Skip to content

avSercanKoc/superlex-skills

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

⚖️ SuperLex Skills

The open-source agentic skill library for lawyers.

The open-source jurisdiction-aware legal engineering framework for AI agents, built on the principles of Compliance by Design.


What SuperLex Skills Solves

General-purpose AI agents are excellent at writing text, but in the legal domain, they are often "confidently wrong." Generic agents suffer from Legal AI Slop: uydurma (fabricated) article numbers, mixing of legal regimes (KVKK vs. GDPR), and an absence of professional guardrails.

SuperLex Skills transforms your AI agent from a creative writer into a disciplined Legal Associate. It enforces:

  • Strict Context Collection: No drafting without knowing the client's sector and legal status.
  • Jurisdiction Routing: Automatic selection of the correct legal framework (TR, EU, etc.).
  • Zero Fabrication: Mandatory verification of statutes against official gazettes.
  • Agentic Verification: A three-layer safety model that forces the agent to stop, flag risks, and offer alternatives before delivery.

The Philosophy: Legal Engineering

SuperLex Skills is built on four core pillars of legal engineering:

  1. Lawyer-in-the-Loop: The agent never makes the final legal call. It identifies risks and presents "Safe Harbor" alternatives for the lawyer's approval.
  2. Motor vs. Fuel Architecture: The logic spine (SKILL.md) is separated from the local legal data (jurisdictions/*.md). This allows for global scalability while maintaining local precision.
  3. Evidence over Claims: Every citation must be verifiable. If the agent cannot cite the specific gazette or article, it must flag the uncertainty.
  4. Agentic but Human-in-the-Loop: The AI agent acts as a disciplined Legal Associate. It drafts, routes jurisdictions, and flags risks, but the final legal sign-off always belongs to a qualified attorney.

The SuperLex Skills Workflow

When you ask SuperLex Skills for a document or analysis, it follows a deterministic path:

  1. Context Intake: The lawyer-context-manager ensures the agent knows who you are representing.
  2. Jurisdiction Injection: The agent loads the specific "Legal Fuel" (e.g., tr.md for Turkish law) into the "Logic Motor."
  3. Drafting & Fact-Check: The agent drafts the document while running an internal SELF-TEST against anti-patterns.
  4. Agentic Verification: The agent triggers a <HARD-GATE>, presenting the 3 highest-risk areas and suggesting safer alternatives.
  5. The Disclaimer: Only after approval is the core/DISCLAIMER.md appended and the document delivered.

Installation & IDE Integration

SuperLex Skills is designed to be a native "plugin" for the world's leading AI coding environments.

Cursor

Install as a repository-based skill:

/add-plugin https://github.com/avSercanKoc/superlex-skills

Uses .cursor-plugin/ for deep agent discovery.

Claude Code

Install via the CLI:

/plugin install https://github.com/avSercanKoc/superlex-skills

Uses .claude-plugin/ and CLAUDE.md for role definition.

OpenCode

Install as a native plugin:

# Add to your opencode.json
"plugin": ["superlex-skills@git+https://github.com/avSercanKoc/superlex-skills.git"]

Uses .opencode/ for universal agentic installation.

Codex

Clone and symlink:

git clone https://github.com/avSercanKoc/superlex-skills.git ~/.codex/superlex-skills
ln -s ~/.codex/superlex-skills/skills ~/.agents/skills/superlex-skills

Uses .codex/ for native skill discovery.

Gemini CLI

Add to your project's GEMINI.md or rely on gemini-extension.json for auto-discovery:

git clone https://github.com/avSercanKoc/superlex-skills.git
# Then reference the repo root in your Gemini CLI workspace settings.

Uses gemini-extension.json + GEMINI.md for context injection at session start.


The Skills Library

Category Skill Risk Jurisdiction Description
Meta using-superlex-skills 🟢 Low tr Bootstrap meta-skill for agent onboarding and rules.
Meta lawyer-context-manager 🟢 Low tr Central client & firm profile management.
Meta specification-before-drafting 🟢 Low tr Pre-drafting legal strategy and risk specification.
Meta writing-superlex-skills 🟢 Low tr TDD-based guide for authoring, testing, and contributing new skills.
Drafting privacy-policy 🟡 Med tr, eu KVKK/GDPR compliant data protection policies.
Drafting terms-of-use 🟡 Med tr Platform-specific (SaaS/E-com) usage terms.
Drafting nda-generator 🟡 Med tr Non-disclosure agreements + legal checklist.
Analysis contract-review 🔴 High tr Clause-by-clause risk matrix & balance report.
Notice legal-letter 🔴 High tr Formal notices (İhtarname) with UETS/Notary rules.
Orchestrator subagent-driven-development 🟢 Low tr Orchestrates multi-task legal workflows with subagents and reviews.

The Guardian: Automated Validation

SuperLex Skills is the only skill library that includes a built-in Linter for Lawyers. Our scripts/validate-skills.sh script is run on every commit to ensure:

  • The Guardian: Automated structural and "leakage" validation via scripts/validate-skills.sh.
  • SessionStart Hooks: Automatic legal context injection via hooks/session-start.
  • Pre-commit Integrity: Automated git-hooks to prevent committing invalid skills.
  • Safety Compliance: Verifies that High-Risk skills have both Pre-Gen and Post-Gen HARD-GATEs.

Run it locally:

npm run validate

Contributing

SuperLex Skills is built to grow. You can contribute by adding new Jurisdiction Packs (e.g., adding us.md or de.md to an existing skill) or by creating new Legal Skills.

Please read the CONTRIBUTING.md, SKILL-ANATOMY.md, and JURISDICTION-ANATOMY.md before submitting a Pull Request.


License & Governance

  • License: MIT License - see LICENSE for details.
  • Changelog: Release history: CHANGELOG.md
  • Future plans: ROADMAP.md
  • Contributor guide: CONTRIBUTING.md
  • Community policy: CODE_OF_CONDUCT.md
  • Security: Responsible disclosure: SECURITY.md
  • Tests: Behavioral pressure scenarios: tests/pressure-tests/

The Architect & Ecosystem

SuperLex Skills is engineered by Sercan Koç and the legal engineering team at Genesis Hukuk — a visionary Law + Tech Studio based in Turkey.

We build infrastructure where legal compliance isn't an afterthought, but a core architectural design (Compliance by Design).

Support Open Source Law

If SuperLex Skills has optimized your legal workflows or saved your firm billable hours, consider supporting our open-source legal engineering efforts through GitHub Sponsors.

About

The open-source jurisdiction-aware legal engineering framework for AI agents. Bringing agentic legal document generation and analysis up to professional standards through a three-layer safety model and extensible architecture.

Resources

License

Code of conduct

Contributing

Security policy

Stars

2 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors