Klotzkette/claude-fuer-deutsches-recht
229 Plugins for German Law, One Giant Disclaimer
An experimental skill set that ports Anthropic’s legal-AI framework to German law practice, bundling 25,000 prompts, 229 plugins, and a small library of compliance warnings.
Collecting fresh signals — velocity needs a few days of history.
collecting data…
star history
What it does
This repository packages Anthropic’s open legal-AI skills into 229 German-law plugins and over 25,000 structured prompt files covering labor, corporate, insolvency, data-protection, and procedural law. It also includes 212 anonymized test cases—real PDFs, Excel sheets, and handwritten notes—to let lawyers stress-test workflows without exposing actual client secrets. Everything is framed around German legal methodology: statutory citations, case-law references with docket numbers, and the classic “Anspruchsgrundlagen” analysis.
The interesting bit
The project treats prompt engineering as legal infrastructure. There is a full legislative-drafting workshop (legistik-werkstatt) that spits out federal bill templates in official Arial house style, and a judgment-relation builder (urteilsbauer-relationsmacher) that formats court rulings to § 313 ZPO standards. Even the “unified mini prompts” are capped at 7,500 characters so they fit into generic chatbots when Claude Code is not an option.
Key highlights
- 229 plugins and 25,913
SKILL.mdfiles spanning everything from BGB general terms to NIS-2 cybersecurity compliance. - 212 deliberately messy, anonymized test cases with PDFs, Word docs, and Excel files for realistic plugin testing.
- A dedicated legislative-workshop plugin that renders draft bills, synopses, and cabinet memos in official German government layout.
- Explicit, repeated warnings that this is experimental, not legal advice, and that users must self-assess compliance with attorney-secrecy rules (§§ 203/204 StGB, § 43e BRAO), GDPR, and the EU AI Act.
Caveats
- The README repeatedly emphasizes this is an untested “technical playground,” not a certified legal product; users must verify current law and compliance themselves.
- All skills require individual calibration for specific law-firm workflows, citation styles, and internal governance rules before any real-world use.
Verdict
German lawyers, legal-tech developers, and policy drafters who want a head start on LLM-assisted workflows should dig in. Anyone looking for a plug-and-play, liability-free legal autopilot should look elsewhere—this project is very clear that it is not that.