<?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>DEV Community: Bruno Arias</title>
    <description>The latest articles on DEV Community by Bruno Arias (@ariasbruno).</description>
    <link>https://dev.to/ariasbruno</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%2F3862523%2F731891b9-952f-4050-861c-191b1a496b82.png</url>
      <title>DEV Community: Bruno Arias</title>
      <link>https://dev.to/ariasbruno</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/ariasbruno"/>
    <language>en</language>
    <item>
      <title>How I stopped AI Context Bloat and built a package manager for my local agents</title>
      <dc:creator>Bruno Arias</dc:creator>
      <pubDate>Sun, 05 Apr 2026 16:56:20 +0000</pubDate>
      <link>https://dev.to/ariasbruno/how-i-stopped-ai-context-bloat-and-built-a-package-manager-for-my-local-agents-2mii</link>
      <guid>https://dev.to/ariasbruno/how-i-stopped-ai-context-bloat-and-built-a-package-manager-for-my-local-agents-2mii</guid>
      <description>&lt;p&gt;I rely heavily on local LLMs and AI coding agents (like Cursor, Copilot, and custom agents) for my daily workflow. But recently, I kept hitting a frustrating wall with how these agents handle their skills.&lt;/p&gt;

&lt;p&gt;I realized I was facing two major problems:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. Context Bloat &amp;amp; Hallucinations&lt;/strong&gt;&lt;br&gt;
If I keep all my agent skills in a global folder, the AI reads &lt;em&gt;everything&lt;/em&gt;—even skills it doesn't need for the current project. If I'm building a simple React frontend, my agent doesn't need access to my database-seeding skills or DevOps skills. Giving it too much context leads to hallucinations and burns through tokens unnecessarily.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. The Security Auditing Nightmare&lt;/strong&gt;&lt;br&gt;
The alternative is to install skills per-project directly from remote sources. But doing this means I waste a ton of time re-auditing each skill for malicious code or weird vulnerabilities every single time I start a new repo.&lt;/p&gt;
&lt;h3&gt;
  
  
  Enter &lt;code&gt;skillbase&lt;/code&gt; 🧠
&lt;/h3&gt;

&lt;p&gt;&lt;code&gt;skillbase&lt;/code&gt; is a Node.js CLI that acts as a local package manager specifically designed for AI agent skills. &lt;/p&gt;
&lt;h3&gt;
  
  
  How it solves the mess
&lt;/h3&gt;

&lt;p&gt;The core idea is simple: &lt;strong&gt;Auditing once, using safely everywhere.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;&lt;code&gt;skillbase&lt;/code&gt; lets you maintain a single, vetted global registry on your machine (defaulting to &lt;code&gt;~/.skillbase/skills/&lt;/code&gt;). Crucially, your AI &lt;em&gt;does not&lt;/em&gt; read this folder by default.&lt;/p&gt;

&lt;p&gt;When you start a new project, you just run:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;skillbase add &amp;lt;skill&amp;gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The CLI safely &lt;strong&gt;copies&lt;/strong&gt; your pre-vetted skills directly into your current workspace's &lt;code&gt;.agents/skills/&lt;/code&gt; folder. If you prefer to keep a single source of truth without duplicating files, you can just pass the &lt;code&gt;-s&lt;/code&gt; or &lt;code&gt;--sym&lt;/code&gt; flag to create a &lt;strong&gt;symbolic link&lt;/strong&gt; instead.&lt;/p&gt;

&lt;h3&gt;
  
  
  A few cool things under the hood:
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;📦 Reproducibility:&lt;/strong&gt; It generates a &lt;code&gt;skillbase.json&lt;/code&gt; manifest (just like a &lt;code&gt;package.json&lt;/code&gt; for AI context). You can commit this file, and your teammates can run &lt;code&gt;skillbase install&lt;/code&gt; to recreate the exact same skill set for that repository.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;🌐 Remote Sources:&lt;/strong&gt; You can fetch new skills directly from public GitHub repos (&lt;code&gt;skillbase install &amp;lt;repo-url&amp;gt; --remote&lt;/code&gt;) to audit and add them to your global registry.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;✨ Auto-detection:&lt;/strong&gt; An &lt;code&gt;init&lt;/code&gt; command scans your project's &lt;code&gt;package.json&lt;/code&gt; dependencies and suggests which skills you might want to inject.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;🔄 Easy Migration &amp;amp; Promotion:&lt;/strong&gt; If you build a custom skill inside a specific project, you can easily promote it to your global registry using the &lt;code&gt;migrate&lt;/code&gt; command. It also lets you import legacy global setups in one go.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Check it out!
&lt;/h3&gt;

&lt;p&gt;I originally built this to scratch my own itch, but it's completely open-source and I'd love to see if it helps other developers dealing with the same AI context chaos.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub Repo:&lt;/strong&gt; &lt;a href="https://github.com/ariasbruno/skillbase" rel="noopener noreferrer"&gt;ariasbruno/skillbase&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;npm:&lt;/strong&gt; &lt;a href="https://www.npmjs.com/package/@ariasbruno/skillbase" rel="noopener noreferrer"&gt;@ariasbruno/skillbase&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;How do you guys manage AI context isolation and skill security in your own setups? I'm all ears for feedback, ideas, or PRs! 👇&lt;/p&gt;

</description>
      <category>ai</category>
      <category>webdev</category>
      <category>productivity</category>
      <category>opensource</category>
    </item>
  </channel>
</rss>
