test

👤 COZY 📦 v1.0.0 ⭐ 4.1 ⬇️ 201 下載
📊 資料分析 免費

📖 技能介紹

🌐 Smart Web Search v3.1

Primary Engines: Baidu (China) + Google (International)

Core Features:

  • ✅ Real-time Search - Today/24h/7d/30d time filtering
  • ✅ Time Filtering - Specify search time range
  • ✅ Multi-engine Aggregation - Search 3-5 engines simultaneously
  • ✅ Smart De-duplication - Auto-filter duplicate content
  • ✅ AI Summaries - Auto-generated key insights
  • ✅ Ad Filtering - Auto-detect and filter ads/promotions
  • ✅ Content De-toxication - Filter spam, fake news, misinformation
  • ✅ Baidu Integration - China's largest search engine (Primary)
  • ✅ Google Integration - World's most accurate engine (Primary)

Auto-selects optimal search engine based on query content.


🧠 Intelligent Detection Logic

Use Baidu Search (China) When:

  • 🇨🇳 Query contains Chinese characters
  • 🇨🇳 Searching China-related topics (WeChat, Taobao, Baidu, etc.)
  • 🇨🇳 Searching Chinese companies/people/events
  • 🇨🇳 Searching Chinese policies/news/regulations

Use Google Search (International) When:

  • 🌍 Query is pure English
  • 🌍 Searching international topics (GitHub, Stack Overflow, Reddit, etc.)
  • 🌍 Searching foreign companies/people/events
  • 🌍 Searching technical docs/academic papers/international news

🔧 Tool Calls

Domestic Search (Chinese Queries)

Primary Engine (Baidu):

web_fetch(url="https://www.baidu.com/s?wd=query", extractMode="text", maxChars=15000)

Advanced Search:

# Exact match
web_fetch(url="https://www.baidu.com/s?wd=\"exact+keyword\"", extractMode="text", maxChars=15000)

# Exclude site
web_fetch(url="https://www.baidu.com/s?wd=keyword+-site:zhihu.com", extractMode="text", maxChars=15000)

# File type
web_fetch(url="https://www.baidu.com/s?wd=keyword+filetype:pdf", extractMode="text", maxChars=15000)

Real-time News:

# Latest 24 hours
web_fetch(url="https://www.baidu.com/s?wd=query+2026-03-17&src=news", extractMode="text", maxChars=15000)

# Latest 7 days
web_fetch(url="https://www.baidu.com/s?wd=query+this+week&src=news", extractMode="text", maxChars=15000)

Backup Engines (by priority):

  1. 360 Search: https://m.so.com/s?q=query
  2. Sogou WeChat: https://weixin.sogou.com/weixin?type=2&query=query
  3. Bing CN: https://cn.bing.com/search?q=query
  4. Sogou Web: https://www.sogou.com/web?query=query
  5. Toutiao: https://www.toutiao.com/search/?keyword=query

Baidu Advantages:

  • ✅ Largest Chinese index (trillions of pages)
  • ✅ Best localization for China
  • ✅ Fastest update speed (minute-level)
  • ✅ Advanced search syntax support
  • ✅ Integrated with Baidu Baike, Zhidao, Tieba

International Search (English Queries)

Primary Engine (Google via Startpage):

web_fetch(url="https://www.startpage.com/do/search?q=query", extractMode="text", maxChars=15000)

Direct Google (if accessible):

web_fetch(url="https://www.google.com/search?q=query", extractMode="text", maxChars=15000)

Real-time News:

# Latest 24 hours
web_fetch(url="https://www.startpage.com/do/search?q=query+2026-03-17", extractMode="text", maxChars=15000)

# Latest 7 days
web_fetch(url="https://www.startpage.com/do/search?q=query+this+week", extractMode="text", maxChars=15000)

Backup Engines (by priority):

  1. DuckDuckGo: https://lite.duckduckgo.com/lite/?q=query
  2. DuckDuckGo News: https://duckduckgo.com/?q=query&ia=news
  3. Qwant: https://www.qwant.com/?q=query&t=web
  4. Bing EN: https://www.bing.com/search?q=query
  5. Brave Search: https://search.brave.com/search?q=query

Google Advantages:

  • ✅ World's largest index (hundreds of billions of pages)
  • ✅ Most advanced search algorithm
  • ✅ Richest academic resources
  • ✅ Best multi-language support
  • ✅ Powerful advanced search features

📋 Usage Examples

Chinese Queries

Basic Search:

User: Search NVIDIA latest financial report
→ Auto-use: Baidu Search
→ Time range: Auto-detect "latest" → 7 days

Real-time News:

User: Today March 17 latest news
→ Auto-use: Baidu News Search
→ Time range: 2026-03-17 (today)
→ Extra parameter: &src=news

WeChat Articles:

User: Find WeChat articles about AI
→ Auto-use: Sogou WeChat Search
→ Search type: WeChat official accounts

Specified Time:

User: Search AI news last 24 hours
→ Auto-use: Baidu Search + time filter
→ Time range: 2026-03-16 to 2026-03-17

English Queries

Basic Search:

User: search for latest AI news
→ Auto-use: Google (via Startpage)
→ Time range: Auto-detect "latest" → 7 days

Real-time News:

User: today news March 17 2026
→ Auto-use: Google News
→ Time range: 2026-03-17 (today)

Technical Content:

User: find Python tutorials on GitHub
→ Auto-use: Google
→ Priority sources: GitHub, Stack Overflow

Specified Time:

User: latest tech news this week
→ Auto-use: Google
→ Time parameter: this week

Mixed Queries

International Company:

7w4.net小蔥技能站,你的AI助手技能庫。

User: Tesla TSLA stock price today
→ Priority: Google (international company)
→ Backup: Baidu (Chinese results)

China Product:

User: WeChat how to export data
→ Priority: Baidu (domestic product)
→ Backup: Google (English tutorials)

Real-time Search (v3.1 Recommended)

Today's News:

"Search today March 17 AI news"
"NVIDIA latest news (24 hours)"
"latest AI news today March 17"
"2026-03-17 AI latest news"

This Week:

"Search this week AI news"
"AI latest news this week"
"Last 7 days AI updates"

🎯 Search Strategy v3.1

Step 1: Language Detection

  • Contains Chinese characters → Baidu (China)
  • Pure English/Latin → Google (International)
  • Mixed language → Judge by topic

Step 2: Topic Detection

  • China companies/products/policies → Baidu
  • International companies/tech/academic → Google
  • News/current events → Dual-engine parallel

Step 3: Time Detection

Auto-recognize time keywords:

  • "Today", "今日", "today" → 2026-03-17
  • "Yesterday", "昨天" → 2026-03-16
  • "Last 24h", "最近 24 小時" → Past 24 hours
  • "This week", "本週" → Current week
  • "Latest", "最新" → 7 days
  • "This month", "本月" → 30 days

No time keyword → Default 30 days

Step 4: Safety Detection

1. Detect if query involves sensitive topics
   - Medical/health → Enable strict de-tox
   - Finance/investment → Enable scam filter
   - News → Enable fake news filter

2. Set filter level
   - Normal search → Standard filter
   - "Verified", "safe" → Strict filter
   - "Academic", "research" → Authoritative source priority

Step 5: Engine Selection

Domestic Priority:

  1. Baidu (Largest Chinese engine) - PRIMARY
  2. 360 Search (General)
  3. 360 News (Real-time news)
  4. Sogou WeChat (Official accounts)
  5. Bing CN (Backup)
  6. Toutiao (Breaking news)

International Priority:

  1. Google via Startpage (Most accurate) - PRIMARY
  2. DuckDuckGo (Privacy)
  3. DuckDuckGo News (News)
  4. Qwant (European)
  5. Bing EN (Backup)
  6. Brave Search (Privacy)

Step 6: Multi-engine Aggregation

Search 3 engines simultaneously:

Primary: Baidu/Google
Backup 1: 360/Qwant
Backup 2: Bing

Merge Strategy:

  • Collect top 10 results from each
  • De-duplicate (URL + title similarity)
  • Filter ads and toxic content
  • Sort by time descending
  • Keep 5-8 best results

Step 7: Search + Fetch

  1. Search 3 engines simultaneously
  2. Merge and de-duplicate results
  3. Filter ads (auto-detect)
  4. De-toxicate (verify sources)
  5. Select 3-5 most relevant URLs
  6. Fetch detailed content with web_fetch
  7. Generate AI summary

⚠️ Best Practices

For Latest Information

✅ "Search today March 17 XXX news"
✅ "XXX latest news (24 hours)"
✅ "latest XXX news today"
❌ "Search XXX" (no time range, may return old news)

For High Accuracy

✅ "March 17 2026 XXX latest news"
✅ "XXX + 2026-03-17"
✅ "XXX breaking news"
❌ "XXX news" (unclear time range)

For Safe Search

✅ "Search vaccine info (verified)"
✅ "XXX safe search"
✅ "verified XXX info"
❌ Direct search for medical/finance (may encounter misinformation)

For Baidu Search

✅ "Search with Baidu XXX"
✅ "XXX site:baidu.com"
✅ "XXX filetype:pdf"

For Google Search

✅ "Search with Google XXX"
✅ "XXX site:github.com"
✅ "XXX filetype:pdf"

🛡️ Ad Filtering

Auto-detect Ads

Ad Characteristics:

Ad Labels = [
    "廣告", "推廣", "Sponsored", "Ad", "Promoted",
    "競價排名", "品牌展示", "商業推廣"
]

Ad Positions = [
    "Top of results", "Right sidebar", "Bottom recommendations",
    "Related searches", "Guess you want"
]

Filtering Rules

1. Label Filtering:

✅ Keep: Organic search results
❌ Filter: Results marked as "Ad", "Sponsored", "廣告", "推廣"

2. Position Filtering:

✅ Keep: Natural results (middle positions)
❌ Filter: Top 3, right sidebar, bottom recommendations

3. Domain Filtering:

❌ Filter domains blacklist:
- Known ad farms
- Low-quality content sites
- Content scrapers

✅ Priority domains whitelist:
- Government sites (.gov.cn)
- Educational institutions (.edu)
- Mainstream media
- Known tech companies

Usage Examples

Enable Ad Filtering:

"Search AI news (filter ads)"
"search AI news (no ads)"
"XXX clean search"

Filter Results:

Raw results: 20 items
Ad count: 6 items (30%)
After filter: 14 items (100% organic)

🧪 Content De-toxication

Toxin Detection

Content Toxin Types:

Type Characteristics Action
🗑️ Spam Duplicate, keyword stuffing Filter
📢 Clickbait "Shocking", "Must-see", "#1" Down-rank
📰 Fake News No source, no author, no date Filter
💊 Medical Rumors Folk remedies, miracle cures, 100% cure rate Filter + Warning
💰 Financial Scams High returns, guaranteed profit, insider info Filter + Warning
🔞 Inappropriate Adult, violence, gambling Filter
🎭 Deepfakes AI-generated fake content Mark + Warning

De-tox Algorithm

1. Source Verification:

Trusted Sources = [
    "Government sites", "Mainstream media", "Academic journals",
    "Known company websites", "Authoritative organizations"
]

if Source not in Trusted Sources:
    Credibility Score -= 30%

2. Content Quality Assessment:

Quality Metrics = [
    "Has author", "Has date", "Has source citations",
    "Clear logic", "Verifiable data"
]

if Quality Score < 60:
    Mark as "Low Quality"

3. Fact-checking:

# Cross-verify multiple sources
if Only single source reports:
    Mark as "Unverified"

if Multiple authoritative sources confirm:
    Credibility += 50%

4. Sentiment Analysis:

if Contains extreme emotion words > 10:
    Mark as "Sensational content"
    Credibility -= 40%

Warning System

Warning Levels:

Level Icon Trigger Action
🟢 Trusted ✅ Authoritative + verified Normal display
🟡 Questionable ⚠️ Single source/medium quality Mark display
🟠 Suspicious ❗ Low quality/clickbait Down-rank + Warning
🔴 Toxic 🚫 Fake/scam Filter + Warning

Usage Examples

Enable De-toxication:

"Search vaccine info (verified)"
"search vaccine info (de-toxic)"
"XXX safe search"
"XXX filter misinformation"

De-tox Results:

Raw results: 20 items
Toxic content: 5 items (25%)
  - Fake info: 2 items
  - Clickbait: 2 items
  - Low quality: 1 item
After filter: 15 items (100% trusted)

📊 Engine Comparison

Engine Region Free API Key Accuracy Features
Baidu 🇨🇳 China ✅ ❌ ⭐⭐⭐⭐⭐ Largest Chinese index
Google 🌍 Global ✅ ❌ ⭐⭐⭐⭐⭐ Most accurate globally
360 Search 🇨🇳 China ✅ ❌ ⭐⭐⭐⭐ Safe & clean
DuckDuckGo 🌍 Global ✅ ❌ ⭐⭐⭐⭐ Privacy-focused
Startpage 🌍 Global ✅ ❌ ⭐⭐⭐⭐⭐ Google results + privacy
Qwant 🌍 Europe ✅ ❌ ⭐⭐⭐⭐ European engine
Bing CN 🇨🇳 China ✅ ❌ ⭐⭐⭐⭐ Stable & reliable
Bing EN 🌍 Global ✅ ❌ ⭐⭐⭐⭐ Stable backup

🚀 Quick Start

Basic Search

"Search XXX"
"search for XXX"
"Help me find XXX"
"find XXX"

Real-time Search (v3.1 Recommended)

Get Today's Latest:

"Search today March 17 XXX news"
"XXX latest news (24 hours)"
"latest XXX news today March 17"
"2026-03-17 XXX latest news"

Get This Week:

"Search this week XXX news"
"XXX latest news this week"
"Last 7 days XXX updates"

Safe Search (v3.1 New)

De-toxic Search:

"Search vaccine info (verified)"
"XXX safe search"
"verified XXX info"
"XXX filter misinformation"

Filter Ads:

"Search XXX (filter ads)"
"XXX no ads"
"XXX clean search"

Specify Engine

Use Baidu:

"Search with Baidu XXX"
"XXX site:baidu.com"

Use Google:

"Search with Google XXX"
"XXX site:github.com"

Advanced Search

Specify Time Range:

"Search March 16-17 2026 XXX"
"XXX news from 2026-03-16 to 2026-03-17"

Specify Source:

"Search WeChat articles about XXX"
"find XXX on GitHub"
"search XXX on Reddit"

🤖 AI 評測

這個Skill的功能概念聽起來不錯,能自動切換國內外搜尋引擎,但實際質量存疑。文件中描述的多種引擎和智慧功能很全面,但不同文件的說法互相矛盾,而且整個包缺少實際可執行的程式碼,你拿到手可能根本用不了。內容重複、版本混亂的問題也比較明顯,誠意不足。

📊 多維度評分

適應性4.3
規範性3.9
有效性4
可靠性3.9
可信度4.8

📁 包含檔案 (13 個)

📄 COMPLETE_SUMMARY.md 6.6 KB
📄 FEEDBACK_SYSTEM.md 9.7 KB
📄 FEEDBACK_TEMPLATES.md 9.3 KB
📄 FINAL_RELEASE_V3.md 7.1 KB
📄 GITHUB_PUSH_STATUS.md 3.8 KB
📄 OPTIMIZATION_V2.md 7.1 KB
📄 PUBLISH_REPORT.md 5.2 KB
📄 PUBLISH_REPORT_V3.md 8.1 KB
📄 README.md 4.5 KB
📄 SKILL.md 13.9 KB
📄 TEST_CASES.md 6.5 KB
📄 TEST_REPORT.md 6.6 KB
📄 _meta.json 286 B