PromptShield v1.0.0 - Official Release

PromptShield v1.0.0 Official Release

The first release of PromptShield, an automated red-blue adversarial testing framework built for LLM application security. This release establishes a complete dual-LLM testing lifecycle, enabling production-grade penetration testing and automated defense reporting.

Core Features

• Dual-LLM Adversarial Framework: Integrated Google Gemini (Red Team) and NVIDIA NIM Llama 3.3 70B (Blue Team).
• Empirical Impact Measurement: Automated leakage detection via specialized Victim LLM environments.
• Adaptive Evolution Capabilities: Enabled multi-round autonomous attack mutation based on defensive feedback loops.
• Industry Standard Mapping: Native support for classifying adversarial strategies under the MITRE ATT&CK framework.
• Actionable Auto-Reporting: One-click production of 7-section vulnerability analysis and system prompt hardening text.

Live Deployment

The application has been successfully deployed and is available for public testing:
https://prompt-shield.streamlit.app