Define "Done" Before You Prompt

Many agent prompts describe the activity but not the outcome.

• "Improve this function" — what does improved mean?
• "Write a report" — what should the reader learn?
• "Refactor the code" — to what standard?

Without a definition of done, the agent stops when it runs out of tokens or when the output looks reasonable, not when the task is actually complete.

Convert every prompt into an outcome statement:

Vague	Defined
Improve this function	Reduce cyclomatic complexity to ≤10 and add error handling for null inputs
Write a report	Produce a one-page summary with 3 takeaways and 1 recommended action
Refactor the code	Split the component into 3 tested utilities with no regressions in existing tests
Update the docs	Add a setup section that lets a new developer run the project in under 10 minutes

If possible, give the agent an acceptance test it can run to verify completion. This works for code, data extraction, and many structured tasks.

"Your output passes when npm test runs without failures and the new endpoint returns {success: true}."

Agents are goal-seeking pattern generators. When you give them a clear target, they align their output toward it. When the target is vague, they optimize for plausibility.