ce-idea-scoutlisted

# Idea Scout **Invocation input:** Domains / constraints (team size, capital ceiling, unfair advantages): $ARGUMENTS Turn a domain / constraint brief into a **ranked shortlist of directions worth pursuing**, through a funnel that spends cheap tokens widely and expensive tokens only where they pay: **generate 15-25 candidates → cheap 3-axis filter → deep-dive the 5-6 survivors on the full `ce-idea-score` rubric → rank, with a top-3 and an avoid-list**. The deep-dive *is* the `ce-idea-score` engine; this skill owns the generation, the triage, and the cross-idea ranking around it. > **A verdict tool, deliberately — and a funnel, deliberately.** Like `ce-idea-score` > (and unlike `/ce-market-scan`), this skill ranks and recommends. Two failure modes it is > built to avoid: (1) **shallow-everywhere** — analyzing 12 ideas at full depth in one > pass produces generic filler and thinner late ideas, then ranks them as if analyzed > equally; the funnel fixes this by deep-diving only survivors. (2) **fabricated > rigor** — ranking ideas on market "facts" the model invented; the evidence discipline > (below) fixes this. The whole artifact is labeled **OPINIONATED decision support — a > ranked shortlist, not validated fact.** > **Where it sits.** Furthest upstream: > **`ce-idea-scout`** (generate + triage → shortlist) → human picks → `ce-idea-score` (deep > verdict on one idea) → human commits → `/ce-market-scan` (optional — the **stakes dial**: > skip for a fast pick, run it to vali