Giving Your AI Agent a Clear Stopping Condition (or Watch It Loop Forever)

An agent without a stopping condition is just an expensive infinite loop. Decide upfront: how many steps, what counts as 'done', and what happens if it can't finish.

Some examples

• You set max_steps=10 in a research agent so it can't keep googling forever.
• Claude is told 'stop when the file passes the test suite' and it knows when to quit.
• An agent has a token budget of 50k and gracefully summarizes what it found when it runs out.
• ChatGPT given 'stop when you see a 200 response' exits as soon as the API call succeeds.

Try it!

Add a max_steps and a 'done when' condition to any agent loop you've written. Test that both actually trigger.

Practice this safely

Try this with a school, hobby, or family example where the stakes are low. Use the AI output as a draft you can question, not as the final answer.

1. 1Ask AI to explain stopping condition in plain language, then underline anything that sounds uncertain or too broad.
2. 2Give it one detail from "Giving Your AI Agent a Clear Stopping Condition (or Watch It Loop Forever)" and ask for two possible next steps plus one reason each step might be wrong.
3. 3Check loop against a trusted source, teacher, adult, expert, or original document before you use it.