Back to Details

autoresearch-genealogy

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

autoresearch-genealogy

autoresearch-genealogy

Skill by ara.so — Daily 2026 Skills collection.
A structured system of autoresearch prompts, Obsidian vault templates, archive guides, and methodology references for AI-assisted genealogy research. Built for Claude Code's autonomous research loops, adaptable to any AI tool or manual workflow.
由ara.so开发的Skill——2026每日技能合集。
这是一套用于AI辅助家谱研究的结构化自主研究提示词系统、Obsidian vault模板、档案指南和方法学参考资料。专为Claude Code的自主研究循环打造，也可适配任何AI工具或手动工作流。

What This Project Does

本项目功能

  • Provides 12 Claude Code 
    /autoresearch
    prompts that autonomously search the web, update your vault, and self-verify results
  • Supplies a complete 19-file Obsidian vault starter kit with YAML frontmatter and markdown templates
  • Includes 24 country/region-specific archive guides (Europe, Americas, Oceania, Jewish genealogy)
  • Offers 9 methodology reference documents covering confidence tiers, DNA guardrails, naming conventions, and source hierarchy
  • Defines 7 step-by-step workflows for OCR pipelines, oral history, discrepancy resolution, and phase planning
  • 提供12个Claude Code 
    /autoresearch
    提示词，可自主搜索网页、更新你的库并自我验证结果
  • 提供包含YAML前置元数据和Markdown模板的完整19文件Obsidian库启动套件
  • 包含24份针对特定国家/地区的档案指南（欧洲、美洲、大洋洲、犹太家谱）
  • 提供9份方法学参考文档，涵盖可信度层级、DNA约束、命名规范和来源优先级
  • 定义7个分步工作流，用于OCR处理、口述历史、差异解决和阶段规划

Installation

安装

bash
undefined
bash
undefined

Clone the repository

克隆仓库

git clone https://github.com/mattprusak/autoresearch-genealogy.git cd autoresearch-genealogy
git clone https://github.com/mattprusak/autoresearch-genealogy.git cd autoresearch-genealogy

Copy vault template into your Obsidian vault

将库模板复制到你的Obsidian库中

cp -r vault-template/ ~/path/to/your/ObsidianVault/genealogy/
cp -r vault-template/ ~/path/to/your/ObsidianVault/genealogy/

Or copy to any markdown editor folder

或复制到任何Markdown编辑器文件夹

cp -r vault-template/ ~/Documents/my-genealogy/

No package manager or build step required — this is a pure markdown/prompt project.

---
cp -r vault-template/ ~/Documents/my-genealogy/

无需包管理器或构建步骤——这是一个纯Markdown/提示词项目。

---

Project Structure

项目结构

autoresearch-genealogy/
├── prompts/              # 12 autoresearch prompt files for Claude Code
├── vault-template/       # 19-file Obsidian vault starter kit
│   ├── Family_Tree.md
│   ├── Research_Log.md
│   ├── Open_Questions.md
│   ├── templates/        # Person, certificate, postcard, region, etc.
│   └── ...
├── archives/             # 24 country/region research guides
├── reference/            # 9 methodology documents
├── workflows/            # 7 step-by-step process guides
└── examples/             # 6 anonymized worked examples
autoresearch-genealogy/
├── prompts/              # 12个用于Claude Code的自主研究提示词文件
├── vault-template/       # 19文件Obsidian库启动套件
│   ├── Family_Tree.md
│   ├── Research_Log.md
│   ├── Open_Questions.md
│   ├── templates/        # 人物、证书、明信片、地区等模板
│   └── ...
├── archives/             # 24份国家/地区研究指南
├── reference/            # 9份方法学文档
├── workflows/            # 7份分步流程指南
└── examples/             # 6份匿名示例

Quick Start Workflow

快速开始工作流

Step 1: Seed your family tree

步骤1：初始化家谱

Open 
vault-template/Family_Tree.md
and fill in what you already know, starting with yourself and working backward:
markdown
---
title: Family Tree
last_updated: 2026-03-19
generations_documented: 3
lines_active: 2
---
打开
vault-template/Family_Tree.md
，填入你已知的信息，从自己开始向上追溯：
markdown
---
title: Family Tree
last_updated: 2026-03-19
generations_documented: 3
lines_active: 2
---

Family Tree

家谱

Generation 1 (Self)

第一代（本人）

  • Name: Jane Smith (b. 1985, Chicago, IL)
  • 姓名: 简·史密斯（生于1985年，伊利诺伊州芝加哥）

Generation 2 (Parents)

第二代（父母）

  • Father: John Smith (b. 1955, Detroit, MI)
  • Mother: Mary O'Brien (b. 1958, Boston, MA)
  • 父亲: 约翰·史密斯（生于1955年，密歇根州底特律）
  • 母亲: 玛丽·奥布莱恩（生于1958年，马萨诸塞州波士顿）

Generation 3 (Grandparents)

第三代（祖父母）

  • Paternal Grandfather: Robert Smith (b. ~1920, unknown)
  • Paternal Grandmother: Helen Kowalski (b. ~1925, Poland?)
undefined
  • 父系祖父: 罗伯特·史密斯（生于约1920年，地点未知）
  • 父系祖母: 海伦·科瓦尔斯基（生于约1925年，波兰？）
undefined

Step 2: Scan physical documents

步骤2：扫描实体文档

Photograph or scan certificates, letters, postcards. Use the OCR workflow:
See: workflows/ocr-pipeline.md
拍摄或扫描证书、信件、明信片。使用OCR工作流：
参考：workflows/ocr-pipeline.md

Step 3: Run autoresearch prompts in Claude Code

步骤3：在Claude Code中运行自主研究提示词

/autoresearch prompts/01-tree-expansion.md
/autoresearch prompts/01-tree-expansion.md

Step 4: Audit and verify

步骤4：审核与验证

/autoresearch prompts/02-cross-reference-audit.md
/autoresearch prompts/02-cross-reference-audit.md

Autoresearch Prompts — Reference

自主研究提示词参考

Each prompt in 
prompts/
follows this structure:
markdown
undefined
prompts/
中的每个提示词遵循以下结构：
markdown
undefined

Goal

目标

[What this iteration should accomplish]
[本次迭代需完成的任务]

Metric

衡量标准

[Measurable success condition — e.g., "increase sourced person files from N to N+10"]
[可量化的成功条件——例如："将已溯源的人物文件数量从N增加到N+10"]

Direction

操作指引

[Step-by-step instructions for the AI]
[给AI的分步指令]

Verify

验证

[Cross-check to run after each iteration]
[每次迭代后需执行的交叉检查]

Guard Rails

约束规则

[What NOT to do — prevent hallucination, preserve source rigor]
[禁止操作——防止幻觉，保证来源严谨性]

Iterations

迭代次数

[How many loops to run before stopping for human review]
[在等待人工审核前需运行的循环次数]

Protocol

协议

[Output format, file naming, YAML fields to populate]
undefined
[输出格式、文件命名、需填充的YAML字段]
undefined

All 12 Prompts

全部12个提示词

FilePurpose
01-tree-expansion.md
Push every branch back using web research
02-cross-reference-audit.md
Find and fix discrepancies between tree and sources
03-findagrave-sweep.md
Locate Find a Grave memorials for deceased ancestors
04-gedcom-completeness.md
Sync GEDCOM file with vault data
05-source-citation-audit.md
Verify every person has ≥2 independent sources
06-unresolved-persons.md
Identify and resolve unnamed people in documents
07-timeline-gap-analysis.md
Find life events where records should exist but don't
08-open-question-resolution.md
Systematically attack every open research question
09-bygdebok-extraction.md
Extract data from digitized local history books
10-colonial-records-search.md
Search pre-1800 colonial American records
11-immigration-search.md
Locate passenger manifests and naturalization records
12-dna-chromosome-analysis.md
Analyze per-chromosome ancestry data
文件用途
01-tree-expansion.md
通过网络研究拓展家谱的每一条分支
02-cross-reference-audit.md
发现并修复家谱与来源之间的差异
03-findagrave-sweep.md
为已故祖先查找Find a Grave纪念页面
04-gedcom-completeness.md
同步GEDCOM文件与库数据
05-source-citation-audit.md
验证每个人物都有≥2个独立来源
06-unresolved-persons.md
识别并解决文档中的无名人物
07-timeline-gap-analysis.md
查找应该存在记录但缺失的人生事件
08-open-question-resolution.md
系统性解决所有未完成的研究问题
09-bygdebok-extraction.md
从数字化地方史书籍中提取数据
10-colonial-records-search.md
搜索1800年前的美洲殖民地记录
11-immigration-search.md
查找乘客清单和入籍记录
12-dna-chromosome-analysis.md
分析染色体层面的祖先数据

Running a prompt in Claude Code

在Claude Code中运行提示词

bash
undefined
bash
undefined

In Claude Code terminal or chat:

在Claude Code终端或聊天中：

/autoresearch prompts/08-open-question-resolution.md
/autoresearch prompts/08-open-question-resolution.md

With a specific vault path context:

指定库路径上下文：

/autoresearch prompts/03-findagrave-sweep.md --context vault-template/Family_Tree.md

---
/autoresearch prompts/03-findagrave-sweep.md --context vault-template/Family_Tree.md

---

Vault Template Files

库模板文件

Person file template (
vault-template/templates/person.md
)

人物文件模板（
vault-template/templates/person.md
）

markdown
---
full_name: ""
birth_date: ""
birth_place: ""
death_date: ""
death_place: ""
father: ""
mother: ""
spouse: ""
children: []
confidence: "Moderate Signal"  # Strong Signal | Moderate Signal | Speculative
sources: []
open_questions: []
last_updated: ""
---
markdown
---
full_name: ""
birth_date: ""
birth_place: ""
death_date: ""
death_place: ""
father: ""
mother: ""
spouse: ""
children: []
confidence: "Moderate Signal"  # Strong Signal | Moderate Signal | Speculative
sources: []
open_questions: []
last_updated: ""
---

[Full Name]

[全名]

Life Events

人生事件

EventDatePlaceSource
Birth
Marriage
Death
事件日期地点来源
出生
婚姻
逝世

Sources

来源

  1. [Source 1 — type, repository, date accessed]
  2. [Source 2 — type, repository, date accessed]
  1. [来源1 — 类型、存储库、访问日期]
  2. [来源2 — 类型、存储库、访问日期]

Open Questions

未解决问题

  • Question 1
  • Question 2
  • 问题1
  • 问题2

Notes

备注

[Narrative summary, naming variant notes, contextual history]
undefined
[叙述摘要、姓名变体说明、背景历史]
undefined

Certificate transcription template (
vault-template/templates/certificate.md
)

证书转录模板（
vault-template/templates/certificate.md
）

markdown
---
document_type: ""        # birth | death | marriage | baptism
document_date: ""
repository: ""
file_reference: ""
transcribed_by: ""
transcription_date: ""
confidence: ""
---
markdown
---
document_type: ""        # birth | death | marriage | baptism
document_date: ""
repository: ""
file_reference: ""
transcribed_by: ""
transcription_date: ""
confidence: ""
---

Certificate: [Type] — [Name] — [Year]

证书：[类型] — [姓名] — [年份]

Transcription

转录内容

[Verbatim transcription of the document]
[文档的逐字转录]

Key Data Extracted

提取的关键数据

  • Subject:
  • Date:
  • Place:
  • Witnesses/Informants:
  • Officiant:
  • 主体:
  • 日期:
  • 地点:
  • 证人/申报人:
  • 主礼人:

Discrepancies

差异

[Note any conflicts with other sources]
[记录与其他来源的冲突]

Image

图片

![[filename.jpg]]
undefined
![[filename.jpg]]
undefined

Research log entry pattern (
vault-template/Research_Log.md
)

研究日志条目格式（
vault-template/Research_Log.md
）

markdown
undefined
markdown
undefined

2026-03-19 — Tree Expansion Session

2026-03-19 — 家谱拓展会话

Prompt run: 01-tree-expansion.md
Iterations: 5
Metric start: 42 sourced person files
Metric end: 51 sourced person files
运行的提示词: 01-tree-expansion.md
迭代次数: 5
初始指标: 42个已溯源的人物文件
最终指标: 51个已溯源的人物文件

Searches Performed

执行的搜索

  • FamilySearch: "Kowalski Poznan 1880–1920" — 3 results, 2 useful
  • Ancestry: "Smith Michigan census 1920" — found Robert Smith (b. 1919)
  • FindAGrave: "Helen Kowalski Detroit" — memorial #12345678
  • FamilySearch: "Kowalski Poznan 1880–1920" — 3条结果，2条有用
  • Ancestry: "Smith Michigan census 1920" — 找到罗伯特·史密斯（生于1919年）
  • FindAGrave: "Helen Kowalski Detroit" — 纪念页面#12345678

Negative Results (Important)

负面结果（重要）

  • No passenger manifest found for Stanislaw Kowalski, searched 1890–1910
  • No church records found for O'Brien line in Cork pre-1850
  • 未找到Stanislaw Kowalski的乘客清单，搜索范围1890–1910
  • 未找到科克地区O'Brien家族1850年前的教会记录

New Open Questions

新的未解决问题

  • Was Robert Smith born in Michigan or Ohio? 1920 census says MI, 1930 says OH.

---
  • 罗伯特·史密斯生于密歇根州还是俄亥俄州？1920年人口普查显示密歇根，1930年显示俄亥俄。

---

Confidence Tier System

可信度层级体系

From 
reference/confidence-tiers.md
:
Strong Signal    — Two or more independent primary sources agree
Moderate Signal  — One primary source, or two secondary sources agree
Speculative      — Logical inference, DNA suggestion, or single secondary source
Apply confidence in every person file YAML:
markdown
---
confidence: "Moderate Signal"
---
来自
reference/confidence-tiers.md
：
强信号    — 两个或更多独立的原始来源一致
中等信号  — 一个原始来源，或两个次级来源一致
推测性    — 逻辑推断、DNA提示，或单个次级来源
在每个人物文件的YAML中设置可信度：
markdown
---
confidence: "Moderate Signal"
---

Archive Guides — Key Countries

档案指南——重点国家

Each guide in 
archives/
covers:
  • Where to find records (free vs paid)
  • What AI tools can access directly vs what requires browser
  • Record types available by era
archives/
├── ireland.md
├── england-wales.md
├── scotland.md
├── norway.md
├── sweden.md
├── poland.md
├── germany.md
├── italy.md
├── france.md
├── spain-portugal.md
├── netherlands.md
├── austria.md
├── hungary.md
├── russia-ukraine.md
├── usa-colonial.md
├── usa-immigration.md
├── usa-census.md
├── usa-vital-records.md
├── african-american.md
├── canada.md
├── mexico-latin-america.md
├── australia-new-zealand.md
├── jewish-genealogy.md
└── ...
Example usage in a prompt:
markdown
undefined
archives/
中的每份指南涵盖：
  • 记录查找地点（免费 vs 付费）
  • AI工具可直接访问的内容 vs 需要浏览器的内容
  • 不同时代可用的记录类型
archives/
├── ireland.md
├── england-wales.md
├── scotland.md
├── norway.md
├── sweden.md
├── poland.md
├── germany.md
├── italy.md
├── france.md
├── spain-portugal.md
├── netherlands.md
├── austria.md
├── hungary.md
├── russia-ukraine.md
├── usa-colonial.md
├── usa-immigration.md
├── usa-census.md
├── usa-vital-records.md
├── african-american.md
├── canada.md
├── mexico-latin-america.md
├── australia-new-zealand.md
├── jewish-genealogy.md
└── ...
提示词中的使用示例：
markdown
undefined

In prompts/09-bygdebok-extraction.md

在prompts/09-bygdebok-extraction.md中

Direction

操作指引

Consult archives/norway.md for Digitalarkivet access patterns. Search Bygdebok collections for the Rogaland region, 1750–1900.

---
参考archives/norway.md获取Digitalarkivet的访问模式。 搜索Rogaland地区1750–1900年的Bygdebok合集。

---

Common Patterns

通用模式

Pattern 1: New ancestor intake

模式1：新祖先录入

When a new ancestor is found during research:
markdown
1. Create person file from vault-template/templates/person.md
2. Set confidence based on source count
3. Add to Family_Tree.md under correct generation
4. Log the discovery in Research_Log.md
5. Add unresolved questions to Open_Questions.md
6. Run 02-cross-reference-audit.md to check for conflicts
当研究中发现新祖先时：
markdown
1. 从vault-template/templates/person.md创建人物文件
2. 根据来源数量设置可信度
3. 将其添加到Family_Tree.md的对应世代下
4. 在Research_Log.md中记录发现
5. 将未解决问题添加到Open_Questions.md
6. 运行02-cross-reference-audit.md检查冲突

Pattern 2: Resolving a date discrepancy

模式2：解决日期差异

markdown
undefined
markdown
undefined

Open_Questions.md entry

Open_Questions.md条目

Q-042: Robert Smith birth state conflict

Q-042: 罗伯特·史密斯出生地冲突

  • 1920 census: born Michigan
  • 1930 census: born Ohio
  • Status: Unresolved
  • Next step: Run 07-timeline-gap-analysis.md targeting Robert Smith

Then in Claude Code:
/autoresearch prompts/07-timeline-gap-analysis.md
  • 1920年人口普查：密歇根州
  • 1930年人口普查：俄亥俄州
  • 状态：未解决
  • 下一步：运行07-timeline-gap-analysis.md针对罗伯特·史密斯

然后在Claude Code中：
/autoresearch prompts/07-timeline-gap-analysis.md

Focus: Robert Smith, b. ~1919, discrepancy Q-042

重点：罗伯特·史密斯，生于约1919年，差异Q-042

undefined
undefined

Pattern 3: DNA-to-genealogy mapping

模式3：DNA与家谱映射

markdown
undefined
markdown
undefined

In vault-template/Genetic_Profile.md

在vault-template/Genetic_Profile.md中

test_company: AncestryDNA test_date: 2024-11-01 ethnicity_summary:
  • region: Eastern Europe percentage: 38
  • region: Ireland/Scotland percentage: 31
test_company: AncestryDNA test_date: 2024-11-01 ethnicity_summary:
  • region: Eastern Europe percentage: 38
  • region: Ireland/Scotland percentage: 31

Then run:

然后运行：

/autoresearch prompts/12-dna-chromosome-analysis.md
undefined
/autoresearch prompts/12-dna-chromosome-analysis.md
undefined

Pattern 4: Immigration research loop

模式4：移民研究循环

bash
undefined
bash
undefined

Run immigration search prompt

运行移民搜索提示词

/autoresearch prompts/11-immigration-search.md
/autoresearch prompts/11-immigration-search.md

Prompt will:

提示词将：

1. Pull all foreign-born ancestors from Family_Tree.md

1. 从Family_Tree.md提取所有外国出生的祖先

2. Search passenger manifests (Ellis Island, Ancestry, FamilySearch)

2. 搜索乘客清单（埃利斯岛、Ancestry、FamilySearch）

3. Search naturalization records (NARA, Ancestry)

3. 搜索入籍记录（NARA、Ancestry）

4. Update person files with ship name, arrival date, port

4. 更新人物文件的船名、抵达日期、港口

5. Log negative results for each unresolved ancestor

5. 记录每个未解决祖先的负面结果


---

---

Reference Documents

参考文档

FileContents
reference/confidence-tiers.md
Strong / Moderate / Speculative definitions
reference/source-hierarchy.md
Primary vs secondary vs derivative sources
reference/dna-guardrails.md
What DNA can and cannot prove; centimorgan thresholds
reference/naming-conventions.md
Patronymics, farm names, Polish przydomki
reference/gedcom-guide.md
GEDCOM field reference and export instructions
reference/common-pitfalls.md
AI hallucination patterns in genealogy, date traps
reference/glossary.md
Record type definitions, Latin terms, abbreviations
reference/ai-capabilities.md
What AI can access directly vs what requires human
reference/case-for-autoresearch.md
Methodology rationale
文件内容
reference/confidence-tiers.md
强/中等/推测性定义
reference/source-hierarchy.md
原始 vs 次级 vs 衍生来源
reference/dna-guardrails.md
DNA能证明和不能证明的内容；厘摩根阈值
reference/naming-conventions.md
父名制、农场名、波兰语przydomki
reference/gedcom-guide.md
GEDCOM字段参考和导出说明
reference/common-pitfalls.md
家谱研究中的AI幻觉模式、日期陷阱
reference/glossary.md
记录类型定义、拉丁语术语、缩写
reference/ai-capabilities.md
AI可直接访问的内容 vs 需要人工处理的内容
reference/case-for-autoresearch.md
方法学原理

Troubleshooting

故障排除

AI is inventing sources

AI虚构来源

Set guard rails explicitly in your prompt session:
markdown
undefined
在提示词会话中明确设置约束规则：
markdown
undefined

Guard Rails (add to any prompt)

约束规则（添加到任何提示词）

  • Do NOT fabricate census record URLs or Ancestry record IDs
  • If a source cannot be directly linked, mark as "reported" not "confirmed"
  • All new claims require a real URL or repository reference
  • When uncertain, add to Open_Questions.md — do not guess
undefined
  • 不得编造人口普查记录URL或Ancestry记录ID
  • 如果无法直接链接来源，标记为"报告"而非"确认"
  • 所有新主张需要真实的URL或存储库参考
  • 不确定时，添加到Open_Questions.md——不要猜测
undefined

Vault files getting out of sync with GEDCOM

库文件与GEDCOM不同步

Run the completeness audit:
/autoresearch prompts/04-gedcom-completeness.md
This compares every person in your GEDCOM against vault person files and flags mismatches.
运行完整性审核：
/autoresearch prompts/04-gedcom-completeness.md
这会将GEDCOM中的每个人物与库中的人物文件进行比较，并标记不匹配项。

Name variants causing duplicate person files

姓名变体导致重复人物文件

Check 
reference/naming-conventions.md
for your family's relevant region. Common traps:
  • Norwegian farm name changes (Haugen → Bakke on emigration)
  • Polish name Latinization in church records (Stanisław → Stanislaus)
  • Irish anglicization (Ó Briain → O'Brien → Bryan)
  • Spelling variation in census records ("Sakkarias" vs "Zacharias" — both valid)
Add aliases to person file YAML:
markdown
---
full_name: "Stanisław Kowalski"
name_variants:
  - "Stanislaus Kowalski"
  - "Stanley Kowalski"
  - "S. Kowalski"
---
查看
reference/naming-conventions.md
中与你的家族相关的地区。常见陷阱：
  • 挪威移民时的农场名变更（Haugen → Bakke）
  • 教会记录中波兰语姓名的拉丁化（Stanisław → Stanislaus）
  • 爱尔兰语姓名的英语化（Ó Briain → O'Brien → Bryan）
  • 人口普查记录中的拼写变体（"Sakkarias" vs "Zacharias" — 均有效）
在人物文件的YAML中添加别名：
markdown
---
full_name: "Stanisław Kowalski"
name_variants:
  - "Stanislaus Kowalski"
  - "Stanley Kowalski"
  - "S. Kowalski"
---

Autoresearch loop running too long

自主研究循环运行时间过长

Each prompt has an 
## Iterations
field. Set it explicitly:
markdown
undefined
每个提示词都有
## Iterations
字段。明确设置：
markdown
undefined

Iterations

迭代次数

Run 3 iterations maximum, then stop and output a summary for human review.
undefined
最多运行3次迭代，然后停止并输出摘要供人工审核。
undefined

OCR producing poor results on old documents

OCR对旧文档识别效果差

See 
workflows/ocr-pipeline.md
. General guidance:
  1. Photograph at 600 DPI minimum
  2. Use even, diffuse lighting — no flash
  3. Pre-process with a contrast adjustment before running OCR
  4. Use 
    vault-template/templates/transcription.md
    to record both the OCR output and your manual corrections side by side
参考
workflows/ocr-pipeline.md
。通用指南：
  1. 至少以600 DPI拍摄
  2. 使用均匀的漫射光——不要用闪光灯
  3. 运行OCR前先调整对比度
  4. 使用
    vault-template/templates/transcription.md
    同时记录OCR输出和手动修正内容

Contributing

贡献指南

To add a new archive guide or prompt:
  1. Follow the existing file structure and YAML frontmatter patterns
  2. Use placeholder names in all examples (no real family data)
  3. Open a PR with a brief description of what region or record type you've added
License: MIT
添加新的档案指南或提示词：
  1. 遵循现有文件结构和YAML前置元数据格式
  2. 示例中使用占位姓名（不要用真实家族数据）
  3. 提交PR并简要说明你添加的地区或记录类型
License: MIT