caveman-stats

Report actual Claude Code token usage from the JSONL session log.

Updated Jul 2, 2025

Search Tags:#claude code #statusline #token usage #cost savings #hook #session log

Ready to run

npx skills add https://github.com/junechen7414/SpringBoot --skill caveman-stats-junechen7414

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: caveman-stats
Source: https://github.com/junechen7414/SpringBoot/tree/main/.agents/skills/caveman-stats
Command: npx skills add https://github.com/junechen7414/SpringBoot --skill caveman-stats-junechen7414

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill removes guesswork from session cost tracking by reporting the actual Claude Code token usage for the current conversation, along with an estimated baseline comparison.

Core Features & Use Cases

Real usage reporting: Reads the live session log on disk and shows actual input and output token counts.
Savings estimation: Calculates estimated savings versus a non-caveman baseline so you can understand efficiency gains.
Hook-driven delivery: Integrates with the mode-tracker hook to intercept /caveman-stats and return formatted stats immediately.
Statusline support: Writes a lifetime-savings suffix used by the statusline badge for quick at-a-glance feedback.

Quick Start

Ask the assistant to show the current session's caveman stats and savings.

Frequently Asked Questions about caveman-stats

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I check real token usage in a Claude Code session?▼

You can check real token usage in a Claude Code session by triggering the /caveman-stats command, which reads the on-disk JSONL session log to report actual input and output token counts without relying on model-side estimation.

How do I monitor Claude Code cost savings during a live session?▼

Monitoring Claude Code cost savings during a live session is done by comparing actual token usage against an estimated non-caveman baseline, returning formatted efficiency metrics through the mode-tracker hook.

Can I display token savings in the Claude Code statusline?▼

Yes, you can display token savings in the Claude Code statusline by writing a lifetime-savings suffix to the statusline badge, which provides quick at-a-glance feedback on your efficiency gains.

Does the token usage tracker require any external dependencies or components?▼

No external dependencies or components are required. The token usage tracker operates independently by reading the on-disk JSONL session log and integrating directly with the mode-tracker hook to return formatted stats.

How do I get a baseline comparison for Claude Code token efficiency?▼

To get a baseline comparison for Claude Code token efficiency, the tracker calculates estimated savings against a non-caveman baseline using the actual token counts read directly from your live session log.

Why does the session cost tracker read the on-disk JSONL log instead of asking the model?▼

The session cost tracker reads the on-disk JSONL log to ensure accurate efficiency monitoring, bypassing model-side estimation to deliver precise actual input and output token counts for your current conversation.