Scans LLM endpoints for prompt injection and jailbreak vulnerabilities using the TMAS AI Scanner.
TrendAI LLM Scanner is a specialized security tool for Claude Code designed to identify and mitigate vulnerabilities in Large Language Model endpoints. It automatically detects local instances like Ollama or LM Studio, as well as remote OpenAI-compatible APIs, to perform automated red-teaming across four critical attack vectors: system prompt leakage, sensitive data disclosure, tool definition leakage, and malicious code generation. By maintaining a history of scans, it provides unique security posture drift analysis, allowing developers to track improvements or regressions in their model's safety and reliability over time.
主な機能
01Zero-prompt auto-detection of local LLM endpoints and available models
02Automated red-teaming for prompt injection and jailbreak attempts
030 GitHub stars
04Historical drift detection to track security posture changes over time
05Support for Ollama, LM Studio, and OpenAI-compatible API providers
06Comprehensive security reporting with JSON output and automated analysis
ユースケース
01Generating detailed security compliance reports for AI application endpoints
02Monitoring security regressions during model fine-tuning or system prompt updates
03Auditing local LLM deployments for prompt injection vulnerabilities before production release
