IBM Bob Shell + Skills.sh: Supercharging Your Personal AI

If you use **Bob Shell (IBM assistant) daily, you probably already know it’s a massive help for coding. But here’s the reality: most people are only scratching the surface. Bob has a capability that isn’t talked about enough, but it changes everything: Skills**.

We’ve all been there: you try to install a new skill and end up with .bob/skills folders scattered across all your projects like confetti. It’s an organizational nightmare. Today, I want to show you how to move from a "standard" Bob to a truly powerful one by leveraging the entire skills.sh ecosystem. We’re going to centralize its abilities into a Universal Path (.agents/skills) so Bob can work with the same freedom and power you’d see in cutting-edge projects like **OpenClaw**.

1. The Trick: One Path to Rule Them All

The philosophy behind **OpenClaw** is that an agent shouldn’t be limited by its immediate environment. If you install a skill to query a database or automate a report, why shouldn’t it be available to Bob, your Warp terminal, and Cursor all at once?

To achieve this, forget about symlinks (symbolic links). They often fail, break permissions, and are a headache in the shell. The clean way to do it is by modifying Bob’s configuration file directly to integrate into this universal standard.

How to do it:

Find your global configuration file at ~/.bob/settings.json (create it if it doesn't exist) and add this:

{
  "includeDirectories": [
    "/Users/your_username/.agents/skills/"
  ]
}

With this small change, Bob now has “extended radar.” It’s no longer just looking in its small local backpack; it’s scanning everything you put into that global “tool shed,” adopting the modular structure that makes OpenClaw-style agents so potent.

2. Exploiting Skills.sh

This is where things get serious. Many aren’t familiar with **skills.sh**, but it’s essentially the standard for giving “hands” to AIs. Instead of writing code from scratch so Bob learns how to read Jira or clean logs, we use this ecosystem to inject proven capabilities from the community.

To get the most out of it, you don’t need to clone repositories manually; just use npx and let the magic happen:

npx skills add [skill-name]

Why does this make Bob much more powerful? By choosing the Universal (.agents/skills) option during installation, you aren’t just copying files; you’re integrating Bob into a network of tools already used by agents like Amp, Cline, Claude Code, Gemini CLI. You are turning Bob into an agent capable of executing complex tasks on your system, giving it a level of autonomy it simply didn’t have before. It is, quite literally, taking the handcuffs off the AI so it can start working under the same action logic as OpenClaw.

A Security Warning

By enabling this level of integration and autonomy, you are granting “execution permissions” to the AI. A skill installed via **skills.sh** is code that can run commands in your terminal and touch your files.

Don’t install blindly:

• Trust but verify: Only add skills from known authors or repositories.
• Human-in-the-loop: Bob is much more autonomous now, but you are still the boss. Always review what command it plans to execute before giving the green light.
• You hold the control: The AI proposes the action, but the final “Enter” is always yours.