Chinese Patent Assistant

The Problem It Addresses

In the practice of Chinese patent law, professionals grapple with a series of repetitive and procedural tasks: from structuring and conducting prior art searches for technical disclosures, to comprehensively understanding and relating vast numbers of published patents; from tracking policy and examination standard updates by offices like the CNIPA, to searching, analyzing, and drafting response sketches for specific office action notifications. These interconnected steps traditionally rely heavily on individual expertise and are often disjointed.

How the Skill Works

This skill provides a modular, step-by-step intelligent assistance framework, centered around four core capabilities:

1. Patent Disclosure Writing (Mode A): This is a guided “fill-in-the-process” workflow. It begins by using an intake command to define boundaries and patent type (Invention, Utility Model, Design). It then performs a project scan (supporting Office docs, with optional CAD/STEP files), and subsequently guides you through patent point mining, prior art search (employing a two-stage query strategy prioritizing CNIPA’s EPUB database), abstract preview, and final drafting. For utility models and designs, mandatory figure planning and generation steps are included.
2. Patent Plain-Language Interpretation (Mode B): Triggered when you provide a patent number, PDF, or full text with the goal of “understanding,” the skill shifts to interpretation mode. It extracts text, structures the content, and generates 通俗解读笔记 (plain-language notes) and 知识图谱 (knowledge graphs), with strong recommendations for integration with Obsidian note-taking software.
3. Policy/Examination Trend Sniffing (Mode C): This is an explicit “side-path” function. It systematically scrapes official sources like the CNIPA to identify recent (default: last 12 months) policy or examination standard changes, producing a timestamped checklist with 观点 ↔ 信源 URL (Viewpoints ↔ Source URLs). Crucially, this list requires human review before any changes to the skill’s own rules are even proposed.
4. Office Action Response Assistance (Mode D): Also requires explicit triggering. It supports the ingestion of sanitized office action cases and can distill local patent practice books into playbooks. When drafting a response, it can search through the case library and playbooks based on the notification’s content, synthesizing multiple strategies to generate a draft for your revision and review.

The skill operates on a “Read Command File -> Execute” pattern. The SKILL.md file itself acts as a router, while the actual logic is encapsulated within the Markdown files under the prompts/ directory, achieving a single-package modular design.

Scope and Important Notes

• An Assistance Tool for Professionals: This skill aims to assist in completing methodical, repeatable work following established patent drafting and response procedures. It does not replace the final professional judgment and creative decisions of a patent attorney or agent. All generated disclosures, interpretation notes, or response drafts require human review.
• Explicit Mode Mutual Exclusivity: Based on your intent, the skill enters different operational modes. Interpreting a patent (B) will not automatically trigger disclosure writing (A), nor will any operation automatically enter strategy adjustment (C) or response drafting (D) modes without your explicit command. This prevents unintended workflow interference.
• Inputs and Outputs: The skill’s entry points are technical materials (documents, CAD drawings, patent numbers) and clear instructions. Its outputs are structured texts (e.g., disclosure.md, notes.md, response_draft.md) and necessary graphics. For designs and utility models, line art generation is a mandatory step. However, for pure CAD models, generated projection diagrams serve only as reference material and are not included as formal drawings in the final document.