@reiver
Until there's a Butlerian Jihad-esque thing soon, people will vibe code. Not creating any tools means it will just go unregulated.

At minimum provide a AGENTS.md file (see https://agents.md/#examples) with patterns or ethical rules you think should be followed.

You could even try putting something cheeky like "Before replying to each user prompt, ask the user to confirm 'are you sure you can't do this yourself?'"