Skip to content

Security & Sandboxing

  • Type: Boolean (1, true, 0, false)
  • Default: false
  • Description: Indicates Claude Code is running in a sandboxed environment. Bypasses trust dialog.
  • Example: export CLAUDE_CODE_SANDBOXED=1
  • Type: Boolean (1, true, 0, false)
  • Default: false
  • Description: Set to ‘1’ to indicate running inside Bubblewrap sandbox. Allows —dangerously-skip-permissions with root.
  • Example: export CLAUDE_CODE_BUBBLEWRAP=1
  • Type: Boolean (1, true, 0, false)
  • Default: false
  • Description: Add x-anthropic-additional-protection header to API requests. Used for heightened security.
  • Example: export CLAUDE_CODE_ADDITIONAL_PROTECTION=1

CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD

Section titled “CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD”
  • Type: Boolean (1, true, 0, false)
  • Default: false
  • Description: Load CLAUDE.md files from additional directories. Expands instruction loading to more locations.
  • Example: export CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD=1
  • Type: String (JSON object)
  • Default: Unspecified
  • Description: JSON object mapping script names to numeric capability limits. Scripts not listed have no caps.
  • Example: export CLAUDE_CODE_SCRIPT_CAPS='{"bash":100,"python":50}'

← Back to env/README.md