docs: split translation vs proofread workflows in AGENTS.md; cleanup edit-suggestions to manuscript-level only

AGENTS.md

@@ -32,19 +32,108 @@ translate-files/<topic>/<article>/
 
 `.docx` output → `/tmp/`. Don't commit binaries.
 
-## Translation Workflow
+---
 
-1. Load `dharma-translation` and `terms-search` skills
+## Workflow A: Translation（翻译）
-2. Search terms DB with `search.py` for key terms before translating
+
-3. Translate directly — agent IS the model, no external APIs
+Translate Chinese source into English. The agent IS the model — no external APIs.
-4. Write `target.dj`, match source line count exactly
+
-5. TOC: bullet list, no link targets. No page numbers.
+### Input
-6. Create `bilingual.dj` (interleave source/target lines)
+
+Source text in `.dj` or `.docx` (Chinese only).
+
+### Deliverables
+
+- `source.dj` — extracted/cleaned Chinese
+- `target.dj` — English translation, line count matches source
+- `bilingual.dj` — interleaved (source line, target line adjacent, blank between pairs)
+- `edit-suggestions.dj` — terminology/consistency issues flagged for review
+
+### Rules
+
+1. Load `dharma-translation` and `terms-search` skills before starting.
+2. Search terms DB with `search.py` for key Buddhist terms.
+3. TOC: plain bullet lists, no link targets, no page numbers.
+4. Djot formatting:
+   - Emphasis: `*text*` (single asterisks). Never `**` (Markdown bold).
+   - Comments: `{% ... %}`
+   - Em dashes in English text: `---` (three hyphens). Pandoc converts to proper em dash in docx output.
+   - En dashes in English text: `--` (two hyphens).
+5. Preserve source formatting — don't add/remove emphasis.
+6. Translate in-response — never call external translation APIs.
+
+### Review
+
+After translating, run `translation-review` skill (Workflow B) to check:
+- Terminology consistency against terms DB
+- Grammar, fluency, calques
+- Missing content (mid-paragraph truncation)
+- Inconsistency (same term translated differently)
+
+---
+
+## Workflow B: Proofread（校对）
+
+Extract bilingual content from an existing DOCX manuscript and flag issues.
+The DOCX already contains BOTH Chinese and English — the English is the
+authoritative target, not something the agent creates.
+
+### Input
+
+A `.docx` manuscript with Chinese and English in parallel (typically
+paragraph-level alternation: Chinese, blank, English, blank).
+
+### Deliverables
+
+- `bilingual.dj` — extracted from DOCX (interleaved)
+- `edit-suggestions.dj` — mechanical/issues found in the manuscript
+
+### Extraction
+
+1. `pandoc input.docx -f docx -t plain --wrap=none` to plain text
+2. Script to extract CN/EN pairs. Write to `scripts/gen-bilingual-<name>-<hash>.py`.
+   Copy the approach from an existing script (e.g. `gen-bilingual-buddhist-attitude.py`).
+3. Write `bilingual.dj`.
+
+### What to flag (edit-suggestions.dj)
+
+Only flag objective/manuscript-level issues:
+
+- **Garbled text** — merged duplicate edits in source DOCX
+- **Double words** — `the The`, `is is`
+- **Double punctuation** — `warm or cold..`
+- **Numbering mismatches** — CN heading `三` vs EN heading `II`, or body heading
+  number doesn't match TOC
+- **Translator notes** — `（某某翻，某某审）` left in headings
+- **Missing content** — paragraphs present in CN but missing in EN (or vice versa)
+- **Duplicate text** — same name/phrase repeated (`岳麓书院岳麓书院`)
+- **Capitalization typos** — `Philosopher Nietzsche` → `philosopher Nietzsche`
+- **Stray/unusual characters** in either language
+
+### What NOT to flag
+
+Do NOT flag translation quality issues — the English is authoritative manuscript text:
+
+- Terminology choices (e.g. "delusional dreams" for 颠倒梦想)
+- Translation style, calques, fluency preferences
+- Djot formatting conventions (em-dashes, italics)
+- Word order, added concepts, degree weakening
+
+These belong to the **translation** workflow, not proofread.
+
+### Separate translation-quality findings
+
+If the user explicitly asks for a translation review of a proofread article,
+write findings to `translation-findings.dj` (separate from `edit-suggestions.dj`).
+Load the `translation-review` skill and follow its Workflow B.
+
+---
 
 ## Djot
 
 - Comments: `{% ... %}`
-- Dashes: `---` em, `--` en. Pandoc converts in docx output.
+- Emphasis: `*text*` (single asterisks)
+- Dashes in English: `---` em, `--` en. Pandoc converts in docx output.
 - Preserve source formatting — don't add/remove emphasis
 
 ## Scripts
@@ -58,3 +147,4 @@ regenerating the same Python in execute_code each turn.
 - `scripts/dj2docx.fish <target.dj>` — pandoc .dj → .docx in `/tmp/`
 - `scripts/proofread-pdf.py <docx> <pdf>` — word-level diff between manuscript and typeset PDF
 - `scripts/gen-bilingual.fish <dir>` — produce `bilingual.dj` from `source.dj` + `target.dj`
+- `scripts/gen-bilingual-<name>-<hash>.py` — article-specific extraction from DOCX

translate-files/生命也可以被设计的/edit-suggestions.dj

@@ -5,13 +5,12 @@
 - 第47行 `Philosopher Nietzsche`
 + 第47行 `philosopher Nietzsche`
   说明：此处 philosopher 是描述性用语，并非正式头衔，应小写。
-  同一段上文 `the philosopher Zhou Guoping` 已用对。
 
 ## 一、教育是为了育人
 
 - 第130行 CN `我曾和岳麓书院岳麓书院国学研究院朱汉民院长对谈`
 + 第130行 CN `我曾和岳麓书院国学研究院朱汉民院长对谈`
-  说明：Chinese source 中 "岳麓书院" 出现两次，疑为重复。
+  说明：源文 "岳麓书院" 出现两次，疑为重复。
 
 ## 二、人生最重要的设计
 
@@ -28,19 +27,3 @@
 - 第155行 EN heading `2. The Principles of Life's Continuity`
 + 第155行 EN heading `3. The Principles of Life's Continuity`
   说明：TOC 中该条目中英文均为 "3"，正文 EN 标题却写为 "2."，typographical error。
-
-## 四、人性，是设计生命的基础
-
-### 2．人的两面性
-
-- 第211–212行：CN 段落"关于这个问题，孟子进一步提出…"对应的英文有两个段落：
-  (1) "Building on this, Mencius introduced…"（已收录，第212行）
-  (2) "Buddhism teaches that every person possesses Buddha-nature…"（缺失）
-  说明：译者将一个中文段落拆成两个英文段落，bilingual.dj 只保留了第一个。
-  建议：将中文在"这一观点和佛法思想有关。"后拆分为二，使之与英文两段一一对应。
-
-## 全文
-
-- 27处英文 em-dash（Unicode `—`）应替换为三个连字符 `---`
-  说明：djot 规范要求英文 em-dash 用 `---` 表示，pandoc 输出 docx 时自动转换。
-  涉及行号：41, 44, 50, 56, 59, 65, 68, 74, 77, 80, 95, 104, 110, 116, 134, 146, 149, 158, 173, 200, 206, 212, 215, 218, 224, 233, 248

translate-files/生命也可以被设计的/translation-findings.dj

@@ -0,0 +1,52 @@
+# Translation Findings — 生命也可以被设计的
+
+## 1. 内容删减：希特勒举例被移除（四、1．善用理性）
+
+  Chinese: 比如希特勒认为犹太人血统有问题，要建立纯净的血统，就得实施种族屠杀。这正是理性被扭曲为反人道的体现。
+  English: History shows that reason, when distorted by ideology, can even be used to justify atrocities such as genocide.
+
+  CN 用希特勒的具体历史案例说明"理性被扭曲为反人道"，EN 替换为泛泛的 "History shows..."。
+  "such as genocide" 提到了种族屠杀，但原文点名希特勒、犹太人血统的论证链被消解。
+
+## 2. 术语稀释："颠倒梦想"（二、2．心理问题）
+
+  Chinese: 《心经》所说的颠倒梦想
+  English: what the Heart Sutra describes as "delusional dreams"
+
+  "颠倒梦想" 是《心经》核心术语（viparyāsa），四颠倒：常乐我净。指凡夫对无常/苦/无我/不净的四种错误认知。
+  "delusional dreams" 偏文学化，丢失了术语的技术含量。
+  建议："inverted views and delusive thoughts" 或保留 "viparyāsa (the four inverted views)"。
+
+## 3. 双重翻译："本来面目"（二、1．人生问题）
+
+  Chinese: "本来面目"
+  English: "Original Face; Original Mind"
+
+  用分号并列两种译法 "Face" 和 "Mind"，像是译者未决断。
+  禅宗术语通常统一用 "Original Face"（对应"面目"），或 "True Nature"。
+  建议选一种，删掉分号后的备选。
+
+## 4. 添加概念："creator god"（三、2．人有自救能力）
+
+  Chinese: 但佛法不认同世界有一个主宰神
+  English: Buddhism does not acknowledge a creator god who governs the world
+
+  "主宰神" = ruling/controlling deity。译文加了 "creator"（创世神），原文无"创世"义。
+  佛经否定的是梵天式的创世主（creator），但"主宰神"范围更宽（包括印度教的大自在天等）。
+  建议：去掉 "creator"，改为 "a ruling deity who governs the world"。
+
+## 5. 语序颠倒：内圣外王（三、1．人生目的何在）
+
+  Chinese: 进一步，是内圣外王，成就圣贤人格
+  English: at a higher level, it aspires to serve the people with kingly virtue and achieve the character of a sage
+
+  "内圣外王" = 内（圣）→ 外（王）。原文先内后外，译文先外后内。
+  建议调整语序以符合原文逻辑递进。
+
+## 6. 程度弱化：一切 → many forms of（二、2．心理问题）
+
+  Chinese: 三毒是一切心理疾病的根源
+  English: the three poisons are the root of many forms of mental afflictions
+
+  "一切" → "many forms of"，从全称变为部分。
+  建议："the root of all mental afflictions"。