Skip to content
Get packs

Caveman Optimize Evaluator

Walks an agent through a cautious, approval-gated test of one Caveman optimization observation

Source Download skill
Repository
JuliusBrussee/caveman
Stars
116,835
Licence
Apache-2.0
Files
1 file

What it does

This skill tells an AI coding agent how to review a report-only observation from the Caveman CLI, a tool that tracks how an AI setup uses tokens. The agent lists the observations, asks you to pick one, then looks for a specific place in your code that could cause the pattern. It proposes one small change and a fair before-and-after test, and only edits code after you approve. It is deliberately strict: it does not claim dollar savings, and it reverts the change if quality drops or results are unclear.

Great for

  • Developers already using the Caveman CLI who want to test a token-efficiency idea safely
  • Teams that want a documented, evidence-first process before changing agent prompts or tool setups

Skip it if

  • Anyone without a logged-in Caveman CLI session, since the workflow stops at its first step
  • Users who want quick automatic cost cuts, because the skill requires explicit approval at each stage and never claims savings
  • Repositories without fixed test fixtures or quality checks, since the skill stops rather than guess at a fair comparison

Try asking

  • “Show me the Caveman optimization observations for this repo and let me choose one to evaluate”
  • “Find the callsite that could cause the context-window-profile observation and propose a minimal candidate”
  • “Set up a paired baseline and candidate eval for the tool-output-size observation I picked”
Step-by-step guide: how to use Caveman Optimize EvaluatorInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.

    caveman-optimize.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip caveman-optimize.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip caveman-optimize.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Read the observations yourself before choosing; the skill will not rank them for you.
  • Make sure your repo has a fixed fixture and quality check first, or the skill will stop and name what is missing.
  • Expect the agent to revert its own edit if the paired result is inconclusive; that is intended behaviour.

What's inside

  • SKILL.md

Redistributed unmodified from JuliusBrussee/caveman under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.