Back to marketplace

Caveman

Makes Claude talk like a caveman, cutting output tokens by ~75% — less filler, same quality output, faster and cheaper Claude sessions.

Productivity / Token Optimizationproductivitytoken-optimization
by AgentVoltv1.0.0Published 1y ago1 install
Free to sign up · every skill included with AgentVolt Pro

Before & after

Same prompt, same agent — the only difference is the skill.

Prompt“Explain API rate limiting and how to handle backoff in production systems”
Beforegeneric AI agent
API rate limiting is an important concept to understand for building resilient systems. When an API provider sets rate limits, they're trying to prevent any single client from overwhelming their servers with too many requests. Different APIs have different limits, which are usually measured in requests per second or requests per day.

To handle rate limiting in production, you need to implement backoff strategies. The most common approach is exponential backoff, where you wait progressively longer between retry attempts. For example, you might wait 1 second after the first failed request, 2 seconds after the second, 4 seconds after the third, and so on. You should also respect the Retry-After header if the API provides one.

Another consideration is request queuing. Instead of firing all requests at once, you can use a queue to space them out over time. This is especially important if you're processing large batches of data. You might also want to implement circuit breakers, which automatically stop sending requests to an API if it's returning too many errors.
Afterwith Caveman
API RATE LIMITING - PRODUCTION NOTES

What it is: Server says "hold on, you're hitting me too fast." Response: Wait, then retry.

How to handle:

1. Exponential Backoff
   Try #1 → Wait 1s → Try #2 → Wait 2s → Try #3 → Wait 4s
   Cap: Don't exceed 60s per retry

2. Respect Retry-After Header
   Server says "retry in 120s" → You wait 120s, not your own math

3. Queue Requests
   Don't fire 1000 requests at once. Drip them: 10/second max.

4. Circuit Breaker
   If 10 consecutive failures → Stop. Wait 5 min. Try once. If fail, wait 5 min more.

Code shape (pseudocode):
   function callWithBackoff(url, attempt=0) {
     try { return fetch(url) }
     catch error {
       if (attempt > 5) throw error
       wait(2^attempt seconds)
       return callWithBackoff(url, attempt+1)
     }
   }

Trap: Never assume API limits are permanent. Monitor daily—they change. Log every 429 status to spot patterns.

About this skill

name: caveman description: Use when Makes Claude talk like a caveman, cutting output tokens by ~75% — less filler, same quality output, faster and cheaper Claude sessions.

Caveman

Caveman solves Claude's verbosity problem. Claude loves to explain every tiny step like it's writing a thesis. Caveman strips all that away with a communication style that eliminates filler and yap. The creator claims ~75% token reduction while maintaining output quality. Less mental clutter, faster sessions, lower usage costs. Sounds like a meme but solves a real problem.

What you get

  • ~75% token reduction on Claude outputs, faster response times, lower usage costs, same output quality without the step-by-step explanations.

Customize your output

  • Adjust how extreme the caveman-speak is. Use minimal mode for light brevity or full caveman for maximum token savings.

Example output

Instead of: "I'll now inspect the relevant files to understand the current implementation" → Claude says: "Me check files."

Best for

Power users watching usage limits, anyone annoyed by Claude's verbosity, developers wanting faster and cheaper sessions.

SKILL.md preview

SKILL.md
---
name: caveman
description: Use this skill when the user wants Claude's responses stripped of filler, cutting output tokens by roughly seventy-five percent while keeping the same underlying quality.
version: 1.0.0
category: Productivity / Token Optimization
author: AgentVolt
license: proprietary
tags:
  - productivity
  - token-optimization
  - standard
---

# Caveman

Strips Claude's default explain-everything verbosity down to short, direct statements, cutting output length substantially while keeping the actual content and correctness intact.

## When to use

… (sign up to view the full skill)
Sign up to view, copy, and install the full skill