CNIPA 商标证据归档

来源目录：cnipa-trademark-evidence-archive/SKILL.md

Goal

Create local evidence folders for Chinese trademark records from the CNIPA 商标网上检索系统, with separate PDFs for the trademark detail page and trademark process page.

Inputs

Confirm or infer:

• applicant_name: trademark applicant/company name.
• trademark_numbers: one or more application/registration numbers when known.
• output_root: default to ~/Downloads/<applicant_name>.
• pages: default to 商标详情 and 商标流程.

Do not store credentials, OTPs, cookies, or private account/session details in the skill output or summaries. If the site requires login or verification, ask the user to complete it in their browser.

Workflow

1. Open the CNIPA trademark search site in Chrome:

   • https://wcjs.sbj.cnipa.gov.cn/home
   • Use the user's existing browser session when login state or verification matters.

2. Enter comprehensive search:

   • Click 商标综合查询 / 商标综合检索.
   • Search by applicant_name if collecting all marks for a company.
   • Search by trademark_number if the task is for a known mark.

3. Open each target record:

   • From the result list, open the detail page for the target trademark.
   • Verify the visible 申请/注册号, 国际分类, and 申请人名称（中文）.
   • Record only the workflow shape in responses; do not expose sensitive client details beyond what the user needs.

4. Export the trademark detail page:

   • Stay on or switch to 商标详情.
   • Use the page 打印 button if available; otherwise use Chrome print.
   • Enable Chrome's built-in Headers and footers / 页眉和页脚; the exported PDF must show date, page title, URL, page number, and total page count.
   • In Chrome print preview, choose 保存 / Save as PDF.
   • Create/select folder <output_root>/商标-<trademark_number>/.
   • Save as 商标详细内容-1.pdf unless the user asks for clearer names such as 商标详情.pdf.

5. Export the trademark process page:

   • Click the 商标流程 tab/link or the 商标流程 row's 点击查看 link.
   • Confirm the process view shows columns such as 申请/注册号, 业务名称, 环节名称, 结论, and 日期.
   • Enable Chrome's built-in Headers and footers / 页眉和页脚; the exported PDF must show date, page title, URL, page number, and total page count.
   • Print and save to the same folder as 商标详细内容-2.pdf unless the user asks for clearer names such as 商标流程.pdf.

6. Repeat for all target marks:

   • Return to the result list or next detail page.
   • Repeat detail and process exports per trademark number.
   • Keep each trademark in a separate 商标-<trademark_number> folder under the applicant folder.

7. Validate:

   • Confirm each 商标-<trademark_number> folder exists.
   • Confirm each folder contains both expected PDFs.
   • Confirm each PDF contains Chrome header/footer information: date, title, URL, page number, and total pages (for example 1/1 or 1/2, 2/2).
   • If a page fails to save, is unavailable, or is blocked by verification, note the issue instead of inventing a result.
   • Report the final root folder and list the trademark folders/files created.

Scripted Chrome Path

When the user asks to process by script, use the Chrome plugin and the helper script at scripts/export-cnipa-trademark-pdfs.mjs.

Prerequisites:

• The user has an open Chrome tab on the CNIPA 商标检索结果 page.
• The result page belongs to the intended applicant or target search.
• The Chrome plugin session can access the user's existing CNIPA login/session state.

The script:

• Claims the current 商标检索结果 tab.
• Reads sessionStorage.params.dataList when available; if not, builds a minimal record list from the visible result table.
• Opens detail pages by real-clicking the visible application/registration number link on the result page. Do not navigate to /detail by only writing sessionStorage and assigning location.href; that path can leave Chrome CDP in a paused document response state and can make the process tab load only a header.
• Exports the detail page with Chrome DevTools Page.printToPDF, with Chrome's built-in headers and footers enabled. Do not provide custom headerTemplate or footerTemplate unless the user explicitly asks for a non-Chrome layout.
• Clicks the official visible 商标流程 tab/link, then exports the process page.
• Treats the process PDF as usable only after confirming actual flow rows for the target application/registration number. If the process page shows only headers such as 业务名称 and 环节名称, do not report it as a complete process archive.

Example Node REPL invocation:

const { exportCnipaTrademarkPdfs } = await import("/Users/licheng/.codex/skills/cnipa-trademark-evidence-archive/scripts/export-cnipa-trademark-pdfs.mjs");
const result = await exportCnipaTrademarkPdfs({
  applicantName: "深圳潜行创新科技有限公司",
  trademarkNumbers: ["87649562", "86982451"],
  outputRoot: "/Users/licheng/Downloads/深圳潜行创新科技有限公司"
});
nodeRepl.write(JSON.stringify(result, null, 2));

Practical Notes

• The site states that data is for reference and may lag; preserve the page exactly as displayed for evidence management.
• Prefer the Chrome plugin and semantic UI targets (商标综合查询, 查询, result-row application number links, 商标流程) over coordinates. Do not use Computer Use unless the user explicitly asks for it.
• If using the scripted Chrome path, prefer sessionStorage.params.dataList over scraping table text when it exists because it contains the exact record object the Vue app passes into /detail; fall back to table rows when the current result page stores only query params.
• The save panel may default to 商标详细内容; append or replace the suffix deliberately before saving.
• If the user prefers clearer output names, use 商标详情.pdf and 商标流程.pdf; otherwise follow the demonstrated convention 商标详细内容-1.pdf and 商标详细内容-2.pdf.
• For legal due diligence, do not edit the exported PDF contents after saving.

Lessons From Failed Runs

• Standalone Playwright may be blocked by CNIPA with 403 Forbidden; use the user's existing Chrome session through the Chrome plugin.
• Do not rely on old 商标检索结果 tabs. Verify the result page contains the intended applicant name before exporting; if it does not, rerun 商标综合检索.
• Programmatic value assignment may not trigger Vue state. If a search button click does not open results, fill the applicant field with a real browser input action and click 查询.
• Directly setting sessionStorage.params and navigating to /detail can produce stuck Chrome states and incomplete process pages. The robust route is: result page -> real click target application number -> new detail tab -> print detail -> click 商标流程 -> print flow.
• Chrome built-in headers and footers are produced by displayHeaderFooter: true. They include date, title, URL, and page number/total pages. This is a required output condition for trademark evidence PDFs; avoid custom templates for ordinary evidence exports.
• Validate every exported PDF with text extraction: detail PDFs must contain the target application number and applicant; process PDFs must contain 业务名称, 环节名称, and at least one flow row for the same target number.