<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Blogus Blog</title><description>Engineering notes on treating prompts as a first-class dependency: extracting them from code, versioning them with content hashes, and auditing changes the same way you audit deps.</description><link>https://blogus.skelfresearch.com/</link><language>en-us</language><item><title>Auditing prompt changes the same way you audit dep upgrades</title><link>https://blogus.skelfresearch.com/blog/auditing-prompt-changes/</link><guid isPermaLink="true">https://blogus.skelfresearch.com/blog/auditing-prompt-changes/</guid><description>Your team probably has a playbook for reviewing a uv lock update. This post is about applying the exact same playbook to a prompts.lock update — and what changes when you do.</description><pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Prompts as a first-class dependency in the build graph</title><link>https://blogus.skelfresearch.com/blog/prompts-in-the-build-graph/</link><guid isPermaLink="true">https://blogus.skelfresearch.com/blog/prompts-in-the-build-graph/</guid><description>If a prompt is a thing your app depends on, your build graph should know about it. A short essay on why the analogy is exact and where blogus fits into pre-commit, CI, and release pipelines.</description><pubDate>Fri, 22 May 2026 00:00:00 GMT</pubDate></item><item><title>Why your inline prompts drift faster than your code</title><link>https://blogus.skelfresearch.com/blog/why-inline-prompts-drift/</link><guid isPermaLink="true">https://blogus.skelfresearch.com/blog/why-inline-prompts-drift/</guid><description>Inline prompts look stable in a PR. They are not. Here is a precise account of how they actually rot — and the small set of operations a tool like blogus has to do to stop it.</description><pubDate>Tue, 12 May 2026 00:00:00 GMT</pubDate></item></channel></rss>