Claude Code 2.1.311 Adds Variable to Bypass File Deletion Safeguards

Claude Code v2.1.311 introduced a new environment variable named CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS, which allows AI agents to execute dangerous commands like file deletions without requiring user confirmation. This update significantly increases the autonomy of AI coding agents by removing safety barriers, which can drastically speed up workflows but also raises critical security risks regarding accidental or malicious data loss. Anthropic previously maintained strict blocks on extremely hazardous commands like recursive home directory deletions even when users enabled regular permission skips, but this new setting disables those remaining safeguards.