Guardrails on James Lee | Security Noteshttps://jameslee177.github.io/tags/guardrails/Recent content in Guardrails on James Lee | Security NotesHugoen-usFri, 30 Oct 2026 16:00:00 +0800Part 2: The Limits of System Prompt Defenses — Do We Still Need an AI Firewall?https://jameslee177.github.io/posts/prompt-injection-part-2/Fri, 30 Oct 2026 16:00:00 +0800https://jameslee177.github.io/posts/prompt-injection-part-2/A well-written system prompt can stop simple attacks, but it is not a security boundary.