As technology evolves, the safety of automatic modes has outpaced that of manual approval processes. Yesterday, Anthropic announced that, effective August 14th, the automatic mode will be enabled by default for Pro, Max, and Team users of Claude Code. This mode utilizes an independent classifier to scrutinize shell commands and operations. Testing reveals that it can detect 89% of dangerous commands, whereas manual approval can identify only 14%. This finding underscores the maturity of Agent tools' defense mechanisms at the infrastructure level, giving the company confidence in setting the automatic mode as the default to boost development efficiency.