patent-disclosure-skill

An MIT-licensed skill for Chinese patent work, covering invention, utility-model, and industrial-design disclosure drafting, plain-language patent reading, policy monitoring, and assistance with examination responses. For disclosure drafting, it scans project materials such as Markdown, code, DOCX, PPTX, and optionally STEP/CAD files; discusses candidate patent points, performs prior-art searches with China National Intellectual Property Administration sources as the preferred source, and produces a redacted Markdown disclosure with Mermaid diagrams or planned patent figures, with an optional editable DOCX export. Utility-model and design workflows use schemas and a figure plan to organize structural drawings, product line art, views, and component numbering. It saves timestamped drafts, preserves multiple revisions and revision conversations, and supports self-checking, correction, and material-merging workflows.

Its patent-reading workflow extracts text from a patent or PDF into an Obsidian-based knowledge system, producing a claims tree, terminology table, feature-to-description-to-figure comparisons, narrative notes, Canvas relationship graphs, and linked clues from a limited set of public web sources. A separate policy-evolution mode collects recent policy and examination information without changing the workflow until confirmed. Its examination-response mode anonymizes historical notices and responses, indexes them with legal-defect tags, optionally adds local vector similarity retrieval, and drafts a response for human review. The repository documents Python and Node integrations, an Obsidian vault configuration, example cases, and outputs under local output directories or an Obsidian vault.