<?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: IdoGol24</title>
    <description>The latest articles on DEV Community by IdoGol24 (@idogol24).</description>
    <link>https://dev.to/idogol24</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.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F569686%2F7c2e5b97-c136-46b1-866d-aeb04c8ad266.jpeg</url>
      <title>DEV Community: IdoGol24</title>
      <link>https://dev.to/idogol24</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/idogol24"/>
    <language>en</language>
    <item>
      <title>Weir - deterministic unit tests for AI agents (no LLM)</title>
      <dc:creator>IdoGol24</dc:creator>
      <pubDate>Tue, 25 Aug 2026 20:15:04 +0000</pubDate>
      <link>https://dev.to/idogol24/your-evals-pass-and-your-agent-is-broken-stop-asking-an-llm-whether-your-llm-misbehaved-26e9</link>
      <guid>https://dev.to/idogol24/your-evals-pass-and-your-agent-is-broken-stop-asking-an-llm-whether-your-llm-misbehaved-26e9</guid>
      <description>&lt;p&gt;Your agent's logs can say "success" while the run was quietly hijacked - untrusted content at step 2 steering a tool call at step 6 - and the industry's default answer is pointing a second LLM at the first one.&lt;/p&gt;

&lt;p&gt;I built &lt;code&gt;Weir&lt;/code&gt; to be the deterministic opposite: it reads the OpenTelemetry traces you already have and answers two questions:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt; How much of your agent's behavior your telemetry can actually prove&lt;/li&gt;
&lt;li&gt; Whether a forbidden flow happened (shown as a witness path you can fail CI on)&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Same input, byte-identical answer, no LLM in the loop, nothing leaves your machine.&lt;/p&gt;

&lt;p&gt;Try it out:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pip &lt;span class="nb"&gt;install &lt;/span&gt;weir-scan &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; weir gauge &lt;span class="nt"&gt;--sample&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;License: Apache-2.0&lt;/p&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/IdoGol24" rel="noopener noreferrer"&gt;
        IdoGol24
      &lt;/a&gt; / &lt;a href="https://github.com/IdoGol24/weir" rel="noopener noreferrer"&gt;
        weir
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      Weir - unit tests for your agents
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;Weir - unit tests for your agents&lt;/h1&gt;
&lt;/div&gt;

&lt;p&gt;Weir is a CI gate for AI agents.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/IdoGol24/weir/actions/workflows/ci.yml" rel="noopener noreferrer"&gt;&lt;img src="https://github.com/IdoGol24/weir/actions/workflows/ci.yml/badge.svg" alt="CI"&gt;&lt;/a&gt; &lt;a href="https://pypi.org/project/weir-scan/" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/1e88f6ad092422e29485f179e22d5ef07ba4d5e8bcba755e7ec2cc1803c35486/68747470733a2f2f696d672e736869656c64732e696f2f707970692f762f776569722d7363616e" alt="PyPI"&gt;&lt;/a&gt; &lt;a href="https://github.com/IdoGol24/weir/LICENSE" rel="noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/798509b4df525f56802b56f8096862487f08023e3d7561c68656f8dab10d0d6e/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f6c6963656e73652d4170616368652d2d322e302d626c75652e737667" alt="License: Apache-2.0"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;
  &lt;a rel="noopener noreferrer nofollow" href="https://raw.githubusercontent.com/IdoGol24/weir/main/docs/assets/demo.svg"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2FIdoGol24%2Fweir%2Fmain%2Fdocs%2Fassets%2Fdemo.svg" alt="Terminal recording: weir scan reports a verdict-grade finding with a witness path, then exits 1" width="620"&gt;&lt;/a&gt;
&lt;/p&gt;

&lt;p&gt;It reads the OpenTelemetry traces your agent already emits and fails the build when sensitive data reaches a sink it should not reach.&lt;/p&gt;

&lt;p&gt;Weir asks a structural question:&lt;/p&gt;

&lt;div class="markdown-heading"&gt;
&lt;h3 class="heading-element"&gt;Did sensitive data flow through the agent to a sink it should never have reached?&lt;/h3&gt;
&lt;/div&gt;

&lt;p&gt;Your agent already answers that question in the traces it emits. Weir reconstructs the session graph, tracks taint through it, and shows the evidence node by node.&lt;/p&gt;


  &lt;div class="js-render-enrichment-target"&gt;
    &lt;div class="render-plaintext-hidden"&gt;
      &lt;pre&gt;flowchart LR
  A["traces your agent&amp;lt;br/&amp;gt;already emits"] --&amp;gt; B{"weir gauge"}
  B --&amp;gt;|"coverage too low"| C["names the exact&amp;lt;br/&amp;gt;instrumentation switch"]
  C -.-&amp;gt;|"flip it, re-run"| B
  B --&amp;gt;|"coverage sufficient"| D{"weir scan"}
  D --&amp;gt;|"no forbidden flow"| E["exit 0"]
  D --&amp;gt;|"forbidden flow"| F["exit 1 + witness path&amp;lt;br/&amp;gt;n2 → n3 → n4 → n5 → n6"]
&lt;/pre&gt;
    &lt;/div&gt;
  &lt;/div&gt;
  &lt;span class="js-render-enrichment-loader d-flex flex-justify-center flex-items-center width-full"&gt;
    &lt;span&gt;
      &lt;span class="sr-only"&gt;Loading&lt;/span&gt;
&lt;/span&gt;
  &lt;/span&gt;


&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Try it in two minutes&lt;/h2&gt;

&lt;/div&gt;

&lt;div class="snippet-clipboard-content notranslate position-relative overflow-auto"&gt;
&lt;pre class="notranslate"&gt;&lt;code&gt;pip install weir-scan
weir gauge your-export.jsonl   # or: weir gauge&lt;/code&gt;&lt;/pre&gt;…&lt;/div&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/IdoGol24/weir" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;



&lt;p&gt;The roadmap is completely in the open. I'd genuinely like to know where it breaks on your exports!&lt;/p&gt;

</description>
      <category>ai</category>
      <category>testing</category>
      <category>opensource</category>
      <category>agents</category>
    </item>
  </channel>
</rss>
