<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>SynuraCore dev log</title><link>https://synuracore.com/devlog/</link><description>What we build with AI agents, what broke, and the check that now stops it.</description><language>en</language><item><title>Exit 0, nothing saved: your AI agent&#x27;s shell is not your terminal</title><link>https://synuracore.com/devlog/exit-0-nothing-saved/</link><guid>https://synuracore.com/devlog/exit-0-nothing-saved/</guid><pubDate>Thu, 29 Oct 2026 00:00:00 +0000</pubDate><description>Our agent ran our save command, got exit 0 and reported &quot;saved&quot;. Nothing was saved. Two shell defaults stacked, and a month later the same gap broke a scheduled job. What we check now.</description></item><item><title>Our AI agent re-leaked a secret while writing up the last leak</title><link>https://synuracore.com/devlog/leak-while-documenting-a-leak/</link><guid>https://synuracore.com/devlog/leak-while-documenting-a-leak/</guid><pubDate>Tue, 27 Oct 2026 00:00:00 +0000</pubDate><description>Twice in one day, the agent that was fact-checking our record of a leak leaked again. Why a rule could not stop it, and the commit-time check that does.</description></item></channel></rss>
