Does shortening responses hurt technical accuracy?

No. The plugin keeps all technical accuracy while responses are written in clipped shorthand.

How much does it reduce output tokens?

Roughly 65%, based on measurement.

caveman

PackDev tools

Caveman is a plugin that makes an AI agent write its responses in clipped, caveman-style shorthand. The result is roughly 65% fewer output tokens, measured, while technical accuracy is kept intact. It suits anyone who wants shorter agent replies without losing correctness.

Unavailable. Delivery for this kind is on the roadmap — not serving yet.

Have an agent setup that supports plugins.

What your AI can do with it

Switches agent responses to clipped caveman-style shorthand
Cuts output token usage by roughly 65% (measured)
Keeps all technical accuracy in shortened replies

Getting started

Have an agent setup that supports plugins.
Add the caveman plugin to it.
Enable the plugin so the agent's responses use the shorthand style.

Signals

GitHub stars: 108k
Forks: 6k
Last commit: Sep 2026
Hacker News mentions: 20

Questions

Does shortening responses hurt technical accuracy?: No. The plugin keeps all technical accuracy while responses are written in clipped shorthand.
How much does it reduce output tokens?: Roughly 65%, based on measurement.

Advanced

Item type: plugin
Key: juliusbrussee-caveman-caveman
Source: github.com/juliusbrussee/caveman

github.com/juliusbrussee/caveman