ex-USSR Kinship Research: family history that shows its evidence

Tell the agent "my great-grandfather was exiled to Kazakhstan in the 1930s, help me find his parents in church records" and it opens a case: a folder of registers, a log of every search it runs, and a report where each parent-child link carries the records behind it. GitHub user DenisSergeevitch (Denis Shiryaev, per the MIT license notice) publishes it for Claude Code and Codex. Anthropic doesn't publish it. It answers in whatever language you write in, says the README. For how skills load, see what Claude skills are.

What it searches

The SKILL.md covers church and metric books, civil registration, revision lists, military, migration, deportation and repression files, cemetery records and scans nobody has indexed, across the Russian Empire, the USSR and the 15 republics. Regional guides cover places whose borders moved, such as Galicia, Bukovina and Bessarabia, and a separate guide covers the 1944 Chechen deportation. Before searching, the agent works out which jurisdiction a village belonged to on the date of the event and builds a table of name spellings, since one surname can appear in Russian and Polish forms.

How it decides a link is real

Each claim gets one of six statuses from the SKILL.md: hypothesis, single-source, corroborated, strongly-corroborated, contradicted or unresolved. Percentages of certainty are refused. Two people with the same name, age and village stay two people until a record joins them, and a copy or index of a register doesn't count as a second source. A family story stays a family story; it isn't promoted to documentary fact. An audit script checks the registers before the final report is written.

If an archive has to be asked, the agent drafts the letter for the institution's official channel and waits for your go-ahead before sending.

How to install it

With Node.js, one command:

npx skills add https://github.com/DenisSergeevitch/ex-ussr-kinship-research

Or clone it into Claude Code's skills folder (for Codex, use ~/.codex/skills/):

git clone https://github.com/DenisSergeevitch/ex-ussr-kinship-research ~/.claude/skills/ex-ussr-kinship-research

Most scripts need only Python 3. PDF export needs a Chromium-based browser; the default report is a standalone HTML file.

Example request

Use ex-ussr-kinship-research. My [relative, e.g. great-grandmother] [full name], born about [year] in [village, district, region]. Family memory says [what the family remembers]. I have [documents or scans you hold]. Find her parents, and keep the case folder in [path outside the skill folder].

You get the case registers, a search log, a report.html with every link and its status, and on request a GEDCOM file to import into family-tree software.

Limits

• Not for finding living people. The README says it refuses current addresses, phones or accounts of anyone alive.
• No getting past barriers. No CAPTCHA solvers, paywall or geo-block workarounds, or leaked databases. Logins to sites such as FamilySearch are yours to do.
• A missing record proves nothing. An empty search is logged as an empty search, not as proof the person never existed.
• Nothing leaves without your say. Documents, photos and DNA results aren't uploaded anywhere without your permission, and the README asks you to keep case files outside the skill's Git folder.

When to use it

Use it when every link in the line has to be backed by a record you can show a relative or an archive. For a quick sketch of a tree from family memory, it's more process than you need. For general long research jobs kept in files, see Planning with Files.