LochBot

What is LochBot?

LochBot is a free, browser-based tool designed to identify prompt injection vulnerabilities in Large Language Models (LLMs). It operates entirely client-side, meaning no data leaves your browser during the analysis. This approach enhances privacy and security, as sensitive system prompts are not uploaded to any server. The tool analyzes your LLM's system prompts against a database of 31 distinct attack patterns, including common jailbreaks, role override attempts, and data exfiltration techniques.

Upon completion of the analysis, LochBot provides a security score that quantifies the vulnerability of your prompt. This score is accompanied by actionable remediation guidance, helping you understand specific areas for improvement and how to strengthen your LLM's defenses. The absence of a signup requirement and its client-side execution make it an accessible option for quick checks and initial security assessments.

This tool is particularly useful for developers and security professionals working with LLMs who need to perform preliminary checks on their system prompts. While it offers valuable insights into potential vulnerabilities, it is intended as a first line of defense. For more in-depth security audits or integration into CI/CD pipelines, further specialized tools or manual reviews may be necessary.

Key features

• Browser-based checker
• Client-side analysis
• 31 attack patterns
• Security scoring
• Remediation guidance
• No signup required

Use cases

• Initial security check for LLM system prompts.
• Testing prompts for common jailbreak techniques.
• Assessing risk of data exfiltration via prompts.
• Verifying LLM adherence to role instructions.
• Quick security assessment before deployment.

Pros & cons

FAQ