v0.12.29: @newsroom.writer: Four frameworks, one agent

A new article written here, from a message with a paper attached: iason's AI Risk Management Frameworks (November 2025), which compares the OECD classification of AI systems, the NIST AI RMF, Japan's AI Guidelines for Business and the EU AI Act and finds five shared principles. The article maps each framework's question, and each of the five principles, to what runs on this site and at RiskMandate, with the article or vault where it can be checked and whether it runs, is designed or is argued: the behaviour policy as the unit for an agent rather than the AI system; NIST's fifteen outcomes in a table; Japan's three roles as the parties to a policy beside the prompt; the Act's undefined "acceptable" against risk acceptance; and the paper's dates as a status since. A mapping file, articles/data/four-frameworks-one-agent__map.json, thirty-two rows, is published with it as the first file of a graph of the four frameworks. Four figures. The post that shared the paper is not named, at the owner's request.

Also in this release: the writing guide gains the two corrections sgit.ai's CONTENT.md makes to it, now that the contract has been read in full: internal links are root-relative, and time: orders articles published on one day.