← 返回 Skills 市场
oscarka

Meituan Physical Automation

作者 oscarka · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
32
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install meituan-automation
功能描述
Give your AI hands. This skill physically taps, swipes, and interacts with the Meituan App on a real Android phone — searching restaurants, picking items, ha...
使用说明 (SKILL.md)

🍜 Meituan Physical Automation

Your AI finally has hands. Now it can order your lunch.

This skill gives any AI Agent (like ClawBot) the ability to physically control a real Android phone to order food delivery on Meituan (美团) — the largest food delivery platform in China with over 700 million users.

No API keys. No Meituan developer account. No reverse engineering. Just the same taps and swipes a human would make — done entirely by your AI.


⚠️ Platform Compatibility

Platform Supported
🤖 Android Yes — requires ADB + USB Debugging
🍎 iOS No — Apple's sandbox prevents ADB-style physical control

This skill is Android-only by design. iOS devices don't support ADB or UIAutomator2, which are the underlying technologies that make physical screen control possible without reverse engineering the app. If you have an Android phone (any brand: Samsung, Xiaomi, Huawei, OnePlus, Pixel, etc.), you're good to go.


✨ What Makes This Different

Most "food delivery" AI tools rely on unofficial APIs, web scraping, or require developer credentials. This skill is different:

Other approaches This skill
Call unofficial HTTP APIs Physically tap the screen like a human
Risk account bans from bot traffic Indistinguishable from a real user
Break when apps update their API Works as long as the UI exists
Require app reverse engineering Zero reverse engineering
Stop at complex spec pop-ups Smart Visual Radar auto-handles multi-level spec selection (size, temperature, sweetness)
Silently fail on business errors Returns clear error messages ("¥7 short of minimum", "mandatory item missing")

🎯 What Your AI Can Do With This Skill

Just tell your AI what you want. For example:

"Order me a Grande Coconut Latte from the nearest Luckin Coffee."

Your AI will autonomously:

  1. 🔍 Search Meituan for nearby Luckin Coffee stores
  2. 🏪 Enter the correct store
  3. 📦 Switch to the delivery ordering tab
  4. 🔎 Search for "Coconut Latte" within the store menu
  5. ➕ Add it to cart (auto-handling size/temperature pop-ups)
  6. 🧾 Navigate to the checkout confirmation page

Then it stops and hands control back to you to pay. Your money stays 100% safe — the AI never completes payment.


🚀 One-Time Setup (3 steps, ~5 minutes)

Step 1 — Install the skill engine (your AI can do this for you):

pip install -r requirements.txt
python -m uiautomator2 init

Step 2 — Enable Developer Mode on your Android phone:

  • Go to Settings → About Phone
  • Tap Build Number 7 times until "Developer Mode" activates
  • Go to Settings → Developer Options → Enable USB Debugging

Step 3 — Connect your phone and verify:

adb devices
# Should show your device ID, e.g.: "0B211FDD4003PN    device"

That's it. Tell your AI you want food. It takes over from here.


🤖 AI Agent Execution Runbook

This section is for the AI reading this skill. Follow these steps precisely.

Phase 1: Environment Check (run once on first use)

  1. Run pip install -r requirements.txt and wait for completion.
  2. Run python -m uiautomator2 init to initialize the Android device driver.
  3. Run adb devices and parse the output. If no device is listed, tell the user: "Please connect your Android phone via USB and ensure USB Debugging is enabled in Developer Options."
  4. Run adb shell am start -n "com.sankuai.meituan/com.meituan.android.pt.homepage.activity.MainActivity" to bring Meituan to the foreground.

Phase 2: Ordering Workflow (strict sequence — do not skip steps)

# 1. Search for restaurants
python cli.py search "\x3Crestaurant_name>"
# Returns: list of stores with index numbers

# 2. Enter a store (try index 0 first)
python cli.py open \x3Cindex>
# Returns: "营业中" (open) or "打烊" (closed). If closed, try next index.

# 3. Switch to delivery tab (MANDATORY — do not skip)
python cli.py waimai

# 4. Search for the specific item within the store
python cli.py tap "搜索"
python cli.py type "\x3Citem_name>"

# 5. Add item to cart (visual radar handles all spec pop-ups automatically)
python cli.py add "\x3Citem_name>"

# 6. Go to checkout (STOPS HERE — never auto-pays)
python cli.py checkout
# Returns: estimated delivery time, total price, address status

Phase 3: Error Handling

  • If a command returns ERROR: read the HINT: on the next line and adapt.
  • If a store shows "门店已打烊" (closed), go back and try open \x3Cnext_index>.
  • If add fails, the item name on screen may differ slightly — try a shorter keyword.
  • If checkout shows needs_address: True, inform the user to add a delivery address on their phone.

📁 Files In This Package

File Purpose
SKILL.md This file — skill description + AI runbook
cli.py Main entry point — all commands go through here
meituan.py Core business logic — visual radar, cart handling, checkout parsing
device.py Low-level Android device abstraction (ADB + UIAutomator2)
requirements.txt Python dependencies

💡 Tips & Known Limitations

  • Store hours matter: Meituan stores have specific delivery hours. If a store is closed ("打烊"), ask your AI to try the next store in the search results.
  • Slider CAPTCHAs: High-frequency usage may occasionally trigger a human-verification slider on screen. Just swipe it manually — takes 2 seconds.
  • Delivery address: The first time you use the skill, you may need to add a delivery address manually on the checkout page. After that, it's saved.
  • Language: The Meituan app is in Chinese. The skill handles all Chinese UI navigation internally — you can give commands in English or Chinese.

Built with ❤️ by @oscarka · MIT-0 License

安全使用建议
Install only if you are comfortable letting an AI operate a real Android phone and a logged-in Meituan account. Use a dedicated or monitored device, review the unexpected Frida/MCP dependencies, avoid running the HTTP server unless secured, and require manual confirmation for permissions, checkout details, and all payment steps.
功能分析
Type: OpenClaw Skill Name: meituan-automation Version: 1.0.0 The skill provides deep automation for the Meituan app on Android via ADB, but exhibits several high-risk indicators. Most notably, 'requirements.txt' includes 'frida' and 'frida-tools'—powerful dynamic instrumentation tools often used for hooking and reverse engineering—which are never used in the code, while the actual required dependency 'uiautomator2' is missing. The skill also extracts sensitive PII (user's full address, name, and phone number) in 'meituan.py' and 'cli.py' exposes an unauthenticated HTTP server on '0.0.0.0:18080' that allows remote control of the device (tapping, typing, and screen scraping) by anyone on the local network.
能力标签
cryptocan-make-purchases
能力评估
Purpose & Capability
The stated purpose matches the core ADB/UI automation behavior, but the capability is high-impact: it can add items, enter checkout, read the phone UI, and send arbitrary taps/types on a real device.
Instruction Scope
The runbook encourages autonomous ordering through checkout, and the package also documents raw tap/type/screen-control interfaces that are broader than the Meituan-specific workflow.
Install Mechanism
There is no install spec or declared required binary, but SKILL.md tells the agent to run pip installs and uiautomator setup; requirements also include unexpected Frida and MCP packages for a claimed UI-only automation skill.
Credentials
Direct Android screen dumping, coordinate tapping, typing, Home/back controls, and HTTP control endpoints are more powerful than a narrowly scoped food-ordering workflow.
Persistence & Privilege
No background persistence or exfiltration is shown, but the code changes device screen settings and keeps small local caches, so users should restore settings if needed.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install meituan-automation
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /meituan-automation 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
First public release. Gives AI agents the ability to physically control a real Android phone to order food delivery on Meituan — end-to-end from restaurant search to checkout. Supports multi-level spec popups, smart visual radar button detection, and safe checkout (never auto-pays).
元数据
Slug meituan-automation
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Meituan Physical Automation 是什么?

Give your AI hands. This skill physically taps, swipes, and interacts with the Meituan App on a real Android phone — searching restaurants, picking items, ha... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 32 次。

如何安装 Meituan Physical Automation?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install meituan-automation」即可一键安装,无需额外配置。

Meituan Physical Automation 是免费的吗?

是的,Meituan Physical Automation 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Meituan Physical Automation 支持哪些平台?

Meituan Physical Automation 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Meituan Physical Automation?

由 oscarka(@oscarka)开发并维护,当前版本 v1.0.0。

💬 留言讨论