Tech • IA • Crypto
3 Claude Code Concepts 99% of People Don’t Understand
Summary
TL;DR
Mastering the concepts of skills versus commands, hooks, and file structure is essential to effectively using Claude code and optimizing AI-driven workflows.
Key Points
Skills versus Commands
Skills in Claude code are keyword-activated rule sets that run automatically when triggered by specific words in prompts. Commands, however, are manually triggered by users using a slash (/) followed by the command name, allowing for on-demand execution with customizable steps. Skills provide continuous, always-on functionality that can enhance commands or workflows, while slash commands are ideal for creating tailored, step-based processes.
Hooks
Hooks are custom-coded scripts that run automatically under predetermined conditions, enabling dynamic automation beyond basic skills and commands. They represent a more advanced Claude code feature typically used by experienced users to automate repetitive tasks such as renaming files or organizing outputs from integrations (e.g., automatically sorting images created by Nano Banana). Hooks enhance workflow efficiency by triggering specific behaviors without manual input.
File Structure as Claude’s Knowledge Base
File structure functions as the organizational backbone and contextual memory for Claude code operations, effectively acting as "Claude’s brain." Properly structuring folders and files helps define what information Claude accesses during a session, influencing its responses and workflow performance. This structure typically includes general instruction files (like Claude.md), brand-specific contexts, skill and command directories, and administrative folders containing directives about business operations, clients, or projects. Importing a specific file or folder into a conversation scopes Claude’s context to that domain, optimizing output relevance and precision.
Integrating the Concepts
These three core concepts—skills/commands, hooks, and well-planned file structure—form a hierarchy of usability levels in Claude code, from basic keyword triggers and manual commands at level two, to sophisticated automated scripting and file-based contextualization at level three and beyond. Mastery enables the creation of robust AI workflows that combine always-on skills, on-demand commands, and auto-executed hooks, all operating within clearly defined knowledge domains.
Practical Applications
- Skills offer continuous enhancement for prompts and workflows without manual activation, ideal for embedding essential capabilities.
- Commands are suited for specialized, user-initiated processes requiring step-by-step control.
- Hooks automate background tasks such as file management, security protocols, or product testing, saving time and effort.
- A logical and comprehensive file structure allows Claude to maintain situational awareness and apply appropriate rules effectively, enhancing output quality.
Mastering these elements empowers users to unlock advanced Claude code functionality, streamline AI interactions, and build scalable, intelligent automation tailored to specific workflows and business needs.
Full transcript
If you are using Claude code, and you don't understand these three concepts, then you're wasting time and money. And the holy grail of concepts is skills versus commands. When we're talking about skills, they are keyword activated rules. Claude will spot the trigger via keyword, and then run that rule set. And then when we're looking at commands, we trigger those manually. And they're triggered with a slash. So, for example, to see my humanize skill activate, see this prompt? I'm going to let it run. So, we can see the humanize skill was activated off of the keyword trigger. Versus having this slash command. And it's on demand because I added additional steps into the command. So, from my perspective, slash commands are great for custom workflows that you want to build, while skills are great to have ever-present to add to any custom slash command or workflow that you're building. And the second essential concept is hooks. And you can think about hooks as custom-coded scripts that Claude builds that auto-run every time. A level two Claude code user is going to be using skills and commands. But at level three, you're building dynamic hooks. For example, I set up a hook so that every time Nano Banana creates an image for me, it renames them and puts them in the right folder for me. I've set up security hooks, as well as testing hooks when I'm building products. And here's the Nano Banana hook I told you about. Claude is smart enough to know when a hook is needed, so maybe just ask Claude what you can create hooks for. And the most important concept to remember is the importance of file structure. Your file structure with Claude code is your brain in files. This is how you decide what Claude knows in any context. So, my Claude MD, my different brand context, skills, commands. And at deeper levels, you have admin folders for things about your business, maybe your clients, landing pages, some direction files. So, for example, if I reference this file and drag it into the conversation with Claude, it then knows specifically that I want to look at the context of this specific folder. And this is where I host my blog and SEO rules for my website. And at the top level inside the Claude.md file, where my basic instructions live, I teach it how I organize all my files so it knows where to find what. So, yeah, if you want to skip to level three, master these three concepts. And if you want to hit level five, follow me for more, and comment AI agent so I can send you some resources.