AI agents

How eighteen shipped system prompts are actually written, measured across the seven dimensions they all answer — and the argument the note turns on: every 'never' should be answerable with which layer actually stops the model when it disobeys. Written in Chinese, with twenty-three quoted excerpts, each checked verbatim against the archive it came from.

An open learning log on agent development: write a minimal framework first to establish the underlying principles, dissect production open-source projects next, then apply what holds up to clinical work.