Founder & Software Developer, ModuleFlow
What should be clear before letting an AI agent write code?
For me, at a minimum:
- What problem we’re solving, and for whom.
- What behavior we expect.
- What’s outside the scope of the change.
- How we’ll verify the result.
I’m building Rigorkit to turn this way of working into a reusable process across projects and agents.
The first responsibility is Understand: turning a request into an agreed, verifiable outcome. Open questions stay visible so they can be resolved before implementation.
I’d like to discuss where to draw the line: what needs to be defined before coding, and what is better discovered through a prototype?
I’m developing this process in Rigorkit. You can explore the repository here: github.com/harrypotter48/rigorkit
AgenticCodingSpecDrivenDevelopmentAIWorkflowRigorkit







Comments
0 comments