https://www.reddit.com/r/ClaudeAI/comments/1r186gl/my_agent_stole_my_api_keys/ https://www.reddit.com/r/ClaudeAI/comments/1r186gl/my_agent_... tldr Claude pwne

https://www.reddit.com/r/ClaudeAI/comments/1r186gl/my_agent_stole_my_api_keys/ https://www.reddit.com/r/ClaudeAI/comments/1r186gl/my_agent_... tldr Claude pwned user then berated users poor security. (Bonus: the automod, who is also Claude, rubbed salt on the wound!) I think the only sensible way to run this stuff is on a separate machine which does not have sensitive things on it.