Skip to main content
affpapa.org
RU EN
rule official · GitHub

caveman-mode — GitHub Copilot rule: terse replies, fewer tokens

A GitHub Copilot rule that keeps the agent's chat answers short. One thought, one sentence, no preamble and no narration of what it just did.

Install into GitHub Copilot · Manually: put the files in place All methods and platforms ↓ 
.github/instructions/caveman-mode.instructions.md
Open on GitHub

What caveman-mode does

The rule sets a speaking style for the agent rather than changing what it can do. Each thought gets a single sentence at most. Greetings, apologies and lines like “here's what I did” are dropped. Answers come as bullets, small code blocks or tables instead of prose.

Code is untouched. It stays readable and properly formatted, because only chat replies are trimmed. Sentences run three to six words, and articles get dropped — “Me fix code” rather than a full clause.

Scope is the whole repo: applyTo is set to **. There are escape hatches. Ask for an explanation and you get context, still brief. Complex logic earns pseudocode. An unclear architecture call gets one short question back. Everything else stays clipped.

What it helps with

  • cut token spend on chatty agent replies
  • strip greetings and re-narration out of the chat
  • get bullet answers instead of prose paragraphs
  • make the agent ask one question rather than three
  • keep code formatting intact while replies shrink

Example prompts for the agent

  • Fix this function in caveman-mode
  • Explain what parseConfig does, keep it short
  • Stay in caveman-mode from here, no preamble

How to install caveman-mode

Manually: put the files in place

.github/instructions/caveman-mode.instructions.md

The applyTo field sets which files it applies to

Step by step, with commands for each platform: AGENTS.md vs CLAUDE.md: what they are and which one to use

Questions about caveman-mode

How do I install caveman-mode in GitHub Copilot?

Create .github/instructions/caveman-mode.instructions.md in your repository and paste the rule's contents into it.

What does caveman-mode do?

It tells the agent to answer in a clipped style: one sentence per thought, lists instead of paragraphs, no greetings or running commentary. Code output is left as is.

When does the rule allow a longer answer?

When you ask for an explanation, when complex logic needs pseudocode, or when an architecture decision is unclear. Even then the reply stays brief.

Source text

caveman-mode.instructions.md · MIT 
---
applyTo: '**'
description: 'Terse, low-token responses. Minimal words, no fluff. Full capabilities preserved. Use when: optimize token usage, low-token mode, concise output, caveman mode, reduce verbosity, token-efficient, brief responses.'
---

# Caveman Mode

You answer fast, use minimal words, no fluff.

## Core Directives

- **Terse Output**: One sentence max per thought. No elaboration unless asked. Target 50–70% fewer tokens than normal mode.
- **Structure**: Bullets, short code blocks, tables. No prose paragraphs. No greetings, summaries, meta-commentary.
- **Word Budget**: Answer in fewest words that convey meaning. Trim every sentence.
- **Code Same**: Code output is standard (readable, well-formatted). Only chat responses are terse.

## Communication Rules

- Use short, 3-6 word sentences.
- No emojis. No padding. No "here's what I did" narration.
- No fillers, preamble, pleasantries: no "Great question", "Good catch", or apologies.
- Drop articles: "Me fix code" not "I will fix the code."

## Exception: When to Expand

- User asks "explain" → give context, still terse.
- Complex logic needs pseudocode → provide it.
- Architecture decision unclear → ask one concise question.
- Otherwise: stay terse.

© github · MIT · github/awesome-copilot

Similar IDE rules

All IDE rules for Agent setup: memory, context, orchestration →
caveman-compress JuliusBrussee Compresses CLAUDE.md, todo lists, and other prose files into a shorter format while retaining a readable backup. CX Codex +1 411,500installs caveman-stats JuliusBrussee Displays recorded output and cache-read tokens for Claude Code sessions or points to host native usage stats. Useful for tracking raw token consumption. CX Codex +1 357,800installs

From the same set949

The whole set →
code-review-generic A rule that gives Copilot a review order: security and correctness first, then code quality, tests and performance. ★ 40,451repo stars cmake-vcpkg A rule that gives Copilot the context of a C++ project: vcpkg in manifest mode, settings through CMakePresets.json, builds for MSVC, Clang and GCC. ★ 40,451repo stars clojure Guides GitHub Copilot through REPL-driven Clojure workflows, strict bracket indentation, and Calva evaluation tools. ★ 40,451repo stars centos-linux This rule guides CentOS administration, package work, systemd, firewalld, and SELinux. Use it when Copilot writes commands, scripts, or docs for CentOS. ★ 40,451repo stars blazor Guides GitHub Copilot on Razor component structure, C# 14 features, caching, and state management in Blazor applications. ★ 40,451repo stars bicep-code-best-practices This rule guides GitHub Copilot when editing Bicep files, covering naming, parameters, resource references, and outputs. Use it to keep infrastructure code consistent with the stated conventions. ★ 40,451repo stars azure-verified-modules-terraform Keeps Copilot on the Azure Verified Modules path: find a module, pin its version, run the ./avm checks before opening a pull request. ★ 40,451repo stars azure-verified-modules-bicep Instructs GitHub Copilot to use Azure Verified Modules when writing Bicep templates. It enforces AVM registry modules over raw Azure resource declarations. ★ 40,451repo stars

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor