Prompt builder On-device
Write a prompt in parts, get it linted and exported.
- Task Task is empty.
Paste on the left to start.
Paste above to start.
Everything runs on your device. Nothing you paste is sent anywhere, and it's cleared when you close the tab.
Build a prompt from separate sections instead of one long paragraph. The builder checks it as you type, flags a missing task, vague wording and placeholders left behind, and exports Markdown, XML or chat-message JSON. Section headings stay in English in every export. Nothing you write is sent anywhere.
About this tool
What you provide
- Text typed into role, context, task, constraints, examples and output format sections.
What you get
- The assembled prompt as Markdown, XML or chat-message JSON.
- Lint findings with fixes where a fix is mechanical.
Limits
- The token figure is an estimate.
- Lint findings are suggestions, not errors.
- Section headings stay in English in every export.
How to use it
-
Fill the sections
Write the task and any role, context, constraints or examples.
-
Fix the findings
Read the lint list and apply or ignore each suggestion.
-
Export
Choose Markdown, XML or JSON, then copy or download.
Frequently asked questions
Why are the section headings in English?
Models were trained mostly on English headings such as Role and Task, so they stay in English in every export, whatever language the content is in.
What does the lint check?
It flags a missing task, vague wording, placeholders left in, unclear JSON formats and prompts that are very long. Each finding explains itself; none blocks you.
Is my prompt saved anywhere?
Only in this tab's session storage, so a reload restores it and closing the tab clears it. Nothing is uploaded.