<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Guardrails on Raja Sleem</title><link>https://www.rajasleem.com/tags/guardrails/</link><description>Recent content in Guardrails on Raja Sleem</description><image><title>Raja Sleem</title><url>https://www.rajasleem.com/images/rajasleem-og-thumbnail-1200x630.png</url><link>https://www.rajasleem.com/images/rajasleem-og-thumbnail-1200x630.png</link></image><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 12 Aug 2026 09:00:00 -0400</lastBuildDate><atom:link href="https://www.rajasleem.com/tags/guardrails/index.xml" rel="self" type="application/rss+xml"/><item><title>The Paperwork Outweighed the Product: What Testing My Own Pipeline Cost</title><link>https://www.rajasleem.com/posts/agentic-workflow-part-7-the-paperwork-problem/</link><pubDate>Wed, 12 Aug 2026 09:00:00 -0400</pubDate><guid>https://www.rajasleem.com/posts/agentic-workflow-part-7-the-paperwork-problem/</guid><description>I finally pointed my autonomous pipeline at two real repos. One paid 4.7x more per line of code than the other, and the culprit was not the model. It was my own specs.</description></item><item><title>Obedience Is Futile: The Guardrail You Build Against Your Own Limited Context Window</title><link>https://www.rajasleem.com/posts/agentic-workflow-part-6-obedience-is-futile/</link><pubDate>Tue, 04 Aug 2026 09:00:00 -0400</pubDate><guid>https://www.rajasleem.com/posts/agentic-workflow-part-6-obedience-is-futile/</guid><description>The bottleneck in an agentic workflow is still you. Not your attention this time, your context window. Here is the guardrail I built so my own limits stop quietly steering what I build.</description></item></channel></rss>