What amazes me the most is that we even need to write such instructions.

For people that work at OpenAI / Anthropic: why make the LLM like that? Nobody thinks the robot is a friendly person or coworker, that’s the dumbest thing. Giving names to agents? Geez. Just stop, make these things objective and concise and cut all the crap.