Token导航 LogoToken导航

Agent Skill · 研究检索

app-navigator(应用程序导航器)

app-navigator 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

app-navigator 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,权限为操作浏览器;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 71,089。

总安装 196周安装 8GitHub Stars 17下载 63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:app-navigator(应用程序导航器)
来源仓库:https://github.com/buildbetter-app/bb-skills
仓库路径:skills/app-navigator
安装命令:
npx skills add https://github.com/buildbetter-app/bb-skills --skill app-navigator
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/buildbetter-app/bb-skills --skill app-navigator

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。

app-navigator 用于查找、检索和筛选相关信息。适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词或任务场景快速定位候选结果。通过 npx skills add 命令从指定 GitHub 仓库安装,需结合原始 README 核验具体用法。

安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。当前功能聚焦于构建应用 UI 地图和 Playwright 测试脚本,支持浏览器端验证。

来源仓库内路径:skills/app-navigator

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/buildbetter-app/bb-skills --skill app-navigato…;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

    它需要操作浏览器,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

app-navigator 怎么安装?

在终端执行 npx skills add https://github.com/buildbetter-app/bb-skills --skill app-navigator。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

app-navigator 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。

app-navigator 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

app-navigator 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。

app-navigator 用的人多吗?

公开数据是累计安装 196 次,近一周安装 8 次,下载 63 次,GitHub 17 星。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。

app-navigator 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
app-navigator

应用程序导航器 · 本页

19617操作浏览器未明确未通过

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

App Navigator

Map your application's UI and build reusable Playwright playbooks for browser-based testing and verification.

Core principle: Know your app before you test it. This skill creates a living knowledge base of routes, pages, components, and interactions that other skills (like trust-but-verify) build on.

When to Use

  • Setting up browser-based testing for the first time in a project
  • After significant UI changes that may have shifted routes or layouts
  • Before running trust-but-verify, if no app-map.md exists yet
  • When you need to understand the app's page structure

Not for: API-only projects, backend-only changes, or projects without a web UI.

Invocation

  • /app-navigator setup -- map the app and write playbooks
  • /app-navigator audit -- UI/UX audit (v2, not yet implemented)
  • If no argument given and no app-map.md exists, runs setup mode automatically

Setup Mode Process

digraph setup {
    "Check credentials" [shape=box];
    "Creds in memory?" [shape=diamond];
    "Ask user + save" [shape=box];
    "Check dev server" [shape=box];
    "Server reachable?" [shape=diamond];
    "Tell user what to start" [shape=box];
    "User authorizes start?" [shape=diamond];
    "Start server" [shape=box];
    "Login via browser" [shape=box];
    "Walk top-level routes" [shape=box];
    "For each page" [shape=box];
    "Screenshot + document" [shape=box];
    "Map to source component" [shape=box];
    "Write playbooks" [shape=box];
    "Save app-map.md + component-map.md" [shape=box];
    "Done" [shape=doublecircle];

    "Check credentials" -> "Creds in memory?";
    "Creds in memory?" -> "Check dev server" [label="yes"];
    "Creds in memory?" -> "Ask user + save" [label="no"];
    "Ask user + save" -> "Check dev server";
    "Check dev server" -> "Server reachable?";
    "Server reachable?" -> "Login via browser" [label="yes"];
    "Server reachable?" -> "Tell user what to start" [label="no"];
    "Tell user what to start" -> "User authorizes start?";
    "User authorizes start?" -> "Start server" [label="yes"];
    "User authorizes start?" -> "Login via browser" [label="no, user starts manually"];
    "Start server" -> "Login via browser";
    "Login via browser" -> "Walk top-level routes";
    "Walk top-level routes" -> "For each page";
    "For each page" -> "Screenshot + document";
    "Screenshot + document" -> "Map to source component";
    "Map to source component" -> "For each page" [label="next page"];
    "Map to source component" -> "Write playbooks" [label="all pages done"];
    "Write playbooks" -> "Save app-map.md + component-map.md";
    "Save app-map.md + component-map.md" -> "Done";
}

Step 1: Credential Check

Read the memory file at ~/.claude/projects/<project>/memory/reference_local_auth.md.

If it exists, extract: Email, Password, App URL, Login path.

If it does NOT exist, ask the user:

"I need login credentials for the local dev environment. Please provide: 1. Email address 2. Password 3. App URL (default: http://localhost:5173) 4. Login path (default: /login) These will be saved to your project memory for future use."

Save their response to ~/.claude/projects/<project>/memory/reference_local_auth.md:

---
name: local-auth-credentials
description: Local dev login credentials and app URL for Playwright verification
type: reference
---

- Email: <user-provided>
- Password: <user-provided>
- App URL: <user-provided, default http://localhost:5173>
- Login path: <user-provided, default /login>

Update MEMORY.md with a pointer to this file.

Step 2: Preflight -- Dev Server Check

Run:

curl -s -o /dev/null -w "%{http_code}" <App URL> 2>/dev/null || echo "unreachable"

If unreachable, tell the user:

"The dev server at isn't reachable. You'll need to start it. Start your development server (e.g., npm run dev, pnpm run dev). Start your API server if needed. Want me to start it for you, or will you handle it?"

If user authorizes, start the server. Otherwise, wait for user confirmation that it's running, then re-check.

Step 3: Login via Browser

Use Playwright MCP tools:

  1. mcp__playwright__browser_navigate to <App URL><Login path>
  2. mcp__playwright__browser_snapshot to see the login form
  3. mcp__playwright__browser_fill_form with email and password fields
  4. mcp__playwright__browser_click on the submit/login button
  5. mcp__playwright__browser_wait_for with text set to a known post-login element (e.g., a dashboard heading or username)
  6. mcp__playwright__browser_snapshot to confirm logged-in state

If the page redirects through an SSO/OAuth provider or shows an organization selector, follow the redirects and select the first organization.

If login fails, ask the user to verify credentials and update the memory file.

Step 4: Navigate & Map

For each top-level navigation route (sidebar items, nav bar links):

  1. mcp__playwright__browser_click on the nav item
  2. mcp__playwright__browser_wait_for with text set to a key element on the target page
  3. mcp__playwright__browser_snapshot to capture the page structure
  4. mcp__playwright__browser_take_screenshot with filename set to an absolute path (resolve ~ via echo $HOME first), e.g., /Users/<user>/.claude/skills/app-navigator/screenshots/<page-slug>.png
  5. Record: URL, page title, key elements visible, available interactions

To map to source components, dispatch a subagent with the setup-prompt.md template, providing the list of discovered routes. The subagent uses Grep and Glob to search src/ for route definitions and component files.

Step 5: Write Playbooks

Write three markdown playbooks based on what was discovered:

playbooks/auth.md -- Document the exact MCP tool call sequence to log in:

  • Which URL to navigate to
  • Which form fields exist (their ref values from snapshots)
  • What to fill in each field
  • Which button to click
  • What to wait for after submission (specific text)
  • How to handle SSO/OAuth redirects or org selector if present

playbooks/navigation.md -- For each mapped page:

  • The nav element to click (with ref or text description)
  • The URL it navigates to
  • What text to wait for to confirm the page loaded
  • Key landmark elements on the page

playbooks/interactions.md -- Common interaction patterns discovered:

  • How to open modals (which buttons trigger them)
  • How to fill forms (common form patterns)
  • How to create/delete items
  • How to use dropdowns, date pickers, etc.

Step 6: Save Documentation

Write app-map.md with this structure:

# App Map

**Generated:** YYYY-MM-DD
**App URL:** <url>

## Routes

### [Page Name]
- **URL:** /path
- **Purpose:** What this page does
- **Key Elements:** List of important UI elements
- **Interactions:** What you can do on this page
- **Screenshot:** ./screenshots/<page-slug>.png
- **Source Component:** src/path/to/Component.tsx

Write component-map.md:

# Component Map

| Route | URL | Component | File |
|-------|-----|-----------|------|
| Page Name | /path | ComponentName | src/path.tsx |

Cleanup

After all mapping is complete, close the browser with mcp__playwright__browser_close.

Red Flags

  • Never hardcode credentials in SKILL.md or playbooks. Always read from memory.
  • Never run services without asking. Check first, ask permission, then start.
  • Don't map parametric sub-pages (like /users/:id). Stick to top-level nav routes.
  • Don't screenshot behind auth walls you can't pass. If a page requires special permissions, note it and skip.

After Setup: What's Next

When setup completes, print a summary of what was created, then offer:

Ready to verify a feature branch? Run /trust-but-verify to check that your UI implementation matches the plan. It uses the app map and playbooks you just created to: 1. Read your ExecPlan + git diff + PR description 2. Build a verification checklist of what to test 3. Login and navigate the app with Playwright 4. Test happy paths, edge cases, error states, and responsive layouts 5. Produce a report at docs/verification/ with findings and screenshots You can run it anytime on any feature branch. Type /trust-but-verify to start.

If the user says yes, invoke the trust-but-verify skill.

Integration

  • Used by: trust-but-verify (reads app-map.md and playbooks)
  • Credentials shared with: Any skill that reads reference_local_auth.md from memory

平台分布

Codex

33.9%
按下载量换算21

Claude

31.23%
按下载量换算20

Cursor

21.08%
按下载量换算13

Gemini CLI

10.17%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

Snyk

未通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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