terry-li-hm

web-search

0
0
# Install this skill:
npx skills add terry-li-hm/skills --skill "web-search"

Install specific skill from multi-skill repository

# Description

Reference for choosing the right web search tool. Not user-invocable β€” use as internal guidance when performing searches.

# SKILL.md


name: web-search
description: Reference for choosing the right web search tool. Not user-invocable β€” use as internal guidance when performing searches.
user_invocable: false


Web Search Tool Selection Guide

Reference for choosing the optimal search tool based on use case. Updated 2026-01-30.

Calling Convention

Claude Code (native MCP):

mcp__perplexity__perplexity_search

OpenClaw (via mcporter):

mcporter call perplexity.perplexity_search query="..."

Config: ~/clawd/config/mcporter.json (imported from Claude Code)

Available Tools

Tool Type Best For
web_search Built-in General purpose, quick searches (Brave API)
Perplexity MCP Deep research, detailed analysis
Serper MCP Google results, regional targeting
Exa MCP Code context, semantic search
Brave MCP Alternative web search, video search
Tavily MCP Content extraction, crawling

Use Case Recommendations

AI News

  1. Quick overview: WebSearch β€” returns synthesized summary with key trends
  2. Deep dive: mcp__perplexity__perplexity_search β€” rich excerpts, analysis
  3. Avoid: Tavily (returns press releases), Brave News (mixed unrelated results)
  1. Overview: WebSearch β€” finds job boards, synthesizes listings
  2. Regional search: mcp__serper__google_search with gl parameter (e.g., gl: "hk")
  3. LinkedIn extraction: mcp__exa__web_search_exa or mcp__tavily__tavily-search

Code & Documentation

  1. Primary: mcp__exa__get_code_context_exa β€” optimized for APIs, libraries, SDKs
  2. Backup: mcp__perplexity__perplexity_search

Deep Research

  1. Primary: mcp__perplexity__perplexity_research β€” comprehensive with citations
  2. Reasoning: mcp__perplexity__perplexity_reason β€” for complex analysis
  1. Default: WebSearch (built-in) β€” fast, good summaries
  2. Google-specific: mcp__serper__google_search
  3. Alternative index: mcp__brave-search__brave_web_search

Tool Details

web_search (Built-in)

  • Best all-around for quick answers
  • Returns synthesized summaries, not just links
  • Good formatting with key points extracted

Perplexity (MCP)

# OpenClaw
mcporter call perplexity.perplexity_search query="..."
mcporter call perplexity.perplexity_ask userContent="..."

# Claude Code
mcp__perplexity__perplexity_search
mcp__perplexity__perplexity_ask
  • Excellent for AI news β€” detailed excerpts with analysis

Serper (MCP)

# OpenClaw
mcporter call serper.search query="..." num=10
mcporter call serper.search query="..." gl=hk hl=zh  # Regional

# Claude Code
mcp__serper__google_search
mcp__serper__scrape
  • Google search via API
  • Supports regional targeting: gl: "hk", gl: "us"
  • Language: hl: "en", hl: "zh"
  • Clean structured JSON output

Exa (MCP)

# OpenClaw
mcporter call exa.web_search_exa query="..." numResults=5 type=deep
mcporter call exa.get_code_context_exa query="React hooks"
mcporter call exa.company_research_exa companyName="Anthropic"

# Claude Code
mcp__exa__web_search_exa
mcp__exa__get_code_context_exa
  • Semantic search, good for finding similar content
  • get_code_context_exa is excellent for programming queries
  • Returns full text snippets

Brave (MCP)

# OpenClaw
mcporter call brave-search.brave_web_search query="..."
mcporter call brave-search.brave_news_search query="..."
mcporter call brave-search.brave_video_search query="..."

# Claude Code
mcp__brave-search__brave_web_search
mcp__brave-search__brave_news_search
mcp__brave-search__brave_video_search
  • Alternative to Google index
  • Video search is useful
  • News search can return unrelated results β€” use with caution

Tavily (MCP)

# OpenClaw
mcporter call tavily.tavily-search query="..."
mcporter call tavily.tavily-extract urls='["https://..."]'

# Claude Code
mcp__tavily__tavily-search
mcp__tavily__tavily-extract
mcp__tavily__tavily-crawl
  • tavily-extract good for scraping specific URLs
  • tavily-crawl for site mapping
  • News search is weak (returns press releases)

Quality Rankings (2026-01-21)

For AI News

  1. WebSearch ⭐⭐⭐⭐⭐
  2. Perplexity ⭐⭐⭐⭐⭐
  3. Serper ⭐⭐⭐⭐
  4. Exa ⭐⭐⭐⭐
  5. Brave News ⭐⭐⭐
  6. Tavily ⭐⭐
  1. WebSearch ⭐⭐⭐⭐⭐
  2. Serper ⭐⭐⭐⭐
  3. Exa ⭐⭐⭐⭐
  4. Tavily ⭐⭐⭐⭐
  5. Perplexity ⭐⭐⭐
  6. Brave ⭐⭐⭐

WeChat Articles (mp.weixin.qq.com)

See the dedicated wechat-article skill for full workflow.

Quick reference:
- Primary: wechat.imagenie.us API (only short URLs: /s/ABC123)
- Fallback: Mirror sites (163.com, zhihu.com, csdn.net)
- Direct access fails: WebFetch, Serper, Jina, Claude in Chrome all get CAPTCHA blocked

Removed Tools

  • Google Search (direct API) β€” Removed 2026-01-21. "Custom Search JSON API not enabled" error. Serper provides same Google results.

Notes

  • For Terry's use cases (AI news, job search), core tools are: WebSearch + Perplexity + Serper
  • Exa worth keeping for code context
  • Brave and Tavily are somewhat redundant but kept for specialized features (video search, crawling)

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.