<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Permissions on Fernando Ruiz</title><link>https://www.fernandoux.com/tags/permissions/</link><description>Recent content in Permissions on Fernando Ruiz</description><generator>Hugo</generator><language>en-us</language><atom:link href="https://www.fernandoux.com/tags/permissions/index.xml" rel="self" type="application/rss+xml"/><item><title>Approving an agent's actions: how to ask permission without stopping the work</title><link>https://www.fernandoux.com/en/wiki/ai/action-approval/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://www.fernandoux.com/en/wiki/ai/action-approval/</guid><description>&lt;div class="info-panel"&gt;
 &lt;div class="info-header"&gt;
 &lt;span class="material-symbols-outlined info-panel-icon"&gt;info&lt;/span&gt;
 &lt;span class="info-panel-label"&gt;Quick Definition&lt;/span&gt;
 &lt;/div&gt;
 &lt;div class="info-content"&gt;
 &lt;strong&gt;Approving an agent&amp;rsquo;s actions&lt;/strong&gt; is the step where the system stops before running a sensitive action and asks a person to authorize or reject it. The action does not happen until a decision is recorded, and that decision can apply once or stay as a rule for similar cases.
 &lt;/div&gt;
&lt;/div&gt;

&lt;h2 id="what-exactly-gets-approved"&gt;What exactly gets approved?&lt;/h2&gt;
&lt;p&gt;An agent does many things, but not all of them deserve a question. What gets approved are actions that are hard to reverse or that leave the system: writing to a project file, running a terminal command, sending an email, creating a payment, deleting a record.&lt;/p&gt;</description></item></channel></rss>