@harrypotter482
Founder & Software Developer, ModuleFlow
1h ago · Vibe-coding · Quito

How do you keep “while we’re at it” out of an agent’s task?

A coding agent can suggest a reasonable improvement that still doesn’t belong in the current change.
Consider a hypothetical task: add a status filter to a table.

During implementation, the agent proposes saved views, a toolbar redesign, and a component refactor. These may be worthwhile, but they introduce new behavior and more things to verify.

In the process I’m developing with Rigorkit, I want the distinction to stay explicit:

  • Required now: work needed to deliver and verify the agreed behavior.
  • Deferred: useful improvements that can be evaluated separately.
  • Scope change: a new requirement that needs an explicit decision before implementation continues.

A dependency that prevents the filter from working deserves attention. An unrelated cleanup deserves a recorded follow-up.

The aim is to preserve useful ideas while keeping the current delivery bounded and reviewable.
What rule do you use to decide whether an agent should expand a task, ask for a decision, or record a follow-up?

github.com/harrypotter48/rigorkit

AgenticCodingScopeManagementAIWorkflowRigorkit

Comments

No comments yet
Start the conversation.
Log in to join the thread.