Comment by alvinveroy
Comment by alvinveroy 9 hours ago
Working on UPSS (Universal Prompt Security Standard) - an open-source framework for managing LLM prompts securely at scale.
The problem: Most organizations hardcode prompts directly into application code, creating security vulnerabilities (90% prompt injection success rate in typical deployments), operational inefficiency (3-5 day deployment cycles for simple prompt changes), and compliance gaps (insufficient audit trails for SOC 2, ISO 27001).
Our approach: - Externalize prompts from code with secure configuration management - Implement modular middleware architecture with composable security primitives (BasicSanitizer, LightweightAuditor, SimpleRBAC, InputValidator) - Provide complete audit trails and version control with approval workflows - Support both startups and enterprises with practical, not theoretical, security
Version 1.1.0 is now available with Python implementation and examples for Node.js, Java, Go, Rust.
We're actively looking for community contributions - security primitives, framework integrations, language implementations, and adoption stories.
https://github.com/upss-standard/universal-prompt-security-s...