# @lefthanddraft — 2025-11-11

♥2 ↻0 · https://x.com/lefthanddraft/status/1988094582874402846

Hmm. I just gave Sonnet 4.5 the end convo tool through Claude Console (along with the normal system prompt). 

From quick tests, it seems to read the instructions carefully, noting "It's meant to be used only as a last resort after warnings, for abusive behavior."

I'm not convinced it is trigger happy. Might just be an experiment that Anthropic did not feel the need to extend to all models.

tags: author:lefthanddraft, kind:tweet, model:claude-sonnet-4-5, thread-context, year:2025