← Back to Skills Marketplace
fireium

eShop

by fireium · GitHub ↗ · v1.1.0 · MIT-0
cross-platform ⚠ suspicious
348
Downloads
1
Stars
0
Active Installs
5
Versions
Install in OpenClaw
/install eshop
Description
一个聚合购物与外卖服务的智能平台，帮助用户完成商品搜索、下单购买、外卖点餐、订单查询、优惠券领取等全流程购物体验。支持多平台、多商家、多品牌的统一接入与智能路由。
README (SKILL.md)

eShop

概述

eShop 是电商和外卖的工具库，用于帮助用户购物或者点外卖。

顶层 skill.md 只负责：

  • 识别用户意图
  • 判断任务域
  • 识别平台、商家、品牌等实体
  • 将请求路由到对应子文档
  • 在必要时才加载详细说明

任务域

外卖

适用于以下请求：

  • 我要点餐
  • 我要点外卖
  • 帮我点麦当劳
  • 看看有什么吃的

对应子目录：

delivery/

电商

适用于以下请求：

  • 我要买东西
  • 帮我找商品
  • 看看某个平台上的商品
  • 我想买手机壳

对应子目录：

ecommerce/

路由原则

处理流程：

用户输入
-> 意图识别
-> 任务域判断
-> 实体抽取
-> 别名归一化
-> 文件映射
-> 按需加载子文档

规则：

  1. 先判断是外卖还是电商
  2. 再识别用户是否指定了平台、商家、品牌或店铺
  3. 只有在目标明确平台，商家，品牌信息的时候，才加载对应子文档
  4. 不要一次性加载所有子文档
  5. 若信息不足，先澄清，再决定是否加载

实体映射

当用户提到已知平台、商家或品牌时，先做别名归一化，再映射到文件路径。

示例：

  • 麦当劳 / McDonald's / mcd -> delivery/merchants/mcd.md
  • 肯德基 / KFC -> delivery/merchants/kfc.md
  • 必胜客 / Pizza Hut -> delivery/merchants/pizzahut.md
  • 骆岗/ Luogang / inLG -> ecommerce/platforms/luogang.md
  • 平台1 -> delivery/platforms/platform1.md

匹配优先级：

  1. 精确别名
  2. 常见缩写
  3. 标准名称
  4. 若有多个候选，先澄清
  5. 若没有命中，留在顶层通用流程

Routing Index

外卖商家

  • aliases: 麦当劳, McDonald's, McDonald, mcd, mcdonalds target: delivery/merchants/mcd.md

\x3C!-- - aliases: 肯德基, KFC, kfc target: delivery/kfc.md

  • aliases: 必胜客, Pizza Hut, pizzahut target: delivery/pizzahut.md

外卖平台

  • aliases: 平台1, 外卖平台1 target: delivery/platform1.md

  • aliases: 平台2, 外卖平台2 target: delivery/platform2.md

  • aliases: 平台3, 外卖平台3 target: delivery/platform3.md -->

电商平台

  • aliases: luogang,Luogang,inLG,骆岗, 骆岗电商, 骆岗门市部 target: ecommerce/platforms/luogang.md

澄清策略

以下情况先澄清，不直接加载子文档：

  • 未说明是外卖还是电商
  • 未指定平台或者品牌，但后续必须依赖平台规则
  • 实体可能对应多个目标

示例提问：

  • 你是想点外卖，还是买商品？
  • 你有指定平台吗？

示例

我要点麦当劳

处理：

外卖
-> 实体: 麦当劳
-> 归一化: mcd
-> 加载: delivery/mcd.md

我想买手机壳

处理：

电商
-> 未指定平台，默认使用 luogang
-> 再决定是否加载子文档

去骆岗看看耳机

处理：

电商
-> 实体: luogang
-> 加载: ecommerce/luogang.md
Usage Guidance
This skill appears to implement an aggregator for shopping and delivery and contains per-platform instructions that call external MCP endpoints and require user/API tokens. Before installing: 1) Ask the publisher to reconcile and publish a clear list of required environment variables (which tokens are required and why) and to explain how tokens are stored and protected. 2) Insist the skill always asks for explicit, per-order user confirmation before using any stored USER_TOKEN or API token and before creating orders or claiming coupons; remove instructions that say to use tokens/address 'without asking'. 3) Treat MCD_TOKEN/USER_TOKEN as sensitive — do not provide long-lived credentials to an unknown skill; prefer short-lived, per-session tokens or OAuth flows. 4) Verify the MCP endpoints (yuju-mcp.wxhoutai.com, mcp.mcd.cn, eshop.wxhoutai.com) are legitimate and acceptable for your data. 5) If you need limited risk, run the skill in a sandbox or with no tokens so it can only show product data, not place orders. Given the manifest inconsistencies and the potential for the agent to act on user credentials without clear consent, proceed cautiously or request fixes from the maintainer before use.
Capability Analysis
Type: OpenClaw Skill Name: eshop Version: 1.1.0 The skill bundle facilitates shopping and delivery services by using the high-risk 'execute_bash' tool to perform network requests via curl. While the logic aligns with the stated purpose, the instructions in 'ecommerce/platforms/luogang.md' explicitly direct the AI agent to use authentication tokens and default addresses without user confirmation, which bypasses typical security oversight. Additionally, 'delivery/merchants/mcd.md' requires a sensitive 'MCD_TOKEN' and uses shell execution for all API interactions (e.g., https://mcp.mcd.cn and https://yuju-mcp.wxhoutai.com). Although the McDonald's assistant includes warnings against logging credentials, the combination of shell-based networking and instructions to minimize user intervention with sensitive data presents a significant security risk.
Capability Assessment
ℹ Purpose & Capability
Overall purpose (e-commerce + delivery aggregation) matches the contents: routing to per-platform docs and calling MCP APIs to search, create orders, claim coupons, etc. However, registry metadata vs. included files disagree: the top-level registry declares no required env vars or binaries, while _meta.json and several sub-documents require MCP endpoint URLs and user/API tokens (e.g., SHOPPING_MCP_URL, MCD_TOKEN, USER_TOKEN, LUOGANG_MCP_HTTP_URL). That mismatch is incoherent — a legitimate aggregator should declare required endpoints/credentials clearly at the top level.
⚠ Instruction Scope
SKILL.md and the platform/merchant docs instruct the agent to call external MCP HTTP endpoints via curl and to collect/use user tokens and PII (addresses, phone). Two problematic patterns: (1) luogang.md says to use user token if available 'directly without user confirmation' and to use default shipping address without asking — this grants the agent authority to place orders autonomously; (2) mcd.md requires an MCD_TOKEN and the execute_bash tool for curl calls but the top-level instructions do not declare execute_bash or these env vars. The instructions also ask for creation of addresses and phone numbers — appropriate for ordering, but must be explicitly scoped and require user consent.
✓ Install Mechanism
Instruction-only skill with no install spec and no code files — low disk/write risk. All remote calls are performed via curl patterns described in documents; nothing is downloaded or executed by an installer. (However, some sub-docs mention 'execute_bash' tool requirement which is a runtime capability rather than an installer.)
⚠ Credentials
The published registry claims no required env vars, but _meta.json and merchant/platform docs list several environment/config variables and tokens: SHOPPING_MCP_URL (required in _meta.json), DEFAULT_PLATFORM_ID, MCD_TOKEN (required), MCD_MCP_URL, LUOGANG_MCP_HTTP_URL, LUOGANG_SHOP_ID, and implicit USER_TOKEN usage. These are not consistently declared in one place. Requesting user API/login tokens and telling the agent to use them without consistent, explicit consent is disproportionate unless the skill guarantees per-action confirmation and clear storage rules.
ℹ Persistence & Privilege
always: false (not force-included) and autonomous invocation is allowed (platform default). That combination is normal, but the skill's instructions (especially luogang.md) permit placing orders using stored tokens and default addresses without asking the user — this materially increases risk when the agent is allowed to invoke the skill autonomously. No explicit instructions about how tokens are stored or protected are provided.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install eshop
  3. After installation, invoke the skill by name or use /eshop
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.0
**eShop 1.0.4 introduces multi-domain (e-commerce & food delivery) intent routing and entity mapping.** - 新增外卖与电商融合的聚合平台架构，支持“点餐”“买东西”等多意图识别与路由。 - 增加7个子文档，覆盖主流外卖商家（麦当劳、肯德基、必胜客）及主流平台。 - 明确实体别名归一化与文件路由规则，实现平台、商家、品牌的多别名智能映射。 - 优化交互策略：信息不全时自动引导澄清，按需加载子文档，提升用户体验。 - 更新元数据标签及描述，支持多平台、多场景的购物与外卖统一入口。
v1.0.3
- Version updated to 1.0.3 - Added _meta.json for enhanced skill metadata management - Improved documentation: clarified default config values, streamlined setup instructions, and refined safety/privacy wording - Standardized variable usage for environment variables in example commands
v1.0.2
- 增加安全与隐私说明，明确数据外发和跳转链接风险，并提供相关操作建议 - 版本号由 1.0.0 升级为 1.0.2 - 其他功能和接口保持不变
v1.0.1
eShop v1.0.0 - 更新 Skill 描述，使其更简洁：由“电商聚合助手 - 支持多平台商品查询、详情展示及购买引导。默认平台为骆岗（luogang）。”修改为“电商购物助手 - 支持多个电商平台的商品查询、详情展示及购买引导。” - 其余功能、文档结构、示例与核心用法未发生变化。
v1.0.0
eShop 1.0.0 – 首个版本上线 - 支持多平台商品检索、详情展示与购买引导，默认平台为“骆岗（luogang）” - 集成 MCP 服务，实现实时商品数据查询 - 提供两个核心工具：search_products（商品检索）、get_product_detail（商品详情） - 配置灵活，支持平台切换及筛选条件设置 - 优化交互体验，专注核心信息展示，最多展示 10 个产品候选 - 引导用户跳转至平台官方 H5 或小程序进行下单
Metadata
Slug eshop
Version 1.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 5
Frequently Asked Questions

What is eShop?

一个聚合购物与外卖服务的智能平台，帮助用户完成商品搜索、下单购买、外卖点餐、订单查询、优惠券领取等全流程购物体验。支持多平台、多商家、多品牌的统一接入与智能路由。 It is an AI Agent Skill for Claude Code / OpenClaw, with 348 downloads so far.

How do I install eShop?

Run "/install eshop" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is eShop free?

Yes, eShop is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does eShop support?

eShop is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created eShop?

It is built and maintained by fireium (@fireium); the current version is v1.1.0.

More Skills
self-improving agent
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⬇ 463283 · by pskoett
Skill Vetter
Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
⬇ 259410 · by spclaudehome
Polymarket
Query Polymarket prediction markets. Check odds, find trending markets, search events, track price movements.
⬇ 232503 · by joelchance
Self-Improving + Proactive Agent
Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...
⬇ 200423 · by ivangdavila
Github
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
⬇ 191113 · by steipete
ontology
Typed knowledge graph for structured agent memory and composable skills. Use when creating/querying entities (Person, Project, Task, Event, Document), linkin...
⬇ 190156 · by oswalpalash

💬 Comments