📉 Real token savings from the Caveman skill are lower than expected
JetBrains conducted a benchmark of the Caveman skill for AI agents, which promised up to 65% token savings. Tests showed that in real-world tasks, savings do not exceed 8.5%, as the bulk of tokens is consumed by code and commands rather than dialogue.
🌍 The results highlight the gap between chat optimization and actual agentic processes.
👤 The profit will be around 8-10%, which is safe for work quality.
Source 1: https://blog.jetbrains.com/ai/2026/07/speak-to-ai-agents-like-cavemen-tosave-tokens