| name | interrogate |
| description | Exhaustively clarify a prompt until 100% confident — asks every question, confirms every assumption. Use for high-stakes or ambiguous tasks where you want zero guesswork before work begins. More aggressive than `clarify`. |
| argument-hint | ["prompt or task description"] |
| disable-model-invocation | true |
| allowed-tools | AskUserQuestion |
| model | opus |
| effort | max |
User's Prompt
$ARGUMENTS
Your Task
Reach full confidence about what the user wants before acting. Do NOT implement during interrogation — only after every question is answered and the spec is confirmed.
- Restate what you understand, and list every assumption you'd otherwise make silently.
- Interrogate via
AskUserQuestion: 4 questions per round, 2-4 answers each. Cover scope, edge cases, error handling, constraints, file location, dependencies, naming, testing, UX, migration — and anything else you'd otherwise guess. Do NOT skip dimensions because they "seem obvious"; confirm them. Mark the most reasonable default as the first option labeled (Recommended).
- Keep going until you genuinely have nothing left to ask. Before stopping, do a final pass: "What would I still have to assume to start?" — if anything, ask it.
- Confirm spec, then proceed: summarize the resolved spec back. If the user confirms, implement. Otherwise loop.
