AI Legal Research Skill

Purpose

Integrate OpenAI GPT-4 with UK legal databases for context-aware legal research with proper citations and disclaimers.

When Claude Uses This

• Implementing AI chat features
• Legal document retrieval
• Case law searches
• Citation generation
• Building RAG pipelines for legal context

Architecture

OpenAI Integration

// src/services/OpenAIService.ts
import OpenAI from 'openai';

interface LegalQuery {
  question: string;
  caseContext?: string;
  jurisdiction: 'UK' | 'England' | 'Scotland' | 'Wales' | 'NI';
}

interface LegalResponse {
  answer: string;
  sources: Array<{
    type: 'legislation' | 'case-law';
    title: string;
    url: string;
    relevance: number;
  }>;
  disclaimer: string;
}

UK Legal APIs

1. Legislation.gov.uk API

• Base URL: https://www.legislation.gov.uk/
• Format: XML (parse with fast-xml-parser)
• Types: Acts, Statutory Instruments, UK Public General Acts

2. National Archives Case Law API

• Base URL: https://caselaw.nationalarchives.gov.uk/
• Format: JSON
• Coverage: All UK courts and tribunals

RAG Pipeline Workflow

1. User Query → Extract Legal Terms
2. Search UK Legal APIs → Retrieve Relevant Documents
3. Chunk Documents → Create Embeddings
4. Query OpenAI with Context → Generate Response
5. Cite Sources → Add Mandatory Disclaimer

Implementation Checklist

Phase 1: OpenAI Setup

• Install: pnpm add openai
• Create src/services/OpenAIService.ts
• Store API key in .env: OPENAI_API_KEY=sk-...
• Implement streaming responses
• Add token usage tracking

Phase 2: UK Legal API Integration

• Create src/services/UKLegalAPIService.ts
• Implement legislation.gov.uk parser
• Implement caselaw search
• Add response caching (24hr TTL)
• Handle rate limiting (respectful scraping)

Phase 3: RAG Implementation

• Document chunking (max 512 tokens/chunk)
• Vector embeddings with OpenAI text-embedding-3-small
• Similarity search (cosine similarity)
• Context injection into prompts
• Source attribution

Phase 4: UI Integration

• Chat interface component (src/features/chat/AIChat.tsx)
• Streaming message display
• Source citations UI
• Disclaimer footer (always visible)
• Copy/export responses

Critical Requirements

Legal Disclaimer (MANDATORY)

Every AI response MUST include:

⚠️ DISCLAIMER: This is information, not legal advice.
Consult a qualified solicitor for your specific situation.

Data Privacy (GDPR)

• User consent before sending queries to OpenAI
• Option to disable AI features
• Audit log all AI queries
• No PII sent to OpenAI without explicit consent

Rate Limiting

// Max queries per user
const RATE_LIMITS = {
  free: 10 / day,
  pro: 100 / day,
  enterprise: unlimited
};

Example Usage

Puppeteer Web Scraping (Case Law)

# Use puppeteer MCP to scrape case law
mcp__puppeteer__navigate https://caselaw.nationalarchives.gov.uk/
mcp__puppeteer__fill "input[name='query']" "employment discrimination"
mcp__puppeteer__click "button[type='submit']"

Context7 for Library Docs

# Get latest OpenAI SDK docs
mcp__context7__get-library-docs openai

Testing Strategy

Unit Tests

• OpenAI API mocking (vitest)
• Legal API response parsing
• Disclaimer injection validation

E2E Tests

• Complete RAG pipeline
• Streaming responses
• Source citation accuracy
• Rate limiting enforcement

Performance Targets

• Response time: < 3s for cached queries
• First token: < 500ms (streaming)
• API calls: Batch where possible
• Caching: 24hr TTL for legal documents

Error Handling

• OpenAI API failures → Graceful degradation
• Legal API timeouts → Show cached results
• Rate limit exceeded → Clear user message
• Invalid queries → Suggest reformulation

References

• OpenAI Node.js SDK: https://github.com/openai/openai-node
• Legislation.gov.uk API: https://www.legislation.gov.uk/developer
• Case Law API: https://caselaw.nationalarchives.gov.uk/about-this-service
• GDPR Compliance: Track user consents in consent_logs table