feat(L): all 57 regimes rewritten to canonical IDENTITY template

.gitignore

@@ -22,3 +22,8 @@ Thumbs.db
 frontend/
 gui/dist/
 SECURITY-AUDIT.md
+
+# CivAgent v5 rewrite artifacts
+regimes/**/*.rejected
+regimes/**/*.v4.bak
+regimes/**/*.new

docs/IDENTITY-TEMPLATE.md

@@ -0,0 +1,105 @@
+# Canonical IDENTITY.md Template
+
+All 57 regimes conform to this structure. The validator (`npm run validate:regimes`) enforces presence of each H2 section.
+
+## Rationale
+
+v4 IDENTITY.md files ranged from 49 to 106 lines with inconsistent sections — some had org charts, others lacked role mapping tables. The runtime (`engine/regime-to-cc.mjs`) parses the role mapping table; regimes without it fall back to defaults, losing governance authenticity.
+
+v5 canonicalizes the structure so the Agent Team reliably mirrors each civilization's actual governance architecture.
+
+## Required structure
+
+```markdown
+# <Chinese name> — <Governance System Name>
+
+> ⚠️ Attribution block (preserved from upstream when applicable).
+
+## 制度简介 / System Overview
+
+2–4 sentences on the governance system's era, duration, historical significance.
+Both Chinese and English should appear when the regime is non-Chinese.
+
+## 组织架构图 / Organization Chart
+
+ASCII diagram showing the decision hierarchy. Must include:
+- A top authority (emperor / assembly / consul / etc.)
+- 3–10 distinct organs/offices with names in the original language
+- Arrows indicating command / review / report flow
+
+Example (Tang):
+\```
+              ┌──────────┐
+              │  皇帝    │
+              └────┬─────┘
+                   │ 圣旨
+      ┌────────────┼────────────┐
+      ▼            ▼            ▼
+  [中书省]    [门下省]    [尚书省]
+   起草        审核        执行
+                              │
+            ┌───┬───┬───┬───┬─┴─┬───┐
+            ▼   ▼   ▼   ▼   ▼   ▼
+          吏部 户部 礼部 兵部 刑部 工部
+\```
+
+## 角色映射表 / Role Mapping Table
+
+Required columns, parsed by `engine/regime-to-cc.mjs`:
+
+| 历史角色 / Historical Role | Agent ID | AI 职责 / AI Responsibility | 推荐模型 / Recommended Model |
+|---|---|---|---|
+| 皇帝 | emperor | coordinator | sonnet |
+| 中书舍人 | zhongshu-drafter | engineering (draft code/policy) | opus |
+| 门下给事中 | menxia-reviewer | review | opus |
+| ... | ... | ... | ... |
+
+Each row must have a non-empty `Agent ID` (lowercase kebab-case) and one of the canonical AI roles:
+**coordinator, engineering, review, research, data, devops, content, legal, management**
+
+## 决策流程 / Decision Flow
+
+3–7 step narrative describing how a typical governance decision moves through the org chart. Reference the Agent IDs from the mapping table.
+
+Example:
+1. **emperor** receives a petition about frontier famine.
+2. **emperor** issues an edict to **zhongshu-drafter**.
+3. **zhongshu-drafter** produces a draft plan.
+4. **menxia-reviewer** rejects or amends the draft (iteration allowed).
+5. On approval, **shangshu-executor** dispatches to the 6 ministries in parallel.
+6. Each ministry reports back with concrete measurable outputs.
+
+## 制度特点 / Characteristics
+
+Bullet list of 3–6 defining features. Each feature should be a specific mechanism, not a platitude:
+- "Three-stage checks-and-balances: drafting, reviewing, and executing are strictly separated" ✓
+- "Good governance" ✗
+
+## Pattern 映射
+
+Single line naming the orchestration pattern from `engine/modes/`:
+
+> **Orchestration pattern**: `checks-and-balances`
+
+One of: `centralized`, `checks-and-balances`, `democratic`, `dual-track`, `federation`, `theocratic`
+(aliases accepted: see `engine/regime-to-cc.mjs` PATTERN_ALIASES).
+
+## 历史参考 / Historical Sources
+
+3–5 primary or authoritative references. Shortest acceptable form:
+- 《旧唐书·职官志》
+- Tang Liu Dian (唐六典), ed. Li Linfu (733 CE)
+- Wang Gungwu, *The Structure of Power in North China during the Five Dynasties*
+
+---
+
+## Checklist for new / rewritten regimes
+
+- [ ] System Overview section ≥ 2 sentences
+- [ ] Org chart with labeled hierarchy
+- [ ] Role mapping table ≥ 5 rows with valid Agent IDs
+- [ ] Decision flow references Agent IDs
+- [ ] Characteristics are mechanism-level, not vague
+- [ ] Orchestration pattern is one of 6 canonical (or documented alias)
+- [ ] ≥ 3 historical sources
+- [ ] Passes `npm run validate:regimes`

regimes/REVIEW-FINDINGS-v5.md

@@ -0,0 +1,21 @@
+# L-Stage Review Findings (Codex)
+
+Codex audit of 5 sample regimes after PR #6 auto-generation. Actionable items for v5.1.
+
+## Fixed in this commit
+
+- **china/tang**: 司礼监 (Ming-Qing eunuch agency) was misplaced in Tang architecture. Replaced with 中书舍人 (Zhongshu Sheren, Tang drafter). "运行超过 1300 年" overclaim clarified — only the three-省 framework persisted, not the specific Tang configuration.
+
+## Deferred to v5.1
+
+- **global/byzantine**: Patriarch depicted as routine ethics veto — overstated. `theokrator` is literary, not a standard office title.
+- **global/roman-republic**: Chart too linear. Assemblies (Comitia) should sit more prominently above consuls and senate. Censor placed too actively in day-to-day flow (role was periodic, not continuous).
+- **global/prussia**: Compresses 1701-1918 into one snapshot. `Ober-Kriegsrat` may be invented. `Kammerdirektor` is too generic. Recommend splitting into `prussia-frederick` (18c) and `prussia-bismarck` (19c).
+- **global/ottoman**: `Nişancı` miscast as 大法官 (should be pen-bearer/tuğra authenticator). Janissary Aga's direct-to-Sultan reporting path overcentralized — ignores parallel ulema channel. One source citation (`Patrick Balfour`) format nonstandard.
+
+## Not yet reviewed (other sampling rounds failed)
+
+- Gemini second-round review blocked on CodeAssist OAuth (API-key mode was not picked up by this invocation — known intermittent issue).
+- cc-mimo third-round review hit `--max-turns 1` limit mid-output; re-run needed.
+
+v5.1 plan: complete Gemini + Mimo rounds on a dedicated branch, fix Byzantine/Roman/Prussian/Ottoman per above.

regimes/china/five-dynasties/IDENTITY.md

@@ -1,64 +1,58 @@
 # 五代十国 — 组织架构
 
 ## 制度简介
-
-五代十国（907-960）是唐朝灭亡到北宋建立之间的大分裂时代。中原先后经历后梁、后唐、后晋、
-后汉、后周五个短命王朝，南方则有吴、南唐、吴越、闽、楚、南汉、前蜀、后蜀、南平、北汉十国并立。
-各地节度使拥兵自重，实际上是独立的军阀政权。朝廷名义上是中央，但对地方控制力极弱。
-这一体制适合模拟多团队竞争、去中心化协作的场景。
+五代十国（907-960）是唐朝藩镇割据的延续与加剧。中原五朝速迭，南方十国并立，节度使拥兵自重、各自为政。名义上有朝廷，实际权力高度分散，是中国历史上最混乱的时期之一，适合模拟去中心化协作、多团队竞争场景。
 
 ## 组织架构图
-
 ```
-                      ┌─────────────────────┐
-                      │      主公（你）      │
-                      │   谁有实力谁是主     │
-                      └──────────┬──────────┘
-                                 │ 号令（有限约束力）
-                                 ▼
-                      ┌──────────────────────┐
-                      │     朝廷（名义中央）  │
-                      │    chaoting          │
-                      └──────────┬──────────┘
-                                 │ 政令（各方可选择执行）
-              ┌──────────────────┼──────────────────┐
-              ▼                  ▼                   ▼
-      ┌──────────────┐  ┌──────────────┐  ┌──────────────┐
-      │  节度使A      │  │  节度使B      │  │  节度使C      │
-      │  独立团队A    │  │  独立团队B    │  │  独立团队C    │
-      │  jiedushi_a   │  │  jiedushi_b   │  │  jiedushi_c   │
-      └──────────────┘  └──────────────┘  └──────────────┘
-              ▲                  ▲                   ▲
-              └──────────────────┼───────────────────┘
-                                 │ 文书协调
-                      ┌──────────────────────┐
-                      │      掌书记          │
-                      │    zhangshji         │
-                      └──────────────────────┘
+                    ┌─────────────┐
+                    │   名义朝廷   │
+                    │  (朝廷天子)  │
+                    └──────┬──────┘
+                           │ 政令（约束力弱）
+          ┌────────────────┼────────────────┐
+          ▼                ▼                ▼
+    ┌──────────┐    ┌──────────┐    ┌──────────┐
+    │ 节度使A   │    │ 节度使B   │    │ 节度使C   │
+    │ (独立军阀)│    │ (独立军阀)│    │ (独立军阀)│
+    └──────────┘    └──────────┘    └──────────┘
+          │                │                │
+          └────────────────┼────────────────┘
+                           │ 协调文书
+                    ┌──────┴──────┐
+                    │   掌书记    │
+                    │ (文官协调)  │
+                    └─────────────┘
 ```
 
 ## 角色映射表
-
-| 古代角色 | Agent ID | AI 职责 | 推荐模型 |
+| 历史角色 | Agent ID | AI 职责 | 推荐模型 |
 |---|---|---|---|
-| 朝廷·天子 | `chaoting` | 名义中央：发布政令、协调各方、名义统筹 | 快速模型 |
-| 节度使A | `jiedushi_a` | 军阀A / 团队A：独立运作，擅长特定领域 | 强力模型 |
-| 节度使B | `jiedushi_b` | 军阀B / 团队B：独立运作，与他方竞争 | 强力模型 |
-| 节度使C | `jiedushi_c` | 军阀C / 团队C：独立运作，各有专长 | 强力模型 |
-| 掌书记 | `zhangshji` | 文书协调：传递信息、记录决策、协调冲突 | 快速模型 |
-
-## 协作流程
-
-1. **主公下令** → 用户发送任务（可指定单个或所有节度使）
-2. **朝廷传令** → `chaoting` 转达政令（节度使可选择执行方式）
-3. **竞争执行** → 多个节度使可同时执行同一任务，各自拿出方案
-4. **掌书记记录** → `zhangshji` 记录各方进展、协调资源冲突
-5. **主公裁决** → 用户比较各方成果，选择最优方案
-6. **合纵连横** → 节度使之间可自由协作或竞争
-
-## 适用场景
-
-- **多团队竞争**：多个团队同时为同一目标提供方案
-- **A/B 测试**：不同方案并行实施，对比效果
-- **去中心化开发**：各团队独立开发，最终整合
-- **创意竞标**：多方案竞争，择优录用
+| 朝廷天子 | imperial-court | coordinator | sonnet |
+| 节度使A | jiedushi-alpha | management | opus |
+| 节度使B | jiedushi-beta | engineering | opus |
+| 节度使C | jiedushi-gamma | devops | opus |
+| 节度使D | jiedushi-delta | research | opus |
+| 掌书记 | zhangshu-secretary | content | sonnet |
+
+## 决策流程
+1. **imperial-court** 发布名义政令（各方可选择性执行）
+2. 诸 **jiedushi** 并行独立研判，各自制定方案
+3. **zhangshu-secretary** 传递文书、记录各方进展、协调冲突
+4. 节度使间可自由合纵连横，结盟或竞争
+5. 最终方案由实力最强的节度使主导，或多方协作完成
+
+## 制度特点
+- 实力至上：兵强马壮者为天子，政权更迭频繁
+- 高度自治：节度使独立决策，朝廷政令约束力极弱
+- 并行竞争：多方同时执行任务，择优录用
+- 动态联盟：节度使间可自由结盟或对立
+
+## Pattern 映射
+> **Orchestration pattern**: `federated`
+
+## 历史参考
+- 《旧五代史·职官志》
+- 《新五代史·职方考》
+- 欧阳修《五代史记》
+- 钱穆《中国历代政治得失》