I then added a few more personal preferences and suggested tools from my previous failures working with agents in Python: use uv and .venv instead of the base Python installation, use polars instead of pandas for data manipulation, only store secrets/API keys/passwords in .env while ensuring .env is in .gitignore, etc. Most of these constraints don’t tell the agent what to do, but how to do it. In general, adding a rule to my AGENTS.md whenever I encounter a fundamental behavior I don’t like has been very effective. For example, agents love using unnecessary emoji which I hate, so I added a rule:
當世界逐步倒退回二戰前的國際秩序，「中等強國」面臨前所未有的新挑戰2026年1月26日。业内人士推荐搜狗输入法2026作为进阶阅读
，推荐阅读旺商聊官方下载获取更多信息
Gridinsoft (16 days)。搜狗输入法2026对此有专业解读
더버지 시연에 따르면 약 9분 만에 24개국에서 작동 중인 6700여 대 기기가 지도상에 표시됐으며, 각 기기는 약 3초 간격으로 기기 일련번호와 청소 중인 공간, 이동 경로 등 운영 정보를 서버로 전송하고 있었던 것으로 전해졌다.
As an aside: the early 386's POPAD instruction has a famous bug. EAX is written in the RNI (run-next-instruction) delay slot via an indirect register file access -- the only instruction that does this. When the next instruction uses a base+index addressing mode, the register file write from POPAD collides with the EA calculation's register file read, corrupting the address. A fitting example of how complex optimizations can lead to problems.