How to use patent-disclosure-skill: setup and Chinese patent drafting

What patent-disclosure-skill does

The project is a collection of skills, not a standalone app. Each skill is meant to be used from an AI coding agent environment. The README lists several sub-skills, including patent-disclosure for writing disclosure documents, patent-reader for plain-language reading, patent-oa for review-opinion replies, and patent-chart for claim comparison tables.

• patent-disclosure: turns materials into a disclosure document (invention, utility model or design)
• patent-reader: turns a publication number or PDF into plain-language notes and graphs
• patent-oa: breaks down review opinion clauses and drafts reply text
• patent-chart: builds an Excel claim comparison table; the README says it does not constitute legal opinion
• patent-exam-policy: summarises recent CNIPA policy news into a brief

Check requirements and clone the repository

The source states Python 3.9+ as a prerequisite. The README points to INSTALL.md for detailed installation steps, which this guide does not repeat. The only install command given in the source is the clone command below.

• Confirm Python 3.9 or newer is available
• Make sure you have an AI agent environment where skills can be loaded
• Read INSTALL.md in the repository for the full setup

Write a patent disclosure document from project materials

Start with the patent-disclosure skill. The README describes it as a way to pull out the points worth protecting from design documents and code, check prior art, and produce a disclosure document. Outputs are described as timestamped files in an outputs folder, with mermaid diagram directories and revision history when you iterate.

The README also mentions that after a first version is finalised, you can do a protection-oriented 1+N patent layout. Check the patent-disclosure README in the repository for the exact trigger phrases and steps.

• Gather design documents, code descriptions and structure or appearance drawings
• Ask the agent to write the disclosure, using the trigger word the skill documents (the README lists 「交底书」)
• Review the timestamped output and iterate; earlier versions are kept

Read a Chinese patent in plain language with Obsidian

The patent-reader skill converts a single patent into plain-language notes and graphs. The README says these can be stored in Obsidian, where notes are linked and shown in a graph view. Obsidian setup for Windows, including optional community plugins, is documented in skills/patent-reader/docs/obsidian-setup-guide.md.

The README describes this as a way to build a private, linked patent knowledge base over time. It is optional: you need Obsidian only if you want the linked note system.

• Provide a publication number or PDF
• Store the resulting notes in your Obsidian vault
• Use the graph view to see links between terms, claims and figures

Draft a reply to a patent office review opinion

The patent-oa skill is for review opinions (审查意见). According to the README it breaks down clauses, answers questions and drafts a reply. It can also store historical cases in anonymised form so the agent can compare with past approaches.

Treat the draft as a starting point. The README says these tools are aids rather than professional patent services, so a patent agent should check the reply.

• Paste or load the review opinion text
• Ask for clause-by-clause questions and a draft reply
• Check the draft with a qualified patent agent before filing