What is the exact threat being addressed here? If the proxy detects a sensitive key and then places it in a .env file, Claude code would pull the value from the env file and then still send it to Anthropic servers wouldn’t it?
jricramc•3h ago
Correct, it doesn't protect against a malicious LLM actor, but rather places guardrails on the developer as well as the agent.
jricramc•6h ago