Justice Matrix Research Agent

When to Use

• Discovering new global youth justice cases
• Finding advocacy campaigns worldwide
• Updating the Justice Matrix database
• Checking source health and coverage gaps
• Running scheduled research cycles

Commands

| Command | Purpose | Duration | |---------|---------|----------| | /ralph-matrix-scan | Run discovery across active sources | 10-30 min | | /ralph-matrix-extract | Process pending discoveries | 5-15 min | | /ralph-matrix-status | Show pipeline status and stats | Instant | | /ralph-matrix-source [name] | Deep dive specific source | 10 min |

Workflow

DISCOVER → EXTRACT → QUEUE → REVIEW → APPROVE
   ↓          ↓         ↓        ↓         ↓
Sources    AI Parse  Pending   Admin    Database

1. Source Scan (/ralph-matrix-scan)

// Query active sources by priority
const sources = await supabase
  .from('justice_matrix_sources')
  .select('*')
  .eq('is_active', true)
  .order('scrape_priority', { ascending: true });

// For each source:
// 1. Fetch pages
// 2. Extract case/campaign links
// 3. Check for duplicates
// 4. Queue new discoveries

2. AI Extraction (/ralph-matrix-extract)

For each pending discovery:

1. Fetch full content from source URL
2. Run AI extraction using patterns from PRD
3. Geocode to lat/lng
4. Categorize by legal themes
5. Calculate confidence score
6. Check for duplicates

3. Review Queue

Discoveries go to /admin/justice-matrix/discoveries where admins can:

• Preview extracted data
• Edit before approval
• Approve → creates case/campaign
• Reject → marks as rejected with notes
• Merge → links to existing item

Source Types

| Type | Extraction Focus | |------|------------------| | court_database | Citation, year, court, holding, outcome | | advocacy_org | Campaign name, orgs, goals, tactics, status | | legal_database | Case analysis, precedent value | | regional_body | Regional jurisdiction, treaty basis |

Extraction Patterns

Court Database

{
  "required": ["case_citation", "jurisdiction", "year", "court"],
  "extract": ["strategic_issue", "key_holding", "outcome"],
  "infer": ["categories", "precedent_strength", "lat/lng"]
}

Advocacy Organization

{
  "required": ["campaign_name", "country_region", "lead_organizations"],
  "extract": ["goals", "tactics", "outcome_status"],
  "infer": ["is_ongoing", "categories", "lat/lng"]
}

Quality Scoring

| Signal | Weight | Description | |--------|--------|-------------| | Relevance | 30% | Youth justice / child rights focus | | Precedent | 25% | Significance for future work | | Coverage | 20% | Fills geographic/thematic gap | | Recency | 15% | How current is this? | | Actionable | 10% | Useful for practitioners |

Database Tables

| Table | Purpose | |-------|---------| | justice_matrix_sources | Scraping source configuration | | justice_matrix_discovered | Pending review queue | | justice_matrix_scrape_logs | Run history and stats | | justice_matrix_cases | Approved cases | | justice_matrix_campaigns | Approved campaigns |

PRD Reference

Full source configuration: ralph/justice-matrix-prd.json

Sacred Boundaries

Never scrape:

• Sealed juvenile records
• Private case files
• Paywalled content without permission
• Individual identifying information

Always mark:

• Indigenous rights cases
• First Nations advocacy
• Community-controlled campaigns

Always check:

• Publication permissions
• Child privacy protections
• Cultural sensitivity

Example Usage

# Quick status check
/ralph-matrix-status

# Run full discovery cycle
/ralph-matrix-scan

# Process discoveries
/ralph-matrix-extract

# Deep dive on specific source
/ralph-matrix-source "ECHR HUDOC"