Simple Websearch

极简网络搜索技能 - 支持通用搜索 + 社交媒体搜索（小红书/知乎/微博）

Security Scan

The skill mostly does what it says (web and social-media search) but contains inconsistencies: it calls hardcoded, external helper scripts via subprocess (absolute /home/lin paths) that are not declared in SKILL.md or in requirements, which is unexpected and risky.

Purpose & Capability

Name/description match the code's intent (general + social-media search). However the Python code invokes external helper scripts at absolute paths (/home/lin/.openclaw/scripts/exa-search.sh and /home/lin/.openclaw/workspace/skills/dynamic-webfetch/scripts/fetch.py). Those external scripts are not included, not documented in SKILL.md, and are environment-specific — requiring local helper tools is disproportionate to a self-contained search skill and may cause it to execute arbitrary local code.

Instruction Scope

SKILL.md describes pure HTTP-based searching and optional Playwright. The runtime code does perform HTTP scraping (requests) which is expected, but also runs subprocesses to call external scripts for 'Exa' results and dynamic webfetch. The SKILL.md does not disclose these subprocess invocations or the absolute filesystem paths referenced, so the documented runtime behavior is incomplete and grants the skill the ability to execute local programs outside the package.

Install Mechanism

There is no install specification in the skill registry (instruction-only), but the package contains setup.py, requirements.txt, and multiple scripts — so it is not purely instruction-only. Installation is local (pip requirements only). No remote downloads or URL-based installers were found, which lowers installer risk, but the mismatch between 'no install spec' and presence of installable package is inconsistent and worth noting.

Credentials

The skill declares no environment variables or credentials (good). However, it relies on absolute paths to other local skill scripts and will execute them via subprocess. That behavior does not require new credentials but does assume and potentially execute code outside its package, which is not proportional to the documented requirements and increases risk (especially if those external scripts send data off-host).

Persistence & Privilege

The skill is not marked always:true and does not request persistent elevated privileges. It does not modify other skills' configs in the observed files. Autonomous invocation is allowed (platform default) but is not combined here with additional privilege requests.

What to consider before installing

This skill largely implements web scraping/search logic and is probably useful for searches, but there are notable red flags you should consider before installing or running it: - Hardcoded external script calls: The code calls /home/lin/.openclaw/scripts/exa-search.sh and /home/lin/.openclaw/workspace/skills/dynamic-webfetch/scripts/fetch.py via subprocess. These files are not included in the package or documented in SKILL.md. If those paths exist on your system they will be executed with your privileges; if they do not exist the skill will silently degrade. - Undocumented execution surface: SKILL.md describes HTTP scraping and optional Playwright, but does not mention executing local helper scripts. Treat the skill as executing local programs in addition to making network requests. - Inspect external helpers: If you plan to use this, locate and review the referenced exa-search.sh and dynamic-webfetch fetch.py scripts on your system (or ask the author for them). Ensure they are trustworthy before allowing the skill to run. - Run in isolation first: Execute the skill in a sandboxed or ephemeral environment (container/VM) to observe its behavior, especially the subprocess calls and any outbound network traffic. - Optional Playwright: If you enable Playwright, it will install and run a browser (chromium) which increases resource/attack surface; only enable if needed. - Package vs. documentation mismatch: The skill is deliverable as a pip package (setup.py) but registry metadata claimed instruction-only — be cautious about installing it system-wide. Confidence is medium because the visible code explains the suspicious behavior (subprocess calls) but the referenced helper scripts are not present so the full runtime effects cannot be determined without those files. If you can provide the contents of the referenced exa-search.sh and dynamic-webfetch scripts (or confirm they don't exist on your system), I can raise or lower the risk assessment accordingly.

Simple Web Search Skill V2

极简网络搜索技能增强版，无需任何 API Key，开箱即用。

🆕 V2.0 新特性

• 🔍 通用搜索 - Exa AI + 必应 + 百度，三引擎智能降级
• 📱 社交媒体搜索 - 小红书/知乎/微博，一键搜索全网热点
• ⚡ 稳定可靠 - 国内网络优化，保证可用性
• 📦 零依赖 - 仅需 requests，可选 Playwright 增强

---

特点

• 🚀 Exa AI 优先 - OpenCode 同款搜索接口，免费无需 API Key
• 🔍 多引擎备用 - Exa → 必应 → 百度，自动降级保证有结果
• 📱 社交搜索 - 支持小红书/知乎/微博内容搜索
• 🎯 结果精准 - 优化过的正则提取，过滤广告
• 📦 开箱即用 - 无需配置，安装即可搜索
• ⚡ 轻量快速 - 仅需 requests 一个 Python 依赖

---

搜索引擎策略

通用搜索

1. Exa AI 优先 - OpenCode 同款，免费、快速、质量好
2. 必应备用 - Exa 无结果时自动切换
3. 百度补充 - 中文内容补充

社交媒体搜索

1. 小红书 - 种草笔记、产品评测、生活分享
2. 知乎 - 专业问答、深度讨论、知识分享
3. 微博 - 实时热点、新闻动态、话题讨论

---

使用方法

基本搜索（通用）

result = main({
    "action": "search",
    "query": "Python 教程",
    "num_results": 5,
    "type": "general"  # 默认值，可省略
})

社交媒体搜索

# 搜索所有平台
result = main({
    "action": "search",
    "query": "AI Agent",
    "num_results": 10,
    "type": "social_media"
})

# 只搜索小红书
result = main({
    "action": "search",
    "query": "护肤品推荐",
    "platform": "xiaohongshu",
    "type": "social_media"
})

# 只搜索知乎
result = main({
    "action": "search",
    "query": "如何学习编程",
    "platform": "zhihu",
    "type": "social_media"
})

# 只搜索微博
result = main({
    "action": "search",
    "query": "今日热搜",
    "platform": "weibo",
    "type": "social_media"
})

命令行

# 通用搜索
python3 scripts/search_v2.py "搜索关键词" 5 general

# 社交媒体搜索
python3 scripts/search_v2.py "AI Agent" 10 social_media

# 指定平台
python3 scripts/search_v2.py "护肤" 5 social_media

---

输入参数

参数	类型	必填	说明
action	string	是	操作类型：search / test
query	string	是	搜索关键词
num_results	int	否	结果数量，默认 5，最大 20
type	string	否	搜索类型：general / social_media
platform	string	否	社交媒体平台：xiaohongshu / zhihu / weibo / all

---

输出格式

通用搜索结果

{
    "success": true,
    "query": "搜索关键词",
    "type": "general",
    "engine": "exa+bing",
    "num_results": 5,
    "results": [
        {
            "title": "结果标题",
            "href": "https://...",
            "body": "摘要内容"
        }
    ],
    "message": "搜索完成",
    "timestamp": "2026-04-09T08:00:00Z"
}

社交媒体搜索结果

{
    "success": true,
    "query": "搜索关键词",
    "type": "social_media",
    "platforms": ["xiaohongshu", "zhihu", "weibo"],
    "num_results": 3,
    "results": [
        {
            "title": "[小红书] 结果标题",
            "href": "https://www.xiaohongshu.com/...",
            "body": "小红书笔记",
            "source": "xiaohongshu"
        },
        {
            "title": "[知乎] 问题标题",
            "href": "https://www.zhihu.com/...",
            "body": "知乎问答",
            "source": "zhihu"
        }
    ],
    "message": "搜索完成 (平台：xiaohongshu, zhihu)"
}

---

示例

示例 1：搜索新闻

result = main({
    "action": "search",
    "query": "AI 最新进展 2026",
    "num_results": 8
})

if result["success"]:
    for r in result["results"]:
        print(f"- {r['title']}: {r['href']}")

示例 2：搜索股票信息

result = main({
    "action": "search",
    "query": "三六零 601360 股价",
    "num_results": 5
})

示例 3：小红书种草

result = main({
    "action": "search",
    "query": "敏感肌护肤品",
    "platform": "xiaohongshu",
    "type": "social_media"
})

示例 4：知乎专业问答

result = main({
    "action": "search",
    "query": "如何评价 GPT-5",
    "platform": "zhihu",
    "type": "social_media"
})

示例 5：微博热点

result = main({
    "action": "search",
    "query": "今日热搜",
    "platform": "weibo",
    "type": "social_media"
})

示例 6：测试搜索引擎可用性

result = main({
    "action": "test"
})

# 输出：
# {
#     "success": true,
#     "results": {
#         "exa": true,
#         "bing": true,
#         "baidu": true
#     }
# }

---

依赖

仅需一个 Python 包：

pip install requests

可选依赖（社交媒体搜索增强）

pip install playwright
playwright install chromium

安装后可启用动态网页抓取，获取更丰富的社交媒体内容。

---

与 web-search-ex-skill 的区别

特性	simple-websearch	web-search-ex-skill
依赖	requests	requests + baidusearch + crawl4ai + playwright
安装大小	~100KB	~200MB+
启动速度	毫秒级	秒级（Playwright 冷启动）
搜索深度	基础搜索 + 社交媒体	支持深度搜索/爬虫
社交媒体	✅ 小红书/知乎/微博	❌ 无
适用场景	日常快速搜索 + 社交热点	需要抓取网页详情

---

搜索引擎对比

引擎	优点	缺点	使用场景
Exa AI	免费、快速、质量好	中文内容较少	技术搜索、英文内容
必应	中英文兼顾、结果丰富	需要网络请求	通用搜索
百度	中文内容最全	广告较多	中文内容搜索
小红书	种草笔记、真实体验	需要动态抓取	产品评测、生活分享
知乎	专业问答、深度讨论	需要动态抓取	知识学习、问题分析
微博	实时热点、新闻动态	内容较碎片	热点追踪、舆情监控

---

注意事项

1. 首次运行会自动安装 requests
2. 搜索结果可能因网络环境而异
3. 建议单次查询不超过 10 条结果
4. 社交媒体搜索需要动态网页抓取支持（可选）
5. 国内用户建议优先使用 Exa + 必应组合

---

常见问题

Q: 为什么搜索结果有时为空？

A: 可能是网络问题或搜索引擎限流。建议：

1. 检查网络连接
2. 减少结果数量
3. 更换搜索关键词
4. 使用 action: test 测试引擎可用性

Q: 社交媒体搜索为什么只有链接没有内容？

A: 这是正常行为。社交媒体有反爬机制，为了稳定性：

• 基础模式：返回搜索链接，用户点击查看详情
• 增强模式：安装 Playwright 后可抓取部分内容

Q: 如何提高搜索质量？

A:

1. 使用更具体的关键词
2. 指定搜索平台（如只搜知乎）
3. 结合多个平台结果对比

---

💬 反馈与支持

如果你觉得这个技能有用：

• ⭐ 给个 Star - 在 ClawHub 上收藏支持一下
• 💬 留个评论 - 你的反馈帮助我改进
• 🐛 报告问题 - 遇到问题欢迎提 issue

你的支持是我更新维护的动力！🦆

---

简洁，但不简单。 🦆