shadow-ai-monitor（影子 AI 监视器）

Shadow AI Monitor

A professional web dashboard system for tracking employee AI tool usage, identifying data exposure risks, and measuring PIPEDA compliance.

What This Skill Does

Generates enterprise-grade HTML dashboards showing:

• 📊 AI Tool Usage Analytics - Track which tools employees use (ChatGPT, Claude, Gemini, Copilot, Midjourney, Grok, Perplexity, Cursor, GitHub Copilot)
• ⚠️ Risk Analysis - Classify data exposure as Low/Medium/High risk
• 📋 Compliance Scoring - PIPEDA compliance score (0-100) with detailed requirement breakdown
• 👥 Employee Insights - Anonymized usage patterns and heatmaps
• 📈 Trend Tracking - Daily usage trends and event logs
• 📄 Board-Ready Reports - PDF export with professional styling

Perfect For

• Security Teams monitoring shadow AI adoption
• CTOs assessing compliance gaps
• Legal/Compliance teams tracking data exposure
• Sales Teams demoing AI governance products
• Consultants delivering AI risk assessments

Quick Start

Generate a demo dashboard:

node scripts/generate_demo_data.js
node scripts/generate_dashboard.js shadow-ai-data.json

Open shadow-ai-dashboard.html in your browser.

Features

🎯 Interactive Dashboard

Click to Explore:

• Employee Bars → See top 3 riskiest events, tools used, data categories, dates
• Compliance Score → View 5 PIPEDA requirements, pass/fail status, specific fixes
• Export Button → Generate PDF with letterhead styling

Key Metrics:

• PIPEDA compliance score (color-coded: green ≥85, yellow ≥70, red <70)
• Risk event breakdown (High/Medium/Low)
• Top 3 AI tools by usage
• Top 10 employees (anonymized)

Visualizations:

• Bar charts: Tool usage, employee activity
• Doughnut chart: Risk distribution
• Line chart: Daily trends
• Heat map: Employee usage intensity
• Scrollable event log (last 20 events, high-risk highlighted)

📋 Personalized Recommendations

Auto-generated advice based on actual data patterns:

• References specific employees and percentages
• Tool-specific recommendations (e.g., "ChatGPT accounts for 48% of usage")
• Compliance-specific action items
• Example: "Employee 5 generated 8 high-risk events involving client matters — review recommended"

🔒 PIPEDA Compliance Analysis

Detailed breakdown of 5 core requirements:

1. Consent for Collection
2. Limiting Use, Disclosure & Retention
3. Safeguards
4. Openness
5. Individual Access

Each shows:

• Pass/Fail status
• Requirement description
• Specific action items using your actual data

📊 Demo Data Generator

Creates realistic data for a 50-person Canadian organization:

Configurable:

• Company name (default: Morrison & Associates)
• Employee count
• AI tools tracked
• Data categories (Client Legal, Financial, Health, Personal, etc.)
• Risk distribution
• Time period (last 7 days default)

Realistic Patterns:

• Mix of 9 different AI tools
• Varied risk levels (~25% high, ~35% medium, ~40% low)
• Employee anonymization (Employee 1, Employee 2, etc.)
• Concerning patterns for demo impact

Files

shadow-ai-monitor/
├── SKILL.md                        # This file
├── scripts/
│   ├── generate_demo_data.js      # Demo data generator
│   └── generate_dashboard.js      # HTML dashboard generator

Installation

Via ClawHub:

clawhub install shadow-ai-monitor

Manual:

mkdir -p ~/.openclaw/skills
cd ~/.openclaw/skills
# Download and extract skill files

Usage

Basic Demo Dashboard

cd ~/.openclaw/workspace  # or your working directory
node ~/.openclaw/skills/shadow-ai-monitor/scripts/generate_demo_data.js
node ~/.openclaw/skills/shadow-ai-monitor/scripts/generate_dashboard.js shadow-ai-data.json
open shadow-ai-dashboard.html  # or double-click in Finder/Explorer

This generates:

• shadow-ai-data.json - Raw usage data
• shadow-ai-dashboard.html - Interactive dashboard

Customizing Demo Data

Edit scripts/generate_demo_data.js to customize:

Company Name:

const output = {
  company: 'Your Company Name',
  // ...
};

Employee Count:

const employees = Array.from({ length: 100 }, (_, i) => ({
  id: `Employee ${i + 1}`,
  // ...
}));

AI Tools Tracked:

const AI_TOOLS = [
  'ChatGPT', 'Claude', 'Gemini', 'Your-Custom-Tool'
];

Data Categories:

const DATA_CATEGORIES = [
  { name: 'Your Category', risk: 'High' },
  // ...
];

Automated Weekly Reports

Set up with OpenClaw cron:

{
  "name": "Shadow AI Weekly Report",
  "schedule": {"kind": "cron", "expr": "0 9 * * 1", "tz": "America/Toronto"},
  "payload": {
    "kind": "agentTurn",
    "message": "Generate Shadow AI dashboard: 1) Run demo data generator 2) Generate dashboard 3) Send WhatsApp notification with metrics"
  },
  "sessionTarget": "isolated",
  "delivery": {"mode": "announce", "channel": "whatsapp"}
}

Technical Details

Data Structure

{
  "generated": "2026-02-22T15:00:00Z",
  "company": "Morrison & Associates",
  "employeeCount": 50,
  "period": "Last 7 days",
  "events": [
    {
      "timestamp": "2026-02-22T10:30:00Z",
      "employee": "Employee 1",
      "employeeRole": "Senior Associate",
      "tool": "ChatGPT",
      "dataCategory": "Client Legal Matters",
      "risk": "High"
    }
  ],
  "metrics": {
    "totalEvents": 268,
    "topTools": [["ChatGPT", 89], ["Claude", 56]],
    "riskCounts": {"Low": 95, "Medium": 83, "High": 90},
    "complianceScore": 68,
    "recommendations": ["..."]
  }
}

Risk Scoring

High Risk:

• Client legal matters
• Financial records
• Health information

Medium Risk:

• Personal information
• Internal memos
• Proprietary data

Low Risk:

• General questions
• Code templates
• Public research

Compliance Score Calculation

Base: 100 points
Deduct: -2 points per % of high-risk events
Deduct: -0.5 points per % of medium-risk events
Range: 0-100

Example:

• 268 total events
• 90 high-risk (34%) → -68 points
• 83 medium-risk (31%) → -15.5 points
• Score: 16/100

Dashboard Technology

• Pure JavaScript - No npm install required
• Chart.js via CDN - Loaded from jsDelivr
• Dark Theme - Professional enterprise styling
• Responsive Design - Works on desktop and tablet
• Print-Optimized - Clean PDF export with letterhead

Use Cases

1. CTO Demo

Script:

"This is Morrison & Associates, a 50-person law firm. Over 7 days, we detected 268 AI tool events. Their PIPEDA compliance score is 16/100 because 34% of interactions involved client legal matters and health records shared with unapproved AI tools."

*Click compliance score* → Show specific PIPEDA failures *Click Employee 38* → Show their risky events *Click Export* → Board-ready PDF

2. Security Assessment

Generate dashboard with your organization's data:

1. Monitor Slack/Teams/Email for AI tool mentions
2. Log to JSON format matching data structure
3. Run dashboard generator
4. Present findings to security team

3. Sales Presentation

Demo realistic data showing:

• Concerning compliance gaps
• Specific risk patterns
• Clear ROI through risk reduction
• Professional, board-ready output

4. Compliance Audit

Track progress over time:

• Generate weekly snapshots
• Compare compliance scores
• Identify trends (improving/worsening)
• Document remediation efforts

Security & Privacy

✅ No External API Calls - All processing local ✅ No Data Collection - Demo data never leaves your machine ✅ No Credentials Required - Pure JavaScript execution ✅ Anonymized Data - Employee IDs, no PII in demo ✅ Open Source - Review all code before running

Requirements

• Node.js (any recent version)
• Web Browser (Chrome, Firefox, Safari, Edge)
• No npm packages - Uses only Node.js built-ins

Roadmap

Future enhancements:

• [ ] Real-time monitoring integration
• [ ] Multi-company comparison
• [ ] Historical trend analysis
• [ ] Custom compliance frameworks (GDPR, HIPAA, SOC 2)
• [ ] Email delivery of reports
• [ ] Slack/Teams integration
• [ ] Live data ingestion from DLP tools

Support & Contribution

• Issues: Report bugs or request features on ClawHub
• Questions: OpenClaw Discord community
• Contributions: Submit improved visualizations, compliance frameworks, or data sources

License

MIT - Free for personal and commercial use

Credits

Built with OpenClaw by the automation engineering community.

Version: 1.0.0 Author: Automation Engineers Category: Security, Compliance, Analytics Tags: AI monitoring, PIPEDA, compliance, security, dashboard