AIRS-具身智能招投标采集器

Name: AIRS-具身智能招投标采集器
Author: airs-guest

功能描述

AIRS 具身智能天眼查招投标数据查询工具。查询企业在天眼查平台的招投标/中标公示信息，导出结构化 CSV 报表，基于浏览器自动化技术实现。 Keywords: AIRS, 具身智能, 天眼查, 招投标, embodied intelligence, bidding, tianyancha

使用说明 (SKILL.md)

When to Use

当用户需要以下场景时触发此技能：

查询、导出企业在天眼查平台上的招投标/中标/投标公示信息
批量查询一批企业的招投标历史记录
按时间范围和金额筛选企业中标信息
基于企业名单文件进行招投标数据查询
交互式查询单个企业的招投标记录

典型用户请求：

"查询宇树科技的招投标记录"
"导出这些企业的中标信息"
"采集 2026 年第一季度的招投标数据"
"搜索乐聚机器人的中标项目"

Requirements

系统要求

Node.js: >= 18.0.0
操作系统: macOS / Windows / Linux
Chrome 浏览器: 已安装并可运行

前置准备

1. 安装 Node.js

如未安装，请访问 https://nodejs.org/ 下载 LTS 版本。

验证安装：

node --version  # 应显示 v18 或更高版本
npm --version

2. 启动 Chrome 远程调试

⚠️ 必须先关闭所有 Chrome 窗口，然后运行以下命令：

macOS:

/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome \
  --remote-debugging-port=9222 \
  --user-data-dir=/tmp/chrome_debug_profile

Windows:

"C:\Program Files\Google\Chrome\Application\chrome.exe" \
  --remote-debugging-port=9222 \
  --user-data-dir=%TEMP%\chrome_debug_profile

Linux:

google-chrome \
  --remote-debugging-port=9222 \
  --user-data-dir=/tmp/chrome_debug_profile

3. 登录天眼查

Chrome 启动后，访问 https://www.tianyancha.com 并完成登录。

Quick Start

# 进入脚本目录
cd scripts

# 安装依赖
npm install

# 检查环境状态
node cli.js status

# 搜索并确认企业信息
node cli.js search

# 下载招投标记录
node cli.js download --start-date 2026-01-01 --end-date 2026-03-31

# 交互式查询单个企业
node cli.js query "宇树科技"

Commands

`status` - 环境状态检查

检查 Node.js、Chrome 连接、npm 依赖等环境状态。

node cli.js status

`search` - 企业搜索确认

从企业名单搜索天眼查信息，补全企业全称和链接。

本地优先匹配： 如果企业在 assets/具身智能中游企业数据库.md 中已有「天眼查企业全称」和「天眼查链接」，则直接使用本地数据，不再访问天眼查搜索。仅对缺少天眼查信息的企业才发起在线搜索。

# 使用默认企业名单
node cli.js search

# 使用自定义名单
node cli.js search --company-file /path/to/custom.md

输出： data/company_list.csv

`download` - 批量下载招投标记录

基于已确认的企业列表，批量下载招投标记录。

# 使用默认参数（本季度）
node cli.js download

# 指定时间范围和金额门槛
node cli.js download \
  --start-date 2026-01-01 \
  --end-date 2026-03-31 \
  --min-amount 100

参数：

参数	类型	默认值	说明
`--start-date`	string	本季度第一天	开始日期 (YYYY-MM-DD)
`--end-date`	string	今天	结束日期 (YYYY-MM-DD)
`--min-amount`	number	0	最低金额门槛（万元）

输出： data/bidding_records.csv

`query` - 交互式单企业查询

交互式查询单个企业的招投标记录，支持模糊匹配。

# 交互式输入
node cli.js query

# 直接指定企业名称
node cli.js query "宇树科技"

# 指定查询参数
node cli.js query "宇树科技" --start-date 2026-01-01 --min-amount 50

交互流程：

输入企业名称（支持简称/全称/模糊匹配）
在本地企业数据库（assets/具身智能中游企业数据库.md）中匹配，命中则直接使用已有的天眼查全称和链接，无需在线搜索
如有多个匹配，显示列表供选择
输入时间范围和金额门槛
自动采集并保存结果

Data Format

输入：企业名单（Markdown 表格）

| 索引 | 企业名称 | 所属领域 | 产品名称 | 城市 | 天眼查企业全称 | 天眼查链接 |
|------|----------|----------|----------|------|----------------|------------|
| 1 | 宇树科技 | 本体 | Unitree H1 | 杭州 | 宇树科技股份有限公司 | https://... |

输出 1：企业列表（CSV）

路径：data/company_list.csv

字段	说明
索引	企业编号
企业简称(MD)	输入的简称
企业全称(天眼查)	天眼查完整名称
公司ID	天眼查公司ID
天眼查链接	企业详情页 URL
所属领域	行业领域
产品名称	主要产品
城市	所在城市
搜索状态	已确认/未找到/失败

输出 2：招投标记录（CSV）

路径：data/bidding_records.csv

字段	说明
企业名称	天眼查企业全称
项目名称	招投标项目标题
公告类型	中标公告/招标公告等
采购人	招采单位
中标金额	原始金额文本
发布日期	YYYY-MM-DD
天眼查详情页链接	项目详情 URL

Error Handling

错误场景	错误信息	解决方案
Chrome 未连接	`未检测到 Chrome 远程调试服务`	按前置准备步骤启动 Chrome
需要安全验证	`天眼查平台安全验证已触发`	在 Chrome 窗口中手动完成验证码/滑块
企业未找到	`未找到企业"XXX"的搜索结果`	检查企业名称准确性，或尝试简称
npm 依赖缺失	`Cannot find package`	运行 `npm install`
天眼查信息缺失	`该企业没有天眼查信息`	先运行 `node cli.js search`

Performance Guidelines

单次查询建议: 不超过 200 家企业
时间范围建议: 不超过 1 年，可分季度采集
请求频率: 自动 3-6 秒间隔，避免触发风控
安全验证: 如遇频率过高（>5次/50家），建议暂停 30 分钟

Expected Output

企业搜索确认报告

企业搜索完成：共 161 家国内企业
  已确认: 161 家
  未找到: 0 家
  海外跳过: 16 家

招投标记录摘要

招投标记录下载完成
  时间范围: 2026-01-01 至 2026-03-31
  金额门槛: 无门槛
  有记录企业: 28 / 161 家
  符合条件记录: 156 条

安全使用建议

This skill appears to do what it says (browser-driven scraping of Tianyancha and CSV export). Before installing/running, consider the following: 1) Run in an isolated environment (VM, container, or throwaway account) rather than on your main workstation. 2) Do NOT reuse your primary Chrome profile: follow the instructions but point --user-data-dir to a dedicated directory you control. 3) Be aware that enabling Chrome remote debugging (default port 9222) exposes the DevTools protocol on localhost — ensure no untrusted local users/processes can access that port, and bind it to localhost only. 4) Inspect package.json and package-lock.json, run npm install yourself and consider npm audit / reviewing dependencies. 5) Review the included assets file (assets/具身智能中游企业数据库.md) for any sensitive data before use. 6) The tool expects manual intervention when anti‑bot checks trigger; this is normal for scraping but may require human handling of captchas and can trigger site rate limits or ToS issues. 7) If you need stronger guarantees, run the tool inside a disposable VM or container and use a throwaway Tianyancha account.

功能分析

Type: OpenClaw Skill Name: embodied-bidding-tracker Version: 1.0.6 The skill bundle is a legitimate browser automation tool designed to scrape bidding and procurement data from the Tianyancha platform. It uses Puppeteer to connect to a user-initiated Chrome instance via remote debugging (port 9222), which is a standard technique for bypassing anti-bot measures by leveraging the user's active session. The code in `scripts/modules/biddingDownload.js` and `scripts/modules/companySearch.js` is focused entirely on DOM interaction and data extraction, saving results locally to CSV files in the `data/` directory. There is no evidence of data exfiltration, credential theft, or malicious prompt injection.

能力评估

✓ Purpose & Capability

Name/description match the implementation: the code uses puppeteer-core to automate Chrome, searches Tianyancha, parses results and writes CSV/Excel. Declared requirements (Node >=18, Chrome) align with the actual code and dependencies.

⚠ Instruction Scope

Runtime instructions require you to close Chrome and restart it with --remote-debugging-port and a user-data-dir, then log in to Tianyancha manually. This is necessary for puppeteer-core to attach to an existing Chrome, but it elevates local attack surface (exposes the DevTools protocol on a TCP port and creates/uses an alternate profile in a temp dir). The SKILL.md and antiCrawl code also direct you to manually complete captchas/logins and the tool reads/writes the local assets and data files; these behaviors are within scope but the remote debugging instruction is the primary security concern.

ℹ Install Mechanism

No remote download/install spec in the registry manifest; installation is local via npm (package.json and package-lock.json included). Dependencies are typical (puppeteer-core, xlsx, winston, csv-writer). Using npm is standard but brings the usual supply-chain considerations — the package-lock references a registry mirror (npmmirror) which is common in some regions but you should verify registry/trust if that matters to you.

✓ Credentials

The skill does not request environment credentials or config paths beyond an optional TIANYANCHA_DEBUG_PORT. All environment access (reading assets MD, writing data/*.csv, logging to data/tool.log) is proportional to the stated functionality.

✓ Persistence & Privilege

The skill is not always-enabled, does not claim elevated platform privileges, and only writes to its own project/data and assets files. It does not attempt to modify other skills or system-wide agent settings.

版本历史

v1.0.6

- Reverted skill version from 2.0.0 to 1.0.6 in SKILL.md. - No other content or functionality changes detected.

v1.0.5

- Added _meta.json file to the project. - No changes to functionality or documentation.

v1.0.4

- No file changes detected in this version. - Version remains unchanged; all features and documentation are identical to previous release. - No user-visible updates.

v1.0.3

No changes detected in code or documentation for this version. - No file or documentation updates in this release. - Functionality remains identical to the previous version.

v1.0.2

优化技能描述

v1.0.1

- Removed legacy scripts: `download_bidding.js`, `step1_search_companies.js`, and `step2_download_bidding.js`. - All core functionality consolidated under the unified `cli.js` entry point. - Streamlined codebase for easier maintenance and enhanced reliability.

v1.0.0

embodied-bidding-tracker v2.0.0 introduces a major refactor and new features: - Unified CLI entry point (`cli.js`) for all operations. - Added unified config management (`config.js`). - New interactive single-company query command (`query`). - Improved fuzzy matching and batch querying. - Enhanced cross-platform support (macOS/Windows/Linux). - Robust error handling and logging. - Now conforms to Agent Skill standards.

元数据

Slug embodied-bidding-tracker

版本 1.0.6

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 7

常见问题