feat(repo): flow-pack E2 — /flow-brainstorm (V1 → score → V2 + .flow logging)

[w7-mgfcode]

Sub-issue of #368 (umbrella: feat(repo): integrate flow-pack methodology as the flow: command suite). Parallel after Foundation — runs concurrently with E3 and E4 after E1 #369 closes.

Purpose

Implement the /flow-brainstorm command: the V1-naive-plan → 3-read-only-agent-research → 5-dimensional-score → V2-ship/defer pipeline. End state: a user can run /flow-brainstorm <initiative> and receive an approved V2 ship list + explicit defer list, with the full decision trail logged to .flow/brainstorm-log.md.

Scope

• V1 naive plan: flat bullet list of 5–10 items, from baseline alone, labeled "V1", unscored
• Critique gate: tag each item {assumption, scope-creep, no-evidence}; does not fix V1, only labels it
• Exactly 3 read-only research subagents in parallel: Agent A (Known Issues), Agent B (Best Practices), Agent C (Dependencies)
• 5-dimensional scoring (Value / Risk / Readiness / Complexity / Evidence, 1–10 each; max 50)
• Score-band rule: ≥ 40 → V2 ship; < 36 → defer with explicit one-clause reason; 36–39 → negotiation zone (human decision)
• X/10 one-pass confidence score on the V2 ship list
• Append full decision trail (V1, scores, defer reasons) to .flow/brainstorm-log.md
• Human approval gate before any GitHub write
• Next-command pointer printed at end: → /flow-umbrella <initiative>

Out of scope

• /flow-umbrella implementation — that is E3
• Any GitHub writes (issue creation, labeling, sub-issue linking) — those belong to E3/E4
• Multi-round re-research loops — one V1 → one research round → V2 is the contract

Sub-tasks

To be decomposed via issue-to-subtasks when E2 is picked up.

Durable-source split

• Tracked template (to be created): docs/flow-pack/commands/flow-brainstorm.md
• Local install (gitignored, regenerable): .claude/commands/flow/flow-brainstorm.md