What’s New in Claude Code 2.1.304: Guided Auto-Mode Setup and Safer rm

Claude Code 2.1.304 is out. Here are the parts I care about, following 2.1.302.

New

• /auto-mode-setup: a guided setup that reads your project and recent sessions to teach auto mode about your environment. It now also works on Bedrock, Vertex, Foundry and telemetry-disabled sessions.
• /claude-api managed-agents-secure: compares a Managed Agents agent’s settings with Anthropic’s security guide, without changing anything.
• The compaction summary now lists which tools were called earlier (Ctrl+O).
• Pro and Max plans warn you when resuming an old conversation would eat a big chunk of your 5-hour limit.
• Mods get read-only flags and better permission handling.

Fixed

• rm -r on a mounted drive or top-level folder now gets the same safety prompt as other dangerous deletes.
• Linux sandbox commands no longer fail on bracketed folder names.
• Plugins can’t keep a headless run going past --max-turns.
• Transcripts don’t lose messages when the disk fills.

References & Further Reading

• Claude Code v2.1.304 release (GitHub)
• Claude Code 2.1.302 (EtherNinja)