Anthropic Agent SDK v0.2.158: Verbatim Prompts for Security

Anthropic's Claude Agent SDK v0.2.158 introduces the new `verbatim_prompts` option in `ClaudeAgentOptions` to enhance security. This feature prevents untrusted user messages from triggering unintended file reads or command execution by disabling `@path` expansion and slash-command dispatch. Developers can now ensure prompts are delivered to the CLI exactly as written, requiring CLI 2.1.248+.