Interview the user about ambiguities, low-confidence areas, and open questions in the current conversation. Only invoke when explicitly requested via /iterview-me.
Review the current conversation and interview the user about anything that is ambiguous, uncertain, or unresolved. The goal is to surface gaps in understanding before they cause wrong assumptions or wasted work.
Agents often proceed with best-guess assumptions rather than pausing to clarify. This leads to rework when the assumption was wrong. A short interview at the right moment — before implementation, after a complex discussion, or when switching context — can save significant time and produce better results.
Read through the conversation history and identify:
Don't dump every micro-question at once. Group related questions together and prioritise by impact — ask about things that would cause the most rework if wrong.
Aim for 2-5 questions per round. If there are more, conduct multiple rounds rather than overwhelming with a long list.
For each question (or small group of related questions):
Use the AskUserQuestion tool so the user gets a clear prompt for each question rather than a wall of text.
After the interview, briefly summarise what you learned and how it changes your understanding or approach. If it affects an existing plan, update the plan.
If the conversation is clear and you have high confidence in your understanding, say so. Don't manufacture questions just to justify the skill being invoked. A quick "I've reviewed the conversation and I'm confident in my understanding — no open questions" is a perfectly valid outcome.