how-to-guide

Patent Disclosure Skill Responsibilities: A Complete Guide to Each Sub-Skill

September 5, 2026 handsomestWei/patent-disclosure-skill ↗

The patent-disclosure-skill suite contains six specialized sub-skills—patent-disclosure, patent-application, patent-reader, patent-oa, patent-search, and patent-exam-policy—each handling a distinct phase of the patent lifecycle from prior-art search through policy analysis.

The handsomestWei/patent-disclosure-skill repository provides a modular, open-source framework for automating patent workflow tasks. Rather than a single monolithic tool, it decomposes complex patent work into granular, composable sub-skills that can be invoked independently or chained together. This architecture follows the Agent Skills specification, enabling seamless integration with AI assistants and automation pipelines.

According to the main README, each sub-skill maps to a specific user intent and trigger phrase, with its own tool implementations and manifest files.

Patent Disclosure Sub-Skill: Generate Invention Disclosure Documents

The patent-disclosure sub-skill is responsible for creating invention disclosure documents (交底书) across all three Chinese patent types: invention patents, utility models, and design patents.

Core Functions

This sub-skill performs five sequential operations:

  • Patentable point extraction – Identifies novel technical contributions from raw invention descriptions
  • Prior-art contextualization – Incorporates search results to position the invention against existing technology
  • Data de-identification – Removes sensitive business information before document generation
  • Draft generation – Produces the initial disclosure document with structured sections
  • Iterative refinement – Accepts feedback loops to improve claim scope and technical depth

Trigger and Entry

Attribute Value
Trigger phrase 「交底书」
Entry point [skills/patent-disclosure/README.md](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-disclosure/README.md)
Source reference README.md#L99-L126

Key Implementation Files

Patent Application Sub-Skill: Transform Disclosures into Filing Documents

The patent-application sub-skill converts completed disclosure documents into formal patent application packages ready for CNIPA submission.

Core Functions

  • Claims drafting – Structures independent and dependent claims with proper antecedent basis
  • Specification generation – Expands technical disclosure into full patent specification format
  • Abstract composition – Creates concise summaries meeting formal requirements
  • Figure preparation – Generates black-and-white line drawings and reference numerals
  • Package assembly – Outputs complete .docx or PDF filing packages

Trigger and Entry

Attribute Value
Trigger phrases 「申请文件」or 「申请底稿」
Entry point [skills/patent-application/README.md](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-application/README.md)
Source reference README.md#L128-L133

Key Implementation Files

Patent Reader Sub-Skill: Ingest and Annotate Published Patents

The patent-reader sub-skill transforms published patents into structured, interlinked knowledge artifacts within an Obsidian vault.

Core Functions

  • Patent ingestion – Accepts publication numbers or PDF uploads
  • Claim parsing – Extracts and hierarchically structures all claims
  • Terminology extraction – Identifies key technical terms and definitions
  • Knowledge graph generation – Creates linked notes connecting patents, inventors, assignees, and technology classes
  • Figure annotation – Embeds patent drawings with reference numeral callouts

Trigger and Entry

Attribute Value
Trigger phrase 「读专利」
Entry point [skills/patent-reader/README.md](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-reader/README.md)
Source reference README.md#L135-L140

Key Implementation Files

Patent OA Sub-Skill: Assist with Office Action Responses

The patent-oa sub-skill provides structured assistance for responding to patent examiner rejections and objections.

Core Functions

  • OA parsing – Extracts rejection grounds, cited prior art, and examiner arguments from Office Action documents
  • Response strategy suggestion – Recommends amendment approaches and argument frameworks
  • Prior-art vector retrieval – Searches for distinguishing features or secondary considerations evidence
  • Draft response generation – Produces formatted rebuttal documents with claim amendments

Trigger and Entry

Attribute Value
Trigger phrases 「审查答复」or 「审查意见」
Entry point [skills/patent-oa/README.md](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-oa/README.md)
Source reference README.md#L142-L147

Key Implementation Files

Patent Search Sub-Skill: Execute Bibliographic Searches

The patent-search sub-skill interfaces directly with CNIPA databases to perform structured bibliographic searches.

Core Functions

  • Inventor/applicant search – Retrieves patent families by human or organizational entity
  • Classification search – Queries by IPC, CPC, or LOC codes
  • Keyword and semantic search – Natural language and Boolean query construction
  • Structured report generation – Exports results in analyzable formats (CSV, JSON, markdown tables)

Trigger and Entry

Attribute Value
Trigger phrase 「著录检索」
Entry point [skills/patent-search/README.md](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-search/README.md)
Source reference README.md#L149-L154

Key Implementation Files

Patent Exam Policy Sub-Skill: Generate Policy Briefs

The patent-exam-policy sub-skill produces analytical briefs connecting current examination guidelines to drafting strategy.

Core Functions

  • Guideline monitoring – Tracks updates from the State Intellectual Property Office examination standards
  • Impact mapping – Correlates policy changes to claim drafting and prosecution tactics
  • Strategic brief generation – Creates actionable reports for patent attorneys and agents
  • Self-evolution capability – Designed to upgrade into an autonomous policy-tracking agent

Trigger and Entry

Attribute Value
Trigger phrases 「政策简报」or 「政策雷达」
Entry point [skills/patent-exam-policy/README.md](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-exam-policy/README.md)
Source reference README.md#L156-L162

Key Implementation Files

Typical Workflow Integration

The sub-skills are designed to compose into complete patent workflows. A standard progression:

  1. patent-search – Establish prior-art landscape before drafting
  2. patent-disclosure – Create invention disclosure grounded in search results
  3. patent-application – Convert disclosure to formal filing package
  4. patent-oa – Respond to examiner objections during prosecution
  5. patent-reader – Incorporate competitive patents into knowledge base
  6. patent-exam-policy – Adapt strategy to evolving examination standards

Each transition passes structured data—search results feed into disclosure generation, disclosure documents seed application drafting, and examiner citations inform OA response research.

Code Examples: Invoking Sub-Skills Programmatically

The following Python snippets demonstrate direct invocation of each sub-skill's core tools. These assume the repository is installed with pip install -r requirements.txt and the skills/ directory is on PYTHONPATH.

Patent Search Execution

from skills.patent_search.tools.cnipa_search import cnipa_search

results = cnipa_search(
    inventor="张三",
    applicant="华为技术有限公司",
    ipc_class="G06F17/00"
)
print(f"Found {results['total']} patents")
print(results['records'][0]['title'])  # First result title

Patent Disclosure Generation

from skills.patent_disclosure.tools.build_disclosure import build_disclosure

disclosure = build_disclosure(
    invention_title="一种基于深度学习的语音识别系统",
    technical_points=["端到端声学模型", "噪声环境下的鲁棒性提升"],
    prior_art_results=results['records']
)
print(disclosure['markdown_path'])  # Path to generated disclosure

Patent Application Transformation

from skills.patent_application.tools.emit_application_docx import emit_application_docx

app_package = emit_application_docx(
    disclosure_md=disclosure['markdown_path'],
    output_dir="./filing_ready/"
)
print(app_package['claims_docx'])   # Path to claims document

print(app_package['spec_docx'])     # Path to specification

Patent Reader Vault Integration

from skills.patent_reader.tools.vault.write_patent_obsidian_note import write_patent_obsidian_note

note_path = write_patent_obsidian_note(
    publication_number="CN11223344A",
    vault_root="/home/user/Obsidian/Patents",
    include_claims=True,
    include_figures=True
)
print(f"Created note: {note_path}")

OA Response Drafting

from skills.patent_oa.tools.emit_opinion_docx import emit_opinion_docx
from skills.patent_oa.tools.search_cases import search_cases

distinguishing_cases = search_cases(keywords=["深度学习", "语音识别", "端到端"])
response_doc = emit_opinion_docx(
    office_action_path="./office_action_cn11223344.pdf",
    cited_prior_art=["CN10987654A", "US2020123456A1"],
    supporting_cases=distinguishing_cases
)
print(response_doc['response_path'])

Policy Brief Generation

from skills.patent_exam_policy.tools.generate_policy_brief import generate_policy_brief

brief = generate_policy_brief(
    topic="人工智能领域创造性审查标准的最新变化",
    relevance_to_draft="权利要求中算法特征的撰写策略"
)
print(brief['summary'])
print(brief['recommended_claim_amendments'])

Summary

  • patent-disclosure generates invention disclosure documents (交底书) with prior-art integration and iterative refinement
  • patent-application transforms disclosures into formal CNIPA filing packages including claims, specification, and figures
  • patent-reader ingests published patents into structured, linked Obsidian knowledge graphs
  • patent-oa parses examiner rejections and drafts strategic Office Action responses
  • patent-search executes bibliographic queries against CNIPA databases with structured export
  • patent-exam-policy tracks examination guideline evolution and produces strategy briefs

Each sub-skill follows the Agent Skills specification with a SKILL.md manifest, modular tool implementations, and clear trigger phrases for conversational invocation.

Frequently Asked Questions

What is the difference between patent-disclosure and patent-application sub-skills?

The patent-disclosure sub-skill produces internal invention disclosure documents (交底书) that capture the technical invention in narrative form with extracted patentable points. The patent-application sub-skill takes a completed disclosure and reformats it into the strict formal structure required for patent office submission—including properly formatted claims, specification sections, and abstract. Think of disclosure as the technical foundation and application as the legal packaging.

Can I use patent-search independently without the other sub-skills?

Yes. The patent-search sub-skill is fully self-contained and exposes a direct API through [cnipa_search.py](https://github.com/handsomestWei/patent-disclosure-skill/blob/main/skills/patent-search/tools/cnipa_search.py). You can invoke it for standalone prior-art research, competitive intelligence, or invalidity searches without engaging the disclosure or application workflows. The search results are returned in structured JSON that any downstream process can consume.

How does patent-exam-policy differ from a simple news aggregator?

The patent-exam-policy sub-skill performs operationalized analysis rather than passive monitoring. According to the topic prompt map, it maps specific guideline changes to concrete claim drafting tactics and amendment strategies. It is designed as an upgradable skill—the architecture supports evolution into an autonomous agent that proactively alerts users to policy shifts affecting their pending applications.

Which sub-skill should I start with for a new invention?

Start with patent-search to understand the prior-art landscape, then proceed to patent-disclosure to document your invention's novel contributions in that context. This sequence ensures your disclosure is grounded in actual competitive technology and strengthens subsequent claim drafting. The repository's README#L99-L126 explicitly documents this recommended workflow progression.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"
Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →