The most common reason an agent fails is a task that is too big or too vague.
A good task names the files, says what done looks like and can be checked by a test. If you cannot write the check, the task is not ready.
Our template has four lines: goal, files, constraints and how to verify. Copy it and adjust it to your project.
