Show HN: SentinelGate – Drop prompt injections at Token 0 before LLM inference

Category: security

Tags: prompt-injection, llm-security, ai-firewall, token-filtering

Score: 5.3/10 (Innovation: 6, Technical: 5, Documentation: 3, Utility: 5)

SentinelGate claims to be an in-memory firewall that blocks prompt injections at token 0 before LLM inference, using a kinematic veto mechanism with Ed25519 attestations. The concept of pre-inference prompt filtering is interesting, but the README is cryptic and lacks concrete implementation details or evidence of effectiveness.

Target audience: backend devs, ai engineers, security engineers

Repository: https://sentinelgate.dev

View on Hacker News