Master of Tokens v2.1: How I cut 92% (24,800+ tokens/session) across 200+ Claude Code skills & Antigravity agents

Description and use

Implementing master‑of v2.1.0 cut token usage by 92 % across 200+ Claude Code and Antigravity skills, saving 24,877 tokens per session.

Problem

Claude Code and Antigravity agents consume excessive system prompt tokens (27,159 tokens) when 200+ skills, subagents, and slash commands are always-on, leading to high token usage per session.

Outcome

Implemented master‑of v2.1.0 reduced per‑session token usage by 92%, saving 24,877 tokens across 200+ components.

Workflow

• Install master‑of via git clone and bun run
• Run mo sync to batch‑park all raw skills, agents, and commands into dormant domain gates
• Use mo classify --all to confirm categories in bulk
• Execute skills via gated commands or agents, reducing prompt noise
• Repeat as needed for new components

Evidence

27,159 tok before gating (all 200+ components always‑on) − 2,282 tok after gating (domain gates exposed instead) ─────────── 24,877 tok saved per session (92% reduction)

Source