Patent Kit

AI-Native Patent Analysis Kit, designed for Claude Code.

[!IMPORTANT] > Disclaimer: This tool is provided for informational purposes only. The outputs do not constitute legal advice or professional patent opinions.

Overview

A complete toolkit that empowers AI agents to autonomously search, analyze, and evaluate patents with human-level precision.

This kit provides structured commands to automate:

• Concept Interview: Define product concept and identify competitors.
• Targeting: Create a target population from patent databases.
• Screening & Evaluation: Filter and analyze patents for relevance.
• Claim Analysis: Compare product features against patent elements.
• Prior Art Research: Search for prior art references for high-risk patents.
• Investigation Reporting: Track progress across all phases.

Install

Add this repository as a marketplace and install the plugin to your Claude Code environment:

# 1. Add this repository as a marketplace
claude plugin marketplace add sonesuke/patent-kit

# 2. Install the plugin (automatically loads required MCPs)
claude plugin install patent-kit@patent-kit-marketplace

Prerequisites

You must have the following CLI tools installed and accessible in your system PATH. When this plugin is loaded, it will automatically connect to these tools as built-in MCP servers.

• Google Patent CLI (google-patent-cli)
• arXiv CLI (arxiv-cli)
• sqlite3 — Required for database operations (screening, evaluating, etc.)

Quick Start

Navigate to your working directory and start Claude:

mkdir my-patent-project && cd my-patent-project
claude

Then run the skills in order:

/patent-kit:concept-interview
# Output: specification.md

Workflow

Phase 0: Concept Interview

Define product concept and identify competitors.

/patent-kit:concept-interview
# Output: specification.md

Phase 1: Targeting

Generate search queries and create a target population.

/patent-kit:targeting
# Output: targeting.md, keywords.md, csv/

Phase 2: Screening & Evaluation

Screen patents for relevance and evaluate claims.

/patent-kit:screening
# Output: patents.db (screening results)

/patent-kit:evaluating
# Output: patents.db (claims and elements)

Phase 3: Claim Analysis

Compare product features against patent elements.

/patent-kit:claim-analyzing
# Output: patents.db (similarity results)

Phase 4: Prior Art Research

Search for prior art references for patents with Moderate/Significant similarities.

/patent-kit:prior-art-researching
# Output: patents.db (prior art references)

Progress Report

Track progress across all phases at any time.

/patent-kit:investigation-reporting
# Output: PROGRESS.md (overall progress)
# Output: <patent_id>.md (specific patent report)

Output Structure

.
├── specification.md          # Phase 0: Product definition
├── targeting.md               # Phase 1: Search strategy
├── keywords.md                # Phase 1: Search keywords
├── csv/                       # Phase 1: Target patent data
│   └── *.csv
└── patents.db                 # SQLite database for all investigation data

Skills

User-Invocable Skills

Skill	Purpose
concept-interviewing	Define product concept and identify competitors
targeting	Create target population from patent databases
screening	Filter patents by legal status and relevance
evaluating	Decompose claims and elements for relevant patents
claim-analyzing	Compare product features against patent elements
prior-art-researching	Search for prior art references
investigation-reporting	Generate progress reports

Internal Skills

These skills are automatically invoked by other skills and should not be used directly.

Skill	Purpose
investigation-preparing	Initialize SQLite database and import CSV files
investigation-fetching	Retrieve data from SQLite database
investigation-recording	Record data to SQLite database
legal-checking	Review documents for legal compliance violations

License

MIT