<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Workflow on brtkwr.com</title><link>https://brtkwr.com/tags/workflow/</link><description>Recent content in Workflow on brtkwr.com</description><generator>Hugo</generator><language>en</language><lastBuildDate>Thu, 30 Jul 2026 23:45:00 +0000</lastBuildDate><atom:link href="https://brtkwr.com/tags/workflow/index.xml" rel="self" type="application/rss+xml"/><item><title>Git-managed agent routines that propose their own fixes</title><link>https://brtkwr.com/posts/2026-07-30-git-managed-self-improving-agent-routines/</link><pubDate>Thu, 30 Jul 2026 23:45:00 +0000</pubDate><guid>https://brtkwr.com/posts/2026-07-30-git-managed-self-improving-agent-routines/</guid><description>&lt;p>&lt;strong>TL;DR&lt;/strong> This is about Claude&amp;rsquo;s scheduled cloud routines specifically, which is the only agent platform I have tried it on. I moved my routine prompts out of the scheduler and into a git repo, leaving a fifteen-line stub behind that reads its instructions from &lt;code>main&lt;/code> at runtime. Changing a routine is now a commit rather than a form edit, and each routine has standing authority to open a PR against its own prompt when it trips over a problem. I set this up today, so I have no idea yet whether the self-improvement PRs will be worth merging.&lt;/p></description></item></channel></rss>