Your AI Agent Doesn’t Need More Prompts. It Needs Skills!

A developer highlights the use of 'Skills' as a solution to repetitive prompting in AI coding agents, describing a folder-based standard that works across tools like Claude Code, Cursor, and Copilot. The post explains how to create skills via SKILL.md files, tooling interfaces, or marketplaces like skills.sh, and warns about token waste and unused skills due to poor descriptions. The author shares personal experience with the 'explain-check' skill, a customized version of Geoffrey Litt's 'explain-diff'.

Tired of explaining the same things again and again to your AI Agent? Frustrated because the AI keeps forgetting minute things custom to your codebase which needs to be kept in mind in each change? This is the current scenario for most people using AI agents to build their software. You handoff a task to it, it gives back the solution but misses something. You explain that to it, it nods back and then does it again. I myself did it until i came to know about Skills. Remember the CONTRIBUTING.md file we find in almost every open source repository? The file which explained anyone coming to the repo what to check, understand and keep in mind when contributing to it so that you don’t break it. The Skills works like that for any AI Agent who is going to make changes in your codebase. Its a folder that your AI checks anytime it needs to perform a specific task, specialized jobs or multi-step workflows without requiring you to prompt every time. And the best thing is, it follows an open standard that works with almost every AI agent be it Claude Code, Cursor, Copilot and more. It follows a folder-based structure around a SKILL.md file containing YAML metadata about that skill and instructions for that in markdown. Skills can vary from simple instructions to multi-step workflows depending on your need and there are 3 ways limited by my knowledge to build a skill: First you need to create a dedicated folder for your skill and place a SKILL.md file inside it. This file needs to have 2 things: Below is a basic sample SKILL.md file for your reference: — - name: word-counter description: Counts the total number of words in a given text. — - Word Counter Instructions - Take the user’s input text. - Count the total number of words. - Return only the final word count as a number. It is a tooling interface command-line or script which can take your requirements and build a Skill for you accordingly. Its like talking through what you want and having it make the file for you. Most AI Agents like Claude, Cursor and Gemini have one of their own but their are also others like Crew CLI and more. This is way i have used the most. So there are marketplaces like “skills.sh”. Just like we install libraries from NPM, this is a registry from where we can browse and install the skills we need from the numerous options available. I have tried the ‘explain-diff’ by Geoffrey Litt and the famous ‘matt-pocock’ skills and they both have helped me really well. You can check out here https://www.skills.sh/ https://www.skills.sh/ Here is the part that nobody mentions about Skills. There is a specific way your skill should be written. It is not something you can turn on/off to be used everytime, its your AI agent who decides when to use it. Now, there are 2 main friction points here: As you don’t control the use of your skill, you may have just created a skill but your AI will never use it. It is because the description of your skill doesn’t match the way you converse with your agent. So the agent cannot map your chat to the description and know exactly when to use the skill. Irrelevant skills getting loaded everytime and silently eating away your tokens. This is even worse, as without even you knowing it scrapes away your tokens resulting in rising costs. This can be caused by a description too generic causing it to be loaded almost in every session. I came to know about these when my skills weren’t getting used and i had to dig inside why. So once i got the hang of it, skills started working like a charm. The skill I use the most is something called “explain-check”. It is a customized version of the one by Geoffrey litt which can explain any feature made in HTML with quizzes. Mine also has questions to check my understanding but not quizzes, a predefined paragraph of answer. I believe the potential for skills can huge and using them effectively can make your work even polished with AI. So if a friend asked me if skills are worth it, I would say “Yes” without any hesitation. Coming back to my problems stated in the beginning of this article, skills did fix my repeating problem. But just like any other software, it needs understanding. You cannot vibecode any skill and expect it to work perfectly as you need. Its not something you set and forget, you have to create it how you work, how you chat with AI agent and what you exactly need to achieve from that skill. Thank you for reading If you have any feedback or notice any mistakes, please feel free to leave a comment below. I’m always looking to improve my writing and value any suggestions you may have. If you’re interested in working together or have any further questions, please don’t hesitate to reach out to me at fa1319673@gmail.com mailto:fa1319673@gmail.com .