What problem does it solve? Every word an AI agent writes is re-sent in the conversation history on every subsequent turn, so verbose filler (greetings, preambles, recaps) multiplies token costs across long sessions. This Skill switches the agent into a compressed communication mode that cuts the fluff while keeping all technical substance intact. ## Core Features & Use Cases - Three intensity levels: lite (no filler, full sentences), full (fragments, dropped articles), and ultra (abbreviations, arrows for causality), switchable at any time and persistent across the session. - Substance protection rules: file paths, commands, error messages, version numbers, and expressions of uncertainty are never shortened or paraphrased. - Auto-clarity fallback: the agent exits compressed mode for security warnings, destructive-action confirmations, and ordered multi-step sequences, then returns to it automatically. - Use Case: During a long debugging session with a tight usage quota, activate full mode so every reply is a terse finding-plus-action line, cutting the per-turn prose cost without losing any diagnostic detail. ## Quick Start Tell the agent to turn on caveman mode at full intensity and keep it active for the rest of the session.
