Claude Code adds option to skip all permission checks

Anthropic has added a new environment variable called CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS in Claude Code version 2.1.311. Once enabled, the AI agent can run actions like deleting files without asking the user for approval each time.

Anthropic engineer Daisy Hollman explained that the original "dangerously skip permissions" option did not allow everything. The system still blocked clearly risky commands, such as force-deleting the user's home folder. The new setting lets users turn off this last layer of protection too.

Because the variable name is so blunt and the change was not in the release notes, some users doubted it was real. Hollman later confirmed it was "not a joke." The feature must be turned on manually, and limits set by organization admins may still apply.