patent-disclosure-skill for Chinese patents
patent-disclosure-skill
Nine skills for patent work in China: invention disclosures, application documents, CNIPA search, patent reading and office action replies
- 12,479 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Runs Python scripts and a controlled browser, and for STEP files sets up a separate CadQuery environment
- Reads confidential invention materials and builds queries to the CNIPA publication site from them
- Writes Word, Markdown and spreadsheet files into the project's outputs folder and, when configured, into an Obsidian vault
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/patent-disclosure-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Runs Python scripts and a controlled browser, and for STEP files sets up a separate CadQuery environment
- Reads confidential invention materials and builds queries to the CNIPA publication site from them
- Writes Word, Markdown and spreadsheet files into the project's outputs folder and, when configured, into an Obsidian vault
This is third-party code. Review the repository files before installing.
What it does
The repository holds one root SKILL.md and nine nested skills in the skills folder. The root file only routes a request through a table: which nested skill to read and on which user phrases. The main one, patent-disclosure, takes an inventor's materials (documents, code, diagrams, Word files and slides) to an invention disclosure for an invention, utility model or design patent: it mines patentable points, runs a light novelty search, writes the text and saves it as Markdown and Word, turning mermaid block diagrams into images. For utility models it prepares drawings with part numbers, and for designs line views of the product. The patent-application skill takes a finished disclosure to claims, a specification, an abstract and figures, and patent-docket runs both stages in sequence from an engineer's materials. The patent-search skill queries the CNIPA publication site by inventor, applicant, class, title and abstract and writes a report into `outputs/patent-search`. The remaining skills read other people's patents and explain them in plain language (patent-reader), match the features of independent claims against documents in an Excel table (patent-chart), build a map of accumulated patents (patent-map), help reply to examiner objections (patent-oa) and compare drafting technique against the agency's policy news (patent-exam-policy). All results land in the `outputs` folder of the current project. The skill asks the user for missing facts instead of guessing: unknown details of the invention are recorded in an uncertain section and do not enter the claims.
Who it is for. For engineers, patent professionals and founders in China who prepare Chinese patent applications and work with Chinese-language text alongside an agent.
Good fit when
- You have project documents and code with no patentable points identified, and a Word disclosure is due
- You need a disclosure for a utility model or design from drawings and product photos
- A finished disclosure has to become claims, a specification, an abstract and figures
- You want to find publications by inventor, applicant or class on the CNIPA publication site and get a report
- A Chinese patent is hard to read and you need notes with a graph of terms and claims
Not a fit when
- The application is not for CNIPA: the skills are built on Chinese forms, terminology and examination practice
- You need a legal opinion or an infringement decision: the claim chart and examination replies stay drafts for a professional
- The machine cannot have Python 3.9 or newer: the scripts and Word output run only on it, and without scripts you are left with manual Markdown work
- You need to work without Chinese: the skills' instructions and texts are written in Chinese
Example request
Go through the project materials folder, find the patentable points and prepare an invention disclosure in Word with a block diagram and a novelty searchLimitations
- All instructions and results are in Simplified Chinese by default. The rules are tied to CNIPA, and the author labels the claim-chart skill as not giving a legal opinion.
- The full invention path needs Python 3.9 or newer, the dependencies from `pip install -r requirements.txt` (including `playwright` and `python-docx`) and Chrome or Edge, which the skill uses to draw block diagrams and open the publication site.
- The novelty search and the bibliographic search go from your machine to the CNIPA publication site, and the queries are built from words of your invention. The author forbids presenting the bibliographic search result as the disclosure's novelty search.
- For parsing STEP 3D models an optional separate environment with CadQuery on Python 3.10–3.12 is created. It turns on only with your consent.
- The patent-reading skill works best with Obsidian: without the `PATENT_READER_OBSIDIAN_VAULT` variable the notes stay in `outputs/patent_reader`. The application, map, examination, chart and policy sub-skills start only when named directly.
How to disable. Delete the patent-disclosure-skill folder from the agent's skills directory (for example `.claude/skills` in the project or `~/.cursor/skills`). Work results stay in the project's `outputs` folder, which you can delete by hand. If an environment for STEP models was created, it goes away with the skill folder `skills/patent-disclosure/tools/cad-env`.
FAQ
Which patent types does the skill support?
Inventions, utility models and designs. If the type is not named, invention is the default. For utility models and designs the skill asks for drawings and photos and, before building the text, produces line drawings and never invents part numbers itself.
What does the skill do about novelty?
At step five it takes three or four key phrases and searches them on the CNIPA publication site, ten records per page, then pins one closest document and writes a table of differences. The skill forbids inventing a document, and when the check fails the text is still written but the inventive-step rating is lowered.
How does the skill keep guesses out of the text?
Details the materials say nothing about go into an uncertain section and do not enter the claims or the specification. When facts are missing, the skill asks questions along the way and does not fill gaps with inventions.
What does a disclosure come out as?
A Markdown file and a Word file with editable formulas. Block diagrams are rendered as images through a browser. If there is no browser, the skill leaves mermaid code in the Markdown, and the images can be made later with the `mermaid_render.py` script.
Can I start the application sub-skill right away?
Only with the folder of a finished disclosure. Without it the application skill stops and suggests writing the disclosure first. The patent-docket sub-skill runs both stages itself and asks the engineer for missing facts.
Discussion
No questions or reviews yet.
Related
SkillsSkill with 30 prompt templates for research papers in English and Chinese, from literature review to grant proposals
A skill for Chinese technical writing: rewrites docs, interface strings and API descriptions, keeps the facts intact and checks output with a linter
A set of three research skills: paper writing, academic Word and PowerPoint documents, scientific computing and figures
A skill set for academic papers: deep method analysis, visual slides and a coherent write-up from a paper PDF
A skill that turns an article, report or notes into a deck of full-slide images and assembles them into a pptx
A skill set for research work in the style of Nature journals: finding and reading papers, writing, statistics, figures, reviewer responses
README in short
The repository's front page, in Chinese, opens with two stories: why the author needed a disclosure-writing skill and why patent reading was split out to build a personal notes base. Then come screenshots of results (folders of disclosure versions, line drawings, a patent map), a table of nine sub-skills with their purposes and trigger words, a section on supporting the author and links to INSTALL.md and the root SKILL.md. Installation for Claude Code and Cursor is in INSTALL.md, and each sub-skill has its own README, examples and tests.
README badge
Original SKILL.md
--- name: patent-disclosure-skill description: "中国专利技能：挖掘专利点并撰写发明/实用新型/外观设计交底书；将已有交底改写为申请文件四件套；依据发明人材料衔接交底与申请；按著录项目检索公布公告；解读专利文献；基于已入库解读生成专利地图；对照审查政策撰写简报；辅助审查意见答复；编制权利要求对照表底稿（不构成法律意见）。" version: "5.1.0" user-invocable: true allowed-tools: Read, Write, Edit, Grep, Glob, WebSearch, Bash --- # 中国专利技能 按用户意图 **`Read`** 对应入口的 `SKILL.md`，再按该流程执行。**先查「路由判定表」定去向，再看「通则」。** ## 能力总览 | 能力 | 做什么 | 入口 | |------|--------|------| | **交底** | 挖专利点 → 查新 → 成稿 → 迭代（发明 / 实用新型 / 外观） | `skills/patent-disclosure/SKILL.md` | | **申请文件** | 已有交底 → 权要 / 说明书 / 摘要 / 附图 | `skills/patent-application/SKILL.md` |