phone_extract — Claude Code prompt: what it does, how to install
The phone_extract Claude Code prompt reads visible accessibility data without taking a screenshot. It handles one screen at a time, including text, list rows, or editable fields.
The prompt extracts visible text nodes, list rows, or editable form fields. It groups list rows by their scrollable parent and includes field labels with their parent context.
It reads a single screen. Secret field values are omitted; phone_collect is the prompt for gathering content across screens.
What it helps with
read visible interface text without a screenshot
group list rows under their scrollable parent
inspect editable fields with labels and parent context
Example prompts for the agent
Extract the visible text elements from this screen.
Collect the list rows and group them by scrollable parent.
Show editable fields with their labels and parent context.
How to install phone_extract
Manually: put the files in place
.claude/commands/phone_extract.md
How to use
Paste the text at the start of a chat or into project instructions
Questions about phone_extract
How do I install phone_extract in Claude Code?
Add the prompt file at .claude/commands/phone_extract.md.
What does phone_extract do?
It extracts visible accessibility data from one screen: text nodes, list rows, or editable fields. Secret field values are left out.
How is phone_extract different from phone_collect?
phone_extract reads one screen, while phone_collect gathers content across screens.
Source text
phone_extract.md · Apache-2.0
Extract visible accessibility data without a screenshot: text nodes, list rows grouped by their scrollable parent, or editable form fields with labels and parent context. Secret field values are omitted. This reads one screen; use phone_collect for content across screens.