AI Model Token & Cost Audit

★ 7.7 · ai-tooling

audit is a Claude Code skill that generates a token usage and cost optimization report across MiniMax M2.1, MiniMax-lightning, Claude Sonnet, and other configured models. Invoked via `ralph audit`, it displays query distribution by model, daily and weekly usage trends, and a breakdown of actual versus potential costs. Pricing references are built in: Claude Sonnet at $3.00/$15.00 per 1M tokens (input/output), MiniMax M2.1 at $0.30/$1.20 (~92% savings), and MiniMax-lightning at $0.15/$0.60 (~96% savings). The active model is determined by `~/.claude/settings.json` or `ANTHROPIC_DEFAULT_*_MODEL` environment variables, requiring no extra flags. It's designed for developers and teams who want to track AI model spending and identify where switching to a lower-cost model translates into measurable budget savings.