Start with your task
Make a simple event web page
Request one static page and check it locally step by step.
- Prepare materials
- Choose your AI tool
- Get instructions
- Check the result
Describe the result, add only the materials it needs, then choose what counts as done.
What to ask for
One self-contained HTML file returned as text, manual saving instructions and a test sheet. The page only presents event information.
- Plan + complete HTML
- Manual local preview
- Facts + mobile + keyboard checks
Follow these steps
- Open the example and replace the event facts. Keep the first version to four information sections.
- Choose a tool and copy the instruction. Ask for a plan, one HTML file and a manual test sheet.
- Review the code for scripts and external links. If it matches the brief, manually save the HTML as index.html and open it in a browser.
- Check the facts, 360px layout and keyboard section links. If a check fails, report the action, expected result and actual result; request one repair and rerun the whole sheet.
Bring these inputs
- Confirmed event facts; mark missing contact and capacity
- Four sections: event, details, what to bring and questions
- A text editor, browser and a place to manually save the file
Check before using
- Facts match the input and missing values remain To confirm.
- At 360px, no horizontal scrolling; Tab and Enter work on section links.
- No scripts, forms, external requests or unperformed test claims.
A common gap → a clearer request
The assistant adds a working sign-up form or says tests passed without running them.
Restate the static-page limit, remove the form and require each unperformed check to say Not yet run. Report layout failures with width, action and what you saw.
Choose a tool
Use Codex or Claude to request HTML as text. LM Studio requires a locally installed model capable of producing HTML. Copy the matching generated instruction; use a text editor and browser for the manual preview.
What this flow covers
This is a static information page for local preview. No sign-up, account, database or deployment is included. The prompt asks for code as text and grants no shell, file-writing or network permission. Review code before you save and open it yourself; generated code and its tests remain unverified until checked.
The example prepares an instruction using synthetic material. You choose an AI tool and verify its answer yourself.