Using agents in
daily research.

Recipes for giving research tasks to agents: the steps, how to check the output, and the errors reported so far. Skills come as briefs your agent can turn into a version for your own institute. Anyone can contribute.

Comment submitted Theis, F. J. & Helmholtz Munich Computational Health Center. Delegate what you can check: ten lessons on AI agents in research. A link follows on publication. Best practices: aiforscience.eu →

AI for Scientists

What is on this site

Hand agents the work whose results can be checked cheaply. Every recipe states how to check its output.

Recipes →

One task per recipe: the steps, how to check the output, and known problems.

Skill briefs →

Copy a brief, give it to your agent, answer its questions about your institute, and get your own version of the skill.

What went wrong →

Errors reported with the recipes, and which recipe now prevents each one.

Contribute

The first recipes come from one centre. They get better when other groups test them.

Report an error →

An agent made an error others should know about. Names and details are removed by default.

Also: which work to give to agents · where to draw the line · the idea behind the site.