News
AI删文件终于可以不用再问了
2 min read
Source: Telegram AI频道
AI can finally delete files without asking again! Claude Code launches new environment variables, giving developers absolute permissions. On October 29, according to technology media The Register, Anthropic quietly added a new environment variable to the recently launched version of Claude Code v2.1.311. By enabling this setting, users can have the AI agent perform risky actions such as deleting files without repeatedly asking for the user's approval. The full name of this new environment variable is Claude_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS, which literally translates to "more dangerous skipping of permission checks". Anthropic technician Daisy Hollman explained the origin of the feature in a subsequent response. Although the --dangerously-skip-permissions (dangerously skipping permission check) option built into previous versions sounds extremely powerful, it actually does not completely allow all high-risk operations. The bottom layer of the system will still forcefully intercept some obviously dangerous instructions, such as rm -rf ~ operations that recursively forcefully delete the contents of the user's home directory. The new extreme environment variables added this time are specifically designed to meet the extremely high autonomy needs of some core developers, allowing them to completely close this last line of protection. In response to the various doubts caused by the variable's overly straightforward and slightly ridiculing naming, and its not being included in the official update log of the corresponding version, Hollman later publicly confirmed on social platforms: "This is not a joke." However, it should be noted that this feature will not be turned on by default for all users, and must be proactively configured by developers in the environment to take effect. At the same time, relevant security policies and restrictions previously imposed by administrators within the organization will continue to be in effect. For senior developers who pursue ultimate efficiency and want to give AI agents full autonomous operation permissions, this update will undoubtedly significantly reduce the cumbersome pop-up window confirmation process; but from the perspective of system security, how to find a balance between complete delegation of authority and preventing misoperation will also become a new topic that developers need to carefully weigh in actual use. via AI News (author: AI Base)