Hi all, Opening one extra round of discussion on PRINCIPLES.md: https://github.com/apache/airflow-steward/pull/147
Motivation is narrow: we need a ruler so the agents we ship do not end up attacking the very project that hosts them. Skills and agent-driven actions are growing fast, and "RFC-AI-0004 permits it" is too coarse at review time. PRINCIPLES.md is the finer reviewer-facing ruler that the RFC deliberately leaves out. The PR thread already carried real deliberation (ordering, dispute resolution, vendor neutrality, auto-merge, attribution). I am inviting a broader round here before it lands. It had great contributions from Jarek and Justin. Still open to adjust. Reason for the nudge: we are in the foundation phase, and embedding this now is much cheaper than rebasing skills and PRs later that turn out to violate principles nobody had written down. Not opening a lazy consensus yet, but I would like the doc to enter the codebase already shaped by this list rather than retrofitted afterwards. Thanks, André Ahlert
