<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>mpl blog</title><description>Notes on contracts, quality measurement, and audit trails for AI agent communication.</description><link>https://mpl.skelfresearch.com/</link><item><title>Reading mpl audit trails: forensic patterns</title><link>https://mpl.skelfresearch.com/blog/reading-audit-trails/</link><guid isPermaLink="true">https://mpl.skelfresearch.com/blog/reading-audit-trails/</guid><description>An MPL audit record is small. What you can do with a stream of them is not. A field guide to the queries that matter once the trail exists.</description><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Quality measurement vs guardrails: a different layer of the stack</title><link>https://mpl.skelfresearch.com/blog/quality-vs-guardrails/</link><guid isPermaLink="true">https://mpl.skelfresearch.com/blog/quality-vs-guardrails/</guid><description>Guardrails block harmful output. Quality measurement scores correctness against the contract. They are not the same job and they shouldn&apos;t share a budget.</description><pubDate>Wed, 20 May 2026 00:00:00 GMT</pubDate></item><item><title>Why agent-to-agent calls need contracts, not just JSON</title><link>https://mpl.skelfresearch.com/blog/contracts-not-just-json/</link><guid isPermaLink="true">https://mpl.skelfresearch.com/blog/contracts-not-just-json/</guid><description>JSON is a wire format, not an agreement. Once two agents start talking, the absence of a contract becomes the most expensive thing in the stack.</description><pubDate>Wed, 06 May 2026 00:00:00 GMT</pubDate></item></channel></rss>