Hey Product Hunt community! 👋 While building autonomous AI agents and Model Context Protocol (MCP) integrations, we ran into a scary reality: standard LLMs blindly execute malicious instructions hidden in resumes, customer tickets, or scraped web pages. An injection isn't just an offensive text output anymore—it's arbitrary bash code execution (rm -rf), SSRF targeting internal cloud metadata (169.254.169.254), or runaway subagent loops that blow API budgets. We built AgentExploitDB to provide an end-to-end defense kit conforming to the OWASP Top 10 for LLM Applications: 🛡️ 1. Zero-Dependency Runtime Shield (MCPGuardrail): Pure Python middleware intercepting shell injections, SSRF, and unicode steganography in sub-millisecond latency (< 1ms). 📊 2. Open Benchmark (150 Vectors): 100% free MIT-licensed dataset on GitHub with executable unit test assertions. 📄 3. Automated HTML Audit Certificate: 1-Click executive security report generator designed for client deliverables and CI/CD gates. We also packaged an exhaustive 1,200+ vector Pro Vault on Gumroad ($9) for security teams needing complete compliance coverage. 🔗 Free GitHub Repo: https://github.com/trobasuj-cpu/... 🔗 Pro Vault: https://beatsprom.gumroad.com/l/... Would love to hear your feedback! How are you currently protecting your AI agent tool-call pipelines in production?

AgentExploitDB

Zero-dependency security shield & exploit benchmark for AI

Zero-dependency security shield & exploit benchmark for AI