메뉴
将已有 Zotero PDF 或历史 MinerU 输出归档为可追踪的 ResearchVault Fulltext Markdown：先复用已成功的 MinerU 链路，再写入统一 frontmatter、整理安全图片路径、保留页面映射并执行只读校验。此技能不做中文总结、不改写论文正文、不负责用户检索。
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/cheneternity/Zotero-Analytical-Workflow-Skills --skill zotero-fulltext-archiver명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
Orchestrate the verified ResearchVault single-paper ingest workflow from Zotero identity through Fulltext, Analytical Note, Knowledge decision, validation, and append-only logging. Use only when the user explicitly identifies paper(s) to add or asks to assess the ingest state of specified paper(s).
Maintain the persistent Research Knowledge Wiki in D:\ResearchVault\knowledge from existing analytical literature notes. Use when creating or incrementally updating cross-paper themes, concepts, methods, relationships, controversies, syntheses, the knowledge index or append-only log; or when linting traceability, duplicates, conflicts, and stale claims. Preserve note-first retrieval and use linked MinerU fulltexts only for targeted verification of precise claims.
ResearchVault 文献知识问题的默认检索技能。先从 02vault/_index 和 D:\ResearchVault\02vault 的 Analytical Notes 定位相关论文，再按需要定向进入对应 D:\ResearchVault\03fulltext MinerU Markdown，必要时回到 Zotero PDF 验证。若用户消息以‘基于当前 ResearchVault 项目文件检索’（或明确要求基于当前 ResearchVault 项目文件作答）开头，必须优先执行严格的项目文件检索后再回答。纯 Skill/Python/Git/文件整理/MinerU 调试等操作任务不自动触发文献检索。
SKILL.md 표시 중
| name | zotero-fulltext-archiver |
| description | 将已有 Zotero PDF 或历史 MinerU 输出归档为可追踪的 ResearchVault Fulltext Markdown：先复用已成功的 MinerU 链路，再写入统一 frontmatter、整理安全图片路径、保留页面映射并执行只读校验。此技能不做中文总结、不改写论文正文、不负责用户检索。 |
When a validated Zotero PDF must be run through MinerU on the current Windows CPU environment, use an ASCII-only working copy in a scoped temporary run directory. The Zotero attachment remains read-only. Record the source and working-copy SHA-256 values and do not invoke MinerU until they match.
Use the explicitly validated pipeline backend for this environment. This is
an environment-specific fallback, not a general claim that the hybrid backend
is unsuitable. Give a complete article a hard limit of at least 60 minutes.
A no-progress stop may be used only after 10--15 minutes during which stdout,
stderr, output files, and process CPU time have all remained inactive. Abort
for persistently available RAM below 2 GB only after a sustained observation,
and record the resource samples.
Every invocation must stream timestamped stdout and stderr to the run
directory, record stage transitions, and clean only the MinerU CLI tree and
new mineru.cli.fast_api descendants created by that invocation. Recheck for
those exact processes after cleanup; never terminate unrelated Python work.
Archive only after the raw Markdown gate passes: non-empty full-document front/middle/end samples, source identity, image-file/reference checks, and no missing image targets. Formal archiving may add schema frontmatter and rewrite image paths, but must otherwise preserve the extracted body verbatim.
Before invoking MinerU, inspect the formal 03fulltext path by zotero_key.
If a formal Fulltext already has matching zotero_key/pdf_key, valid
frontmatter, resolved images, and a non-empty article body, reuse it and run
targeted validation. A Note-template repair is not a reason to rerun MinerU.
If page_mapping is unknown, retain that value; the Analytical Writer may
still verify quotation pages directly against the read-only Original PDF.
执行：Zotero PDF → MinerU → Fulltext Markdown → 图片整理 → metadata → Note 关联 → validation。
不执行中文翻译、分析笔记写作、批量检索或 Zotero 数据库改写。
不要重新安装 MinerU。先搜索 D:\ResearchVault、D:\research 和相关项目中的 MinerU、mineru、magic-pdf、批处理脚本、配置和历史输出。
当前已发现的可复用调用链是：
D:\research\mineru_batch_runner.py → D:\MinerU\.venv\Scripts\mineru.exe → 系统临时输出目录 → D:\ResearchVault_Archive\mineru-staging\。批量输出只能作为外部暂存；逐篇补齐 frontmatter、图片路径和 Note 关联并验证后，才复制到 D:\ResearchVault\03fulltext\<collection>\。
历史 MinerU_test 的页码辅助文件已归档到 D:\ResearchVault_Archive\2026-08-10\MinerU_test\；如需核验历史页码映射可定向读取，但不足以证明所有论文都可可靠映射。
正式全文：
D:\ResearchVault\03fulltext\<collection>\<zotero_key>.md
D:\ResearchVault\03fulltext\<collection>\images\<zotero_key>\<image-file>
旧的 MinerU_batch 已归档到 D:\ResearchVault_Archive\2026-08-10\MinerU_batch\，不作为运行时输入或正式全文检索目录。当前 Vault 的分析笔记仍位于 论文库/ 时，不移动它们；仅在全文 frontmatter 中写准确的 note_path。
若 MinerU_batch 已有与 zotero_key 唯一对应的 Markdown 和图片：
03fulltext/<collection>/<zotero_key>.md；分析笔记中的 Obsidian 链接仍使用 fulltext/<collection>/<zotero_key>。images/<zotero_key>/，不得使用完整论文标题作为目录名。。若没有可复用结果，才调用已确认的 MinerU 可执行文件处理单篇 PDF；不得批量重跑整个库。
每个正式全文顶部至少包含：
---
type: literature-fulltext
title: "..."
zotero_key: "Q22PFLNV"
pdf_key: "4RMSR7ZR"
doi: "..."
collection: "创新经济地理"
note_path: "论文库/创新经济地理/论文标题.md"
fulltext_path: "03fulltext/创新经济地理/Q22PFLNV.md"
zotero_item: "zotero://select/library/items/Q22PFLNV"
zotero_pdf: "zotero://open-pdf/library/items/4RMSR7ZR"
source_type: mineru
page_mapping: unknown
---
Vault 内部路径统一使用 /。缺失的 DOI 可留空，但不得伪造。
content_list.json/middle.json 等信息与真实 PDF 通过单篇测试可靠对应时，才写 page_mapping: reliable 或 <!-- pdf_page: N -->。page_mapping: unknown，不要猜 page。归档完成后：
fulltext_path，并可增加 [[fulltext/<collection>/<zotero_key>]] 入口；不因全文归档重写整篇笔记，模板化重排由 zotero-analytical-writer 单独负责。note_path，确认双方 zotero_key、pdf_key 一致。D:\research\zotero_batch\validate_research_vault_literature_links.py，只报告，不自动删除。