Proposal: LLM-Friendly Brand Guidelines via brand.json / brand.txt | Not Hacker News!