SkillRepo
Sign InSign Up
Catalog
juliusbrussee/caveman-stats

juliusbrussee

caveman-stats

Show real token usage and estimated savings for the current session. Reads directly from the Claude Code session log — no AI estimation. Triggers on /caveman-stats. Output is injected by the mode-tracker hook; the model itself does not compute the numbers.

v1.0Latest
New~274Updated Aug 17, 2026

This skill is delivered by hooks/caveman-stats.js (read by hooks/caveman-mode-tracker.js on /caveman-stats). The model does not need to do anything when this skill fires — the hook returns decision: "block" with the formatted stats as the reason. The user sees the numbers immediately.

Output also includes Est. rule overhead and Est. net lines wherever a savings estimate exists with a known turn count. Rule overhead is the estimated per-turn INPUT-token cost of the injected caveman rules (default 1,250 tokens/turn, override with CAVEMAN_RULE_OVERHEAD_TOKENS) times the turn count. Net is savings minus that overhead — when negative, the output says so plainly and suggests turning caveman off for that workload, rather than hiding the net-negative regime behind a gross-savings number (see docs/HONEST-NUMBERS.md).

Files2
2 files · 3.5 KB

Select a file to preview

Overall Score

82/100

Grade

B

Good

Safety

85

Quality

84

Clarity

88

Completeness

78

Summary

The caveman-stats skill displays real token usage and savings data from the current Claude Code session by reading directly from the session log. It is triggered via the `/caveman-stats` command, intercepted by the mode-tracker hook, and returns formatted statistics as a blocked decision—the model itself does not compute the numbers. Output includes actual input/output tokens, estimated baseline, savings percentage, rule overhead cost, and net savings (accounting for the injected rule cost).

Detected Capabilities

session log file readhook-driven output injectionenvironment variable read (CAVEMAN_RULE_OVERHEAD_TOKENS)lifetime-savings file writejson data parsingarithmetic computation (token deltas, percentages)

Trigger Keywords

Phrases that MCP clients use to match this skill to user intent.

session token usagecaveman mode savingstoken cost analysisrule overheadnet cost benefit

Risk Signals

INFO

Environment variable read for configuration override (CAVEMAN_RULE_OVERHEAD_TOKENS)

SKILL.md and README.md
INFO

Direct session log file read from disk

README.md: 'Reads the current Claude Code session log directly'
INFO

Lifetime-savings file write for statusline badge

README.md: 'writes a lifetime-savings suffix file'

Use Cases

  • >Display real session token usage without AI estimation
  • Calculate actual token savings from caveman mode activation
  • Assess net cost-benefit of caveman rules for the current workload
  • Monitor per-turn rule overhead impact on token efficiency
  • Determine whether caveman mode is beneficial for terse vs. verbose reply patterns

Quality Notes

  • Documentation is clear and practical: explains what the skill does, how to invoke it, and provides a realistic example
  • Honest cost-benefit framing: the skill explicitly reports when caveman mode net-negative and recommends turning it off, rather than masking true cost
  • Good referential structure: cross-links to SKILL.md for mechanics and docs/HONEST-NUMBERS.md for methodological background
  • Limitations well-stated: savings estimates only unambiguous with single benchmarked mode and known turn count; mixed or unattributed spans are excluded from net calculation
  • Edge case handling: correctly handles short, terse sessions where output savings don't cover input rule cost
  • Potential gap: README does not document how to interact with or configure the hook system itself; assumes hooks/caveman-stats.js and hooks/caveman-mode-tracker.js are already in place
Model: claude-haiku-4-5-20251001Analyzed: Aug 17, 2026

Reviews

Add this skill to your library to leave a review.

No reviews yet

Be the first to share your experience.

Use juliusbrussee/caveman-stats in your dev environment

Start a Developer trial

Command Palette

Search for a command to run...