Show HN:I built a safety shield for AI agents that intercepts dangerous commands · HackerLangs