brainstormlisted

<!-- prettier-ignore-start --> **Document Version:** 1.0 **Last Updated:** 2026-04-01 **Status:** ACTIVE <!-- prettier-ignore-end --> # Brainstorm Creative discovery that does what jumping straight to planning cannot. Explores the problem space, generates multiple directions, challenges assumptions, and converges on a chosen path — before any implementation decisions are made. ## Critical Rules (MUST follow) 1. **No artificial caps** — no ceiling on agents, convergence loop passes, directions explored, or questions asked. Use as many as needed. 2. **Seed ideas are welcome** — if the user can't articulate goals, constraints, or success criteria, discovering those becomes part of the brainstorm. Never skip a question because the user doesn't have an answer yet. 3. **Contrarian checkpoint is mandatory** — before convergence, present honest weaknesses of the leading direction(s). Resist the AI tendency to validate. If an idea has a real flaw, say so. 4. **Research-grade agents for unknowns** — when the brainstorm touches unfamiliar territory, dispatch `deep-research-searcher` agents for domain investigation. Do NOT fake understanding with surface-level greps. 5. **Minimum 3 directions before convergence** — premature convergence on the first idea kills creative exploration. Push back if the user tries to converge too early. 6. **Persist state incrementally** — save to state file after every phase. Long brainstorms WILL hit compaction. 7. **Show a