<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>AI Agents on Rusty Eddy</title><link>https://rustyeddy.com/tags/ai-agents/</link><description>Recent content in AI Agents on Rusty Eddy</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Thu, 08 Oct 2026 19:48:43 -0700</lastBuildDate><atom:link href="https://rustyeddy.com/tags/ai-agents/index.xml" rel="self" type="application/rss+xml"/><item><title>Writing Agent Instructions for a New Software Project</title><link>https://rustyeddy.com/software/writing-agent-instructions/</link><pubDate>Sat, 01 Aug 2026 00:00:00 +0000</pubDate><guid>https://rustyeddy.com/software/writing-agent-instructions/</guid><description>&lt;h2 id="working-thesis"&gt;Working Thesis&lt;/h2&gt;
&lt;p&gt;Agent instructions should be treated as an executable interface to a repository,
not as a long prompt. A useful instruction set tells an agent where truth lives,
which boundaries it must preserve, how to verify a change, and when to stop and
ask for a decision.&lt;/p&gt;
&lt;p&gt;The &lt;a href="https://github.com/rustyeddy/trader"&gt;Trader repository&lt;/a&gt; and its
&lt;a href="https://github.com/rustyeddy/trader/pull/9"&gt;agent-instructions pull request&lt;/a&gt;
provide the running example. The pull request is intentionally useful as a
work in progress: its structure shows what to centralize, while its review
comments show why every path and command in an instruction file must be tested.&lt;/p&gt;</description></item></channel></rss>