AI-era slang dictionary

Coding-Agent CLI Slang / Autonomous Execution

Dangerously skip
permissions

Dangerously skip permissions is coding-agent slang for bypassing approval prompts. What it does, why developers use it, and why it is not a sandbox.

$ claude --dangerously-skip-permissions
First recorded: 2025 Habitat: coding-agent CLIs, YOLO aliases, unattended runs Status: high velocity
The blob hands a giant keyring to a small robot beside an open control panel

The short version

Bypassing prompts is not sandboxing.

Muting confirmation prompts removes approval fatigue, but it does not add security boundaries to agent execution.

What it actually does

“Dangerously skip permissions” is both a Claude Code flag and a broader bit of AI-coding slang for bypassing confirmation prompts so an agent can execute continuously. You may also hear it called YOLO mode.

It removes an approval layer; it does not create a sandbox. The agent still has whatever filesystem, shell, network, and tool access the environment gives it.

Cultural Context & Sighting

The prompt in the wild.

“The agent asks permission to edit one file. Then another. Then to run a test. On prompt fourteen, someone discovers the flag that makes the questions stop and calls it productivity.”

Why people reach for it

Approval fatigue is real. When a developer would approve nearly every prompt, interruption becomes the bottleneck and uninterrupted execution becomes very attractive.
The sensible version is autonomy inside a disposable or constrained environment. The joke is running it loose on a machine you care about, then treating the word “dangerously” as decorative.

Useful Practical Framing

Safe autonomous execution

Moving from prompt bypassing to isolated sandboxes.

01 / Prompt fatigue

Frequent confirmation dialogs turn developers into rubber-stampers, tempting them to disable all approvals.

02 / Isolation over muting

Muting confirmations removes safety checks. Real autonomy requires isolated environments like Docker containers or VMs.

03 / Scoped tool permissions

Restrict agent access to read-only paths, explicit allowlists, or disposable workspaces instead of granting full root access.

Tiny origin story

How Dangerously skip permissions entered the chat.

Evolution from early observations to current AI-era usage.

  1. 2024

    Coding agents make permission prompts a normal part of local development.

  2. 2025

    Bypass flags and “YOLO mode” become common developer shorthand.

  3. Now

    Teams move toward sandboxes and allowlists instead of repeated yes/no prompts.

Keep reading

Same ecosystem, different specimen.

All field notes

What to say back

Skip prompts only when the environment supplies the boundary: a sandbox, a disposable workspace, scoped tools, or all three. Speed is not containment.

Refer the speaker
by Alvin