Set a project goal with an AI agent
Turn a loose intention into a result someone can verify. Name what should change, what evidence counts and when the agent should come back to you.
To set a project goal with an AI agent, describe what you want to be able to check at the end. In Codex, define-goal turns that description into a proposed goal. Making a course pack easy to find is different from renaming every file in the course. A clear goal tells the agent which change would count as success.
A good criterion can be numerical, but it does not have to be. A reviewer being able to trace each claim to a source may be more meaningful than a page count. The guide also checks for an existing goal so the agent does not silently replace it.
What it does
- Define an observable outcome
- Choose evidence that demonstrates completion
- Set a boundary around the work
- Identify when a decision needs your input
- Check whether the chat already has a matching goal
Install and use
- Install the skill in a Codex conversation with the goal tools enabled. Describe the outcome and a constraint that matters to you.
- Try “Help me define a goal for reorganising our course folder. Published files must keep their names.”
- Review the completion criteria. Explicitly ask to record the goal when you want it active.
FAQ
Source
Made by OpenAI. Original skill: define-goal, published on GitHub under the Apache-2.0 licence.
View on GitHub