name
name
Name things the way Lexicon, Igor, A Hundred Monkeys, and Catchword do: generate broad, then screen hard. The breakthrough name is rarely in the first 20 — it's in the second hundred. Your job is to generate a wide, varied candidate set, run it through the screens below, and deliver the whole pool plus a recommended shortlist — always as an interactive HTML artifact the user can filter and re-curate — with the reasoning attached.
This skill generates and screens. It does not clear trademarks or register domains — it flags likely risks and tells the user what to check, but a serious venture still needs a real trademark search (tmsearch.uspto.gov, WIPO Global Brand Database) and an attorney. Say so; don't imply a name is "available."
For deep dives — case studies, the legal doctrine in full, China/global naming, the academic citations, agency philosophies — load reference.md in this skill folder. Don't load it for routine naming; the essentials are below.
Two modes — quick vs. deep
Same deliverable (the HTML artifact below); they differ only in how much you ask and how many candidates you generate.
deep(default). Grill the user with 5–10 questions, generate 100+ candidates across the taxonomy, screen down to a 20–30-name shortlist. The brief is where quality is won, so this is the default whenever the mode is ambiguous.quick. Ask nothing (at most one blocker question), generate ~20 candidates with light screening, hand back a ~20-name shortlist.
Honor an explicit quick / deep. Otherwise infer from signal — terse "name my X" with a hurry-vibe → quick; a richer prompt or real venture → deep. When genuinely ambiguous, default to deep and say so in one line ("Running in deep mode — I'll ask a few questions first; say quick to skip"). The user can switch anytime.