Show HN: MindShield v1.1.5 – Local AI guardrails tuned to reduce false positives | Hacker News Reader