Multi Search

This is a smart search wrapper that tries DuckDuckGo first, then falls back to Tavily or Bing if needed based on your network and API quota. It auto-detects when you're behind a firewall and switches engines accordingly, caching network checks for five minutes. The quality mode prioritizes Tavily for better results at the cost of burning through your monthly API limit faster. Includes helpers to fetch full page content from search results, which pairs well with a summarization step. Main win here is not babysitting which search engine works in your current environment, especially if you toggle VPNs or work across different networks regularly.

skills.sh installs: 1.1K
GitHub stars: 11
last updated: Jul 14, 2026

View on GitHub

Install

$ npx -y skills add nex-zmh/agent-websearch-skill --skill multi-search --agent claude-code

Installs into .claude/skills of the current project.

Files

Multi-Search Skill - 智能多引擎搜索

本技能整合多个搜索引擎，自动检测网络环境，智能选择最佳可用引擎。

引擎优先级

质量优先模式 (prefer_quality=True)

Tavily API (1000次/月) - 质量最高，需 API Key
DuckDuckGo (无限免费) - 无需 API Key
Bing Web Search API (1000次/月) - 需 API Key
Bing 爬虫 (无限免费) - 最终回退

平衡模式 (prefer_quality=False, 默认)

DuckDuckGo (无限免费) - 优先免费引擎
Tavily API (1000次/月) - 如果配置了 API Key
Bing Web Search API (1000次/月)
Bing 爬虫 (无限免费)

核心能力

智能网络检测与引擎切换
自动配额管理（Tavily/Bing API）
支持网页内容抓取
5分钟网络检测缓存

使用方式

基本搜索

from multi_search import search

# 平衡模式 - 优先免费引擎
results = search("Python tutorial", max_results=5)

# 质量优先模式 - 优先使用 Tavily
results = search("AI research", max_results=5, prefer_quality=True)

# 强制重新检测网络（切换 VPN 后使用）
results = search("OpenClaw skills", max_results=5, force_network_check=True)

搜索技能（自动质量优先）

from multi_search import search_skills

results = search_skills("OpenClaw AI agent automation", max_results=10)

查看系统状态

from multi_search import get_status

status = get_status()  # 使用缓存
status = get_status(force_network_check=True)  # 强制重新检测

SKILL.md on GitHub

Details

Repository: github.com/nex-zmh/agent-websearch-skill
Category: AI & Agent Building
First seen: Apr 16, 2026