[Team VELOCITY] PromptShield AI – Unified Enterprise AI Firewall Across Browser, Desktop & CLI

[ruthviksharma-d]

🚀 Team VELOCITY – PromptShield AI

Overview

This PR contains Team VELOCITY's complete submission for the Mutagent Hackathon.

PromptShield AI is an Enterprise AI Firewall designed to prevent sensitive information from being unintentionally shared with AI assistants. Instead of replacing existing AI tools, PromptShield integrates directly into existing developer workflows, protecting AI interactions across browsers, desktop AI applications, IDEs, and command-line interfaces using a unified security pipeline.

---

🔗 Links

• 🌐 GitHub Repository: https://github.com/ruthviksharma-d/mutagent-hackathon
• 🎥 Demo Video: https://drive.google.com/file/d/1bUyAIJRPFzvCIZHPxW0gDTasqXw1tMZ0/view?usp=sharing

---

✨ Key Features

🌐 Browser Protection

Real-time protection for:

• ChatGPT
• Claude
• Gemini

Features include:

• Prompt interception
• File upload scanning
• Risk analysis
• Policy enforcement
• Explainable security findings

---

💻 Desktop & IDE Protection

PromptShield extends protection beyond the browser by securing AI interactions inside desktop applications and AI-powered IDEs.

Supported integrations include:

• Claude Desktop
• Claude Code
• Gemini CLI
• AI-powered IDE workflows
• VS Code Extension

This enables developers to safely use AI during software development while maintaining the same enterprise security policies.

---

⚙️ CLI Protection

PromptShield provides secure AI access through protected CLI wrappers.

Supported providers:

• Claude CLI
• Gemini CLI

Developers interact normally while every request is intercepted, analyzed, and policy-checked before reaching the AI model.

---

📄 Secure File Analysis

PromptShield scans uploaded files before they reach an AI model.

Supported formats:

• TXT
• Markdown
• PDF
• DOCX
• JSON
• CSV
• XLSX
• .env

---

🛡 Unified Security Pipeline

All integrations share a centralized backend that performs:

• Sensitive information detection
• PII detection
• Secret detection
• Prompt analysis
• Risk scoring
• Policy evaluation
• Audit logging

Supported policy actions:

• ✅ ALLOW
• ⚠️ WARN
• ✂️ REDACT
• ⛔ BLOCK

---

📊 Enterprise Dashboard

Centralized management interface featuring:

• Prompt audit logs
• Risk analytics
• Investigation history
• Policy management
• Security insights

---

🏗 Architecture

PromptShield follows a single reusable security architecture.

Browser Extension
        │
Desktop / IDE
        │
Claude CLI
        │
Gemini CLI
        │
VS Code Extension
        │
────────▼────────
 PromptShield Backend
────────▼────────
 Investigation Engine
 Risk Engine
 Policy Engine
 Audit Engine
────────▼────────
 Decision
(ALLOW / WARN / REDACT / BLOCK)