patent-search（专利检索）

Patent Search Skill

Two powerful patent search methods:

1. BigQuery Search (Recommended) - 100M+ worldwide patents, zero local storage
2. PatentsView API - Detailed US patent metadata (inventors, assignees, classifications, citations)

FOR CLAUDE: All files and dependencies installed.

• Go directly to Quick Test section
• Script at: .claude/skills/patent-search/bigquery_search.py
• Run from skill directory
• Windows: Use cmd syntax (dir, set, &&)

Quick Test

# Windows
cd ".claude\skills\patent-search"
set GOOGLE_APPLICATION_CREDENTIALS=%APPDATA%\\gcloud\\application_default_credentials.json
python bigquery_search.py search "voice biometric" 5

# Linux/macOS
cd .claude/skills/patent-search
export GOOGLE_APPLICATION_CREDENTIALS="$HOME/.config/gcloud/application_default_credentials.json"
python bigquery_search.py search "voice biometric" 5

Expected: 5 patent results in ~4 seconds

Quick Start

BigQuery Search

# Keyword search (2-3 keywords for best results)
python bigquery_search.py search "voice biometric authentication" 20

# Get specific patent (hyphenated format: US-XXXXXXX-XX)
python bigquery_search.py get US-12424224-B2

# CPC classification search
python bigquery_search.py cpc G10L 15

PatentsView API

Choosing the Right Method

Use BigQuery When:

• Quick keyword search needed
• Worldwide patents (not just US)
• Fast results (3-4 seconds)
• Zero local storage
• CPC classification search
• Budget-conscious (free tier: 1TB/month)

Use PatentsView When:

• Need detailed US patent metadata
• Searching by inventor/assignee
• Citation analysis required
• Complex boolean queries
• Exact field matching
• Patent family analysis

Combined Workflow

1. Start with BigQuery (broad keyword search)
2. Identify relevant patents and CPC codes
3. Switch to PatentsView (detailed metadata/citations)
4. Export final results

Example:

# Step 1: BigQuery broad search
python bigquery_search.py search "voice biometric authentication" 20

# Step 2: Found CPC G10L17, search more
python bigquery_search.py cpc G10L17 50

# Step 3: Use PatentsView for inventor/assignee analysis

Instructions for Claude

When user requests patent searches:

1. Understand Goal: Technology, time period, prior art vs competitive analysis, US vs worldwide
2. Check Dependencies: Verify BigQuery/PatentsView setup
3. Choose Method: Default BigQuery for broad, PatentsView for detailed
4. Optimize Queries:

- BigQuery: 2-3 keywords, simplify if zero results, use CPC codes - PatentsView: Verify API key, construct JSON queries, handle pagination

1. Present Results: Parse JSON, highlight key info, provide Google Patents URLs
2. Offer Next Steps: Suggest refinements, related classifications, citation analysis

Common Use Cases

Prior Art Search

1. BigQuery keyword search
2. Identify CPC codes
3. BigQuery CPC search
4. PatentsView citation analysis
5. Document findings

Competitive Intelligence

1. PatentsView search by assignee
2. Filter by date range
3. Group by CPC
4. Identify key inventors
5. Trend report

Technology Landscape

1. BigQuery CPC search worldwide
2. Analyze by country/date
3. Identify patent families
4. PatentsView US details
5. Summary report

Freedom to Operate

1. BigQuery keyword + CPC search
2. Filter by jurisdiction/active status
3. PatentsView claim analysis
4. Review forward citations
5. Risk assessment

Performance & Coverage

*Free tier: 1TB queries/month

Quick Reference

BigQuery Commands

python bigquery_search.py search "query" <limit>
python bigquery_search.py get <PATENT-NUMBER>
python bigquery_search.py cpc <CODE> <limit>

Common CPC Codes

Troubleshooting

FOR CLAUDE: Only run diagnostics if Quick Test fails.