中国专利.skill is a Chinese-language agent skill for finding patentable ideas, writing patent disclosure documents, explaining published patents, tracking examination policies, and helping prepare examination responses. It is for developers and researchers working with Chinese invention, utility-model, and design patents, with workflows that can store related research in Obsidian.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add handsomestWei/patent-disclosure-skill --skill patent-searchgit clone --depth 1 https://github.com/handsomestWei/patent-disclosure-skillWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/handsomestwei/patent-disclosure-skill/patent-search)<a href="https://agentmods.dev/skills/handsomestwei/patent-disclosure-skill/patent-search"><img src="https://agentmods.dev/badge/skills/handsomestwei/patent-disclosure-skill/patent-search/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/handsomestwei/patent-disclosure-skill/patent-search"><img src="https://agentmods.dev/badge/skills/handsomestwei/patent-disclosure-skill/patent-search.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00046 | $0.00840 |
| Opus 5 | $0.00023 | $0.00420 |
| Sonnet 5 | $0.00009 | $0.00168 |
| Haiku 4.5 | $0.00005 | $0.00084 |
Grade A, and why
patent-search scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
著录检索
通用公布站高级查询，不是「个人清单技能」。个人公开清单只是一种用法。
先 Read prompts/patent_search.md。 用户给单图或权要时再 Read prompts/derived_query.md。
默认少翻页
- 阈值在
config.yaml：max_pages（默认 3）；max_pages_hard（默认 20）仅在探测不到「共 N 页」时启用。 page_size只是探测不到「每页 N 条」时的回退，不要用它估总条数。- 对话里用户说「多翻几页 / 翻 5 页」→ 传
--max-pages N（普通检索仍受 hard 限制）；或改config.yaml。 - 不要一上来爬完全部分页。只有用户明确要求穷举清单时才加
--complete。 --complete第一页若读到「共 N 页」，按总页数翻完（硬上限随总页数）；读不到才用max_pages_hard。- 采到总页数才能说「已翻完全部分页」或「还剩 x 页」；没采到只谈还能不能点「下页」。
complete: false时禁止称为「全部」。退出码3表示分页不完整。
命令
python skills/patent-search/tools/cnipa_search.py --inventor "姓名" --applicant "单位"
python skills/patent-search/tools/cnipa_search.py --title "数据处理" --abstract "吸附 and 再生" --class B01J20 --max-pages 2
python skills/patent-search/tools/cnipa_search.py --abstract "折叠 and 杯盖" --type design --derived-from image --type-inferred --derived-note "折叠杯盖"
python skills/patent-search/tools/cnipa_search.py --inventor "姓名" --complete
单图必须 --type design|utility_model|invention，不能 all。用户没说类型时看图推断并加 --type-inferred。权要按文本选类型，可用 all。
单独拷走本包时：python tools/cnipa_search.py …。
结果默认落到 outputs/patent-search/SEARCH-YYYYMMDD-HHMMSS.md（gitignore）。改版式只动 tools/emit_search_report.py。
机读前缀：EPUB_SEARCH_MD: / EPUB_SEARCH_JSON: / EPUB_SEARCH_NOTE: / EPUB_SEARCH_INCOMPLETE:。对话里告诉用户 Markdown 路径，不要只倒 JSON。
必须走高级查询字段（#e72 发明人、#ti 名称、#abs 摘要/简要说明、#e51 分类号等），禁止把发明人姓名填进首页综合框代替。申请号填表前去掉校验点。
不做：Google Patents / 学术检索与跨库去重、PSS 登录站、权利要求全文/同族深挖、按附图视觉相似检索、按权利要求语义检索。单图/权要只生成公布站布尔式。
禁止被交底 Step 5 当查新引擎调用。
What ships with it
18 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- config.yaml 883 B
- prompts/derived_query.md 4.6 KB
- prompts/patent_search.md 5.3 KB
- README.md 1.8 KB
- references/patent_type_search.yaml 961 B
- tests/__init__.py 24 B runs code
- tests/test_cnipa_search.py 27 KB runs code
- tests/test_emit_search_report.py 4.7 KB runs code
- tools/browser.py 5.0 KB runs code
- tools/cnipa_crawler.py 34 KB runs code
- tools/cnipa_parse.py 21 KB runs code
- tools/cnipa_search.py 14 KB runs code
- tools/derived_query.py 3.0 KB runs code
- tools/emit_search_report.py 12 KB runs code
- tools/patent_type.py 9.1 KB runs code
- tools/requirements-cnipa.txt 230 B
- tools/search_config.py 2.4 KB runs code
- tools/stdio_utf8.py 1.4 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago Changed · +3 lines · +16 tokens per session dc1019c6d780
- 8d ago First seen · 41 lines · 30 tokens per session scan A 49b621bbc3aa
patent-search is a skill published in the GitHub repository handsomestWei/patent-disclosure-skill (9,078 stars, last pushed yesterday), licensed MIT. It adds 46 tokens to every session and 840 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
clawhub-content-rights-correspondence
Use when drafting, sending, or preserving email correspondence for an existing ClawHub content rights case.
compliance-checker
Check affiliate content for FTC compliance and platform rules. Triggers on: "check my content for compliance", "FTC disclosure check", "is this legal", "review for compliance", "check affiliate disclosure", "am I FTC compliant", "audit my content", "compliance review", "legal check", "platform rules check", "check…
claim-analysis
Deep claim scope analysis — decompose independent/dependent claims and assess protection scope breadth. Budget: 30 patent families, 30 claim parses, 20 web searches.
competitive-intelligence
Analyze competitor IP portfolios — comparative patent portfolio reports with strategy inference. Budget: 120 patent families, 15 claim parses, 40 web searches.
appropriateness-bounding
Establish acceptability standards through RAND/UCLA Appropriateness Method or Consensus Conference protocols.
knowledge-base-management
A lifecycle system for managing an Obsidian knowledge base, which is a folder of linked notes. It organizes raw material, AI-maintained wiki pages, and generated views into separate layers.