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 fnjialun/cnipa-patent-writer --skill cnipa-patent-writergit clone --depth 1 https://github.com/fnjialun/cnipa-patent-writerWrote 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/fnjialun/cnipa-patent-writer/cnipa-patent-writer)<a href="https://agentmods.dev/skills/fnjialun/cnipa-patent-writer/cnipa-patent-writer"><img src="https://agentmods.dev/badge/skills/fnjialun/cnipa-patent-writer/cnipa-patent-writer/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/fnjialun/cnipa-patent-writer/cnipa-patent-writer"><img src="https://agentmods.dev/badge/skills/fnjialun/cnipa-patent-writer/cnipa-patent-writer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00297 | $0.04138 |
| Opus 5 | $0.00148 | $0.02069 |
| Sonnet 5 | $0.00059 | $0.00828 |
| Haiku 4.5 | $0.00030 | $0.00414 |
Grade B, and why
cnipa-patent-writer scanned grade B with 1 finding 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 12d 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
外**还需系统二进制 `dot`**：`sudo apt-get install -y graphviz`，否则渲染报错。不画 graphviz 图则无需安装。 How it starts
The opening of the file, as written. The whole thing — 139 lines — stays where its author put it; the contents beside it link to each section on GitHub.
中国发明专利撰写（套用模板格式 / 无模板按 CNIPA 标准）
帮助撰写一篇新的发明专利申请文档（.docx）：套用页眉/页脚、分节、字体字号、缩进行距、分页与附图编排
等版式，但内容据待保护的技术方案重新写——模板只提供格式与行文风格，其技术内容一律不抄。
何时用 / 输入
- 有模板：用户给一份模板专利
.docx（作格式参照）+ 一个要写成专利的"技术方案/系统/项目"（代码库、设计 文档或口头描述）。→ 用克隆装配器build_patent.py，版式 100% 继承模板。 - 无模板：用户只给技术方案。→ 用兜底装配器
build_patent_cnipa.py，按 CNIPA 标准版式生成。 - 目标产物：一篇排版规范的
.docx（含配图），可直接提交/继续打磨。
依赖（缺则先装；装包注意 proxy）
- 必装：
python-docx、matplotlib、Pillow（pip install python-docx matplotlib Pillow）。 - 可选：
graphviz（用gen_figures_graphviz.py画复杂流程/架构图时）——除pip install graphviz（python 封装） 外还需系统二进制dot：sudo apt-get install -y graphviz，否则渲染报错。不画 graphviz 图则无需安装。 - CJK 字体：matplotlib/graphviz 画中文必须有中文字体（如 Noto Sans CJK / 宋体），否则中文渲染成豆腐块。
- 可视化校验工具（强烈建议）：
libreoffice-writer（提供soffice，docx→pdf）+poppler-utils（提供pdftoppm，pdf→png）+ 中文字体（如fonts-noto-cjk，或把 Windows 宋体/黑体拷进~/.fonts后fc-cache -f）。这些一般需sudo apt-get install；无 sudo 时请用户手动装。
核心原则（先读这几条，避免返工）
- 只借格式、绝不抄技术内容（最重要的红线）：模板只提供版式、行文风格、分节/页眉页脚/分页等通用 架构；它的技术方案、技术架构、层级划分、模块名、步骤编排、配图的数量与种类（哪怕看着很"通用"）一律 不得搬进新专利——因为换一个用户就是另一套技术，照搬既写错又涉抄袭。一切文本与配图都据用户提供的 材料/项目代码实际生成；模板里出现的具体技术（某种特定级联结构、某种模型分工、某些字段/JSON 示例等）只 能当"格式长什么样"的参照，不能当内容来源。动笔与画图前先自问：这句话/这张图是不是来自用户的真实方案？
- 结构校验 ≠ 视觉校验（最容易栽的坑）：python-docx 查到"样式名/字号全对"，用户在 Word/WPS 打开仍可能 错（页眉错、图越界、中文豆腐块、标题当正文）。完稿必须渲染成页面逐页看（见下"第7步"）。这是本技能 反复返工后总结出的第一教训。
- 权利要求一律宽泛，真实参数/产品名只进实施例：claims 用上位概念（如"预设阈值""通信通道""存储模块"）， 具体数值/角度/距离/倍率/型号/接口/消息通道（你这篇方案里的真实值）只在"实施例"里以"例如/优选"出现。
- 正文 solo 写、术语锁定：claim N ↔ 发明内容里对应的"进一步地" ↔ 实施例里的实现，三处用词必须一致。 先定一份术语表再写，全篇复用；不要并行分写正文（会术语漂移）。
- 创新点放进权利要求1：把最硬的发明点（解决的关键技术问题）写进独立权利要求1，不要藏在从属里。
工作流
1. 学习格式基准（务必先做）
读 references/docx-format.md（克隆复刻法）与 references/cnipa-format-spec.md（CNIPA 精确版式）。然后：
- 有模板：跑
python scripts/inspect_template.py 模板.docx一次扒出全部版式——分节数与分节符、各节页眉 文字、页脚是否有 PAGE 域页码及起始页、页面尺寸/页边距、各类段落（标题/正文/权项/发明名称/实施例子标题/ 图号）的字体字号对齐缩进段距、正文段长分布。逐项对齐（尤其确认 4 节 4 页眉：说明书摘要 / 权利要求书 / 说明书 / 说明书附图，以及附图节页码怎么处理）。这一步决定build_patent.py能否 1:1 复刻。 - 无模板：直接采用
cnipa-format-spec.md的标准参数（已内置进build_patent_cnipa.py），跳到写作。 - 注意：
inspect_template.py只读模板的版式，不读取也不复制模板的技术内容。
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.
- 12d ago First seen · 139 lines · 297 tokens per session scan B ec8d0aa96315
cnipa-patent-writer is a skill published in the GitHub repository fnjialun/cnipa-patent-writer (9 stars, last pushed 3mo ago), licensed MIT. It adds 297 tokens to every session and 4,138 once invoked, about $0.0015 per session on Opus 5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
deepread-pii
Redact PII from documents before sharing or sending to LLMs. 14 PII types (names, SSN, credit cards, medical records, etc.) detected with context-aware AI — not regex. Knows patient vs. doctor, personal vs. institutional. Black bar redaction on PDFs, scanned images, and text files. Free tier: 2,000 pages/month.
deepread-legal
Extract structured data from contracts, legal agreements, court filings, and compliance documents. Pre-built schemas for parties, clauses, dates, obligations. PII redaction for privilege review. 97%+ accuracy. Free 2,000 pages/month.
deepread-insurance
Extract structured data from insurance claims, policies, EOBs, and loss reports. Pre-built schemas for claimant info, coverage, damages, adjuster notes. PII redaction for compliant sharing. 97%+ accuracy. Free 2,000 pages/month.
deepread-id-documents
Extract structured data from passports, driver's licenses, and national ID cards for KYC/onboarding — name, document number, dates, nationality — as typed JSON. Vision AI reads scans and photos. Per-field confidence flags. PII redaction built in for compliant storage. Free 2,000 pages/month.
docx-template-translator
Adaptive conversion of LaTeX, PDF, or Markdown sources into a complete Word .docx that follows a user-supplied .docx template. Use when pandoc --reference-doc alone is not enough — for thesis, dissertation, report, or institutional Word formatting that needs cover pages, declarations, TOC, heading numbering, captions…
patent-application
A patent-application document workflow that turns supplied invention materials into claims, a description, an abstract, and black-and-white drawings. It can also produce Markdown and Word files.