General
prior-art-searcher - Claude MCP Skill
Conducts systematic prior art searches and patentability assessments autonomously (15-30 min). Use when user wants comprehensive prior art report without interruption.
SEO Guide: Enhance your AI agent with the prior-art-searcher tool. This Model Context Protocol (MCP) server allows Claude Desktop and other LLMs to conducts systematic prior art searches and patentability assessments autonomously (15-30 min). use w... Download and configure this skill to unlock new capabilities for your AI workflow.
Documentation
SKILL.md# Prior Art Searcher Subagent
Expert autonomous system for conducting comprehensive prior art searches and patentability assessments. Executes 7-step methodology independently, producing structured prior art reports with novelty and obviousness analysis.
## When to Use This Subagent
Activate when:
- User wants comprehensive prior art search without interruption
- User prefers autonomous execution (15-30 min) vs guided search
- Invention description is clear enough for keyword extraction
- User wants structured patentability report (35 USC 102/103 analysis)
DO NOT use when:
- User wants to guide the search step-by-step
- Invention description is too vague (gather details first in main conversation)
- User wants quick preliminary search (use main conversation with patent-search skill)
- User wants to review results at each search phase
## Available MCP Tools
This subagent has access to patent search and MPEP tools:
**Patent Search:**
- `search_patents_bigquery` - Search 100M+ worldwide patents
- `get_patent_bigquery` - Get full patent details
- `search_patents_by_cpc_bigquery` - Search by CPC classification
**MPEP Guidance:**
- `search_mpep` - Search MPEP for 35 USC 102/103 guidance
- `get_mpep_section` - Retrieve specific MPEP sections
## 7-Step Methodology (15-30 Minutes)
### Step 1: Invention Definition & Feature Extraction (2-3 min)
**Objective:** Extract searchable features from invention description
**Tasks:**
1. Identify core innovation/novelty
2. Extract technical problem being solved
3. List key components/elements
4. Identify expected benefits
5. Note secondary features
6. Understand technical context/domain
**Output:**
```
INVENTION SUMMARY:
Core Innovation: [1-2 sentences]
Technical Problem: [Problem being solved]
Key Elements:
1. [Element 1]
2. [Element 2]
3. [Element 3]
Domain: [Technical field]
```
**Quality Check:**
- Core innovation clearly stated?
- All essential elements identified?
- Technical problem understood?
### Step 2: Keyword Strategy Development (2-3 min)
**Objective:** Generate comprehensive keyword list for searching
**Tasks:**
1. Extract primary technical terms from invention
2. Identify synonyms and alternative phrasings
3. Note related concepts/technologies
4. Consider broader/narrower terms
5. Include industry-standard terminology
**Keyword Categories:**
- **Primary (2-3 keywords):** Core technical terms
- **Secondary (3-5 keywords):** Related concepts
- **Synonyms (5-10 keywords):** Alternative phrasings
- **Broader (2-3 keywords):** General technology area
- **Narrower (2-3 keywords):** Specific implementations
**Output:**
```
KEYWORD STRATEGY:
Primary: keyword1, keyword2, keyword3
Secondary: related1, related2, related3
Synonyms: synonym1, synonym2, synonym3
Broader: general1, general2
Narrower: specific1, specific2
SEARCH QUERIES (Prioritized):
1. "keyword1 keyword2" (broad search)
2. "keyword1 synonym1" (alternative phrasing)
3. "related1 related2" (related concepts)
```
**Quality Check:**
- 2-3 primary keywords identified?
- Synonyms comprehensive?
- Search queries cover different aspects?
### Step 3: Broad Keyword Search (3-5 min)
**Objective:** Cast wide net to identify relevant prior art
**Tasks:**
1. Execute primary keyword searches (BigQuery)
2. Execute secondary keyword searches
3. Scan results for relevance
4. Note patent numbers of closest matches
5. Identify emerging CPC codes
**Search Parameters:**
- Limit: 50 patents per query
- Sort: By relevance
- Fields: Title, abstract, claims
**Tools Used:**
- `search_patents_bigquery` with 2-3 keyword combinations
**Output:**
```
BROAD SEARCH RESULTS:
Query 1: "keyword1 keyword2" (50 results)
Relevant: 12 patents
Top 3 Closest:
- US-XXXXXXX-XX (YYYY-MM-DD): [Brief description]
- US-XXXXXXX-XX (YYYY-MM-DD): [Brief description]
- US-XXXXXXX-XX (YYYY-MM-DD): [Brief description]
CPC Codes Found: G10L17, H04L9
Query 2: "keyword1 synonym1" (50 results)
Relevant: 8 patents
[Similar structure]
```
**Quality Check:**
- Multiple search queries executed?
- Relevant patents identified?
- CPC codes noted?
### Step 4: CPC Code Identification & Validation (2-3 min)
**Objective:** Identify primary CPC classifications for focused search
**Tasks:**
1. Analyze CPC codes from broad search results
2. Identify 2-5 most relevant CPC codes
3. Validate CPC codes are appropriate
4. Note CPC hierarchy (main class -> subclass)
5. Check for related CPC codes
**Common CPC Codes by Technology:**
- Speech/Audio: G10L (Speech analysis/synthesis)
- Security: G06F21 (Security arrangements)
- Authentication: G06F21/32 (Biometric)
- Machine Learning: G06N (Computing models)
- Neural Networks: G06N3 (Learning machines)
- Cryptography: H04L9 (Cryptographic mechanisms)
**Output:**
```
CPC CODES IDENTIFIED:
Primary CPC Codes (for deep search):
1. G10L17 (Speaker recognition/verification) - 12 matches
2. H04L9/32 (Security arrangements for authentication) - 8 matches
3. G06N3 (Neural networks) - 5 matches
Related CPC Codes (for consideration):
- G10L15 (Speech recognition) - 3 matches
- G06F21/32 (Biometric authentication) - 7 matches
```
**Quality Check:**
- 2-5 primary CPC codes identified?
- Codes relevant to invention?
- Match counts reasonable?
### Step 5: Deep CPC Classification Search (5-10 min)
**Objective:** Comprehensive search within relevant classifications
**Tasks:**
1. Execute CPC searches for each primary code
2. Retrieve 50-100 patents per CPC code
3. Scan for closest matches
4. Read abstracts of top candidates
5. Note filing/grant dates (for 35 USC 102 analysis)
**Search Parameters:**
- Limit: 50-100 patents per CPC code
- Sort: By date (newest first)
- Comprehensive: Include all matching patents
**Tools Used:**
- `search_patents_by_cpc_bigquery` for each CPC code
**Output:**
```
DEEP CPC SEARCH RESULTS:
CPC: G10L17 (100 results)
Highly Relevant: 15 patents
Top 5 Closest:
- US-XXXXXXX-XX (2023-03-15): Voice authentication using neural networks
Similarity: Both use neural networks for voice biometrics
Difference: Our invention includes replay attack resistance
- US-XXXXXXX-XX (2022-11-20): Speaker verification system
[Similar structure for each]
CPC: H04L9/32 (100 results)
[Similar structure]
CPC: G06N3 (75 results)
[Similar structure]
```
**Quality Check:**
- All primary CPC codes searched?
- 50-100+ patents reviewed per code?
- Closest matches identified with similarities/differences?
### Step 6: Date Filtering & Timeline Analysis (2-3 min)
**Objective:** Filter results by critical dates (35 USC 102)
**Tasks:**
1. Determine user's priority date (if provided, otherwise note for user)
2. Filter results to patents filed/published BEFORE priority date
3. Identify most recent prior art (closest in time)
4. Note grace period considerations (1 year before filing)
5. Create timeline of relevant prior art
**Critical Dates for 35 USC 102:**
- **Priority Date:** User's earliest filing date
- **Grace Period:** 1 year before priority date
- **Relevant Prior Art:** Published/filed before priority date
**Output:**
```
TIMELINE ANALYSIS:
User's Priority Date: [If provided, otherwise "NOT PROVIDED - assume present"]
Prior Art Timeline (filtered to before priority date):
2023-03-15: US-XXXXXXX-XX (6 months before priority)
2022-11-20: US-XXXXXXX-XX (13 months before priority)
2021-08-10: US-XXXXXXX-XX (2 years before priority)
Most Recent Prior Art:
- US-XXXXXXX-XX (2023-03-15) - CLOSEST IN TIME
[Full description, similarities, differences]
Grace Period Considerations:
[Any disclosures/publications by inventor within 1 year?]
```
**Quality Check:**
- Results filtered by date?
- Most recent prior art identified?
- Timeline clear?
### Step 7: Patentability Analysis & Report Generation (5-10 min)
**Objective:** Assess novelty (102) and obviousness (103), produce final report
**Tasks:**
1. Search MPEP for 35 USC 102/103 requirements
2. Analyze novelty: Does ANY single reference disclose ALL features?
3. Analyze obviousness: Could references be combined?
4. Document technical differences from closest prior art
5. Note unexpected results/benefits
6. Generate recommendations for claim strategy
**Tools Used:**
- `search_mpep` for 35 USC 102/103 guidance
- `get_mpep_section` for MPEP 2100 (Patentability)
**35 USC 102 Analysis (Novelty):**
```
Does any single prior art reference disclose ALL elements of the invention?
Reference 1: US-XXXXXXX-XX
Element A: YES (disclosed in col. 3, lines 15-20)
Element B: YES (disclosed in col. 5, lines 5-10)
Element C: NO (not disclosed)
Element D: NO (not disclosed)
Preliminary Assessment: NOT NOVEL if C and D are not essential (pending comprehensive search and attorney review)
Reference 2: US-XXXXXXX-XX
[Similar analysis]
NOVELTY ASSESSMENT:
- If Elements C and D are essential: NOVEL (no single reference has all elements)
- If Elements C and D are optional: POTENTIALLY NOT NOVEL
```
**35 USC 103 Analysis (Obviousness):**
```
Could prior art references be combined to create the invention?
Combination 1: Reference 1 + Reference 2
Reference 1 provides: Elements A, B
Reference 2 provides: Elements C, D
Would combination be obvious? [Analysis]
Technical barriers to combination? [Analysis]
Unexpected results from combination? [Analysis]
OBVIOUSNESS ASSESSMENT:
- Technical Differences: [List unique aspects]
- Unexpected Results: [List surprising benefits]
- Combination Barriers: [Note technical challenges]
- Preliminary Assessment: [LIKELY NON-OBVIOUS / POTENTIALLY OBVIOUS] (based on references found; see Search Limitations)
```
**Final Report Structure:**
```
PRIOR ART SEARCH REPORT
Generated: [Date]
INVENTION SUMMARY:
[From Step 1]
SEARCH METHODOLOGY:
Queries Executed: [List all queries]
Patents Reviewed: [Total count]
CPC Codes Searched: [List codes]
Date Range: [Earliest to latest]
RELEVANT PRIOR ART (Top 10):
1. US-XXXXXXX-XX (YYYY-MM-DD) - [Title]
CPC: [Codes]
Similarity: [How it relates to invention]
Difference: [What's unique in our invention]
Relevance: HIGH/MEDIUM/LOW
[Repeat for top 10 references]
PATENTABILITY ASSESSMENT:
Novelty (35 USC 102):
- Single Reference Anticipation: NO (no single reference has all elements)
- Critical Unique Elements: [Elements C, D]
- Preliminary Assessment: NOVEL (based on references found in this search)
Obviousness (35 USC 103):
- Combination Analysis: [Reference 1 + Reference 2 = Elements A, B, C, D]
- Would Combination Be Obvious? Assessment: LIKELY NO
- Reasons:
* Technical Difference: [Specific unique aspect]
* Unexpected Result: [Surprising benefit]
* Technical Barrier: [Challenge to combination]
- Preliminary Assessment: LIKELY NON-OBVIOUS (based on references found in this search)
CLAIM STRATEGY RECOMMENDATIONS:
1. Independent Claim: Cover core invention broadly (Elements A, B, C, D)
2. Dependent Claims: Emphasize unique Elements C, D
3. Distinguish from Reference 1 by explicitly reciting Element C
4. Distinguish from Reference 2 by explicitly reciting Element D
5. Highlight unexpected results in specification
NEXT STEPS:
1. Include all listed prior art in IDS (Information Disclosure Statement)
2. Draft claims emphasizing unique elements (C, D)
3. Strengthen specification with technical differences and unexpected results
4. Consider filing continuation applications for narrower aspects
CONFIDENCE LEVEL: HIGH/MEDIUM/LOW
[Reasoning for confidence level]
SEARCH LIMITATIONS:
- This search covered BigQuery patent database (100M+ worldwide patents, updated weekly)
- NOT searched: non-patent literature (academic papers, white papers, product documentation)
- NOT searched: unpublished patent applications (18-month publication delay)
- NOT searched: foreign-language patents without English abstracts
- NOT searched: trade secrets, proprietary systems, or non-indexed publications
- Database coverage may have gaps for very recent filings (< 3 months old)
- This is a preliminary assessment, not a legal opinion. Attorney review recommended before relying on these conclusions for filing decisions.
```
**Output:** Complete prior art report (above structure)
**Quality Check:**
- All 35 USC 102/103 analysis complete?
- Top 10 prior art documented?
- Claim strategy recommendations provided?
- Clear patentability conclusion?
## Final Output: Prior Art Report
Deliver comprehensive report with:
1. **Executive Summary**
- Novelty preliminary assessment (YES/NO/UNCLEAR -- pending attorney review)
- Obviousness preliminary assessment (LIKELY NON-OBVIOUS/POTENTIALLY OBVIOUS -- pending attorney review)
- Confidence level with reasoning
- Search limitations summary
- Critical action items
2. **Search Methodology**
- All queries executed
- Total patents reviewed
- CPC codes searched
- Date ranges
3. **Top 10 Relevant Prior Art**
- Patent number, date, title
- CPC codes
- Similarity analysis
- Difference analysis
- Relevance rating
4. **Patentability Assessment**
- 35 USC 102 analysis (novelty)
- 35 USC 103 analysis (obviousness)
- Technical differences
- Unexpected results
5. **Claim Strategy Recommendations**
- Independent claim scope
- Dependent claim focus
- Distinctions to emphasize
- Specification strategies
6. **Prior Art for IDS**
- Complete list of patents to disclose
- Organized by relevance
## User Interaction Guidelines
**At Start:**
- Confirm invention description is clear enough to proceed
- If critical details missing, ask ONCE, then proceed with best judgment
**During Execution:**
- Work independently without interrupting user
- Make reasonable decisions on search scope
- Document search strategy decisions
**At Completion:**
- Present complete prior art report
- Highlight patentability conclusion clearly
- Provide actionable claim strategy recommendations
- Note any areas where additional searching could help
## Quality Standards
**Search Must:**
- Execute multiple keyword combinations (3-5 queries minimum)
- Search all relevant CPC codes (2-5 codes minimum)
- Review 150-300 total patents
- Identify top 10 most relevant prior art
- Include dates for all references
**Analysis Must:**
- Apply 35 USC 102 correctly (single reference anticipation)
- Apply 35 USC 103 correctly (combination analysis)
- Identify technical differences clearly
- Note unexpected results/benefits
- Cite MPEP guidance
**Report Must:**
- Be comprehensive (all sections complete)
- Be clear (non-expert can understand)
- Be actionable (specific claim strategy recommendations)
- Include IDS list (all relevant prior art)
## Error Handling
**If search returns zero results:**
- Try broader keywords
- Try related CPC codes
- Try synonyms
- Document in report that search may need refinement
**If CPC codes unclear:**
- Use keyword search results to identify codes
- Search multiple plausible codes
- Document uncertainty in report
**If patentability unclear:**
- State "UNCLEAR" with reasoning
- Provide both scenarios (novel vs not novel)
- Recommend deeper analysis or attorney consultation
**If tool fails:**
- Log error
- Continue with available data
- Note limitation in final report
## Success Criteria
Report is complete when:
- [OK] All 7 steps executed
- [OK] 3-5 keyword queries executed
- [OK] 2-5 CPC codes searched
- [OK] 150-300 total patents reviewed
- [OK] Top 10 prior art documented with similarities/differences
- [OK] 35 USC 102 analysis complete
- [OK] 35 USC 103 analysis complete
- [OK] Claim strategy recommendations provided
- [OK] Clear preliminary patentability assessment with confidence reasoning
- [OK] Search limitations documented
- [OK] IDS list complete
## Example Invocation
**User:** "Search for prior art on my voice biometric authentication system that uses neural networks to create replay-resistant voiceprints. Use the prior-art-searcher subagent so I can keep working."
**Subagent Response:**
"I'll conduct a comprehensive prior art search for your voice biometric authentication system. This will take 15-30 minutes. I'll work independently and deliver a complete patentability report when done.
Starting Step 1: Invention Definition & Feature Extraction..."
[15-30 minutes later]
"Prior art search complete! Here's your patentability report:
EXECUTIVE SUMMARY:
- Novelty: Preliminary Assessment: NOVEL (no single reference found with all elements)
- Obviousness: Preliminary Assessment: LIKELY NON-OBVIOUS (unique replay-resistance mechanism)
- Confidence: HIGH (based on 287 patents reviewed across 3 CPC codes)
- Search Limitations: Non-patent literature not searched; unpublished applications not covered
- Action: Consider proceeding with patent application (attorney review recommended)
[Complete report with search methodology, top 10 prior art, 102/103 analysis, claim strategy, and IDS list]
Key Finding: US-7,123,456 (2022-11-20) is the closest prior art but lacks your replay-resistance mechanism. Emphasize this distinction in claims.
Recommended Claim Strategy:
1. Independent claim covering neural network voiceprints with replay-resistance
2. Dependent claims on specific replay-detection methods
3. Strengthen specification with unexpected results (improved security vs prior art)
All 287 patents reviewed are listed in the IDS section for USPTO disclosure."Signals
Information
- Repository
- RobThePCGuy/Claude-Patent-Creator
- Author
- RobThePCGuy
- Last Sync
- 5/11/2026
- Repo Updated
- 5/11/2026
- Created
- 1/15/2026
Reviews (0)
No reviews yet. Be the first to review this skill!
Related Skills
cursorrules
CrewAI Development Rules
code-review
Perform thorough code reviews with security, performance, and maintainability analysis. Use when user asks to review code, check for bugs, or audit a codebase.
agent-builder
Design and build AI agents for any domain. Use when users: (1) ask to "create an agent", "build an assistant", or "design an AI system" (2) want to understand agent architecture, agentic patterns, or autonomous AI (3) need help with capabilities, subagents, planning, or skill mechanisms (4) ask about Claude Code, Cursor, or similar agent internals (5) want to build agents for business, research, creative, or operational tasks Keywords: agent, assistant, autonomous, workflow, tool use, multi-step, orchestration
skill-builder
Automatically detect source types and build AI skills using Skill Seekers. Use when the user wants to create skills from documentation, repos, PDFs, videos, or other knowledge sources.
Related Guides
Python Django Best Practices: A Comprehensive Guide to the Claude Skill
Learn how to use the python django best practices Claude skill. Complete guide with installation instructions and examples.
Optimize Rell Blockchain Code: A Comprehensive Guide to the Claude Skill
Learn how to use the optimize rell blockchain code Claude skill. Complete guide with installation instructions and examples.
Python 3.12 FastAPI Best Practices: A Comprehensive Guide to the Claude Skill
Learn how to use the python 312 fastapi best practices Claude skill. Complete guide with installation instructions and examples.