판 이력 — RoguePrompt: Dual-Layer Encoding for Self-Reconstruction to Circumvent LLM Moderation | AIChainDay