<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Forem: Octomay </title>
    <description>The latest articles on Forem by Octomay  (@mayzaynara).</description>
    <link>https://forem.com/mayzaynara</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3569864%2F61c8c027-a779-4f7d-97dc-44f1d90d8a7a.webp</url>
      <title>Forem: Octomay </title>
      <link>https://forem.com/mayzaynara</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://forem.com/feed/mayzaynara"/>
    <language>en</language>
    <item>
      <title>Fundamentals of Modern AI: Agents, MCP, Skills, and Orchestration</title>
      <dc:creator>Octomay </dc:creator>
      <pubDate>Thu, 23 Apr 2026 16:36:00 +0000</pubDate>
      <link>https://forem.com/mayzaynara/fundamentals-of-modern-ai-agents-mcp-skills-and-orchestration-1elm</link>
      <guid>https://forem.com/mayzaynara/fundamentals-of-modern-ai-agents-mcp-skills-and-orchestration-1elm</guid>
      <description>&lt;p&gt;Every day a new AI appears — new models, new features, budget cuts here, fewer tokens there… and with that, a lot of people start wondering where all this is going. For juniors, they say there are no more job openings. For seniors, they say it won't be long before there aren't any either. Lots of "blah blah blah." These days, AI is the new JavaScript: every day a new library, "the best one," "the revolution."&lt;/p&gt;

&lt;p&gt;But let's get into it.&lt;/p&gt;

&lt;p&gt;The principle of AI is the same as frameworks. If you know programming logic, any other language has a smaller learning curve. This is the same territory. If you know the fundamentals, it doesn't matter whether you'll use Copilot, Codex, Claude, all of them together or separately — you can adapt.&lt;/p&gt;

&lt;p&gt;In the previous article, I talked about how I felt the first time I used Copilot. And today, clearly, our way of working has changed. We're going through a technological evolution.&lt;/p&gt;

&lt;p&gt;Every day we see an influencer claiming they haven't written a single line of code in months. And somehow, that creates a certain FOMO — especially for those just starting out.&lt;br&gt;
And many think: &lt;em&gt;"Does it even make sense for me to learn this, if AI does it for me?"&lt;/em&gt;&lt;br&gt;
So, little grasshopper, I'm going to explain this with analogies, and I hope that by the end of this article you'll understand some AI fundamentals, step out of the hype, and start truly understanding this new world.&lt;br&gt;
After reading, drop your insight in the comments. I really want your feedback. Enjoy.&lt;/p&gt;




&lt;p&gt;Imagine a head chef.&lt;br&gt;
She needs to serve 80 people. But there's a small detail: she doesn't know how to cook. She knows how to make instant noodles… and that's it. Bless her heart. haha.&lt;br&gt;
So she hires a junior chef. And the first order comes in: a chocolate cake.&lt;br&gt;
The junior asks:&lt;br&gt;
— Does the cake have flour? Yes.&lt;br&gt;
— Does it have eggs? Yes.&lt;br&gt;
— Does it have chocolate? Yes.&lt;br&gt;
— Does it have garlic? Yes.&lt;br&gt;
— Does it bake at 400 degrees? Yes.&lt;br&gt;
She fully trusts that junior. After all, he boils water like a pro… so baking a cake should be easy, right?&lt;/p&gt;

&lt;p&gt;And that's exactly the point, little grasshopper: if you don't know how to do it, how are you going to run a kitchen you're responsible for?&lt;br&gt;
That's why it's so important to know how to code, understand fundamentals, have seen many different coding contexts, and already fixed plenty of bugs before delegating anything to an AI.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;AI doesn't replace understanding — it amplifies execution.&lt;/strong&gt;&lt;br&gt;
When you use AI without a foundation, you become that head chef: someone who approves everything without any criteria. You don't know what's right, what's wrong, what's efficient or dangerous. Because AI doesn't think. It predicts patterns.&lt;br&gt;
Without a foundation, you become a hostage. With a foundation, you become a multiplier.&lt;br&gt;
And in a scenario that changes every day, another question arises: what if the machines go down?&lt;br&gt;
In the end, the question isn't whether AI will replace you. The question is: &lt;strong&gt;do you know how to work without it?&lt;/strong&gt;&lt;br&gt;
Because whoever understands what they're doing uses AI as a shortcut. Whoever doesn't… uses it as a crutch. And a crutch doesn't hold anyone up when the ground shifts.&lt;/p&gt;




&lt;p&gt;If the challenge used to be knowing how to code, now it's something else: &lt;strong&gt;do you know how to orchestrate intelligence?&lt;/strong&gt;&lt;br&gt;
Because today it's no longer just about writing code. It's about coordinating agents, defining context, controlling execution, and knowing exactly what to delegate.&lt;br&gt;
And that's where concepts like skills, agents, subagents, MCP, worktrees, hooks come in…&lt;br&gt;
If these names sound complex, relax — at their core, they all solve the same problem: giving structure to something that, on its own, just improvises.&lt;/p&gt;




&lt;p&gt;Back to the kitchen.&lt;br&gt;
Now imagine that, in addition to the junior chef, she has an entire automated kitchen: robots that chop, bake, plate dishes, and even suggest recipes.&lt;br&gt;
It sounds perfect… until you realize that someone still needs to: decide who does what, organize the order of tasks, and validate the final result.&lt;br&gt;
That's exactly what we're doing today with AI.&lt;/p&gt;

&lt;p&gt;Now this chef did things differently. She studied. She tested recipes at home. She failed dishes, burned others, adjusted seasonings. Over time, she stopped just following instructions and started understanding what she was doing. She built a repertoire.&lt;/p&gt;

&lt;p&gt;And in the middle of that process, she realized that the chaos in the kitchen wasn't a lack of recipes. It was a lack of connection.&lt;br&gt;
Each supplier, for example the flour supplier, brought something different.&lt;br&gt;
Some flour already came with baking powder mixed in, another had higher protein content. Others absorbed more water.&lt;/p&gt;

&lt;p&gt;Even following the same recipe, the result was never the same.&lt;/p&gt;

&lt;p&gt;She realized that each flour supplier had its own protocol. The egg supplier, another one. The ordering system, yet another. And each tool with its own standard.&lt;/p&gt;

&lt;p&gt;So she thought: &lt;em&gt;what if I created a standardized ordering system? A single protocol where every part of the kitchen operated on the same standard. Suppliers, stations, tools, inventory — all integrated through the same system.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;She called it &lt;strong&gt;MCP Delicious&lt;/strong&gt; — no more translating between systems. The system already bridged everything.&lt;/p&gt;

&lt;p&gt;And that's exactly what we technically call MCP.&lt;br&gt;
The MCP, or Model Context Protocol, is the structure that organizes how AI operates. It allows AI to stop just responding… and start operating within a system. It can access real tools, retrieve data, interact with APIs, and execute tasks based on a defined flow.&lt;br&gt;
In practice, as presented in the Copilot documentation, these connectors allow AI to integrate with external systems, real data, and real functionality — going far beyond text-based responses. Without it, AI responds with what it "thinks." With it, AI works with what it "accesses."&lt;/p&gt;




&lt;h2&gt;
  
  
  〔AGENTS〕
&lt;/h2&gt;

&lt;p&gt;You already understand the kitchen. Now you need to understand who works in it.&lt;br&gt;
The agent is the cook. Not the one who just executes orders — the one who makes decisions. It receives an objective, analyzes the context, chooses which tools to use and in what order, and executes. If something goes wrong along the way, it adjusts.&lt;br&gt;
In practice: you say "prepare dinner for 30 people." The agent doesn't wait for you to list every step. It decides what to do, uses what's available, and delivers the result.&lt;br&gt;
That's what sets an agent apart from a simple prompt: it doesn't respond. &lt;strong&gt;It operates.&lt;/strong&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  〔SKILLS〕
&lt;/h2&gt;

&lt;p&gt;To operate well, the agent needs abilities.&lt;br&gt;
In our scenario: knowing how to make a sauce, how to chop an ingredient properly, how to bake a cake to perfection. These are reusable capabilities.&lt;br&gt;
In AI, skills are specific abilities that combine instructions, context, and access to tools to solve tasks. But what does that mean in practice?&lt;br&gt;
Starting from a prompt like "given a repository, analyze the commit history and identify performance bottlenecks." The agent doesn't need you to explain everything from scratch — it loads the skill that encapsulates the &lt;em&gt;how&lt;/em&gt;: which tools to use, which context to load, which format to deliver. You define it once. The agent reuses it.&lt;br&gt;
In GitHub Copilot, for example, you can configure custom instructions that work like skills: specific guidance on how the agent should behave in certain contexts, which tools it can trigger, and what response format to deliver. Because having every recipe in the world is useless… if nobody knows when to use each one.&lt;/p&gt;




&lt;h2&gt;
  
  
  〔SUBAGENTS〕
&lt;/h2&gt;

&lt;p&gt;And when the kitchen grows, another need emerges. You can't have everyone working in the same space. It becomes chaos.&lt;br&gt;
And here's a detail that seems obvious but isn't: a subagent isn't a different kind of entity. It's an agent. The difference lies in the role it plays and the specialty it develops.&lt;br&gt;
That's where subagents come in — the specialists.&lt;br&gt;
One handles desserts, another handles sauces, another handles finishing. Each one takes on a specific part of the process. They don't do everything. But they do one thing very well.&lt;br&gt;
In practice: imagine you have a main agent responsible for deploying an application. It can trigger one subagent to run the tests, another to check for security vulnerabilities, and another to document the changes. Each subagent has its own scope and the main agent simply orchestrates the results.&lt;br&gt;
In GitHub Copilot in agent mode, this is reflected in the ability to chain tasks: the main agent breaks a complex objective into subtasks and delegates them, waiting for results before proceeding. This increases output quality because each part is handled with focus, not with generality.&lt;/p&gt;




&lt;h2&gt;
  
  
  〔WORKTREE〕
&lt;/h2&gt;

&lt;p&gt;And as the operation grows even further, you need organization.&lt;br&gt;
The kitchen is no longer just one space. It starts having separate stations. Each one with its own context, its own prep, its own timing.&lt;br&gt;
A dish can be created from scratch at one station while another is being finished at a different one — without interference.&lt;br&gt;
&lt;strong&gt;That's worktree.&lt;/strong&gt;&lt;br&gt;
The concept comes from Git. &lt;code&gt;git worktree&lt;/code&gt; is a feature that allows you to have multiple checkouts of the same repository in different directories at the same time. In practice: you can be fixing a critical bug on the &lt;code&gt;hotfix&lt;/code&gt; branch while developing a new feature on &lt;code&gt;feat/new-screen&lt;/code&gt;, without needing to stash anything, without switching branches, without the risk of mixing up context.&lt;/p&gt;

&lt;p&gt;But real quick:&lt;br&gt;
A &lt;strong&gt;branch&lt;/strong&gt; is like a parallel line of work. You "copy" the project at a given moment and can make changes without touching the main version. It's like creating a separate version just to test or develop something new.&lt;br&gt;
A &lt;strong&gt;stash&lt;/strong&gt; is like temporarily saving changes. Imagine you started something, but you need to switch tasks: you "store" those changes without finishing them and can retrieve them later exactly where you left off.&lt;br&gt;
In the world of AI agents, the concept is the same: context isolation. Each agent (or subagent) operates in its own workspace, without interfering with what the other is doing. It's working in parallel without breaking what already works.&lt;/p&gt;




&lt;p&gt;In the end, none of this is about isolated technology. It's about structure.&lt;br&gt;
Because whoever just uses AI requests. Whoever understands, organizes. And when the scenario changes — because it always does — whoever depends on it freezes. Whoever understands, figures it out.&lt;br&gt;
And being completely honest: I'm also still learning. This world is new to me, to you, to everyone.&lt;br&gt;
The difference isn't who already knows everything. It's who is trying to understand while everything is still being built.&lt;/p&gt;




&lt;p&gt;But the truth is, this is still just the beginning. We talked about structure, agents, context, and organization. But there's still one important piece of this kitchen missing: what makes all of this work on its own.&lt;br&gt;
Because up until now, there's still someone deciding every step. And when you scale that… you don't just want to organize. You want to automate. You want the system to react, make simple decisions, and keep running without you having to intervene all the time.&lt;br&gt;
And that's exactly where hooks come in. But that… is a topic for the next article. 👀&lt;/p&gt;




&lt;h2&gt;
  
  
  〔CONCLUSION〕
&lt;/h2&gt;

&lt;p&gt;Before you go: what you just read isn't science fiction theory. It's the real skeleton of how AI-powered systems organize themselves today.&lt;br&gt;
MCP as a connection structure. Skills as reusable abilities. Agents as orchestrators. Subagents as specialists. Worktree as context isolation.&lt;br&gt;
Each of these concepts solves a specific problem, and together, they form the foundation of any AI system that goes beyond "chatting with a bot."&lt;br&gt;
If you understood this, you're already ahead of a lot of people who are still debating whether AI will or won't replace developers. While they argue, you're learning.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;See you at the next commit.&lt;/em&gt;&lt;/p&gt;




&lt;p&gt;&lt;strong&gt;References&lt;/strong&gt;&lt;br&gt;
If you want to keep diving deeper, here are some of the materials I used as a foundation:&lt;br&gt;
Mentorship with Pedro Lacerda&lt;br&gt;
GitHub Copilot documentation (agent customization): &lt;a href="https://docs.github.com/en/enterprise-cloud@latest/copilot/reference/customization-cheat-sheet" rel="noopener noreferrer"&gt;https://docs.github.com/en/enterprise-cloud@latest/copilot/reference/customization-cheat-sheet&lt;/a&gt;&lt;br&gt;
Model Context Protocol (MCP) – Anthropic: &lt;a href="https://www.anthropic.com/news/model-context-protocol" rel="noopener noreferrer"&gt;https://www.anthropic.com/news/model-context-protocol&lt;/a&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>mcp</category>
      <category>agents</category>
      <category>githubcopilot</category>
    </item>
    <item>
      <title>Fundamentos de IA moderna: agentes, MCP, skills e orquestração</title>
      <dc:creator>Octomay </dc:creator>
      <pubDate>Tue, 21 Apr 2026 21:21:13 +0000</pubDate>
      <link>https://forem.com/mayzaynara/fundamentos-de-ia-moderna-agentes-mcp-skills-e-orquestracao-2136</link>
      <guid>https://forem.com/mayzaynara/fundamentos-de-ia-moderna-agentes-mcp-skills-e-orquestracao-2136</guid>
      <description>&lt;p&gt;Todos os dias surge uma IA nova, modelos novos, funcionalidades novas, um recorte de gastos daqui, menos tokens dali… e com isso, muita gente começa a se perguntar onde isso tudo vai parar. Para júnior, dizem que não há mais vagas. Para sênior, dizem que em breve também não haverá. Muito "blá blá blá". Hoje em dia, a IA é o novo JavaScript: todo dia uma biblioteca nova, "a melhor", "a revolução".&lt;/p&gt;

&lt;p&gt;Mas vamos lá.&lt;/p&gt;

&lt;p&gt;O princípio da IA é o mesmo de frameworks. Se você sabe lógica de programação, qualquer outra linguagem tem uma curva de aprendizagem menor. Aqui é o mesmo terreno. Se você conhece a base, independente de se vai usar Copilot, Codex, Claude, juntas ou separadas, você consegue se adaptar.&lt;br&gt;
No artigo anterior, eu disse como me senti na primeira vez que usei o Copilot. E hoje, claramente, o nosso modo de trabalho mudou. Estamos passando por uma evolução na tecnologia.&lt;/p&gt;

&lt;p&gt;Todos os dias vemos um influenciador que diz que faz meses que não escreve uma linha de código. E, de alguma maneira, isso gera um certo FOMO,  principalmente para quem está começando.&lt;br&gt;
E muitos pensam: "Faz sentido eu estar aprendendo isso, já que a IA faz por mim?"&lt;br&gt;
Então, gafanhoto, vou te explicar com analogias,  e espero que você saia deste artigo entendendo alguns fundamentos da IA, saindo do hype e passando a entender de verdade esse mundo novo.&lt;br&gt;
Depois que ler, deixa aqui nos comentários o insight que você teve. Quero muito seu feedback. Boa leitura.&lt;/p&gt;




&lt;p&gt;Imagine uma chef de cozinha.&lt;br&gt;
Ela precisa servir 80 pessoas. Mas tem um pequeno detalhe: ela não sabe cozinhar. Ela sabe fazer um miojo… e é isso. e Fé. jajaja.&lt;br&gt;
Agora ela contrata um chef júnior. E vem o primeiro pedido: um bolo de chocolate.&lt;br&gt;
O júnior pergunta:&lt;br&gt;
 — O bolo vai farinha? Yes. &lt;br&gt;
— O bolo vai ovo? Yes. &lt;br&gt;
— O bolo vai chocolate? Yes. &lt;br&gt;
— O bolo vai alho? Yes&lt;br&gt;
. — O bolo vai a 400 graus? Yes.&lt;br&gt;
Ela confia plenamente naquele júnior. Afinal, ele esquenta água como ninguém… então fazer um bolo deve ser tranquilo, né?&lt;/p&gt;

&lt;p&gt;E aí que está, pequeno gafanhoto: se você não sabe fazer, como você vai orquestrar a cozinha da qual você é responsável?&lt;br&gt;
Por isso é tão importante saber programar, entender fundamentos, ter visto vários contextos de código e já ter consertado muitos bugs antes de delegar qualquer coisa para uma IA.&lt;/p&gt;

&lt;p&gt;IA não substitui entendimento, ela amplifica execução.&lt;br&gt;
Quando você usa IA sem base, você vira essa chef : alguém que aprova tudo sem critério. Você não sabe o que está certo, o que está errado, o que é eficiente ou perigoso. Porque a IA não pensa. Ela prevê padrões.&lt;br&gt;
Sem base, você vira refém. Com base, você vira multiplicador.&lt;br&gt;
E, em um cenário que muda todos os dias, surge outra pergunta: e se as máquinas desligarem?&lt;br&gt;
No fim, a questão não é se a IA vai substituir você. A questão é: você sabe trabalhar sem ela?&lt;br&gt;
Porque quem entende o que está fazendo usa a IA como atalho. Quem não entende… usa como muleta. E muleta não sustenta ninguém quando o chão muda.&lt;/p&gt;




&lt;p&gt;Se antes o problema era saber programar, agora o problema é outro: você sabe orquestrar inteligência?&lt;br&gt;
Porque hoje não se trata mais só de escrever código. Se trata de coordenar agentes, definir contexto, controlar execução e saber exatamente o que delegar.&lt;br&gt;
E é aqui que entram conceitos como skills, agentes, subagentes, MCP, worktrees, hooks…&lt;br&gt;
Se esses nomes parecem complexos, relaxa, no fundo, todos eles resolvem o mesmo problema: dar estrutura para algo que, sozinho, só improvisa.&lt;/p&gt;




&lt;p&gt;Voltando ao chef de cozinha.&lt;br&gt;
Agora imagina que, além do júnior, ele tem uma cozinha inteira automatizada: robôs que cortam, assam, montam pratos e até sugerem receitas.&lt;br&gt;
Parece perfeito… até você perceber que alguém ainda precisa: decidir quem faz o quê, organizar a ordem das tarefas, validar o resultado final.&lt;br&gt;
É exatamente isso que estamos fazendo hoje com IA.&lt;/p&gt;

&lt;p&gt;Agora essa chef fez diferente. Ela estudou. Testou receitas em casa. Errou pratos, queimou outros, ajustou temperos. Com o tempo, deixou de só seguir instruções e passou a entender o que estava fazendo. Ela construiu repertório.&lt;/p&gt;

&lt;p&gt;E, no meio desse processo, ela percebeu que o caos da cozinha não era a falta de receita. Era falta de conexão. &lt;br&gt;
E cada fornecedor, por exemplo, de farinha, trazia algo diferente.&lt;br&gt;
Uma farinha já vinha com fermento, outra com mais proteína. Outras absorvem mais água. &lt;/p&gt;

&lt;p&gt;Mesmo seguindo a mesma receita, o resultado nunca era igual. &lt;/p&gt;

&lt;p&gt;Ela percebeu que cada fornecedor de farinha tinha o seu próprio protocolo. O de ovos, outra. O sistema de pedidos, outra. E cada ferramenta com o seu próprio padrão. &lt;/p&gt;

&lt;p&gt;Então pensou: e se eu criasse um sistema de pedidos padronizados? Um único protocolo  onde todas as partes da cozinha operassem no mesmo padrão. Fornecedores, estações, ferramentas, estoque, tudo integrado pelo mesmo sistema. &lt;/p&gt;

&lt;p&gt;Ela chamou isso de MCP Delicious , não precisava ficar mais traduzindo, O sistema já fazia ponte. &lt;/p&gt;

&lt;p&gt;E é exatamente isso que, tecnicamente, chamamos de MCP.&lt;br&gt;
O MCP, ou Model Context Protocol, é a estrutura que organiza como a IA trabalha. Ele permite que a IA deixe de apenas responder… e passe a operar dentro de um sistema. Ela pode acessar ferramentas reais, buscar dados, interagir com APIs e executar tarefas com base em um fluxo definido.&lt;br&gt;
Na prática, como apresentado na documentação do Copilot, esses conectores permitem integrar a IA com sistemas externos, dados e funcionalidades reais — indo muito além de respostas baseadas apenas em texto. Sem isso, ela responde com o que "acha". Com isso, ela trabalha com o que "acessa".&lt;/p&gt;




&lt;p&gt;〔AGENTES〕&lt;br&gt;
Você já entendeu a cozinha. Agora precisa entender quem trabalha nela.&lt;br&gt;
O agente é o cozinheiro. Não o que só executa ordens, o que toma decisões. Ele recebe um objetivo, analisa o contexto, escolhe quais ferramentas usar, em que ordem, e executa. Se algo der errado no meio do caminho, ele ajusta.&lt;br&gt;
Na prática: você diz "prepara o jantar para 30 pessoas". O agente não espera você listar cada passo. Ele decide o que fazer, usa o que tem disponível e entrega o resultado.&lt;br&gt;
É isso que diferencia um agente de um simples prompt: ele não responde. Ele opera.&lt;/p&gt;




&lt;p&gt;〔SKILLS〕&lt;br&gt;
Para operar bem, o agente precisa de habilidades.&lt;br&gt;
No nosso cenário:  Saber fazer um molho, cortar um ingrediente corretamente, assar um bolo no ponto certo. São capacidades reutilizáveis.&lt;br&gt;
Na IA, skills são habilidades específicas que combinam instruções, contexto e acesso a ferramentas para resolver tarefas. Mas o que isso significa na prática?&lt;br&gt;
A partir de um prompt como "dado um repositório, analise o histórico de commits e identifique gargalos de performance". O agente não precisa que você explique tudo do zero, ele carrega a skill que encapsula o como fazer: quais ferramentas usar, qual contexto carregar, qual formato entregar. Você define uma vez. O agente reutiliza.&lt;br&gt;
No GitHub Copilot, por exemplo, é possível configurar instruções personalizadas que funcionam como skills: direcionamentos específicos sobre como o agente deve se comportar em determinados contextos, quais ferramentas pode acionar e qual formato de resposta entregar. Porque não adianta ter todas as receitas do mundo… se ninguém souber quando usar cada uma.&lt;/p&gt;




&lt;p&gt;〔SUBAGENTES〕&lt;br&gt;
E quando a cozinha cresce, surge outra necessidade. Você não pode ter todo mundo trabalhando no mesmo espaço. Vira caos.&lt;br&gt;
E aqui tem um detalhe que parece óbvio mas não é: subagente não é um tipo diferente entidade. É um agente. A diferença está no papel que ele ocupa e na especialidade que ele desenvolve. &lt;br&gt;
É aí que entram os subagentes, os especialistas.&lt;br&gt;
Um cuida da sobremesa, outro dos molhos, outro da finalização. Cada um assume uma parte específica do processo. Eles não fazem tudo. Mas fazem uma coisa muito bem feita.&lt;br&gt;
Na prática: imagine que você tem um agente principal responsável por fazer o deploy de uma aplicação. Ele pode acionar um subagente para rodar os testes, outro para verificar vulnerabilidades de segurança e outro para documentar as mudanças. Cada subagente tem seu escopo e o agente principal apenas orquestra os resultados.&lt;br&gt;
No GitHub Copilot em modo agente, isso se reflete na capacidade de encadear tarefas: o agente principal quebra um objetivo complexo em subtarefas e as delega, aguardando os resultados antes de prosseguir. Isso aumenta a qualidade do resultado porque cada parte é tratada com foco, não com generalidade.&lt;/p&gt;




&lt;p&gt;〔WORKTREE 〕&lt;br&gt;
E conforme a operação cresce ainda mais, você precisa de organização.&lt;br&gt;
A cozinha deixa de ser uma só. Ela passa a ter estações separadas. Cada uma com seu contexto, seu preparo, seu momento.&lt;br&gt;
Um prato pode estar sendo criado do zero em uma estação, enquanto outro já está sendo finalizado em outra, sem interferência.&lt;br&gt;
Isso é worktree.&lt;br&gt;
O conceito vem do Git, git worktree é um recurso que permite ter múltiplos checkouts do mesmo repositório em diretórios diferentes, ao mesmo tempo. Na prática: você pode estar corrigindo um bug crítico na branch hotfix enquanto desenvolve uma nova feature em feat/nova-tela, sem precisar fazer stash, sem trocar de branch, sem risco de misturar contexto.&lt;/p&gt;

&lt;p&gt;Mas, rapidinho:&lt;br&gt;
Uma branch é como uma linha paralela de trabalho. Você “copia” o projeto naquele momento e pode fazer alterações sem mexer na versão principal. É como criar uma versão separada só para testar ou desenvolver algo novo.&lt;br&gt;
Já o stash é como guardar alterações temporariamente. Imagine que você começou algo, mas precisa mudar de tarefa: você “guarda” essas mudanças sem finalizar e pode recuperá-las depois exatamente de onde parou.&lt;br&gt;
No universo de agentes de IA, o conceito é o mesmo: isolamento de contexto. Cada agente (ou subagente) opera em seu próprio espaço de trabalho, sem interferir no que o outro está fazendo. É trabalhar em paralelo sem quebrar o que já funciona.&lt;/p&gt;




&lt;p&gt;No fim, nada disso é sobre tecnologia isolada. É sobre estrutura.&lt;br&gt;
Porque quem só usa IA pede. Quem entende, organiza. E, quando o cenário muda, porque muda o tempo todo, quem depende trava. Quem entende, se vira.&lt;br&gt;
E sendo bem honesta: eu também estou aprendendo. Esse mundo é novo pra mim, pra você, pra todo mundo.&lt;br&gt;
A diferença não está em quem já sabe tudo. Está em quem está tentando entender enquanto tudo ainda está sendo construído.&lt;/p&gt;




&lt;p&gt;Mas a verdade é que isso ainda é só o começo. A gente falou sobre estrutura, agentes, contexto e organização. Mas ainda falta uma peça importante dessa cozinha: o que faz tudo isso funcionar sozinho.&lt;br&gt;
Porque até agora, ainda existe alguém decidindo cada passo. E quando você escala isso… você não quer só organizar. Você quer automatizar. Quer que o sistema reaja, tome decisões simples e continue funcionando sem você precisar intervir o tempo todo.&lt;br&gt;
E é exatamente aí que entram os hooks. Mas isso… é assunto pro próximo artigo. 👀&lt;/p&gt;




&lt;p&gt;〔CONCLUSÃO 〕&lt;br&gt;
Antes de ir: o que você acabou de ler não é teoria de ficção científica. É o esqueleto real de como sistemas com IA se organizam hoje.&lt;br&gt;
MCP como estrutura de conexão. Skills como habilidades reutilizáveis. Agentes como orquestradores. Subagentes como especialistas. Worktree como isolamento de contexto.&lt;br&gt;
Cada um desses conceitos resolve um problema específico, e juntos, eles formam a base de qualquer sistema de IA que vai além do "chat com o bot".&lt;br&gt;
Se você entendeu isso, você já saiu na frente de muita gente que ainda está discutindo se a IA vai ou não substituir devs. Enquanto eles debatem, você está aprendendo.&lt;/p&gt;

&lt;p&gt;Nos vemos no próximo commit &lt;/p&gt;




&lt;p&gt;Fique por dentro da OctoMay — minha série de artigos sobre Git, GitHub e a evolução da profissão no ecossistema do GitHub.&lt;/p&gt;

&lt;p&gt;Aqui estão alguns textos já publicados:&lt;/p&gt;

&lt;p&gt;🔹  A sensação curiosa de usar o GitHub Copilot&lt;/p&gt;

&lt;p&gt;&lt;a href="https://www.linkedin.com/pulse/sensa%C3%A7%C3%A3o-curiosa-de-usar-o-github-copilot-mayza-ynara-rodrigues-pi8lf/" rel="noopener noreferrer"&gt;https://www.linkedin.com/pulse/sensa%C3%A7%C3%A3o-curiosa-de-usar-o-github-copilot-mayza-ynara-rodrigues-pi8lf/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;🔹 Quando o CEO do GitHub sai, quem salva seu código? Aprenda Git!&lt;/p&gt;

&lt;p&gt;&lt;a href="https://www.linkedin.com/pulse/quando-o-ceo-do-github-sai-quem-salva-seu-c%C3%B3digo-git-rodrigues-ynhyf/" rel="noopener noreferrer"&gt;https://www.linkedin.com/pulse/quando-o-ceo-do-github-sai-quem-salva-seu-c%C3%B3digo-git-rodrigues-ynhyf/&lt;/a&gt;&lt;/p&gt;




&lt;p&gt;Referências &lt;br&gt;
Se você quiser continuar se aprofundando, esses foram alguns dos materiais que usei como base:&lt;/p&gt;

&lt;p&gt;Mentoria com Pedro Lacerda&lt;/p&gt;

&lt;p&gt;Documentação do GitHub Copilot (customização de agentes): &lt;a href="https://docs.github.com/en/enterprise-cloud@latest/copilot/reference/customization-cheat-sheet" rel="noopener noreferrer"&gt;https://docs.github.com/en/enterprise-cloud@latest/copilot/reference/customization-cheat-sheet&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Model Context Protocol (MCP) – Anthropic: &lt;a href="https://www.anthropic.com/news/model-context-protocol" rel="noopener noreferrer"&gt;https://www.anthropic.com/news/model-context-protocol&lt;/a&gt;&lt;/p&gt;

</description>
      <category>agents</category>
      <category>ai</category>
      <category>career</category>
      <category>mcp</category>
    </item>
  </channel>
</rss>
