Jev Skill Gate
shivam pansuriya@PansuriyaShivam
This project optimizes Claude Code by reducing the number of tokens required to load skill descriptions at session start, using Jev to score skills in parallel.
SignalOne call, $0.0009, scores all 217 in parallel. 12,750 → 3,185 tokens. https://github.com/ShivamPansuriya/jev-skill-gate
Claude Code loads every skill description into context at session start.
217 skills on my machine = 12,750 tokens spent before I type a word.
Built a gate on @typesafeai Jev. One call, $0.0009, scores all 217 in parallel.
12,750 → 3,185 tokens.
github.com/ShivamPansuriya/jev-skill-gate