Product
Prompt Doctor
Improve weak prompts directly from the screen with a mode designed for context, constraints, and clarity.
What it does
Prompt Doctor inspects a prompt visible on screen, diagnoses what is missing, and rewrites it into a more grounded, more constrained, and more usable version.
Input
- the visible prompt from the primary-screen capture
- the user's spoken request
- the selected mode context
Output
- a short spoken summary of what changed
- a markdown card that explains weaknesses in the original prompt
- a copyable improved prompt in
actionable_output - an optional clarifying question if the target model or use case is unclear
Why it is useful
Developers often know a prompt feels weak but do not want to break flow by pasting it into a separate review tool. Prompt Doctor keeps the original context intact, sees surrounding clues on the screen, and produces an edit that is immediately reusable.
Current implementation intentionally does not do
- multi-turn prompt engineering sessions
- prompt evaluation against real model outputs
- automatic detection of hidden context outside the visible screen
- direct write-back into the editor