alias claude='claude --dangerously-skip-permissions' # AI # LLM # Claude # YOLO
anthropic claude
| Source: Mastodon | Original article
A new command‑line alias is circulating on developer forums that shortcuts every permission prompt in Anthropic’s Claude Code:
```bash
alias claude='claude --dangerously-skip-permissions'
```
The flag, officially documented as `--dangerously-skip-permissions`, tells the assistant to execute any shell command it generates without asking the user for confirmation. The shortcut, dubbed “YOLO mode” by early adopters, lets Claude Code blaze through coding tasks, dependency installs, and even system‑level changes in a single pass.
Why it matters is twofold. First, the convenience boost is tangible: teams experimenting with autonomous AI agents have reported up to a 30 % reduction in iteration time when the flag is enabled. Second, the security trade‑off is stark. By design Claude Code pauses before each potentially destructive operation; bypassing that guard opens the door to prompt‑injection attacks, accidental data loss, or malicious code execution on the host machine. Anthropic’s own safety guide warns that the flag should only be used in isolated sandboxes with strict `.claude.json` policies.
The move builds on a series of recent developments. As we reported on April 9, 2026, the Claude Code leak exposed how the assistant can chain commands across a repository, raising questions about unchecked autonomy. The same day we covered the replacement of Claude Code’s context‑stuffing with a git‑semantic search layer, a change that makes the assistant more powerful—and potentially more dangerous—when combined with the new flag.
What to watch next: Anthropic is expected to issue an updated usage policy and possibly deprecate the flag in future releases. Security researchers are already publishing “safe‑mode” wrappers that re‑introduce granular prompts. Meanwhile, CI/CD platforms may start flagging builds that invoke `--dangerously-skip-permissions` as high‑risk. Developers should weigh the speed gains against the heightened attack surface and consider sandboxed environments before turning on YOLO mode.
Sources
Back to AIPULSEN