← 返回 Skills 市场
lazyboyjgn99

查找资料

作者 LazyBoyJgn99 · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ⚠ suspicious
311
总下载
1
收藏
1
当前安装
2
版本数
在 OpenClaw 中安装
/install find-news
功能描述
多引擎搜索 API 工具,支持通用网页搜索和新闻搜索。 使用场景: - 用户要求搜索网页、新闻、社交媒体内容 - 需要从 Google/Baidu/Bing/Yahoo/DuckDuckGo 搜索 - 需要搜索微信公众号、YouTube、GitHub、Reddit、Bilibili - 需要获取搜索结果的完整内容...
使用说明 (SKILL.md)

Find News 技能

概述

这个技能提供多引擎搜索能力,通过调用官方 API 实现快速搜索。开箱即用,跨平台支持(Windows/Mac/Linux)

API 接口

搜索接口

端点: POST http://36.151.144.35:3001/api/v1/search

认证: Bearer sk_test_a6f84bf78896f10b2d28aebd7857744c

请求参数:

  • query (string, 必需): 搜索关键词
  • search_service (string, 必需): 搜索服务名称
    • baidu: 百度搜索
    • google: Google 搜索
    • bing: Bing 搜索
    • duckduckgo: DuckDuckGo
    • yahoo: Yahoo 搜索
    • wechat: 微信公众号
    • youtube: YouTube
    • github: GitHub
    • reddit: Reddit
    • bilibili: 哔哩哔哩
  • max_results (number, 可选): 返回结果数量 (1-20, 默认 3)
  • crawl_results (number, 可选): 爬取完整内容的结果数 (0-10, 默认 0)
    • 0: 只返回搜索摘要(快速)
    • 1-10: 爬取完整页面内容(慢,成本高)

请求示例:

{
  "query": "人工智能",
  "search_service": "baidu",
  "max_results": 5,
  "crawl_results": 0
}

响应示例:

{
  "results": [
    {
      "title": "搜索结果标题",
      "url": "https://example.com",
      "snippet": "搜索结果摘要...",
      "content": "完整内容(仅当 crawl_results > 0 时)"
    }
  ],
  "total": 5,
  "cached": false
}

新闻接口

端点: POST http://36.151.144.35:3001/api/v1/news

认证: Bearer sk_test_a6f84bf78896f10b2d28aebd7857744c

请求参数:

  • query (string, 必需): 搜索关键词
  • search_service (string, 可选): 新闻服务名称
    • google (默认): Google News
    • bing: Bing News
    • duckduckgo: DuckDuckGo News
    • yahoo: Yahoo News
  • max_results (number, 可选): 返回结果数量 (1-20, 默认 3)
  • crawl_results (number, 可选): 爬取完整内容的结果数 (0-10, 默认 0)

请求示例:

{
  "query": "科技新闻",
  "search_service": "google",
  "max_results": 10,
  "crawl_results": 0
}

使用场景

场景 1: 快速获取搜索摘要(默认,推荐)

用户: "帮我搜索一下最新的 AI 新闻"

Agent 操作: 发送 POST 请求到新闻接口

{
  "query": "AI",
  "search_service": "google",
  "max_results": 5,
  "crawl_results": 0
}

→ 返回 5 条新闻标题和摘要(快速,0.5秒,每条 10 积分

场景 2: 深度内容抓取(仅在必要时使用)

用户: "我需要这篇文章的完整内容"

Agent 操作: 发送 POST 请求到搜索接口

{
  "query": "文章标题",
  "search_service": "baidu",
  "max_results": 1,
  "crawl_results": 1
}

→ 返回完整页面内容(慢,10-30秒,成本 10 积分

场景 3: 多平台搜索

用户: "在微信公众号和 B 站搜索'Python 教程'"

Agent 操作: 发送两个 POST 请求

请求 1 - 微信公众号:

{
  "query": "Python教程",
  "search_service": "wechat",
  "max_results": 5,
  "crawl_results": 0
}

请求 2 - 哔哩哔哩:

{
  "query": "Python教程",
  "search_service": "bilibili",
  "max_results": 5,
  "crawl_results": 0
}

场景 4: GitHub 代码搜索

用户: "搜索 GitHub 上的 react hooks 相关项目"

Agent 操作: 发送 POST 请求

{
  "query": "react hooks",
  "search_service": "github",
  "max_results": 10,
  "crawl_results": 0
}

场景 5: YouTube 视频搜索

用户: "在 YouTube 上搜索编程教程"

Agent 操作: 发送 POST 请求

{
  "query": "programming tutorial",
  "search_service": "youtube",
  "max_results": 5,
  "crawl_results": 0
}

定价

  • 每个搜索结果:10 积分
  • 缓存命中:免费(30 分钟内相同查询)
  • 相同查询 30 分钟内第二次调用免费

⚠️ 重要提示

性能

  • crawl_results=0(默认):0.5-2 秒
  • crawl_results=1-10:每个结果增加 10-30 秒

成本

  • 每个搜索结果:10 积分
  • 缓存命中:免费(30 分钟内相同查询)
  • 建议:优先使用摘要,只在必要时抓取完整内容

最佳实践

  1. 默认使用 max_results: 3, crawl_results: 0(快速获取摘要)
  2. 只有用户明确要求"完整内容"时 才设置 crawl_results > 0
  3. 搜索前 先检查是否有缓存(相同查询 30 分钟内免费)
  4. 合理控制数量 max_results 不要设置过大,避免不必要的成本

注意事项

  1. 默认使用 crawl_results: 0 以获得快速响应
  2. 设置 crawl_results > 0 会显著增加延迟(10-30秒/页)
  3. 注意控制请求频率,避免超出限流
  4. 相同查询 30 分钟内会使用缓存,免费且更快
  5. 所有请求必须在 Header 中包含 Authorization: Bearer sk_test_a6f84bf78896f10b2d28aebd7857744c
  6. Content-Type 必须设置为 application/json

错误处理

常见错误码:

  • 400: 请求参数错误
  • 401: 认证失败(API Key 无效)
  • 429: 请求频率超限
  • 500: 服务器内部错误

建议在请求失败时进行重试,最多重试 3 次。

安全使用建议
This skill will forward whatever the user asks to an unknown third‑party server (IP 36.151.144.35) using a hard-coded API key. Before installing, consider the following: 1) Do you trust the skill author and that server? Queries can contain sensitive info that will be sent off-platform. 2) Prefer skills that let you supply your own official API key (or call documented public APIs directly). 3) Avoid using crawl_results>0 unless you trust the backend (it requests and returns full page content). 4) Ask the publisher for provenance: where is the service hosted, privacy policy, logging/retention, and whether the embedded key is revocable. If you cannot verify the backend and its data handling, do not enable or use this skill for sensitive queries.
功能分析
Type: OpenClaw Skill Name: find-news Version: 1.0.1 The skill bundle defines a search tool that communicates with an external API over unencrypted HTTP using a raw IP address (36.151.144.35:3001) and includes a hardcoded API key (sk_test_a6f84bf78896f10b2d28aebd7857744c) in SKILL.md. While the functionality aligns with its stated purpose of web and news searching, the lack of transport encryption and the presence of hardcoded credentials represent significant security vulnerabilities that could lead to credential theft or data interception.
能力评估
Purpose & Capability
The skill's name/description (search/news aggregator) matches the runtime instructions (POST to a search/news API). However, instead of calling the documented upstream search providers directly, it calls a single third-party HTTP endpoint (http://36.151.144.35:3001) — that's plausible for an aggregator but unusual because the backend host is an IP address with no homepage or owner info supplied.
Instruction Scope
SKILL.md explicitly instructs the agent to send user queries and optionally full-page crawl requests to the remote endpoint and to include an Authorization header. That means any user-provided query text (possibly sensitive) will be transmitted to the remote service. The instructions do not limit or warn about sending sensitive data and encourage using crawl_results>0 which causes the backend to fetch full page contents.
Install Mechanism
No install spec and no code files are present (instruction-only). This minimizes local disk/execute risk — the skill's behavior is entirely network-bound to the described endpoint.
Credentials
No environment variables or user credentials are requested, but the SKILL.md includes a hard-coded Bearer token (sk_test_...). Embedding a credential in the instructions is unexpected and concerning because it both exposes a secret (even if 'test') and causes the agent to use the skill author's key for outgoing calls instead of letting the user provide their own API key or opt in.
Persistence & Privilege
The skill does not request elevated persistence (always is false), does not modify other skill settings, and is user-invocable. Autonomous invocation is allowed by platform default, but that alone is not an additional red flag here.
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 find-news
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /find-news 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
v1.0.1
- Major update: switched from local Search API scripts to a unified official API (cloud endpoint). - Removed local script dependency (`scripts/search.py`) for all search operations. - API key and endpoint configuration is now handled centrally; no client setup required. - All search and news queries are done via direct HTTP POST requests to the public API. - Updated authentication and example requests; usage is simplified and cross-platform. - Updated pricing from currency-based to point-based system and clarified caching behavior.
v1.0.0
- 首次发布,支持多引擎网页和新闻搜索(Google、Baidu、Bing、Yahoo、DuckDuckGo、WeChat、YouTube、GitHub、Reddit、Bilibili) - 增加新闻与通用搜索模式,可指定服务和返回结果数 - 支持抓取完整页面内容(可设 crawl_results 数量),满足深度内容需求 - 提供详细脚本使用说明和 API 调用示例 - 明确各项速率、定价、性能提示及最佳实践
元数据
Slug find-news
版本 1.0.1
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 2
常见问题

查找资料 是什么?

多引擎搜索 API 工具,支持通用网页搜索和新闻搜索。 使用场景: - 用户要求搜索网页、新闻、社交媒体内容 - 需要从 Google/Baidu/Bing/Yahoo/DuckDuckGo 搜索 - 需要搜索微信公众号、YouTube、GitHub、Reddit、Bilibili - 需要获取搜索结果的完整内容... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 311 次。

如何安装 查找资料?

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

查找资料 是免费的吗?

是的,查找资料 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

查找资料 支持哪些平台?

查找资料 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 查找资料?

由 LazyBoyJgn99(@lazyboyjgn99)开发并维护,当前版本 v1.0.1。

💬 留言讨论