universal-dev-standards 6.8.0 → 6.9.0

package/bundled/locales/zh-TW/skills/commands/brainstorm.md

@@ -1,9 +1,10 @@
 ---
 source: ../../../../skills/commands/brainstorm.md
-source_version: 4.0.0
-source_hash: 3212203924de
-translation_version: 4.0.0
-last_synced: 2026-06-22
+source_version: 4.2.0
+source_hash: 319f8ad4f02f
+translation_version: 4.2.0
+last_synced: 2026-08-21
+status: current
 ---
 
 ---
@@ -36,7 +37,7 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 | **PRE-FLIGHT** | 防止 AI 錨定 | 使用者先寫 3 個想法；無類比種子；**BQS 第 0 層意圖（explore/exploit）** |
 | **FRAME** | 清楚定義問題 | 5 Whys、HMW、Stakeholder Map；**D1 框架純度** |
 | **DIVERGE** | 逼出視角多樣性 | persona 集成 + 多樣性透鏡；**BQS 第 1 層 D1–D4，硬閘隱藏 D5–D8** |
-| **CONVERGE** | 降偏誤選擇 | 多評審面板 + Devil's Advocate／Steelman；**BQS 第 2 層 D5–D8 施於 Top 3，D4 判官≠產生者** |
+| **CONVERGE** | 降偏誤選擇 | 多評審面板 + Devil's Advocate／Steelman；**BQS 第 2 層 D5–D8 施於推薦集（Agg. ≥ 3.5），D4 判官≠產生者（兩軸：獨立 context＋宣告模型層級）** |
 | **OUTPUT** | 可執行的報告 | Brainstorm Report + Seeds + 爭議區 + **第 3 層 Judgment Override** |
 
 > **BQS v1（XSPEC-296）：** v4 在 v3 上疊加四層 × 時間軸品質契約。完整 oracle 與結構規則見 skill：[Brainstorm Assistant Skill](../brainstorm-assistant/SKILL.md#bqs-v1--品質契約)。
@@ -70,7 +71,7 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 ## Problem Statement
 [Refined problem from FRAME phase]
 
-## Top 3 Recommendations
+## Recommendations (Agg. Score ≥ 3.5 — uncapped)
 1. **[Idea]** — Agg. X.X ✓ Passed rebuttal — Persona: [..] — [Why recommended]
 
 ## Diversity Note
@@ -102,7 +103,7 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 | `/brainstorm "topic"` | 以指定主題啟動 PRE-FLIGHT，再進入 FRAME |
 | `/brainstorm --personas "a,b,c"` | 以自訂 persona 組進入 DIVERGE |
 | `/brainstorm --lens <name>` | 以指定多樣性透鏡為主進入 DIVERGE |
-| `/brainstorm --enhanced` | 若宿主支援子代理則平行跑 persona／評審，否則靜默退回 baseline；此模式下 BQS D4 方可 pass |
+| `/brainstorm --enhanced` | 若宿主支援子代理則平行跑 persona／評審，否則靜默退回 baseline；此模式滿足 BQS D4 的座位軸（獨立 context），pass 另需宣告模型層級（心智軸，見 skill D4） |
 | `/brainstorm --intent exploit` | 以 exploit 意圖啟動（BQS 第 0 層），D2 低覆蓋不扣分 |
 
 ### 互動腳本
@@ -127,12 +128,13 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 3. 至少套用一個多樣性透鏡（類比／假設反轉／形態矩陣）
 4. 以多樣性（覆蓋的 persona／透鏡數，D2/D3）而非數量為繼續門檻
 5. **硬序列閘**：D5–D8 在最後一個 persona 產完前**不得揭示或評分**；CONVERGE critic 不得提前呼叫
+6. **模型層級（發散側）**：以 **Standard 或更高**層級執行——依 core/model-selection.md 的 MS-004／Criterion 2（引用，不複述；不是「最高層級」）
 
-#### CONVERGE Phase（BQS 第 2 層 — D5–D8 僅施於 Top 3）
-1. 以 3 個獨立評審（工程可行性／使用者影響／策略一致）各自評分後聚合
-2. 對前 3 名跑硬角色 Devil's Advocate（論證會失敗）+ Steelman；使用者以 (a)修改／(b)反駁／(c)移除 回應。**D4 判官≠產生者**：評審須在獨立 context（`--enhanced`）方可 pass，否則標 `[degraded]` 不得 pass
-3. **僅對 Top 3** 套用 D5 接地（外部事實需 file:line／來源、跨級地板；假說免接地）、D6 淨值（解誰問題／真有嗎／不做代價＋掛 lagging 欄）、D7 二態證偽（「需先做 X」轉 next-step、不算 fail）、D8 next-step 裁決
-4. **Meta 停止規則**：維度全綠且再跑一輪 Top 3 集合成員不變→停；硬上限 2 輪
+#### CONVERGE Phase（BQS 第 2 層 — D5–D8 僅施於推薦集）
+1. 以 3 個獨立評審（工程可行性／使用者影響／策略一致）各自評分後聚合；**評審提示以「目標＋約束」表達**（判準，不是步驟清單——理由見 skill 引用的 core/model-selection.md MS-009），加權公式與逐準則指南為參考格式
+2. 對**推薦集**（Agg. Score ≥ 3.5，不限筆數）跑硬角色 Devil's Advocate（論證會失敗）+ Steelman；使用者以 (a)修改／(b)反駁／(c)移除 回應。**D4 判官≠產生者，兩軸**：(1) 評審在獨立 context（`--enhanced`）執行，且 (2) 宣告模型層級（core/model-selection.md 的廠商中立標籤；指名具體廠商模型即違規）——缺任一軸標 `[degraded]` 不得 pass
+3. **僅對推薦集**套用 D5 接地（外部事實需 file:line／來源、跨級地板；假說免接地）、D6 淨值（解誰問題／真有嗎／不做代價＋掛 lagging 欄）、D7 二態證偽（「需先做 X」轉 next-step、不算 fail）、D8 next-step 裁決——四者為判準、不限順序
+4. **Meta 停止規則**：維度全綠且再跑一輪推薦集集合成員不變→停；硬上限 2 輪
 
 🛑 **STOP**: 展示 Brainstorm Report（含 Seeds／爭議區／Judgment Override）後等待使用者決定下一步
 
@@ -153,13 +155,15 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 | 使用「像 X 但給 Y」種子 | 改寫為底層問題再繼續（反種子 guardrail） |
 | 指定的 persona／透鏡不存在 | 列出可用選項供選擇 |
 | `--enhanced` 但宿主不支援子代理 | 靜默退回 baseline，照常進行 |
-| 前 3 名全來自同一 persona／透鏡 | 標示並在輸出前再跑一個透鏡 |
+| 推薦集全來自同一 persona／透鏡 | 標示並在輸出前再跑一個透鏡 |
 | baseline 單 context 跑評審（無獨立 context） | BQS D4 標 `[degraded]`，不得標 pass |
+| 評審未宣告模型層級（或以具體廠商模型名宣告） | BQS D4 標 `[degraded]` 不得 pass；模型名改為 core/model-selection.md 的層級標籤 |
 | 外部事實宣稱無 file:line／來源 | D5 接地 fail，要求補來源或改標 `[假說]` |
 
 ## 參考
 
 * [Brainstorm Assistant Skill](../brainstorm-assistant/SKILL.md)
 * [Brainstorm Assistant Guide](../brainstorm-assistant/guide.md)
+* [Model Selection](../../core/model-selection.md) — BQS D4 與發散側的模型層級標籤（引用，不複述）
 * [Requirement Assistant](../requirement-assistant/SKILL.md)
 * [Spec-Driven Development](../spec-driven-dev/SKILL.md)

package/bundled/locales/zh-TW/skills/commands/config.md

@@ -171,7 +171,6 @@ Current Skills status:
 | 模式 | 說明 |
 |------|------|
 | `standard` | 摘要 + 任務映射，AI 知道何時讀取哪個標準（推薦）|
-| `full` | 將所有標準完整內嵌於整合檔案 |
 | `minimal` | 僅內嵌核心規則 |
 
 ## 設定變更的影響

package/bundled/locales/zh-TW/skills/commands/init.md

@@ -42,7 +42,7 @@ argument-hint: "[--level N | --yes]"
 | **OpenCode** | `AGENTS.md` | ✅ | ✅ |
 | **Gemini CLI** | `GEMINI.md` | ✅ | ✅ |
 | **Codex** | `AGENTS.md` | ✅ | ❌ |
-| **Antigravity** | `INSTRUCTIONS.md` | ✅ | ❌ |
+| **Antigravity** | `.agents/AGENTS.md` | ✅ | ❌ |
 
 預選環境中偵測到的工具。注意：Codex 和 OpenCode 共用 `AGENTS.md`。
 
@@ -211,7 +211,6 @@ Options:
 | 選項 | 說明 |
 |------|------|
 | **Standard（建議）** | 摘要 + 任務對應，AI 知道何時讀取哪項標準 |
-| **Full Embed** | 嵌入所有規則，AI 可立即使用但檔案較大 |
 | **Minimal** | 僅檔案參考，適合搭配 Skills 使用 |
 
 ### 步驟 13：確認並執行

package/bundled/locales/zh-TW/skills/commit-standards/SKILL.md

@@ -126,6 +126,8 @@ Closes #123
 
 ## 參考
 
+- [Conventional Commits 指南](./conventional-commits.md) — 需要確切的 type、scope 與格式規則時讀它。
+- [Commit Message 語言選項](./language-options.md) — 要決定 commit message 用哪一種語言時讀它。
 - 詳細指南：[guide.md](./guide.md)
 - 核心規範：[commit-message-guide.md](../../../../core/commit-message-guide.md)
 

package/bundled/locales/zh-TW/skills/contract-test-assistant/SKILL.md

@@ -2,7 +2,7 @@
 name: contract-test-assistant
 source: ../../../../skills/contract-test-assistant/SKILL.md
 source_version: 1.0.0
-source_hash: f67623eb913a
+source_hash: 35b654f0b277
 translation_version: 1.1.0
 last_synced: 2026-08-17
 status: current
@@ -103,6 +103,7 @@ ASSESS ──► CHOOSE ──► DEFINE ──► IMPLEMENT ──► VERIFY
 
 ## 參考
 
+- 詳細指南：[guide.md](./guide.md) — 合約測試的策略與實作。要設定 Pact／OpenAPI，或把合約測試接進 CI 時讀它。
 - 詳細指南：[contract-testing.md](../../options/testing/contract-testing.md)
 - 相關：[api-design-assistant](../api-design-assistant/SKILL.md)
 

package/bundled/locales/zh-TW/skills/dev-methodology/SKILL.md

@@ -102,6 +102,8 @@ argument-hint: "[動作] [參數]"
 
 ## 參考
 
+- [建立自訂 Methodology 指南](./create-methodology.md) — 要自建方法論而不是用現成的那幾套時讀它。
+- [Methodology Runtime 指南](./runtime.md) — 需要知道方法論在執行當下如何運作時讀它。
 - 詳細指南：[guide.md](./guide.md)
 
 ## AI 代理行為

package/bundled/locales/zh-TW/skills/dev-workflow-guide/SKILL.md

@@ -4,7 +4,7 @@ source: ../../../../skills/dev-workflow-guide/SKILL.md
 source_version: 1.0.0
 translation_version: 1.1.0
 last_synced: 2026-08-17
-source_hash: 5bf83f0db544
+source_hash: f90f1fe1f6f2
 status: current
 description: |
   [UDS] 把目前的軟體開發階段對應到正確的 UDS 指令與 Skill。

package/bundled/locales/zh-TW/skills/knowledge-graph/guide.md

@@ -2,8 +2,8 @@
 source: ../../../../skills/knowledge-graph/guide.md
 source_version: 1.0.0
 translation_version: 1.0.0
-last_synced: 2026-06-02
-source_hash: 9c86b5a2a71d
+last_synced: 2026-07-30
+source_hash: 7921f66da47d
 status: current
 ---
 

package/bundled/locales/zh-TW/skills/migration-assistant/SKILL.md

@@ -2,7 +2,7 @@
 name: migration-assistant
 source: ../../../../skills/migration-assistant/SKILL.md
 source_version: 1.0.0
-source_hash: 67b6f33f825e
+source_hash: 317e388b2b82
 translation_version: 1.1.0
 last_synced: 2026-08-17
 status: current

package/bundled/locales/zh-TW/skills/observability-assistant/SKILL.md

@@ -47,6 +47,7 @@ description: |
 
 ## 參考
 
+- 詳細指南：[guide.md](./guide.md) — 日誌、指標、追蹤三大支柱的完整說明。要建立監控、挑選指標型別、導入追蹤或評估成熟度時讀它。
 - 核心規範：[observability-standards.md](../../../../core/observability-standards.md)
 - 核心規範：[alerting-standards.md](../../../../core/alerting-standards.md)
 - 核心規範：[slo-standards.md](../../../../core/slo-standards.md)

package/bundled/locales/zh-TW/skills/project-discovery/SKILL.md

@@ -123,4 +123,5 @@ docs/risks/
 
 ## 參考
 
+- 逐步流程：[workflow.md](./workflow.md) — 探索的各個步驟與順序（程式碼健康度、架構理解、文件盤點、審查快照、相依與安全檢查）。真的要跑一次探索時讀它。
 - 詳細指南：[guide.md](./guide.md)

package/bundled/locales/zh-TW/skills/project-structure-guide/SKILL.md

@@ -151,6 +151,7 @@ project/
 
 ## 相關標準
 
+- [各語言的專案結構](./language-patterns.md) — 逐語言的實際目錄配置（Node.js/TypeScript、Python、Go 等）。要為某個特定語言擺放專案結構時讀它。
 - [核心：專案結構](../../core/project-structure.md)
 - [核心：文件結構](../../core/documentation-structure.md)
 - [指南：檔案歸檔決策指南](../../core/guides/file-placement-guide.md)

package/bundled/locales/zh-TW/skills/release-standards/SKILL.md

@@ -71,5 +71,8 @@ description: |
 
 ## 參考
 
+- [變更日誌格式指南](./changelog-format.md) — 要撰寫或審查變更日誌時讀它。
+- [發布流程指南](./release-workflow.md) — 真的要發一版、需要逐步流程時讀它。
+- [語意化版本指南](./semantic-versioning.md) — 要判斷一個變更算 major、minor 還是 patch 時讀它。
 - 詳細指南：[guide.md](./guide.md)
 - 核心規範：[versioning.md](../../../../core/versioning.md)

package/bundled/locales/zh-TW/skills/requirement-assistant/SKILL.md

@@ -74,5 +74,7 @@ So that [benefit].
 
 ## 參考
 
+- [需求完整性檢查清單](./requirement-checklist.md) — 要在需求被接受之前檢查它完不完整時讀它。
+- [需求撰寫指南](./requirement-writing.md) — 要從零開始寫一份需求時讀它。
 - 詳細指南：[guide.md](./guide.md)
 - 核心規範：[requirement-engineering.md](../../../../core/requirement-engineering.md)

package/bundled/locales/zh-TW/skills/reverse-engineer/SKILL.md

@@ -126,6 +126,9 @@ description: |
 
 ## 參考
 
+- [BDD 提取工作流程指南](./bdd-extraction.md) — 要從既有規格裡萃取 BDD 場景時讀它。
+- [TDD 分析工作流程指南](./tdd-analysis.md) — 要對照場景分析測試涵蓋率、找出缺口時讀它。
+- 逐步流程：[workflow.md](./workflow.md) — 逆向工程的各階段與順序（程式碼掃描、資料模型、設定、測試分析、涵蓋率）。真的要跑一次逆向時讀它。
 - 詳細指南：[guide.md](./guide.md)
 - 核心規範：[reverse-engineering-standards.md](../../../../core/reverse-engineering-standards.md)
 

package/bundled/locales/zh-TW/skills/runbook-assistant/SKILL.md

@@ -46,6 +46,7 @@ description: |
 
 ## 參考
 
+- 詳細指南：[guide.md](./guide.md) — 運維手冊的撰寫、組織與驗證。要寫手冊、審查運維程序、規劃演練或評估涵蓋範圍時讀它。
 - 核心規範：[runbook-standards.md](../../../../core/runbook-standards.md)
 - 相關：[alerting-standards.md](../../../../core/alerting-standards.md)
 - 相關：[postmortem-standards.md](../../../../core/postmortem-standards.md)

package/bundled/locales/zh-TW/skills/slo-assistant/SKILL.md

@@ -46,6 +46,7 @@ description: |
 
 ## 參考
 
+- 詳細指南：[guide.md](./guide.md) — SLI／SLO／錯誤預算的定義與管理。要定義 SLO、挑選 SLI、計算錯誤預算或設定可靠度目標時讀它。
 - 核心規範：[slo-standards.md](../../../../core/slo-standards.md)
 - 相關：[observability-standards.md](../../../../core/observability-standards.md)
 - 相關：[alerting-standards.md](../../../../core/alerting-standards.md)

package/bundled/locales/zh-TW/skills/tdd-assistant/SKILL.md

@@ -75,5 +75,7 @@ description: |
 
 ## 參考
 
+- [TDD 語言範例](./language-examples.md) — 需要某個特定語言的實際範例時讀它。
+- [TDD 工作流程指南](./tdd-workflow.md) — 要完整跑一輪紅-綠-重構時讀它。
 - 詳細指南：[guide.md](./guide.md)
 - 核心規範：[test-driven-development.md](../../../../core/test-driven-development.md)

package/bundled/skills/atdd-assistant/SKILL.md

@@ -97,6 +97,8 @@ After `/atdd` completes, the AI assistant should suggest:
 
 ## Reference | 參考
 
+- [Acceptance Criteria Writing Guide](./acceptance-criteria-guide.md) — Read it when writing or reviewing acceptance criteria.
+- [ATDD Workflow Guide](./atdd-workflow.md) — Read it when running the ATDD cycle end to end.
 - Detailed guide: [guide.md](./guide.md)
 - Core standard: [acceptance-test-driven-development.md](../../core/acceptance-test-driven-development.md)
 

package/bundled/skills/bdd-assistant/SKILL.md

@@ -92,6 +92,8 @@ After `/bdd` completes, the AI assistant should suggest:
 
 ## Reference | 參考
 
+- [BDD Workflow Guide](./bdd-workflow.md) — Read it when running discovery, formulation and automation as a cycle.
+- [Gherkin Quick Reference Guide](./gherkin-guide.md) — Read it when writing Given/When/Then and you need the exact syntax.
 - Detailed guide: [guide.md](./guide.md)
 - Core standard: [behavior-driven-development.md](../../core/behavior-driven-development.md)
 

package/bundled/skills/brainstorm-assistant/SKILL.md

@@ -29,6 +29,14 @@ v3 把發散從「單一 AI 衝數量」改為**persona 集成**（每個角色
 
 **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](../../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](../../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](../../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](../../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.
@@ -81,7 +89,7 @@ At the start, declare the **explore/exploit ratio** and the bet type (incrementa
 | **D1 Frame purity** | Problem embeds a specific solution or a "like X but for Y" framing → fail; must run 5-Whys to the root cause | 問題內嵌特定方案或「像 X 給 Y」→fail；須 5-Whys 到根因 |
 | **D2 Divergence coverage** | Surviving ideas span < 3 independent personas/lenses → fail (modulated by Layer 0 weight); only a round with zero new ideas counts as saturation | 存活想法跨 <3 獨立 persona/透鏡→fail（受第 0 層權重調節）；連一輪零新增才算飽和 |
 | **D3 Cross-session diversity** | Seed is a competitor analogy → fail; Top all from a single lens → fail | 種子是競品類比→fail；Top 全來自單一 lens→fail |
-| **D4 Evaluation de-bias** | Requires ≥3 critics + hard-role Devil's Advocate + Steelman; **pass requires an independent context**, otherwise mark `[degraded]` — must NOT be marked pass | 須 ≥3 評審 + 硬角色 Devil's Advocate + Steelman；**pass 須在獨立 context 執行**，否則標 `[degraded]` 不得標 pass |
+| **D4 Evaluation de-bias** | Requires ≥3 critics + hard-role Devil's Advocate + Steelman; **pass requires both axes**: (1) an independent context, **and** (2) a declared **model level** requirement using the vendor-neutral labels of [model-selection](../../core/model-selection.md) — naming a concrete vendor model is a violation. Missing either axis → mark `[degraded]` — must NOT be marked pass | 須 ≥3 評審 + 硬角色 Devil's Advocate + Steelman；**pass 須同時滿足兩軸**：(1) 獨立 context，**且** (2) 以 [model-selection](../../core/model-selection.md) 的廠商中立標籤宣告**模型層級**要求——指名具體廠商模型即違規。缺任一軸→標 `[degraded]` 不得標 pass |
 
 > **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.
 >
@@ -109,7 +117,7 @@ A reserved space no oracle enters: a human's intuition to "keep / kill" needs on
 ### 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. | D2/D4/D5/D7 判定須獨立視角；單 context 自評只能 `[degraded]`，不得 pass。
+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。
@@ -180,6 +188,10 @@ Define the problem space clearly before generating ideas.
 >
 > **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](../../core/model-selection.md#rules) / [Criterion 2](../../core/model-selection.md#criterion-2--specification-definiteness規格明確度) 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](../../core/model-selection.md#rules)／[Criterion 2](../../core/model-selection.md#criterion-2--specification-definiteness規格明確度) 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`.
@@ -238,11 +250,15 @@ The "best ideas appear in the second half" pattern (Nijstad) is a **human-group*
 >
 > **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](../../core/model-selection.md#rules) 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](../../core/model-selection.md#rules)。
+
 #### 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. Each critic uses the weighted formula below.
+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。每位評審皆套用下方加權公式。
+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。
 
 | Critic lens | Weighted criteria it owns | 評審透鏡 |
 |-------------|---------------------------|----------|
@@ -250,7 +266,7 @@ Run **three independent critics**, each scoring every idea 1–5 on its own lens
 | **User impact** | Impact 70% · Alignment 30% | 使用者影響 |
 | **Strategic alignment** | Alignment 60% · Impact 40% | 策略一致性 |
 
-Per-criterion guide (1–5): 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).
+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.
 >
@@ -262,9 +278,9 @@ A soft "please critique this" instruction yields mostly agreement (sycophancy).
 
 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。
 
-Each counterargument must take the form: "This idea will fail in [specific context] because [specific reason]." Vague concerns ("this might be hard") are rejected.
+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.
 
-每個反對理由須為：「在 [具體情境] 下，此想法會失敗，因為 [具體原因]。」模糊顧慮（「這可能有點難」）不接受。
+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:
 
@@ -278,20 +294,21 @@ The user **must** respond to each before advancing:
 
 **Flag:** `--no-rebuttal` skips this step; report section marked "Rebuttal: skipped".
 
-> **BQS D4 — judge ≠ generator | BQS D4 — 判官≠產生者:** D4 passes **only when the critics/Devil's Advocate run in an independent context** (the `--enhanced` isolated-agent host). On a baseline single context, the panel runs as same-context self-evaluation, which is marked `[degraded]` and **must NOT be marked pass** — it is honest but not independent. Do not silently treat a baseline run as a D4 pass.
+> **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](../../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**評審/Devil's Advocate 在獨立 context 執行**（`--enhanced` 隔離 agent 宿主）才可 pass。baseline 單 context 下，面板是同 context 自評，標 `[degraded]`、**不得標 pass**——誠實但非獨立。不得把 baseline 跑當成 D4 pass 靜默通過。
+> 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](../../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. For each idea in the Recommended Set:
+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。對推薦集中每個想法：
+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](../../core/deferred-item-exit.md). | `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](../../core/deferred-item-exit.md) 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.**
 >
@@ -350,9 +367,9 @@ Produce a Brainstorm Report ready for `/requirement` or `/sdd`. Each surviving i
 - [ ] 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).
+> **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。
+> **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
 
@@ -441,6 +458,7 @@ After `/brainstorm` completes, the AI assistant should suggest:
 ## Reference | 參考
 
 - Detailed guide: [guide.md](./guide.md)
+- Model level basis: [core/model-selection.md](../../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
 

package/bundled/skills/brainstorm-assistant/guide.md

@@ -10,8 +10,8 @@ description: |
 
 > **Language**: English | [繁體中文](../../locales/zh-TW/skills/brainstorm-assistant/guide.md)
 
-**Version**: 4.1.0
-**Last Updated**: 2026-07-09
+**Version**: 4.2.0
+**Last Updated**: 2026-08-21
 **Applicability**: All software projects
 **Scope**: universal
 **Type**: Utility Skill (no core standard)
@@ -286,7 +286,7 @@ A single LLM is a weak, biased evaluator (Li et al., 2025: LLMs are strong at ge
 | **User impact** | Impact 70% · Alignment 30% |
 | **Strategic alignment** | Alignment 60% · Impact 40% |
 
-**Per-criterion guide (1–5):**
+**Per-criterion guide (1–5, reference format — not a script the critic must walk through):**
 
 | Criterion | 5 | 3 | 1 |
 |-----------|---|---|---|
@@ -312,7 +312,7 @@ A soft "please critique this" yields mostly agreement — LLMs are sycophantic u
 - **Devil's Advocate**: "Your job is to argue this idea WILL fail. Produce 2 specific failure conditions."
 - **Steelman**: "State the strongest, most charitable version of the counterargument — the one a thoughtful opponent would actually make."
 
-Each counterargument must take the form: **"This idea will fail in [specific context] because [specific reason]."**
+Constraint: each counterargument must name a **specific context** and a **specific reason**. The sentence pattern **"This idea will fail in [specific context] because [specific reason]"** is a reference format, not a required phrasing (v4.2 — the critic brief is goal + constraints, see SKILL.md "Critic brief shape").
 
 **NOT acceptable** (too vague): "This might be difficult." / "There could be edge cases."
 
@@ -348,6 +348,7 @@ Each idea that remains receives a badge: `✓ Passed rebuttal — [one-line summ
 **Personas Used**: [domain expert, skeptic, analogist, ...]
 **Lenses Used**: [analogical, reversal, ...]
 **Pre-flight**: [Completed / Skipped]   **Rebuttal**: [Completed / Skipped]   **Tier**: [Baseline / Enhanced]
+**Critic model level (D4 mind axis)**: [vendor-neutral level label per core/model-selection.md — naming a vendor model is a violation; not declared → D4 [degraded]]
 
 ## Problem Statement
 [Refined problem + root cause from 5 Whys]
@@ -561,9 +562,9 @@ BQS v1 is an **additive quality contract**, not a rewrite. Everything from v3 is
 
 - **All flags**: `--personas`, `--lens`, `--enhanced`, `--skip-preflight`, `--no-rebuttal`, `--quick`, `--technique` behave exactly as in v3. v4 adds one optional flag, `--intent`, for the Layer 0 declaration.
 - **All mechanisms**: PRE-FLIGHT anti-anchoring, FRAME 5-Whys/HMW, the persona ensemble, diversity lenses, the multi-critic panel, the hard-role rebuttal (Devil's Advocate + Steelman), the Diversity-Collapse Guardrail, and the Enhanced Tier are unchanged.
-- **The `--enhanced` isolated-agent host** is what lets BQS **D4 pass** (judge ≠ generator); on a baseline single context the panel is marked `[degraded]`, never silently passed.
+- **The `--enhanced` isolated-agent host** is what satisfies BQS **D4's seat axis** (judge ≠ generator); since v4.2 (XSPEC-388) a D4 pass additionally requires the review to declare its **model level** (mind axis — vendor-neutral labels per core/model-selection.md). On a baseline single context the panel is marked `[degraded]`, never silently passed.
 
-BQS v1 是**疊加的品質契約**，不是打掉重練。v3 的一切都保留：所有旗標（`--personas`/`--lens`/`--enhanced`/`--skip-preflight`/`--no-rebuttal`/`--quick`/`--technique`）行為不變，v4 僅新增一個可選的 `--intent` 旗標供第 0 層宣告；所有機制（PRE-FLIGHT 防錨定、FRAME 5-Whys/HMW、persona 集成、多樣性透鏡、多評審面板、硬角色反駁、多樣性崩塌防護、Enhanced 層）不變；`--enhanced` 的隔離 agent 宿主正是讓 BQS **D4 pass**（判官≠產生者）的條件，baseline 單 context 標 `[degraded]`、不靜默通過。
+BQS v1 是**疊加的品質契約**，不是打掉重練。v3 的一切都保留：所有旗標（`--personas`/`--lens`/`--enhanced`/`--skip-preflight`/`--no-rebuttal`/`--quick`/`--technique`）行為不變，v4 僅新增一個可選的 `--intent` 旗標供第 0 層宣告；所有機制（PRE-FLIGHT 防錨定、FRAME 5-Whys/HMW、persona 集成、多樣性透鏡、多評審面板、硬角色反駁、多樣性崩塌防護、Enhanced 層）不變；`--enhanced` 的隔離 agent 宿主滿足的是 BQS **D4 的座位軸**（判官≠產生者），自 v4.2（XSPEC-388）起 D4 pass 另需評審宣告**模型層級**（心智軸——依 core/model-selection.md 的廠商中立標籤），baseline 單 context 標 `[degraded]`、不靜默通過。
 
 ---
 
@@ -598,6 +599,7 @@ BQS v1 是**疊加的品質契約**，不是打掉重練。v3 的一切都保留
 | [Requirement Engineering](../../core/requirement-engineering.md) | Brainstorm output feeds requirement writing |
 | [Spec-Driven Development](../../core/spec-driven-development.md) | Brainstorm output feeds SDD proposals |
 | [Anti-Hallucination](../../core/anti-hallucination.md) | Critic feasibility claims must be evidence-based, not asserted |
+| [Model Selection](../../core/model-selection.md) | The D4 mind axis (declared model level) and the divergence-side level (Standard or higher) defer to this standard's criteria — cited, never restated |
 
 ---
 
@@ -605,6 +607,7 @@ BQS v1 是**疊加的品質契約**，不是打掉重練。v3 的一切都保留
 
 | Version | Date | Changes |
 |---------|------|---------|
+| 4.2.0 | 2026-08-21 | XSPEC-388: D4's pass condition gains a second axis — the seat (independent context, unchanged) plus the mind (the review must declare its required **model level**, expressed only with the vendor-neutral labels of core/model-selection.md; naming a concrete vendor model is a violation; missing declaration → `[degraded]`, never pass). CONVERGE critic briefs re-shaped as goal + constraints per MS-009: D5–D8 presented as criteria the judgment must satisfy, never as an ordered procedure; the fixed rebuttal sentence pattern, per-criterion 1–5 guides, and critic weight formulas are kept as reference formats, not mandatory steps. Divergence side cites MS-004/Criterion 2 (Standard or higher — deliberately not "highest"), validated through the calibration loop's existing Diversity (D2/D3) lagging column. Terminology: the model-capability axis is written "model level" throughout, because "tier" is already taken by BQS tiering and the Enhanced Tier. All v3 flags and mechanisms preserved. |
 | 4.1.0 | 2026-07-09 | XSPEC-325: Replaced the fixed Top-3 recommendation cap with a score-threshold **Recommended Set** (Agg. Score ≥ 3.5, uncapped in count) throughout BQS Layer 2 — the product gate (D5–D8), the hard-role rebuttal round (Devil's Advocate + Steelman), the Meta stop rule (set-membership stability), the Diversity-Collapse Guardrail, and the OUTPUT Recommendations block all now key off the Recommended Set instead of a fixed 3. If no idea reaches 3.5, the single highest-scoring idea is still shown, marked "below threshold — shown for reference", so the report is never empty. |
 | 4.0.0 | 2026-06-22 | XSPEC-296: Brainstorm Quality Standard (BQS v1) — a four-layer × timeline quality contract layered additively on v3. Layer 0 explore/exploit intent (modulates D2 weight); Layer 1 process leading dimensions D1–D4 with a hard sequence gate (D5–D8 forbidden during divergence); Layer 2 product leading dimensions D5–D8 applied to Top 3 only, with a Seeds column and high-variance contested zone; Layer 3 Judgment Override (overrides aggregate score). D4 judge≠generator (independent context else `[degraded]`); D5 claim split (external-fact floor cross-tier); D7 two-state falsifiability ("need to do X" → next-step, not fail); Meta stop rule (Top-3 set stability + hard cap 2 rounds). Session Self-Evaluation re-positioned as the calibration loop's lagging end (Adoption→D6, Diversity→D2/D3, Cognitive Load→cost; two parallel evaluations forbidden). First principle refined to "decisions use leading, calibration uses lagging". Tiering bound to objective Mode Selection triggers. New flag `--intent`. All v3 flags/mechanisms preserved. |
 | 3.0.0 | 2026-06-01 | XSPEC-247: DIVERGE re-centred on persona ensemble + diversity lenses (analogical/reversal/morphological); CONVERGE re-centred on multi-critic panel + hard-role rebuttal (Devil's Advocate + Steelman); Diversity-Collapse Guardrail; Enhanced Tier (parallel persona/critic agents, graceful fallback); Research Foundations rebuilt on 6 verified 2024–2026 sources; Validity Caveats re-rated (pre-flight LOW, Nijstad/Nemeth demoted); new flags `--personas`/`--lens`/`--enhanced`; anti-seed guardrail |

package/bundled/skills/code-review-assistant/SKILL.md

@@ -82,6 +82,7 @@ After `/code-review` completes, the AI assistant should suggest:
 
 ## Reference | 參考
 
+- [Pre-Commit Checklist](./checkin-checklist.md) — Read it before committing, as the pre-commit pass.
 - Detailed guide: [guide.md](./guide.md)
 - Core standard: [code-review-checklist.md](../../core/code-review-checklist.md)
 

package/bundled/skills/commands/brainstorm.md

@@ -28,7 +28,7 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 | **PRE-FLIGHT** | Prevent AI anchoring | User writes 3 ideas first; no analogy seed; **BQS Layer 0 intent (explore/exploit)** | 防止錨定 |
 | **FRAME** | Define problem clearly | 5 Whys, HMW, Stakeholder Map; **D1 frame purity** | 清楚定義問題 |
 | **DIVERGE** | Force viewpoint diversity | Persona ensemble + diversity lenses; **BQS Layer 1 D1–D4, hard gate hides D5–D8** | 逼出視角多樣性 |
-| **CONVERGE** | Bias-checked selection | Multi-critic panel + Devil's Advocate/Steelman; **BQS Layer 2 D5–D8 on Top 3, D4 judge≠generator** | 降偏誤選擇 |
+| **CONVERGE** | Bias-checked selection | Multi-critic panel + Devil's Advocate/Steelman; **BQS Layer 2 D5–D8 on the Recommended Set (Agg. ≥ 3.5), D4 judge≠generator (two axes: independent context + declared model level)** | 降偏誤選擇 |
 | **OUTPUT** | Actionable report | Brainstorm Report + Seeds + Contested Zone + **Layer 3 Judgment Override** | 可執行的報告 |
 
 > **BQS v1 (XSPEC-296):** v4 layers a four-layer × timeline quality contract on v3. Full oracles and structural rules are in the skill — see [Brainstorm Assistant Skill](../brainstorm-assistant/SKILL.md#bqs-v1--quality-contract--bqs-v1--品質契約).
@@ -64,7 +64,7 @@ PRE-FLIGHT ──► FRAME ──► DIVERGE ──────────► C
 ## Problem Statement
 [Refined problem from FRAME phase]
 
-## Top 3 Recommendations
+## Recommendations (Agg. Score ≥ 3.5 — uncapped)
 1. **[Idea]** — Agg. X.X ✓ Passed rebuttal — Persona: [..] — [Why recommended]
 
 ## Diversity Note
@@ -96,7 +96,7 @@ After brainstorming, the typical workflow is:
 | `/brainstorm "topic"` | 以指定主題啟動 PRE-FLIGHT，再進入 FRAME |
 | `/brainstorm --personas "a,b,c"` | 以自訂 persona 組進入 DIVERGE |
 | `/brainstorm --lens <name>` | 以指定多樣性透鏡為主進入 DIVERGE |
-| `/brainstorm --enhanced` | 若宿主支援子代理則平行跑 persona/評審，否則靜默退回 baseline；此模式下 BQS D4 方可 pass |
+| `/brainstorm --enhanced` | 若宿主支援子代理則平行跑 persona/評審，否則靜默退回 baseline；此模式滿足 BQS D4 的座位軸（獨立 context），pass 另需宣告模型層級（心智軸，見 skill D4） |
 | `/brainstorm --intent exploit` | 以 exploit 意圖啟動（BQS Layer 0），D2 低覆蓋不扣分 |
 
 ### Interaction Script | 互動腳本
@@ -121,12 +121,13 @@ After brainstorming, the typical workflow is:
 3. 至少套用一個多樣性透鏡（類比／假設反轉／形態矩陣）
 4. 以多樣性（覆蓋的 persona/透鏡數，D2/D3）而非數量為繼續門檻
 5. **硬序列閘**：D5–D8 在最後一個 persona 產完前**不得揭示或評分**；CONVERGE critic 不得提前呼叫
+6. **模型層級（發散側）**：以 **Standard 或更高**層級執行——依 core/model-selection.md 的 MS-004／Criterion 2（引用，不複述；不是「最高層級」）
 
-#### CONVERGE Phase (BQS Layer 2 — D5–D8 on Top 3 only)
-1. 以 3 個獨立評審（工程可行性／使用者影響／策略一致）各自評分後聚合
-2. 對前 3 名跑硬角色 Devil's Advocate（論證會失敗）+ Steelman；使用者以 (a)修改/(b)反駁/(c)移除 回應。**D4 判官≠產生者**：評審須在獨立 context（`--enhanced`）方可 pass，否則標 `[degraded]` 不得 pass
-3. **僅對 Top 3** 套用 D5 接地（外部事實需 file:line／來源、跨級地板；假說免接地）、D6 淨值（解誰問題／真有嗎／不做代價＋掛 lagging 欄）、D7 二態證偽（「需先做 X」轉 next-step、不算 fail）、D8 next-step 裁決
-4. **Meta 停止規則**：維度全綠且再跑一輪 Top 3 集合成員不變→停；硬上限 2 輪
+#### CONVERGE Phase (BQS Layer 2 — D5–D8 on the Recommended Set only)
+1. 以 3 個獨立評審（工程可行性／使用者影響／策略一致）各自評分後聚合；**評審提示以「目標＋約束」表達**（判準，不是步驟清單——理由見 skill 引用的 core/model-selection.md MS-009），加權公式與逐準則指南為參考格式
+2. 對**推薦集**（Agg. Score ≥ 3.5，不限筆數）跑硬角色 Devil's Advocate（論證會失敗）+ Steelman；使用者以 (a)修改/(b)反駁/(c)移除 回應。**D4 判官≠產生者，兩軸**：(1) 評審在獨立 context（`--enhanced`）執行，且 (2) 宣告模型層級（core/model-selection.md 的廠商中立標籤；指名具體廠商模型即違規）——缺任一軸標 `[degraded]` 不得 pass
+3. **僅對推薦集**套用 D5 接地（外部事實需 file:line／來源、跨級地板；假說免接地）、D6 淨值（解誰問題／真有嗎／不做代價＋掛 lagging 欄）、D7 二態證偽（「需先做 X」轉 next-step、不算 fail）、D8 next-step 裁決——四者為判準、不限順序
+4. **Meta 停止規則**：維度全綠且再跑一輪推薦集集合成員不變→停；硬上限 2 輪
 
 🛑 **STOP**: 展示 Brainstorm Report（含 Seeds／爭議區／Judgment Override）後等待使用者決定下一步
 
@@ -147,13 +148,15 @@ After brainstorming, the typical workflow is:
 | 使用「像 X 但給 Y」種子 | 改寫為底層問題再繼續（反種子 guardrail） |
 | 指定的 persona/透鏡不存在 | 列出可用選項供選擇 |
 | `--enhanced` 但宿主不支援子代理 | 靜默退回 baseline，照常進行 |
-| 前 3 名全來自同一 persona/透鏡 | 標示並在輸出前再跑一個透鏡 |
+| 推薦集全來自同一 persona/透鏡 | 標示並在輸出前再跑一個透鏡 |
 | baseline 單 context 跑評審（無獨立 context） | BQS D4 標 `[degraded]`，不得標 pass |
+| 評審未宣告模型層級（或以具體廠商模型名宣告） | BQS D4 標 `[degraded]` 不得 pass；模型名改為 core/model-selection.md 的層級標籤 |
 | 外部事實宣稱無 file:line／來源 | D5 接地 fail，要求補來源或改標 `[假說]` |
 
 ## References | 參考
 
 * [Brainstorm Assistant Skill](../brainstorm-assistant/SKILL.md)
 * [Brainstorm Assistant Guide](../brainstorm-assistant/guide.md)
+* [Model Selection](../../core/model-selection.md) — model-level labels for BQS D4 and the divergence side (cited, not restated)
 * [Requirement Assistant](../requirement-assistant/SKILL.md)
 * [Spec-Driven Development](../spec-driven-dev/SKILL.md)

package/bundled/skills/commands/config.md

@@ -170,7 +170,6 @@ Supported tools for Commands:
 | Mode | Description | 說明 |
 |------|-------------|------|
 | `standard` | Summary + task mapping, AI knows when to read which standard (Recommended) | 摘要 + 任務映射（推薦）|
-| `full` | Embed all standards in integration files | 完整內嵌所有標準 |
 | `minimal` | Only core rules embedded | 僅內嵌核心規則 |
 
 ## Effects of Configuration Changes | 設定變更的影響