# Stack Measure

> Measure token usage and savings via ccusage — establish a baseline and compare before/after. Use right after init (to record the baseline) and again after about a week to quantify the input-token reduction against the ≥40% target. Wraps `agent-stack measure`, which reads ccusage's local JSONL — never vendor self-reported numbers.

- Skill: `drmahdikazempour/stack-measure` (Agent Skill)
- Install (CLI): `npx skillmds@latest add drmahdikazempour/stack-measure`
- Raw SKILL.md: https://api.skillmd.com/api/skills/drmahdikazempour/stack-measure/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: drmahdikazempour (https://skillmd.com/u/drmahdikazempour)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/drmahdikazempour/stack-measure

---


# stack-measure

Quantify token usage with the neutral `ccusage` tool. Savings are always measured, never claimed.

## What to do
- **Baseline:** `npx @drmahdikazempour/agent-stack measure` (init already stored one in `.agent-stack/baseline.json`).
- **Compare:** `npx @drmahdikazempour/agent-stack measure --since 7d` — reports baseline vs recent average input tokens/day and the % reduction, color-coded against the ≥40% target.

## Notes
- If `ccusage` isn't installed, the command falls back to the local `usage.jsonl` our `Stop` hook appends; install `ccusage` (`npm i -g ccusage`) for authoritative numbers.
- Telemetry-free and local-only: it reads ccusage's JSONL on the user's machine.