pithlisted

# PITH — Inter-Agent Payload Compressor > *Compresses what agents say to each other — the gap no other tool fills.* Removes token waste from agent-to-agent handoffs. Uses Zipf word-density scoring to rank sentences by information content, Benford's Law as a structural integrity validator, and a feedback loop that prevents over-compression from damaging meaning. --- ## Origin PITH was created by **Albert** ([@VjAlbert](https://github.com/VjAlbert)), a developer and builder of Claude skills with a background in **game theory** and **Benford's Law**. The motivation came from a simple observation: every existing compression tool targets either the input (LLMLingua) or the output (Caveman). The enormous token payload exchanged *between* agents in a pipeline — tool results, reasoning traces, intermediate context — was being passed raw and untouched. From a game theory perspective, this is a deviation from equilibrium. In any multi-player system, the Nash equilibrium of communication is the strategy where each agent transmits the minimum information necessary for the next agent to act optimally. Every token above that minimum is pure cost. The mathematical insight: Zipf's Law governs which words carry information (rare = dense), and Benford's Law governs whether compressed text still has the statistical signature of natural writing. Together they form a compression-validation loop that is principled, fast, and requires no model calls. The empirical test confirmed it: human