Las 7 SKILLS de ChatGPT Work que uso cada día (CÓPIALAS)

0:00If I ask ChatGPT how a mission to Mars changes when I modify the ship's speed, payload, or launch date, it might give me an explanation like this. It's correct, but in the end, it's still just a wall of text. Now I select a skill and pose the same problem, and I get this: an interactive simulator with

0:17Earth and Mars moving around the Sun. Controls to change each variable and metrics that update in real time. It's the same ChatGPT. The difference is that for the second task, I indicated to it via a skill how I wanted it to approach the work. In this video, I’m going to show you seven ChatGPT Team skills that help to visualize ideas, verify information before publishing, move projects between different AIs, repurpose content, and automate multi-step processes. And something important: all the skills you're going to see are below the video. I have prepared a practical guide in PDF. For those that are already official, I explain where to find them and how to

0:56use them. And for the custom ones, you have the complete prompt to create them , the test prompt, and all the demo files. You won't need to pause the video to copy anything. You download the resource, follow the steps, and you can end up with the same system running on your computer. A skill is basically a set of reusable instructions. Instead of explaining the same process to ChatGPT every time, you teach it once, save it, and then you can run it again whenever you need it. That saves time, but the major advantage is something else: it makes the result much more consistent. GPT remembers the order of

1:35the steps, the format it must deliver, which sources to prioritize, and what things it cannot do without asking you. Before we see them, I'll show you where they are. To start using them, go into the ChatGPT desktop application and click here on the left where it says plugins or add-ons. Then, at the top, go into skills. Here you will be able to see all the skills you already have installed. There are some that come pre-installed by default and others that you can add yourself. And to add

2:02new skills, you have several options. For example, you can click on the top right where it says add. But the way I recommend for creating a new skill is simpler. Simply open a new chat. Make sure you are within ChatGPT Team and describe directly to it what skill you want ChatGPT to create for you. ChatGPT will take care of creating it and saving it as a new skill so you can use it again whenever you want in the future. And now that you know how to create a skill, I'm going to show you some of the ones I use practically every day. The first skill is one of my

2:42favorites, and you probably haven't used it yet. Even though it comes pre-installed in ChatGPT, it's called Visualize and it turns responses like this into something like this. To find it, we open a new task and type a forward slash. This command allows us to search for any of the skills we already have installed. I start typing "visualize," select it, and from that moment on, ChatGPT knows it shouldn't give a conventional response, but rather a much more visual and interactive representation. To check the difference, I first asked ChatGPT how a spacecraft's speed, payload, and launch window affect a mission between Earth and Mars. Without using any skill

3:23, it returns a correct explanation with different scenarios and various data, but it's still a block of text that I have to read and interpret. Now I'm going to perform the same query using this skill. And look at the difference. Instead of another written explanation, it has created an interactive control center to simulate the mission. On the left, we can see the trajectory around the Sun. Earth and Mars move while the ship follows its route, and I can pause the animation or manually scroll through the different days of the journey. On the right, I can choose between three modes: efficient, balanced, and fast. I can also modify

4:01the speed, payload mass, and launch window. Every time I change a variable, the simulation recalculates the entire mission. For example, in efficient mode , the trip takes 290 days and requires less fuel. If I select fast mode, it drops to 181 days, but consumption increases and the difficulty is also greater. And further down, we have a timeline and a direct comparison of the three scenarios. In seconds, I can understand the cost of getting to Mars sooner without having to manually

4:29compare several paragraphs. Additionally, it has included the formulas, assumptions, and limits of the model. Obviously, it's an educational simulation and not a real aerospace planning tool, but it works perfectly to understand the concept. And this is what's really interesting about this skill. It isn't limited to decorating a response with a chart. It can convert information into simulations, interactive dashboards, maps, diagrams, comparators, or timelines. If you are a visual person, try this skill because it is one of the functions hidden inside ChatGPT that can completely change the way you understand complex information. The second skill solves a less visually spectacular problem, but it's probably more important. Publishing information

5:16that seems correct but no longer is. When I prepare a video about artificial intelligence, prices, dates, limits, compatibility, or features appear that can change from one week to the next. Asking ChatGPT to review the text doesn't always work because it might focus on the style and overlook the very claims that should be checked. That is why I created Verific. This skill first extracts all verifiable claims, separates facts from opinions, and prioritizes what could make the video outdated. To test it, I have prepared a draft with intentional errors that includes correct claims, others that need nuance, and some that are outright false. So I attach the document, invoke the skill, and ask it

5:58this. Review the attached file before I publish it. Check all claims regarding availability, operation, pricing, and compatibility. Prioritize official documentation and return the errors that could make the video outdated to me first. Then, give me the corrected sentences ready to replace. And here we have the result. It has extracted all verifiable claims from the draft and organized them into an audit table. For each sentence, it gives me the verdict, the evidence it found, the source used, the risk level, and, most usefully, an exact correction ready to replace the original text. For example, the draft stated that ChatGPT only works from the browser. The skill verifies that it is

6:41also available on compatible mobile and desktop applications. It marks the sentence as incorrect and gives me the corrected version directly. It also detects claims that are not completely false but need context. In those cases, it classifies them as needing nuance and explains exactly what condition or limitation we should add. Furthermore, it doesn't just stick with the first result it finds. For each check, it tries to use official documentation and leaves me the link along with the date it performed the query. And after the entire audit, it gathers only the paragraphs I need to replace. I don't need to rebuild the script. By looking at the table, I can simply copy these

7:20corrected versions and replace them directly. Finally, it separates the doubts it could not resolve. In this case, for example, it warns that some prices or limits may depend on the plan , contract, and workspace configuration . This is useful for reviewing scripts, commercial proposals, presentations, or training materials, especially when they contain prices, dates, compatibility, or features that can change over time. And this is indeed a personal skill. And remember that for this skill, I leave you the full Prompt to install it under the name Verifica and use it with your own documents. By the way, if you have an idea for a website but don't know how to code or

7:57don't want to pay an agency and wait weeks, watch this. This video is in collaboration with Hostinger, a platform already used by more than 5 million customers in 150 countries. And I am going to create a complete website using only artificial intelligence. We enter through the link that I will leave below in the description and here we can compare the different plans and check exactly what each one includes. I am going to choose the Unlimited plan, which in Spain is listed from € 3.79 per month. And before paying, I enter the code Migena10 to apply the discount . If you notice, that's nearly € 20 in extra savings if you use the link

8:33below in the description. I am going to create a website for an AI automation studio. I explain the business, the sections I need, and the visual style. And in a few minutes, we will have a fully functional first version. We can choose between two ways of working. Manual mode allows you to control the design directly, while agentic mode lets you simply describe what you need

8:54and the AI builds it for you. Furthermore, we don't have to settle for the first result. We can keep talking to the AI and ask it to improve the header, change the colors, or add a contact form. We can also incorporate payments. an online store, email marketing, or even an AI chatbot, all from the same platform and without

9:13connecting external tools separately. Finally, we check how it looks on a computer and on mobile. We connect the custom domain that is included for free for the first year and click publish. If you want to try it, go through the link in the description and use the code Miguena10. The Unlimited plan starts at € 3.79 a month, has a 30- day money-back guarantee, and you can cancel whenever you want. Now, let's continue with the video. If you use ChatGPT along with Claude, Gemini, or other models, you have surely had this happen: you start a project in one artificial intelligence and after making significant progress, you want

9:51to continue in another because it might solve a specific part better. That is what this next skill, which I've called relay, is for. Relay analyzes all the work done and prepares a package with the project's objective. What is already finished, the decisions made, the necessary files, the pending problems, and the exact next step. Instead of starting from scratch in Claude and explaining how everything works again, I am going to execute the relay skill within this same task. To show it to you, I have created this small 3D video game. We control a ship inside a space tunnel. We can dodge obstacles, collect energy, and increase our score. Visually it works well, but

10:30there are two errors I haven't been able to fix. On the mobile phone, the touch controls overlap with the interface, and after restarting several times, duplicate ships appear. Instead of starting from scratch in Claude and re-explaining how everything works, I'm going to run the relay skill within this same task and ask it to prepare the project for me to continue in Claude. The game works, but I need to fix those issues appearing in the mobile version. And here we have the relay prepared. It has created a complete package so that Claude can continue the project. Inside, it includes a file called relaycl.md where it explains what we are building, which

11:07parts already work, which decisions to keep, and what the two pending errors are. At the end, it generates this instruction ready to copy and paste, asking it to fix the mobile interface and the appearance of duplicate ships, but also to check the game in portrait and landscape, test simultaneous controls, and perform at least 20 restarts. Now I download the package, open Claude, attach all the files, and paste this instruction. With this, Claude already knows the objective, the project status, the pending errors, and the tests it must perform. I don't have to reconstruct the entire conversation

11:42or explain how the game works again. And here we have the result. Claude did indeed reproduce the control issue. Some horizontal phones would disappear completely, and several interface elements would overlap. Then, it adapted the controls for phones and tablets, corrected the overlaps in portrait and landscape, and checked simultaneous movements with two fingers . Finally, it provides me with two versions: the complete project in a ZIP file and a single HTML file that, in theory, can run locally without installing anything. This demonstrates another advantage of a good relay. The second artificial intelligence doesn't just receive an order; it also receives the context, the files, the observed errors, and specific criteria to verify

12:23the result. And if I want to continue in ChatGPT later, I can create another relay from Claude and repeat the process in the opposite direction without losing all this work. The next skill solves another quite annoying behavior of ChatGPT, which is that you give it a still poorly defined idea, and it starts researching or building before understanding exactly what you need. To avoid this, there is "Question Me." When you use it, it forces ChatGPT to interview you before starting. The goal is to discover missing decisions, the assumptions it is making, and any

13:00detail that could change the result. And it doesn't start working until we both agree on the assignment. The original idea for this skill was developed by Mat Pocock. His version conducts a fairly intense interview, so this adaptation is designed to reach the same point more quickly. I am going to give it a deliberately open-ended request. I want to create a tool that analyzes my videos and tells me what I should post next. Instead of starting to build a random application, it stops and asks me five very specific questions. First, it wants to know where I post and which metric I want to improve. Then, it asks me what exactly

13:35"analyzing my videos" means, whether it should study the content, the statistics, or both. It also needs to know the channel's theme, what output I want to receive, and the scope of this first version—a tool where I upload the data, or an application connected directly to my accounts. Additionally, within each question, it offers several examples so I can respond faster, and if I don't care about something, I can tell it to use its own judgment. My version is limited to five main questions and cannot start researching or building anything yet. When I respond, it will summarize the objective, the deliverable, and the decisions made so that I can confirm it

14:13understood correctly. Only after receiving my approval will it be able to start working. This way, we add a short pause at the beginning, but we avoid wasting much more time correcting something that ChatGPT started creating without fully understanding the task. The next skill is also new and serves to convert a single source into several different materials. We can give it a meeting and get tasks, an email, and a summary; convert a report into a presentation and an action plan; or transform a class into a complete study system. For example, here I have the transcript of a 28-minute class on how

14:50memory works and how to study better. Normally, I would have to reread it, prepare notes, create an outline, write flashcards, and design an exam to check if I have actually learned the subject. Well, with this skill, I can generate all those materials from a single source. I attach the transcript, select the skill, and use this prompt. Convert the attached class into a complete study package. Create well-organized notes, a one-page visual outline, 30 study cards, a practice exam, and a

15:21separate document with the solutions. Also add a 7-day review plan. And there we have the result. From a single transcript, it has created a complete study package with all the files organized inside a ZIP. First, we have four pages of structured notes. it has separated the concepts, the relationships, and the examples, while also keeping timestamps to locate each explanation within the original lecture . Then, it creates this one-page visual outline titled "From Understanding to Remembering." Here we can quickly see the entire process: preparing for understanding, retrieving information, spacing reviews, interleaving concepts, and connecting what has been learned. I have also generated 30 study flashcards in CSV format, each with its question,

16:06answer, difficulty level, and reference to the lecture. The next file is a practice exam with 14 questions and no answers. The solutions appear in a separate PDF and include both the correct answer as well as an explanation and the grading criteria. And finally, it prepares a 7-day plan that distributes the notes, flashcards, and exam into concrete, adjustable study sessions. And think about the real value of this. You can record any class, always with permission, or pull the transcript from an educational YouTube video, attach it, and use the skill to automatically convert it into notes, outlines, study cards, exams, and a complete review plan. From a single source, you move to having an

16:49entire system to understand, study, and grasp what you have learned. The next category is multi-step skills. They are useful for those repetitive processes where we always open the same files, check the same information, and perform the same actions in the same order. To create them, I have prepared " CreateFlow." You explain a process you repeat, it asks you the necessary questions, and it turns it into a new reusable skill. I am going to use reviewing a sponsored video as an example. Normally, I have to open the brief, check its requirements against the script, correct any non-compliance, retrieve the review link, prepare the email, and update the tracker. Instead

17:30of explaining all these steps every time, I am going to turn the entire process into a skill. Since the process is already sufficiently clear, CreateFlow doesn't need to ask me any more questions and moves directly to designing the skill. It organizes the work into six steps. First, it will review all the files without modifying anything. Then it will show the non-compliances and the proposed corrections, and will wait for my

17:52approval before applying any changes. It also makes it clear that it will review the script again, prepare the email draft, and update the tracker, but it will never send anything without my authorization. Now it shows me the complete flow and asks me to approve it before creating the new skill. Review sponsorships. Now I have to approve this design so it can build and install the new skill. Once installed, I attach the brief, the script, the review link, and the tracking table for a brand called Novaflow. First, it compares the four files and detects three issues in the script. For each non-compliance, it shows the affected requirement and

18:27directly proposes the text we should replace it with. It also separates what does comply and what it still cannot verify, such as the integration duration, the visual demo, or if the disclaimer appears correctly in the final edit. And most importantly, it hasn't modified any file or prepared the email. It has stopped and is asking for my authorization to apply only these three corrections. Let’s approve the corrections. Now it can proceed. It applies only the three approved corrections to a copy of the script and keeps the original files intact. Then it performs a second review and provides me with four results: the corrected script, the updated tracking, the new verification

19:06report, and the email draft. Nothing has been sent. Obviously, this skill is adapted to my work, but think about any tedious process you repeat. It could be preparing meetings, reviewing deliverables, logging invoices, or closing projects. If you use the same files and follow the same steps, you can explain them once to this skill and turn them into a skill that you then execute with a single instruction. The last one is not a personal skill, but a complete package of skills included within a plugin. It's called Remotion and it allows us to create videos and motion graphics using code. Normally, we would have to prepare the project,

19:47program each animation, fix errors, and render the result. But within ChatGPT, we can describe what we want using plain language. To install it, I go back to plugins or add-ons, search for Remotion, and click install. When opening its profile, we see that it incorporates different specialized skills, a total of 12, to create animations, work with text, preview the project, and render the video. And we don't have to learn their names. When we ask it to create an animation with Remotion, ChatGPT automatically selects

20:17the skills it needs. Now let's test it. I'm going to take advantage of the fact that OpenAI just introduced GPT-4o Astra to ask for something much more ambitious: a cinematic launch piece that looks like it was created by a professional studio. ChatGPT creates the entire project, programs the different scenes, and opens a preview so I can check the result before rendering it. And here is the finished animation. We start with a small core of light that grows and forms a huge sphere of particles. Around it, some of Astra's capabilities appear, such as research, programming, computer use, and professional work. And then all the particles connect, form a large letter

20:57A, and end up exploding to reveal the name GPT6 AS from a single description, without me having to program the animation, manually place every element , or use a video editor. Furthermore, since it keeps the project, I can ask it to change the colors, speed up a scene, replace any text, or modify the duration. ChatGPT works on the existing composition and re-renders it. But the important thing about this example is not just Remotion. Some add-ons or plugins incorporate their own skills so that ChatGPT knows how to use them correctly. We don't even have to invoke many of them. They are activated

21:33automatically depending on the task. That is why, when you install an add-on , also check the skills it includes. You can add very advanced capabilities without having to personally create a skill for every part of the process. Well, these are the seven skills. As I promised you, below you have a practical guide with everything you need. You will find the instructions for use, the complete prompts to create the skills, and the files I used during the demonstrations. It is not just a summary of the video. The idea is that you can open ChatGPT, follow the document step by step, and end up with these same skills working in your

22:06account.