China Patent Full-Lifecycle Assistant
Paste the following prompt into your AI chat to install this skill:
Please install @davtime/zhongguozhuanli-zhuanxiecailiaojiedushenchazhengce according to the official installation guide at https://skillhub.cn/install/skillhub.md.
About this skill
The Specific Problem It Solves
Patent work involves a series of specialized and repetitive tasks: from drafting and novelty searching for technical disclosure documents, to rapidly understanding published patents, conducting multi-faceted reviews, responding to office actions, and tracking policy changes from the China National Intellectual Property Administration (CNIPA). Each step has its own methodology, formatting requirements, and document templates, but stitching them together requires significant manual effort in switching contexts and organizing information.
The China Patent Full-Lifecycle Assistant aims to digitize and systematize this workflow. It does not replace expert judgment but rather encapsulates expert experience into six operational modes and an automated pipeline. Its goal is to ensure the correct templates, scripts, and checklists are invoked at each stage, minimizing formatting errors and procedural oversights.
How the Skill Works: Core Modes and Key Steps
The skill is menu-driven and will force-display a mode selection upon first activation. Its core capabilities are realized through these modes:
- A · Disclosure Drafting: Starts from the technical concept (
挖点), performs an automated novelty search, and generates a compliant initial draft. It auto-detects the patent type (invention / utility model / design) based on input materials and calls workflow scripts underreferences/disclosure-workflow.md. - C · Patent Interpretation: Given a patent number or PDF, it produces a plain-language summary, a glossary of terms, and can attempt to build a knowledge map. It executes instructions in
references/patent-reading.mdand strongly recommends integration with note-taking tools like Obsidian. - D · Multi-Agent Review: This is a key differentiating capability. Upon receiving a
.docxapplication file, it launches over 13 parallel review agents to perform hard checks from various dimensions (e.g., claim clarity, novelty argumentation, formal compliance), ultimately generating an annotated, revised copy. This process mandates the use of built-in scripts, prohibits sub-agents from asking user questions, and must pass structural and paragraph-count validations. - E · Office Action Response: Handles office actions (OA). Through case de-identification and storage (via
templates/oa-case-template.md) and vector search for similar precedents, it assists in drafting a response. This mode is disabled by default and requires explicit triggering. Drafts must undergo human review before submission. - F · Policy Radar: Sniffs out the latest CNIPA policies or examination trends and provides an informational list with URLs. This mode is also disabled by default and must be triggered explicitly; the generated list is for reference only and cannot directly modify the skill's own configuration.
Key Execution Principle: The skill follows a "thin routing" design. SKILL.md handles mode dispatch only; specific instructions reside in the references/ and templates/ directories. In environments like Windows, information output to stderr does not necessarily indicate failure; success is determined by the exit code and specific prefixes like EPUB_HITS_JSON: or REVIEW:. All file paths must be absolute.
Application Boundaries and Considerations
- Explicit Triggers and Human Review Gates: Operations involving final deliverables (OA drafts, E) or affecting skill configuration (policy lists, F) must be triggered by explicit user instructions and their output requires human review. The skill will not enter these modes automatically.
- File Format Dependencies: Mode D strictly requires a
.docxfile as input; Mode C supports public numbers, PDFs, etc. The skill validates input formats. - Environment Compatibility: On Windows PowerShell, scripts may flag
stderrstreams asNativeCommandError, which does not affect script execution and should not be a reason to reinstall or downgrade the skill. - Not an Omnipotent Tool: The skill is an auxiliary tool designed to structure and accelerate known processes. Final legal judgment and the distillation of the core innovative aspects of a technical solution still depend on the user's (inventor, agent, or examiner) professional expertise and decision-making. It does not generate entirely new technical solutions nor replace formal communication with the CNIPA.
Use Cases
- A patent engineer extracts core innovations from lab notebooks and uses Mode A to generate a compliant draft invention disclosure, with an initial automated novelty search.
- A R&D personnel receives a competitor's published patent PDF and needs to quickly grasp its technical scheme, using Mode C to produce a plain-language summary, claim dependency tree, and key terminology list.
- A patent agent receives a .docx patent application file for submission and uses Mode D to initiate multi-agent parallel review, focusing on logical flaws in the claims and formal compliance.
- An IP specialist receives an office action from the CNIPA and uses Mode E, leveraging a case library to help draft a preliminary response addressing novelty or inventive step objections.
Best For
- Hardware or algorithm R&D engineers who need to write 1-2 technical disclosure documents weekly but are unfamiliar with the formal format of patent legal texts.
- Patent agents or paralegals responsible for patent drafting, office action responses, and case management, seeking tools to batch-process format checks and case retrieval.
- Technical managers or IP specialists who manage a company's IP portfolio, needing to regularly monitor policy changes, understand competitor patent landscapes, and write analysis reports.
- Tech media editors or industry analysts who frequently need to read and parse domestic and international technical patents to identify industry trends and competitive dynamics.
Related Skills
Systematically identify and evaluate the economic moats of listed companies using the Tang Shu Fang investment methodology for long-term investment analysis.
A skill that converts natural language questions into A-stock data queries and returns verifiable structured analysis conclusions, covering multi-dimensional analysis of market trends, fundamentals, and news.
An AI financial copilot by Wind, integrating financial databases and multimodal analysis to provide end-to-end support across investment research, asset allocation, risk control, quant, and report generation.
Searches for latest policies and generates GB/T 9704-2020 compliant official documents (requests or reports) in Word and PDF formats.