Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

caveman

caveman: A skill and proxy that helps AI coding agents say less and read less while keeping code exact.

What problem it solves

Coding agents can consume substantial input and output tokens while processing verbose logs, test output, web pages, and repetitive explanations. The project addresses this by shortening agent responses, compressing data before models read it, and preserving exact code, commands, paths, and error messages.

Key features

  • Concise agent responses — The caveman voice produces direct responses while preserving code, commands, paths, and error messages verbatim.

    Source: README

  • Input compression — Logs, JSON, CSV, YAML, test output, and memory files can be compressed before the agent processes them, with original files retained on disk.

    Source: README, shrink

  • Model and effort routing — Auto selects the model and effort for each request, using a more capable configuration for difficult tasks while retaining the ability to use additional providers.

    Source: README

  • Reusable code-mode blocks — Blocks retain reusable scripts in the repository and direct agents toward existing blocks when they cover a repeated operation.

    Source: README, .blocks

  • Agent-wide installation and controls — The installer detects supported coding agents and provides commands for status, individual feature switches, diagnostics, token history, and setup changes.

    Source: README, installer

Quick start

Copied verbatim from the project's README.

npx @caveman-ai/cli@latest

Who it may suit

This may suit developers who use AI coding agents and want shorter responses, smaller inputs, reusable scripts, or model-routing controls. It may also suit teams evaluating agent token use, answer quality, and cost.

Topics to discuss

  • How should compressed input preserve enough detail for reliable coding decisions?

  • Which compression and routing methods work best across different coding agents and workloads?

  • How can teams measure token savings without compromising answer quality?


Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/JuliusBrussee/caveman#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.

github.comhttps://github.com/JuliusBrussee/caveman

Comments

…