What problem does it solve? Long, verbose AI responses waste tokens and accelerate weekly usage cap exhaustion, especially when dispatching sub-agents or running batch tasks. This Skill enforces an ultra-compressed response style that cuts filler words while preserving full technical accuracy. ## Core Features & Use Cases - Persistent Terse Mode: Once triggered, every response stays compressed across multiple turns until the user explicitly says to stop. - Precision-Preserving Compression: Removes articles, filler, hedging, and pleasantries while keeping technical terms, code blocks, and error messages verbatim. - Auto-Clarity Exceptions: Automatically pauses compression for security warnings, irreversible operation confirmations (DROP, rm -rf, git reset --hard), and ambiguous multi-step sequences. - Use Case: When dispatching Sonnet sub-agents for batch tasks, append "respond in caveman mode" to the prompt so each agent's status report shrinks to lines like "Batch 03 done. pushed:12 failed:1 429:0", cutting per-agent token usage by an estimated 40-60%. ## Quick Start Say "caveman" or "省 token" to activate compressed response mode for the rest of the conversation.
