Chinese Patent Skill for Claude Code: Drafting Invention Disclosure Documents
WHY IT MATTERS
A new Chinese-language skill for Claude Code assists with patent mining and drafting invention disclosure documents for Chinese patents. It supports invention, utility model, and design patents and includes policy tracking.
What Happened
A Chinese-language skill for Claude Code has been published on GitHub that automates patent mining and the drafting of invention disclosure documents. The skill supports three Chinese patent categories — invention (发明), utility model (实用新型), and design (外观设计) — and bundles a policy-tracking component that monitors regulatory updates. Distribution is via a public GitHub repository as a Claude Code skill, installable into existing agent workflows.
Why It Matters
IP documentation has historically sat outside general-purpose agent tooling because it requires jurisdiction-specific structure, statutory phrasing, and continuous awareness of rule changes from CNIPA. This skill collapses that gap for the Chinese market, where filing volume and procedural specificity are both high. For operators building agents in or targeting China, jurisdiction-bound, compliance-heavy document generation is now a commodity function rather than a specialized vendor contract. The policy-tracking layer addresses a recurring cost center: manual monitoring of examination guidelines, fee schedules, and category-specific rule updates. The practical beneficiary is small engineering teams that previously could not justify legal-ops overhead for first-pass drafting.
Technical Details
The skill operates as a Claude Code extension, meaning it inherits the host agent's file system access, tool calls, and context window rather than running as a standalone service. It classifies candidate disclosures into invention, utility model, or design categories and produces structured invention disclosure documents aligned to each category's expected fields. The policy-tracking module watches for regulatory changes and surfaces them into the drafting context, though the update cadence and source list are repository-defined rather than API-backed. Constraints are inherent to the approach: drafting quality depends on the completeness of engineering inputs, and the skill does not perform prior-art search, freedom-to-operate analysis, or formal claim strategy. Output is a preliminary disclosure, not a filing-ready specification, and jurisdiction coverage appears limited to Chinese practice.
Operational Impact
Builders can now wire engineering artifacts — design docs, commit histories, internal RFCs — into a pipeline that emits preliminary invention disclosures without human drafting labor. The discovery-to-disclosure cycle shrinks from days of back-and-forth with IP staff to an automated first pass, with attorneys repositioned to review, claim strategy, and prosecution rather than intake drafting. Policy tracking that previously required a subscription or a paralegal's attention becomes an in-repo dependency, reducing both cost and latency on rule changes. The workflow change is concrete: an agent that already ingests technical outputs gains a downstream IP-capture step, and the marginal cost of producing a disclosure approaches zero. What becomes obsolete is the first-pass drafting layer of legal-ops SaaS for teams already running Claude Code.
SOURCE
GitHub
SHARE
MORE FROM STUFFINSIDER
Orkas Launches Local-First Multi-Agent Desktop Team With 16 Specialists
Oct 31AGENTSClaude Code Game Studios: 49 Agents, 72 Workflow Skills
Oct 31AGENTSLangChain Releases DeepAgents: A Batteries-Included Agent Harness
Oct 30AGENTSCopilotKit Releases OpenBot: Open-Source AI Coworkers With Computers
Oct 29