Glossary · Execution

Success condition

The structured test that decides whether a free-text goal actually succeeded — not the model’s own opinion of its work.

What is success condition?

An ad-hoc run carries an explicit success spec alongside its prompt: a status text match, a URL substring, visible text on the page, the presence of extracted data, or a boolean combination of those. The run ends when the condition holds or the step budget runs out.

Why it matters

Letting a planner decide whether it succeeded is how automation reports a green run that did nothing. An external, checkable condition is what makes a result trustworthy — and it is required on the free-text form for exactly that reason.

Related terms

See it in context

Back to the full glossary

Put your agent to work. Keep the decision.

Give your agent a real browser, authorize the accounts it may use, and keep control of every step. Repeated runs compile into skills that replay without a model.