmation Pattern

For update/delete bulk actions in other workflows (for example, finalrun-update-test):

Step 1: Call without confirm=true -> receive preview + confirmationToken
Step 2: Call with confirm=true + confirmationToken -> execute

Always show preview to user before confirm.

This two-phase pattern does not apply to run_test_by_name_on_devices, run_test_suite_by_name_on_devices, run_test_locally, or run_test_suite_locally.

Error Recovery

  • Auth/connectivity issue -> call ping
  • "No test found matching" -> call list_tests with search
  • Upload failure -> verify absolute filePath
  • No devices found -> verify platform filter and casing

Quick Checklist

  • MCP connectivity verified via ping
  • Test or suite identified via list_tests / list_test_suites
  • Target device(s) selected (cloud or local)
  • Platform compatibility verified (Android↔Android, iOS↔iOS)
  • App mapping resolved via available_apps
  • Run plan confirmed with user
  • Test run executed

Related Skills

  • Creating tests: See finalrun-generate-test workflow
  • Updating or deleting tests: See finalrun-update-test workflow

References

平台分布

Codex

35.12%
按下载量换算38

Claude

28.88%
按下载量换算31

Cursor

18.27%
按下载量换算20

Gemini CLI

9.74%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

未通过

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills