Brainstorm

Brainstorm Assistant | 腦力激盪助手

Language: English | 繁體中文

Structured ideation before specification writing. Transform vague ideas into actionable feature proposals through guided brainstorming — grounded in 2024–2026 research on AI-assisted ideation.

在撰寫規格前進行結構化發想。以 2024–2026 年 AI 輔助發想研究為基礎，透過引導式腦力激盪，將模糊構想轉化為可執行的功能提案。

Implements: XSPEC-296 Brainstorm Quality Standard (BQS v1) — brainstorm v4, layered on XSPEC-247 brainstorm v3 (Multi-Persona Ensemble + Multi-Critic Convergence).

What changed in v3 | v3 的核心改動: Divergence is no longer a single AI voice racing to a count — it is a persona ensemble (each role reasons via chain-of-thought, in isolation) crossed with diversity lenses. Convergence is no longer one AI scorer plus one devil's-advocate — it is a multi-critic panel plus a hard-role rebuttal (Devil's Advocate + Steelman). This directly targets the strongest finding in the literature: multiple personas beat a single pass, and a single LLM critic is weak and prone to sycophancy.

v3 把發散從「單一 AI 衝數量」改為persona 集成（每個角色以思維鏈獨立推理）× 多樣性透鏡；把收斂從「單一 AI 評分 + 單一反駁」改為多評審面板 + 硬角色反駁（Devil's Advocate + Steelman）。直接對應文獻最強結論：多 persona 勝過單一 pass，單一 LLM 評審既弱又易諂媚。

What changed in v4 | v4 的核心改動: v3 supplied strong mechanisms but no decidable pass/fail quality gate. v4 layers a time-sequenced quality contract on top — the Brainstorm Quality Standard (BQS v1) — without removing any v3 behaviour. BQS is a four-layer × timeline structure: Layer 0 declares explore/exploit intent; Layer 1 (process, leading, visible during divergence) runs dimensions D1–D4; Layer 2 (product, leading, applied to the Recommended Set only after convergence — ideas with Agg. Score ≥ 3.5, uncapped) runs D5–D8 plus a Seeds column and a contested zone; Layer 3 is an ungoverned Judgment Override. First principle: decisions use leading signals; calibration uses lagging signals — these are ordered in time, not a right/wrong trade-off. The hard sequence gate forbids D5–D8 from being revealed or scored during divergence. See "BQS v1 — Quality Contract" below.

v4 的核心改動： v3 提供了強健的機制，卻沒有可判定 pass/fail 的品質閘。v4 在其上疊加一道時序化品質契約——腦力激盪品質標準（BQS v1）——且不移除任何 v3 行為。BQS 是四層 × 時間軸結構：第 0 層宣告 explore/exploit 意圖；第 1 層（過程、leading、發散期全程可見）跑維度 D1–D4；第 2 層（產物、leading、收斂後僅施於推薦集，即 Agg. Score ≥ 3.5、不限筆數的想法）跑 D5–D8 加 Seeds 欄與爭議區；第 3 層是不受治理的 Judgment Override。第一原理：決策用 leading 訊號、校準用 lagging 訊號——兩者是時間前後，而非對錯取捨。 硬序列閘禁止 D5–D8 在發散期被揭示或評分。詳見下方「BQS v1 — 品質契約」。

What changed in v4.2 (XSPEC-388) | v4.2 的核心改動: D4 pinned where the judge sits and said nothing about whether it can think. An independent context running a literal-following model satisfies the old wording verbatim and rubber-stamps its way through D5–D8 — indistinguishable on paper from a review where someone actually objected. v4.2 gives D4 a second axis: the review must also declare its required model level (see the terminology note below). The CONVERGE briefs are re-shaped as goal + constraints — D5–D8 are criteria the judgment must satisfy, not steps to execute in order; the fixed sentence patterns and per-criterion scoring guides are kept as reference formats. The divergence side cites core/model-selection.md (Standard or higher) instead of adding a rule of its own. No v3 mechanism is removed.

v4.2 的核心改動（XSPEC-388）： D4 原本只釘住判官坐在哪裡，一個字都沒說它得想得動。一個獨立 context 跑字面遵循型模型會逐字滿足舊條文，然後照 D5–D8 逐條打勾放行——與一場真的有人反對過的評審在報告上無從分辨。v4.2 給 D4 加上第二個軸：評審還必須宣告其模型層級要求（見下方術語說明）。CONVERGE 的提示改為目標＋約束形狀——D5–D8 是判斷必須滿足的判準，不是按順序執行的步驟；既有固定句型與逐準則分數指南保留為參考格式。發散側引用 core/model-selection.md（Standard 或更高）而不另立規則。不移除任何 v3 機制。

Terminology — "model level", never "tier" | 術語——「模型層級」，不用「tier」: inside this skill, tier is already taken twice — the BQS tiering table (which dimensions apply) and the Enhanced Tier. The model-capability axis defined in core/model-selection.md (called model tier there) is therefore always written model level（模型層級） in this skill. Model levels are that standard's vendor-neutral labels; this skill never names a concrete vendor model (see D4).

在本 skill 內，tier 已被佔用兩次——BQS 分級表（套用哪些維度）與 Enhanced Tier。core/model-selection.md 定義的模型能力軸（該標準稱 model tier）在本 skill 一律寫作模型層級（model level）。模型層級是該標準的廠商中立標籤；本 skill 絕不指名具體廠商模型（見 D4）。

Mode Selection | 使用前先選模式

Apply these objective triggers before starting. Default is full v3 — routing rules are shortcuts to skip phases, not barriers to add.

使用前套用以下客觀觸發條件。預設為完整 v3，路由規則是跳過階段的快捷鍵，而非額外障礙。

Rule of thumb: If you are unsure which row applies, use full v3. The cognitive overhead of deciding is higher than just running the full flow.

判斷原則： 不確定適用哪一行時，直接用完整 v3。判斷本身的認知成本高於直接跑完整流程。

Workflow | 工作流程

[Mode Selection] ─► PRE-FLIGHT ─► FRAME ─► DIVERGE ───────────► CONVERGE ──────────► OUTPUT
   客觀路由          防止錨定      定義問題   persona 集成+透鏡       多評審面板+硬角色反駁    輸出提案
   ▲ Layer 0 intent              ▲ Layer 1 (D1–D4, leading)        ▲ Layer 2 (D5–D8, Recommended Set)   ▲ Layer 3 override

BQS v1 — Quality Contract | BQS v1 — 品質契約

First principle | 第一原理: A brainstorm's output is a hypothesis, not an answer. Whether an idea is good is unknowable at ideation time, so quality can only be judged on leading signals (process + epistemic integrity); the standard's legitimacy is calibrated by lagging signals (later outcomes). Decisions use leading; calibration uses lagging; they are ordered in time, not a right/wrong choice. (This corrects any "use only leading, never lagging" absolutism.)

腦力激盪的產物是假說，不是答案。一個點子好不好在發想當下不可知，所以品質只能用 leading 訊號（過程＋認識論完整性）判；標準的正當性靠 lagging 訊號（事後結果）校準。決策用 leading、校準用 lagging，兩者是時間前後，不是對錯取捨。（此處修正任何「只用 leading 不用 lagging」的絕對說法。）

BQS is a four-layer × timeline contract. It is additive to v3 — every v3 flag and mechanism is preserved (see "Backward Compatibility").

BQS 是四層 × 時間軸契約。它是 v3 的疊加——所有 v3 旗標與機制都保留（見「向後相容」）。

Layer 0 — Intent (opening declaration) | 第 0 層 — 意圖（開場宣告）

At the start, declare the explore/exploit ratio and the bet type (incremental vs barbell long-tail). This layer modulates dimension weights: under an exploit-leaning session, a low D2 (divergence coverage) score is correct and is NOT penalised.

開場宣告本次 explore/exploit 配比 與賭注類型（漸進 vs barbell 長尾）。此層調節維度權重：偏 exploit 的工作階段，D2（發散覆蓋）低分是正確的、不扣分。

Layer 1 — Process (divergence-period leading, fully visible) | 第 1 層 — 過程（發散期 leading，全程可見）

Hard sequence gate | 硬序列閘: D5–D8 are forbidden from being revealed or scored during the divergence period. The CONVERGE critics MUST NOT be invoked before the last persona has produced its set.

D5–D8 禁止在發散期揭示或評分；CONVERGE 的 critic 不得在最後一個 persona 產完前被呼叫。

Layer 2 — Product (post-convergence leading, applied to the Recommended Set only) | 第 2 層 — 產物（收斂後 leading，僅施於推薦集）

Evaluative dimensions are confined to after convergence × the Recommended Set only (ideas with Agg. Score ≥ 3.5, uncapped in count) — never as a full gate on every divergence idea (that would retroactively pollute divergence, kill evidence-free future ideas, and create form-filling theatre). | 評判維度限縮在收斂後 × 僅推薦集（Agg. Score ≥ 3.5 的想法，不限筆數），絕不在發散期對全部想法當硬閘（否則回溯污染發散、扼殺無證據的未來想法、製造填表劇場）。

Layer 3 — Ungoverned zone (Judgment Override) | 第 3 層 — 不可治理區（Judgment Override）

A reserved space no oracle enters: a human's intuition to "keep / kill" needs only a one-line reason and overrides the aggregate score — it passes through no dimension. This admits the standard is incomplete and avoids "brainstorming to pass dimensions".

明文留一塊 oracle 不進入的地：人類直覺「保留/斃掉」附一句理由即可，凌駕聚合分數、不需過任何維度。承認標準不完備，避免「為過維度而 brainstorm」。

Structural rules | 結構規則

1. Meta stop rule | Meta 停止規則: stop when the layer's dimensions are all green and, after one more round, the Recommended Set membership is unchanged (set membership, not internal ranking). Hard cap: 2 rounds. This replaces the vague "decision didn't flip" criterion. | 該層維度全綠 且 再跑一輪後 推薦集成員不變（看集合成員、不看內部排序）→停。硬上限 2 輪。取代模糊的「不翻決策」。
2. Judge ≠ generator | 判官≠產生者: D2/D4/D5/D7 judgements need an independent viewpoint; single-context self-evaluation may only be [degraded], never pass. D4 additionally requires a declared model level — independence is one of its two axes (see D4). | D2/D4/D5/D7 判定須獨立視角；單 context 自評只能 [degraded]，不得 pass。D4 另需宣告模型層級——獨立性只是它兩軸之一（見 D4）。
3. Calibration loop (consumes lagging) | 校準回路（吃 lagging）: v3's three Session Self-Evaluation metrics are folded in as the lagging end — Adoption Rate = D6 lagging validation, Diversity = D2/D3 lagging observation, Cognitive Load = a cost constraint. Two parallel evaluation systems are forbidden — there is one loop, not a separate self-eval. | v3 的 Adoption Rate＝D6 滯後驗證、Diversity＝D2/D3 滯後觀測、Cognitive Load＝成本約束；三者收編為此回路 lagging 端，禁兩套平行評估。
4. BQS self-evolution (lightweight) | BQS 自我演化（輕量）: versioned + an evidence last-reviewed that flags when overdue; "periodically brainstorm BQS itself" is optional. | 版本化 + 證據清單 last-reviewed 逾期 flag；「定期對 BQS 自身 brainstorm」列可選。
5. Minimal sufficiency | 最小充分原則: apply only the fewest dimensions the tier requires + confining Layer 2 to the Recommended Set (Agg. Score ≥ 3.5) instead of every divergence idea — this is the gate for cognitive economy; no separate cost dimension is added. | 套用該 tier 所需的最少維度 + 把第 2 層限縮在推薦集（Agg. Score ≥ 3.5），而非全部發散想法——認知經濟性的守門，不另設成本維度。

Tiering (bound to v3 objective triggers, not self-assessment) | 分級（綁 v3 既有客觀觸發，非自評）

Tiers are selected by the objective Mode Selection triggers (word count, flags, existence of a spec) — not by the initiator's self-assessment. | 分級由客觀模式選擇觸發表（字數、旗標、是否有規格）決定，而非發起者自評。

Phase 0: PRE-FLIGHT | 防止 AI 錨定

Why this phase exists: Independent ideas written before the AI generates anything consistently produce more diverse results. In AI-assisted contexts this matters more, not less: research on design fixation shows that AI output — being fluent and high-fidelity — deepens fixation rather than relieving it (Wadinambiarachchi et al., CHI 2024).

本階段存在的原因： 在 AI 生成任何內容之前先寫下自己的想法，能持續產出更多樣的結果。在 AI 情境下這更重要：設計固著研究顯示，流暢、高擬真的 AI 輸出反而加深固著（Wadinambiarachchi 等，CHI 2024）。

Before the AI generates any content, the user completes three items:

在 AI 生成任何內容之前，使用者完成三件事：

After the user submits, the AI reads all three inputs and proceeds to FRAME. The AI's first DIVERGE output MUST explore directions the user did not mention and MUST NOT duplicate the user's three ideas.

使用者提交後，AI 讀取全部三項輸入再進入 FRAME。AI 的第一批 DIVERGE 輸出必須探索使用者未提及的方向，且不得重複使用者已提交的想法。

Anti-seed guardrail (new in v3): Do NOT accept or generate a "like X but for Y" framing as the seed (e.g. "Slack but for doctors"). Such analogical seeds lock the LLM into one solution space and measurably reduce idea variety. Capture the underlying problem, not a product analogy.

反種子 guardrail（v3 新增）： 不要用「像 X 但給 Y」的框架當種子（如「給醫生用的 Slack」）。這類產品類比種子會把 LLM 鎖進單一解空間、明顯降低想法多樣性。請捕捉底層問題，而非產品類比。

BQS Layer 0 — Intent declaration (new in v4): Before FRAME, state the explore/exploit ratio and bet type (incremental vs barbell long-tail) for this session. This modulates D2 weighting downstream — under an exploit-leaning session, low divergence coverage is correct and is not penalised. Default when unstated: explore-leaning.

BQS 第 0 層 — 意圖宣告（v4 新增）： 在 FRAME 之前，宣告本次的 explore/exploit 配比 與賭注類型（漸進 vs barbell 長尾）。此宣告下游調節 D2 權重——偏 exploit 時低發散覆蓋是正確的、不扣分。未宣告時預設：偏 explore。

Flag: --skip-preflight bypasses this phase with a one-line warning: ⚠ Skipping Pre-flight may cause AI anchoring

Phase 1: FRAME | 定義問題

Define the problem space clearly before generating ideas.

在產生想法之前，先清楚定義問題空間。

Phase 2: DIVERGE | 發散思考（v3：persona 集成 + 多樣性透鏡）

Core mechanism in v3: a persona ensemble, each persona reasoning via chain-of-thought in isolation, crossed with diversity lenses. Meincke, Mollick & Terwiesch (2024) found that chain-of-thought + personas produces the highest idea diversity of any prompting strategy — close to human groups. A raw idea count is a weak proxy; structurally forcing distinct viewpoints is the real lever.

v3 核心機制： persona 集成——每個 persona 以思維鏈在隔離狀態下推理——再乘上多樣性透鏡。Meincke、Mollick、Terwiesch（2024）發現「思維鏈 + persona」的想法多樣性高於所有受測提示策略，接近人類團體。光衝數量是弱代理；結構性逼出不同視角才是真正槓桿。

Model level (divergence side) | 模型層級（發散側）: a divergence brief gives only a goal and constraints — exactly the trigger of MS-004 / Criterion 2 in core/model-selection.md: run divergence at Standard or higher. Not "the highest level" — that would overstate both what MS-004 says and our evidence. Lagging registry: this choice is validated through the calibration loop's existing Diversity (D2/D3 lagging observation) column — no second registry.

發散提示只給目標與約束——正是 core/model-selection.md 的 MS-004／Criterion 2 的觸發條件：發散以 Standard 或更高層級執行。不是「最高層級」——那會同時超出 MS-004 所說的與我們的證據。lagging 登記：此選擇由校準回路既有的 Diversity（D2/D3 滯後觀測） 欄驗證，不另開第二套登記。

Step 2a — Persona ensemble | persona 集成

Generate ideas through a default ensemble of personas. Each persona reasons step by step (chain-of-thought) and produces 2–4 ideas from its own lens only. The user may add, drop, or rename personas via --personas.

透過預設 persona 組生成想法。每個 persona 逐步推理（思維鏈），只從自己的視角產出 2–4 個想法。使用者可用 --personas 增減或改名。

Branch isolation: In baseline mode, generate each persona's ideas without showing it the other personas' output — this prevents intra-session anchoring. Present all personas' ideas together only after every persona has produced its set. (In the Enhanced tier, run personas as parallel isolated agents — see "Enhanced Tier" below.)

分支隔離： baseline 模式下，生成每個 persona 的想法時不讓它看到其他 persona 的輸出，以防止 session 內錨定。等所有 persona 都產完才一起呈現。（Enhanced 層以平行隔離 agent 跑——見下方「Enhanced Tier」。）

Step 2b — Diversity lenses | 多樣性透鏡

Apply at least one lens across the ensemble to push past the "obvious answer zone." Connecting disparate concepts measurably increases originality (Mehrotra, Parab & Gulwani, 2024).

在 persona 組上至少套用一個透鏡，以突破「顯而易見答案區」。連結異域概念能可量測地提升原創性（Mehrotra、Parab、Gulwani，2024）。

Force --lens analogical|reversal|morphological to make a specific lens the primary one.

Step 2c — Continue nudge (auxiliary) | 繼續發散提示（輔助）

The "best ideas appear in the second half" pattern (Nijstad) is a human-group finding and is not confirmed for LLMs (which tend to plateau / exhaust). So a fixed idea-count gate is demoted to an auxiliary nudge: if fewer than ~8 distinct ideas exist across the ensemble, prompt "Continue — add a persona or lens you haven't used." Diversity (distinct lenses covered), not raw count, is the gate.

「好點子在後半」（Nijstad）是人類群體現象，未在 LLM 證實（LLM 多為高原/枯竭）。故固定數量門檻降為輔助提示：若全組少於約 8 個相異想法，提示「繼續——加一個還沒用過的 persona 或透鏡」。真正的門檻是多樣性（覆蓋了幾個不同視角），而非數量。

BQS Layer 1 applies here (D1–D4, leading, visible) | 此處生效 BQS 第 1 層（D1–D4，leading，全程可見）: during divergence, only D1 (frame purity), D2 (divergence coverage, weighted by Layer 0), D3 (cross-session diversity), D4 (evaluation de-bias) are scored and shown. Hard sequence gate: the evaluative dimensions D5–D8 must NOT be revealed or scored before the last persona has produced its set — the CONVERGE critics are not invoked until divergence is complete.

發散期只對 D1（框架純度）、D2（發散覆蓋，依第 0 層加權）、D3（跨會話多樣性）、D4（評估去偏） 評分與呈現。硬序列閘： 評判維度 D5–D8 在最後一個 persona 產完前不得揭示或評分——CONVERGE 的 critic 在發散完成前不被呼叫。

Classic techniques (still available) | 經典技法（仍保留）

Phase 3: CONVERGE | 收斂（v3：多評審面板 + 硬角色反駁）

Core mechanism in v3: a multi-critic panel replaces the single weighted scorer. A single LLM is a weak, biased evaluator (Li et al., 2025: LLMs are strong at generation/refinement but weak at evaluation — keep the human as final arbiter). Three independent critic lenses score each idea; their scores are aggregated.

v3 核心機制： 多評審面板取代單一加權評分者。單一 LLM 是弱且有偏的評估者（Li 等，2025：LLM 強於生成/精煉、弱於評估——人類保留最終裁決權）。三個獨立評審透鏡各自評分後聚合。

Critic brief shape — goal + constraints (v4.2) | 評審提示形狀——目標＋約束（v4.2）: when dispatching the panel, the Devil's Advocate, or the Steelman — above all into independent contexts (--enhanced) — the brief MUST be phrased as a goal plus the constraints the judgment must satisfy, never as an ordered procedure. D5–D8 and the specificity requirement below are criteria ("your judgment must satisfy these"), not steps ("do these in this order"). The fixed sentence patterns, the per-criterion 1–5 guides, and the critic weight formulas below stay available as reference formats — none of them is a mandatory step. Rationale: MS-009 in core/model-selection.md.

對面板、Devil's Advocate、Steelman 派工時——尤其派入獨立 context（--enhanced）——提示必須以目標＋其判斷必須滿足的約束表達，不得寫成按順序執行的步驟清單。D5–D8 與下方的具體性要求是判準（「你的判斷必須滿足這些」），不是步驟（「照這個順序做這些事」）。既有固定句型、逐準則 1–5 指南與評審權重公式保留為參考格式，皆非必經步驟。理由：core/model-selection.md 的 MS-009。

Step 3a: Multi-critic panel | 多評審面板

Run three independent critics, each scoring every idea 1–5 on its own lens. Aggregate (mean) to reduce single-critic bias. The weighted criteria below define the default lenses — a reference format, not a script the critic must walk through.

跑三個獨立評審，各自以自己的透鏡對每個想法打 1–5 分；取平均聚合以降低單評審偏誤。下方加權準則定義預設透鏡——是參考格式，不是評審必須逐步照走的腳本。

Per-criterion guide (1–5, reference format): Feasibility (5=trivial … 1=near-impossible); Impact (5=transformative … 1=negligible); Effort (5=hours … 1=quarters, inverted so lower effort scores higher); Alignment (5=core mission … 1=off-mission).

Optional — RICE / ICE (product features): for prioritising shippable features, score RICE = (Reach × Impact × Confidence) / Effort or the lighter ICE = Impact × Confidence × Ease. Let engineers — not the LLM — estimate Effort; the LLM lacks codebase knowledge. RICE favours incremental wins, so don't use it alone for strategic bets.

可選——RICE / ICE（產品功能）： 排序可出貨功能時用 RICE =（Reach × Impact × Confidence）/ Effort 或較輕的 ICE = Impact × Confidence × Ease。Effort 交由工程師估、不要讓 LLM 估（它無程式庫知識）。RICE 偏好漸進式勝利，別單獨用於策略性押注。

Step 3b: Hard-role Rebuttal Round | 硬角色反駁輪

A soft "please critique this" instruction yields mostly agreement (sycophancy). v3 assigns hard roles: for each idea in the Recommended Set (Agg. Score ≥ 3.5, uncapped), run a Devil's Advocate ("Your job is to argue this idea WILL fail") and a Steelman ("State the strongest charitable version of the counterargument"). Together they stress-test resilience rather than merely poke.

軟性「請批評一下」只會得到附和（諂媚）。v3 指派硬角色：對推薦集（Agg. Score ≥ 3.5，不限筆數）中的每一個想法各跑一個 Devil's Advocate（「你的任務是論證此案會失敗」）與一個 Steelman（「說出反方最強而善意的版本」）。兩者一起壓力測試韌性，而非只是戳。

Constraint: each counterargument must name a specific context and a specific reason — vague concerns ("this might be hard") are rejected. The sentence pattern "This idea will fail in [specific context] because [specific reason]" is a reference format, not a required phrasing.

約束：每個反對理由須指出具體情境與具體原因——模糊顧慮（「這可能有點難」）不接受。句型「在 [具體情境] 下，此想法會失敗，因為 [具體原因]」是參考格式，不是必用句式。

The user must respond to each before advancing:

使用者必須對每個給出回應才能繼續：

Flag: --no-rebuttal skips this step; report section marked "Rebuttal: skipped".

BQS D4 — judge ≠ generator, two axes | BQS D4 — 判官≠產生者，兩軸: D4 passes only when both axes hold. (1) Seat — independence: the critics/Devil's Advocate run in an independent context (the --enhanced isolated-agent host). (2) Mind — declared model level: the review declares its required model level, expressed with the vendor-neutral labels of core/model-selection.md; selection basis is that standard's criteria (link, don't restate). Naming a concrete vendor model is a violation — the level labels are the only legal form. Missing either axis → the panel is marked [degraded] and must NOT be marked pass: an independent context running a literal-following model satisfies the seat axis verbatim and is a fully compliant rubber stamp — recognizing "the same thing under a different name" is judgment work that box-ticking cannot do. On a baseline single context, the panel is same-context self-evaluation — [degraded], never pass, honest but not independent. Do not silently treat a baseline run as a D4 pass.

D4 唯有兩軸同時成立才可 pass。(1) 座位——獨立性：評審/Devil's Advocate 在獨立 context 執行（--enhanced 隔離 agent 宿主）。(2) 心智——宣告模型層級：評審宣告其模型層級要求，以 core/model-selection.md 的廠商中立標籤表達；選層依據為該標準的判準（引用，不複述）。指名具體廠商模型即違規——層級標籤是唯一合法的表達方式。缺任一軸→面板標 [degraded]、不得標 pass：獨立 context 跑字面遵循型模型會逐字滿足座位軸，成為一枚完全合規的橡皮圖章——認出「不同名字下的同一件事」是打勾做不到的判斷工作。baseline 單 context 下，面板是同 context 自評，標 [degraded]、不得 pass——誠實但非獨立。不得把 baseline 跑當成 D4 pass 靜默通過。

Step 3c: BQS Layer 2 — product gate on the Recommended Set | BQS 第 2 層——對推薦集的產物閘

After convergence, apply D5–D8 to the Recommended Set only (ideas with Agg. Score ≥ 3.5, uncapped — never to every divergence idea). If no idea reaches 3.5, keep the single highest-scoring idea and mark it [below threshold — shown for reference] so the report is never empty. D5–D8 are criteria the judgment must satisfy — in any order; they are not a procedure. For each idea in the Recommended Set, the judgment must satisfy:

收斂後，僅對推薦集（Agg. Score ≥ 3.5 的想法，不限筆數）套用 D5–D8（絕不對全部發散想法）。若無想法達 3.5，仍保留分數最高的 1 個，標記 [未達門檻——僅供參考]，避免報告空白。D5–D8 是判斷必須滿足的判準——不限順序，不是程序。對推薦集中每個想法，判斷必須滿足：

• D5 Grounding: split each claim into [current state / external fact] (needs a file:line or source, else fail) vs [future / hypothesis] (no grounding needed, mark [hypothesis]). An external-fact claim is a cross-tier floor — it must be grounded even at the creative tier. | 將每個主張分流為 [現狀/外部事實]（需 file:line 或來源，否則 fail）vs [未來/假說]（免接地、標 [假說]）。外部事實宣稱為跨級地板——creative 級也須接地。
• D6 Net benefit: each idea in the Recommended Set must answer "whose problem / do we actually have it / cost of not doing it"; at least one idea must be eliminated as "not worth doing"; attach a lagging registry field (which later signal validates this). | 推薦集中每個想法須答「解誰問題／我們真有嗎／不做的代價」；至少一個淘汰為「不值得做」；掛 lagging 登記欄（事後哪個訊號驗證）。
• D7 Falsifiability (two-state): mark [falsifiable now] or [need to do X first to define falsification]. The latter is routed to a next-step that feeds D8 — it does NOT count as fail. | 標 [現可陳述證偽] 或 [需先做 X 才能定義證偽]。後者轉 next-step 餵 D8——不算 fail。
• D8 Actionability: every surviving idea needs a next-step decision (including an explicit "defer / not now"); none → fail. | 每個存活想法需 next-step 裁決（含明確「暫不做」）；無 → fail。
  • A defer next-step, and every idea parked in Seeds (from killed ideas), are deferred items under deferred-item-exit. | defer 的 next-step，以及**Seeds（來自被淘汰的想法）**中每個被擱置的想法，皆為 deferred-item-exit 所指的延後項目。

Meta stop rule (BQS structural rule 1) | Meta 停止規則: stop iterating when the applied dimensions are all green and one more round leaves the Recommended Set membership unchanged (set membership, not internal ordering). Hard cap: 2 rounds.

當套用的維度全綠 且 再跑一輪後 推薦集成員不變（看集合成員、非內部排序）→停。硬上限 2 輪。

Phase 4: OUTPUT | 輸出提案

Produce a Brainstorm Report ready for /requirement or /sdd. Each surviving idea is marked ✓ Passed rebuttal with a one-line summary of the user's response, its originating persona/lens, and its aggregated critic score.

產生可直接對接 /requirement 或 /sdd 的腦力激盪報告。每個存活想法標記 ✓ Passed rebuttal、使用者回應摘要、來源 persona/透鏡、以及聚合評審分數。

Output Format | 輸出格式

# Brainstorm Report: [Topic]

## Problem Statement
[Refined problem + root cause from FRAME]

## HMW Questions
1. How might we ...?

## Ideas Generated
| # | Idea | Persona | Lens | Critic-Feas | Critic-Impact | Critic-Align | Agg. Score |
|---|------|---------|------|-------------|---------------|--------------|-----------|
| 1 | ...  | Skeptic | Reversal | 4.0 | 4.5 | 4.0 | 4.2 |

## Recommendations (BQS Layer 2 applied, Agg. Score ≥ 3.5 — uncapped)
1. **[Idea]** (Agg. X.X) ✓ Passed rebuttal — [Why] — Persona: [..] — [User rebuttal response]
   - D5 grounding: [current-state claims with file:line | future claims marked [hypothesis]]
   - D6 net benefit: [whose problem / do we have it / cost of not doing] — lagging signal: [..]
   - D7 falsifiability: [falsifiable now | need to do X first → next-step]
   - D8 next-step: [action | defer]
2. **[Idea]** (Agg. X.X) ✓ Passed rebuttal — ...
<!-- List every idea with Agg. Score ≥ 3.5, sorted descending — do not cap at 3. If none reach 3.5, list only the single highest-scoring idea marked "[below threshold — shown for reference]". -->

## Contested Zone (high critic-variance ideas)
[Ideas whose critic variance exceeded threshold — surfaced, NOT eliminated by mean ranking (barbell long-tail)]

## Seeds (from killed ideas)
[For each killed idea: what was wrong, what real problem it points to. Non-empty if ≥1 idea was killed.]

## Judgment Override (Layer 3, ungoverned)
[Human keep/kill decisions that override the aggregate score, each with a one-line reason. Optional.]

## Diversity Note
[How many distinct lenses/personas the surviving ideas span — flag if all from one cluster]

## Discarded Ideas (with reasons)
| Idea | Reason |

## Next Steps
- [ ] Proceed to `/requirement` with top idea
- [ ] Proceed to `/sdd` if requirements are clear

BQS output additions (v4): the Seeds section (rule: non-empty when ≥1 idea was killed), the Contested Zone (high-variance ideas not eliminated by mean ranking), and the Judgment Override channel (human decision overriding the aggregate score) are required by BQS Layer 2/3. The Recommendations block records the D5–D8 status per idea, for every idea in the Recommended Set (uncapped). Since v4.2, the report header also records the review's declared model level (vendor-neutral label) — without it, D4 is [degraded] and the declaration has nowhere to be verified.

BQS 輸出新增（v4）： Seeds 區（規則：被殺 ≥1 則非空）、爭議區（高方差想法不按 mean 淘汰）、Judgment Override 通道（人類裁決凌駕聚合分）為 BQS 第 2/3 層所要求。Recommendations 區塊為推薦集（不限筆數）中每個想法逐項記錄 D5–D8 狀態。自 v4.2 起，報告表頭並記錄評審宣告的模型層級（廠商中立標籤）——沒有它，D4 為 [degraded]，且宣告無處可驗。

Diversity-Collapse Guardrail | 多樣性崩塌防護

Using a single LLM for ideation reduces the diversity of ideas across users, even when each individual feels more creative (Anderson, Shah & Kreminski, 2024; corroborated by the widely-cited Doshi & Hauser, Science Advances 2024). Guard against it:

用單一 LLM 發想會降低跨使用者的想法多樣性，即使個人覺得更有創意（Anderson、Shah、Kreminski，2024；與廣為引用的 Doshi & Hauser，《Science Advances》2024 同向）。防範方式：

• Never seed with a competitor or product analogy ("like X but for Y"). | 絕不用競品/產品類比當種子。
• Vary the lens, not just the wording — reword ≠ diversify. | 變的是透鏡而非措辭——換句話不等於多樣化。
• If the surviving Recommended Set all came from one persona/lens, flag it and run one more lens before OUTPUT. | 若推薦集全來自同一 persona/透鏡，標示並在輸出前再跑一個透鏡。

Enhanced Tier — Parallel Personas | 強化層——平行 persona

Multi-agent ideation (independent agents conversing/contributing) outperforms a single agent on perceived quality and novelty (Quan et al., 2025, MultiColleagues). Where the host supports parallel subagents (e.g. Claude Code's Agent/Workflow tools), --enhanced runs each persona — and each critic — as a parallel, context-isolated agent, then merges and de-duplicates the results.

多 agent 發想（獨立 agent 互相對話/貢獻）在感知品質與新穎度上勝過單 agent（Quan 等，2025，MultiColleagues）。在支援平行子代理的宿主（如 Claude Code 的 Agent/Workflow 工具），--enhanced 會把每個 persona 與每個評審當作平行、context 隔離的 agent 跑，再合併去重。

Graceful degradation: This tier is optional. On hosts without subagents, --enhanced silently falls back to baseline (single-context simulated personas). The skill remains scope: universal.

優雅降級： 此層為可選。在無子代理的宿主，--enhanced 靜默退回 baseline（單 context 模擬 persona）。本 skill 維持 scope: universal。

Technique Quick Reference | 技法速覽

Calibration Loop — Lagging Signals (BQS) | 校準回路——lagging 訊號（BQS）

This is the lagging end of the single BQS quality loop — NOT a second, parallel evaluation system. The leading decision is made by BQS Layers 0–2 during the session; these three metrics calibrate the standard afterwards. Do not run a separate "self-evaluation" alongside BQS.

這是同一道 BQS 品質回路的 lagging 端——不是第二套平行評估系統。 工作階段中由 BQS 第 0–2 層做 leading 決策；這三個指標事後校準標準。禁止在 BQS 之外另跑一套「自評」。

After each session, record three metrics (1–5). Each maps onto a BQS dimension as its lagging validation:

每次工作階段結束後記錄三個指標（1–5 分），各自對應一個 BQS 維度作為其滯後驗證：

Collect 3 sessions before drawing conclusions. See guide.md for the full A/B experiment protocol.

收集 3 次工作階段資料再下結論。完整 A/B 實驗協議見 guide.md。

Flags | 旗標

Usage | 使用方式

• /brainstorm — Start interactive brainstorming session
• /brainstorm "user retention" — Brainstorm around a specific topic
• /brainstorm --enhanced "user retention" — Parallel persona ensemble (if host supports it)
• /brainstorm --personas "designer,economist,skeptic" "pricing" — Custom personas
• /brainstorm --lens analogical "onboarding" — Force the analogical lens
• /brainstorm --quick "reduce checkout friction" — Fast 3-idea mode
• /brainstorm --no-rebuttal "topic" — Skip rebuttal round

Next Steps Guidance | 下一步引導

After /brainstorm completes, the AI assistant should suggest:

腦力激盪完成。建議下一步 / Brainstorming complete. Suggested next steps:

• 執行 /requirement 將最佳構想轉為使用者故事 — Convert top idea to user stories
• 執行 /sdd 直接建立規格（若需求已明確）⭐ Recommended / 推薦 — Create spec directly
• 針對特定構想進行更深入探索 — Explore a specific idea further

Reference | 參考

• Detailed guide: guide.md
• Model level basis: core/model-selection.md — the D4 mind axis and the divergence-side level defer to that standard's criteria. Cite it, don't restate it — a summary rots when its source is corrected; a pointer does not. | 模型層級依據：D4 心智軸與發散側層級皆以該標準的判準為準。引用、不複述——摘要會隨來源被更正而腐壞，指標不會。

AI Agent Behavior | AI 代理行為

完整的 AI 行為定義請參閱對應的命令文件：/brainstorm

For complete AI agent behavior definition, see the corresponding command file: /brainstorm