phone_type_text — Claude Code prompt for text entry
The phone_type_text Claude Code prompt enters a complete text string into a phone field in one call. It can target a field by label or coordinates.
Apache-2.0 ★ 9,488 updated 2026-11-16
.claude/commands/phone_type_text.mdWhat phone_type_text does
The prompt types text into the focused field or targets one by its label, accessible name, or placeholder. A ref and snapshotId can identify the field directly. The name must match exactly one current input; otherwise, the result reports possible matches or asks for another screen observation.
Set pressEnter to submit after typing. The action reads the field back and reports whether its contents match the intended text; if the text did not appear, it retries once. On screens that cannot be inspected through the element list, provide the input box’s point or an @vN reference from phone_find_visually. Password values are hidden in readback.
What it helps with
- enter a full message in one call
- target an input by its label
- type into a field on an unreadable phone screen
- check whether the entered text matches
Example prompts for the agent
- Enter “The file will be ready today” in the message field and submit it.
- Type “Review complete” into the field labelled “Comment”.
- Find the input box on screen and enter “New address”.
How to install phone_type_text
Manually: put the files in place
.claude/commands/phone_type_text.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about phone_type_text
How do I install phone_type_text in Claude Code?
Add the command file phone_type_text.md to .claude/commands/.
What does phone_type_text do?
It enters text into a focused or selected input, then checks what the field contains. It can also submit the text with Enter.
How does phone_type_text target a field on an unreadable screen?
Provide the field’s coordinates or an @vN reference from phone_find_visually. The prompt types at that point and checks the result.
Source text
phone_type_text.md · Apache-2.0
Type the whole text into the focused field in one call; tap the field first, pass ref+snapshotId to target it, or name the field in element (its label, accessible name or placeholder as listed; it must match exactly one input field on the screen now, otherwise the result says which ones match or asks you to observe; a partial match is reported as resolvedFrom and matched). pressEnter submits after the text was entered. The result's effect reads the field back: readback is what the field shows now (hidden for password fields) and readbackMatches whether it is the intended text; text that did not land is retried once automatically. effect.kind is done, small_change, no_effect or place_changed as for other actions. On a screen the element list cannot read (an app that draws its own screen), give the input box's point instead of a ref: x+y (a visibleText center) or the @vN ref from phone_find_visually. The point is tapped and the text pasted into the box there, keyboard showing or not, then checked against the screen text: verification match means it went in; point_not_typed means that point did not take the text, so find the box again instead of repeating it. Without a ref or point, the text goes into whichever box holds focus the same way; no_focus means give the box's point. Without its point, name the box in element in a few words (e.g. the message box at the bottom): when the element list cannot read this screen and has no matching field, the box is found by sight (with phone_find_visually's checks) and typed into at its point the same way; foundBySight and seen say what was used.
© wangzan101 · Apache-2.0 · wangzan101/Mika