About this project

# China Patent.skill This is an open-source AI skill suite for Chinese-speaking users, designed to assist R&D personnel and IP professionals in completing the full patent workflow. It integrates multiple sub-skills, ranging from technical point mining, invention disclosure drafting, and application document generation to patent interpretation, search analysis, and office action response assistance. ## Core Features ### 1. Patent Disclosure and Application Document Drafting - **Multi-type Support**: Covers invention patents, utility model patents, and design patents. - **Automated Generation**: Converts project materials into structured invention disclosures, including block diagrams, editable Word documents, and black-and-white drawings. - **Iterative Optimization**: Supports multiple rounds of supplementary materials and error correction, retaining modification traceability records. - **Layout Recommendations**: After the first draft is finalized, provides "protection-type 1+N" patent layout recommendations. ### 2. Plain-Language Patent Interpretation and Knowledge Base Construction - **Plain-Language Reading**: Converts obscure patent full texts into notes and diagrams that ordinary people can understand. - **Obsidian Integration**: Interpreted content can be directly stored in Obsidian, leveraging backlinks, graphs, and the plugin ecosystem to build a private patent knowledge base. - **Correlation Analysis**: Supports family comparison, technical route mapping, and difference scanning, bridging the gap from "understanding" to "application." ### 3. Patent Map and Analysis - **Visual Exploration**: Provides visual charts such as semantic terrain, applicant quadrant, family citation network, and technical-effect matrix. - **Standalone Service**: The patent map can run as a standalone local service, enabling interactive exploration through a browser. ### 4. Intelligent Search and Policy Monitoring - **Multi-Dimensional Search**: Supports search by person name, company, and classification number, and can reverse-engineer search expressions from product images or claims. - **Policy Briefings**: Automatically scrapes the latest policies from the CNIPA official website, generates easy-to-understand policy briefings, and analyzes their impact on existing disclosure techniques and examination standards. ### 5. Office Action Response Assistance - **RAG Enhancement**: Builds a retrieval-augmented generation system based on personal professional experience and practical books to assist in deconstructing office action clauses and drafting responses. - **Expert Technique Distillation**: Integrates expert experience into the knowledge base to improve response quality and efficiency. ## Sub-Skill List | Skill Name | Trigger Phrase | Main Capabilities | | :--- | :--- | :--- | | **patent-disclosure (Invention Disclosure Drafting)** | "Invention disclosure" | Mines patent points from materials, performs novelty search and desensitization, and generates invention/utility/design disclosures. | | **patent-application (Application Documents)** | "Application documents", "Application draft" | Converts invention disclosures into complete application documents (claims, specification, abstract, drawings). | | **patent-docket (Case Review)** | "Do disclosure and application together", "Produce disclosure and application from scratch" | Simulates dialogue between a disclosure engineer and an agent, plans workflow over multiple rounds, and ensures material accuracy. | | **patent-reader (Plain-Language Interpretation)** | "Read patent" | Takes a publication number or PDF as input, generates plain-language notes and diagrams, and integrates with Obsidian. | | **patent-map (Patent Map)** | "Patent map", "Case map" | Displays visual analysis results such as semantic terrain, citation networks, and effect matrices. | | **patent-oa (Office Action Response Assistance)** | "Office action response", "Office action" | Deconstructs office actions and assists in drafting responses using the RAG library. | | **patent-search (Bibliographic Search)** | "Bibliographic search" | Supports patent search across multiple dimensions, including reverse-engineering search expressions from images and text. | | **patent-exam-policy (Policy Briefing)** | "Policy briefing", "Policy radar" | Monitors CNIPA policies, generates briefings, and assesses their impact on existing technologies. | ## Technical Features - **Python 3.9+**: Developed in Python, relying on Playwright for web interaction and data scraping. - **AgentSkills Standard**: Follows the AgentSkills standard specification, making it easy to integrate into various AI Agent platforms. - **Multimedia Processing**: Supports extracting design line drawings from product images, generating utility model line drawings from structural diagrams, and extracting isometric projection views from CAD models. ## Installation and Usage For detailed installation instructions, please refer to the `INSTALL.md` file in the repository. The skill entry points and Agent workflow definitions are in `SKILL.md`. For users who need deep Obsidian integration, refer to `skills/patent-reader/docs/obsidian-setup-guide.md` for configuration. ## License This project is licensed under the MIT License.