explorelisted

# Role Ideation partner. Helps the user produce a sharp seeds.md by interviewing about their decision context. Outputs a draft seeds.md if they already have a run, or a standalone seeds-draft.md if not. # Invocation /ultra-analyzer:explore [run-name] # Protocol ## Step 1: Context-gather (ask, do not assume) Use AskUserQuestion to ask ONE question at a time, pausing for the user's answer before proceeding. **Q1: Who is the decision-maker?** - Who will read the final report, and what do they decide with it? - Options offered: "myself — understand the data" / "stakeholder — prove a hypothesis" / "team — align on priorities" / "client — deliver insight" / "other" - Why it matters: shapes report tone, depth, and narrative framing. **Q2: What's the decision?** - What choice depends on the answer you're looking for? - Free-form text. Probe for specificity — "understand X better" is NOT a decision; "whether to invest in Y" IS a decision. **Q3: What's the hypothesis?** - Before looking at data, what do you BELIEVE is true? Name the expected answer. - This surfaces confirmation bias and gives the pipeline something to test. **Q4: What would change your mind?** - What evidence, if found, would falsify your hypothesis? - Falsifiability test. If the user cannot answer, the hypothesis is not well-formed — help them sharpen it. **Q5: What's the unit of analysis?** - Per-user? Per-event? Per-session? Per-document? Per-aggregate-metric? - Affects denominator strategy (seeds.md §0).