PlugWiki

caveman-stats Skill for Claude: See Your Session Token Usage

caveman-stats shows how many output and cache-read tokens your Claude Code session used, split by caveman mode.

Installs
364.9K
on skills.sh
Stars
116.7K
on GitHub
Files
1
besides SKILL.md
Updated
Nov 2026
last repo commit

caveman-stats runs when you type /caveman-stats in Claude Code. A hook in the caveman package reads the session log on disk and Claude prints the generated report as it is. You can add --all, --since 7d or --share.

The report shows recorded output and cache-read tokens, response counts and mode attribution. It does not show tokens saved: SKILL.md says savings are unknown because there is no matching session without caveman to compare. The skill alone is not enough, because the report comes from a hook script, so install the full caveman package.

Good for

  • Claude Code users who want to see real token counts for a session
  • People testing caveman who want honest numbers, not estimates

Install caveman-stats

Claude Code

  1. 1

    Copy this prompt.

    Install the caveman-stats skill in Claude Code. Run this command:
    
    /plugin marketplace add JuliusBrussee/caveman
    /plugin install caveman@caveman
    
    Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Use the plugin, not the single skill folder. SKILL.md says /caveman-stats is served by hook scripts (caveman-mode-tracker.js and caveman-stats.js) that live in the caveman package. The installer needs Node.js 22.13 or newer.

From the project README. Source.

What's in the skill

When it loads

Show recorded output and cache-read token usage and mode attribution for the current Claude Code session, or locate the host's native usage report. Trigger: /caveman-stats.

Key facts

License
Apache-2.0
Installs on skills.sh
364,926 as of 2026-11-02
Last commit to the repo
2026-11-01

What it includes

  • Current session: output tokens, cache reads, responses

    The caveman mode-tracker hook runs a stats script and hands the report to Claude to print verbatim in a code block. Claude does not recompute the numbers.

    Try: “/caveman-stats”

    Runs code · /caveman-stats

  • Adds up the latest count from every recorded session

    Aggregates the latest recorded output count per session.

    Runs code · /caveman-stats --all

  • Limits the totals to a recent window

    Same as --all, for sessions in the last 7 days.

    Runs code · /caveman-stats --since 7d

  • A shareable summary of observed usage

    Reports observed usage with savings marked unknown.

    Runs code · /caveman-stats --share

  • A SessionEnd hook records each session on exit

    Per the bundled README, every Claude Code session writes a snapshot when it ends so --all and --since cover all sessions.

    Runs code · Session snapshot

  • Gemini CLI points to /stats; others say usage is unavailable

    In Gemini CLI it tells you to use /stats model or /stats session. In other hosts it uses a native report if one exists.

    Read · Other hosts

From the official docs

Permissions

Low risk

It installs hooks that run a script and write a snapshot file at the end of every session. Check the hook files in the caveman package if you manage Claude Code settings carefully.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • Hooks run on every session

    Per the bundled README, a SessionEnd hook runs the stats script silently and writes a snapshot. You can see them in your Claude Code hook settings.

  • It reads your session transcripts

    The report is built from the session log on disk. Nothing in SKILL.md says it is sent anywhere.

Technical details
Scopes
  • A hook runs a Node script on /caveman-stats

  • Reads the Claude Code session log on disk

  • A SessionEnd hook writes a usage snapshot when a session ends

Questions

What does caveman-stats do?

It shows token usage for your current Claude Code session: output tokens, cache-read tokens, response counts and which caveman mode was active. The numbers come from the session log, not from Claude's estimate.

How do I install caveman-stats?

Use the install box at the top of this page and pick the full caveman package. SKILL.md says the report is produced by hook scripts in that package, which the single skill folder does not contain.

Does it show how many tokens caveman saved?

No. SKILL.md says savings are unknown, because the transcript has no matching session without caveman. It also says not to infer saved tokens, percentages or dollars from the counts.

Does it work outside Claude Code?

Only the Claude Code report is generated by the hook. In Gemini CLI the skill points you to /stats model or /stats session. In other tools it says usage is unavailable unless a native report exists.

Does it send my session data anywhere?

SKILL.md describes a local script reading the session log on disk. It does not say anything is sent out. Read the hook scripts in the repo if that matters for your setup.

Is caveman-stats safe?

It reads your session log and writes a snapshot at the end of each session, through hooks the caveman package installs. On skills.sh the audits read Gen Agent Trust Hub pass, Socket pass and Snyk pass.

Information

updated Nov 2, 2026

Folder
skills/caveman-stats

juliusbrussee/cavemanskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5
  1. caveman README · github.com
  2. caveman-stats SKILL.md · github.com
  3. caveman-stats SKILL.md · github.com
  4. caveman-stats README · github.com
  5. skills.sh listing and audits · skills.sh

Updated Nov 2, 2026.