Agent permissions
What an agent may touch, tool by tool: Read and Write checkboxes where Write reaches back along a drawn line to bring Read with it, and a risk meter that climbs from Low to Medium.
Toggles and settingsNo. 0542 variationsFree with an account
Playing on its own
Variation
- The idea
- Tools by Read and Write, with risk
- Autoplay
- Read cascades on for Calendar and Files; the cursor ticks Write on Email and a line arcs back to Read, which springs on as the risk meter moves from Low to Medium; then Read off takes Write with it, and Write goes back on.
- When you click
- Real checkboxes: click or Space. Write turns Read on with it; Read off turns Write off. The risk meter and the summary follow every change.
- Look
- grid, cascade and link lines, sans with mono labels type
- Variations
- Sand, Terminal
- Files
- index.html, style.css, script.js and prompt.md. No framework, no CDN, works offline.
The prompt and the files
Paste the prompt into Claude, ChatGPT, Lovable or any coding agent to rebuild it in your own product, or take the three files as they are.
Loading the files