patent-skills
这是一个开源的AI代理技能库,专注于专利分析和原理提取。它遵循Agent Skills标准,旨在扩展AI编码代理的能力,帮助用户高效地从专利文档中识别关键信息和技术原理。
git clone https://github.com/Obviously-Not/patent-skills && cp -r patent-skills/code-patent-scanner .claude/skills/Before / After 效果对比
1 组传统专利分析耗时费力,需人工阅读大量文档,难以快速定位核心技术原理,易遗漏关键信息,效率低下。
AI代理自动进行专利分析和原理提取,显著缩短分析时间,提高信息捕获的准确性,加速研发决策。
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.
| Skill | What It Does | Start Here? |
|---|---|---|
| Code Patent Scanner | Scans your codebase for distinctive patterns. Returns structured scoring and evidence for patent consultation. | Yes — for code |
| Code Patent Validator | Takes scanner findings and generates search queries to research existing implementations. | Use after Code Patent Scanner |
| Patent Scanner | Describe your concept and discover what makes it distinctive. Structured analysis for patent consultation. | Yes — for concepts |
| Patent Validator | Takes 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.
| Skill | What It Does | When to Use |
|---|---|---|
| PBE Extractor | Extracts invariant principles from any text. Finds the ideas that survive rephrasing. | First step — extract before comparing or synthesizing |
| Principle Comparator | Compares two sources to find shared and divergent principles. | After extraction — with 2 sources |
| Principle Synthesizer | Synthesizes invariant principles from 3+ sources. Finds the core that survives across all expressions. | After extraction — with 3+ sources |
| Essence Distiller | Finds what actually matters in your content. The ideas that survive any rephrasing. | Extract first, then compare or refine |
| Pattern Finder | Discovers what two sources agree on. Finds the signal in the noise. | After extraction — with 2 sources |
Advanced
| Skill | What It Does | When to Use |
|---|---|---|
| Core Refinery | Finds the core that runs through everything. The ideas that survive across all your sources. | After 3+ extractions |
| Golden Master | Tracks 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
- Skills page: obviouslynot.ai/skills
- GitHub: github.com/Obviously-Not/patent-skills
- ClawHub: clawhub.ai (search "leegitw")
- Agent Skills standard: agentskills.io
- ObviouslyNot: obviouslynot.ai
License
MIT License - see LICENSE
Obviously Not - obviouslynot.ai
forum用户评价 (0)
发表评价
暂无评价
统计数据
用户评分
为此 Skill 评分