Quick Overview
Claude Code is Anthropic's coding agent. You can ask it to edit files, fix bugs, build features or create a simple webpage from the terminal.
A skill is a set of reusable instructions that changes how Claude approaches or presents a task. Installing one does not give Claude a new model or make it more intelligent. It changes its working style.
Caveman and I-Have-ADHD are both communication skills. Their job is to make Claude's output less frustrating to read.
Caveman is primarily a compression skill. It changes how much Claude says and how densely it says it.
I-Have-ADHD is an actionability skill. It changes what information appears first and how Claude helps you continue the task.
That’s an important distinction because shorter output is not always easier output.
What You’ll Learn?
What Caveman and I-Have-ADHD change in Claude Code
Why they solve different problems despite looking similar
What happened when we gave both the same build task
Which skill fits autonomous versus interactive work
How to test them without changing your workflow
Caveman vs I-Have-ADHD
Both skills reduce unnecessary commentary, but they solve different problems.
Caveman | I-Have-ADHD | |
|---|---|---|
Main goal | Reduce words and output tokens | Make responses easier to act on |
How it works | Removes filler, hedging and unnecessary narration | Leads with the action, numbers steps and gives a clear next move |
Output style | Compressed and terse | Structured and action-first |
Best for | Long runs, batch jobs, work you check after Claude finishes | Interactive work, debugging, projects you leave and return to |
Controls |
| One main mode |
Cost | Free | Free |
The simplest distinction is:
Caveman reduces how much Claude says.
I-Have-ADHD reduces how much effort it takes to figure out what to do next.
What Our Test Results Showed
We ran an experiment pitting these two popular Claude Code skills in three fresh runs: build a one-page coffee shop website in a single HTML file.
Run | Mid-build chatter | Final summary | Total words |
|---|---|---|---|
Control | 98 | 283 | 381 |
Caveman | 26 | 191 | 217 |
I-Have-ADHD | 0 | 120 | 120 |
Caveman cut total assistant output by 43%. I-Have-ADHD cut it by 69%.
More importantly, all three runs produced a working page. The shorter outputs did not produce obviously worse execution in this test.
Caveman removed most of the play-by-play but kept a substantial final summary. I-Have-ADHD removed the mid-build narration entirely, completed the work, checked it and reported once.
This was one run per setup, so treat the percentages as a directional result rather than a benchmark. Both skills are prompt-driven, so results can vary.
The simplest rule from the test is:
Machine reads it: use Caveman. You read it: use I-Have-ADHD
If your main problem is... | Try |
Too many output tokens | Caveman |
Long autonomous runs | Caveman |
Batch work you check later | Caveman |
Not knowing what to do next | I-Have-ADHD |
Losing your place between sessions | I-Have-ADHD |
Following the work interactively | I-Have-ADHD |
The two can also be combined: I-Have-ADHD for structure, Caveman for compression. Our test did not measure that setup, so treat it as an experiment rather than a proven recommendation.
How To Install and Try Them
You need Claude Code installed first. Then open Terminal, which is the same app you use to start Claude Code.
Refer to our guide on How to Install a Claude Skill from GitHub to understand how to install any skill.
Install Caveman
Paste this into Terminal and press Enter:
claude plugin marketplace add JuliusBrussee/caveman && claude plugin install caveman@cavemanAfter it finishes, start Claude Code and type:
/caveman liteStart with lite. It removes filler while keeping normal sentence structure, so it is easier to read than the stronger modes.
Install I-Have-ADHD
Paste these commands into Terminal one at a time:
claude plugin marketplace add ayghri/i-have-adhdclaude plugin install i-have-adhd@i-have-adhdThen start Claude Code and type:
/i-have-adhdThat activates the skill.
Test it on a real task
Choose something you already understand. Run it once with normal Claude Code and once with the skill enabled.
Compare three things:
how much you had to read,
how quickly you understood the next action, and
whether you still had enough information to trust the result.
You can also try one skill for a few real sessions before testing the other. For instance, if you like I-Have-ADHD's structure but still want shorter replies, you can later experiment with combining it with a lighter Caveman mode.
This combination was not tested in our experiment, so treat it as something to try rather than a proven setup.
What These Two Skills Cannot Do
Neither skill improves Claude's reasoning or guarantees better code. They change presentation, not the underlying model.
They also cannot guarantee a fixed amount of token savings. Different tasks and longer conversations can produce different results.
Caveman and I-Have-ADHD both make Claude Code less verbose, but they solve different frustrations.
Choose Caveman when you want less narration. Choose I-Have-ADHD when you want clearer actions. Run the same task with and without the skill, compare the experience, and keep the one that makes Claude easier for you to work with.
If you want more practical guides for using AI at work without turning your job into a collection of prompts, subscribe to Practicaly AI.
How useful was this guide for you?
💌 We’d Love Your Feedback
If you need any guidance while implementing this, or if something isn't quite clear, feel free to ping the team. We're here to support you and clear things up.
Until next time,
Team PracticalyAI
