Home/数据 & AI/patent-skills
P

patent-skills

by @Obviously-Notv
3.5(0)

This is an open-source AI agent skill library focused on patent analysis and principle extraction. It adheres to the Agent Skills standard, aiming to expand the capabilities of AI coding agents, to help users efficiently identify key information and technical principles from patent documents.

AI AgentPatent AnalysisInformation ExtractionOpen SourceAgent SkillsGitHub
Installation
git clone https://github.com/Obviously-Not/patent-skills && cp -r patent-skills/code-patent-scanner .claude/skills/
compare_arrows

Before / After Comparison

1
Before

Traditional patent analysis is time-consuming and labor-intensive, requiring manual review of numerous documents, making it difficult to quickly identify core technical principles, prone to missing key information, and inefficient.

After

AI agents automatically perform patent analysis and principle extraction, significantly reducing analysis time, improving the accuracy of information capture, and accelerating R&D decisions.

description SKILL.md

AI Agent Skills

Open-source skills for patent analysis and principle extraction. Built on the Agent Skills standard.

What Are Agent Skills?

Agent skills are portable instructions that extend what AI coding agents can do. Each skill is a SKILL.md file — a Markdown document with YAML frontmatter that any compatible agent can read and execute.

The Agent Skills standard was originated by Anthropic and adopted by 27+ tools including Claude Code, Gemini CLI, Codex CLI, Cursor, GitHub Copilot, Goose, and more. Write once, use everywhere.

Format

skill-name/
  SKILL.md           # Required — instructions + YAML frontmatter
  scripts/           # Optional — executable code
  references/        # Optional — detailed docs loaded on demand
  assets/            # Optional — templates, data files

The SKILL.md file contains:

---
name: skill-name
description: What this skill does and when to use it.
---

# Instructions

Step-by-step instructions the agent follows...

Agents read the description field to decide when to activate the skill. The Markdown body contains the full instructions, loaded on demand to keep context usage efficient.

Installation

Claude Code / Gemini CLI / Cursor

Clone the repo and copy any skill into your agent's skills directory:

git clone https://github.com/Obviously-Not/patent-skills
cp -r patent-skills/code-patent-scanner .claude/skills/

The skill becomes available as a /code-patent-scanner slash command. Claude Code (and other compatible agents) will also auto-invoke it when it detects a relevant task.

Project-level (shared with your team via version control):

your-project/.claude/skills/code-patent-scanner/SKILL.md

Personal (available in all your projects):

~/.claude/skills/code-patent-scanner/SKILL.md

Any LLM Agent (Copy/Paste)

Open the SKILL.md file on GitHub, copy the contents, and paste directly into your agent's chat. The agent will follow the instructions immediately. This works with any LLM — no special setup required.

OpenClaw

clawhub install leegitw/code-patent-scanner

Skills install to ./skills/ and OpenClaw loads them automatically.

Skills Catalog

Patent Analysis

These skills help you discover and validate patentable patterns in code and concepts.

SkillWhat It DoesStart Here?
Code Patent ScannerScans your codebase for distinctive patterns. Returns structured scoring and evidence for patent consultation.Yes — for code
Code Patent ValidatorTakes scanner findings and generates search queries to research existing implementations.Use after Code Patent Scanner
Patent ScannerDescribe your concept and discover what makes it distinctive. Structured analysis for patent consultation.Yes — for concepts
Patent ValidatorTakes concept analysis and generates search queries to research the landscape.Use after Patent Scanner

Recommended workflow: Scanner → Validator → Research → Attorney

Start with the appropriate Scanner to identify patterns, then use the matching Validator to generate search strategies. Use those strategies for your own research, then bring your findings to a patent attorney.

Principle-Based Extraction

These skills extract and compare the invariant ideas that survive any rephrasing — useful for understanding what's actually novel in your work.

SkillWhat It DoesWhen to Use
PBE ExtractorExtracts invariant principles from any text. Finds the ideas that survive rephrasing.First step — extract before comparing or synthesizing
Principle ComparatorCompares two sources to find shared and divergent principles.After extraction — with 2 sources
Principle SynthesizerSynthesizes invariant principles from 3+ sources. Finds the core that survives across all expressions.After extraction — with 3+ sources
Essence DistillerFinds what actually matters in your content. The ideas that survive any rephrasing.Extract first, then compare or refine
Pattern FinderDiscovers what two sources agree on. Finds the signal in the noise.After extraction — with 2 sources

Advanced

SkillWhat It DoesWhen to Use
Core RefineryFinds the core that runs through everything. The ideas that survive across all your sources.After 3+ extractions
Golden MasterTracks source-of-truth relationships between files. Knows when derived content becomes stale.After synthesis — track derivations

Important Disclaimer

These skills are not legal advice. They identify technical patterns, not patentability.

What they DO:

  • Identify distinctive patterns in code and concepts
  • Generate search queries for prior art research
  • Suggest research strategies
  • Document technical evidence

What they DON'T:

  • Search patent databases
  • Assess patentability
  • Provide legal conclusions
  • Replace professional counsel

Always consult a patent attorney for IP guidance.

Free Tier Options

These skills work with any LLM. You don't need a paid subscription to use them.

Option 1: Local LLM (Privacy-First)

Run entirely on your machine with Ollama:

# Install Ollama
brew install ollama  # macOS
# or download from https://ollama.ai

# Pull a capable model (32B recommended for patent analysis)
ollama pull qwen2.5:32b

# Run with your agent
# Skills work with any agent that can call local models

System requirements: 32GB RAM recommended for 32B models, 16GB minimum for 7B models.

Model recommendations: For patent analysis, larger models (32B+) produce better results. qwen2.5:32b and llama3.1:70b work well. See Ollama's model library for options.

Option 2: BYOK (Bring Your Own Key)

Use your own API key with any provider:

OpenRouter (access to 100+ models):

export OPENROUTER_API_KEY="sk-or-..."
# Skills automatically use OpenRouter when key is set

Google AI Studio (free tier available):

export GEMINI_API_KEY="..."
# Get free key at https://aistudio.google.com/apikey

Option 3: Agent's Built-in Model

Most AI coding agents (Claude Code, Cursor, Copilot) include their own model access. Just install the skill and use it — no additional API key needed.

Related Projects

  • Live Neon Skills - Open-source PBD skills for principle extraction and synthesis. The 7 PBD skills were migrated from this repo to live-neon/skills.
  • NEON-SOUL - Soul synthesis for AI agents, using PBD skills for principle extraction and compression.

Links

License

MIT License - see LICENSE


Obviously Not - obviouslynot.ai

forumUser Reviews (0)

Write a Review

Effect
Usability
Docs
Compatibility

No reviews yet

Statistics

Installs7
Rating3.5 / 5.0
Version
Updated2026年4月7日
Comparisons1

User Rating

3.5(0)
5
0%
4
0%
3
0%
2
0%
1
0%

Rate this Skill

0.0

Compatible Platforms

🔧Claude Code
🔧Gemini CLI
🔧Cursor

Timeline

Created2026年4月7日
Last Updated2026年4月7日