explain-permission

A Claude Code mod. When Claude stops to ask "is this ok?" before running something, a pane opens beside the conversation and explains the request in plain English:

• WHAT (cyan): what it will actually do on your computer or online
• WHY (magenta): why it's needed for what you asked, based on the conversation
• RISK (green / yellow / red): low, medium or high, and anything that gets sent, deleted, published or paid for

The dialog appears right away; the explanation fills in a few seconds later. Read the pane before you answer.

Use it

claude --plugin-dir /path/to/explain-permission

Type /explain-permission to open the pane yourself. It docks on the right in fullscreen mode, and otherwise sits above the prompt. It opens by itself on terminals at least 144 columns wide.

Good to know

• Needs a recent Claude Code with mods (function hooks), an early-access feature.
• Each explanation is one extra model call on your own account. It asks Claude, with the conversation so far, and falls back to Haiku without the conversation if that fails.
• It only explains. It never approves or denies anything for you.

Made by Hermann Peterscheck · https://hermann.fyi

Install from GitHub

This folder is also a plugin marketplace. Once it's in a GitHub repo:

claude plugin marketplace add petershk/explain-permission
claude plugin install explain-permission@hermann-mods

or the same with /plugin inside Claude Code. Needs Claude Code 2.1.287 or later.