We just deployed some real-time protections to Claude Code and it's starting to flag calls that are risky! Just saw it stop an attempt to read .env.development – lots of keys in there, many of which we wouldn't want to be sent over to the LLM.

→ View original post on X — @jiquanngiam