Last commit on July 3, 2026·Created on July 3, 2026
Kulaxyz/token-diet
“Cuts token waste across the session without cutting correctness.”
Combined rank
#294
across all skills
In Token Optimization
#2
category rank
Stars
503
-1.9% in last 7d
Forks
4
-20.0% in last 7d
Watchers
0
0 watchers in last 7d
Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of real attention, adoption, and active maintenance.
Token Diet applies a structured set of rules to every part of a coding agent's interaction: user replies, written artifacts, tests, code, context reads, tool calls, and sub-agent instructions. Users can choose from three enforcement levels — on, lite, and ultra — while guardrails protect critical test coverage, exact identifiers, and reasoning depth. It helps heavy AI users, engineering teams, and administrators lower token consumption and costs without losing substance.